Está en la página 1de 764

Nuclear Physics B 581 (2000) 3–33

www.elsevier.nl/locate/npe

SUSY Higgs at the LHC:
effects of light charginos and neutralinos
G. Bélanger a , F. Boudjema a , F. Donato a , R. Godbole b , S. Rosier-Lees c
a Laboratoire de Physique Théorique, Chemin de Bellevue, B.P. 110, F-74941 Annecy-le-Vieux, Cedex, France
b Centre of Theoretical Studies, Indian Institute of Science, Bangalore 560 012, India
c Laboratoire de Physique des Particules, Chemin de Bellevue, B.P. 110, F-74941 Annecy-le-Vieux, Cedex,

France
Received 3 March 2000; accepted 19 April 2000

Abstract
In view of the latest LEP data we consider the effects of charginos and neutralinos on the twophoton and bb¯ signatures of the Higgs at the LHC. Assuming the usual GUT inspired relation
between M1 and M2 we show that there are only small regions with moderate tan β and large stop
mixings that may be dangerous. Pathological models not excluded by LEP which have degeneracy
between the sneutrino and the chargino are however a real danger because of large branching fraction
of the Higgs into invisibles. We have also studied models where the gaugino masses are not unified
at the GUT scale. We take M1 = M2 /10 as an example where large reductions in the signal at the
LHC can occur. However we argue that such models with a very light neutralino LSP may give a
too large relic density unless the sleptons are light. We then combine this cosmological constraint
with neutralino production with light sfermions to further reduce the parameter space that precludes
observability of the Higgs at the LHC. We still find regions of parameter space where the drops in
the usual Higgs signals at the LHC can be drastic. Nonetheless, in such scenarios where Higgs may
escape detection we show that one should be able to produce all charginos and neutralinos. Although
the heavier of these could cascade into the Higgs, the rates are not too high and the Higgs may not
always be recovered this way.  2000 Elsevier Science B.V. All rights reserved.

1. Introduction
Uncovering the mechanism of symmetry breaking is one of the major tasks of the high
energy colliders. Most prominent is the search for the Higgs particle. Within the standard
model, SM, this scalar particle poses the problem of naturalness and its mass is a free
parameter. Current data 1 seem to indicate a preference for a light Higgs with a mass
1 The limit on the mass of the SUSY Higgs depends on the SUSY parameters. For an update on the limits on
the Higgs mass see [1–4].

0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 2 4 3 - 1

4

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

that can nicely fit within a supersymmetric version of the SM. In fact an intermediate
mass Higgs, IMH, is one of the most robust prediction of SUSY, since one does not have
strict predictions on the large array of the other masses and parameters in this model.
Another, perhaps circumstantial, evidence of SUSY is the successful unification of the
gauge couplings at some high scale. Add to this the fact that the neutralino can provide
a good dark matter candidate explains the popularity of the model. Even so the search
for the lightest Higgs is not so easy. LEP2 where the Higgs signature is easiest may
unfortunately be some 20–30 GeV short to be able to cover the full range of the minimal
SUSY lightest Higgs mass. Searches at the Tevatron need very good background rejection
and in any case need to upgrade the present luminosities quite significantly. At the LHC,
most analyses have relied extensively on the two-photon decay of the IMH either in the
dominant inclusive channel through gg → h → γ γ or in associated production. Only
recently has it been shown that associated production of the Higgs with tops with the
former decaying into bb¯ can improve the discovery of the Higgs, albeit in the region mh <
120 GeV 2 . Unfortunately, until recently [5], most simulations for Higgs searches have in
effect decoupled the rest of the supersymmetric spectrum from the Higgs sector, like in the
much advertised ATLAS/CMS MA − tan β plane [5,6].
This assumption of a very heavy SUSY spectrum can not be well justified. First,
naturalness arguments require that at least some of the SUSY masses be below 1 TeV
or even much less. Second, it has been known [7–9] that relaxing this assumption can
have some very important consequences on the Higgs search at the LHC. This is not
surprising considering the fact that the most important production channel gg → h is loop
induced as is the main discovery channel h → γ γ . One of the most dramatic effect is
that of a light stop with large mixing which drastically reduces the production rate [7–10].
Fortunately, when this happens, a careful analysis [11] shows that the Higgs signal can
be rescued in a variety of channels that become enhanced or that open up precisely for
the same reason that the normal inclusive channel drops, so that in a sense there is a
complementarity. For instance with all other sparticles but the stops heavy, one can show
that whenever the production rate in the inclusive channel drops, the branching ratio into
two photons increases with the consequence that associated W h/Zh and t t¯h where the
Higgs decays into two photons becomes a very efficient means of tracking the Higgs.
Moreover, associated t˜1 t˜1 h production [11–13] becomes important through the cascade of
the heavier stop t˜2 , t˜2 → t˜1 h. At the same time since the hbb¯ coupling is hardly affected
t t¯h production could play an important role. Similar sort of complementarity has also
been pointed out in supersymmetric scenarios where the coupling hbb¯ can be made very
small [14].
In our investigation of the effects of light stops with large mixing, all other particles but
the stops were assumed rather heavy. It is then important to ask how the overall picture
changes had we allowed other sparticles to be relatively light. Considering that the present
LEP and Tevatron data precludes the decay of the lightest Higgs into sfermions, the effect
of the latter on the properties of the lightest Higgs can only be felt through loops. These
2 For an update on Higgs searches in CMS see [5].

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

5

effects can therefore be considered as a special case of the stop that we studied at some
length and apart from the sbottom at large tan β the effects will be marginal. One can
then concentrate on the spin-half gaugino–higgsino sector. In order to extract the salient
features that may have an important impact on the Higgs search at the LHC, we leave
out in this study the added effects of a light stop. Compared to the analysis with the
stop, this sector does not affect inclusive production nor the usual associated production
mechanisms. The effect will be limited to the Higgs decay. First, if the charginos are not
too heavy they can contribute at the loop level. We find however, by imposing the present
limits on their masses, that this effect is quite small. On the other hand we show that the
main effect is due to the possible decay of the Higgs into the lightest neutralino. This is
especially true if one relaxes the usual so-called unification condition between the two
gaugino components of the LSP neutralino. Although at LEP an invisible Higgs is not so
much of a problem 3 , since it can be easily tagged through the recoiling Z, it is a different
matter at the LHC. Few studies have attempted to dig out such an invisible (not necessarily
supersymmetric) Higgs at the LHC, in the associated Zh, (W h) [16,17] channel. Even with
rather optimistic rejection efficiencies the backgrounds seem too overwhelming. It has also
been suggested [18] to use associated t t¯h production but this requires very good b-tagging
efficiencies and a good normalisation of the backgrounds. Recently Ref. [19] looked at how
to hunt for an invisible Higgs at the Tevatron. For mh > 100 GeV a 5σ discovery requires
a luminosity in excess of 30 fb−1 .
¯ where a sort of compensation
Compared to the effects of the stop or a vanishing hbb,
occurs in other channels, the opening up of the invisible decay reduces all other channels
¯ Previous studies [20–24] have mainly concentrated if
including the branching ratio in bb.
not on a mSUGRA scenario then on a scenario based on the mSUGRA inspired relation
between the electroweak gaugino masses, M1 , M2 . Moreover, LEP searches and limits
refer essentially to the latter paradigm. In the course of this analysis we had to re-interpret
the LEP data in the light of more general scenarios. We therefore had to take recourse to
various limits on cross sections rather than absolute limits on some physical parameters
quoted in the literature. We have also tried to see whether new mechanisms come to the
rescue of the Higgs search at the LHC when the invisible channel becomes substantial and
reduces the usual signal significantly. Much like in our analysis of the stop [11] where
we found that the Higgs could be produced through the decay of the heavier stop into the
lighter one, we inquired whether a similar cascade decay from the heavier neutralino or
charginos to their lighter companions can take place. This is known to occur, for instance
for some mSUGRA points [5,25], but its rate is found not to be substantial when an
important branching ratio into invisibles occurs. Even if it were substantial it would be
difficult to reconstruct the Higgs since again at the end of the chain the Higgs will decay
predominantly invisibly.
Considering the dire effects of a large invisible branching ratio occurring for rather
light neutralinos, we have investigated the astrophysical consequences of such scenarios,
specifically the contribution of such light neutralinos on the relic density. We find that these
3 For an updated analysis at √s = 189 GeV see [15].

6

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

models require rather light slepton masses. In turn, with such light sleptons, neutralino
production at LEP2 provides more restrictive constraints than the chargino cross sections.
Taking into account the latter constraints helps rescue some of the Higgs signals.
The paper is organised as follows. In the next section we introduce our notation for the
chargino–neutralino sector and make some qualitative remarks concerning the coupling of
the lightest Higgs as well as that of the Z to this sector. This will help understand some of
the features of our results. In Section 2 we review the experimental constraint and discuss
how these are to be interpreted within a general supersymmetric model. Section 3 presents
the results for Higgs detection at the LHC within the assumption of the GUT relation for
the gaugino masses. Section 4 analyses the “pathological” cases with a sneutrino almost
degenerate with the chargino, leading to lower bounds on the chargino mass. Section 5
analyses how the picture changes when one relaxes the GUT inspired gaugino masses
constraint and the impact of the astrophysical constraints on the models that may jeopardise
the Higgs search at the LHC. Section 6 summarises our analysis.

2. The physical parameters and the constraints
2.1. Physical parameters
When discussing the physics of charginos and neutralinos it is best to start by defining
one’s notations and conventions. All our parameters are defined at the electroweak scale.
The chargino mass matrix in the gaugino–higgsino basis is defined as
√  

2MW cos β
√ M2
,
(2.1)
2MW sin β
µ
where M2 is the soft SUSY breaking mass term for the SU(2) gaugino while µ is the socalled higgsino mass parameter whereas tan β is the ratio of the vacuum expectation values
for the up and down Higgs fields.
Likewise the neutralino mass matrix is defined as


M1

0

 −M sin θ cos β
Z
W
MZ sin θW sin β

0
M2
MZ cos θW cos β
−MZ cos θW sin β

−MZ sin θW cos β
MZ cos θW cos β
0
−µ

MZ sin θW sin β
−MZ cos θW sin β 
, (2.2)
−µ
0

where the first entry M1 (corresponding to the bino component) is the U (1) gaugino mass.
The oft-used gaugino mass unification condition corresponds to the assumption
5 2
M2
tan θW M2 '
.
(2.3)
3
2
Then constraints from the charginos alone can be easily translated into constraints on the
neutralino sector. Relaxing Eq. (2.3), or removing any relation between M1 and M2 means
that one needs further observables specific to the neutralino sector.
The other parameters that appear in our analysis emerge from the Higgs sector. We base
our study on the results and prescription of [26,27] for the improved two-loop calculations
M1 =

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

7

based on the effective Lagrangian. 4 The parameters here are, apart from the ubiquitous
tan β, the mass of the pseudo-scalar Higgs, MA , At the trilinear mixing parameter in the
stop sector, as well as MS a supersymmetric scale that may be associated to the scale in the
stop sector. Since we want to delimit the problem compared to our previous study on the
stop effects, we will set the stop masses (and all other squarks) to 1 TeV. We will also be
working in the decoupling limit of large MA that we also set at 1 TeV. The lightest Higgs
mass is then larger than if we had taken a lower MA . As we will see the most important
effect that results in small branching ratio for the two-photon width is when the invisible
decay opens. This occurs if one has enough phase space and therefore if the mass of the
Higgs is made as large as possible.
√ Thus for a given tan β the effect is maximal for what
is called maximal mixing: At ∼ 6MS in the implementation of [26,27]. One would also
think that one should make tan β large, however this parameter also controls the masses
of the neutralinos and for the configuration of interest, those leading to the largest drops
in the two-photon signal, one needs to keep tan β as low as possible to have the lightest
neutralino as light as possible.
In principle we would have liked to decouple all other sparticles, specifically sfermions
as stated in the introduction. However sleptons (in particular selectrons and sneutrinos)
masses determine also the cross sections and the decay signature of the charginos and the
neutralinos. Therefore, allowing for smaller sfermions masses does not so much directly
affect the two-photon width but can relax quite a bit some of the limits on the chargino–
neutralino sector which in turn affect the Higgs search. We thus allow for this kind of
indirect dependence on the sfermion mass.
Often, especially in the case of neutralinos, LEP analyses set absolute bounds on masses.
Ideally, since one is using bounds that are essentially set from the couplings of neutralinos
to gauge bosons, to translate to couplings of these neutralinos and charginos to the Higgs,
one needs to have access to the full parameter space µ, tan β, M1 , M2 . Thus absolute
bounds are only indicative and it is much more informative to reinterpret the data. In the
case of limits set solely from the chargino data, the re-interpretation is quite straightforward
since no assumption on the parameters in Eq. (2.1) is made and the limits ensue from
e+ e− → χ˜ 1+ χ˜ 1− . Limits on the neutralinos are a bit more involved. To make some of these
points clearer and to help understand some of our results it is worth reviewing the couplings
to neutralinos.
2.2. Couplings of neutralinos to the Higgs and Z
The width of the lightest Higgs to the lightest neutralinos writes [20–23]

 GF MW mh 
3/2

C 0 0

2 ,
1 − 4m2χ˜ 0 /m2h

Γ h → χ˜ 10 χ˜ 10 =
hχ˜ 1 χ˜ 1
1
2 2π

(2.4)

where [33]
4 There is now a two-loop diagrammatic calculation ([28], the limiting case of vanishing stop mixing and large
MA and tan β has been considered by R. Hempfling and A. Hoang [29]) which is in good agreement with an
updated version of the two-loop effective Lagrangian approach [30,32].

8

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33  

N
N
N
N
Chχ˜ 0 χ˜ 0 = O12
− tan θW O11
+ cos αO14
sin αO13
1 1  

N
N
N
N
− tan θW O11
− cos βO13
sin βO14
.
' O12

(2.5)

OijN are the elements of the orthogonal (we assume CPconservation) matrix which
diagonalises the neutralino mass matrix. α is the angle that enters the diagonalization of the
CP-even neutral Higgses which in the decoupling limit (large MA and ignoring radiative
N 2
| defines the composition of the lightest
corrections) is trivially related to the angle β. |O1j
0
N
N |2 + |O N |2 is the gaugino
2
neutralino χ˜ 1 . For instance, |O11 | is the bino purity and |O11
12
purity. It is clear then, apart from phase space, that the LSP has to be a mixture of gaugino
and higgsino in order to have a large enough coupling to the Higgs. The same applies for
the diagonal coupling of the charginos (hχ˜ i− χ˜i+ ).
In Fig. 1 we show the strength C 2 0 0 assuming the GUT unification condition between
hχ˜ 1 χ˜ 1

M1 and M2 for tan β = 5 and tan β = 15. One should note that the coupling is much larger
for positive values of µ. The largest effect (peak) occurs for small values of µ and M2
which however are ruled out by LEP data on the chargino mass. Note also, by comparing
the tan β = 5 and tan β = 15 case in Fig. 1, that especially for µ > 0, as tan β increases the
Higgs coupling to the LSP gets smaller. At the same time the neutralino LSP gets heavier.
Thus large tan β values corresponding to higher Higgs masses will not lead to the largest
h → χ˜ 10 χ˜ 10 . Similar behaviour is also observed for the coupling of the chargino to Higgs,
the largest coupling sits in the µ > 0 and small M2 region. However, it turns out that the
effect of charginos in the loop never becomes very large.
As seen in Fig. 2 for the case with M1 = M2 /10, the same kind of behaviour persists:
µ > 0 and moderate tan β lead to stronger couplings. On the other hand, the constraints on
M2 , M1 , µ which are derived for instance from neutralino production are more sensitive to
the higgsino component of the neutralino. Indeed the Z coupling to these writes 

N N
N N 2
Oj 3 − Oi4
Oj 4 .
(2.6)
Z ∗ → χ˜ i0 χ˜ j0 ∝ Oi3
Chargino production in e+ e− is not as much critically dependent on the amount of mixing
since both the wino and (charged) higgsino components couple to the Z and the photon.
Some interference with the t-channel sneutrino exchange may occur in the case of a wino
component (i.e., |µ|  M2 ), therefore the kinematic limit can be reached quite easily,
except the situation where the signature of the chargino leads to almost invisible decay
products.
2.3. Accelerator constraints
This brings us to how we have set the constraints.
Higgs mass
In the scenarios we are considering with large MA and large stop masses the ZZh
coupling is essentially SM-like and LEP2 limits on the mass of the SM Higgs apply with
little change even for an invisible Higgs. In any case, as discussed earlier, to make the
chargino–neutralino effect most dramatic we will always try to maximise the Higgs mass

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

(a)

(b)

(c)

(d)

9

Fig. 1. Strength of the lightest Higgs coupling to χ˜ 10 χ˜ 10 for tan β = 5 and tan β = 15. In the second
set (bottom) of figures we have imposed mχ˜ + > 94.5 GeV (the current limit on the lightest chargino)
1
and mχ˜ 0 < 65 GeV (this corresponds to the threshold for the lightest Higgs with a maximum mass
1
of about 130 GeV to decay into the LSP neutralino). Here Chn1 n1 ≡ Chχ˜ 0 χ˜ 0 .
1 1

independently of tan β by choosing an appropriate At . The LEP2 mass limit are thus always
evaded. For M1 = M2 /10 we stick to tan β = 5, considering there is always enough phase
space for h → χ˜ 10 χ˜ 10 , it is sufficient to discuss the case with At = 0. For the canonical
unification case, the effect of maximising the Higgs mass through At is crucial.
Chargino cross section
Typically when no sparticle is degenerate with the chargino, the lower limit on the
chargino mass reaches the LEP2 kinematic limit independently of the exact composition of
the chargino and does not depend much on the sneutrino mass as explained earlier. Latest
LEP data give [34],
mχ + > 94.5 GeV.
1

(2.7)

10

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

(a)

(b)

Fig. 2. Strength of the lightest Higgs coupling to χ˜ 10 χ˜ 10 for tan β = 5 and tan β = 15 in the case
M1 = M2 /10 . As in Fig. 1, we have imposed mχ˜ + > 94.5 GeV and mχ˜ 0 < 65 GeV.
1

1

Very recent combined preliminary data [1–4] suggest mχ + > 100.5 GeV. We will also
1
comment on how our results can change by imposing this latter limit.
Degeneracy with the LSP
Even when slepton masses can be large, in which case the chargino cross section is
larger, the chargino mass constraint weakens by a few GeV when the lightest chargino
and neutralino are almost degenerate [34]. The χ˜ 1+ → χ˜ 10 f f¯0 decay leads to soft “visible”
products that are difficult to detect. Recent LEP data has greatly improved the limits in this
small 1Mχ˜ + −χ˜ 0 mass difference region. However within the assumption of gaugino mass
1
1
unification the highly degenerate case with a light chargino/neutralino occurs in the region
µ  M2 , M2 > 2 TeV. In this region the light (and degenerate) neutralino and chargino are
almost purely Higgsino and therefore as seen from Eq. (2.5) do not couple strongly to the
Higgs. Their effect on the Higgs invisible width as well as indirectly on the two-photon
width is negligible. We will not consider this case.
Degeneracy with a light sneutrino
There is another degeneracy which is of more concern to us. It occurs for small slepton
masses that are almost degenerate with the chargino, rendering the dominant two-body
decay mode χ˜ + → ν˜ l + undetectable (the three flavours of sneutrinos are also degenerate).
When this occurs, for ∆deg = mχ˜ + − mν˜e < 3 GeV, neutralino production is also of no use
1
since the neutralinos will also decay into invisible sneutrinos. Since SU(2) relates the mass
of the sneutrinos to that of the left selectrons, the search for the latter will then set a limit
on the charginos in this scenario. The explorable mass of the selectron is a few GeV from
the LEP2 kinematical limit. In fact the LEP Collaborations make a stronger assumption to
relate the mass of the sneutrinos to those of the selectrons. Left and right sleptons masses
are calculated according to a mSUGRA scenario by taking a common scalar mass, m0 ,
defined at the GUT scale. This gives
2
− sin2 θW Dz ,
me2˜R = m20 + 0.15 M1/2

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

11 

2
me2˜L = m20 + 0.52 M1/2
− 0.5 − sin2 θW Dz ,
2
+ Dz /2
m2ν˜e = m20 + 0.52 M1/2

with Dz = MZ2 cos(2β),

(2.8)

where M1/2 is the common gaugino mass at the GUT scale also, which we can relate to
the SU(2) gaugino mass as M2 ∼ 0.825M1/2. With these assumptions, me˜R gives the best
limit. One thus arrives at a limit [34]
mχ˜ + ' mν˜e > 70 GeV (tan β = 5).
1

(2.9)

The above reduction in the chargino mass limit compared to Eq. (2.7) will have dramatic
effects on the Higgs two-photon width. In this very contrived scenario the conclusions we
will reach differ significantly from the general case. This very contrived scenario will be
discussed separately in Section 4.

Neutralino production and decays
LEP2 also provides a constraint on the mass of the neutralino LSP from the search
for a pair of neutralinos, specifically e+ e− → χ˜ i0 χ˜ j0 . This constraint is relevant for the
small (µ, M2 ) and also when we relax the unification condition. We have implemented
the neutralino constraint by comparing the cross-section for neutralino production with
the tables containing the upper limit on the production cross-section for χ˜ 10 χ˜ j0 obtained

by the L3 collaboration at s = 189 GeV [35]. These tables give an upper limit on the
cross-section for the full range of kinematically accessible χ˜ 10 + χ˜ 20 masses. The limits
depend in a non-trivial manner on the masses of the produced particles. Moreover, the
limits are slightly different depending on whether one assumes purely hadronic final states
from the decay of the heavier neutralino or whether one assumes leptonic final states. 5
Under the same assumptions we have also used these tables for setting the upper limit on
χ˜ 10 χ˜ 30 production. In all models where gaugino mass unification is imposed, the virtual Z
0
→ χ˜ 10 Z ∗ , constitutes the main decay mode when the neutralinos are
decay mode, χ˜ 2,3
light enough to be accessible at LEP2. In models where the gaugino mass unification is
relaxed and very light neutralinos exist, as will be discussed in section “no-unification”,
other decay channels may open up, for example χ˜ 30 → χ˜ 10 h. The analysis, and hence the
derived constraints, is made more complicated if one allows for light sleptons as will be
suggested by cosmology in these models. Though light sleptons enhance the neutralino
cross section quite significantly, in the case of left sleptons the efficiency is degraded
because the branching ratio of the heavier neutralino into invisible (through a three body
or even two-body χ˜ 20 → ν ν˜ ∗ ) may be important.
5 The combination of the various selections, the leptonic and hadronic ones, is an a priori optimisation using
Monte Carlo signal and background events. This optimisation procedure is defined to maximise the signal
efficiency (including the leptonic and hadronic branching ratios) and minimise the background contamination.
∞ k(b) P (b, n)/ε, where k(b) is
This consists in the minimisation of κ, expressed mathematically by κ = 6n=0
n
n
the 95% confidence level Bayesian upper limit, P (b, n) is the Poisson distribution for n events with an expected
background of b events, and ε is the signal efficiency including the branching ratios.

12

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

As just discussed one also needs to take into account the various branching ratios
of the neutralinos and charginos. These were also needed when considering production
of neutralinos and charginos at the LHC. We have taken into account all two-body
and three-body decay modes of gauginos, including fermionic and bosonic final states,
˜ and χ˜ + → χ˜ + Z, χ˜ + h, χ˜ 0 W + , ll,
˜ νν.
˜ The analytical
χ˜ j0 → χ˜ i0 Z, χ˜ i0 h, χ˜ ± W ∓ , ll
j
j
i
1
formulas were checked against the outputs of programs for automatic calculations such
as GRACE [36] and COMPHEP [37]. For channels involving a Higgs boson, the
radiatively corrected Higgs mass as well as the Higgs couplings, sin α, following the same
implementation as in [38] were used.

Invisible width of the Z and single photon cross section at LEP2
In the case were we lift the unification condition that leads to rather small neutralino
masses which are kinematically accessed through Z decays we have imposed the limits on
the invisible width of the Z [39]: 

Z
≡ Γ Z → χ˜ 10 χ˜ 10 < 3 MeV.
(2.10)
Γinv
In view of the limits on the single photon cross section which can translate into limits on
σ (e+ e− → χ˜ 10 χ˜10 γ ), with cuts on the photon such that Eγ > 5 GeV and θbeam−γ > 100 ,

we used σ (e+ e− → χ˜ 10 χ˜ 10 γ ) < 0.1 pb at s = 189 GeV. In fact L3 gives a limit of 0.3 pb
[40], foreseeing that similar analysis will be performed for the other collaborations and the
results will be combined we conservatively took 0.1 pb. However this constraint turned out
not to be of much help.
2.4. Cosmological constraints
Scenarios with M1 = M2 /10 have very light neutralino LSP into which the Higgs can
decay, suppressing quite strongly its visible modes. Accelerator limits still allow for such a
possibility. However, it has been known that a very light neutralino LSP can contribute
quite substantially to the relic density if all sfermions are light. In the last few years
constraints on the cosmological parameters that enter the calculation of the relic density
have improved substantially. Various observations [41–47] suggest to take as a benchmark
Ωχ h20 < 0.3 where we identify Ωχ with the fraction of the critical energy density provided
by neutralinos. h0 is the Hubble constant in units of 100 km s−1 Mpc−1 . This constraint is
quite consistent with limits on the age of the Universe [48], the measurements of h0 [49,
50], the measurements of the lower multipole moment power spectrum from CMB data
and the determination of Ωmatter from rich clusters, see [41–47] for reviews. It also,
independently, supports data from type Ia supernovae [51,52] indicative of a cosmological
constant. For illustrative purposes and to show how sensitive one is on this constraint we
will also consider a higher value, Ωχ h20 < 0.6 that may be entertained if one relies on
some mild assumptions based on the age of the Universe and the CMB result only [53].
In this scenario the calculation of the relic density is rather simple since one only has to
take into account annihilations into the lightest fermions. Keeping with our analysis, we

G. Bélanger et al. / Nuclear Physics B 581 (2000) 3–33

13

required all squarks to be heavy but allowed the sleptons to be light. To calculate the relic
abundance we have relied on a code whose characteristics are outlined in [54]. To help
with the discussion we will also give an approximate formula that agrees better than 30%
with the results of the full calculation.
2.5. LHC observables
The principal observables we are interested in are those related to the Higgs production
and decay. Since we are only considering the effects of non-coloured particles and are in
a regime of large MA , all the usual production mechanisms (inclusive direct production
through gluon-gluon as well as the associated production W (Z)h and t t¯h) are hardly
affected compared to a SM Higgs with the same mass. Contrary to the indirect effects
of light stops and/or sbottoms, the main effects we study in this paper affect only decays
of the Higgs. The main signature into photons is affected both by the indirect loop effects
of light enough charginos (and in some cases sleptons) and by the possible opening up of
the Higgs decay into neutralinos. When the latter is open it leads to the most drastic effects
¯ hence posing a problem even
reducing both the branching into photons as well as into bb,
¯
¯
for the search in t t h with h → b b. To quantify the changes of the branching ratios we
define, as in [11], the ratio of the Higgs branching ratio into photons normalised to that of
the SM, defined for the same value of the Higgs mass:
Rγ γ =

BRSUSY (h → γ γ )
.
BRSM (h → γ γ )

(2.11)

Likewise for the branching ratio into bb¯
Rbb¯ =

¯
BRSUSY (h → b b)
.
SM
¯
BR (h → b b)

(2.12)

The latter signature for the Higgs has only recently been analysed within a full ATLAS
simulation and found to be very useful for associated t t¯h production, but only for mh <
120 GeV [5]. With 100 fb−1 the significance for a SM Higgs with mh = 100 GeV is
6.4 but drops to only 3.9 for mh = 120 GeV. Since this is the range of Higgs masses that
will interest us, we will consider a drop corresponding to Rbb¯ = 0.7 to mean a loss of this
signal.
As concerns the two-photon signal, we take Rγ γ < 0.6 as a benchmark for this range
of Higgs masses. This is somehow a middle-of-the-road value between the significances
given by ATLAS [5] and the more optimistic CMS simulations [6].
For the computation of the various branching ratios of the Higgs and its couplings we
rely on HDECAY [38] in which the Higgs masses are determined following the two-loop
renormalisation group approach [26].
Since appreciable effects in the Higgs search occur for relatively light spectra, this means
that the light particles should also be produced at an appreciable rate at the LHC even
though they are electroweak processes. We have calculated, at leading order, all associated
chargino and neutralino cross sections:

Therefore. With heavy sleptons. 3. 1 6 K-factors for these processes have been computed in [55]. (2.5 GeV. neutralino production does not constrain the parameter space any further.5 GeV (tan β = 5). 2. The available parameter space In the case of no-degeneracy of the lightest chargino with the sneutrino. 2. All these limits map into the M2 −µ parameter space for a specific tan β.1) Fig.1. 3. We have therefore calculated all branching ratios for all the charginos and neutralinos. Q2 = sˆ /4.13) Neutralino pair production 6 pp → χ˜ j0 χ˜ k0 is much smaller with the heavy squark masses that we assume. Gauginos masses unified à la GUT 3. It is also possible for the heaviest of these neutralinos to cascade into the lighter ones and the lightest Higgs. (2. 4. LEP2 excluded region (inside the branch) in the M2 −µ parameter space for tan β = 5 and tan β = 30 from mχ˜ + > 94. For the structure function we use CTEQ4M at a scale. the tan β independent limit Eq. These processes have been calculated with the help of CompHEP [37]. 3 The absolute limit on the lightest neutralino for tan β = 5 turns out to be: mχ˜ 0 > 47. Bélanger et al.14 G.7) applies. 3. / Nuclear Physics B 581 (2000) 3–33 pp → χ˜ i± χ˜ j0 i = 1. j = 1. 1 (3. . In principle other means of neutralino/chargino production are possible through cascade decays of heavy squarks. The available parameter space for tan β = 5 and 30 is shown in Fig. if these are not too heavy to be produced at the LHC. the constraint comes essentially from the chargino cross section.

that critical regions are for moderate tan β. as expected. and maximal stop mixing. let us however not forget that especially in the two-photon signal at the LHC the significance increases with increasing Higgs mass. 3. We see that the maximum drop is for tan β ∼ 5.6. 4 and our benchmark Rγ γ > 0. (2. The case with At = 0 and tan β = 5 We now go into more detail and choose tan β = 5 in the case of no mixing. Scanning over µ(> 0).3. but this is mainly due to the loop effects of the charginos. M2 and tan β we show.2. 4. increasing tan β also increases the neutralino masses and reduces the hχ˜ 10 χ˜ 10 couplings as we discussed earlier. Below this value of tan β the Higgs mass is small compared to the neutralino threshold. For the lower limit on the neutralino mass the reduction factor brought about by the β 3 P -wave factor in Eq. One can already conclude on the basis of Fig. The At tan β dependence The above mass of the Higgs for tan β = 5 corresponds to a mixing angle in the stop sector At = 0.4 TeV. We can therefore increase At and tan β. Fig. However. Bélanger et al. (b)At for tan β = 2. 3. The results are summarised in Fig.G. 5. 4. . We have already taken MA = MS = mt˜ = mg˜ = 1 TeV. This said. for mh = 109 GeV. while above this value the LSP gets heavier “quicker” than does the Higgs.1. (a) (b) Fig. / Nuclear Physics B 581 (2000) 3–33 15 Therefore.5−50. the variation with At affects essentially the maximal reduction curve. Moreover. the Higgs coupling to the LSP gets weaker as tan β increases. Obviously to maximise the effect of the neutralinos through the opening up of the Higgs decay into neutralino one should increase the mass of the Higgs. the extremal variation of the Rγ γ as a function of tan β for maximal mixing and taking the available constraints into account. On the other hand the increase Rγ γ > 1 grows with smaller tan β. Rγ γ (shaded area) as a function of (a) tan β for At = 2. tan β ∼ 5. Also. in the non degenerate case there is a very small window for the Higgs to decay into neutralinos.4) factor amounts to about 0.

M2 = 50−300 GeV. All plots are for tan β = 5. observability of the Higgs in this channel. A decrease for positive µ is strongly correlated with the opening up of the little window for h → χ˜ 10 χ˜ 10 . last frame. The latter channel leads to a branching ratio which is at most some 20%. with our benchmark for the branching into bb¯ and thus the channel t t¯h → t t¯b b. µ = 100−500 GeV and At = 0. the largest drop occurs for negative values of µ and is due to the contribution of the light charginos in the two-photon width (see also the dependence with mχ˜ + and M2 ). At this stage one can conclude that the effect of light charginos/neutralinos. First of all note that in this scenario the ratio Rγ γ can vary at most by 15% and that this can lead to either a slight increase or a slight decrease. Contrary to what we will see for other scenarios. 5. Bélanger et al. mχ˜ 0 and Br(h → χ˜ 10 χ˜10 ). Also shown. Rbb¯ > 0.7. the Higgs should still be observed in this channel. Variation of Rγ γ and Rbb¯ vs µ. . mχ + . / Nuclear Physics B 581 (2000) 3–33 (a) (b) (c) (d) (e) (f) Fig. M2 . especially in view of the theoretical uncertainty (higher order QCD corrections) in predicting the signal.16 G. The sign of µ is also that of the interference between the dominant W loop 1 and the chargino loop contribution. 1 1 the variation of Rbb¯ with µ. When this occurs (only for positive µ) it will affect also ¯ However.

It is only in a small region M2 6 160 GeV and µ 6 400 GeV that h → χ˜ 10 χ˜ 10 exceeds 10%.4 TeV. We have therefore imposed the constraint mχ˜ + > 100 GeV. at least at tan β = 5. µ = 100−500 GeV for tan β = 5 in the maximal mixing case. In this case the two-photon rate and the h → χ˜ 10 χ˜ 10 branching ratios are well correlated as shown Fig. can get. The reduction in Rγ γ is no longer more than 80%. Nevertheless one should enquire how large any additional production mechanism. The Higgs branching ratio into neutralinos can reach as much as 40%. / Nuclear Physics B 581 (2000) 3–33 17 is very modest. 3. was also performed. an increase of just a few GeV’s has a drastic effect. The case with maximal At and tan β = 5 Increasing the mass of the Higgs through as large At as possible for the same value of tan β changes the picture quite substantially. it is interesting to enquire about the consequence of an improved lower limit of the chargino masses in the last runs of LEP2. reduction in the usual two photon signals was due essentially to a drop in the main production mechanism through gluons and occurred when the stops developed strong couplings to the Higgs. Also since the largest drops are when the branching ratio of the Higgs into invisible is appreciable. This means that there might be problems with Higgs detection especially in the t t¯h channel. 6(b). At = 2. leading to a reduction of Rγ γ and Rbb¯ of about 60%. When this occurs. A scan over a wider range. Associated chargino and neutralino production at the LHC In our previous study of the effects of light stops [11] on the Higgs search at the LHC. as a lever.4. will only be marginal at the LHC if LEP2 does not observe any charginos or neutralinos before the end of its final run. The contour plots of constant h → χ˜ 10 χ˜ 10 in the M2 −µ plane are displayed in Fig. M2 6 2 TeV and |µ| 6 1 TeV. With our implementation of the corrections to the Higgs mass the increase is about 10 GeV and leaves enough room for h → χ˜ 10 χ˜ 10 in the small µ−M2 region. the reconstruction of the Higgs will be more difficult. this means that even if one triggers Higgs production through charginos and neutralinos. the result of a scan over the parameters M2 = 50−300 GeV. 6(c)–(d). In conclusion. one has large production of stops as well as associated stop Higgs production. the effect of gauginos/higgsinos on the crucial branching ratio of the Higgs. In the present scenario with a common gaugino . with an increase of a few GeV on the lower limit on charginos.5. 6(a). if any. thus recovering a new mechanism for Higgs. Bélanger et al. when one assumes the unification condition and no degeneracy. Furthermore the small window for Higgs decaying into LSP will be almost closed. First the effects are due to weakly interacting particles whose cross sections at the LHC are smaller than those for stops. 3. In the present case uncovering a new effective Higgs production mechanism will be more complicated. As the maximum 1 reduction occurs for the lightest allowed value for the chargino mass. As the results presented here depend critically on the minimum allowed value for the mass of the lightest chargino and neutralino. see Fig.G. The points for larger values of M2 −µ all cluster around Rγ γ ≈ 1 allowing for only a few percent fluctuations.

3. The shaded area is the allowed region. mass at the GUT scale and no (accidental) degeneracy between the chargino and the sneutrinos.18 G. with M2 = 140 GeV.6 (for maximal mixing). Moreover. the rates are more than reasonable in the parameter space that leads to the largest drops. the cross section χ˜ 20 χ˜ 1+ is about 6 pb and is mildly dependent on µ. (c) Variation of Rγ γ with the mass of the LSP Mχ 0 and (d) mass of the 1 chargino Mχ + . 0. µ) region. (a) Contour plot of Br(h → χ˜ 10 χ˜ 10 ) = 0. since the Br(h → bb) about 0. Since the reduction in the usual inclusive two photon channel always occurs in the small (M2 . 4 however.6. 1 1 M2 = 50−300 GeV. 0. 0. all gauginos are relatively light and therefore have reasonable production rates. With the first process. 7 shows. considering the rather large cross section. In fact as Fig. 6. is some 100 fb (with mχ˜ 0 ∼ 250 GeV) when Rγ γ = 0.4 (from right to left respectively) in the M2 −µ plane. the Higgs should ¯ does not drop below be discovered in the usual channels. while production of χ˜ 40 χ˜ 2+ . Rγ γ (and Rbb¯ ) being at worst 0. All plots are for tan β = 5. (b) Correlation between Rγ γ and Br(h → χ˜ 10 χ˜ 10 ). / Nuclear Physics B 581 (2000) 3–33 (a) (b) (c) (d) Fig. and decreases quickly with increasing µ (where. we could use this signature in the cascade decay of the heavier neutralinos and charginos into Higgs.6.4 TeV. Rγ γ increases). Bélanger et al. The vertical line corresponds to mχ˜ + = 100 GeV. µ = 100−500 GeV and At = 2. For instance. .1.2.

(b) as a function of M2 for µ = 200 GeV. In the case of gaugino mass unification without degeneracy with the sneutrinos.G. it should be possible through measurements of the masses and some of the signatures of χ˜ 20 and χ˜ 2+ to get some information on the parameters of the neutralinos and charginos. In the case of degeneracy with the sneutrinos all points in the figure are valid. / Nuclear Physics B 581 (2000) 3–33 19 (a) (b) Fig. Bélanger et al. Associated Production of chargino and neutralino at the LHC at LO (a) as a function of µ for M2 = 140 GeV. 7 we would then know that one might have some difficulty with the Higgs signal in the inclusive channel. the LEP limit means that for M2 = 140 GeV. As for the latter process. . 7 See. for instance. [56]. M2 > 130 GeV. µ > 176 GeV and for µ = 200 GeV. 7. it has more chance to trigger light Higgs than the former.

through cascade decays provide an additional source of Higgs.2 2 1 much as 25% branching ratio for χ˜ 40 → h + anything when Rγ γ is lowest. It rests to see whether a full simulation with a reduced branching ratio of h into b’s can dig out the Higgs signal from such cascade decays. for the parameters of interest. always open is the mode χ30 → χ1. Since in our scenario there is not enough phase space for χ˜ 20 → χ˜ 10 h. without much effect on the decoupling scenario we have assumed. where χ˜ 20 → χ˜ 10 h and h → b b¯ is advocated. We should make another remark. 4. So many gluinos could. Less effective and not 0 h where the branching never exceeds a few percent.2 pb. 8. if we had taken mg˜ = 500 GeV. about 20 fb. Still. which by the way corresponds to a situation where the gaugino mass unification extends also to M3 . Much higher branching are of course possible. For the former one obtains as are potentially interesting: χ40 → χ1. but they occur for higher values of mχ˜ 0 4 where there is no danger for Higgs discovery in the usual modes. The following modes 0 h and χ + → χ + h. the gluino cross section jumps to about 20 pb. However. we see from Fig. therefore. The scan over M2 and µ is as in Fig. Gauginos masses unified à la GUT degenerate with sleptons In the so-called sneutrino-degenerate case where charginos can be as low as 70 GeV. 8. In [25]. the absolute lower limit on the neutralino LSP mass: .20 G. 9 that the largest cross sections originate from the decays of the heaviest neutralino χ˜ 40 while the chargino helps also. Branching ratio of χ˜40 into h for tan β = 5. At = 2400 GeV. As advertised. see Fig. For instance.2 We are now in a position of folding the different branching ratios for the heavier neutralinos and chargino into Higgs (h) with the corresponding cross sections to obtain the yield of Higgs in these channels. the neutralinos themselves are produced through cascade decays of gluinos and squarks which can have large cross sections. / Nuclear Physics B 581 (2000) 3–33 (a) (b) (c) Fig. In our case we have taken these to be as heavy as 1 TeV and thus their cross section is rather modest. 5. Bélanger et al. the yield is quite modest. gluino pair production at the LHC with this mass is about 0.

1) This lower bound rises by roughly 1 GeV for tan β = 2. there is an immediate 1 fall for mχ˜ + < 100 GeV. χ1+ has a cross section in excess of 10 pb! These processes can trigger Higgs production. (4. These modes will probably not help much. / Nuclear Physics B 581 (2000) 3–33 21 Fig. 11. As seen for Rγ γ vs mχ˜ + . In fact. Higgs yield through charginos and neutralinos decays as a function of µ. Results Relaxing the chargino mass by some 20 GeV has quite impressive effects that result in dramatic drops. 9. because of the decays into light sleptons the rates are modest as seen in Fig. But then this should be compensated by the production of plenty 1 of charginos and sleptons while some of the heavier neutralinos and chargino should still be visible. As indicated by Fig.5 GeV (tan β = 5).1.5 and never goes below 34 GeV for larger values of tan β. the largest rates occur when the Higgs has the largest branching into invisible. For these situations clearly the Higgs would be difficult to hunt at the LHC in both the two-photon and (associated) bb¯ channels. . 10. M2 = 140 GeV and tan β = 5 and maximal mixing. 7. 4.G. The subscript for the parentheses ( )j indicates the parent neutralino or chargino. in this situation all charginos and all neutralinos will be produced with cross sections exceeding 100 fb. However. Bélanger et al. The branching fraction into invisibles can be as large as 90%. mχ0 > 34. see Fig. We will only study the case with At maximum.

(f) Variation of Rbb¯ with µ.3 < Rγ γ < 1. (a) Density plot for Rγ γ in the allowed M2 −µ plane. one could make mχ˜ 0 small enough without running into conflict with the chargino mass limits. However in the canonical model with M1 ' M2 /2 and no pathological degeneracy the effect is never a threat. (b) Variation of Rγ γ with µ. 1 The LSP could then be very light and almost bino. µ = 100−500 GeV. 10. / Nuclear Physics B 581 (2000) 3–33 (a) (b) (c) (d) (e) (f) Fig. Largest couplings will be for smallest values of µ which are however.22 G. Bélanger et al. Results in the degenerate scenario with tan β = 5 and maximal At . Basically this is because the almost model independent limit on the chargino translates into values of M2 (hence M1 ) and µ large enough that the neutralinos are not so light that they contribute significantly a large invisible Higgs width. again. 5. (c) with M2 . On the other hand if M1 were made much smaller than M2 . constrained by the . (e) Correlation between Rγ γ and the 1 branching of h into LSP. Relaxing the gaugino mass unification As we have seen. (d) with the mass of the chargino Mχ + .1 from left to right. In all plots the scans are over M2 = 50−300 GeV. To make it couple to the Higgs though one still needs some higgsino component and thus µ should not be too large. having light neutralinos can very much jeopardise the Higgs discovery at the LHC. The different shadings correspond to 0.

The GUT-scale relation which equates all the gaugino masses at high scale need not be valid in a more general scheme of SUSY breaking.1 (5. It is important to stress that the kind of models we investigate in this section are quite plausible. 9 but in the case of a chargino degenerate with the sneutrino and M2 = 120 GeV. though most models in the literature lead rather to r > 1 which is of no concern for the Higgs search. 11. Models with r > 1 would not affect the Higgs phenomenology at the LHC. To investigate such scenarios we have studied the case with M1 = rM2 with r = 0. LEP data already excludes such a neutralino to contribute to the invisible width of the Higgs and therefore the situation is much more favourable to what we have just studied assuming the usual GUT relation. In fact even within SUGRA this relation need not necessarily hold since it requires the kinetic terms for the gauge superfields to be the most simple and minimal possible (diagonal and equal). chargino mass limit for instance. The recent so-called anomaly-mediated SUSY breaking mechanisms [64–71] are also characterised by non-universal gaugino masses. although dilaton dominated manifestations lead to universal gaugino masses. . In superstring models. since their lightest neutralino should be of the order of the lightest chargino. moduli-dominated or a mixture of moduli and dilaton fields lead also to non universality of the gaugino masses [59] and may or may not (multimodulii [61–63]) lead to universal scalar masses.58]. / Nuclear Physics B 581 (2000) 3–33 23 Fig.1) and have limited ourselves to the case with tan β = 5. Bélanger et al. As in Fig. One can easily arrange for a departure from equality by allowing for more general forms for the kinetic terms [57.G.

the outer curve is for me˜L = 150 GeV and me˜R = 100 GeV. LEP2 allowed region in the M2 −µ parameter space in the case M1 = M2 /10 for tan β = 5 and µ > 0. 5. Since there is plenty of phase space for the decay of the lightest Higgs into such neutralinos we will only consider At = 0 for the stop mixing. The available parameter space In the case of heavy sleptons we find that the µ−M2 allowed parameter space is still determined from the chargino mass limit through e+ e− → χ˜ 1+ χ˜ 1− production. Most importantly LSP have masses in the range ∼ 10−20 GeV for the cases of interest. Fig. In fact. the dotted curve is with me˜L = 150 GeV and me˜R = 1 TeV. we have considered three cases: (i) me˜R = 100 GeV with large me˜L . The limits in the case of light sleptons are derived from data on neutralino production. is therefore essentially the same as the one with the GUT relation. The contour plot. 12. The squeezing comes from limits on χ˜ 10 χ˜ 20 and χ˜ 10 χ˜ 30 cross sections properly folded with branching ratios where two-body and three-body decays involving the relatively light sleptons play an important role. / Nuclear Physics B 581 (2000) 3–33 With r = 1/10 the main feature is that the neutralino mass spectrum is quite different. In all cases the leptonic final state signature can be enhanced at the expense of the hadronic signature which usually have a better efficiency.24 G. Bélanger et al. The dashed curve is for me˜R = 100 GeV and me˜L = 1 TeV. this enhancement can be counterbalanced by the fact that a non negligible branching ratio into invisible neutrinos can occur with small enough left selectrons. while light sleptons generally enhance the neutralino cross sections. To illustrate this. (ii) me˜L = 150 GeV with large me˜R .1. (iii) me˜R = 100 GeV. see text. Neutralino pair production χ˜ 10 χ˜ 20 and χ˜ 10 χ˜ 30 . 12. . 12. we show on the same figure. see Fig. Since cosmological arguments will drive us to consider light sleptons masses. although kinematically possible do not squeeze the parameter space further. The inside curve corresponds to the case where all sleptons are heavy (1 TeV) and is derived essentially from the chargino cross section. One sees that. me˜L = 150 GeV. how the µ−M2 parameter space is squeezed in this case. with a very mild M2 dependence. light right selectrons eliminate smallest |µ| values that are otherwise still Fig.

5. 1 . see Fig. although this region does not cover all the available neutralino phase space. however even for chargino masses as high as 200 GeV the drop can (a) (b) (c) (d) (e) (f) Fig. more so for the smallest M2 values. (c) with M2 . These drops are due essentially to a large branching ratio of the Higgs into invisibles. That e˜R do not cut on M2 values can be understood on the basis that they do not have any SU(2) charge. Bélanger et al. (f) Variation of Rbb¯ with µ.3 < Rγ γ < 0. / Nuclear Physics B 581 (2000) 3–33 25 allowed by chargino searches.G. Effects of neutralinos from M1 = M2 /10 with tan β = 5 and At = 0 with heavy selectrons. one carves out an important region. Heavy sleptons The main message is that there are some dangerous reductions in the branching ratios of the Higgs both into photons and into bb¯ which can be only a 1/5th of what they are in the SM. (d) with the mass of the chargino Mχ + . The different shadings correspond to 0. 13. With light left selectrons the gain with respect to the chargino limit is appreciable and occurs across all M2 values.9 (right band). (b) Variation of Rγ γ with µ. The most dramatic reductions occur for chargino masses at the edge of the LEP2 limits.4 (left band) to 0. µ = 100−500 GeV. (a) Density plot for Rγ γ in the allowed M2 −µ plane.2. When both left and right selectrons are relatively light. Since smallest values of µ are the ones that enhance h → χ˜ 10 χ˜ 10 these limits are important.8 < Rγ γ < 0. 13. scans are over M2 = 50−300 GeV. In all the plots. (e) Correlation between Rγ γ and the branching into LSP.

an approximate requirement is q (5.3) Note in passing that Eq. χ˜ 30 have a large higgsino component. thus assuming unification for the scalar masses. specifically the stops. This calls for light (right) sfermions. m0 . neglecting Yukawa couplings one may write me2˜R = m20 + 0. while left sleptons have an added M22 contribution and would then be “much heavier”. As for the gaugino masses to obtain M1 = M2 /10 at the electroweak scale one needs M 1 ' M 2 /5 at the GUT scale. R 1 with all masses expressed in GeV. this means that one has an almost direct limit on m0 . Putting all this together the parameter space still allowed by requiring that the relic density be such that Ωχ h2 < 0.006M 22 − sin2 θW Dz . 8 One knows that for a very light LSP bino the annihilation cross section is dominated by sfermions with largest hypercharge. (5. sufficiently heavy we would not have had large enough radiative corrections to the Higgs mass and would have been in conflict with the LEP2 constraint on the Higgs mass.2) m2l˜ < 103 (Ωχ h2 )max mχ˜ 0 . (5. In these configurations the lightest chargino and χ˜ 20 . Since the limit on the relic density in these scenarios with M1 = M2 /10 constrain essentially the right slepton mass. that is right sleptons [74.3. defined at the GUT scale. We have therefore relied on a full calculation. in the M2 −µ plane the decrease in the ratios is strongly dependent on µ. the bino purity is around 90% for the worst case scenarios. Indeed. Cosmological constraint Considering these large reductions and the fact that the LSP is very light. 10−20 GeV. This scheme leads to almost no running of the right slepton mass.75].48M 22 − (0. M 2 is the SU(2) gaugino mass at the GUT scale which again relates to M2 at the electroweak scale as M2 ∼ 0. We assumed all squarks heavy and took a common mass for the SUSY breaking sfermion mass terms of both left and right sleptons of all three generations. otherwise it would not couple to the Higgs. Bélanger et al. 5.3 and by taking into account all accelerator constraints listed in Section 2 is shown in Fig.825M 2 . This explains why. r3 > 1 at the GUT scale one could make the squarks “naturally heavy” as we have assumed. / Nuclear Physics B 581 (2000) 3–33 reach 60%. m2ν˜e = m20 + 0.48M 22 + Dz /2. me2˜L = m20 + 0. In our case the LSP is not a pure bino. Note also in this respect that had we not taken the squarks.3) can be extended to squarks and if we take M3 = r3 M2 .73] but not from the perspective followed in this paper. As a rule of thumb. 14. since the contribution from the running is of order M12 . with all sfermions heavy but the three right sleptons.26 G.5 − sin2 θW )Dz . The most important message is that sleptons must be 8 Cosmological consequences of non-unified gaugino masses have been investigated in [72. . we investigated whether the most dramatic scenarios are not in conflict with a too large relic density.

Even with the very mild constraint Ωχ h2 < 0. µ = 100−500 GeV and m0 = 50−1000 GeV. . (c) me˜R .2 would have meant me˜R < 125 GeV. (a) Allowed region in the M2 −µ plane. 14. The grey areas represent the allowed parameter space after including the constraints √ from neutralino cross sections at s = 189 GeV while the extra dark area does not include the constraint on neutralino production. As expected the latter cut on smallest µ values (and also a bit on smaller M2 values). but also cut on the amount of the higgsino component in χ˜ 10 and thus on the contribution of χ˜ 10 to the invisible decay of h. Ωχ h2 < 0. (5. We therefore see that a combination of LEP2 neutralino cross sections with improved constraints from the relic density are important. Had we imposed a lower Ωχ h2 . that not only allow accessible χ˜ 20 . All points that pass the constraint Ωχ h2 < 0.G. In these plots. Bélanger et al.2) is therefore quite good and explains the various behaviours of Fig. (d) Scatter plot of µ vs me˜R . χ˜ 30 . / Nuclear Physics B 581 (2000) 3–33 (a) (b) (c) (d) 27 Fig. The same figure also shows the effect of not taking into account the constraint from the LEP2 neutralino cross sections. lighter than about 140 GeV.6 in the case M1 = M2 /10 with tan β = 5 and At = 0.6 right selectron masses are below 160 GeV. 14. The approximate rule of thumb given by Eq. (b) Ωχ h2 vs µ. scans are over M2 = 50−300 GeV.

3). Effects of neutralinos from M1 = M2 /10 with tan β = 5 and At = 0 with light selectrons. 5. (c) with the mass of the chargino Mχ˜ + . Bélanger et al. . The different shadings correspond to 0.3 < Rγ γ < 0. In all plots the scans are over M2 = 50−300 GeV. Right and left charged sleptons of equal masses contribute almost equally and interfere destructively with the dominant W loop hence reducing the width h → γ γ . Once again large drops are possible with reduction factor as small as 0.9 (right band). (e) Variation of Rbb¯ with µ.8 < Rγ γ < 0.4.3 in both the branching ratio of ¯ the Higgs into photons and bb.28 G. (a) Density plot for Rγ γ in the allowed M2 −µ plane. µ = 100−500 GeV. (5. The masses are calculated according to Eq. (b) Variation of Rγ γ with µ. / Nuclear Physics B 581 (2000) 3–33 (a) (b) (c) (d) (e) (f) Fig. 1 (d) Correlation between Rγ γ and the branching into LSP. 15.4 (left band) to 0. (f) Rbb¯ vs the branching of the Higgs into the LSP neutralino. Light sleptons We now allow for light sleptons with masses such that ml˜ > 90 GeV but take into account all cosmological and accelerator constraints. Although one has reduced the µ−M2 parameter space somehow one has also allowed for light sleptons that indirectly contribute to h → γ γ beside the light charginos.

17 that. 9 Fig. They account for some 10−15% drop as can be seen when h → χ˜ 10 χ˜ 10 is closed and by comparing with the heavy slepton case. This also occurs in models that do not assume the GUT inspired gaugino mass. / Nuclear Physics B 581 (2000) 3–33 29 The loop effects of the sleptons are rather marginal compared to the effect of the opening up of the neutralino channels. through essentially χ˜ 30 decays. apart from the charginos and neutralinos. The loop effects of charginos in the two-photon width are small compared to the theoretical uncertainties. could probably allow the determination of the parameters of the higgsino–gauginos sector “sending an early warning signal” that indicates difficulty in the detection of the Higgs.5. 5.6. suppressing all other signatures. These in turn contribute quite significantly to the cross section for neutralino production at LEP2 which then constrains the parameter space in the gaugino–higgsino sector where the invisible branching ratio of the Higgs becomes large. specifically those where. we have shown that current LEP limits on charginos imply that there should be no problem finding the lightest SUSY Higgs at the LHC in the two-photon mode or even bb¯ in the associated t t¯h channel. In scenarios “on the fringe” with a conspiracy between the sneutrino mass and the lightest chargino mass. at the weak scale. 16 shows that. Conclusions In a model that assumes the usual common gaugino mass at the GUT scale and where. all other supersymmetric particles are heavy. The LEP data in this scenario mean that the decay of the Higgs into invisibles is almost closed. Associated chargino and neutralino production In cases where there are very large reductions in the usual bb¯ and γ γ signatures of the Higgs. as constrained from cosmology in these scenarios. This is because the (invisible) Higgs decay into light neutralinos may become the main decay mode. it is in these regions with highest yield that the Higgs has a large branching ratio into invisible and would be difficult to track. If we now look at the (lightest) Higgs that can be produced through cascade decays in these processes. However we point out that limits from the relic density in these types of models require rather light right selectron masses. Although large reductions in the usual channels are still possible. all neutralinos and charginos can be produced. they amount to less than about 15% and can either increase or decrease the signal. the cross section for χ˜ 40 χ˜ 2+ is in excess of 100 fb while χ˜ 20 χ1+ is above 1 pb. for values of µ−M2 where Rγ γ is below 0. again. the Higgs signal can be very much degraded in both the two-photon and the b final states. associated Higgs cross sections of about 30 fb are possible. the U (1) gaugino mass is much smaller than the SU(2) gaugino mass.G. is on the other hand quite modest at the LHC. For instance with M2 = 250 GeV. . production of charginos and neutralino at the LHC is quite large. Bélanger et al. 6. the combination of LEP2 data and cosmology means that observation of the Higgs signal 9 Production of light sleptons. one sees from Fig. Therefore early observations of these events. Nonetheless.

Some of the heavier of these particles may even trigger Higgs production through a cascade decay into their lighter partner. Moreover. one is assured of having a quite healthy associated chargino and neutralino cross section at the LHC. at the LHC is jeopardised in only a small region of the SUSY parameter space. Associated Production of chargino and neutralino at the LHC at LO for M2 = M1 /10 (a) as a function of µ for M2 = 250 GeV. 16. Bélanger et al. .30 G. It rests to see whether the Higgs can be seen in this new production channels. considering that it will predominantly have an “invisible” signature. (b) as a function of M2 for µ = 150 GeV. / Nuclear Physics B 581 (2000) 3–33 (a) (b) Fig. we show that in these scenarios where the drops in the Higgs signals are most dramatic.

hep-ph/9806315. [9] G.ps. P. Osland. November 1999. K. Marco. D. J. http://delphiwww. March 1999. ATLAS Technical Design Report.ch/ALPUB/seminar/ lepc_nov99/lepc.cern.P. Phys. Phys. Rahal-Callot. Bélanger. http://l3www. Phys.L. D 50 (1996) 213. 1999.ps. Djouadi. G.pdf.ps. [8] B. Kribs.D. [5] ATLAS Collaboration. Phys. http://atlasinfo.ch/Atlas/ GROUPS/PHYSICS/HIGGS/higgs-www/Analyses-notes. Pandita. [7] B. As in Fig. Martin.gz. Kileng. [6] R. Lett. Z. Acknowledgements We would like to thank Sacha Pukhov and Andrei Semenov for help and advice on the use of CompHEP.cern. chõffline/physics_links/lepc. Nucl. hep-ph/9506455. ch/conferences/ps/RahalCallot_LEPC_L3_Nov99. http://alephwww. This work is done under partial financial support of the Indo-French Collaboration IFCPAR-1701-1 Collider Physics.html.D. Aleph talk for the LEP Committee. L3 talk for the LEP Committee. April 1997. hep-ph/9904348.cern. ch/Opal/talks/pward_lepc99.N. Phys. Z. Kileng.cern. Blondel. http://www1. 9 but with M2 = 250 GeV and M1 = M2 /10. C 71 (1996) 87. Wells. B 568 (2000) 3.G. November 1999. Bélanger et al. November 1999. Boudjema. B 435 (1998) 101. F. [2] J. [4] P. [3] G. acknowledges the hospitality of LAPTH where this work was initiated.html/higgs-sugra.G. CMS Note 1997/057. References [1] A. Denegri. Delphi talk for the LEP Committee. [10] A. S. Kane.cern. Ward. . Kinnunen. R. Opal talk for the LEP Committee. P.gz. Rev. [11] G. C 63 (1993) 87. / Nuclear Physics B 581 (2000) 3–33 31 Fig. 17.gz. Sridhar.

M. hep-ph/9907422. to be published in the David Schramm Memorial Volume. D 48 (1993) 4062. [39] LEP Electroweak Working Group. Astrophys. Lett. Lett. Phys. [41] W. Phys. Astrophys. J. Moretti. hep-ph/9704448. Rev. Rev. 80 (1998) 1830. astro-ph/9912211.R. Phys. Lett. [47] C.G. Haber. [43] N. Lineweaver. Proc. Djouadi. H. [29] R. to appear in the Nat. Phys. Freedman. [45] M. Phys.E. Carena. J. Lett. Perlmutter et al. [15] ALEPH Collaboration. astro-ph/9804082. Phys. B 440 (1998) 296. [49] A. [26] M.M.. D 50 (1994) R4244. Phys. Freedman. Kneur. Carena. Bahcall. [48] C. [14] M. Riess et al. Lett. Phys. Djouadi.E. [21] A. Kalinowski. astro-ph/9911493. J. hep-ph/9903218. G. R. [28] S.M.L. Addison–Wesley.L. [50] W. Lett. B 331 (1994) 99. Zerwas.3. [27] M. hep-ph/0001002. C 9 (1999) 343. White. 460 (1996) L15. Zhang. Commun. B 376 (1996) 220. Frederiksen.. Djouadi. C. K.. Wagner. 1990. Nucl. B. P.. Phys. Lett. J. Wagner. J. Phys. Gunion. S. / Nuclear Physics B 581 (2000) 3–33 [12] A. hep-ph/9904491. S. Comp. Weiglein. X. Astron. M. Turner. in press. Wells. T. Eur. R. P. astro-ph/9805326. Y. Moultaka. Rev. J.. [32] J. S. The GRACE manual Ver. CERN. Reid. J.M. Lopez et al. D 60 (1999) 035006.cern.. The Higgs Hunter’s Guide. hep-ph/9711244.E. Rev. Eur.html (Talks given at the International Conference of Cosmology and Particle Physics (CAPP98). CompHEP user’s manual.F. Lineweaver. G. J. astro-ph/9911493.M. [42] M.ch/LEPEWWG/ [40] L3 Collaboration. Acad. M. Lett. June 1998). D 60 (1999) 015007. B 461 (1996) 407. v3. in press. B 322 (1994) 368. in press. Zerwas.P. Wagner. Rev. [35] L3 Collaboration. Lett. CERN-EP/99-129. hep-ph/9912336. S. [19] S. Carena. Phys. Djouadi. A 14 (1999) 359. hep-ph/9903404. Wagner.R. J. Woodside. Kato. [38] A. Heinemeyer. J.A. G. G. Rev. M. J. Kaneko. L3 preprint 201. Phys. Spira. Griest. Lett. W.M. KEK Report 92-19. Phys. X. Janot. 1993. Espinosa. N. Perlmutter. B. Ishikawa.F.P. K. hep-ph/9807423.L. http://www. Preprint INP MSU 98-41/542. hep-ph/9912223.. Kawabata. [13] A. B 455 (1999) 179. Redwood City. Lineweaver. Pukhov et al. Choudhury. J. hep-ph/9903259. Rep. Astrophys. R. Bisset.J. Phys. Martin. [20] K. Acciarri et al. Z. Lett. Quiros. Kirshner. [33] J. Heinemeyer. [22] J. Lineweaver. Hempfling. C.M. [25] H. hep-ph/9903382. J. Acciarri et al. P. Phys. 506 (1998) 495. Phys. Phys. astro-ph/9906463. J. Phys. Astrophys. D 37 (1988) 719.E. [24] A. [46] N. Acciarri et al. Rev. S. submitted to Phys.A. [18] J.. B 355 (1995) 209. [23] A. [31] M. Dawson.P.E. Bahcall. H. Phys. Johnson. Sandage et al. Lett. Roy. 1. Fan. Kalinowski.G. Phys. Sci. 1998. [51] S. Rev. hepph/9908288.0. Espinosa. CA. CERN-EP/99-127. Kane. hep-ph/9803277. 72 (1994) 199. S. M. hep-ph/9812328. astro-ph/9909076. hep-ph/9909526. Haber. Shimizu.J. [36] T. C. hep-ph/9812472. C.E. Hollik. Gunion.. Phys.D. Phys. [37] A. Quiros. Djouadi. M. B 569 (2000) 53. Carena. Rev.cern. Mrenna.L. C 57 (1993) 569. Tata. [34] L3 Collaboration. 108 (1998) 56. 505 (1998) L69.. Phys. D 46 (1992) 303. Hoang. B 470 (1999) 268. Bélanger et al.. [17] D. P. wwwth. Nucl. D. Steinhardt. M. astro-ph/9904051. Mod. Kane. [16] S. [52] A. C 10 (1999) 515. [30] M. Weiglein. A. [44] C. Lett. Tanaka. Baratte et al. Baer. CERN-EP/99-125. Dedes. G. J.ch/capp98/programme. .. 116 (1998) 1009. D 58 (1998) 091701. Phys. astro-ph/9812133. C. Phys. Carena et al. J.32 G.. Kalinowski. Ostriker.

B 158 (1985) 409. Gherghetta. Wells.D. Pomarol. Drees. Ibáñez. Z.D. M. C. D. Jungman. M. [54] A.A. Beenakker et al. Phys. D 47 (1992) 376. [59] A. hep-ph/9910202. [68] Z. Sundrum. R. [57] J. V. Griest. X. Moroi. Maksymyk. Astropart. D 46 (1992) 3309. Sundrum. Lett. D. [63] Y. Phys. [55] W. Ibáñez. N. Phys. Yamada. hep-ph/9906298. Kobayashi. H. G. Rattazzi. R.E. hep-ph/9905390. Griest. Brignole.M. Rev. B 348 (1995) 402. Scheich. Murayama. Rep. Kamionkowski. T.A. Phys. Mignola and S. Nucl. Chacko. Luty. Brignole. B 155 (1985) 381. Phys. Phys. Phys.D. Rattazzi.E. [64] L. B 559 (1999) 27. Kobayashi. Rev. Wells. Luty. C 74 (1997) 157. T. JHEP 9905 (1999) 5. B 502 (1997) 37. [70] J. hep-ph/9508258. Y. [56] M. Enqvist. hep-ph/9903448. Khalil. Nucl. JHEP 9812 (1998) 27. Phys. Giudice. Roszkowski. [67] T. and references therein. [62] A.G. Rattazzi. 267 (1996) 195. Wells.) together with N.V. Muñoz. and references therein. M. Nojiri. [72] M. Phys. B 422 (1994) 125. Rev. Scopel). M. Mignola.F. Nucl. de Alfaro.A. hep-ph/0001267. Kobayashi. hep-ph/9809504. Rev.E. Phys. A. L. G. M. [58] M. Tata. [61] T. Nucl. Suematsu. Fornengo. C. Phys. / Nuclear Physics B 581 (2000) 3–33 33 [53] J. [66] A. K. Lett. Yamagishi. Phys. [60] A. Bottino. K. Nojiri. hep-th/9810155. Phys. Ellis. K. S. [69] M. Phys. Giudice. hep-ph/9703239. R. K.. D. Brignole. . Muñoz. 83 (1999) 3780. Bagger.D. Strumia. Phys. Lett. Muñoz. Lett. Phys. J. [74] M. I. Ibáñez. hep-ph/9707209. Lett. [75] G. E. Kawamura. T. Nucl. B 557 (1999) 79. Drees. hep-ph/9904378. R. D 43 (1991) 1971. Rev. Toya. [73] K. G.M. Pignone. Poppitz. Luty. Randall. D 33 (1986) 1486. L. hep-ph/9903448. Tamvakis. hep-th/9911029. B 443 (1998) 196. Drees. Bélanger et al. [71] R. C. hep-ph/9912390. Ponton. Phys.F. J. L. Fornengo.A. 2 (1994) 67 (The code we used here was developed by one of the authors (F. L. C. [65] G. E. B 436 (1995) 747. Nanopoulos.

nl/locate/npe Associated H − W + production in high energy e+e− collisions A.b .X . 2 CP-even H .fr.3 2 1 3 ( 0 0 ) 0 0 1 9 8 . 416. Faculty of Sciences. Arhrib a. F-34095 Montpellier Cedex 5. Introduction The discovery of the standard Higgs boson is one of the major goals of the present and future searches in particle physics. Hollik d . Moultaka c a Département de Mathématiques. D-76128 Karlsruhe.Bx. Moreover. Morocco c Laboratoire de Physique Mathématique et Théorique. From the 8 degrees of freedom initially present in the 2 Higgs doublets. Capdequi Peyranère c. M.Fr.80. Rare decay PACS: 12.Cp 1.∗ . Charged Higgs boson. The two most popular versions of the THDM (type I and II).15. We present the analytical results and give a detailed discussion for the total cross section predicted in the context of a general Two Higgs Doublet Model (THDM).V. Faculté des Sciences et Techniques. the problematic scalar sector of the Standard Model (SM) can be enlarged and some simple extensions such as the minimal Two Higgs Doublet Model (THDM) versions [3] are intensively studied.Lk. 12. Universität Karlsruhe. CNRS-UMR. W. Tanger. B. accepted 27 March 2000 Abstract We study the associated production of charged Higgs bosons with W gauge bosons in high energy e+ e− collisions at the one loop level.P.V. is noticeably different from the other SM or THDM Higgs particles. 3 correspond to masses of the longitudinal gauge bosons and we are left with 5 degrees of freedom which manifest themselves as 5 physical Higgs particles (2 charged Higgs H ± .Nuclear Physics B 581 (2000) 34–60 www.38. 5825 Université Montpellier II. Physics Department. Fax: 33 4 67 144234 0550-3213/00/$ – see front matter  2000 Elsevier Science B. Direct or indirect results give stringent lower and upper bounds on its mass [1. France d Institut für Theoretische Physik. All rights reserved.60. Germany Received 26 January 2000. Morocco b UFR–High Energy Physics. The charged Higgs H ± . differ in some Higgs couplings to fermions. 12. All rights reserved. PII: S 0 5 5 0 . PO Box 1014. E-mail: micapey@lpm. the Higgs spectrum is the same.elsevier. G. 14.  2000 Elsevier Science B. Keywords: THDM. but in both types and after electroweak symmetry breaking [4]. h and one CP-odd A). Rabat.2].univ-montp2. In other words the discovery of a charged ∗ Corresponding author. Phone: 33 4 67 143512. because of its electrical charge.

scalar bosons and fermions). In our study. loop or/and threshold effects can give a substantial enhancement. Arhrib et al. This process is kinematically better suited than the H ± pair production when the Higgs mass is larger than the W mass. qb → q 0 bH − [22– (i) single charged Higgs production via gb → tH − .11] and shown that e+ e− machines will offer a clean environment and in that sense a higher mass reach.11]. the simplest way to get a charged Higgs is through H ± pair production. Section 3 contains the notations and conventions while Section 4 is devoted to the on-shell renormalization . especially if the 1–2 TeV options are available. the physical predictions are sometimes very different from typical ones in the SM and or the THDM. Note also that charged Higgs bosons can be produced in hadronic machines: ¯ − .).A. / Nuclear Physics B 581 (2000) 34–60 35 scalar Higgs boson would attest definitively that the Standard Model is overcome. Moreover. In models of the HTM type. The aim of this paper is the study of the production of the charged Higgs boson in association with a W gauge boson through electron positron annihilation. Of course. the particle spectrum of the MSSM doubles that of the THDM. On the contrary a neutral Higgs experimental evidence has to be refined to go beyond the SM [2. one can construct models with an even larger scalar sector than in the THDM. once worked out. but the scalar Higgs sectors. H ± → h0 W ± . A noteworthy difference between THDM and HTM is that the HTM contains a tree level ZW ± H ∓ coupling while in the THDM this coupling is only generated at one loop level [3. otherwise one would have to look at bosonic decay modes (e. any experimental deviation from the results within such a model should bring some fruitful information on the new physics.9]. In any case. etc. one of the most popular being the Higgs Triplet Model (HTM) [6]. up to some mass constraints and couplings in the MSSM. we will only consider models with two Higgs doublets. Although it is a rare process in the THDM. The paper is organized as follows. Furthermore. the study of the various production mechanisms of charged Higgs bosons would give information about new physics beyond the SM.6].8. We mention also that charged Higgs bosons pair production through laser back-scattered γ γ collisions has been studied in the literature [19–21] and found to be prominent to discover the charged Higgs boson. responsible for the electroweak symmetry breaking. H ± → A0 W ± . The study of THDM is also very interesting insofar as it provides a general framework to test the minimal supersymmetric extension of the SM (MSSM) [10. we review all the charged Higgs boson interactions (with gauge bosons. are analogous.g. 5. gg → t bH 27] (ii) single charged Higgs production in association with W gauge boson via gg → W ± H ∓ or bb¯ → W ± H ∓ [28. In Section 2. Such studies have been already undertaken at tree-level [9] and one-loop orders [10.. In future colliders or linear accelerators devoted to the electron–positron annihilations.29] (iii) H ± pair production through q q¯ annihilation would be feasible only if the charged Higgs decay in top–bottom is kinematically forbidden [30–33].

7). The combination v12 + v22 is thus fixed by the electroweak scale through v12 + v22 = v22√ (2 2 GF )−1 . From Appendix A. and we are left with 7 free parameters in Eq. The trilinear vertices H 0 H + H − and h0 H + H − depend. 2 2mW sin 2β λ6 = g 2 m2A 2m2W . namely λi and tan β = v2 /v1 .1). besides on the Higgs boson masses. (2. Eqs. A. where g and g 0 are the SU (2)weak and U (1)Y gauge couplings and v 2 = v12 + . (A. (2. λ5 =  g 2 sin 2α 2 mH − m2h − 4λ3 . We will assume the arbitrary additive constant λ7 to be vanishing. using straight-forward algebra. The trilinear self couplings required for our study are listed in the Appendix A. H ± [16] in the following way:    g2 2 2 2 2 cos(2α + β) + m + m − m + λ3 (−1 + tan2 β). Φ2 = λ1 Φ1+ Φ1 − v12 + λ2 Φ2+ Φ2 − v22 + λ3 Φ1+ Φ1 − v12 2      + Φ2+ Φ2 − v22 + λ4 Φ1+ Φ1 Φ2+ Φ2 − Φ1+ Φ2 Φ2+ Φ1   2  2 (2. mH . (2. In our analysis we will take into account the following constraints when the masses and the coupling parameters are varied: . and Section 6 contains our conclusions.2).36 A. one can relate the coefficients λi of the scalar potential to the masses of the physical Higgs bosons h. H. the corresponding vacuum expectation values v1 and v2 . In Section 5 we present our numerical results. Via the Higgs mechanism. Eqs. 2.1).3) + m + m − m m λ2 = H h h H sin β 16 sin2 βm2W λ4 = g 2 m2H ± 2m2W .1)–(A. As shown in [11]. it is obvious that one would get into conflict with the requirements from perturbative unitarity when tan β and/or λ3 are large. / Nuclear Physics B 581 (2000) 34–60 scheme we will use. mh .1. Recalling the most general (dimension 4) SU (2)weak × U (1)Y gauge-invariant and CP-invariant scalar potential for the THDM  2 2   V Φ1 . Arhrib et al. [16]. mA ). α. (A. (A.2) m λ1 = H h H h cos β 16 cos2 βm2W    g2 2 2 2 2 sin(2α + β) + λ3 (−1 + cot2 β). and the coefficients λi as real-valued parameters. the W and Z gauge bosons acquire masses given by m2W = 12 g 2 v 2 and m2Z = 12 (g 2 + g 0 2 )v 2 . on λ3 and tan β. Charged Higgs boson interactions 2. (2.4) Consequently one can choose as free parameters of the Higgs sector: the four Higgs boson masses (mH ± . Charged Higgs boson interactions with scalar bosons In this section we will follow the notation of Ref. we have the doublet fields Φ1 and Φ2 with weak hypercharge Y = 1. tan β and λ3 .1) + λ5 Re Φ1+ Φ2 − v1 v2 + λ6 Im Φ1+ Φ2 + λ7 .

Models I and II lead to similar results as far as the effects originating from the top–bottom loops for tan β close to 1 are considered. with the coupling structure copied from the Standard Model. . invoking the charge-conjugate of Φ2 for the uptype quarks.0017 6 δρ 6 0. • Lower bounds on tan β in the general THDM have been obtained from the experimental limits on the processes e+ e− → Z ∗ → h0 γ and/or e+ e− → A0 γ [36. Conventionally they are classified in terms of the following categories. (2. Φ2 gives mass to both up. Arhrib et al.7) Vud is the CKM matrix element.A. 37]. which we will approximate by unity.37].0027. For our study we will restrict the discussion to values tan β > 0. one assumes that Φ1 couples only to down-type quarks and charged leptons and Φ2 to up-type quarks (and eventually to neutrinos). (2.and down-type quarks. The type II model is the pattern found in the MSSM.5) on each trilinear scalar vertex H H H entering the process at one loop. these bounds can be rather low [36. following [17]. for model II. The general structure of the charged-Higgs boson interaction with a doublet of up. the Yukawa couplings have the form mu tan β mu = tan β md tan β L = Yud R and Yud =− L Yud R and Yud = md tan β for model I. 2. type I models cannot be realized within the MSSM.2.and down-type fermions is given by the vertex    −  igVud L (1 − γ5 ) R (1 + γ5 ) + Yud . owing essentially to the form of the Yukawa couplings Y R and Y L . Charged Higgs boson interactions with fermions In the two-Higgs-doublet extension of the Standard Model there are different ways to couple the Higgs fields to the fermions. For light h masses. by imposing the condition |H H H | < 3g (1 TeV)2 2mW (2. (ii) Model type II: To avoid the problem of flavor changing neutral currents (FCNC) [21].6) Yud H ud¯ = √ 2 2 2 mW In the specific models mentioned above. • The extra contribution δρ to the ρ parameter [19] should not exceed the current limits from precision measurements [20]: −0. / Nuclear Physics B 581 (2000) 34–60 37 • Unitarity constraints will be respected in a simplified way. Since supersymmetry forbids the appearance of the complex conjugate.5. labeled as type I and type II models: (i) Model type I: All quarks and leptons couple exclusively to the second Higgs doublet Φ2 .

p2 . tree-level contributions to e+ e− → W ± H ∓ come only from neutrino exchange in the tchannel and from CP-even Higgs mediated s-channel. The Mandelstam variables are defined as follows: s = (p1 + p2 )2 = (k1 + k2 )2 . The momenta of the incoming electron and positron. ±κ sin θ. Notations and conventions We will use the following notations and conventions. are listed in Appendix C. respectively. the process under study is essentially a loop-mediated process. These interactions follow from the kinetic term. t = (p1 − k1 )2 = (p2 − k2 )2 = . involving the covariant derivative. k1. g 0 ) the associated coupling constants. ±κ cos θ . p1.  s s 1 2 m + m2H ± − + κ cos θ.2 = 2 √   m2 − m2H ± s . 0. and g (resp. (2. 2 2 2 s + t + u = m2W + m2H ± . the momenta. The γ H ∓ W ± and the ZH ∓ W ± vertices vanish at the tree level. All these contributions are strongly suppressed by the small electron mass. Therefore.2 = 1± W 2 s √ where s denotes the center of mass energy. 3. Wµa denote the SU (2)L gauge fields. Bµ the U (1)Y gauge field. outgoing gauge boson W + and outgoing Higgs boson H − are denoted by p1 . 0.8) ∂µ + ig Ta Wµ + ig ∂µ + ig Ta Wµ + ig = 2 2 i where TEa is the isospin operator and YΦi the hypercharge of the Higgs fields. The individual couplings in the various vertices.3. Hence.38 A. in the e+ e− center-of-mass system. θ the scattering angle between e− and W + . and κ is determined by   κ 2 = s − (mH ± + mW )2 s − (mH ± − mW )2 /s 2 . are given by: √ s (1. Neglecting the electron mass. ±1). k1 and k2 . Charged Higgs boson interactions with gauge bosons The interactions between the charged Higgs bosons of a two-doublet model and the gauge bosons are completely dictated by local gauge invariance and thus independent of the assumption whether the model is supersymmetric or not. 2 W 2 2  s s 1 u = (p1 − k2 )2 = (p2 − k1 )2 = m2W + m2H ± − − κ cos θ. derived from Eq. Arhrib et al. / Nuclear Physics B 581 (2000) 34–60 2. (2.8). 0. in the Higgs Lagrangian X +  Dµ Φi Dµ Φi i  +    X 0 YΦi 0 YΦi a a E E E E Bµ Φi Bµ Φi .

expressed in terms of the one-loop amplitude M1 :  κ X .A. Arhrib et al. / Nuclear Physics B 581 (2000) 34–60 39 The differential cross section reads.

.

1 .

.

2 1 − γ5 u(p1 ).1) M .ε(k1 ) . ¯ 2 )/ε(k1 ) A2 = v(p 2  1 + γ5 u(p1 ) p1 .ε(k1 ) .ε(k1 ) .3) A6 = v(p 2 The squared amplitude.2 dσ + − e e → W ±H ∓ = (3. dΩ 64π 2s Pol can be decomposed into six invariant matrix elements Ai and scalar coefficient functions Mi as follows: M1 M1 = 6 X Mi Ai . ¯ 2 )/k1 (3. gets the following form:  X.2) i=1 The basic matrix elements Ai are given by the following expressions: 1 + γ5 u(p1 ).ε(k1 ) . ¯ 2 )/k1 A4 = v(p 2  1 + γ5 u(p1 ) p2 . ¯ 2 )/k1 A5 = v(p 2  1 − γ5 u(p1 ) p2 . after summation over the polarizations of the W + boson. ¯ 2 )/k1 A3 = v(p 2  1 − γ5 u(p1 ) p1 . (3.

.

 (m2 m2 − tu) .

M1 .

The types of Feynman diagrams are depicted in Fig.2 = 2s |M1 |2 + |M2 |2 − H ± W 4 |M1|2 + 4 |M2|2 2 4m W Pol   2    2 ∗ + 4 mW − t Re M1 M3 + M2 M∗4 + m2W − t |M3 |2 + |M4 |2   2    + 4 m2W − u Re M1 M∗5 + M2 M∗6 + m2W − u |M5 |2 + |M6 |2     − 2 m2H ± m2W + m2W s − tu Re M3 M∗5 + M4 M∗6 .2). and using dimensional regularization [41. according to the decomposition (3. It displays the corrections to the γ –W ± H ∓ and to the Z–W ± H ∓ vertices. box diagrams. Note that the s-channel . contributions coming from the various mixings H − – W + and H − –G+ in t and s channels. On-shell renormalization We have evaluated the one-loop induced process e+ e− → W ± H ∓ in the ’t Hooft– Feynman gauge. 1.42]. (3.4) ¯ 2 )/ε(k1 ) A1 = v(p The coefficient functions Mi can be read off from the explicit expressions in Appendix C. 4. and finally the counter-terms.

H and tadpole counter-terms δh. 1. The Higgs fields and vacuum expectation values vi are renormalized as follows: Φi → (ZΦi )1/2 Φi . Although the amplitude for our process vanishes at the tree level.6) ReΣ H (the b labels the renormalized quantity which is given in Appendix C). / Nuclear Physics B 581 (2000) 34–60 H − –W + mixing (Fig.40 A. These conditions guarantee that v1. In this scheme. TH + δtH = 0. Arhrib et al. All the Feynman diagrams have been generated and computed using FeynArts [23] and FeynCalc [24] packages.2 is assigned to each Higgs doublet Φ1. for the Higgs sector. i.   µ (4. We also use the fortran FF-package [25] in the numerical analysis.H vanish: Th + δth = 0. 1.. we obtain all the counter-terms relevant for our process: a counter-term for the W − H 2-point function. vi → (ZΦi )1/2 (vi − δvi ). ∆= 2 v1 v2 2 (4. (4. This renormalization condition determines the term ∆. expanding the renormalization constants Zi = 1 +δZi to the one-loop order. the sum of tadpole diagrams Th.16) vanishes for an on-shell transverse W gauge boson. The explicit expressions are given in Appendix C. the following renormalization conditions are imposed: • The renormalized tadpoles. (4.2 .e. . We adopt the on-shell renormalization scheme of [26].4) δ Zν Wµ H = iegµν cW with   δv1 δv2 1 sin 2β mW − + (δZΦ2 − δZΦ1 ) . complications like tadpole contributions and vector boson-scalar mixings require a careful treatment of renormalization.5) In the on-shell scheme.   sW ± ∓ ∆.2 appearing in the renormalized Lagrangian LR are located at the minimum of the one-loop potential. visualized in Fig. 1. There is no contribution from the W + –G− mixing because the γ –G± H ∓ and Z–G± H ∓ vertices are absent at the tree level.2) δ Wν± H ∓ = −k1 ∆.26 to Fig.8).3) δ Aν Wµ H = iegµν ∆.28 (k1 denotes the momentum of the outgoing W + ). bH − W + (k 2 ) vanishes for • The real part of the renormalized non-diagonal self-energy Σ an on-shell charged Higgs boson:  bH − W + m2 ± = 0 (4. which is an extension of the on-shell scheme in [27]. and consequently δ[Aν Wµ± H ∓ ] and δ[Zν Wµ± H ∓ ] are also fixed. A field renormalization constant ZΦ1. and counter-terms for the γ W H and ZW H vertices.1) With these substitutions in the Lagrangian (2. field renormalization is performed in the manifest-symmetric version of the Lagrangian.   ± ∓ (4.

/ Nuclear Physics B 581 (2000) 34–60 41 Fig. 1. Feynman diagrams relevant for the e+ e− → W + H − . Fig. Arhrib et al. 1.11 are vertex diagrams.1–1. Fig.12 is box diagram. . 1.A.

1. Fig. Fig.21–1.42 A. Arhrib et al. Continued.25.26 and 1. / Nuclear Physics B 581 (2000) 34–60 Fig. 1.28 are the counter-terms for the photon–W + –H − and Z–W + –H − vertices. 1.20. 1. self-energies of the mixing G+ –H − . Fig. Fig.16 are typical diagrams contributing the self energies mixing of W + –H − and G+ –H − . 1. . 1. Fig. Feynman diagrams relevant for the e+ e− → W + H − . self-energies of the mixing W + –H − .13–1.27 are the counter-term for the mixing G+ –H − and W + –H − .17–1.

We will take the following experimental input for the physical parameters [30]: • The fine structure constant: α = e2 /(4π) = 1/137.03598.5 6 tan β < 10. Arhrib et al. For small values of tan β the cross sections are approximately the same for both models of type I and II (Fig. In this interval for √ tan β the total cross section varies from 7. ReΣ H In particular.6). each of the remaining form factors should be UV finite by itself. s  e2 sW  = − ∆ (gA + gV )A2 + (gV − gA )A1 MZ CT 2 s − mZ cW γ MCT = − (4. (4. . As a consequence. the cross section decreases to very small values in Model-I.187 GeV and MW = 80. following the b Yukawa coupling proportional to tan β. so that it receives by definition scheme we define sin2 θW by sin2 θW ≡ 1 − MW Z no loop corrections. 4.4) contain only the matrix elements A1 and A2 as follows: e2 ∆(A1 + A2 ). for s = 500 GeV and mH ± = 220 GeV. For tan β > 10 (Fig.5 GeV. which provides a good analytical check of the calculations. In the on-shell 2 /M 2 . In the following more explicit discussion we consider the range 0. it follows that bH − G+ (m2 ± ) = 0. / Nuclear Physics B 581 (2000) 34–60 43 The last renormalization condition is sufficient to discard the real part of the H − –G+ mixing contribution as well. The amplitudes corresponding to the two counter-terms (photon and Z boson exchanges) in Eqs.8) and sW ≡ sin θW . 1. In Model-II. being purely real valued. corresponding to the Yukawa coupling proportional to 1/ tan β. the Feynman diagrams depicted in Fig. cW ≡ cos θW . Numerical results and discussion The difference between the predicted cross sections in Model-I and Model-II is essentially due to the fermion loops containing the different Yukawa couplings. where both type I and II models give practically the same numerical results. 4sW cW (4. (4.012 fb. together with Eq.5 fb to 0. 5.41 GeV.3. using the Slavnov–Taylor identity [28.A. 4sW cW gA = 1 .29] k 2 ΣH − W + (k 2 ) − mW ΣH − G+ (k 2 ) = 0 at k 2 = m2H ± valid also for the renormalized quantities. • The top and bottom quarks masses are taken to be: mt = 175 GeV and mb = 4. 2(a)). only the form factors M1 and M2 can contain UV divergences. 2(b)).7) with the electron–Z coupling constants gV =  1 2 1 − 4sW .25 will not contribute with the above renormalization conditions. the cross section has a minimum for tan β ≈ 30 and increases again for large tan β. Indeed. • The gauge boson masses: MZ = 91.

/ Nuclear Physics B 581 (2000) 34–60 (a) (b) Fig. 2. and s = 500 GeV. . Arhrib et al. Top–bottom contribution to √ the integrated cross section as a function of tan β in Model type I and II for mH ± = 220 GeV.44 A.

mH = 280. Arhrib et al. One can see that the cross section reaches a minimum for moderate values of tan β while it gets much larger (a few orders of magnitude) for both low and high values of this parameter. tan β = 2. respectively.5 TeV.2. mH = 180. tan α = 1.6. (c) we show the total cross section (including all virtual boson and fermion contributions) √ as a function of the center of mass energy s for the three configurations listed above and for four different values of λ3 which satisfy the nominal unitarity constraints (2.6 and 1. 1 TeV.5 fb for a charged Higgs mass of about 220 GeV. the cross section increases with λ3 . This contribution remains generically suppressed (0.6 (Fig. (b). 3(a)) the top quark effect is enhanced. the nominal unitarity condition (2. 3 shows the top–bottom contribution to the integrated cross section as a function of the √ center of mass energy s for four values of mH ± and for two values of tan β. As tan β increases the top quark effect decreases. for a fixed tan β.A. 5. The effect of λ3 arises essentially via the trilinear vertices H 0 H + H − and h0 H + H − . mA = 80. The first enhancement (small tan β) is due to the top quark effect as discussed above. tan α = 3. One should. For the general THDM we will present our numerical results in the following three configurations (where all the masses are in GeV): • C1: • C2: • C3: mH ± = 220.2. mh = 120. For large values of tan β(' 50) the bottom quark contribution becomes leading and of comparable magnitude to that of the top quark in the small tan β region. (c) start violating the unitarity constraints on the H 0 H + H − and h0 H + H − couplings. One can reach a cross section of 3. (2. stress that a different choice of the Higgs masses and λ3 closer to the MSSM configurations would suppress the Higgs sector effect in the large tan β regime. tan β = 1. It can be seen that for a small tan β (Fig. 0.6. mA = 370. Figs. 4(a). 5(b) and 5(c) show the total cross section as a function of tan β in the C1. 5(a).1.5). (b). tan β = 3. / Nuclear Physics B 581 (2000) 34–60 45 Fig. tan α = 2. even in the favourable light Higgs mass configurations such as in case C1 . One can see from those curves that. .5) forbids tan β larger than 14. and 1.4. 3(b)).3 √ configurations with λ3 = 0.01–0. 4(a). mh = 90. mA = 220. The cross section is enhanced for mH ± = 170 GeV owing to the proximity of the normal threshold cut of the three-point function at mH ± = mt + mb . In those three cases the present experimental bound on δρ is respected.5). 4(d) shows the (λ3 independent) box contributions to the cross section in the C1. thus allowing even larger tan β values to be consistent with Eq. and for s = 500 Gev. With the parameter choice of Fig.4. mH ± = 300. Note that values of λ3 larger than the ones chosen in Fig. mh = 120. In Figs. Fig. mH = 380. In this case large effects from the bottom quark sector become dominant. leading to almost an order of magnitude suppression of the cross section for tan β = 1. while the second enhancement comes from the effect of large tan β in H 0 H + H − and h0 H + H − couplings.3 configurations.6.005 fb for √ s =500–1000 GeV). however. mH ± = 400.

6. 3. 300 GeV. Arhrib et al. / Nuclear Physics B 581 (2000) 34–60 (a) (b) √ Fig.6. Top–bottom contribution to the integrated cross section as a function of s for four values of mH ± = 140 GeV.46 A. and 400 GeV. (a) tan β = 0. 185 GeV. . (b) tan β = 1.

tan β = 1. tan α = 1. Integrated total cross section as a function of s for four values of λ3 . . mH = 180. mH = 280.A. mA = 80. mh = 90.4.6. (b) C2 case: mH ± = 300. mA = 220. Arhrib et al. tan β = 3. mh = 120. (a) C1 case: mH ± = 220. tan α = 2. / Nuclear Physics B 581 (2000) 34–60 47 (a) (b) √ Fig.4. 4.6.

mA = 370. Arhrib et al.2. / Nuclear Physics B 581 (2000) 34–60 (c) (d) √ Fig.48 A. . tan β = 2 (where all masses are in GeV). (d) Box contributions to the total cross section as function of the center of mass energy for C1.3 configurations. Continued. mH = 380. mh = 120. (c) C3 case: mH ± = 400. tan α = 3. 4. Integrated total cross section as a function of s for four values of λ3 .

2.5 TeV (c). respectively. 5. Arhrib et al.A.3 configurations with λ3 = 0. 1 TeV (b) and 1. / Nuclear Physics B 581 (2000) 34–60 49 (a) (b) Fig.1 and for s = 500 GeV (a). . Integrated total√cross section as a function of tan β in the case of the C1.

tan α = 3. and 1500 GeV. 1000. with tan β = 2. 5.) √ Fig. / Nuclear Physics B 581 (2000) 34–60 (c) Fig. λ3 = 0. Integrated total cross section as a function of mH ± for s = 500. mh = 90 GeV and mA = 80 GeV . (Continued.1. mH = 180 GeV.50 A. 6. Arhrib et al.

inasmuch as we can compare.2) − 2 sin 2β gH 0 H ± G∓ = −ig sin(β − α)(m2H ± − m2H ) 2mW . 6 we show the dependence of the cross section on the charged Higgs mass for √ s = 500 Gev. in the general THDM.5 TeV in the case where tan β = 2. The smallness of the cross section would require.1) − 2 sin 2β      m2 cos(α + β) g 1 4λ3 v 2 + m2H + m2h sin(β − α) m2H ± − h + gh0 H + H − = −i mW 2 sin 2β 2     1 sin 2α + 2 sin(α + β) cos(α − β) m2H − m2h . λ3 = 0. One notes the high sensitivity of the cross section to the Higgs mass when the latter is above the W − h and bt¯ decay thresholds (which happen to roughly coincide owing to our input values. another paper [33] appeared dealing with the same subject. after this work was completed. come from the trilinear H H ± H ∓ and hH ± H ∓ vertices. a very high luminosity option. however. The study is done in the general two Higgs doublet model taking into account constraints on the ρ parameter and unitarity constraints on the trilinear Higgs couplings. and also the renormalization scheme in [32] was not explicitly defined. while the leading contributions in the large tan β regime. / Nuclear Physics B 581 (2000) 34–60 51 In Fig. (A. tan α = 3. however. mH = 180 GeV. (A.      m2 g 1 sin(α + β) 4λ3 v 2 + m2H + m2h cos(β − α) m2H ± − H + gH 0 H + H − = −i mW 2 sin 2β 2   2  1  sin 2α + 2 sin(α − β) cos(α + β) mH − m2h . 6. one should keep in mind that this region corresponds to the opening of the charged Higgs pair production which yields a much more interesting event rate. Note however that we did not assume any Higgs mass rules. The calculation is performed within dimensional regularization in the on shell scheme. [32]. the cross-section is much less sensitive to the Higgs mass. We mention that our results are in qualitative agreement with those of Ref. Below this threshold. 1 TeV and 1.3) . We have shown that in the small tan β regime the top effect is enhanced leading to important cross sections (about 1 fb). (A.1. the associated production with a W boson would be the only means to look for direct evidence for it.A. mh = 90 and mA = 80 GeV. If the charged Higgs is too heavy to be produced in pairs in e+ e− future machines. and correspond to the kink in the plots). Conclusions To summarize. we have computed the associated production of a charged Higgs boson with a gauge boson W at high energy e+ e− collisions. Arhrib et al. Appendix A In this appendix we list the Feynman rules for the 3-point vertices involving charged Higgs bosons. Quite recently.

4) 2mW m2 ± − m2 (A. m20 .5) gA0 H ± G∓ = ∓ H √ A .µ = (2πµ)(4−D) iπ 2 Z dD q {1. proportional to m2H and m2h . with the scalar functions Ci.52 A. di = (q + pi )2 − m2i .52] we use. Three-point functions: C0. one gets for the vector integral Bµ = p1 µ B1 with the scalar function B1 (p12 .2) 2 . p12 2 0 1 2 . d0 where µ is an arbitrary renormalization scale and D is the space–time dimension.µν = (2πµ)(4−D) iπ 2 Z dD q {1. Two-point functions: B0. Arhrib et al. d0 d1 Using Lorentz covariance.1) Cµν = gµν C00 + p1µ p1ν C11 + p2µ p2ν C22 + (p1µ p2ν + p2µ p1ν )C12 . where the pi are the momenta of the external particles (always incoming). (B. qµ } .7) gh0 G+ G− = −im2h √ 2v The parameter λ3 enters only gH 0 H + H − and gh0 H + H − . (A. One-point function: A0 m20  (2πµ)(4−D) = iπ 2 Z dD q 1 . m21 ). qµ qν } . d0 d1 d2 where pij2 = (pi + pj )2 . while the vertices gH 0 H ± G∓ .µ. (B. qµ . (A. Appendix B. Lorentz covariance yields the decomposition Cµ = p1µ C1 + p2µ C2 . m2 ). v 2 cos(β − α) √ . One-loop functions Let us briefly recall the definitions of scalar and tensor integrals [51. m2 . respectively.ij (p12 . (A. gh0 H ± G∓ and gA0 H ± G∓ have a particularly simple form.6) gH 0 G+ G− = −im2H 2v sin(β − α) . The inverse of the propagators are denoted by d0 = q 2 − m20 . / Nuclear Physics B 581 (2000) 34–60 gh0 H ± G∓ = ig cos(β − α)(m2H ± − m2h ) . p2 . m2 .

m2 .1.2) for each V boson exchange (V = γ . 2 e ) (1 − 2sW 2s 2 ef u fR .1) udW 2 2 sW in the following notation: gγf L = gγf R = −ef .1.4) D0. d-d-u exchange The diagram with the ddu triangle. Fermionic loops C.µν = Dµν = gµν D00 + p1µ p1ν D11 + p2µ p2ν D22 + p3µ p3ν D33 + (p1µ p2ν + p2µ p1ν )D12 + (p1µ p3ν + p3µ p1ν )D13 + (p3µ p2ν + p2µ p3ν )D23 (B.1. p2 . we use the fermion–vector boson coupling constants as defined in terms of the neutral-current and charged-current vertices     f L 1 − γ5 f R 1 + γ5 + gV . eu = −2ed = |e|. p2 . p2 . yields the following contribution to the one-loop amplitude (3. Vertex diagrams C. qµ qν } (2πµ)(4−D) . m2 ). Appendix C. m2d .1 = √ 2 2(s − mV )sW    L dL R L dR gV + md mu Yud gV C0 gVdL − gVdR − m2d Yud − mu m2u Yud . C0 and D0 .1.1. m2 . Fig. 1.A.2) C. m2 . p12 23 3 2 13 0 1 2 3 All the tensor coefficients can be algebraically reduced to the basic scalar integrals A0 . gZdL = 2sW cW 2 ee = −|e|.µ. The analytical expression of all the scalar functions can be found in [51.ij (p12 . gZ = W .1.5) 2 . 52]. One-loop amplitude In this appendix. (B. 2sW cW 2sW cW 2 e ) (1 + 2s W d . 3 gZuL = − (C. dD q 2 iπ d0 d1 d2 d3 Again. B0 .3) Dµ = p1 µ D1 + p2µ D2 + p3µ D3 . m2d M1. p2 . f f¯Vµ = γµ gV 2 2     ¯ µ+ = − √ 1 γµ 1 − γ5 (C. / Nuclear Physics B 581 (2000) 34–60 53 Four-point functions: Z {1.1. with the scalar 4-point functions Di. Z): n    2NC α 2 L dL R − mu Yud gV + md Yud gVdL − gVdR B0 s. Lorentz covariance allows the decomposition (B. Arhrib et al. qµ .

m2W . m2d . gVdR ←→ gVuR and t ←→ u. Arhrib et al.4) + (4C1 + C2 − C12 − C22 ) gVeR (A3 + A5 ) + gVeL (A6 + A4 ) .2. m2W .1. gVdL ←→ gVuL . and also A3 ←→ A5 and A6 ←→ A4 . given in the associated tables.54 A. Yud mu ←→ md . The couplings are summarized in the following table (S is a generic notation for one of the Higgs bosons): .2 is obtained from the previous one by making the following replacements: L R ←→ Yud . m2H ± . 1. (C. Fig. / Nuclear Physics B 581 (2000) 34–60 o  dL  L R + 2 mu Yud + md Yud gV C00 gVeR A1 + gVeL A2 n  L dL  R L dL gV + md Yud gV C0 m2W gVdL − tgVdR C1 + mu uYud − mu m2W + u Yud o    L dL R + mu m2H ± + u Yud gV + md Yud ugVdL − m2H ± gVdR C2 gVeR A1 n  L dL  R L dL + mu m2W + t Yud gV + md Yud gV C0 m2W gVdL − ugVdR C1 − mu tYud o    L dL R + mu −m2H ± − t Yud gV + md Yud m2H ± gVdR − tgVdL C2 gVeL A2 n   L L L R − 2 gVdL mu Yud C0 + mu Yud C1 + 2mu Yud + md Yud C2 o     L R L R + mu Yud + md Yud + md Yud C12 + mu Yud C22 gVeR A3 + gVeL A6 n  R dR L gV C1 − gVdL mu Yud C2 + 2 md Yud  o eL   L R eR + mu Yud + md Yud (C12 + C22 ) gV A4 + gV A5 . m2H ± . m2W − m2S − m2W + s C0 W 2 s − mV     + m2H ± − m2W − s C1 + m2H ± − m2W + s C2 + C00 gVeR A1 + gVeL A2  o (C. m2S .3 = − All the Ci and Cij have the same arguments: (m2W . C. yields the corresponding contribution to the one-loop the matrix element (3. dd2 ).1. u-u-d exchange The corresponding expression for the the diagram with the uud triangle in Fig.3) Therein.2. m2u . M1.3: n   α 2 gV W + W − + W − S gH − W + S g −B0 s.2). m2W ). Bosonic loops Listed are always the analytic expression for each generic diagram of Fig. s.1. The sum over all the particle contents. in practice only the third quark generation is relevant. Summation has to be performed over all fermion families. all the Ci and Cij have the same arguments: (m2W . 1. s. 1. C.

H ) cβα msWW Fig. m2W . H ± . m2H ± . H ± . Arhrib et al.5: M1. 1. h) −1 (G± .Sj ) (A0 .A. / Nuclear Physics B 581 (2000) 34–60 S gZZS gW + W − S h mZ cWβα sW mW sβα W H c mZ cWβα sW c mW sβα W s s 55 gH − W + S gZW + W − gγ W + W − cβα 2sW −sβα 2sW − scWw −1 − scWw −1 Therein. The couplings are summarized in the following table: gZSi Sj gH − Si W gW − W + Sj cβα 2sW cW s − 2s βαc W W 1 2sW 1 2sW sβα msWW (Si . H . m2Sj ).5 = α2 s − m2V n   gV Si Sj gW + Sj Sk gH − Si Sk C00 gVeR A1 + gVeL A2  o − (C2 + C12 + C22 ) gVeR (A3 + A5 ) + gVeL (A6 + A4 ) . 1. (C. Fig. G± ) 0 (A0 .4: M1. G± .6) All the Ci and Cij have the same arguments: (m2W . s. sβα ≡ sin(β − α) and cβα ≡ cos(β − α). m2Sj ). s.5) where the arguments of Ci and Cij are now as follows: (m2W . H ) −1 gZSi Sj c i 2s βαc W W s −i 2s βαc W W W) − cos(2θ 2sW cW W) − cos(2θ 2sW cW W) − cos(2θ 2sW cW W) − cos(2θ 2sW cW gH ± Si Sk igH − A0 G+ igH + A0 G− gH − hG+ gH − H G+ gH − hH + gH − H H + where the gH ± Si Sk couplings have been defined in Appendix A. H ) −1 (H ± . h) (A0 . G± ) 0 (G± . gW + Sj Sk sβα 2sW cβα 2sW sβα 2sW cβα 2sW cβα 2sW s − 2sβα W .Sj . m2Sk . G± .Sk ) gγ Si Sj (A0 . h. m2H ± . The couplings are summarized in the following table: (Si . h) −1 (H ± .4 = n   2α 2 gV Si Sj gW − W + Sj gH − W + Si C00 gVeR A1 + gVeL A2 2 s − mV  o − (2C0 + 2C1 + 3C2 + C12 + C22 ) gVeR (A3 + A5 ) + gVeL (A6 + A4 ) (C. m2Si . m2Si .

H ) mW −mZ sW cβα 2sW s − 2sβα W sβα 2sW cβα 2sW Fig. m2V 0 ). H .56 A. m2W . M1.8 = − where all the Ci and Cij have the same arguments: (m2W . (C.Sj ) gγ W + Si gZW + Si gH − W + Sj gW + Si Sj (G± . m2H ± .7) where C0 = C0 (m2W . W ± ) (G± . m2S ). m2Si . m2Sj . Z) 0 (H . h) mw −mZ sW (G± . / Nuclear Physics B 581 (2000) 34–60 Fig. 1.7: M1. s.V 0 ) gγ V 0 Si gZV 0 Si (h. 1. m2H ± .Sj .8) All the Ci and Cij have the same arguments: (m2W . m2V 0 ). s. W ± ) gH + Si Sj gW + V 0 Sj mZ sWβα cW gH − G+ h −mZ sW 0 c mZ sWβα cW gH − G+ H −mZ sW mW −mZ sW gH − G+ h sβα msWW mW −mZ sW gH − G+ H cβα msWW s Fig. h.9) (2C0 − 2C1 + 3C2 + C12 + C22 ) gVeR (A3 + A5 ) + gVeL (A6 + A4 ) . The couplings are summarized in the following table: (Si .8: n   α2 gZZS gZW − W + gH − W + S B0 s. s. m2Z .6 = −   α2 gV V 0 Si gW + V 0 Sj gH − Si Sj C0 gVeR A1 + gVeL A2 . 2 s − mV (C. m2H ± .7 = 2α 2 s − m2V n   gV V 0 Si gW + Si Sj gH − V 0 Sj −C00 gVeR A1 + gVeL A2  o + (−C2 + C12 + C22 ) gVeR (A3 + A5 ) + gVeL (A6 + A4 ) . m2Sj . The couplings are summarized in the following table: . Arhrib et al. The couplings are summarized in the following table: (Si . m2Si . m2Z + 2m2H ± + 3m2W − 2s C0 2 s − mZ       + m2H ± + 3m2W − s C1 + 3m2H ± + m2W − s C2 − C00 gVeR A1 + gVeL A2  o (C. G± . G± .6: M1. m2S . 1. Z) (G± .

m2H ± gVeR A1 + gVeL A2 . s. (C. 1. s. m g A + g A g . (C. m2W . m2e . 1. m . 1.17–1. H − ) (h.20. S2 ) gZW + S1 S2 gγ W + S1 S2 (h. MH (C. / Nuclear Physics B 581 (2000) 34–60 S gZZS h mZ sWβα cW H mZ sWβα cW s c gH − W + S gZW + W − cβα 2sW −sβα 2sW − scWw 57 − scWw Fig. m2W + m2h − u D0 + m2W − u D1 o   + m2W − s − t D3 A2 − 4D1 A4 .2. Counter-term diagrams Denoting by k µ ΣH W (k 2 ) the bare H ± W ± 2-point function.10: M1.10 = − n  o α2 gV W + S1 S2 gH − S1 S2 B0 m2H ± .13) where all the Di functions have as arguments: (m2W .9: M1. m2e . G+ ) (H. The corresponding box graph with a heavy neutral scalar is obtained by the substitution Box = −MhBox(mh → mH ). m A + g A B m g . 1. m2e . m2W . Box diagrams Fig.11: M1. G+ ) W sβα 2cW s − 2cβα W c − 2cβα W gH − S1 S2 ghH − H − gH H − H − ghH − G+ gH H − G+ C. The renormalized non- . 0.2). the renormalized one is obtained by adding the counter-term (4. 1 2 ZW S Z V V 1 1 s − m2Z (C.1.2.11) 2 s − mV Fig.A.9 = n  eR o α2 2 2 2 eL + − + − g g . m2S .12: MhBox = − α2 2 4sW cβα sβα mW n    C0 m2e . 0. m2h .12) where the couplings are given in the table (S1 . m . Arhrib et al. 1. m2W ). m2H ± . H − ) − 2cβα c cβα 2sW s − 2sβα W sβα 2sW sβα 2sW (H. (C.11 = n  eR o α2 2 2 eL + H − S gSZZ B0 s.14) C. Fig.10) 0 1 2 V W H S1 W W S1 W S1 W V V s − m2V Fig.

1. Arhrib et al. m2H ± − cβα mW sβα B0 m2H ± . C.2.20) M1. m2W + 2cβα gH H + G+ sW B1 m2H ± .2. m2W   + 2ghH + G+ sβα sW B1 m2H ± . t-channel W + –H − mixing The generic diagram for this topology is depicted in Fig. m2H . (C.2.R defined in (C.1).28. 1.18) Then the amplitudes containing the counter terms.19) with gZ = −sW /cW . m2h . m2d . m2h . m2W − gH H + H − sβα sW B0 m2H ± . m2H . m2H . m2W 2 8πsW   + ghH − G+ sβα sW B0 m2H ± . m2h . Amplitudes with non-diagonal self-energies C. m2h . + mu Yud (C.16) ∆ = ReΣH W m2H ± . The amplitude contains only the invariant A2 :   ferm + Im Σ bos e2 Im ΣH W HW A2 . m2W − 2gH H + H − sβα sW B1 m2H ± . 1. requiring that the renormalized W + –H − mixing vanishes.15) According to the condition (4. Fig. The self-energy ΣH W (m2H ± ) ≡ ΣH W has the following explicit form: ferm bos ΣH W = ΣH W + ΣH W . The fermionic part is the sum over the individual fermion families. m2H ± + cβα mW sβα B1 m2H ± .2.58 A.17) and the bosonic part is given by bos ΣH W =   α  cβα ghH − H − sW B0 m2H ± . / Nuclear Physics B 581 (2000) 34–60 bH W (k 2 ) is thus given by: diagonal self-energy Σ bH W (k 2 ) = ΣH W (k 2 ) − ∆. m2W   + 2cβα ghH + H − sW B1 m2H ± . m2W + cβα gH H − G+ sW B0 m2H ± . m2H ±   − cβα mW sβα B1 m2H ± . m2h .13. m2H . m2d . m2H . m2H . Σ (C. ∆ is determined by:  (C. (C. m2u ΣH W = −NC √ 2 2 sW π   L R + md Yud B1 m2H + . each one contributing   α ud R md Yud B0 m2H + . read: Mδ(V W H ) = 4παgV s − m2V   Re(ΣH W ) gVeR A1 + gVeL A2 (C. m2H ±   + cβα mW sβα B0 m2H ± . m2u .6). gγ = −1 and gVeL. m2W .13 = 2 2sW m2H ± − m2W . m2h .

. m2h . [4] H. Phys. Mrenna.15 = 2 2 2 (s − mV ) mH ± − mW where gZW W = cW /sW and gγ W W = 1. J. MA. m2H . Wagner. CERN-TH/98-262. Phys. m2H ±   + ghG+ H − ghG+ G− B0 m2H ± . hep-ph/9808312. s-channel W + –H − mixing The generic diagram for this topology is depicted in Fig.2. Addison-Wesley. 1. (C. Eur.24) where gZW + G− = −mW sW /cW and gγ W + G− = mW . / Nuclear Physics B 581 (2000) 34–60 59 C.15.E.3. m2d . m2u + mu −m2d + m2u Yud  o L R − md Yud B1 m2H ± . References [1] ALEPH Collaboration. m2d . CERN-EP/98-173.E.M. Gunion. OPAL Collaboration. Ackerstaff et al. m2W − 2m2H ± B1 m2H ± .14 = 2 2 2 (s − mV ) mH ± − mW (C. Reading.2. S. Phys. B 263 (1986) 265. Komatsu. + m2H ± mu Yud (C. m2H ± = ΣGH 4π  + gH H + H − gH G+ H − B0 m2H ± . Kubo. Nucl. [2] G..2.22) and the bosonic part is given by  α n bos ghH + H − ghG+ H − B0 m2H ± . 1990.F. G. H. Arhrib et al. Phys. OPAL Collaboration.14 can be written in the form   ud + Im Σ bos  e2 gV W + G− Im ΣGH GH gVeR A1 + gVeL A2 . m2H .2. C. m2W − 2 4sW    o + A0 m2W + m2H + m2H ± B0 m2H ± . (C. s-channel G+ –H − mixing The non-diagonal Goldstone–H ± self-energy is ΣGH evaluated in the same way as done before for the W –H ± case (k 2 = m2H ± ): ferm bos + ΣGH . Dawson.23) Then the amplitude of the diagram Fig. [3] J. m2u . . Carena. m2h . hep-ex/9811025. m2H . m2W − 2m2H ± B1 m2H ± . C. B 157 (1985) 90. m2h .A. M1.2 as follows:   ferm + Im Σ bos  Im ΣH  e2 gV W W W HW 2 mW − s gVeR A1 + gVeL A2 . m2W . ΣGH = ΣGH where the fermionic part is obtained by summing over the families with n   −αNC ud L R mu Yud = √ − md Yud A0 m2d ΣGH (2 2 mW πsW )  L  B0 m2H ± . Haber. m2W     sβα cβα  A0 m2W + m2h + m2H ± B0 m2H ± . S. C 5 (1998) 19. m2W + gH G+ H − gH G+ G− B0 m2H ± . [6] M. 1. B 440 (1998) 403. m2h . Kane.21) M1. The Higgs Hunter’s Guide. [5] K. Abbiendi et al.L.2. The amplitude can be projected on A1. m2H . Lett. Lett. J.

Willenbrock. D 15 (1977) 1958. D 53 (1996) 1304. C 47 (1990) 75. Nucl. Phys. R. J. V. Tung. Rev. Roy. Phys. J. ’t Hooft. J.A. Rev. S. B.A. Weinberg. J. Haber. Eur.A. W. Capdequi Peyranère. Phys. B 324 (1994) 236. Proceedings of the Workshop on e+ e− Collisions at 50 GeV The Physics Potential. November 6–9. Sampayo. Phys. Barger. Rev.E. Veltman. D 55 (1997) 5627. B 558 (1999) 3. A. Phys. Eur. G. Khoze.L. Dicus.P. D 38 (1988) 2158. B 153 (1979) 365. H. Rev. M. Capdequi Peyranère. Iogansen.D. Caso et al. Passarino. M. Tung. 1992. Langacker. B 272 (1986) 1. Arhrib et al. Nilles. P. Phys. R.K.E. DESY 92-123. Zerwas (Ed. T. B 299 (1988) 117.E. D. Fortschr. S. V. Guasch. Rev. G.A. hep-ph/9901221. Phys. G. Z. H.L.. Moultaka. Dabelstein. J. Guide to FeynCalc 1. Rev. Chong Sheng Li. D 59 (1999) 015009. Phys. A. Zhu. Rev. Mohapatra. hep-ph/9809324. J. Rev. Mättig.N.A. M. Haber. C 57 (1993) 569.0. Kanemura. Dominici. Guasch.A. hep-ph/9812536. P. Phys.H. J. Ma et al.M. B 44 (1972) 189. 52 (1977) 11. Denner.P. H. Commun. University of Würzburg. Moultaka.M Barnet. 117 (1985) 75. Phillips. Rev. KA-TP-19-1999. Pati. Phys.F. Djouadi. hep-ph/9808317. Mertig.P. hep-ph/9803484. D. J. J. 46 (1987) 890. Olness. B 315 (1995) 313.G. D 43 (1991) 3640. University of Würzburg. Maison. R. 1992. D. D.P. Soper. D 40 (1989) 787. C. hep-ph/9911541.S. Gatto. Zochowski. Nucl. Uraltsev. Capdequi-Peyranère. Hollik. W. M. Roy. ’tHooft. J. Diaz.J.A.A.E. W. Phys. hep-ph/9811256. 110 (1984) 1. Lett. Bawa. C 2 (1998) 373. Rev. Garcia. Talk given at the 5th International Wein Symposium (WEIN 98). J. J.60 [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] A. IL. Komamiya. Paige.J. Lett. Nucl. Phys. hep-ph/9911452. Mod. F. Phys. in: P. Nucl. Zerwas. Chong Shou Gao. Lett.. Phys. J. Lett.0. B 160 (1979) 151..K. Phys. Int. Casalbuoni. Drees. Presented at Workshop on Physics at the First Muon Collider and at the Front End of the Muon Collider. Hollik.N. V. Odagiri.M. Phys.H. Irulegui.. Rizzo. Haber. M. Coarasa. T. K. Kim. Phys. Veltman. C 3 (1998) 1.J. Phys. Sov. H. M.). D 50 (1994) 6820. Kalinowski. H.A. Martin. Phys. A. 1997. G.E. Glashow. R. Phys. Phys. Kraft. D. B 294 (1987) 621. Shou Hua Zhu. W. Eck. A. Nucl.M. Phys. Math. Hollik. G. Z. S. hep-ph/9712367. Phys. M. Rev. hep-ph/9501315. D 11 (1975) 566. hep-ph/9807480. A. Moretti. B 306 (1988) 697. Kao. M. P. . C. Z. Phys. Brignole et al. Kniehl.G. A 14 (1999) 429. Küblbeck. Arhrib.E. R. Feruglio. Nucl. A. Gunion. S. Z. G. C 51 (1991) 695. A. P. Phys. Nucl. Batavia. P. 1992. Nucl. R. D 44 (1991) 191. 66 (1991) 1. DESY-98-177. J. N. Haber. C. Phys. M. hep-ph/9301205 (Errata). J.L.E. Phys. Comput. Santa Fe. Veltman. O. hep-ph/9809352. D. S. Phys. June 1998. Breitenlohner. G. Arhrib. Krawczyk. Godbole. H.N. Phys.G. S. Nucl. F. A. Haber. Phys. B 308 (1988) 813. D 41 (1990) 884. Jimenez. DPF94 Proceedings. 34 (1986) 11. Kane. J. C 67 (1995) 495. D. Gunion. Barger. Krawczyk.J. M. D. J. Bowser-Chao et al.C. B 269 (1991) 155. R. Phys. Phys. D. Rep. Phys. R. van Oldenborgh. Barrientos Bendezú. Guide to FeynArts 1. A.F. Rep. B 341 (1995) 313. Phys. Kühn. A.A. W. ter Veldhuis. A. Diaz-Cruz. Erler. Roy. / Nuclear Physics B 581 (2000) 34–60 R. Böhm. Phys. Nucl. M. H. S. Hewett. W. R.S.I. Commun. Phys.D. Phillips. Sola. F. J. Spiesberger. Guth. H.

28049 Madrid.Gh.  2000 Elsevier Science B.A.V. 1 This instability is a drawback of the Standard Model. for the case of a light Higgs (i. Di Clemente. or indirectly.. All rights reserved.Bn Keywords: Renormalization schemes. requires the presence of new physics to stabilize the SM vacuum. then. Madrid. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. with the present lower bounds on the Higgs mass. PACS: 11.1 .60. In particular. C-XVI Universidad Autónoma de Madrid. Casas. by imposing perturbativity of the new physics.10. V.elsevier. Spain Instituto de Física Teórica. which is unable to describe physics at scales beyond ΛSM and. 140 GeV) ΛSM can be closer to values that could directly. All rights reserved.-i. Spain Received 23 February 2000. ΛSM . MH .80.nl/locate/npe The Standard Model instability and the scale of new physics ✩ J.e. We find. In this paper we will fix Mt to its experimental mean value and disregard the effect due to the experimental error 1Mt . For large values of the Higgs mass the scale ΛSM is larger than the Planck scale and thus has no impact on the physics at present and future accelerators.3 2 1 3 ( 0 0 ) 0 0 1 9 9 . 12. PII: S 0 5 5 0 . This implies that. that M can be as large as one order of magnitude higher than the instability scale of the Standard Model. which solely depends upon the value of the (yet undiscovered) Higgs mass MH [1]. Quirós Instituto de Estructura de la Materia (CSIC) Serrano 123. 28006.Nuclear Physics B 581 (2000) 61–72 www. the new physics could easily (but not necessarily) escape detection in the present and future accelerators. and to guarantee the absolute stability of ✩ Work supported in part by the CICYT of Spain (contract AEN98-0816). accepted 23 March 2000 Abstract We apply a general formalism for the improved effective potential with several mass scales to compute the scale M of new physics which is needed to stabilize the Standard Model potential in the presence of a light Higgs. thus having an impact on the physics at the corresponding scales. M. Efective potential and physics beyond Standard Model 1. be detected at future accelerators. 1 The original studies considered Λ SM as a function of MH and Mt (the top-quark mass) [2–7].V. Introduction The Standard Model (SM) effective potential is unstable beyond a scale. 14. However.

thus implying independence of the (complete) effective action with respect to the scale µi (similar to the usual scale invariance). as deduced from the stability requirement. Obviously. as. M. that is normally associated to its mass. . say µt .. of the new involved fields. which is the ordinary MS renormalization scale. In Section 3 we will apply the ideas and results of Section 2 to the model we are considering to remove the SM instability. Multi-scale effective theory We will consider the effective potential of a theory with N different scales in a massindependent renormalization scheme. it is clear that the simplest and most economic SM extension that could circumvent the instability problem is the addition of just scalar fields coupled to the SM Higgs field [8. new physics must be introduced. that should be identified with the mass of the new scalar fields (Φ). A general formalism to evaluate the effective potential in this kind of multi-scale scenario was presented by the authors in Ref.e. we have to consider two different mass scales: the SM scale. However. µi . In our case. A similar analysis of this kind of model was performed by Hung and Sher in Ref. This will give rise to a set of RGE with respect to the scales µi . for the previous reasons. e. In Section 4 we will present our numerical results and in Section 5 our conclusions. the MS scheme. This will be the model considered in the present paper. i. / Nuclear Physics B 581 (2000) 61–72 the electroweak vacuum. the scale invariance of the complete effective potential indicates that the results should not depend on the details of the chosen decoupling scale µi . where the effective potential was considered in the one-loop approximation. can be much more realistic. On top 2 For the sake of simplicity we will consider µ as the only SM scale. as the masses of the Higgs and the gauge bosons. or indirect. say µΦ .g. detection of the new physics depends on the mass. 2 and the scale of new physics. e. and will not make a distinction with the t other (nearby) SM scales. 2. [13]. The outline of the paper is as follows: in Section 2 we will briefly summarize our general proposal of multi-scale effective potential.62 J. Casas et al. where the general procedure for decoupling fields [14] was incorporated. This is the issue that will be considered in this paper. We will use this formalism throughout the paper. where the decoupling is not automatically incorporated in the theory. However improving the effective potential by the renormalization group equations (RGE) yields very important corrections and must be taken into account in any realistic calculation. Since the presence of extra bosonic (fermionic) degrees of freedom tend to stabilize (destabilize) the SM Higgs potential [8–12]. the simple model under consideration basically represents the most efficient way to cure the SM instability. The direct.9].A.g.. Improving the effective potential in a SM extension generally implies considering a multi-scale problem. such as the MSSM. that can be identified with the top-quark mass.. The usual procedure is to decouple every field at a given scale. but. by means of. other SM extensions. a step function. thus giving reliable upper bounds on the scale of new physics. as well as with respect to µ. [8]. on the mass of the extra particles.

7) i=1 3 In addition. This translates into the decomposition: X dλa ≡ µ βλ a = µi β˜λa θ (µ − µi ). in the model at consideration in the present paper. / Nuclear Physics B 581 (2000) 61–72 63 of that we will use a simple step function for decoupling 3 .4) where si is the spin of the ith field. although it could be easily smoothed. i 64π 2 µ2 µ2i i=1 (2. dVeff =0 (2.6) it follows that µ βλ a + N X i=1 µi βλa = N X µi β˜λa = βλMS .2) and (2. dµ N µ (2.3) where V0 is the tree-level potential. when computing the one-loop β-functions corresponding to λa (where λa denotes all dimensionless and dimensional couplings of the theory.1) i=1 which provides the definition of the factors µi β˜λa . the theory is non-supersymmetric and beyond it supersymmetric. the contribution from decoupled fields should not be counted. On the other hand.J. Casas et al. Invariance of the complete effective potential with respect to the MS scale µ simply reads dVeff = 0. (2. This kind of complications will not appear.5) µi dµi leads to the running of the parameters with respect to the scale µi : dλa ≡ µi βλa = µi β˜λa θ (µi − µ).6) µi dµi From Eqs. [15] and [16]. the one-loop effective potential can be written as: Veff = V0 + V1 . (2. and   N µ2i Mi2 1 X 2si 4 (−) M + θ (µ − µ ) log − C V1 = log i i . including the bosonic and fermionic fields). For µ < µi the term Mi4 log(Mi2 /µ2i ) can be taken (at the one-loop level) as frozen at the scale µ = µi . at a given threshold the corresponding symmetry of the system may change. (2. Mi are the (tree-level) mass eigenvalues and Ci depends on the renormalization scheme. Below it.A. In the MS-scheme it is equal to 3/2 (5/6) for scalar bosons and fermions (gauge bosons).4) coincides with the usual MS effective potential when there are no decoupled particles. [13] and similar ones can be found in Refs. A typical example is the threshold of supersymmetric particles in the MSSM. so it does not run with respect to µ. however. a (2. the invariance of the effective potential with respect to µi . . Notice that for µ > µi (for all i) the potential (2. In this decoupling approach. in which case the corresponding matching conditions have to be taken into account at the corresponding thresholds.2) µ dµ On the other hand. These ideas were presented in Ref. (2.

4). it simply reads 4 This is similar to the ordinary MS-scheme where the scale µ can be fixed to a field dependent value. (2. the evaluation of the effective potential requires to evaluate all the λa parameters at the same values of µi (note that λa run with the µi scales). the situation is also intermediate: some of the µ˜ i become equal to the MS-scale µ. we have µ˜ i = µi . This gets rid of all the dangerous logarithms in Eq. (2. These can be constant.10) Finally. Lkin. It can be checked that the one-loop effective potential obtained from the Lagrangian (2. int all interaction and potential terms. int . respectively.64 J. int = (2.8) Lbare = i j where we have included in Lbare. They should be chosen so that the βand γ -functions obtained from the bare Lagrangian satisfy the decomposition given by Eqs. (2.1) and (2. 2 −ε/2 fφi (µ˜ ` ) Zφi ∂µ φi . the bare Lagrangian. (2. as it is usually done [17. and Ob (φi .4). ψj ) represent interaction operators between the fields. so the preferred value of the µi scales are. (2.9) b where Zφi (Zψj ) and Zλb are the bosonic (fermionic) wave function renormalizations and the coupling renormalizations.6).18]. φi = 12 µ˜ i −ε/4 Lkin. Zψj ψj . (2. in order to improve the validity of perturbation theory. ψj = i µ˜ j fψj (µ˜ ` ) Zψj ψ¯ j ∂µ γ µ ψj . X  1/2 1/2 Zλb fλb (µ˜ ` )λb Ob Zφi φi . It is interesting to note that the previous decoupling approach can be obtained starting with the bare Lagrangian written in an appropriate renormalization scheme. or fielddependent.16]. Lbare. They correspond to finite wave-function and coupling renormalizations. . / Nuclear Physics B 581 (2000) 61–72 where βλMS is the usual (complete) β-function in the MS-scheme. we recover the ordinary MS-scheme.1). In terms of the µ˜ i scales defined in Eq. while for µ > µi (all i). 4 A choice that is particularly convenient to greatly improve the validity of perturbation theory is µi ' Mi and µ . we have µ˜ i = µ. the functions fφi . Let us notice that for µ < µi (all i). More precisely. a The invariance of the effective potential with respect to µ and µi allows to choose any values for these scales. At intermediate values of µ. The latter is a generalization of the so-called multi-scale renormalization scheme proposed in Refs. (2. ψj + Lbare. Casas et al. In the case of the SM this value is usually ∼ Mt .8) has precisely the form of the proposed one-loop effective potential of Eq. This implies a knowledge of the µi β˜λa functions defined in Eq. Notice here that since Mi are in general functions of the fields. In addition. φi + Lkin.10). [15. min {µi } [13]. The µ˜ i scales are appropriate combinations of the independent scales µ and µi . fψj and fλb are dimensionless functions of the ratios µ˜ i /µ˜ j which are constant and finite and can be expanded as: f = 1 + O(h¯ ). namely log µ˜ i = log µ θ (µ − µi ) + log µi θ (µi − µ). ψj ) and renormalized couplings and masses λb .A. In the latter region. in terms of renormalized fields (φi . Given a set of bosonic φi and fermionic ψj fields. can be written as: X X Lkin.

Casas et al. and V1 as given by (2. µt . µΦ . the one-loop effective potential in the decoupling approach explained in the previous section can be decomposed as in Eq. Of course. δ is the only coupling that involves the SM with the new physics and will play a relevant role in our analysis.11) In the next section we will apply this approach to our simple extension of the Standard Model. thus giving reliable upper bounds on the scale of the required new physics. In the present case we have two relevant scales: the SM scale. ΛSM . (3. but. λ and Λc are the SM mass term. which can be and the scale of conventionally chosen to be that corresponding to the top-quark. H 0 = hc / 2. quartic coupling respectively.2) and vacuum energy. equivalently. by (2. i. where m2 . Consequently. Consequently..e. where LSM is the SM Lagrangian. Effective theory of the Standard Model extension The presence of extra bosonic (fermionic) degrees of freedom tends to stabilize (destabilize) the SM Higgs potential. when the SM Higgs potential presents an instability at a certain scale. this model may be unrealistic.11)]   1 Mt2 (hc ) µ2t 4 (h ) log + θ (µ − µ ) log − −12M V1 = c t t 64π 2 µ2 µ2t  M 2 (hc ) µ2 4 + θ (µ − µΦ ) log Φ2 + NMΦ (hc ) log Φ 2 µ µΦ (3.4) . i i 64π 2 µ˜ 2i i=1 (2. 2 2 2 MΦ (hc ) = M + δ h2c . In the background of the Higgs field.3) The masses of the top-quark and the ΦE field are defined by λt h2 Mt2 (hc ) = √ c . / Nuclear Physics B 581 (2000) 61–72 V1 = 65   N Mi2 1 X 2si 4 (−) M − C log . the most economic cure is the presence of just extra bosonic fields.4) [or. (2. for the previous reasons. E its quartic coupling and δ provides the mixing between the Higgs and Φ fields.A. the mass of the extra particles. H the SM Higgs doublet and ΦE N real scalar fields E λs transforming under the vector representation of O(N).3) with V0 = − 12 m2 h2c + 14 λh4c + Λc . 3. (3. √ the new physics. M is the invariant mass of Φ. Since we are assuming that ΦE is a SM singlet. it basically represents the most efficient way to cure a SM instability. 3 2  3 − 2  .1) L = LSM + 12 ∂µ ΦE − 12 δ|H |2 ΦE 2 − 12 M 2 ΦE 2 − 4!1 λs ΦE 4 . in this section we will apply the results of Section 2 to a very simple extension of the Standard Model defined by the Lagrangian 2 (3.J.

which depends explicitly on the RGE scale µ. µΦ ' MΦ (hc ). µΦ β˜m2 = − κ N 4 M .1) and (2. (3. However. For for the SM couplings.2). for an improved evaluation of the potential it is much more convenient to choose µ ' µt ' Mt (hc ). κ   1 1 2 ˜ 8δ + (N + 2)λs δ . . (3.12) Here V1SM is the usual SM one loop-effective potential in the MS-scheme. µt γ˜h = γhSM . Expressions (3. except the top-quark field. µt β˜m2 SM = βm 2 . β SM . To evaluate the potential given by Eqs. where κ ˜ couplings corresponding to new physics the β-functions are: 24 = 2γh δ + λδ.11) µΦ βM 2 = 3κ As noted in the previous section.6) (3.. µt β˜Λ SM = βΛ . 2N 2 δ . 64 π 2 Φ 2 µ2Φ (3.8) γ SM and are the SM β and γ -functions. This is given by µt β˜λ = βλSM . all the masses. evaluate explicitly the effective potential. µt β˜λs = κ 1 (N + 8)λ2s .5) (3. µt . (3.3).9) µΦ βδ = κ 3 48 2 δ .10) µΦ β˜λs = 3κ 8 2 µt β˜M 2 = − δm .3) reduces to the ordinary MS effective potential.2). where the one-loop effective potential can be written as: µt β˜δ V1 = V1SM + V1Φ .e. V1Φ   MΦ2 (hc ) 3 N 4 = M (hc ) log − . Hence we need the β. µΦ β˜Λ = 2κ µΦ γ˜h = 0 µΦ β˜λ = = 16π 2 . κ 2N δM 2 . thus getting rid of dangerous logarithms.7) (3. (3.A. Casas et al. (3. (3.3) we also need to know the λa parameters (i. while V1Φ corresponds to the contribution of the decoupled E which runs with µΦ . and is apparent from (3.12) will be used in the next section to field Φ.66 J. µΦ .3) the contribution to the one-loop effective potential of all SM fields.and γ -function decomposition defined in (2. µΦ scales. µ < µΦ . (3. for scales µ > µt . (3. as it is usually done in SM studies. (In the next section we will be more precise about the exact values of these choices. We are neglecting in Eq.6) for the relevant parameters of our model. κ 1 ˜ λs (N + 2)M 2 . the effective potential (3. couplings and fields) at the corresponding values of the µ. / Nuclear Physics B 581 (2000) 61–72 λt being the SM top-quark Yukawa coupling.) These values belong to the range µt .

(4. Numerical results In this section we will study the potential given by Eqs. which have been taken into account in the numerical computations. as it represents the kind of drawback of the pure MS-scheme in the presence of several scales that we wanted to avoid with our approach. 1c V (hc .15) and.8).9)) which leads to the effective ˜ potential (3. Casas et al. (3.12) and analyze the conditions for stability of the electroweak minimum at the vacuum expectation value (VEV) hc = v = 246 GeV. (2. it is interesting to write the explicit form of the bare Lagrangian in a multi-scale renormalization scheme (see Eqs. which is free of such dangerous logarithms.15) which grows logarithmically with the scales ratio µ/µΦ . µ < µΦ (which. This diagram is proportional to δ 2 and plays a major role in our construction. 2 16 π µΦ (3. 6 With the definition of Eq. (3. In principle.   µ 4δ/κ (3. where the za factors are those of the MS scheme.5)–(3. in the sense that it contains internal lines corresponding to the SM and to new physics.12). the contribution of the latter (after renormalization in the MS-scheme) precisely cancels that of the counterterm in Eq. and all constant factors Za have the form Za = 1 + za /ε + · · ·. this seems to contradict the form of the effective potential (3. when turning the background field Φ on and computing the one-loop diagram with external legs |H |2ΦE 2 and internal propagators corresponding to a Higgs and a ΦE field 5 . For given (fixed) values of the Higgs VEV v and the Higgs mass squared m2H . the minimum conditions of the potential read 6 5 In fact this is the only non-trivial one-loop diagram.J.14) fδ = µΦ is necessary in particular to satisfy the RGE given by Eqs.12) and the set (3. when expanded to one-loop order.11) of β-functions. . the presence of the dangerous log(µ/µΦ ) term. all couplings and fields are renormalized. (3. µt .13) where LSM bare is the SM bare Lagrangian. Besides. However. / Nuclear Physics B 581 (2000) 61–72 67 To finish this section. The introduction of the finite renormalization of the coupling δ. In the interesting range of scales.12). (2.2).9). 4. f = 1 + (4 δ/κ) log(µ/µΦ ) + · · · generates a (finite counterterm) contribution to the effective potential in the presence of background fields hc and Φ as. and the non-appearance of a (destabilizing) deeper minimum at larger values of the field. The term in fδ . Φ) = 1 2 2 E2 µ δ hc Φ log . this is worrying. altogether. (3. this is explicitly given by 2 1 −ε/2 ZΦ ∂µ ΦE − 1 ZΦ ZM 2 M 2 ΦE 2 Lbare = LSM bare + µ − 2 Φ 1 2 E4 Z 4! Φ Zλs λs Φ 2 − 12 ZΦ ZH Zδ fδ δ|H |2ΦE 2 (3. leading to the one-loop effective potential given in Eq. (2.A. µ˜ Φ = µΦ ).1) the physical (pole) squared Higgs mass M 2 is equal to m2 plus some (small) H H radiative corrections. by Eq. (3.10) implies µ˜ t = µ.

Casas et al.68 J. / Nuclear Physics B 581 (2000) 61–72 .A.

dVeff .

.

= 0. dhc .

hc =v .

d2 Veff .

.

= m2H . dh2c .

(4. we see that for M 2  m2H (which is the usual case) the third term of the right hand side will amount to a huge contribution. (4.12) and will improve it by the RGE (3.6) with the choice (4. λa . along with the SM vacuum. can be traded by v and m2H as:   M2 Nδ 2 2 Nδ 2 v + M log 2Φ − 1 .2) and (3. as expected. (3.5) to (3. The initial conditions for all parameters will be taken at the boundary scales: . the effective potential parameters. acquire an implicit hc -dependence through their dependence on µΦ . κ Nδ 2 . which must be fine-tuned with the value of m2 . for values of N and δ such that Nδ 2 & κ so we will restrict ourselves to the range of values such that Nδ 2  κ.4) of the µΦ scale.hc =v (4.2) where m2SM . m2 = m2SM − κ κ µΦ λ = λSM − M2 Nδ 2 log 2Φ .5)–(3.3) λSM = 2 + κ 2v µ From the second equality in Eq. (3.. must be done.11) using   M2 dλa (hc ) = 1− 2 (4. in order to keep the right scale for mH . d log hc MΦ (hc ) m2 = m2SM − We will consider now the effective potential given by Eqs.1) Using these conditions.4) Then the parameter fixing (4. log MΦ2 µ2Φ = 1. In fact this dependence can be obtained from the second equalities of Eqs.5) λ = λSM − κ Still perturbation theory can be spoiled. Nδ 2 2 v . i.e. Furthermore.6) µΦ β˜λa θ (µΦ − µ). An immediate consequence of the choice (4. (4.2) is cancelled. 2 κ m2H 3λ4t M2 log 2t . (4. µΦ ' MΦ .2).2) we see that in order to preserve the validity of perturbation theory. from the first equality in Eq. κ µΦ (4.4) for µΦ is that all the couplings of the theory. is avoided by choosing µΦ in such a way that the annoying term in (4. (4. which reflects a hierarchy problem. a choice of the scale. m2 and λ. λSM represent the corresponding values as obtained in the pure SM 1 3λ4 m2SM = m2H + t v 2 .11) and (4.2) becomes. This technical problem.

M. where the relevant multiplicity is N = 12. Casas et al. as a function of the number of extra degrees of freedom. e. Actually. In Fig.. Fig. MH . The dotted line corresponds to Veff for the SM extension with δ = 1.g. there appears a lower bound on M. Conversely. can be larger than the scale ΛSM at which the SM instability develops. This lower bound may seem paradoxical. arbitrarily larger. Plot of the effective potential as function of the Higgs field for mH = 100 GeV. but show that the ratio Mmax /ΛSM can be substantially larger than the one estimated there. Smaller values of M also work. Fig. responsible for the cure of a SM instability. / Nuclear Physics B 581 (2000) 61–72 69 Fig. For a typical value of the multiplicity. for the SM (solid line).J. M. 1. M could be as large as ' 10ΛSM . however. µ20 = Mt2 (v). which corresponds to M = 420 TeV. 1 we show a plot of the effective potential Veff for values of the Higgs and topquark masses. It cannot be. This is illustrated by the dashed line.0 = Mφ2 (v)/e (4. The dashed line corresponds to the result of Ref. Also. The system of partial differential equations (3. with the Higgs mass. . which allows to evaluate the effective potential for any value of hc . which shows an instability for values of the field hc ' 125 TeV ≡ ΛSM . 2 shows (solid line) the ratio Mmax /ΛSM for δ = 1. µ2Φ. This is. mH = 100 GeV and Mt (v) = 175 GeV. [8]. Our results confirm the trend observed in that paper.5). we get the conservative bound M . ΛSM and Mmax . coming from the requirement of perturbativity up to the ΛSM scale. the SM results are recovered in the limit when M → ∞ or δ → 0. the case of the MSSM. which puts an upper bound on the scale of new physics. 3 shows the smooth increasing of both.7) and those for m2 and λ will be fixed by (4. δ = 1 and N = 10. Finally.11) is solved by a step-wise procedure. Clearly. corresponding to the stops. 4ΛSM . The previous example explicitly shows that the scale of new physics.A.5) to (3. which shows how the instability is cured by the new physics. N (recall that Nδ 2  κ to preserve perturbativity). N = O(10). obtained in a cruder approximation (oneloop instead of RGE improved). N = 10 and M = 400 TeV.

Other SM extensions. 1–3 show the relation between the scale ΛSM . it is a feature of the particular SM extension we have chosen: the lower M.5). Plot of the ratio Mmax /ΛSM for δ = 1 and mH = 100 GeV as function of the multiplicity N . The shadowed region shows the values not allowed for M. do not present such lower limits on M.A. The solid line and the dashed one shows the RGE improved and one-loop approximation. Fig. ΛSM and Mmin as a function of the Higgs mass. Casas et al. 2. due to the RGE (3.70 J. as has been explained at the beginning of Section 3. may drive more quickly the quartic Higgs coupling λ into a non-perturbative regime. at which the SM develops a instability. which. in particular supersymmetric extensions. 3. On the other hand. Plot of Mmax . / Nuclear Physics B 581 (2000) 61–72 Fig. The numerical results presented in Figs. the sooner the new physics enters. and the maximum value allowed for the scale . respectively. the upper bound on M is quite robust for any conceivable SM extension.

J. implies that the new physics could easily (but not necessarily) escape detection in the present and future accelerators. in particular in Figs. B 317 (1993) 159. Addendum. Lett. Rep. thus giving reliable upper bounds on the scale of the required new physics. Young. . B. Zaglaner. Zhang. A. imply that ΛSM is at least 102 TeV. a relevant question what is the relation between the Higgs mass (or. Unfortunately. we obtain that Mmax ' 4ΛSM . B 385 (1996) 225. H. Quirós. consisting of N extra scalar fields with an invariant mass M. although unrealistic. Phys. B 228 (1989) 139. arguably represents the most efficient way to cure a SM instability. Sher. Quirós.. therefore. B 382 (1996) 374. With the present experimental lower bounds on MH . Feruglio. This model. References [1] [2] [3] [4] [5] [6] [7] [8] [9] M. The numerical results.R. G.W. Lindner. M. Phys. Lett. coupled to the Higgs field with a coupling δ.16]. Lett. a reliable approach to evaluate the effective potential when several different mass scales are present.A. Phys. Sher. / Nuclear Physics B 581 (2000) 61–72 71 of new physics required to cure it.A. J. We have followed the decoupling approach exposed in a previous paper [13]. Casas et al. Casas.e. Phys. M. Phys. B 353 (1995) 257. J. Mmax . B 342 (1995) 171. ΛSM ) and the maximum allowed value of the scale of the new physics. show the relation between ΛSM and Mmax . Haber and F. M. MH & 105 GeV. in the first place. Lett. the scale at which the instability develops. thus signaling the existence of new physics able to cure it. P. Zwirner for very useful discussions.R. Lett. J. Lett. Then. 5. Datta. G. Lett. equivalently. Altarelli. presented in Section 4. X. Conclusions The possible detection of a relatively light Higgs (MH . [15. together with the previous upper bound on M. it is clear that the possible new physics could easily escape detection in the present and future accelerators. J. which puts an upper bound on the scale of new physics. Hung. which can also be reformulated as a multi-scale renormalization approach.L. Phys. M. i. 140 GeV) would imply an instability of the Higgs effective potential. More precisely. similar to those of Refs. B 337 (1994) 141. M. 179 (1989) 273. Espinosa.J. Acknowledgements We thank F. In this paper we have examined this question in a rigorous way. Lett. It is. MH & 105 GeV. M. the present lower bounds on the Higgs mass. Sher. This requires.A. B 331 (1994) 448.R. Sher. Lett. Phys. the mass of the extra particles. Phys. 1–3. Quirós.Q. Casas. H. B 374 (1996) 138. Espinosa. Espinosa. This fact. for δ = O(1) and N = O(10) (similar to the stop sector in the MSSM case). M. we have considered a simple extension of the Standard Model. Isidori. Phys. Phys.

O. Nakano. Carena. M. J.B. M.A. Wiesendanger. Casas et al. J.V. Ford. A. Lett. C. Zheng. Appelquist. Jones. Maekawa. N. Zenin. C 10 (1999) 629.T. Einhorn. B 398 (1997) 342. M. B 553 (1999) 511. Kugo. D.F. Quirós. Nucl. Quirós. H. Ibarra. Eur. M. hep-ph/9602340. H. Phys. Phys. Di Clemente.A. 90 (1993) 405. Phys. hep-ph/9904295. . Bando.M. V. M. Casas. Rev. M. Nucl. Quirós. Prog.R. Phys. B 230 (FS10) (1984) 261.A. Phys. Pirogov. J. C. Di Clemente. Nucl. D 11 (1975) 2856. / Nuclear Physics B 581 (2000) 61–72 Yu.72 [10] [11] [12] [13] [14] [15] [16] [17] [18] J. Theor. Wagner. Phys. T. Phys.E. Rev. D 55 (1997) 2202. Casas. Carazzone. T. B 461 (1996) 407. J. V. Phys. C.

: 06-6721-2332 ext. for hadron–nucleus interactions. VENUS.2]. tel. Osaka 577-8502. Exotics 1. 4025. Chiba 277-8582.Nuclear Physics B 581 (2000) 73–90 www. A parameter Z.8.3 2 1 3 ( 0 0 ) 0 0 2 7 9 .ac. which is related to the inelasticity of hadron–Pb interactions.kindai. The homogeneous-type thick lead chamber has the advantage that we are able to observe the shower development of hadron-induced showers over a range of a few nuclear mean free paths. HDPM and the modified UA5 model.V.6.elsevier. Tamada a. Ohsawa b a Faculty of Science and Engineering.V. Kinki University.∗ . PII: S 0 5 5 0 . accepted 8 May 2000 Abstract Structures of hadron-induced showers observed by the Pamir thick lead chambers are compared with simulations. All rights reserved. revised 12 April 2000. The simulations are made for several models.40Rc Keywords: Emulsion chamber. whereas that for multi hadrons is close to the HDPM model which gives a smaller inelasticity of hK h–Pb i = 0. A. University of Tokyo. Introduction A series of exposures of homogeneous-type thick lead emulsion chambers was carried out on the Pamir mountains by the Pamir collaboration experiment [1.nl/locate/npe A study of hadron–Pb collisions through hadron-induced showers in thick lead chambers M. QGSJET and modified UA5) which give an average inelasticity hK h–Pb i = 0. part of the chambers were shipped to Japan to be analyzed under the agreement of a joint analysis between Moscow State University and Waseda University. The difference of the characteristics between the two categories of hadron-induced showers is discussed.jp. All rights reserved. which enables us to study in detail characteristics of hadron–Pb interactions ∗ tamada@me2. Some results on the analysis of atmospheric families observed by the chamber have been reported elsewhere [3].  2000 Elsevier Science B. Inelasticity. Japan b Institute for Cosmic Ray Research. QGSJET.5– 0. After the photographic processing of X-ray films. is defined and the Z-distribution of experimental data is compared with those of the simulations applying the same procedure of data processing to both sets of data.0 . High energy cosmic-ray hadrons. Japan Received 18 February 2000. fax: 06-6727-4301 0550-3213/00/$ – see front matter  2000 Elsevier Science B. PACS: 94. The Z-distribution for single-isolated hadrons is well reproduced by the models (VENUS.7–0.

there are several attempts to estimate the inelasticity of hadron–Pb collisions [7–9]. and determines their energies. motivated by the following points: (1) The analytic calculation gives only the average distribution of Z. kinel ≡ where Es is the energy of the surviving particle.. Ohsawa / Nuclear Physics B 581 (2000) 73–90 of hadron energy Eh > 102 TeV. 1 The inelasticity of hadron–carbon collisions is estimated by the Pamir emulsion chambers. produced in the respective collisions. E0 . mainly through decay of π 0 ’s. which is defined by X j /E0 = 1 − xs (xs ≡ Es /E0 ). There exist several reports based on data from the thick lead emulsion chambers. Denoting the (γ ) (γ ) energy released into γ -rays in the interaction at the depths t1 < t2 < · · · by E1 . we defined a parameter Z as X (γ ) (γ ) Ei . [7] where it was concluded that hkinel less than 0. Tamada.8 and that the value agrees with the model predictions. We compared the experimental distribution of Z with those by the analytical calculations h–Pb i= where various types of inelasticity distributions were assumed. A hadron. E2 . 2 It is of advantage that the parameter Z is almost independent from the absolute value of the shower energy. Those are on the observed short attenuation mean free path of high energy hadrons in super-families of visible energy sum 6E(γ ) > 700 TeV [2] and on the extraordinary long penetrating showers beyond simple fluctuations of ordinary hadroninduced showers. in the case when the incident particles have a steep energy spectrum like hadrons of our concern. because the distribution is strongly affected by the energy spectrum of incident hadrons. at Since E1 is proportional to the energy sum of produced secondary particles. [9]. .74 M.05 for hadron–Pb collisions in the energy region of Eh ' 10 eV. A. The surviving hadron and the produced hadrons undergo collisions again in the chamber. The value is h–Pb i is not appreciably smaller than the one in Ref. An analysis using simulations is necessary to discuss the fluctuations of the distribution. the parameter Z gives a measure of the inelasticity kinel of the first (i = 1) interaction. We obtain hkinel +0. In this article we discuss the issue again using the simulated events. in particular. . though many of the parameters to describe the processes are distributed.6−0. . Z ≡ E1 / P (γ ) j . P (γ ) the first hadron–Pb interaction and Ei is proportional to the incident hadron energy. incident upon the chamber. [8. . too [10].9] we discussed the inelasticity of hadron–Pb collisions at 1014 eV by analyzing the structure of observed hadron-induced showers. The significance of the inelasticity and the validity of the view are discussed in detail in Ref. This is important. The emulsion chamber detects only showers which are initiated by the γ -rays. 2 One should notice that we assume a view that the final state of multiple particle production consists of one surviving particle and produced particles. 1 In Refs. Apart from the above studies of anomalous characteristics of hadron interactions. The process is repeated until all the hadrons leave the chamber.03 14 0. [4–6]. etc. causes a nuclear collision — multiple particle production — with a Pb nucleus in the chamber.

Eh = kγ Eh . The γ -rays among the produced particles. where kγ is distributed in 0 6 kγ 6 1 (with hkγ i ' 1/6). It is not self-evident that the estimated energies correspond exactly to those radiated in the respective interactions. initiate a cascade in the chamber. piled up alternatingly. produces a shower in the chamber through a chain of electromagnetic interactions. On the other hand.) Hence we can discuss the Z-distributions of the singleisolated hadrons and hadrons in the family separately with reasonable statistics. See the Section 5.0 (D > 0. Therefore there still remains an ambiguity whether the experimental data can be compared directly with theoretical calculations or not. Emulsion chamber The emulsion chamber is a detector which consists of sensitive layers (X-ray films and/or nuclear emulsion plates) and lead plates of 1 cm thickness. In order to make a direct comparison possible. while single-isolated hadrons consist of nucleons and pions with equal abundance. most of which are decay products of π 0 ’s. which consists of 54 “hadrons in the family” and 85 “single-isolated hadrons”. if any. (See Section 2 for the definition of “hadrons in the family” and “single-isolated hadrons”. incident upon the chamber. 3 We use the criterion of P D / cos θ > 8. the darkness D of the shower spots is measured by a micro-photometer with square slit of 200 × 200 µm2 on every X-ray film. One should notice (γ ) that the observed energy of hadron-induced showers Eh is not the energy of hadron Eh . causes a nuclear collision — multiple particle production — with the lead of the chamber. An electron or a photon. That is.3) in the present analysis. Revision of the selection criteria of the events 3 increased the number of events to 139 hadrons in total.e. produces a small dark spot (∼ 0. Tamada..1 mm in radius) on the developed X-ray films at several successive layers (see Fig. The energy of the shower is determined by photometric measurements of the dark spots on the X-ray films. between nucleon and pion collisions. A. The longitudinal development of spot darkness is then compared with the calculated profile of electromagnetic cascades initiated by a single electron–positron pair. (3) In our previous analysis the number of events is 66 with a total observed energy exceeding 30 TeV. we carry out simulations for hadron-induced showers and apply just the same procedure both to experimental and simulated data. the shower. while P E > 30 TeV in the γ i i previous one. γ ) hereafter). 1). 2.M. . the hadrons in the family consist of charged pions except one nucleon. a hadron. (γ ) i. incident upon the emulsion chamber. a bundle of electrons and photons (abbreviated to (e. Roughly speaking. Since the X-ray film is sensitive to charged particles. Therefore one can see the difference of the nuclear interaction characteristics. These spots are visible by the naked-eyes. Experimental procedure 2.1. Ohsawa / Nuclear Physics B 581 (2000) 73–90 75 (γ ) (2) In our previous analysis the energies Ei were estimated by decomposing manually the observed shower transition into respective interactions.

In the course of the development of 4 The typical performance of X-ray films are 1x ' 50 µm for position resolution and 1E/E = 0.2 for relative energy resolution. because they are produced in the atmosphere by a single primary cosmic-ray particle.2. 4 2.2. γ )-induced and hadron-induced showers in the emulsion chambers. (b) Hadron-induced shower: a hadron.1 ∼ 0.5 nuclear mean free paths. 1. because the collision mean free path of nucleons is ∼ 17 cm Pb. It starts at shallow depth in the chamber and develops over several cm of Pb. and it is possible to determine the positions and energies of the showers with high precision. Dth . Ohsawa / Nuclear Physics B 581 (2000) 73–90 Fig. Russian RT6-type X-ray films. produces a shower in the chamber. The detection threshold of the shower is 1 ∼ 4 TeV.3.2 unambiguously. produced in the collision. Each unit of the chamber has an area of 10 m2 and is composed of 60 layers of lead plates of 1 cm thickness. 5 Details of the chambers are given in Ref. A. of the shower spot is ∼ 0. . depending on experimental conditions. producing several aligned showers in the chamber. equivalent to ∼ 3. Tamada. Pamir thick lead chambers Homogeneous-type lead chambers of our present concern have been exposed at the Pamir for one year beginning from July 1988. In this way the emulsion chamber is sensitive both to electromagnetic and hadronic components. The darkness detection threshold. undergoes a nuclear collision with a Pb nucleus in the chamber. we are able to pick up only shower spots with D > 0. 5 Because of the high background of X-ray films after one year of exposure. with photographic emulsion coated on both sides of a plastic base (200 µm thick). and γ -rays. It normally starts deep in the chamber. 2.76 M. γ )-induced shower: an electron or a photon. incident upon the chamber.2 in one or two layers around the shower maximum in the present chambers. The γ -rays among them produced in this process initiate showers. Illustration of (e. the surviving hadron and the produced hadrons again undergo nuclear interactions in the chamber. incident upon the chamber. The hadrons. (c) Showers in a family are parallel to each other. which corresponds to a shower energy of 2–3 TeV approximately. [2]. collide again in the chamber. Events of successive interactions After the first interaction of an incident hadron with a Pb-nucleus. Showers with energy ' 2 TeV have a spot darkness D > 0. (a) (e. are inserted under every 1 cm lead plate except for the first 2 cm of lead. initiate a cascade. created in the collision. which is called “successive interactions”.

the shower spots on one film coincide exactly with those on the other. 2.e. 6 which is called an event of “successive interactions” (Fig. which we call a cosmic-ray “family” (Fig. Tamada. 3). taking into considerations their respective positions and arrival directions. (2) QGSJET [12. Single-isolated showers and showers in a family Shower spots. On the other hand γ -rays produced in the collision. The arrival direction of a shower is determined by measuring. only when the shower energy exceeds the detection threshold. 7 The identification of showers which are members of a family is accomplished by superimposing two X-ray films at different depth in the chamber. Inelastic reactions are simulated by cutting Pomerons.4. 1).13]: both of which are based on the Gribov–Regge theory of multiple Pomeron exchanges. The produced hadrons collide again in the chamber creating further hadrons. We made simulations of these processes. consists of these showers aligned on a line. (1) VENUS [11]. Since only shower spots with darkness exceeding the detection threshold can be found. 3. are tracked through the entire depth of the chamber by constructing a projection map of all the showers. the emulsion chamber detects usually several showers. i. produce a cascade shower in the chamber through the electromagnetic cascade process. which produces many hadrons. Simulations A cosmic-ray hadron incident upon the chamber causes a nuclear collision with a Pb nucleus in the chamber. separated vertically by 200 µm of plastic base. the horizontally projected direction and the distance between a pair of dark spots in the top and the bottom emulsion layers of the X-ray film.M. A. Ohsawa / Nuclear Physics B 581 (2000) 73–90 77 the hadron cascade further γ ’s are created which start electromagnetic showers at various depths in the chamber. 1). The transition curve of shower spot darkness in high energy hadron interactions often shows a wave-like structure with multiple peaks (Fig. 7 A family is produced by one and the same nuclear and electromagnetic cascade process in the atmosphere originating from an arrival of a single primary particle. mainly via decays of π 0 ’s. γ )’s and local nuclear interactions of hadrons. A shower detected in the chamber is either a single-isolated shower. if the showers concerned are parallel. in this case. This process is called a “nuclear cascade”. (3) HDPM [14]: a semi-empirical model extrapolating experimental data based 6 A hadron-induced shower. aligned on a line. which is not accompanied by any other parallel shower of energy above the detection threshold. creating two color strings per Pomeron which subsequently fragment into color neutral hadrons. or a member of a bundle of parallel showers with the same arrival direction. . That is. We use the following four models for hadron–Pb interactions. These showers are detected by the X-ray films.. The differences between the two models is how the string formation and fragmentation is treated. produced by (e. under microscope.

Therefore “the highest energy particle” is often used instead of the surviving particle in order to clarify the definition in the experiments and in the simulations. respectively. The HDPM model gives a smaller inelasticity in hadron– Pb interactions. the probability pce is energy dependent (pce = 0.5–0.58 0. However. Here.53 a Here the inelasticity is given by K h–Pb ≡ 1 − x where x is defined by the energy fraction carried l l by the highest energy baryon in proton–Pb interactions.62 0.7–0. A.76 0. we define xl as the energy fraction carried by the highest energy baryon in proton–Pb interactions and correspondingly as that carried by the highest energy meson identical to the incident one in pion–Pb interactions.78 M.0 log(E) (mb) inel Table 1 Average inelasticity hK h–Pb ia in hadron–Pb interactions at E0 = 100 TeV p–Pb π–Pbb π–Pbc VENUS QGSJET Modified UA5 HDPM 0.4 log(E) (mb) σ π –Pb (E) = 1329 + 81. In the case with charge exchange. 8 In most of the collisions the highest energy particle is identical with the surviving particle and the relation xs = xl holds.66 at E0 = 100 TeV) in the HDPM model and it is energy independent (pce = 0. c No charge exchange process is taken into account.89 0. pce . hK h–Pb i = 0.82 0.85 0. b Charge exchange process is taken into account.75 – 0.6. it is not always possible to discriminate the surviving particle from the produced secondary particles. The charge exchange probability. is an adjustable parameter in HDPM and the modified UA5 model. the charge-exchange process of the surviving pions increases the inelasticity appreciably. For the cross-sections of hadron–Pb interactions of energy E (GeV). The other three models yield hK h–Pb i = 0. that carried by the highest energy meson identical to the incident one in pion–Pb interactions. . Ohsawa / Nuclear Physics B 581 (2000) 73–90 on the Dual Multichain Parton model and (4) a phenomenological UA5 algorithm [15] modified for hadron–nucleus interaction using a geometrical approach [16]. The difference in the average value of hK h–Pb i between p–Pb and π –Pb interactions is small.75 – 0. Then a parameter defined by K ≡ 1 − xl is closely related to the inelasticity kinel defined in Section 1. Tamada. In Table 1 we summarize the average value of K in p(π)–Pb interactions at E = 100 TeV for these four models.3) in the modified UA5 model. close to that of hadron-p collisions.73 0.8. we use the following approximations: p–Pb σinel (E) = 1565 + 68. In π –Pb interactions. 8 Hereafter we call the parameter K “inelasticity”. In the table we show hK h–Pb i for the two cases with and without charge exchange. The inelasticity kinel is usually defined by the formula in Section 1.

D.6 log(E) (mb) inel for E < 1000 GeV. and all hadrons. Λcoll where A is the atomic number of lead and N Avogadro’s number. which are mainly decay products of π 0 ’s. Simulating the electromagnetic cascade development initiated by all γ -rays of Eγ > 1 GeV.   d = d0 1 − exp(−αρe ) with d0 = 8. Ohsawa / Nuclear Physics B 581 (2000) 73–90 79 for E > 1000 GeV and p–Pb σinel (E) = 1672 + 29. depth T throughout the chamber.2 for hadrons in a family. by using the characteristic relation for the Russian RT6-type X-ray film. The electron number density. For γ -rays of Eγ > 1 GeV. Tamada. Shibata [17]. is converted to the local spot darkness. We calculate the development of nuclear and electromagnetic cascades in lead initiated by hadron–Pb interactions under the following assumptions: (1) the energy spectrum of hadrons arriving at the chamber is of power-law type. are followed until their energy falls below 80 GeV or they leave the chamber. is given by  Z −d 10 dS (S is the area of the slit) D = − log10 S and finally we obtain the transition curve of the spot darkness D vs. Electrons and photons are followed until their energies fall below 1 MeV.0 and α = 3. . respectively. (2) the zenith angle distribution of arriving hadrons is I (6 cos θ ) ∝ (cos θ )−8 and (3) the total thickness of the chamber is 60 cm Pb and the sensitive layers are inserted under every 1 cm Pb. Protons and pions of Eh > 30 TeV. In this code. we further calculate the threedimensional electromagnetic cascade development using the Monte-Carlo code formulated by M. and effects of ionization loss and Compton scattering for low energy particles are also taken into account. of X-ray film. The results are p–Pb π –Pb (E) = 17. the LPM (Landau–Pomeranchuk– Migdal) effect on the cross-section for high energy particles.5 (cm Pb) Λcoll (E) = 15.8 for single-isolated hadrons and β = 1. measured by a 200 × 200 µm2 slit. I (> E) ∝ E −β with β = 1.25 µm2 . are sampled from the above energy spectra. A. produced in the collisions during passage through the chamber. The cross-section is then transformed into a collision mean free path by  h–Pb h–Pb (E) = A/ Nσinel (E) .5 log(E) (mb) σ π –Pb (E) = 1456 + 35. Okamoto and T. ρe .9 (cm Pb) and Λcoll at E = 1014 eV. d.M. The spot darkness. we obtain the lateral distribution of the electron number density under every lead plate for the whole depth of the chamber.

Tamada. Estimation of total released energy.1. Ohsawa / Nuclear Physics B 581 (2000) 73–90 The value for the low energy cut on hadrons and γ -rays is motivated by the fact that (1) the emission angle of secondary particles produced by low energy hadron–Pb interactions is large compared with the slit size 200 ×200 µm2 of spot measurement. released in form of γ -rays during the passage through the chamber is estimated by the sum. . A. 6Eγ . which is well above the detection threshold Dth . 6Di (Di > 0. Diagram on total energy.3)/ cos θ . Open circles are average dependences of 6Eγ on 6Di / cos θ . The other three models give similar results.80 M. 2 for the QGSJET model. 6Di . because the shower-spot formation is mainly determined by high energy γ -rays. E1 . Here we set Dmin = 0. θ is zenith angle of the incident hadron. at the first interaction 4. and released energy. (γ ) 4. We observe a fairly good correlation between 6Eγ and 6Di as shown in Fig.3. and (2) the contribution of these secondaries to the formation of a shower spot is negligibly small. 2. Total released energy ΣEγ The total observed energy of the hadron. in its longitudinal development for p–Pb interactions using the QGSJET model. 6Eγ . released in form of gamma-rays and the sum of shower spot darknesses. of the shower-spot darknesses Di > Dmin in its longitudinal development. 6Eγ . Fig.

M. E1 . Tamada. Ohsawa / Nuclear Physics B 581 (2000) 73–90 (γ ) 4.2. The dotted curve is the best-fit for the first 6 layers in which the spot darkness exceeds D > 0. E1 . by a computer algorithm employing the gradient descent method for a search of the chi-square minimum. Energy E1 81 released in the first interaction In order to extract the energy released in the first interaction from the whole shower transition curve of the event. indicating that the present procedure works but overestimates the energy released in the “first interaction” by ∼ 20% on average. and the first pair-creation depth. calculated for showers of electron–positron pair origin taking into account the exact (γ ) chamber structure. Fig. E1 is 16. The overestimation of E1 in case of hadron-induced showers indicates that even in the first peak of the shower transition there exists a contribution from secondary interactions of low energy hadrons produced at “the first interaction”. at the “first interaction”.2 TeV in this example. Here the “first interaction” is defined as the first interaction among all interactions in which the released energy is larger than the detection threshold. A large value of the deviation δ indicates that there are contributions from successive interactions.3 (shown by J (γ ) symbol ). the following fitting procedure is applied for the first 6 layers in which the spot darkness exceeds D > 0. T . 4. An example for the estimation of the released energy. Fig. 3. which is shown in Fig. 1T . 4 shows a distribution of E1 /E1true (γ ) where E1true (γ ) is the true energy released into γ -rays at the “first interaction”.2.2 TeV whereas the true energy E1true (γ ) is 13. (γ ) The distribution has a clear peak at E1 /E1true (γ ) = 1. The estimated energy is very close to the true (γ ) energy in this case. too. where the summation is made for the layers beyond the fit range and σi is the dispersion of darkness of the standard transition curve at the ith layer. The same fit procedure is also applied to γ -ray-induced pure electromagnetic cascades. We also calculate the deviation of the observed shower transition from the best-fit by δ ≡ 6(Di − Distd )/σi .3. the best fit is selected by choosing the energy value. Eth = 4 TeV. From a set of standard transition curves Dstd vs. . A. Fig. 3 shows an example of the above procedure which is applied to a simulated hadron(γ ) induced shower.

E1 and true energy E1true (γ ) released into γ -rays at the “first interaction” for p–Pb interactions using the QGSJET model (solid histogram). Ohsawa / Nuclear Physics B 581 (2000) 73–90 (γ ) (γ ) Fig. in the shower transition is less than 3. (γ ) 6. The energy spectra of hadrons in both categories are different. is larger than 8. Among them 54 are found as members of atmospheric families and the other 85 are isolated ones. 9 and that the sum of spot darknesses. . 4..0. 6Di (Di > 0. we have 139 showers which satisfy the above criteria. Here we identify a shower as a hadron-induced one if 1T > 4 cm Pb or δ > 50. Tamada. E1 /E1true (γ ). of the estimated energy. The process is not straightforward.8 for the latter. and hence we introduce the above criteria to standardize the data. Selection of the events Emulsion chambers detect both (e. Distribution of Z ≡ E1 /6Eγ P (γ ) Applying the procedure described in the Section 4. We demand further that the darkness at shower maximum.82 M.3)/ cos θ . γ )-induced and hadron-induced showers. because our standard procedure of photometric measurement is limited up to D = 3–3.2 for the former and ∼ 1. A. We have to find a way to estimate the darkness on double-sided emulsion layers from the single-sided darkness. In the Pamir thick lead chamber of 66 m2 year exposure. exceeding ∼ 3. The same selection criteria are also applied to simulated hadron-induced showers. Distribution of the ratio. γ )-induced showers are rejected by these criteria.5. 5. The dotted histogram is the result applying exactly the same method to γ -ray induced pure electromagnetic cascades. e. we obtain E1 and Eγ for the above selected events both in the experiment and in the simulations. According to the simulation. almost all (e. We re-define the inelasticity-related parameter Z by 9 In some events the spot darkness is measured after stripping off the emulsion layer on one side of the X-ray film when the darkness on both layers is very high. and the integral power index is ∼ 1. Dmax . which corresponds to a total released energy 6Eγ > 20–30 TeV.g.

and the experimental distribution is close to that of the HDPM model and not compatible with 10 A hadron-induced shower which consists of only one shower gives Z = 1. are nucleons and the other half are mesons. QGSJET and the modified UA5 model. although the statistics of the events is poor. The distributions are compared with those of simulated hadron-induced showers for the single-isolated hadrons (Fig.M. 5.8. the experimental distribution of single-isolated hadrons is described well.2. (γ ) Z ≡ E1 /6Eγ . Ohsawa / Nuclear Physics B 581 (2000) 73–90 83 Fig.6. is found to be small. 5 shows the experimental distributions of Z ≡ E1 / Eγ for single-isolated hadrons and for hadrons in families. Experimental distribution of Z for single-isolated hadrons (closed circles) and for hadrons in families (open circles) observed by Pamir thick lead chambers.7–0. 6 we assume that half of the incident hadrons are protons and the other half are pions both of which have the integral energy spectrum of the power index β = 1. 7 we assume that all incident hadrons are pions with an energy spectrum of spectral index β = 1. 7). of visible energy Eh > 20 TeV. For the HDPM model calculations are carried out also for the case that the charge exchange process is excluded (pce = 0) in pion–Pb interactions.19] employing QGSJET for hadron–air interaction. A. But the data deviate from the HDPM model which predicts a smaller inelasticity hK h–Pb i = 0. 6) and for hadrons in families (Fig. As is seen in the figure. In Fig. On the other hand.20 [18. For hadrons in families.8.5–0. According to the simulations of atmospheric nuclear and electromagnetic cascades using CORSIKA5. on the contrary. The dependence of the Z-distribution on the power index β and the nature of incident particles. Tamada. On average hadrons in families have a smaller value of the parameter Z. According to the calculations more than 80% of hadrons in families are pions. except for the range of Z > 0. more events populate the region of smaller Z. by VENUS. all of which give an average inelasticity of hK h–Pb i = 0. . half of the single(γ ) isolated hadrons. 10 The figure shows that there exist some differences between the two distributions. however. in Fig.8. (γ ) P Fig. mainly pions.

together with a number of (e. 7. Methodological problems In high energy cosmic-ray families we usually find several hadron-induced showers. 6. Simulations are carried out assuming one half of the events are induced by proton–Pb interactions and the other half induced by pion–Pb interactions. VENUS.84 M. those of the other three models. If there exist at least three shower spots created by family members on each layer of X-ray film which continue to the next layer. In the . Ohsawa / Nuclear Physics B 581 (2000) 73–90 (γ ) Fig. we can determine exactly their relative position and construct a projection map of all showers in the family using these shower spots as references. The power index of the integral energy spectrum is assumed to be β = 1. π –Pb (pce = 0) means no charge exchange processes in pion–Pb interactions.8. A. the modified UA5 and the HDPM model. Tamada. Distribution of Z ≡ E1 /6Eγ for single-isolated hadrons. The histograms are those for simulated showers induced by hadron–Pb interactions using QGSJET. Thus none of the models can describe simultaneously the experimental Z-distribution of the two categories of hadrons. γ )-induced showers of parallel incidence.

A.M. In such a case it is not easy to decide whether the shower observed at a deeper depth is a continuation of the one observed at shallower depth or whether it is a newly produced one. In the following. experiment. we have not always these reference showers over the whole depth of the chamber. Consequently the accuracy of the projection map is much worse for single-isolated hadrons compared with that for hadrons in families. The power index of the integral energy spectrum is assumed to be β = 1. . In the case of single-isolated hadrons. we have no such references and we align the shower spots by constructing a projection map of all observed showers so that every shower has the correct direction within errors.2. Ohsawa / Nuclear Physics B 581 (2000) 73–90 85 Fig. 7. Tamada. however. The histograms are those for simulated showers induced by pion–Pb interactions. Distribution of Z for hadrons in families. we examine whether the differences of tracking accuracy are responsible for the difference between the Z-distributions of the single-isolated hadrons and hadrons in families.

A. Possible reason decreasing Z for hadrons in the families We also cannot rule out the possibility that we misidentify two independent hadroninduced showers as a shower of successive interactions induced by a single hadron if the relative distance between them two is small.e. In Fig. The Z-distribution of single-isolated hadrons agrees well. the number of misidentifications is expected to be ∼ 2. Their Z-distribution. Suppose we misidentify a hadron-induced shower of successive interactions as two independent hadron-induced showers when there exists a gap of 5 or more successive layers where the spot darkness is less than Dmin (= 0. along with the results obtained by the simulations using CORSIKA/QGSJET [18. This misidentification tends to make the value of the parameter Z larger.. Possible reason causing Z to be larger for single-isolated hadrons Due to the limited accuracy in constructing shower projection maps.e. The results are shown in Fig. 9 for hadrons in families.86 M. The probability that a hadron accompanies accidentally another hadron in close vicinity. with QGSJET. Ohsawa / Nuclear Physics B 581 (2000) 73–90 7. as showers of successive interactions produced by a single incident hadron. but the HDPM model still fails to describe the experimental distribution.1. The procedure is applied both to experimental and simulated data. as can be seen from the figure. produced by two independent hadrons. those induced by two hadrons. 7.04. This misidentification causes the value of the parameter Z to be smaller. i.. is too small to explain the excess of events with smaller values of Z for hadrons in families. VENUS and the modified UA5 model. The figure indicates that our measurement on relative distances is limited to the region Rmin (hh) & 1 mm. In such a case we apply the selection criteria described in Section 4 and calculate a value of the parameter Z for each of the two showers. if the relative distance between two hadron-induced showers is less than ∼ 1 mm. between a hadron and its nearest neighbor in families of 6E(γ ) > 100 TeV observed in the lead chamber. over the whole range of Z.. which satisfy the selection criteria. their relative distance being less than 1 mm. from 54 to 38. i. 10 we show the distribution of the relative distance. does not change appreciably. The difference in the Z-distribution between the HDPM model and the other three models is reduced in this case. considerably. we misidentify these two showers. Consequently the difference in the Z-distribution between the two categories of hadrons is not due to the differences of their tracking accuracies.2. however. Tamada.3). Having 54 hadrons in families. we cannot rule out the possibility that we misidentify two showers of successive interactions induced by a single hadron as two independent showers.e. i. however. That is. 8 for single-isolated showers and in Fig. This number. is ∼ 0. Rmin (hh). . if the distance between two successive interactions is large and no shower spots can be seen in many layers.19]. The correction for this effect reduces the number of hadrons in families.

The models for hadron–Pb interactions used in the present analysis are VENUS. Tamada. The symbols are same as those of Fig. The former three models predict average inelasticities in hadron–Pb interactions of hK h–Pb i = 0. A. 6. .M. We have investigated the (γ ) Z(≡ E1 /6Eγ )-distribution which is related to the inelasticity of hadron–Pb interactions. QGSJET.7–0. Distribution of Z for single-isolated hadrons assuming that we misidentify one and the same hadron-induced shower as two independent hadron-induced showers if there is a gap of 5 or more successive layers in which the spot darkness is less than 0. whereas the HDPM model yields hK h–Pb i = 0. 8. the modified UA5 and HDPM.8. The simulations are carried out assuming half of the events to be induced by proton–Pb interactions and the other half by pion–Pb interactions.6.5–0. 8. Summary and discussions We have studied in detail the characteristics of hadron–Pb interactions by observing the structure of hadron-induced showers in thick lead chambers. The difference of the average inelasticity is well displayed in the Z-distribution. The Z-distribution was compared with simulations applying just the same procedure both to experimental and simulated data. Ohsawa / Nuclear Physics B 581 (2000) 73–90 87 Fig.3.

8 for hadrons in families. Tamada. 9. 6. The symbols are the same as those of Fig. Same as Fig. Ohsawa / Nuclear Physics B 581 (2000) 73–90 Fig. . The histograms are those for simulated showers induced by pion–Pb interactions. The histogram presents the results for simulated families using CORSIKA/QGSJET. Fig. Distribution of the relative distance between a hadron and its nearest neighbor of visible energy E(γ ) > 10 TeV in cosmic-ray families of total visible energy 6E(γ ) > 100 TeV. A.88 M. 10.

e. Extraordinary correlated hadron-bundles remind us of exotic shower-clusters. we would possibly misidentify two independent hadron-induced showers as those produced by successive interactions.8 for the latter. Ohsawa / Nuclear Physics B 581 (2000) 73–90 89 A comparison of experimental data with simulations shows that the Z-distribution of single-isolated hadrons is well reproduced by VENUS. however. Grupen (CERN. require the number of those hadron bundles to be 3–4 times more frequent than expected simply by chance.. The calculations were carried out using IBM RS/6000SP at the computer center of Kinki University. i. The Z-distributions of single-isolated hadrons and hadrons in families can be distorted by the poor accuracy of the projection map of the showers. the power index of the integral spectrum is β ∼ 1. In summary none of the models of hadron–nucleus interactions used in the present analysis can describe all the characteristics of the observed two categories of hadron-induced cascades at the same time. these cannot be made responsible for the observed difference of the Z distribution in the two categories of incident hadrons. C. The authors are grateful to Prof. The particle nature is different in the two categories of hadrons. . Acknowledgements The authors would like to express their sincere thanks to all the members of the Pamir and Chacaltaya collaboration experiments. University of Siegen) for carefully reading the manuscript of the paper. named “mini-clusters”. however. The dependence of the Z-distribution on the power index and the nature of incident hadrons.. It would produce a smaller value of the parameter Z even if the average inelasticity is as large as that of single-isolated hadrons. That is. we assume that the observed difference is not due to an observational bias but rather to the characteristics of both categories of hadrons. One of the possible explanations is to assume the existence of hadron bundles in high-energy cosmic-ray families. Since at present we cannot think of any other causes for the difference. are found to be small. The Z-distribution of hadrons in families. however. if there exists a hadron bundle in which the mutual distance of the constituent hadrons is extremely small. especially the members of the Moscow State University who have been involved in using the thick lead chambers from the beginning of the Pamir experiment. i. Almost the same conclusions are obtained also in the analysis on the width of hadron-induced showers [5]. The energy spectrum of incident hadrons is also different. is close to that of the HDPM model but not compatible with the other three models.M. However. according to the examinations presented above. incident hadrons are mainly pions for hadrons in families while half of them are pions and the other half are nucleons for single-isolated hadrons.e.2 for the former and β ∼ 1.e. QGSJET and the modified UA5 model.. The experimental data. in which hadrons and γ -rays are closely correlated. i.9] and the present one comes from the fact that the previous analysis was made for all hadrons without discriminating single-isolated hadrons and hadrons in families which have different Z distributions. A. which are observed in high-energy cosmic-ray families detected in carbon-chambers of the Chacaltaya and Pamir experiments [20–22]. Tamada. The difference of the conclusion between the previous work [8. less than ∼ 1 mm.

Nuovo Cimento C 19 (N6) (1996) 1017.. Rev. 104. G: Nucl. Wilk. [19] D.R. Schatz. J. Nucl. B (Proc.V. Sharaska. XII (N 111–120). 17th Int. Suppl. Schatz. recheres sur les deformations. 119. Capdeville. 1998. Werner. for Cosmic Ray Research. [9] C.L. S. Z. [7] Pamir Collaboration. B (Proc. Paris. Phys. Nauka. Nucl. A. Phys. 21 (1995) 1387. Gladisz. Nucl. Wlodarczyk. p. Part. A 257 (1987) 155.N. Part. M. Acad. Ostapchenko. Phys. 1984. [20] Chacaltaya and Pamir Collaboration. [5] M. Vol. 1. L. Phys.. Arisawa et al. D. B 370 (1992) 365. Rep. Nucl. on Cosmic-Ray and Particle Physics (Tokyo). Part. Nucl. S. Tamada. Capdeville. [15] UA5 Collaboration. B 424 (1994) 241. V. Kopenkin. Barroso et al. Tamada. G: Nucl. G. Phys.G. Madison Workshop on Proton-Antiproton Collider Physics.S. Int. Pavlov. Knapp. Nucl. Baybrina et al. Lodz. p. in: Trudy FIAN. Phys. J. [18] J. Meth.N. Augusto et al. Wlodarczyk. Rev. T. in: Bulletin de la societe des sciences et des lettres de Lodz. ser. of Tokyo. [12] N. Phys. [3] V. D 61 (1999) 012003. in: Proc. 56 (1993) 105. D 59 (1998) 014025. (Physics) 58 (1994) 1966.. Phys. S. Kalmykov. Fujimoto. B 494 (1997) 3. Suppl. [6] E. Hasegawa. Okamoto. Thow. 319. Heck. 232 (1993) 87. in: Proc. Sci. At. [2] Chacaltaya and Pamir Collaboration.T. Buya. p.N. [14] J. Heck. Z. Gladisz-Dziadus. Bull. G. Instr. Phys. Fiz. Zawiejski. [13] N. 85 (1981) 500. Mikocki. Phys. Y. Suppl. 15 (1989) 909.A.V. L. [11] K. 139. Proc.. J. Cosmic Ray Conf. E. 23 (1997) 2057. G.) 52 (1997) 136. Yad. 1981.. [8] S. B (Proc. Karlsruhe. Phys. [22] S. GmbH Karlsruhe. Kalmykov. Tamada. B 291 (1987) 445.S. Russ.C. Ohsawa / Nuclear Physics B 581 (2000) 73–90 References [1] Pamir Collaboration.. Phys. Kopenkin.N. T. Phys. A. Nucl. Phys. J. Phys.I. AIP Conf. Nucl. Ostapchenko. Alner et al. S. Shibata. Univ. Baradzei et al. J. Proc. Phys. [16] M. T. Knapp. Moscow. Forschungszentrum. [21] S. in Russian. FZKA report 6019. [4] Z.90 M.L.) 52 (1997) 221. 1992. 11. p. G: Nucl.) 52 (1999) 171. 1984. Vol. 56 (3) (1993) 346. Nucl.. . Phys. Hasegawa. [17] M. Vol. [10] G. 154. Inst. Symp.

L. Introduction The subject of this paper can be looked at as an example of time asymmetric quantum theory [1–5]. TX 78712. Bohm 1 .Nuclear Physics B 581 (2000) 91–115 www.V. For the Z-boson this means that MR ≈ MZ − 26 MeV and ΓR ≈ ΓZ − 1. All rights reserved.edu 0550-3213/00/$ – see front matter  2000 Elsevier Science B. This presentation corresponds more 1 bohm@physics. N.80. USA Received 5 January 2000. accepted 25 April 2000 Abstract The ambiguity in the definition for the mass and width of relativistic resonances is discussed. Harshman 2 The University of Texas at Austin.2 .nl/locate/npe On the mass and width of the Z-boson and other relativistic quasistable particles Arno R.-m. starting with the “axiom” of Appendix A for the relativistic case.2 MeV where MZ and ΓZ are the values reported in the particle data tables.utexas. PII: S 0 5 5 0 .utexas. 11. revised 15 March 2000.30.Cp. 13. 14. Then. in particular for the case of the Z-boson. then one should read it (or write it) from the end to the beginning.38. Relativistic Gamow vector.  2000 Elsevier Science B. starting from the problems of defining the resonance amplitude and the mass and width of a quasi-stable particle for the analysis of experimental data — in particular for relativistic particles and focusing on the lineshape of the Z-boson. All rights reserved. PACS: 11.70. one will be led to a particular form for the S-matrix Breit– Wigner and will arrive at the conclusion that the Z-boson mass and width must be the particular parameterization (MR . Relativistic Breit–Wigner. This ambiguity can be removed by requiring that a resonance’s width Γ (defined by a Breit–Wigner lineshape) and lifetime τ (defined by the exponential law) always and exactly fulfill the relation Γ = h¯ /τ . using the Poincaré semigroup representations. Poincaré semigroup. Relativistic resonances. ΓR ) or else the generally desired width-lifetime relation τ = h¯ /ΓR will be violated.elsevier. one needs relativistic Gamow vectors √ which in turn define the resonance’s mass MR as the real part of the square root Re sR of the S-matrix pole position sR .Hp. To justify this. We have chosen instead to present the paper in a heuristic way.V. Time asymmetry 1.3 2 1 3 ( 0 0 ) 0 0 2 4 9 .Dg Keywords: Z-boson lineshape.edu 2 harshman@physics. Austin.

25] based on the Weisskopf–Wigner (WW) methods [26]. (Γ /m)π ± ∼ 10−15 . For Γ /m ∼ 10−14 and less.. The ratio of the resonance characterization parameters (Γ /m) runs over a wide range: from (Γ /m) ∼ 10−2 –10−1 for hadron resonances and (Γ /m) ∼ 3 × 10−2 for the Z-boson to much smaller values for other electroweakly decaying particles. the lineshape cannot be resolved and instead of Γ one measures the lifetime τ . Γ comes from the BW energy distribution and τ from the exponential counting rate. Statements that the exponential law is not exact and the use of alternate forms for the relativistic BW have further confused the connection between the width and the lifetime. . N. spin-parity j P and the center of mass energy squared s. 2. Observationally. One measures τ by a fit to the exponential law [19–23]. Conventionally. though for a long time it had been known that there was no consistent theory of resonance and decay phenomena [18]. These can be combined as s = (m − i(Γ /2))2 among other ways. for the Z-boson in ee¯ → Z → f f¯ [7]).R. the exact definition of m and Γ and the relation of Γ to τ have always been of fundamental importance. Both are characterized by species labels or internal quantum numbers.92 A. elementary particles [6] and are not considered qualitatively different from the latter. Relativistic resonances and quasi-stable particles are listed along with stable. 1.] can be considered as approximations” [18].i a rigorous theory to which these various methods [WW. . no generally accepted theory precisely related h¯ /Γ and τ because there did “not exist h. (Γ /m)π 0 ∼ 10−7 .L. For Γ /m > 10−7 . Γ and τ are different qualities.g. The only difference between the characterizations is that stable particles have a real value for s = m2 > 0 and quasi-stable particles and resonances have a complex value for s. e. Connected with the definition of resonance width is the definition of resonance mass. Γ and m are determined as the lineshape parameters of relativistic Breit–Wigner (BW) amplitudes from the cross sections and asymmetries (e. The definition of τ is based on the exponential law for P the partial decay rates P˙η (t) and/or the decay probability P (t) = η Pη (t). A rigorous description of non-relativistic resonances and decaying states by Gamow kets ψ G = |E − iΓ /2− i has been provided during the last two decades by the rigged Hilbert . The experimental definition or method of determination of Γ and m is quite different for different magnitudes of (Γ /m). where the extraordinary accuracy of the LEP data [6–10] set into motion developments that questioned the standard definitions of a resonance [11– 17]. Only since the LEP data for the Z-boson has accuracy reached a level where these differences have become of practical importance. Nevertheless. In what follows we discuss two different relativistic BW’s.. (Γ /m)µ ∼ 3 × 10−18 and (Γ /m)K 0 ∼ 10−14 . So far.g. which is often also parameterized by the two real values “mass” m and “width” Γ . the identification h¯ /Γ ≈ τ is justified by some heuristic and approximate arguments [24. Bohm. etc. S-matrix theoreticians associate a resonance to a pole sR in the S-matrix √ √ and prefer to define the mass as Re sR = MR instead of the peak Re sR = Mρ or the on-shell definition MZ . Experimentalists have in the past identified it with the peak of the invariant mass distribution which was sufficient since the data were not accurate enough. Harshman / Nuclear Physics B 581 (2000) 91–115 closely to the historical process.

in particular for the Z-boson [6–16. 1 − Re Π 0 (MZ2 ) (3) . D(s) is written D(s) = 1 (s − MZ2 )[1 − Re Π 0 (MZ2 )] − i Im Π(s) 1 1 × . Bohm. For these Gamow kets one can derive both an exact Breit–Wigner with width Γ for the lineshape and an exact. From the latter follows the exact Golden Rule for the partial decay rate of the resonance R into the channel η.A. we explore the use of relativistic Gamow kets to provide unambiguous definitions of mass and width. is based on the on-shell definition of mass and width. The part of the Z-boson propagator proportional to ηµν is given by D(s) = s 1 . − Π(s) (1) − M02 The on-shell renormalization scheme defines the (real) mass MZ and the width ΓZ by the renormalization conditions:  MZ2 = M02 + Re Π MZ2 . the precise relation τ = h¯ /Γ between the differently defined quantities τ and h¯ /Γ was established for non-relativistic resonances.L. 29. Harshman / Nuclear Physics B 581 (2000) 91–115 93 space (RHS) formulation of quantum mechanics [5. exponential time evolution ψ G (t) = e−iEt e−(Γ /2)t ψ G (0).30]. two different values for m and Γ have been obtained from the same experimental data [6–10]. P˙ η (t) = Γη e−Γ t . which we shall denote MZ and ΓZ . The standard approach. (2a) where the prime indicates differentiation with respect to s. = √ 1 − Re Π 0 (MZ2 ) s − MZ2 + i s ΓZ (s) (2b) where the “s-dependent width” ΓZ (s) has been defined by √ s ΓZ (s) ≡ − Im Π(s) . The theoretical foundations of the relativistic Gamow vectors are based on the time asymmetric Poincaré transformations of relativistic spacetime and have recently been presented elsewhere [31]. Guided and motivated by recent discussions concerning the ambiguities in the definition of mass and width of relativistic quasi-stable particles.28]. N.R. Lineshape parameters and relativistic Breit–Wigners The determination of the Z-boson mass and width from the lineshape has been performed with two different definitions of mass and width and two different relativistic Breit–Wigner amplitudes. Using these conditions and expanding Re Π(s) about s = MZ2 . The Z-boson mass and width are defined in perturbation theory by the (transverse) self-energy Π(s) of the Z-boson propagator.27. MZ ΓZ = − Im Π(MZ2 ) 1 − Re Π 0 (MZ2 ) . Thus with the properties of the Gamow vectors. As a result. 2. followed by practically all experimental analyses of the LEP and SLC data [6].

16]. one obtains the standard expression for the j th partial wave amplitude used in the line shape analysis of all experiments: −MZ Bef ΓZ RZ = . (m1 . p.g. (4) aj (s) = √ s − MZ2 + i s ΓZ (s) where the partial widths have the same s dependence: Γe. With the pole position sR .R. Γ2 ) [16] and (M¯ Z . (5) 2 0 MZ 1 − Re Π (MZ ) Inserting this in (4) (and neglecting the irrelevant s-dependence in the numerator). (Mρ . 189]) for the lineshape of the Z-boson √ p − s Γe (s)Γf (s) Z . m1 (9a) (9b) 3. Γρ ) (also called (m2 . to the “relativistic Breit–Wigner with energy-dependent width” (e. (7) Since sR is the pole position of the S-matrix. (MR . it should be gauge invariant. for the Z-boson one calculates near the resonance peak (except for some gauge dependent bosonic contributions) s Im Π(s) =− ΓZ . N.. Harshman / Nuclear Physics B 581 (2000) 91–115 This leads. ΓR ) [14] which is suggested (but not dictated) by the analyticity of the S-matrix as a function of the Mandelstam variable s:   ΓR 2 . Γ1 ) [16] which can be defined in terms of (Mρ .8–10]): sR = Mρ2 − iMρ Γρ . Neglecting the fermion mass. Γ¯Z ) [6.L. another approach has been suggested in which the definition for the mass and width is based on the complex-valued position of the propagator pole: sR − M02 − Π(sR ) = 0. (8) 2. Bohm. Therefore. Γρ ) by     1 ΓR 2 1/2 Mρ = MR 1 − 4 MR     1 ΓR 2 −1/2 and Γρ = ΓR 1 − . (10a) sR = MR − i 2 which is related to (Mρ .f (s) = (const) × ΓZ (s). [6.94 A. Γρ ): q m1 Γρ m1 = Mρ2 + Γρ2 and Γ1 = Mρ or vice versa   2 −1 Γ1 Mρ2 = m21 1 + m1   2 −1 Γ1 and Γρ2 = Γ12 1 + . several alternatives for the mass and width of the Z-boson have been defined [13–16]: 1. (10b) 4 MR . (6) ajZ (s) ≈ 2 2 s − MZ + i(s/MZ )ΓZ s − MZ + i(s/MZ )ΓZ This on-shell renormalization scheme is arbitrary [12–17] and the on-shell mass MZ is gauge dependent at O(g 4 ) and higher and the on-shell width ΓZ is gauge dependent at O(g 6 ) and higher [11.

Γρ ) (or (MR . to compare the parameterizations (MZ . cf. but it also does not rely at all on a perturbation series. the Z-boson pole term of the j th partial amplitude is given by ajR (s) = RZ RZ RZ = = . i.e. Γρ ) we align the positions of the maxima of the two functions |ajZ (s)|2 and |ajR (s)|2 . ΓR2 = ΓZ2 1 − + ··· (12b) MR2 = MZ2 1 − 4 MZ 4 MZ and m1 = MZ .35]. Γρ2 = ΓZ2 1 + . The problems with the use of useful for Γ /m  1 and for the initial decay rate P(t asymptotically free states for resonances in the conventional mass renormalization scheme have been mentioned by Passera and Sirlin [34. 2 s − sR s − Mρ + iMρ Γρ s − (MR − i(ΓR /2))2 (11) where (Mρ .L. The amplitudes ajZ (s) and ajR (s) are different functions of s.or out-states and the Born approximation (as employed in Dirac’s Golden Rule) is only ˙ → 0). (7. Γ1 ) are expressed in terms of (MZ .63) of [33]. This is not only a model independent definition of sR (and therefore free of the arbitrariness due to the different choice of renormalization schemes and free of questions concerning gauge dependence). ΓR ) and (m1 . a resonance with spin j is defined by a (pair of) first order pole(s) at the position(s) sR (and sR∗ ) on the second sheet of the analytically continued j th partial S-matrix element [32].49) and (7. We call (11) also a “relativistic Breit–Wigner” or the S-matrix pole Breit–Wigner. scattering resonances are different from their asymptotically free in.36]. ΓR ).. N. .) are basic S-matrix parameters. as eigenvectors of the free Hamiltonian H0 = Hf + Hf¯ and not the full Hamiltonian H = H0 + V . ΓZ )         3 ΓZ 2 5 ΓZ 2 + · · · . The conventional expressions of Im Π and therewith of Γ treat the unstable state as an asymptotically free state. Γ1 ). These problems have been the starting point for the introduction of the Gamow vectors associated to the S-matrix pole [27. The association of the unstable particle to the pole position sR of the S-matrix element does not specify any particular separation of the pole position into mass and width. In analytic S-matrix theory. ΓZ ): 1         ΓZ 2 −1 ΓZ 2 −1 Mρ2 = MZ2 1 + . (m1 . ΓZ ) and (Mρ . (12c) 1 Equivalently we could have aligned the maximum of the cross section derived from (6) and the maximum of the cross section derived from (11). Harshman / Nuclear Physics B 581 (2000) 91–115 95 These are only a few of many possible definitions. etc.28. the other parameterizations (MR . (12a) MZ MZ Using (9b) and (10b). independent of both the center of mass energy squared s and the gauge choice. Using either the S-matrix definition or the pole definition of the propagator (7). Γρ ) in terms of the value most frequently quoted Z-boson parameters (MZ . where V is the decay interaction. The maximum of |ajR (s)|2 is Mρ2 and the maximum of |ajZ (s)|2 is MZ2 (1 + (ΓZ /MZ )2 )−1 and setting these equal gives (Mρ .R. Γ1 = ΓZ . However. Bohm.A.

These fits confirmed the differences (13) and also yielded Γρ − ΓZ ≈ (1 − 2) MeV. Which of the definitions (Mρ . For the first time one could go beyond the level of the Weisskopf– Wigner approximation and one-loop effects. To select one form over the other.R. allowed for the first time to discuss the problem of the definition of mass and width for an unstable relativistic particle.L. Since both formulas (6) and (11) give an equally good fit to the experimental data (as discussed below). Harshman / Nuclear Physics B 581 (2000) 91–115 Numerically one obtains MR = MZ − 26 MeV. Both forms of the partial wave amplitude (6) and (11) were fitted to the experimental cross sections and asymmetries using the LEP data [7–10]. (MR . as witnessed by [11–16] which as a consequence leads to the question: what is a relativistic quasi-stationary particle? . in the case of the W ± -boson.96 A. With the present experimental data it is totally irrelevant which definition of the Z-boson mass one uses for a global fit of the Standard Model since other Standard Model parameters such as MW or 1/αEM (MZ ) have errors which far exceed the differences (13) [29]. the extraordinary accuracy of the Z-lineshape data. (Also. The exact definition of the Z-boson mass (and other resonance masses) is not a central issue of the Standard Model and the prevailing opinion is that there are many ways to parameterize the experimental data of electroweak physics [37] and m1 is as good a value for the Z-boson mass as Mρ and MR . the ‘correct’ choice for the parameterization of (11) in terms of mass and width is unresolved. Re sR = Mρ .35] and have for this reason been suggested as a useful parameterization of the Z-boson lineshape.e. obtained with great expense of time and effort. empirically one cannot distinguish ajZ (s) and ajR (s) (and probably never will). Phenomenologically it will not make any difference which parameterization is used to store and report the experimental data as long as the definitions do not contradict generally accepted consistency conditions.16. the m1 definition is the mass that appears in the conventional calculation of muon and beta decay [34]. N.) We shall in the following section give arguments in favor of the mass and width definition of a resonance by (MR . Γ1 ) or others should be called the mass and width of the Z-boson? Gauge independence and the S-matrix pole definition do not specify whether the mass is the real √ part of the square root of the pole position. confirming to the present level of accuracy the experimental data and also the Standard Model as a theory. Re sR = MR . Bohm. one must invoke a theoretical principle — the requirement of gauge invariance chooses the Breit–Wigner (11) with s-independent width [14–17]. Γ1 ) defined by (9a) are very useful because for the Z-boson they are numerically very close to the conventional standard model values (MZ . However. Mρ = MZ − 34 MeV. The fit is good (χ 2 ≈ 12 for 12 degrees of freedom). (m1 . Γρ ). or some more complicated separation. ΓZ ). The parameters (m1 . (13) using the experimental values of (MZ . ΓR ).. or the square root of the real √ part. ΓZ ) as seen from (12c) [11. i. Even after that decision has been taken. including electroweak radiative corrections. ΓR ). in favor of the separation (10a) of the complex pole position. This opened up a new tier of inquiries. though not presently needed for the Standard Model fits.

Bohm. (MR .2). with sR = (MR − iΓR /2)2 . Harshman / Nuclear Physics B 581 (2000) 91–115 97 This more universal question asks for a more universal consistency condition. for example one other with (complex) mass squared s = sR2 . In the S-matrix approach the amplitude (16) may be considered as a limiting case of the double resonance amplitude (15) for sR2 → 0. then the partial amplitude contains a sum of two BW’s and a background term RZ R2 + + B(s). ¯ (17) ee¯ → Z γ → ff.16]. (m1 . Appendix A. ΓR ).g. This relation has not been tested in the Standard Model or for more general settings. A justification and qualification of (15) is given in Appendix B. The pole term (11) only describes the part of the scattering which goes through the Z resonance ee¯ → Z → f f¯. does not follow from the analyticity assumptions of S-matrix theory but can be justified under a stronger hypothesis of scattering theory in which the in-states and out-states are strictly distinguished. In order to prepare this derivation.L. corrections. The experimental fits of the lineshape data incorporate more than just (11) or (6).A. We want to choose this separation such that the relation Γ = h/τ ¯ can be derived from the resonance part ajR (s) R of the j th partial amplitude aj (s) = aj (s) + B(s). e. Its inclusion in (16) treats the scattering as a double multichannel resonance process (see [36] Section 20. which is often used in phenomenological analysis. a photon term (“γ -Breit–Wigner”) and a slowly varying background amplitude B(s) which is assumed to be constant in the Z-boson energy region [38]: Rγ RZ + B(s). N. in particular the important role of the background and the separation of the scattering amplitude into a pole term (or a superposition of pole terms) and the background. The γ -Breit–Wigner is the analog of the one photon exchange graph of the internal line QED corrections. background and γ –Z-interference need to be incorporated into the analysis. one of which is the width-lifetime relation Γ = h/τ ¯ . Γ1 ). we need to discuss a little more about the phenomenological aspects of the lineshape analysis. etc. The fits to the LEP data [7–10] use for the amplitude an expression which contained. the separation of the amplitude aj (s) into resonant and nonresonant parts is crucial for the selection of the desired mass and width definition from among the many possibilities (Mρ . though it has been used to determine the width of the muon and fix the value of GF . Γρ ). cf. + (16) aj (s) = s − sR s The external line QED corrections are included by a convolution integral with the basic scattering cross sections given by |aj (s)|2 of (16). Why should this relation not also hold for other elementary particles of electroweak physics? As mentioned in [13. though there is really no S-matrix theory . in addition to the Z-boson Breit–Wigner (11) or (6). (15) aj (s) = s − sR s − sR2 This form (15). If there are also other intermediate particles with j P = 1− .. (14) Even if there is only one intermediate particle with j P = 1− there is always a slowly varying background amplitude B(s) so that aj (s) = ajR (s) + B(s).R.

However. Recently the conventional approach. ΓR ) from (16). as stated above. the fitted values of M∆ and Γ∆ from (11) turn out to be significantly smaller than the conventional values from (6). the Z–γ interference term from (16) is important for the fits of various asymmetries [7]. ΓZ ) from (18) and (MR . and the S-matrix approach. Also. A wide assortment of two real numbers can be extracted from sR and identified with the mass and width [43]. the fits to (11) were better than the fits to (6) and they were independent of the background parameterization. 3. using the pole definition (11). This interpretation is in agreement with our RHS theory of resonance scattering as expressed by the results (15) and (16) and explained in detail in Appendix B.g. recent editions of Ref. The interpretation given to these different values by [39. A similar fit to the S-matrix Breit–Wigner (11) was performed for the experimental data on πp scattering in the ∆ resonance region [42]. Γ (Mh2 )) belong to the ∆ and the background together. For baryon resonances like the ∆33 .. e. The fits of the LEP data to (16) and (18) were equally good and they reproduced the expected difference (13) between the differently defined masses and widths. s (18) After the arbitrariness of this form became known.L. for the hadron resonances the pole definition (11) appears to be slightly preferable phenomenologically. N. were compared for hadron resonances. Therefore. Again. When both the S-matrix definition (11) and the on-mass shell definition (6) were applied to the ρ-meson data [41].98 A. the conclusion was that the S-matrix definition of Mρ and Γρ is phenomenologically preferred. The earlier fits to the LEP data use the amplitude with the s-dependent denominator (6) from the on-mass shell definition: aj (s) = RZ s − MZ2 + is(ΓZ /MZ ) + Rγ + B(s). one could write . Thus the experimental lineshape data of the Z-boson do not favor either of these two definitions of Z-boson mass and width. Bohm. The reason given was that these fitted parameters remained largely independent of the parameterization of the background term B(s) and the ρ–ω interference. [6] list both the 2 pole position sh = (1210 − i 100 2 ) MeV in addition to the conventional parameters Mh = 2 1232 MeV and Γ (Mh ) = 120 MeV. the same conclusion does not follow from the Z-boson data [7– 10] where these two fits are equally good and lead just to different values of the Z-boson parameters. using a Breit–Wigner with energy dependent width (6). As with the ρ.40] is that the pole position sh belongs to the ∆-resonance whereas the conventional parameters (Mh . (MZ . Harshman / Nuclear Physics B 581 (2000) 91–115 justification for (16) except the analogy to (15).R. Lifetime parameter and exponential decay rate Neither the treatment of the resonant propagator based on the complex pole position sR = M02 + Π(sR ) nor the phenomenologically equivalent S-matrix theory definition of a quasi-stable particle by the pole at sR define the mass and width separately. the same kinds of fits were made using the S-matrix definition (16).

as mentioned in Section 1.R.L (t) ≡ e−iH eff t f KS. there is a separate physical meaning for the width (and the resonance mass). the relativistic Gamow vector. It is done by a fit of the experimental counting rate N˙ η (t)/N = P˙ η (t)/Γ to the exponential exp(−t/τ ).S t e−(ΓS. π ± . we want to require that a complete theoretical description of a resonance should provide an unambiguous.L = mS. For this we turn to the second definition of Γ given by the inverse lifetime Γ = h¯ /τ . Conventionally state vectors representing decaying states and resonances are obtained using heuristic finite-dimensional models with an n-dimensional complex “effective Hamiltonian matrix” H eff .L . To establish the width Γ of (11) as the inverse lifetime we have to first establish P˙ η (t) = Γ (η) e−ΓR t (20) as a precise result obtained from the second sheet S-matrix pole definition of a resonance.L (22) and the general state vector is taken as the superposition φ = c s f K S + cL f K L with the time evolution (23a) .L. Lineshape by itself. All relativistic quasistable particles should have a width Γ defined by the Breit–Wigner and a lifetime τ defined by the exponential decay law. where t is the time in the center of mass frame of the decay products η. Harshman / Nuclear Physics B 581 (2000) 91–115   ΓR 2 sR = MR − i = Mρ2 − iMρ Γρ 2 99 (19) and call either MR . exact connection between them: the relation Γ = h¯ /τ should be fulfilled exactly and universally. Although both Γ and τ may not be measured for the same particle due to experimental limitations.L − i 2 From this one concludes f KS. the elementary particle state is defined as the eigenvector of the complex Hamiltonian matrix with a complex eigenvalue   ΓS.0 . There is no principle in analytic S-matrix theory that tells us how to separate the complex number sR into the real mass and width. However.26]. N. This cannot be done for the Z-boson because h¯ /ΓZ would be too short to measure. ΓR or Mρ .L = e−imL. Γρ the resonance mass and width. So we need an intermediary that connects the Breit–Wigner (11) to the exponential law (20). The measurement of the lifetime uses the exponential law for the partial decay rates P˙ η (t) of the quasi-stable particle R into any decay products with the channel quantum numbers η. Bohm. in the two-dimensional neutral Kaon system [44]. This will be the “state vector” of the resonance state. does not determine mass and width separately.A. Then. For example. however accurately determined and however precisely the background and corrections are known.L t /2)f KS.L f KS. K ±. the relationship Γ = h¯ /τ will hold precisely and not just as a result of the WW approximate methods [24. (21) H eff f KS. although it has been done for many other decaying elementary particles like µ± .

Bohm. This deduction of the exponential law has the following deficiencies: 1.R. To justify (20) one should start from the fundamental quantum mechanical formula for the probability of decay from R to η . for t → 0) and has only been obtained for the Dirac–Lippmann–Schwinger eigenkets of real energy |E − i and not for vectors like f KS with complex energies (E − iΓ /2)..e. This R(t) one then equates with the time derivative of the decay probability P˙ η (t). N.100 A. Harshman / Nuclear Physics B 581 (2000) 91–115 φ(t) = e−H eff t φ = cS f KS (t) + cL f KL (t) = cS e−imS t e−ΓS t /2 f KS + cL e−imL t e−ΓL t /2 f LS . (23b) The exponential law (20) is then justified in the following way: one inserts (22) for all values of t into Dirac’s Golden Rule and obtains 2 a rate R(t) for all values of t.L. Dirac’s Golden Rule is only an approximation for the initial decay rate (i.

.

 (24) Pη (t) = Tr Λη .

ψ G (t) ψ G (t).

within the RHS formulation of time asymmetric quantum theory by augmenting the conventional assumptions of scattering theory with the new hypothesis described in the Appendix A. etc. From this new hypothesis one derives the form of the j th partial amplitude (15) for two resonance poles of the partial S-matrix and. All these things can be justified.. ΓZ or others) of any relativistic Breit–Wigner. one first has to show that the eigenvectors with complex eigenvalue (21) have a meaning for self-adjoint H . It is unrelated to the width of a lineshape (11) because (21) and its underlying assumptions do not imply a precise relation between ΓS and any width Γ (either ΓR . (6). The widths ΓS are ΓL defined by (21) as the imaginary part of a complex energy and from (22) one sees this is thus just another name for h¯ /τ . usually the representation space of the Poincaré group of transformations of spacetime. e. 2. corresponding to (15) term by term. The energy operator H (or in the rest system. . which in case of the ee¯ of (17) is given by (cf.g. H ≡ P0 = M = (Pµ P µ )1/2 for the relativistic case like the K-meson) is a self-adjoint operator in an infinite-dimensional space. To justify τ = h¯ /Γ one must relate ΓS of (21) to the ΓR (or Γρ . Then the probability rate (20) should result as the time derivative of Pη (t) [28]. that they can be elements of a basis system for the infinite-dimensional space and that the truncation to the two-dimensional subspace gives (23a) as an approximation. 3. To justify the two-dimensional “complex” eigenvector expansion of a general state vector φ (23a). where ψ G describes the decaying state R and Λη is the projection operator on the space of decay products η. Appendix B): X. a complex basis vector decomposition for the prepared in-state φ + . with some qualifications. Γρ .) of the pole definition of the lineshape (11) or any other definition for the lineshape.

.

but this matter is not the subject of the present paper. one inserts (23b) into Dirac’s Golden Rule and obtains the decay rate for the neutral kaon state considered as a superposition of two exponentially decaying states.j. . sR . b− cR + φ bg φ+ = i i i Ri 2 Additionally.

L.R. Harshman / Nuclear Physics B 581 (2000) 91–115 . N. Bohm.A.

.

− − = .

bR cR + . sR .j = 1− .

from which we can then define the fields. 3 The terms in the basis vector decomposition (25) correspond to the terms in the multiresonance partial amplitude (15). s) of P. In the next section we will show how to obtain the decomposition of the in-state into a sum over resonances and a non-resonant part by the analytic continuation of the partial S-matrix. s. The vector φ bg is the component of the (arbitrary) in-state vector φ + in the infinite-dimensional subspace representing the non-resonant continuum and it corresponds to the B(s) term in (15). b− i. sR2 . b+i and which fulfill the Lippmann–Schwinger equation. It is because we can obtain both the BW amplitude (11) and the resonance vector |j. In (25) the quantum numbers b denote a set of degeneracy labels. They are eigenkets of the (self-adjoint) invariant mass squared operator Pµ P µ ≡ M 2 with real eigenvalue s: Pµ P µ |j. s. b−i from the S-matrix pole that we can exactly and precisely establish the relation τ = h¯ /ΓR . b− i in the sum over Ri in (25) comes from the pole at sRi in the second sheet of the S-matrix and the vector φ bg is represented by a background integral (see (32) below).j = 1− . For stable relativistic elementary particles the vector space description is given by the irreducible unitary representation spaces of the Poincaré group P [45]. sRi . sR . are basis vectors of an irreducible unitary representation (j. a few comments are in order about the vectors |j. b−i. bR cR2 + φ bg . This is not restricted to the asymptotic. The ‘out’ and ‘in’ “states” of [46]. which we denote by |j. s. sR . interaction free states [46]. b−i = s|j. s. Before continuing with the details of the derivation in the following section. 2 101 (25) This expression (25) is reminiscent of the heuristic expansion (23a) except that the heuristic expressions do not include the background term φ bg . To each vector in the sum over the resonances Ri in (25) corresponds a BW in the j th partial amplitude (15). which (except for a normalization factor) are the relativistic Gamow vectors ψ G . The vector |j. . b−i and |j.

√ .

− − = s .

s. P0 .j. b = brest .

. sR . In place of (26) and (27) for the Dirac–Lippmann–Schwinger kets |j.j. s. b− i are thus just generalizations of the wellestablished ‘out-states’ |j. sR . b −i and are obtained from them by analytic continuation in s from its “physical value” {s|(me + me¯ )2 6 s < ∞} the the S-matrix pole position sR in the second sheet (or if there are several poles. b = brest (26) (27) We want to consider the Z-boson (or any quasistable relativistic particle) as a fundamental particle in the Wigner sense and therefore give its description in terms of a representation space of Poincaré group transformations. s. a Gamow ket |j. s. sRi . b− i is obtained for each pole sRi ). b±i. b− i of (25) should therefore also be basis vectors of an irreducible representation (j. The Gamow kets |j. The Gamow vectors |j. the Gamow kets should be generalized eigenvectors of the self-adjoint mass-squared operator Pµ P µ ≡ M 2 with complex eigenvalue sR = (MR + iΓR /2)2 : 3 For b one could choose the momentum p and j but we suggest the spatial components of the 4-velocity m 3 pˆ m = pm /s. sR ) of Poincaré transformations. for a detailed discussion see [31]. However this is not important for our discussions here.

b−i = sR |j. Harshman / Nuclear Physics B 581 (2000) 91–115 Pµ P µ |j.R. Bohm.102 A. b− i.L. sR . sR .   . N.

ΓR .

.

b = brest = MR − i . P0 . − − j. sR .

We start with the S-matrix element   ψ out . Because the Gamow vector distinguishes ΓR as the characterizing parameter.1)). The eigenvalue equation (29) agrees with the effective complex Hamiltonian eigenvalue equation (21) and the decomposition (23a) is the truncation of (25) to the resonance subspace if the background φ bg (or equivalently B(s) in (15)) is neglected. More details will be given in the following section. Sφ in = Ω − ψ out . Thus the Lee–Oehme–Yang theory [44] of the neutral kaon system summarized by (21)–(23b) and other finite-dimensional effective theories with complex Hamiltonians [47–50] will be established as approximations of the RHS theory after (25). it implies a preferred separation of the complex value sR into two real parameters (MR . (29) have been justified in the next section. b− i are defined as continuous functionals over the space of observables Φ+ (cf. (28).j. ΓR ): sR = (MR − iΓR /2)2 . Relating the BW lineshape to the exponential decay law: relativistic Gamow vectors The mathematical object that makes the connection between the Γ that appears in the lineshape (11) and the Γ that appears in the exponential decay law (20) is the relativistic Gamow vector ψ G . Appendix (A. Since both come from the same vector characterized by ΓR . The Gamow vector ψ G will on the one hand lead to the lineshape (11) and on the other hand to a precise form of the exponential law (22). but for further details see [31]. sR . b = brest 2 (28) (29) These eigenvalue equations are exact and mathematically rigorous if the |j. the lineshape-Γ and the inverse lifetime-Γ are the same and given by ΓR . Ω + φ in = (ψ − . 4. The vector ψ G fulfills (28) and (29) precisely and is obtained from the pole term of the S-matrix. sR . φ + ) X Z d 3 p0 X Z d 3p .

ds ds 0 ψ − .

j3 j 0 j30 .pj3 [sj ]− = 0 0 2E(p) 2E (p ) j.

.

.

× pj3 [sj ].

S .

p0 j30 [s 0 j 0 ] + p0 j30 [s 0 j 0 ].

For the case under consideration we are only interested in the term of the sum with j = 1− . f f¯). (30) = j (me +me¯ )2 b where we have made use of the symmetry properties of the S-matrix and abbreviated all additional quantum numbers by b.. b|φ +i. b−iSj (s)h+ j. s.φ + ∞ X X Z ds hψ − |j.g. To obtain the Gamow vectors .g. s.. e+ e− ) and ψ − ∈ Φ+ represents the detected out-state of the decay products (e. Here φ + ∈ Φ− represents the prepared in-state (e.

s. Bohm. the integral in (30) splits off a pole term which gives the Gamow vector |j. This we also write as a functional equation in the space Φ+ . Harshman / Nuclear Physics B 581 (2000) 91–115 103 one proceeds in exactly the same way as in the non-relativistic case [27]. This is explained in more detail for the case of two resonances in Appendix B. s. . One deforms the contour of integration from the upper rim of the cut along the positive real axis (me + me )2 6 s < ∞ through the cut into the second (or higher) sheet past the pole at sR (or past the poles at sR .A. For the contour around each resonance pole sR . s − sR (31a) (31b) × : for all ψ − ∈ Φ+ . b−i −∞I I 1 . s.R. .L. sR . if there is more than one resonance pole). the resonance pole: − ψ |j. b−i Z+∞ 1 s − sR ds hψ − |j.sR2 . The Gamow kets are defined (using the Cauchy formula) by the contour integral around sR . b − i ≡− 2π i = 2π I ds hψ − |j. b− i as an analytic continuation of the Dirac–Lippmann–Schwinger kets |j. s = sR . N. . b−i.

.

.g. along the negative real axis on the second sheet).j. By omitting the arbitrary ψ − ∈ Φ+ in (30) one arrives at the complex basis vector decomposition (25) for the in-state vector φ + where φ bg is given by (32). This is the new hypothesis of the Appendix A and here means specifically that the energy wave functions h− j. extends from −∞II to +∞. sR . s. In this way one arrives at the representation (15) for the j th partial amplitude (or similarly for the j th partial Smatrix hb 0 |Sj (s)|bi = Sj (s)). we have obtained by contour deformation the BW amplitude (15) and the basis vector decomposition (25) from the S-matrix element and established the correspondence between the terms of these equations. b−i s − sR (31c) As explained in more details in Appendix B. the integral in (30) becomes a sum of such pole terms (one for each sRi ) plus a background integral Z ds hψ − |j. however for “unphysical” values of s 6 (me + me¯ )2 = 4m2e these values are in the second sheet. s. b|φ +i ≡ hψ − |φ bg i (32) C over a contour C. (15). b−iSj (s)h+ j. Summarizing. From (31b) we see that the variable s in (11). b− ≡ i 2π Z+∞ ds −∞I I |j. s. etc. which we are largely free to choose far away from the resonance poles (e. In order to perform the contour deformation that separates (30) into a sum over resonant terms like (31a) plus the background term (32) and in order to derive the Breit–Wigner energy distribution of (31b) from the pole in (31a) some mathematical properties must be fulfilled in addition to the conventional assumptions. b|ψ − i . s.

g. s.R. Harshman / Nuclear Physics B 581 (2000) 91–115 and h+ j. b− i defined by (31a) satisfy the eigenvalue equations − . Bohm. N. h− s|ψ − i ∈ S ∩ H+ The relativistic Gamow vectors |j..L. sR . e.104 A. b|φ +i must be well-behaved Hardy class functions 4 of the upper and lower half-plane. 2. respectively. in the second sheet of the energy surface of the S-matrix.

2 ×.

s. sR . b − i = 2π Z+∞ ds shψ − |j. b−i (33) × Φ+ . b− i → |j. sR . one can show that the Gamow vectors in the for every ψ − ∈ Φ+ ⊂ H ⊂ − i are generalized eigenvectors of the energy operator rest frame |j. sR . Similarly. s. b−i −∞ − 1 s − sR = sR hψ |j. brest H = P0 − hψ |H . hψ |(M ) j.

×.

− j. brest Z+∞ 1 √ . sR .

i − ds s ψ − .

j. brest = 2π s − sR −∞   ΓR − . s.

.

. The eigenvalue equations (33) and (34) are precise formulations of (28) and (29). . sR . Continuous linear combinations of the Gamow kets with an arbitrary weight function φj3 (b) ∈ S (Schwartz space) XZ G (35) ψj sR = dµ(b)|j. − ψ j.} in the sense of Dirac kets. sR . b−iφj3 (b). The operators H × and (M 2 )× denote the conjugate × of the operators H and M 2 ≡ Pµ P µ in the space Φ+ and are defined operators in Φ+ − − by hH ψ |F i ≡ hψ − |H × |F − i and hM 2 ψ − |F − i ≡ hψ − |(M 2 )× |F − i for all ψ − ∈ Φ+ × and F − ∈ Φ+ (see [27] for more details on their definition and use). b|ψj sR i ∝ (s − sR )−1 . brest = MR − i . s. j3 √ also represent relativistic Gamow states with complex mass sR = (MR − iΓR /2) and a Breit–Wigner energy distribution h− j. . b2 . We calculate for all ψ − ∈ Φ+ iH t − . sR . 2 (34) √ where sR = (MR − i(ΓR /2)). where t is the proper time. b− i are also generalized eigenvectors with generalized eigenvalues b = {b1 . For the time evolution we consider here only the Gamow states at rest. The Gamow kets |j.

× .

− − e ψ .

sR . brest = hψ − |e−iH t .

(33). one obtains (31a). .sR . (34) and more.2) of Appendix A are well-behaved Hardy functions. brest Z × hψ − |e−iH t |s. which according to hypothesis (A. Choosing G− (s) = s n/2 hψ − |s − i. brest− i i ds = 2π s − sR 4 The proof of the second equality in (31a) and of relations (33) and (34) is a consequence of the following Titchmarsh theorem for Hardy functions G− (s) ∈ H2− : for G− (s) ∈ H− (Hardy class in the lower plane) R +∞ and Im z < 0 one has G− (z) = −1/(2π i) −∞ G− (s)/(s − z) ds.

R.L.A. N. Harshman / Nuclear Physics B 581 (2000) 91–115 Z √ s t hψ − |s. Bohm. b − i rest ds s − sR √ .

− = e−i sR t ψ − .

sR . This is fulfilled only for t > 0 and not for t < 0. brest . The dual operator of eiH t . but only for × t > 0. we obtain the time evolution of the relativistic Gamow states at rest: . Thus. we have exponential time evolution. i = 2π e−i 105 (36) √ only then The last equality holds iff hψ − |s − i ∈ H− and e−i st hψ − |s − i ∈ H− because √ can one also use the Titchmarsh formula of footnote 4 for G− (s) = e−i st hψ − |s − i (π < √ arg s 6 2π ) and obtain G− (sR ). due to the new hypothesis of Appendix A. Omitting the arbitrary ψ − ∈ Φ+ in (36). is not defined for t < 0 because eiH t is not a continuous operator in Φ+ . (eiH t )× ≡ e−iH t .

× .

− − = e−iMR t e−ΓR t /2 .

(37) e−iH t . sR .j. brest . t > 0.

the time evolution of the non-resonant background φ bg is nonexponential and depends upon the particular choice of the prepared in-state φ + as seen from (32). 5 It is not universally accepted that the Z-boson or any resonance should have a purely exponential time evolution. 5. [51]. That it holds for t > 0 only does not constitute a limitation for the physics. While the time evolution of the resonance terms in (25) depends only upon the parameters (MRi . brest This is a functional equation over the space Φ+ = {ψ − }. and from (37) it follows that the lifetime is exactly τ = h¯ /ΓR and not h¯ /ΓZ or h¯ /Γρ . But this belief is probably only based on the widely known result that exponential time × evolution is prohibited by the mathematics (topology) of the Hilbert space. described by the Gamow vector |j. The time evolution of |hψ − |φ + (t)i|2 can be very close to exponential if φ + is prepared such that φ bg ≈ 0 5 (and there is only one resonance in (25)). Since these are the signatures of a quasistable particle and a relativistic resonance. e. sR− i. The time evolution (37) holds for every Gamow vector in the basis vector expansion (25).j. t > 0 describes the physical situation correctly because the decay products η. On the basis of (37) we want MR and ΓR to define “mass” and “width” of a quasi-stable relativistic particle. . The Gamov state vectors ψjGsR ∈ Φ+ are not of H. describing the detected decay products. has been created at t = 0 [5]. sR i|2 of the decay products ψ − . described by ψ − . ΓRi ) and is exponential. but the (apparatus-prepared) in-states φ + are. we want to assign the relativistic Gamow vectors as the “state vectors” of quasi-stable relativistic particles.. This means it only makes mathematical sense when used to calculate the proba× bilities |hψ − |e−iH t |j.g. sR . can only be detected after the decaying state R. Summary and conclusion The relativistic Gamow vector has the exact exponential time evolution (37) and the exact S-matrix pole Breit–Wigner energy distribution (11). the “scalar products” of (37) with φ + or ψjGsR are not defined. This is as far as we can establish the heuristic equation (22) by the mathematically rigorous result (37). Just to the contrary. Therewith the time evolution of the heuristic state vector (23b) as a superposition of two exponentials can also be justified for t > 0 if φ bg in (25) can be neglected.

However.0024 GeV is its first prediction. . Fixing the mass and width could only be done using the new hypothesis of time asymmetric quantum physics. N. 6 This prediction was obtained. The accuracy with which the exponential law has been observed in some cases [19–23] shows that the isolation of a microphysical Gamow state ψ G from the background φ bg can be very good. the basis vector expansion (25) or its truncation to the complex “effective” theories like (21)– (23b). they still are considered as separate entities.1541 ± 0. Since the perturbation theoretical understanding of resonances fosters the idea that resonances are complicated phenomena. To define a quantum physical entity entirely by the Smatrix pole alone would be incomplete. The only result which may be difficult to accept is the semigroup property of the time evolution (37). This differs from the conventional mass definition by the lineshape (6) of the on-mass shell renormalization scheme MZ = 91. the representation (15) of interfering Breit–Wigners associated to (25) all have been introduced before as separate assumptions and it may be welcome that here they all follow from the same new hypothesis about the boundary conditions.g. in (23a. Mρ = 91.0031 GeV and the width (by (37) the inverse lifetime) given by ΓR = 2. from the definition of the resonance as a pole of the S-matrix at sR .0021 GeV and it also differs from the definition by the peak position of the relativistic BW (11).L. If the description of resonances by relativistic Gamow vectors is valid. A vector description (or in general a density operator description) is needed because the fundamental probabilities (20) and (24) of quantum mechanics are calculated in terms of operators or vectors. In the relativistic theory this means that Gamow vectors can only undergo Poincaré transformations into the forward light cone [31]. although stable elementary particles also never occur in total isolation. The popularity of the effective theories with finite-dimensional complex Hamiltonian matrices not only in particle physics [44] but also in other areas [47–50] and the phenomenological success of formulas like (15) or (B.106 A. then the lineshape (11) with mass (and by (37)) resonance energy in the rest frame) given by MR = 91. These Poincaré transformations form only a semigroup P+ = {(a.b)).4934 ± 0. Bohm. where Λ is a proper orthochronous Lorentz 6 The value of M quoted here is the ‘Average Value’ in Table 2 of [52]. one may doubt the utility of separating out a well-defined Gamow state vector and a well-defined Breit–Wigner amplitude and of associating these with the resonance state.R.0031 GeV. The other results of this theory like the exponential law with the precise τ = h/Γ ¯ R .1626 ±0. (A. The relativistic Gamow vector is not more nor less of an idealization of reality than Wigner’s unitary representations for stable particles. Harshman / Nuclear Physics B 581 (2000) 91–115 The resonance always occurs with at least some background..1871 ± 0. described by B(s) of the amplitude (15) and alternatively described by the much lesser known φ bg of the prepared state φ + (much lesser known because it is standardly ignored.19). which by itself is insufficient to fix MR = √ √ Re sR and ΓR = −2 Im sR . Λ)}.2) of the appendix. like the relativistic Gamow vector. e.1) and (A. The value of M usually called M ¯Z ρ Z and Γρ usually called Γ¯Z are the ‘S-matrix fit Average Value’ from Table 12 of [53].18) for interfering resonances testifies to the usefulness of separating exponentially evolving Gamow state vectors and ideal Breit–Wigner amplitudes related to them by the correspondence (B.

7 One checks by direct calculations that (a . Bohm. Dicus and G. we admit as physically valid only the one which agrees with our intuitive notion of causality. The semigroup time evolution of the Gamow states is just an example of this general time asymmetry. Λ ) ◦ (a . N.A. Support of the Welch Foundation is gratefully acknowledged. Λ ) = (a + Λ a . Time asymmetric quantum theory of scattering In this section we state the new hypothesis by which time asymmetric quantum theory (TAQT) differs from the assumptions of standard scattering theory.. T.L. In the meanwhile. It expresses a time asymmetry on the microphysical level which is connected with neither the violation of time reversal invariance of the Hamiltonian nor entropy increase (at least not in an obvious way. . Λ) ∈ P+ has an inverse in P+ . we need two rigged Hilbert spaces (RHS) with the same Hilbert space H. The dynamical (differential) equations. the algebras of observables and the physical interpretation (probability) remain the same as in time symmetric quantum theory in the Hilbert space H. Of the two alternate ways of calculating the T -matrix [54]. Riemann for correspondence and discussions which provided further phenomenological background on the Z-boson. and D. The only difference is that TAQT uses time asymmetric boundary conditions. i. Acknowledgements Augusto Garcia first drew our attention to the problem of mass definition for resonances as discussed in [15] and [42]. Appendix A.e. Harshman / Nuclear Physics B 581 (2000) 91–115 107 p transformation and a = (a0 . Mondragon for valuable discussions on the subject. Wickramasekara for some discussions on the mathematics used in Appendix B and I. We also want to thank S. In order to give this a mathematical formulation. We are grateful to the following persons for their contributions: A. P+ is a semigroup. Osipenkov for some calculational help. the irreversible character of quantum mechanical decay has been mentioned in a few textbooks [2. Sirlin for valuable comments and suggestions.R. Lopez Castro for discussions and advice on the perturbation theory definitions of Z-boson mass and width and the problems with gauge invariance and for reading and improving on the manuscript. a) is a four vector which fulfills 7 pˆ · a = 1 + pˆ 2 a0 − pˆ · a > 0 for any pˆ ∈ R3 . Λ ) ∈ P + + i i 1 1 2 2 1 1 2 1 2 but that not every (a. We are also grateful to A.3] and more general considerations support the existence of a fundamental time asymmetry in the quantum theory of cosmology [1] and of microsystems [4]. This semigroup property was a surprising and unintended result when it was first derived for the non-relativistic theory. an assumption that is already implicit in the heuristic Lippmann–Schwinger (integral) equations used for the calculation of the transition matrix (T -matrix). although for a contrary opinion see [1]). Λ Λ ) ∈ P for (a .

in addition to the Dirac–Lippmann–Schwinger kets |E ± i ∈ Φ∓ × . The dual spaces in the RHS’s × . The space Φ− (Φ+ ) is of Hardy class type in the lower (upper) half plane. The notation |R+ means the restriction to the positive real line.108 A. The vectors φ + represent states that are prepared by the accelerator and the vectors ψ − represent observables or out-states that are defined by the detector. the physical values of energy. respectively. N. It is remarkable that from this single new assumption (A.2) of the Hardy class property such a variety of new results follow. At least in the non-relativistic theory [5.2b) − − − ψ ∈ Φ+ ⇔ h E|ψ i ∈ S This condition allows the energy wave functions to be analytically continued into wellbehaved.R. and S denotes the Schwartz space.1b) − × ∈ Φ+ ⊂ H ⊂ Φ+ .e. i.1a) ψ (A. the complex eigenvector decomposition (25) and the correspondence between the terms in the sums.L. also Gamow kets contain. Harshman / Nuclear Physics B 581 (2000) 91–115 The relativistic Gamow vector is a precisely defined mathematical object in the RHS formulation of quantum theory. This new quantum theory is not a drastic departure from the usual Hilbert space (HS) formulation but an extension so that the theory includes Dirac kets |Ei and the solutions of the Lippmann–Schwinger equation |E − i and |E + i which fulfill outgoing and incoming boundary conditions. φ + ∈ Φ− ⇔ h+ E|φ + i ∈ S ∩ H− (A. . rapidly decreasing analytic functions in the lower or upper half plane of the second sheet of the S-matrix. Representing a superposition of resonances with Gamow vectors and with Breit–Wigner amplitudes In this appendix we want to derive the expansion of the scattering amplitude (15).28] one can connect this mathematical assumption to the causality condition that a state must be prepared at a time t0 before an observable can be measured at time t > t0 . this means mathematically that the energy wave functions h+ E|φ + i (h− E|ψ − i) are well-behaved 2| 2 Hardy class functions in the lower (upper) half plane. These results are consequences of the new hypothesis of Appendix A and the assumption (A.2a) 2 ∩ H+ |R + .. The RHS and the HS theory have the same (time symmetric) dynamical equations but different boundary conditions. Bohm. (A. S ∩ H− R+ (S ∩ H+ |R+ ): 2 |R + . φ + ∈ Φ− ⊂ H ⊂ Φ− (A. The standard HS boundary conditions are time symmetric: the space of in-states {φ + } and the space of out-states {ψ − } of scattering theory are identified with the Hilbert space H = {φ + } = {ψ − } or at least with the same subspace thereof {φ + } = {ψ − } ⊂ H. The RHS theory distinguishes meticulously between prepared states (in-states) {φ + } and observables (out-states) {ψ − } for which it uses two RHS’s of Hardy class × . |ER ± i(Γ /2)± i ∈ Φ∓ This mathematical assumption of distinct RHS’s of Hardy class for states and observables is the new hypothesis of TAQT. Appendix B.2) will be employed extensively in the derivation.

this can be immediately carried over to the relativistic case. (B.s.1) or (A.e.A.. φ )j = dE hψ − |Ej − iSj (E)h+ Ej |φ + i. their energy wave ∗ functions hψ − |Ej − i = h− Ej |ψ − i and h+ E|φ + i are not just Lebesgue square integrable as in standard Hilbert space quantum mechanics. (B. i.1) becomes (dropping the j notation) |E ± i = |Ei + Z−∞ bg dE hψ − |E − iSII (E)h+ E|φ + i (ψ . 2. the analytically continued S-matrix Sj (z) has two (or countably many) resonance poles in the second sheet below the cut at positions zRi = ERi − iΓRi /2.g.3) E − H ± i0 With the assumption (B.1) 0 The integration extends over the physical values (spectrum of H ) of energy which starts at E = 0 and runs along the upper edge of the cut in the Riemann surface of Sj (E). j = 1. the eigenvectors |Ei of the free Hamiltonian H0 are connected to the eigenvectors |E ± i of the full Hamiltonian H = H0 + V by the Lippmann–Schwinger equation or the Møller wave operators: 1 V |Ei = Ω ± |Ei.1) can be analytically continued into the lower half plane of the second sheet and the values of the functions are completely determined by their values on the positive real axis R+ .R. the integrand in (B. Now one can deform the contour of integration from the positive real line through the cut into the lower half plane as shown in Fig. hψ out |Ei = hψ − |E − i ∈ S ∩ H− 2 hE|φ in i = h+ E|φ + i ∈ S ∩ H− . e. i = 1. Therefore. φ ) = − + 0 I + C1 dω hψ − |ω− iSII (ω)h+ ω|φ + i . and ignore all the other terms in the sum over j on the r. where |hE|φ in i|2 describes the energy distribution in the beam and the energy resolution of the detector is given by |hE|ψ out i|2 . Bohm. but fulfill a stronger condition: 2 .2). We shall make use of the Hardy class property given by our new hypothesis (causality) (A.2).2) In the above relation. We also ignore the additional quantum numbers b and use the nonrelativistic notation.s.) Then we have for the j th partial S-matrix element − + Z∞ (ψ . (B. (With a suitable change in the mathematical assumption (A. by their physical values hE|φ in i and hE|ψ out i. 1. Harshman / Nuclear Physics B 581 (2000) 91–115 109 We start with the S-matrix element (30) and consider the case that there are two resonances in the j th partial wave — although these results are also proven for a countably infinite number of resonances (with qualifications) under the standard mathematical assumptions used for the S-matrix or potential V [55]. N. We restrict our analysis to only one partial wave. thinking of s as the energy E as we have done in Appendix A.L.h. According to our assumption. The r. of (B.h. of (30)..2): φ + describes the prepared (ee¯) in-state and ψ − describes the observed (f f¯ ) ‘out-state’ registered by the detector.

We now use the expansion around the pole zRi (i) S(ω) = s−1 ω − zRi + s0 + s1 (ω − zRi ) + · · · (B. the cut lies along the positive real axis.8) (B. N. This contour in (b) can be deformed as in (c) and is equivalent to a sum of contours Cbg . (B.6) for each of the two (or countably many) integrals separately.R. they are given by I − + (B. This integral can equivalently be taken −i below the cut in the second sheet (b). 1. Harshman / Nuclear Physics B 581 (2000) 91–115 Fig.7) (ψ . In this figure. The integrals around the poles we call the pole terms.2)).4) C2 The first term has nothing to do with the resonances and we call it the background term Z−∞ bg dE hψ − |E − iSII (E)h+ E|φ + i ≡ hψ − |φ bg i. I + dω hψ − |ω− iSII (ω)h+ ω|φ + i. We begin with (a) the integral is taken i above the cut in the first sheet.5) 0 it is the same as (32). φ )poletermi = dω hψ − |ω− iSII (ω)h+ ω|φ + i Ci I = Ci (i) dω hψ − |ω− i s−1 ω − zRi h+ ω|φ + i (i) hψ − |zRi − ih+ zRi |φ + i = −2πis−1 (B. Bohm.110 A. C1 and C2 since the integral along infinite semicircle in the lower half plane vanishes (because of the Hardy class properties of (B. (B.9) .L.

R.10) It is worthwhile to explain what is used in each step of this calculation. Eq. Bohm. the Cauchy theorem has been applied. It does not matter whether we take the second part of the integral over the physical values of E. More details are given in earlier publications. The major contribution to the integral comes from the physical values 0 6 E < ∞ if zRi is not too far from the real axis. where also the case of higher order S-matrix poles is considered.12)) and (B. using (B. see for example the review [27]. φ + ) in two different ways.9) and (B.10).12) This is equivalent to the definition (31a). along the lower edge of the real axis in the second sheet or in the first sheet immediately above the real axis. To get from (B.A. Section 5. it is an idealized or exact BW whose domain extends to −∞II in the second (unphysical) sheet. to get from (B. E − (ERi − iΓRi /2) h+ zRi |φ + i (B.10) in (B.9).8) to (B. But unlike the conventional BW which (if one worries about these h+ E|φ + i ∈ H− mathematical details) one takes only 0 6 E < +∞. The integral in (B. We shall now write the j th partial S-matrix element (ψ − .10) is a Breit–Wigner amplitude (i) s−1 E − zRi (i) = s−1 E − (ERi − iΓRi /2) . (B. with − ∞II < E < +∞I (B.10) is the Titchmarsh theorem for Hardy class functions (cf. (B. The equality of (B.11) ∗ integrated over along with the smooth and well-behaved apparatus functions h− E|ψ − i 2 . It is this BW that we mean by the “S-matrix Breit–Wigner” (11). This does not violate the condition of semi-boundedness of the Hamiltonian operator H .8) to (B. N. the contour Ci of each integral separately has been deformed into the integral along the real axis (second sheet) from −∞II < E < +∞ and along the infinite semicircle (which vanishes because of the Hardy class property). From (B.4) we obtain: Z∞ 0 dEhψ − |Ej − iSj (E)h+ Ej |φ + i .9) and (B. φ + ) = hψ − |φ bg i + i On the other hand.L. Each of the BW integrals in (B. Harshman / Nuclear Physics B 581 (2000) 91–115 Z+∞ = 111 (i) − − + + dE hψ |E ih E|φ i −∞II s−1 E − zRi . footnote 4).13) (ψ − . (B.10) one obtains the representation − hψ |zRi − i i= 2π Z+∞ dE −∞II + + hψ − |E − i (i) h E|φ i s−1 .1) and (B. using (B.10) extends from E = −∞II in the second sheet along the real axis to E = 0 and then from E = 0 to E = +∞ in either sheet.9) (or (B. On the one hand.5) in (B.4) we obtain: X hψ − |zRi − i(2π/i)h+ zRi |φ + i. 0 6 E < ∞.10) define a Gamow ket |zRi − i corresponding the exact BW.

14) According to the van Winter theorem [56].s. of (B. (B.15) has a corresponding term in (B. where cRi = (2π/i)h+ zRi |φ + i. (B.s.15) carry the same information and each term in (B.17) expresses the in-state φ + in terms of scattering states described by the eigenkets |E + i of the Hamiltonian H with real generalized eigenvalue E. (B.16) φ + = φ bg + i The in-state φ + can be decomposed into a vector representing the non-resonant part φ bg and a sum over the resonance states described by the Gamow kets.h. we cannot be certain about the energy dependence of the background bj (E). However. Omitting the arbitrary ψ − ∈ Φ+ (the observable) one writes this as a functional equation in the space Φ × + and obtains the complex basis vector expansion (25) of the prepared in-state φ + ∈ Φ− : X |zRi − icRi . The first term on the r.2 of [27]. + Z∞ φ = dE |E + ih+ E|φ + i. Appendix A. on the whole half plane) is uniquely determined by its values on the positive real axis. Therefore we can use the Mellin transform to rewrite the first integral on the r.16) is a generalized eigenvector expansion like Dirac’s eigenvector expansion.14) and obtain [57] Z∞ dE hψ − |Ej − iSj (E)h+ Ej |φ + i 0 Z∞ − − + + dE hψ |Ej ibj (E)h Ej |φ i + = +∞ X Z i −∞II 0 − − + + dE hψ |Ej ih Ej |φ i (i) s−1 E − zRi .13) and (B.17) 0 While (B.h. as long as we are only interested in the two resonances in some energy range under investigation. cf. Bohm. The equations (B. If there are no further poles or other singularities besides those included in the sum. N.13). Harshman / Nuclear Physics B 581 (2000) 91–115 Z0 =− dE hψ − |Ej − iSj (E)h+ Ej |φ + i bg −∞ +∞ X Z dE hψ − |Ej − ih+ Ej |φ + i + i −∞II (i) s−1 E − zRi . by the Titchmarsh theorem.L. represents the non-resonant background and the terms in the sum represent the resonant terms. We first discuss (B. The complex spectral resolution (B. then bj (E) is likely to be a slowly varying function of the energy. (B.16) is an . a Hardy class function on the negative real axis (and therefore. we can consider bj (E) as an irrelevant background.15) Without more specific information about the S-matrix.R. (B.112 A.13).

.16) has a corresponding term in (B. Thus we have the correspondence:     Gamow vector |j. brest− i of Breit–Wigner amplitude ajR (s) of (11). but in many cases it may not be relevant for the physical problem and can be ignored.18) for two interfering resonances is another argument in favor of the new hypotheses (A.2) for quantum theory.18) is the full equation (B. (B. Expressions like (B. For an alternative interpretation as a second order Gamow– Jordan state see [60]. We now discuss (B. hψ − |E − ih+ E|φ + i = hψ out |EihE|φ in i ∈ S ∩ H− which describe the resolution of the preparation apparatus and of the registration apparatus 2: and write Eq.15). The representation (15) of the j th partial amplitude is the analogue of (B.18) Each term in (B. according to (B. Note that there must always be such a background term B(s) or φ bg phenomenologically and theoretically. 8 See also references in Chapter 20. This phenomenological success of formulas like (15) or (B.18) are not a consequence of the usual assumptions of analyticity but follow from the new hypothesis of the RHS formulation of quantum mechanics stated in Appendix A.16) corresponds in (B.18) and must therefore be understood as a convolution integral using the apparatus wave functions with the property (B. The vector φ bg describing the non-resonant part of the state φ + (which was prepared in the distant past as the non-interacting prepared in-state φ in ) corresponds to the background part of the Smatrix element S bg (or in the amplitude (15) to the background amplitude B(s) = (bj (E) − 1)/2ip). This equality holds for the whole space of functions 2 .18) has been tested experimentally for the non-relativistic case in nuclear physics [58.2 of Ref. N. (B. Therefore we can omit these arbitrary energy wave functions hψ − |E − ih+ E|φ + i ∈ S ∩H− 2 . Bohm. The interference of resonances as predicted by (B.1) and (A. To the superposition of resonance state vectors described by the Gamow kets |zRi − i in (B.15).18).2).18) the “superposition” of resonance amplitudes described by the exact (with −∞II < E < +∞) Breit–Wigner amplitudes.59] 8 and for the relativistic case in the ρ–ω interference which arises as a consequence of isospin symmetry breaking [41]. (37) with exact exponential decay ⇔ where s extends over −∞II < s < +∞     with lifetime τ = h/Γ with width ΓR = h/τ ¯ R ¯ (B. sr . [36].R.A.L.16) and (B. Harshman / Nuclear Physics B 581 (2000) 91–115 113 expansion in terms of eigenkets |zRi − i of the same self-adjoint Hamiltonian H with complex generalized eigenvalue zRi = ERi − iΓRi /2.18).15) as an equation between distributions over the function space S ∩ H− Sj (E) = bj (E) + X i (i) s−1 E − zRi .19) The precise meaning of the superposition (B.

Phys. Cambridge University Press. H..A. Gadella. Rev. Phys. Kaldass. [34] M. Phys. C 62 (1994) 35. B 203 (1988) 177. Sov. Sirlin. A. Weisskopf. 1345. September 1992. Perseus. pp. Vol. Comm. 1353–1354. Sirlin. [26] V. Kielanowski. Lett. [9] S. 65 (1930) 18. [18] M. 1989. Sachwitz. [22] N. Bardin. 1.W. Cambridge. A 60 (1999) 861. O. [12] F. Z.E. Springer. 1977. Valencia.D. Caso et al. Bohm. Z. Springer. A 264 (2000) 425. Rev. 63. Cambridge. [32] R. M. Riemann.J. C 3 (1998) 1. Harshman / Nuclear Physics B 581 (2000) 91–115 References [1] M. (Eds. MA. Ellis. Phys.. Rep.114 A. [29] M. T. Adriani et al. Nuovo Cimento 13 (1959) 115. Fitch et al. Phys. 322 (1999) 125. B 267 (1991) 240. Z. Haag. Report No. A. van Neerven. Lett. B 259 (1991) 373. Schroeder. [24] M. 11 (1968) 522. Kielanowski (Eds.N. 1995. 236. Levy. Sirlin. Watson. Riemann. Lett.L. R. B 206 (1988) 539.G. Willenbrock. [5] A. Hollik. Harwood Academic. 2000. Lett. Phys. Rep. and references therein. Wiley. E.C. Lectures at the Max-Born-Symposium on “Quantum Future”. Duckeck. [17] H. [19] K. Electroweak precision tests at LEP1. Gell-Mann. D. in: C. Lett. Lett. B 315 (1993) 494. Collision Theory. p. p. Physical Origins of Time Asymmetry. Riemann. Phys. Physica A 236 (1997) 485. B 140 (1965) 1088. Phys. W. P. Math.P. Chapter 13. 67 (1991) 2127. University of California at Santa Barbara. New York. Irreversibility and Causality. Bohm.L. World Scientific. Polkinghorne. [21] V. Rev. [20] K. Veltman. [30] W. L3 note #1233. Phys. N. . B 272 (1991) 353. [27] A. Particle Physics and Introduction to Field Theory. 1986. [3] T. Nikolaev. P. P. Loewe. S. Phys. New York. Maxson. Rev. Goldberger. H. in: J. W. Vol. H. Harshman. in: J. Leike. [15] R.M. 41. Bohm. in: Springer Tracts in Modern Physics 162. Wickramasekara.D. Kleinknecht. Hartle. Phys. Cohen-Tannoudji et al. Phys. [28] A. Lett..V. UCSBTH-95-28 (1995). [33] M. An Introduction to Quantum Field Theory.).L. p. Wiley.D. 1998. Jarlskog (Ed. gr-qc/9509054. Phys. G. II. LANL Archives. Eur. Phys. [23] E.. D 58 (1998) 113010. Stuart. [13] S.L. p. Landshoff. Phys. Lett. unpublished. Bohm. [10] LEP Collaborations. D 55 (1997) 6625. K. Physics at LEP. Bohm. 132 (1990) 245. [4] R. S. Kielanowski (Eds. G.L. 1981. CP Violation. 1998. Peccei (Eds. Phys. Norman. [6] C. Grünwald. Przieka. report CERN 86-02. [7] T. 60 (1988) 2246. J. Berlin. 152 (1958) 546. Lett. Höhler. 1994. Gibbons et al. Doebner. Phys. A Combined Fit to the L3 Data Using the S-Matrix Approach (First Results). [31] A.). Usp.). Phys.. Lett. 157. CERN-preprint 99-15. Hollik. in: A. p.J. S. Irreversibility and Causality.R. Chapter 8. in: A. LEP electroweak working group and the SLD heavy flavour and electroweak groups. [2] C. Rev. Phys. New York. Cambridge University. N. [14] S.J. 63 (1930) 54. Z. The Analytic S-Matrix. Halliwell et al. and references therein.). Olive. 1997. in: Quantum Mechanics. G. Phys. 181.V. J. Reading. Consoli. and references therein. Wigner. Lee. Eden. A. [25] G. Doebner. [11] M. Berends. Kirsch. [16] A. Peskin. B 262 (1991) 113. Rev. 1964. M. [8] The L3 Collaboration. J. New York. Bohm.). Bohm. P.B.B. Passera. Springer. Phys. M. Berlin. and references therein. and references therein. 1996. 236 (1993) 1.D. Burgers. P. Phys. Phys.

Ostrovsky. Newton. Springer. Vol. Brändas (Ed. (Eds. Math. 47 (1974) 633. Bohm. J.2. Weinberg.P. Lett.E. J. [50] V. Browne. hep-th/9912026. [45] E. 1982. Kluwer Academic. Barcelona. VIII.D. Krasnopolsky. Leike. D 45 (11) (1992) VIII. Riemann.10. 162 (1971) 103. LEP electroweak working group and the SLD heavy flavour and electroweak groups. Am. Pestieau. Anal. Reviews of Particle Physics (1994) p. van Winter. Springer. Rev. B 326 (1994) 1. Theor. Lect. Rev. Math.. Rev. V. Harshman / Nuclear Physics B 581 (2000) 91–115 115 [35] A. Horacek.12. [52] M. Stuart. Int. N. Math. [48] C. of the Fourth International Symposium on Radiative Corrections (RADCOR 98. Phys.. App. G. Phys. Weinberg. [46] S. 8–12 September. Bernicha. C. B 273 (1991) 513. Phys. and references therein. Phys. Lett.. Notes in Phys. Vol. Yang. Plenum Press. 1995. Rev. in: E.R. Math. p. [56] C. in: E. New York. Nakamura. J.P. Martin. Scattering Theory of Waves and Particles. D 50 (1994) 4454. López Castro. Trans. [55] M. Z 0 -Physics. [36] A. 41 (1997) 279. R. Mondragon. The Quantum Theory of Fields. 325. Chapter 3. Bohm. Vol. Review of Particle Properties. [39] G.2.. Phys. [42] A. Sola (Ed. Eur. 40 (2) (1939) 149. Pestieau. Nucl. [43] R. [44] T.). 1. Swartz. C. T. private communication. 201. [57] M. C 3 (1998) 624. Phys.G. Spain. Berlin. Hinterberger et al. [38] A. J. [40] R. [60] E. A. p. Gadella. D 56 (1997) 1515. H. [37] S. 2 (1970) 1. 139. 325.I. Brändas (Ed. Bernicha. Quantum Mechanics-Foundations and Applications. J. Lee. 1998. V.). [58] F. Rev. A 597 (1996) 623. Rose. [47] L. 1989. Phys. Phys. 106 (1957) 340.N. Phys. Proc. Mahaux. Berlin.N. Rev. Phys. G. London. Soc. Wigner. [54] R. 36 (1997) 2271.A. Höhler. [59] W. Phys. 1989. in: M. Cutkosky. Ann. Gadella. CERN-preprint 99-15. Höhler.).I.L. Ferreira. Press. Section 7. LANL archives. A 299 (1978) 397.G. Levy et al. Cambridge Univ. 3rd edn.L.. 2nd edn. 58 (1998) 2077. Oehme.S. 483. Lett. J. Cambridge. p. López Castro. Berlin.N. Phys. Lect. Phys. 1991. J. Springer. Sirlin. in: J. G.D. Notes in Phys. Springer. Nucl. . Hernandez. [49] O.). Theory of Resonances. [51] A. Kukulin. 1989. Callendar. Totstkhin. [53] LEP Collaborations. [41] A. 1994.

1.1) The function a 2 (φ) (which represents the running string tension) must be of a special type.V. Princeton University.princeton.8 .edu 2 rytchkov@math.V.Nuclear Physics B 581 (2000) 116–134 www. Rychkov b.elsevier.3 2 1 3 ( 0 0 ) 0 0 1 8 3 . In this duality color–electric flux lines of gauge theory are described as certain relativistic strings. ds 2 = dφ 2 + a 2 (φ) dE (1. Incomplete — since we get these results only in the WKB approximation and only for a special class of contours. accepted 16 March 2000 Abstract We explore gauge fields–strings duality by means of the loop equations and the zigzag symmetry. Princeton University. All rights reserved.edu 0550-3213/00/$ – see front matter  2000 Elsevier Science B. The gauge fields–strings duality was formulated in [1] as an isomorphism between the closed string vertex operators and the gauge invariant operators of gauge theory. it must have a horizon. Princeton.  2000 Elsevier Science B. 1 polyakov@viper.princeton. PII: S 0 5 5 0 .1 . It underwent a rapid and fascinating development in the last three years. In order to properly describe the zigzag-invariant Wilson loop. Princeton. where a 2 (φH ) = 0.2 a Joseph Henry Laboratories. and an infinity where a 2 (φI ) = ∞. The ways to go beyond these limitations and in particular the OPE for operators defined on the loop are also discussed. V. Striking — because we find that the string ansatz proposed in [1] satisfies gauge theory Schwinger–Dyson equations precisely at the critical dimension Dcr = 4. Introduction Gauge fields–strings duality is an old and fundamental subject. Polyakov a. USA b Department of Mathematics. It has been shown in [1] that the “natural habitat” for these strings (and thus for the flux lines) is a five-dimensional curved space with the metric x 2. NJ 08544.nl/locate/npe Gauge fields–strings duality and the loop equation A. The precise form of a 2 (φ) is determined from the condition of conformal invariance on the world sheet. NJ 08544 USA Received 14 February 2000. All rights reserved. The results are striking and incomplete.

which can be used for the renormalized Wilson loop. There is no real understanding (beyond the heuristic arguments of [1–5]) why and. We will give its derivation now. The Wilson loop is given by   I 1 TrP exp Aµ dx µ . It is generally thought that the loop operator is singular and can be applied only to the regularized and non-universal Wilson loop. In the present paper we shall apply the loop operator to the string functional integral. more importantly. (2. In these cases the Schwinger–Dyson equations (formulated as loop equations) is our only tool. In [6] it has already been sketched how to implement the loop equation for the renormalized loops. This would clarify which string theories must be used for the various gauge theories.1) W [C] = N C The averaging in this formula is performed with the Yang–Mills action . It has been shown there that some of the SYM correlation functions (describing absorption of the external particles by the 3-brane) can be calculated by the use of the classical supergravity in which the 3-brane is replaced by the metric (1. when the correspondence works. at the critical dimension D = Dcr = 4 the loop equation is satisfied.A.5] it was shown how to implement the above-mentioned isomorphism in the N = 4 SYM theory. The reason. Ideally. In the subsequent papers [4. It will contain some additional elements to the old works. This we find quite stunning. lies in some common misconceptions concerning loop equations. We will find in this case that. But all is not well. we believe. This is not necessarily so. Rychkov / Nuclear Physics B 581 (2000) 116–134 117 Another development [2] was related to the N = 4 SYM theory and the 3-branes in the type IIB strings. It was further conjectured [3] that the correspondence extends beyond the SUGRA approximation and the function a 2 (φ) in this case must correspond to the AdS space (a 2 (φ) ∝ eαφ ). the loop operator is well defined in any dimension (implying the zigzag symmetry of the string representation). Polyakov. There are reasons to believe that in the most interesting non-supersymmetric cases the D-brane interpretation of the general σ -model metric (1. The results are incomplete but quite stunning.10]). but no conclusive results have been reached so far. one would like to check that the Schwinger–Dyson equations of the Yang–Mills theory can be obtained from the string representation. It is important to stress in this respect that the help of the D-branes in answering the above question is not always available. V. In the vast literature on the subject there have been occasional attempts to explore the loop equations. After that the correspondence in this case has been confirmed in an almost infinite number of papers.1) is not possible. 2.1). The loop equation and the zigzag symmetry The loop equation has been introduced in [7] and further explored in [6] and [8] (see also some related developments in [9. We will be dealing only with the very special contours — wavy lines (suggested in [1]) and only in the WKB approximation. first. and second. This is obviously incomplete.

The second term is regular in this case. Rychkov / Nuclear Physics B 581 (2000) 116–134 S= 1 2 4gYM Z 2 (dx) TrFµν (Fµν is the Yang–Mills field strength). . sN ) = Tr P On1 x(s1 ) On2 x(s2 ) · · · e C Aµ dxµ . Let us assume that the contour C is smooth and non-selfintersecting. δ x(s) − x(s 0 ) = lim (4 − ∆) ∆→4 |x(s) − x(s 0 )|∆ 3 Such operator products were considered before in [16]. Consider a set of operators {On (x)} which are not color singlets. We cannot be sure that on the string side the regularization is the same and thus the comparison of gauge fields and strings becomes strictly speaking impossible.3) ∆n1 +∆n2 −∆m m |x(s1 ) − x(s2 )| where s = (s1 + s2 )/2 and ∆n are anomalous dimensions of the corresponding operators. It has already been noticed in [6] that the way out of this difficult is to use the operator product expansion as s → s 0 . while the first one contains the δ-function which is easy to pick up. Consider as an example a singularity  1 . In this way the OPE in the physical 4d space is transplanted to the one-dimensional contour C. Of course. (2. sN ) have power singularities at the coinciding points. There is no room for the contact terms proportional to δ(s − s 0 ) if we use renormalized correlators. We can in general examine a gauge invariant amplitude D  E H   Gn1 ···nN (s1 . the possible contact terms in the x-space do not give contact terms in the s-space. . . which we do not display. In this case we expect the OPE to have the form 3 Cnm1 n2 (x(s))   X  On1 x(s1 ) On2 x(s2 ) = O x(s) . .118 A. Polyakov. The problem to overcome is the singularity of the second term at s = s 0 . The price to pay is the non-universal arbitrary regularization. V. The idea of the loop equation is to find an operation in the loop space which. consider the second variational derivative of W [C] D   H A dx E δ2W = Tr P ∇ F x(s) e µ µ x˙ν (s)δ(s − s 0 ) µ µν δxµ (s)δxµ (s 0 ) E D   H  (2. Moreover. which makes it difficult to distinguish it from the first one. . . This can be achieved in several different ways. To implement this idea. . . The “brute force” method would be to regularize the gauge theory with some cut-off Λ and the consider |s − s 0 |  1/Λ.2) + Tr P Fµλ x(s) Fµσ x(s 0 ) e Aµ dxµ x˙λ (s)x˙σ (s 0 ).1). and to introduce the “loop Laplacian” which acting on W [C] gives the Yang–Mills equation of motion. being applied to the LHS of (2. The “open string” amplitudes Gn1 ···nN (s1 . in asymptotically free theories there are also powers of logarithms log |x(s1 ) − x(s2 )| in these formulas. The main idea of the loop equation is to separate the first term. where P is the ordering along the contour. will give the Yang–Mills equations of motion at the RHS. .

An exception from this is the operator ∇α Fβγ which can appear in the OPE:  [x(s) − x(s 0 )]µ ∇µ Fλσ  . The loop Laplacian is defined as the coefficient L(s)W in front of the δ-function. {x(s)})  δ2W = b L(s)W δ(s − s 0 ) + . the singular contribution from this term to (2.4) b L(s)W = 0. even when protected by non-renormalization theorem. Z Z  f (s 0 ) ds 0 = 0. However.) However. Fµλ x(s) Fµσ x(s 0 ) ∼ |x(s) − x(s 0 )|2 (We assume here the scenario in which the Fµν operator has the normal dimension as a result of some non-renormalization theorem. Polyakov. . q + p)δxµ ( 2 − p) n (2. Some danger for this analysis would be presented by odd integers ∆n in (2. The above analysis can be extended to the functionals represented by the iterated integrals Z X  Γµ(n) x(s1 ).4) in the momentum space: lim p→∞ δxµ ( q 2 X  δ2 W bq W p0 + = L Cn (q)|p|λn . which would give a pk (log p + const) contribution. {x(s)}) ∝ hTrP (On (x(s)) e A dx )i. Rychkov / Nuclear Physics B 581 (2000) 116–134 119 As it was explained in [6].7) . 0 δxµ (s)δxµ (s ) |x(s) − x(s 0 )|4−∆n n (2.6) does not contain integer even powers of p. the lowest operators which can appear in (2. when one uses test functions. It shows that the Wilson loop belongs to a very special class of functionals.2) vanishes after contraction with x˙λ (s)x˙σ (s 0 ). .4). (2. . for which the asymptotics in (2. . corresponding to the derivatives of the δ-function. V. the nonlocal power-like behavior of the second term for s → s 0 can be separated from the local b s = s 0 singularity of the first one. Since |x(s) − x(s 0 )| ≈ x˙ 2 (s) |s − s 0 |. As a result of this discussion we conclude that the second variational derivative of the renormalized (and analytically regularized) Wilson loop has the form X Cn (s.3).A. have even dimensions. so that the only singularity in the integral comes from s = s 0 . If it does.6) This is an important formula. It is also assumed that the integral is defined by analytic continuation. and the loop equation for non-selfintersecting contours has the form In our applications it will be convenient to write (2.5) H p where Cn (s.1). δ x(s) − x(s 0 ) f (s 0 ) ds 0 = lim (4 − ∆) ∆→4 |x(s) − x(s 0 )|∆ provided that the contour has no self-intersections. x(sn ) W [C] = 1 ···µn n s <s <···<s n 1 2 × x˙ µ1 (s1 ) · · · x˙ µn (sn ) ds1 · · · dsn (2. this means that the functional W [C] cannot be represented as an ordered exponential (2.

g. there are the closed strings amplitudes.6) and check for the terms ∝ pn with n — positive even integer. Z x¨α x˙β − x¨β x˙α . together with the formula (2. if it satisfies some extra analyticity requirements. Such functionals behave well under the application of the loop Laplacian (see [8]). Rychkov / Nuclear Physics B 581 (2000) 116–134 (of which (2. we will use the term “zigzag symmetry” referring to the form (2. If they are present. e. although a possibility is that the functional must be nonsingular with respect to x˙ 2 . It is likely that any zigzag-invariant functional can be represented in the form (2. y).2) ds 2 = ρ(y) dy 2 + dE As usual. (3. e. According to [1] we must consider a 5d background with the metric (1. Z    Vµν = d2 ξ ψpE y(ξ ) ∂a x µ ∂a x ν eipE xE(ξ ) . When we change s ⇒ α(s).7).1) is a special case). We are also considering a fermionic noncritical string. and the functional does not represent any Yang–Mills theory.7) is the zigzag symmetry [1].g. It is important to have the NSR fermions for the gauge fields–strings duality. Without them many other forms are possible. Iαβ = ds x˙ 2 At present the precise form of these requirements is not known. there is a practical way for testing the zigzag symmetry in this sense: take the asymptotics (2. 3.1):  x2 . Polyakov. The σ -model action of this string is given by Z  1 √ d2 ξ g g ab (ξ )GMN z(ξ ) ∂a zM ∂b zN S= 2 (2)  √ + Φ z(ξ ) R(g) g + εab BMN z(ξ ) ∂a zM ∂b zN + other background fields + fermions.. Here we will just remind the reader the underlying logic with a few additional comments. It is also necessary to perform the non-chiral GSO-projection in the functional integral. The string representation: D-branes and/or σ -model Let us summarize the facts about string representation of the Wilson loop. According to the above discussion. (3. In the following. The origin of the fifth dimension is the Liouville degree of freedom which is unavoidable in the non-critical strings.1) We introduced the 5d variable zM = (x µ . There are two types of objects we would like to consider..7). they are invariant even if α 0 (s) changes sign (and is not a diffeomorphism).6).6). there is no place for the zigzag symmetry. The metric GMN has the form obtained by the obvious change of variables in (1. Another feature of the functionals (2. The whole matter has been treated in the paper [12].120 A. Hence. V. This conclusion. We introduce the closed string vertex operators. the background fields are determined from the β-function equation [11]. expressing the g ab -independence of the theory. the second variational derivative for them has the structure (2. First. will be extensively used below.1).

This means that if we find the right action (3. Polyakov. we have to determine y|∂D and the background fields so that the loop equations are satisfied. Indeed. In other words. we create a flux line by acting on the vacuum with the operator I Tr P exp A dx ≈ I + Cµνλσ Tr(Fµν Fλσ ) + · · · . so far there is no real derivation of the isomorphism (3. 3-branes can be presumably replaced by the supergravity background they generate.3) from the first principles.y(ξ )]. represented by small Wilson loops. is to start with the 3-branes of the Type IIB superstring. but not under the zigzag transformations. It was stated there that in order to . These are the only positions where we can place the contour.A. on the one hand. V.3). Also. so that ρ(yH ) = 0 and ρ(yI ) = ∞.1). The basic object in this approach is the Wilson loop. we conjecture an isomorphism Closed string states ⇐⇒ Gauge invariant operators. Another approach [1] is to try to formulate the conditions under which the σ -model satisfies the Schwinger–Dyson equations of the Yang–Mills theory. The main problem with this approach is that it is far from being clear that nonsupersymmetric theories can be described in the D-brane language. it is known that the low energy excitations of the stack of N three-branes are described by the supersymmetric Yang–Mills theory with the SU(N) gauge group. On the other hand. At the boundary of the disc we fixed xE by the condition xE|∂D = xE (s). There are two approaches to the problem. Then. (3. and thus one expects and checks the isomorphism (3. applicable to the supersymmetric Yang–Mills theories. Indeed. we will encounter the lack of the zigzag symmetry coming from the fact that the √ first term in (120) is (due to the presence of g ) invariant only under diffeomorphisms. defined by Z Dy(ξ ) DxE (ξ ) e−S[Ex (ξ ). although numerous checks have been performed. even in the supersymmetric case.3) 2 can be The intuitive reason for this conjecture is that gauge invariant operators like Tr Fµν associated with tiny closed flux lines. But how to discover the right action? This question does not have a complete answer at present. C→0 C The propagation of these loops forms the closed world surfaces with punctures represented by the vertex operators. Hence it must be either zero or infinite at the boundary. A more precise condition was given in [12]. The basic idea of [1] was to notice that the metric ρ(y) must have horizon and/or infinity points. Now. the closed string S-matrix will be equal to the correlation functions of the local gauge invariant operators.3]. One approach [2. Rychkov / Nuclear Physics B 581 (2000) 116–134 121 and claim that for the proper choice of the metric ρ(y) the correlators of V ’s on the string side are equal to the correlators of the gauge invariant operators on the Yang–Mills side. W [E x (s)] = xE|∂D =E x (s) y|∂D =? Here we take the world sheet with a disc topology (appropriate for large N ). if we choose some other y|∂D .

since we need NSR fermions to exclude the boundary tachion (violating the zigzag symmetry).4) ρ(y) = gYM y2 It has been shown in [4. conformal cases it is possible to retreat to the quasiclassical domain by taking gYM In this case the string action takes the form: Z    1q 2 d2 ξ  (∂a xE)2 + (∂a y)2 + O(1). Physically this means that open strings in this theory correspond to vector gluons. The concrete types and forms of background fields depend on the type of the gauge theory we are working with. In this paper we shall adopt this prescription. and they form a condensate in order to stabilize the AdS space. however. Of course to check the complete loop equations we have to solve the above problem first.122 A. In a sense. One should keep in mind. It is therefore natural that the Wilson loop C on the D-brane can be represented as a Wilson loop at the horizon [1]. the Wilson loop is given by [13. form a closed algebra.14] W [C] ∝ e q 2 NA − gYM min [C] .6) . But it is important that the condition be formulated without any reference to D-branes. We arrive at the conclusion that in the WKB limit.5] for the local operators and in [13. gYM N (3. In the latter case this point represents a horizon behind which the D-branes are hidden. and the boundary conditions y|∂D such that the vector vertex operators. For example. We hope that the two types of boundary conditions yield the same result and will be studying the case y = 0. At present we do not know how to determine this field. defined at the boundary. The problem of finding the precise background in each particular case is not yet completely solved. the other background fields. This horizon is physically important because the absorption by the D-branes can be accounted for as the disappearance of the closed strings behind the horizon [2].5) S xE (ξ ). V. although there is no real proof for it. however.14] for the loops that the natural location for the gauge theory quantities is the infinity in this AdS space. our check of the loop equations in this paper can be considered as such (incomplete) proof. the loop generates the Bµν -field with Bµν = Bµν (x. Polyakov. in the 2 N  1. The string representation involves various background fields. we have RR fields in our closed string spectrum. Also. However. When the D-brane description is applicable. {C}) [1]. This point has no geometrical significance in the Euclidean signature (the Lobachevsky space) but is meaningful in the Lorentz signature (that is in the AdS space). the vector vertex operators are selected by the zigzag symmetry. the above condition is satisfied. In the case of the N = 4 SYM the D-branes and conformal symmetry considerations lead to the formula [3] q 2 N 1 . (3. In this case. y(ξ ) = 2 2 y (ξ ) where the terms O(1) contain NSR fermions and RR fields. that there is another zigzag-symmetric location. y = ∞. that is y → 0. (3. Rychkov / Nuclear Physics B 581 (2000) 116–134 describe gauge theory we must find the metric ρ(y). since they are the only ones defined without the world sheet metric.

3)   y = a(σ )τ + 1 b(σ )τ 3 + · · · 3 After substituting this expansion into (4. θ⊥k = 2 (E a . y a τ a and due to (4.2) we obtain   2 dE c   .7) have the form:    1    ∂a 2 ∂a xE = 0.7) 2 2 y is the minimal area of a surface bounded by the loop C. Polyakov. dσ dσ (∂σ cE)2 This guarantees that the leading term in the energy–momentum tensor (4. y We also have to impose the Virasoro constraints ( (∂1 xE)2 + (∂1 y)2 = (∂2 xE)2 + (∂2 y)2 . 0) = cE(σ ).4)  2    ∂σ cE dE c d   fE(σ ) = . Amin [C] = min 4. In the following sections we will analyze the action of the loop operator on this functional. There are some general features of such minimal areas which we discuss in this section.  a 2(σ ) = dσ (4. V. The equations of motion for the action (3.4) the first term vanishes. Rychkov / Nuclear Physics B 581 (2000) 116–134 123 Here Z  1 d2 ξ  (∂a xE)2 + (∂a y)2 (3.2) We are looking for the solutions satisfying the conditions ( xE (σ. (We renamed the variables: ξ 1 = σ .) It is easy to see that the expansion in τ has the form:  1 1   xE = cE(σ ) + fE(σ )τ 2 + gE(σ )τ 3 + · · · 2 3 (4. we have    1 1  1 θ⊥k = 2 ∂τ xE ∂σ xE + ∂τ y∂σ y = 2 fEcE0 + aa 0 + 2 gEcE0 + · · · .2) vanishes.1) and (4. y (4. ξ 2 = τ .A. (4.1)  1  2   ∂ y = (∂a y)2 − (∂a xE )2 . while the second one gives 1 g cE0 ). ∂1 xE∂2 xE + ∂1 y∂2 y = 0. Minimal area in the Lobachevsky space Our contour is located at infinity (the absolute) of the Lobachevsky space. 0) = 0. y(σ.

Indeed. . Namely. and will develop a method of successive approximations for A[C]. δy(s) δ xE (s) y (s) ds ds GMN (z) (5. Because of this difficulty we will choose an alternative way of finding A[C]. Unfortunately. θ⊥⊥ = 5. Rychkov / Nuclear Physics B 581 (2000) 116–134 Analogously: 2 2 2 2  1  ∂τ xE + ∂τ y − ∂σ xE − ∂σ y 2 2y 1 2 (4. To perform this expansion it is convenient to consider the standard Hamilton–Jacobi equation for the minimal surface which has the well-known general form: φ i (s) δA δA dzM dzN = G . MN δzM (s) δzN (s) ds ds For the Poincaré metric this equation becomes     2  2   δA 2 1 dE x dy δA 2 + = 4 + . V.1) . The above would be the case is we were able to find the coefficients gE and b in terms of cE. it is easy to see that gE(σ ) δA = .6) Amin [C] = ε where we introduced the cut-off ymin = a(σ )τmin = ε.1) reveal that the functions gE(σ ) and b(σ ) in (4.6). Further iterations of (4.124 A. δ cE(σ ) a 2 (σ ) and if b were known. There is an unpleasant surprise. i = 2. Polyakov. E are small. D. . x i (s) = φ i (s). we will consider a special type of contours — wavy lines.5). The theory of wavy lines It was already suggested in [1] that it is instructive to consider wavy lines instead of general contours.7) calculated on the classical solution (4. the equation for A would follow from (4. however. We will also see another derivation of (4. .5) = 2 fE 2 + 2ab − cE0 fE0 − a 0 .3) are not determined by the small τ -expansion and can be kept arbitrary. They are fixed by the global condition for the absence of singularities at finite τ .3) has the following structure (encountered previously in [13] in a special case):   L[C] + A cE(σ ) . (4. Our main interest is to derive variational equations for A[C]. One might think that the usual Hamilton–Jacobi equations following from the conditions θ⊥k = θ⊥⊥ = 0 will give us a closed equation for A. a The action (3. namely to look at a curve x 1 (s) = s. Below we shall find the expansion of A[φ(s)] up to the and to assume that fourth order and then explore the action of the loop Laplacian. . In this formula L[C] is the length of the contour C and A[E c(σ )] is a finite functional. and thus it is hard to find them explicitly. life is not so simple.

We can do this by solving (5.A.1) with respect to δA/δy and by noticing that . V. Rychkov / Nuclear Physics B 581 (2000) 116–134 125 We want to explore the limit y(s) = y → 0. Polyakov.

Z δA .

.

2) ∂y δy(s) . (5. ∂A = ds .

∂y y ds δ xE(s) (5. A= n! n Expansion of (5.1) and (5.3) that A(y) behaves like A(y) ≈ y→0 L[C] + O(1).2) s   2  Z 1 dE x δA 2 ∂A 4 = − 2 ds −y . .3) We see directly from (5. sn |y) φi1 (s1 ) · · · φin (sn ) . .3) together with the reparametrization invariance equation . . y(s)=y We have from (5. y To get further information we have to look at the following expansion of A: X 1 Z  ds1 · · · dsn Γi1 ···in (s1 . .

δA .

.

dφE δA dxµ δA =0 = + .

8) . . .3):  Z  L0 1 1 ∂A =− 2 + y 2 πE 2 − 2 (φE 0 )2 ds ∂y y 2 y Z   2  2 1 4 2 1 0 2 2 E0 E + 4y φ πE y πE − (φ ) ds + · · · . . we obtain the following where πE = δA/δ φ.4) (5. . Let us expand (5.  dy 1 We assume here that Z  L0 1 + Γ2 (p) φEp φE−p dp (5. + 8 y2 (5. . . p3 ) − D(p1 . .6) 4 X dΓ  4 2   =y Γ2 (pi ) Γ4 (p1 . p3 ). In this formula (5. equations for the coefficient functions:  p2 dΓ2  2 2  = y Γ − . . ds δxµ (s) δx1 (s) x1 =s ds δ φE will give us equations for the Γ ’s. . . . p4 ) = C4 (p1 . p3 ) − D(p2 . . . p2 ) + D(p3 . . . . p4 )(φEp1 φEp2 )(φEp3 φEp4 )δ pi dp1 · · · dp4 + · · · . .7) A= y 2 Z   X 1 Γ4 (p1 .5): (2π)B4 (p1 . . p4 ) − B4 (p1 . .  2  dy y2 ! (5. . p4 ) − D(p1 . − 8 The only slightly complicated structure is the polynomial B4 which is found from (5. p4 ) − D(p2 . p4 ) + D(p1 . p4 ).5) E If we perform the Fourier transform in s.

the Hamilton– Jacobi method has some general advantages. The equations (5. . we arrive once again at the expression (5. .8): (2π)Γ4 = F (p1 . + ∆p1 p2 p3 p4 p1 p2 p3 p4 1 2 3 4     1 2ε1 ε2 ε1 ε2 1 1 + + (5. . . φi = φi (σ. p4 ). (5.10) and separating the finite part for y → 0 gives    ε1 ε2 ε3 ε4 + 1 ε1 ε2 ε3 ε4 X 1 + F= 2 ∆3 ∆2 |pi | P  |p | · |p | i j ∆ i<j − p2 p2 p2 p2 . . . . . However. Rychkov / Nuclear Physics B 581 (2000) 116–134 C4 = (p1 p2 p3 p4 ) 1 + y 6 (ω1 ω2 ω3 ω4 ). y2 D(p1 . p4 ) + Φ12 + Φ34 − Φ13 − Φ14 − Φ23 − Φ24 .12) + p2 p2 p2 p2 . where we introduced the notation ωi = Γ2 (pi ). (5. .) From the second one. y) ≡ ω(p. From the first one we get Γ2 (p. p4 |y). . y = τ.9) (this solution is the only one which is positive and regular for y → ∞.) This completes the calculation of Γ4 (p1 . τ ). . Polyakov. τ ) = |p|τ 3/2   E E K3/2 |p|τ φ(p) = 1 + |p|τ e−|p|τ φ(p). Expanding (5.13) A= τ2 E The linear equation for φ. p4 |y) P Y (1 + |pi |y) e− |pi |y B4 (p1 . . y) = p2 y(1 + |p|y) (5.   1 1 ∂τ 2 ∂τ φE − 2 ∂σ2 φE = 0 τ τ has the solution φEcl (p. Φ12 = ∆3 ∆2 |p1 | |p2 | ∆ p1 p2 1 2 3 4 P (∆ = |pi |. Z∞ Γ4 = dy e− y Z∞ = dy y Ry 0 dy1 y12 ( P ωi ) B4 (p1 . .6) are easy to solve.126 A. . p2 ) = (p1 p2 ω3 ω4 )y 2 . Z q 2 dτ 1 + φEτ2 + φEσ2 + φEτ2 φEσ2 − φEτ φEσ .10). . V. Let us notice that an alternative way to obtain these formulas is to use the Monge gauge in the expression for the minimal area: x1 = σ. .11) Taking the integral (5.13). . (5.10) i We see that Γ4 also has the structure of (5. εi = sgn pi .

This fact was considered in [15] as a check that the loop equation is satisfied. We have:   x 2 (1 + 4x + x 2 ) x3 x4 5 2 + − 2x (1 + x) + F (k. it makes sense to express them as functions of x = k/p. p1 .6) is absorbed into the mass renormalization of the test particle. p1 ) + Φ(k 0 . p2 ) + (D − 3) Φ(k.3) drops out. The remaining terms have to be expanded for k → ∞.2) The transverse part (in momentum representation) can be read off directly from (5. That is why it is necessary to consider the second term. k 0 ) + Φ(p1 . (2π) 0 E E 2 δ φ(k)δ φ(k )   H = (D + 1)F (k. In this case the third term in the expression for H in (6. y). The second derivative of the minimal area After the divergent part of (4.7). p2 ) + Φ(k 0 . which amounts to taking p = 1 in (5.12). p) is antisymmetric with respect to k → −k. where p1 = −p2 = p. Let us first analyze the case k = −k 0 . V. p2 ) φEp1 φEp2 δ(k + k 0 + p1 + p2 ) dp1 dp2 . For the term quadratic in φ we have δ 2 A(2) = (D − 1)δ(q)Γ2 (p) = (1 − D)δ(q)|p|3 .11) leads to the formula Z  1 δ 2 A(4) =− H (k.   q δA δA δ2 A δ2W 2 2 N = g − N g W. YM YM δxµ (s)δxµ (s 0 ) δxµ (s) δxµ (s 0 ) δxµ (s)δxµ (s 0 ) The first term in brackets has no singularity for s → s 0 . Polyakov. Notice however that in this order an arbitrary functional A will pass this check. = + E E 0) δxµ (s)δxµ (s 0 ) δx1 (s)δx1 (s 0 ) δ φ(s)δ φ(s (6. k 0 . The special structure in (5. p→∞ δ φ( E q + p)δ φ( E q − p) 2 2 lim where Γ2 (p) = −|p|3 is the finite part of Γ2 (p. −k. Let us now consider the quartic term. −p) = |p| 2(1 + x) 2(1 + x)3 2(1 + x)    3 = |p|5 − x 3 − x + O x1 . p. p1 . −k) + Φ(p. The full second variational derivative of A consists of the longitudinal and transverse parts: δ2 A δ2 A δ2 A .6).3) − (D − 1) Φ(k.1) In its second variational derivative. −p) = |p| − 2(1 + x) 2(1 + x) 2(1 + x)3 . because Φ(k. which corresponds to taking q = 0 in (2. k 0 . p2 ) . which we proceed to calculate.A. Rychkov / Nuclear Physics B 581 (2000) 116–134 127 6. (6. p1 ) + Φ(k. we are left with the functional W [C] = e q 2 N A[C] − gYM . p2 )   (6. 2  2  x3 x4 x (1 + x 2 ) 5 − − Φ(k. Because of homogeneity.

Rychkov / Nuclear Physics B 581 (2000) 116–134  1 = |p|5 − x 3 − x + 2 + O 2 1 x   .6) In the approximation we are working with. V. Polyakov. The result is: δ2A + k)δx1 ( q2 − k) .3) and obtain the desired asymptotic expansion in k → ∞: Z  2 3  1 δA(4) = Dp |k| + (D − 1)p4 |k| + (3 − D)|p|5 φEp φE−p dp E E 2π δ φ(k)δ φ(−k)  (6. The required identity can be easily derived from (5. we have to take all the terms in (6.6). (6.2). when we do not assume that q = 0 in (2.3) into account.4) + O 1k . In general. We substitute this into (6.5) To complete the calculation of (6.2)). we have to compute the longitudinal part (the first term in (6. The corresponding formula for the asymptotic expansion (of which (6.128 A. only A(2) contributes to the RHS of (6. This part is not immediately visible from (5.4) is a partial case) can be obtained by means of straightforward but lengthy calculations and has the form: δ 2 A(4) E q + k)δ φ( E q − k) δ φ( 2   Z  2 D−4 3D − 6 2 2 1 p1 p23 + p1 p2 |k| −Dp1 p2 |k|3 + = 2π 2 2     + (4 − D)p12 |p2 |3 + p1 p2 |p2 |3 φEp1 φEp2 δ(p1 + p2 + q) dp1 dp2  + O 1k .7) and has to be recovered by the use of the reparametrization invariance.6). − dp p(p + k + k 0 )φ(p) E δ φ(p + k + k 0 ) (6. ) φ(s) E δx1 (s)δx1 (s 0 ) δφi (s)δφk (s 0 ) ds δ φ(s) In the momentum representation: Z δ2 A δ2A 0 0 0 = − dp dp pp φ (p)φ (p ) (2π) i k δx1 (k)δx1 (k 0 ) δφi (p + k)δφk (p0 + k 0 ) Z δA E .4) and has the form:   δ2 A d δA δ2A 0 ˙E ˙i (s)φ˙ k (s 0 ) = φ − δ(s − s .

3  Z .

.

.

p1 − p2  1 .

− |p2 |3 p1 p2 φEp φEp δ(p1 + p2 + q) dp1 dp2 .

+ k = 1 2 .

.

(6.7) k→∞ 2π 4 δx1 ( q2 The final expression for the asymptotics of the second variational derivative is thus: . 2π 2  Z  3 1 = |k|3 + (p1 − p2 )2 |k| − p1 p2 |p2 |3 × · · · .

8) In the next section we will analyze this result. The presence of those terms would imply that our functional is not zigzag-invariant. The last two terms give no singular contribution to (2. Lq A = 2π This shows that at D = 4 the loop equation is satisfied (at least in our approximation)! We will discuss the significance of this fact in the next section. V. Namely.. 7. Consider once again the OPE (2. Interpretation and discussion of the result Our main result is contained in the formula (6.8). Polyakov. The next result following from (6. that the second variational derivative has the expected form (2. It shows.7). 1 ¨ 2 (C1 + C2 )(x˙ x ) C1 x¨ 2 C2 (x˙ x) + + + . Rychkov / Nuclear Physics B 581 (2000) 116–134 δ 2 A(4) = (1 − D)δ(q)|k|3 + k)δxµ ( q2 − k)   Z  1 D−1 3D − 9 2 2 3 3 p1 p2 + p1 p2 |k| + (1 − D)p1 p2 |k| + 2π 2 2   + (4 − D)p12 |p2 |3 φEp1 φEp2 δ(p1 + p2 + q) dp1 dp2  + O 1k + O(φ 4 ).3) and (2. Using (2.4). (7.2) |x1 − x2 |2 (x = (x1 + x2 )/2). we get Z  4−D b (7.2).1) p12 |p2 |3 φEp1 φEp2 δ(p1 + p2 + q) dp1 dp2 .. If we assume that in conformal theory in the WKB limit the field strength keeps its normal dimension 2 (this will be actually more of the conclusion than of the assumption). Now let us perform another test. that is not presentable in the form (2. first of all. 129 δxµ ( q2 (6. we get δλσ (x1 − x2 )λ (x1 − x2 )σ + C2 |x1 − x2 |4 |x1 − x2 |6 (x1 − x2 )(λ (x1 − x2 )µ Fµσ ) (x) + C3 |x1 − x2 |4 (x1 − x2 )µ ∇µ Fλσ (x) + C4 .A.6) and picking up k 0 terms in (6.4) and let us try to determine its contribution to (6. dangerous terms ∝ δ 00 (s − s 0 ) (which would manifest themselves in (6. In the wavy line approximation we get .8) as terms ∝ k 2 ) cancel for all D.8) concerns the loop operator b Lq .8). The first two give after some calculations: Fµλ (x1 )Fµσ (x2 ) ∼ C1 1 C1 + C2 δ2W = δxµ (s)δxµ (s 0 ) s→s 0 |s − s 0 |4 x˙ 2   . |s − s 0 |2 12x˙ 4 4x˙ 4 4x˙ 4 The derivatives in the RHS are taken at the point s¯ = (s + s 0 )/2.8).

These versions unavoidably contain other fields. Conclusions and outlook The main efforts of this work were directed towards the development of new techniques for checking the loop equations in string theory. Various gauge theories presumably have a string representation with the background (1. C1 2 1 ˙φ + ¨ + O(φ 4 ).4) We see that this formula can be put in complete agreement with (7. we looked at the Wilson loop (2. optimistic view.1). If the theory is conformal. what is the meaning of finding that b LAmin = 0 at D = 4? There are several possible interpretations of this fact. Polyakov. Let us begin with the unpleasant one (in which we do not believe). the structure of the first term cannot be modified by anything and provides a strong check for the consistency of our approach. At the same time. (1.3) by taking C1 = D − 3. Let us take another.. 3 − D 2 1 ˙φ + ¨ . φ φ + 12 4 |s − s 0 |2 (7. we have  1 δ2 A ∝ (1 − D) 1 − φ˙ 2 0 0 4 δxµ (s)δxµ (s ) |s − s |   1 − D . The second term in (7.. where Jν is the current generated by those fields.1) must describe the AdS space. C2 = 2. 8. So. while if we proceed to higher orders in “waviness” of our contour.1) in conformal versions of the Yang–Mills theory. Hence we expect that ∇µ Fµν = Jν 6= 0. Namely. This point perhaps requires some clarifications. φ φ + 12 4 |s − s 0 |2 (7. V. |s − s 0 |4 |k| ↔ − 1 6|s − s 0 |2 to go back to the s-representation. the loop equations will not be satisfied. Rychkov / Nuclear Physics B 581 (2000) 116–134  δ2W 1 = (C1 + C2 ) 1 − φ˙ 2 0 0 4 δxµ (s)δxµ (s ) s→s 0 |s − s |   (C1 + C2 ) .4) may be modified if the theory contains a scalar operator of dimension 2... What distinguishes various theories is not the metric but other background .8). Picking up the terms ∝ |k|3 and |k| and using the Fourier transform identities |k|3 ↔ 1 . We managed to apply the loop Laplacian to the minimal area in the Lobachevsky space and to show for the first time that the equations of motion of gauge theory are satisfied by string theory. It may be that our result is just a fluke.130 A. That can change the constants C1 and C2 . Notice that it also predicts that the dimension of Fµν is not renormalized. Further progress will be difficult in this case.3) We have to compare this behavior with our formula (6. at least in the WKB approximation.

when these theories become conformal.5) by developing the short distance expansion of the Green functions for equations (4. and our result actually checks the universal law (3.1) is conformally invariant for D = 4. Alternatively. (A. in the WKB limit gYM ∞ the asymptotics of the Wilson loop is given by the formula (3. and ρ is some factor. in the asymptotically free theories. Conformal invariance of the loop Laplacian (which we discuss in the Appendix) should play an important role in this analysis. It may be helpful to notice that instead of considering non-conformal case in 4d. V.6) and is the same for all conformal theories. ds = f (log y) y2 The problem for our analysis is not so much the function f (it is easy to generalize our considerations to this case) as the absence of the WKB domain. with the above philosophy this modification is irrelevant in the WKB limit. Acknowledgements We are grateful to Volodya Kazakov for his participation in the early stages of this project as well as for useful discussions. one can. In the case of non-conformal theories the metric has the form  2  dy + dE x2 2 . shift to D = 4 + ε. was partially supported by NSF grant PHY-98-02484. To verify the above assertions it is necessary to go beyond our wavy line approximation.P. in order to check the equation at a point x(s) of a contour C you are allowed to . the method of wavy lines may be useful here.14]. The work of A. say fµ (x) = xµ /x 2 . Rychkov / Nuclear Physics B 581 (2000) 116–134 131 2 N → fields and also the field content on the world sheet. This brings us to a more difficult problem of going beyond WKB approximation. Conformal invariance of the loop equation We are going to show that the loop equation (2. Perhaps. This means that b L(s)W [f (C)] = ρ b L(s)W [C]. As a result. However. It is conceivable that by this method it will be possible to relate OPE on the world sheet and OPE in gauge theory. it means that the current in the above formula is negligible in the WKB limit. Polyakov. Notice also that we are considering the standard Wilson loop and not its modified version suggested in [13. If this is the case. but at the moment we do not know how to evaluate the action of b Lq on quantum corrections to our formula. one can study the second variation of the functional (3. Once again.5) for the Wilson loop (6.6). the classical bq W ∝ ε will be canceled by the quantum fluctuations and will provide us with the part of L “Holy Grail” of this subject — the space-time β-function.A. Again.1). Appendix A.1) where f (x) is a conformal transformation. We believe that this can be done by a more general treatment of the Hamilton–Jacobi equations.

Polyakov. the constant cut-off ε is transformed by F to the variable cutoff ε/x 2 . say with the purpose of simplifying the behavior of the contour at the point you are looking at. the LHS is equal to L[C]/ε + A[C].132 A.3) where Mε is the surface M cut off at the hight ε.6). Although this fact is to be expected. V. Since F is an isometry.ε →0 .3). it is convenient to introduce an auxiliary constant cut-off ε0 on F (M). Rychkov / Nuclear Physics B 581 (2000) 116–134 first apply a conformal transformation.2) (xµ . (A. . so that Area[F (Mε )] becomes equal to Area[F (M)ε0 ] plus the area of a narrow strip between the variable and constant cut-offs. Now (A. we extend the above conformal transformation f to the (D + 1)-dimensional Lobachevsky isometry:   xµ y F . F (M) is the corresponding surface for f (C).2). Moreover. it might become important for general contours.1) is a consequence of two independent observations.3) implies: L[f (C)] L[C] + A[C] = + A[f (C)] + 0 ε ε0 ε. The first one is that the functional A (and hence W ) is conformally invariant for any D [17]. According to (A. To calculate the RHS of (A. Although this observation does not play any significant role when working with wavy lines. y) −→ x2 + y2 x2 + y2 Consider the minimal surface M bounded by C. Area[Mε ] = Area[F (Mε )]. it is not entirely obvious. The relation (A. (A. since isometries of the Lobachevsky space (corresponding to conformal transformations on the boundary) will change the cut-off ε in (4. For small ε. To check the invariance.

I .

Zε0 .

df (x(s)) .

dy .

ds .

. .

.

+ ∂ 2fλ x(s) δfλ (s) . consider the relation:   δ 2 Uf [x(s)] δ2 U 0 = ∂ f f ) x(s) ∂ x(s µ λ µ σ δxµ (s)δxµ (s 0 ) δfλ (s)δfσ (s 0 )  δU δ(s − s 0 ). ds y2 C ε/x(s)2 Calculating the integral. then Uf is also a solution.e. and we get the conformal invariance: A[C] = A[f (C)]. (A.4) The second of the two observations alluded to above is that the loop Laplacian is also conformally invariant (i. In other words. The invariance means that for any (reparametrization invariant) functional U [C]:  b L(s)U [f (C)].5) L(s)Uf [C] = ρ b where Uf [C] = U [f (C)] is the functional U transformed by a conformal transformation. this time only for D = 4. if a functional U satisfies the loop equation. To prove this. we see that the singular terms cancel. commutes with conformal transformations). (A.

Nucl. B 496 (1997) 231. (A. M. Makeenko. B 80 (1979) 372. S. Polyakov. J. .5) is true for D = 4. Polyakov. C. Phys. E. Lett. Math. E. Adv. Phys. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] A. Phys. Maldacena. A. Lett. Phys.6) Substituting fµ = xµ /x 2 . I. Notice that for D 6= 4 the presence of nonzero additional term in the RHS of the transformation law (A. D. J. B 88 (1979) 135. Phys. B 80 (1979) 255. Nucl. Friedlan.5). hep-th/9702076. A. we must collect terms proportional to δ(s − s 0 ).-J. we find that Ω(f ) ∝ (D − 4).1) follows from (A. Polyakov. B 428 (1998) 105. Phys.4). Yu. Phys. B 164 (1980) 171. 80 (1998) 4859. hep-th/9711002 (Proceedings of “Strings 97”). Rey. and thus (A. Phys. L(s)Uf [C] = ρ(f ) b  Ω[λσ ]µ = ∂ 2 f[λ (∂µ fσ ] ) − (∂α ∂µ f[λ )(∂α fσ ] ). Phys. J. A 14 (1999) 645. I. B 262 (1985) 593. Klebanov. hep-th/9803002. Lett. J. Lett. 2 (1998) 253. Phys. Gervais. Nucl. V.A. S. Rev. Yee. Now. A. A. Polyakov. Polyakov. Rychkov / Nuclear Physics B 581 (2000) 116–134 133 For conformal transformations we have ∂µ fλ ∂µ fσ = ρ(f )δλσ . hep-th/9711200. A. δxλ (s) This gives the transformation law  b L(s)U [f (C)] + Ω[λσ ]µ (f )N[λσ ] x˙ µ (s). Perry. J. Lett. hep-th/9802109. Phys. Maldacena. hep-th/9809057. This gives   δU b L(s)U f (C) δ(s − s 0 ) + ∂ 2 fλ δ(s − s 0 ) L(s)Uf [C]δ(s − s 0 ) = ρ(f ) b δfλ   + ∂µ fλ x(s) ∂µ fσ x(s 0 ) N[λσ ] δ 0 (s − s 0 ). δfλ (s)δfσ (s 0 ) 2 From the condition of the reparametrization invariance dxλ δU =0 ds δxλ (s) we have the identity [6]: δU = N[λσ ] (s)x˙σ (s). if we notice that Wf = W by (A. Neveu. Lett. Gubser. hep-th/9802150. Martinec.6) implies that the loop equation for (6. where N[λσ ] is the coefficient in front of the δ 0 -function in the second variational derivative:   s + s0 0 δ2 U = N[λσ ] δ (s − s 0 ) + · · · . Math. Klebanov. Adv. Phys. Witten. Theor. Nambu. Mod. Polyakov. hep-th/9803001.1) cannot be satisfied in this case. Callan. Finally. Y. B 82 (1979) 247. the relation (A. A. Theor. Migdal. 2 (1998) 231.

Nucl. Graham.J. Phys. Gross. Witten. B 546 (1999) 52. 34 (1986) 11. Polyakov. Ooguri. E. Dorn. Fortschr. D 60 (1999) 125006. [17] C. . Rev. [16] H. hep-th/9904191. V. H. Rychkov / Nuclear Physics B 581 (2000) 116–134 [15] N. D. Phys.134 A.R. hep-th/9901021. Phys. Drukker.

Introduction One of the most fascinating consequences of the discovery of D-branes is their intimate connection with gauge theories (good reviews on the subject can be found in [1–3]). The F 2 term is nothing but a U (N) Yang–Mills theory. All rights reserved.3 a Department of Mathematics.9 . B-1050 Brussels.1 . the situation changes.Nuclear Physics B 581 (2000) 135–155 www. we perform a detailed study of the mass spectrum of the non-abelian Born–Infeld theory.vub. 2000: CTP.V. When more than one brane is present. Pleinlaan 2.edu 2 asevrin@tena4.be 3 Aspirant FWO. When the branes coincide. Vrije Universiteit Brussel.ac.3 2 1 3 ( 0 0 ) 0 0 2 7 8 . additional massless states appear. we compare this for several cases to the mass spectrum of intersecting D-branes. gets promoted to a U (N) non-abelian Born–Infeld theory [4]. The strings stretching between two branes have a mass proportional to the shortest distance between those branes. address after September 1. the effective action for N well separated branes. Alexander Sevrin b. it is clear that the trace over the Lie algebra elements should necessarily be taken outside the square root. Subsequently. Columbia University. PII: S 0 5 5 0 . Cambridge. MIT. on tori with constant magnetic fields turned on.nl/locate/npe Non-abelian Born–Infeld versus string theory Frederik Denef a. USA b Theoretische Natuurkunde.vub. Jan Troost b. Belgium Received 23 February 2000.2 . 1 denef@math.be 0550-3213/00/$ – see front matter  2000 Elsevier Science B. while the F 4 term has been obtained directly from both the calculation of a four point open string amplitude [5.  2000 Elsevier Science B. Exact agreement is found in only two cases: BPS configurations on the four-torus and coinciding tilted branes.V. On this basis and from the fact that for the open superstring. The dynamics of the massless fields on an isolated Dp-brane is well described by a U (1) Born– Infeld action. described by a (U (1))N Born–Infeld theory.8].ac.6] and from a threeloop beta-function [7. USA.columbia. defined by the symmetrized trace prescription. As the effective action has to reproduce tree-level open string amplitudes. NY 10027.elsevier. Finally we investigate the fluctuation dynamics of an arbitrarily wrapped Dp-brane with flux. accepted 8 May 2000 Abstract Motivated by the results of Hashimoto and Taylor. The two leading terms are known. The precise form of a nonabelian Born–Infeld theory is still unclear. In this way. All rights reserved. Email: troost@tena4. 1. New York.

In a beautiful paper [10] (see also [11]). while a trace over groupvariables is written as Tr. Denef et al. . A symmetrized trace over groupvariables is denoted by STr.136 F. j. Using these results. In [12] it was shown that the example at hand was not sufficient to uniquely resolve the discrepancy at order F 6 . it is simply 4 4 We denote space–time indices by α. β. We end with some speculations on the true structure of the non-abelian Born–Infeld theory. one obtains a configuration of intersecting D-branes which allows for a string theoretic calculation of the spectrum. the Lie algebra elements should be fully symmetrized. A general prescription is developed which allows to perform the calculation in the presence of arbitrary constant and abelian backgrounds. . . Traces over space–time indices are denoted by tr. For a configuration with no transverse scalars excited. 2. The analysis in [10] was limited to the four-torus. / Nuclear Physics B 581 (2000) 135–155 odd powers of the fieldstrength should vanish. We work in units where 2π α 0 = 1 and we ignore an overall factor. Upon compactifying the theory on a torus and T-dualizing. In Section 4 we apply these results to the simplest case: the two-torus with arbitrary magnetic fields. we systematically study the spectrum of the non-abelian Born–Infeld theory in the presence of constant magnetic background fields on various tori. . Space-like indices are denoted by i. one calculates the spectrum of the Born–Infeld theory. Subsequently. some evidence was found that this was not the case [10]. . In the following section. is a formidable task as it requires at least either the calculation of a six-point string amplitude or a five-loop beta-function. In the next section. The abelian Born–Infeld Lagrangian describes an isolated Dp-brane. we examine the spectrum of the now well-understood abelian case and compare it to the string theory spectrum. as only in that case the spectrum of the off-diagonal fluctuations of the corresponding Yang–Mills theory was known [13]. one of us extended the analysis of [13] to other tori [14]. additional sample calculations are made: general configurations on T 4 and BPS configurations on T 6 are studied. Abelian Born–Infeld As a warm-up exercise. . We compare both Born–Infeld and string theory results. . The outline of this paper is as follows. . . Checking this proposal at the level of F 6 and higher directly. Recently. However. it was shown that certain aspects of the full Born–Infeld can be probed by switching on sufficiently large constant background fields. We work in a Minkowski signature with the “mostly plus” convention. Consistency requires that both results agree. Section 3 is devoted to the calculation of the quadratic fluctuations of the non-abelian Born–Infeld theory. Tseytlin formulated a conjecture for the nonabelian Born–Infeld action [9]: before taking the trace. The subsequent section is devoted to more involved wrappings of Dp-branes on T p . This was the main motivation of the present work in which tools are developed which make the calculation possible for arbitrary configurations. The two-torus provides the clearest and simplest example of the discrepancy. we briefly recapitulate the abelian Born–Infeld theory and its stringy interpretation. A direct test of the non-abelian Born–Infeld theory emerged in this way.

1) becomes. Upon rescaling the space-like directions.6) (2. we can always bring the background fields in a canonical form such that only F12 .2) L = − det(1 + F + δF )  1/2 . (2. We expand the gauge field Aα around a fixed background Aα . F56 .9) k We compactify some of the spacelike directions on a torus. 1/2 (2. Ci . . this reduces to   1 1 ij kl −1/4 ij Lquad = (det G) G δF0i δF0j − G G δFik δFj l . Doing this and introducing λi . we obtain the final form for the Lagrangian.  2 −1 (2. 2 2 (2. Denef et al. / Nuclear Physics B 581 (2000) 135–155 L = − det δαβ + 2πα 0 Fαβ 1/2 . 1 Lquad = (det G)−1/4 tr(GδF GδF ).3) = −(det G)−1/4 det(1 + GδF + BδF ) where G ≡ 1 − F2 −1 B ≡ −F 1 − F . s   1 Y 1 −1 −1 −1 λk λi δF0i δF0i − λi λj δFij δFij . as Li .10) M = λi Li i where the mi are integers. . an ordinary abelian Yang–Mills theory. X 1  2πmi 2 2 . Eq. 137 (2.1) with Fα β ≡ ∂α Aβ − ∂ β Aα . We subsequently determine the spectrum of masses in the uncompactified directions. F0i = 0. (2. 4 When only magnetic background fields are turned on.F.5) We ignore the leading constant term and drop the term linear in the fluctuations as it is a total derivative. . and turn on magnetic fields in these directions only.7) By an orthogonal transformation. F34 . λ2i−1 = λ2i ≡ 1 + (F2i−1 2i )2 .e. modulo an irrelevant overall scale factor. Eq. i. The first non-trivial term is quadratic in the fluctuations and upon partial integration it assumes a simple form. are possibly different from zero. In this way we immediately obtain the spectrum of the abelian Born–Infeld theory. 5 We take the compactified space to be T 2n and write the length of the ith cycle. . Lquad = 2 2 (2. 5 We use the same convention for general spacelike indices and those parametrizing the torus. (2..9) becomes. (2. (2.4) .8) (2. The context should make it clear what is meant. Aα = Aα + δAα and choose the background such that its fieldstrength F is constant.

We parametrize the fieldstrength as F = F + δF and the connection by A = A + δA. both string theory and the Born–Infeld theory yield the same spectrum. / Nuclear Physics B 581 (2000) 135–155 We now make contact with string theory [10]. In other words. respectively. i ∈ {1. To calculate its spectrum from string theory we closely follow [10] and T-dualize along the C2i . and switching back to the original variables.10) and (2.138 F.  1/2 (3. (2. . In order that the transition functions are well defined. Calculating the spectrum of the dual configuration using string theory is now straightforward and yields  n  X 4π 2 m022i−1 m022i L022i−1 n2i L022i + . Quadratic fluctuations of non-abelian Born–Infeld Turning to the non-abelian theory.1) L = − STr det δαβ + Fαβ Taking a symmetrized trace means that upon expanding the action in powers of the fieldstrength. Picking out the term quadratic in the fluctuations gives   1 −1/4 1 tr Bδ2F − tr(Gδ1 F Gδ1 F Lquad = − STr (det G) 2 4 .5). we parametrize the background as. one first symmetrizes all the Lie algebraic factors and subsequently one takes the trace. All background fields take values in the Cartan subalgebra (CSA). 3. We consider a constant background fieldstrength F and denote the background connection by A. L2i−1 L2i A2i−1 = 0. (2. we introduce Lie algebra valued gauge fields Aα and their fieldstrength Fαβ = ∂α Aβ − ∂β Aα − i[Aα . F2i−1 2i = 2πni . The sizes of cycles and ni times around the C2i around the C2i−1 the cycles of the dual torus are L02i−1 = L2i−1 and L02i = 2πL−1 2i .12) M2 = L0 22i−1 + n2i L0 22i L0 22i−1 + n2i L0 22i i=1 Identifying the integers m02i−1 and ni m02i with m2i−1 and m2i . . n} cycles. Following [9].4) and (2. (2.12) match.11) where ni are integers. . shows that Eqs.3) det(1 + GδF + BδF ) = − STr(det G) where G and B were defined in Eqs. In order to achieve this. the Lagrangian becomes. This configuration corresponds to a single D2n-brane wrapped around T 2n with a certain number of lower dimensional D-branes dissolved in it. In terms of these variables. (3. (3. (2. A2i = F2i−1 2i x 2i−1 . the first Chern class should be an integer. we define the nonabelian Born–Infeld Lagrangian by  1/2  . Denef et al. .2) L = − STr det(1 + F + δF )  1/2 −1/4 . The resulting configuration is a single Dn-brane wrapped once 0 0 cycles of the dual torus. Aβ ].

9) 4 4 8 G 00 The terms proportional to B 2 lead. (3. type 2 ≡ STr(δ1 F¯ δ1 F F1 F2 · · · F2n )  1 Tr δ1 F¯ (δ1 F F1 F2 · · · F2n + all permutations) .4) D· ≡ ∂ − i[A. upon partial integration.12) = (2n + 1)! where the presence of the subindices and the bar above the first variation reflects the possibility that these various terms might have different space–time index structures.7) δ2 Fαβ ≡ −i[δAα . 4 where. In the analysis of the Lagrangian at the quadratic level.8) = −1. (3. δF = N X δF ab Eab . δAk] δAl . i. ·]. Denef et al.F.11) We turn now to the contributions quadratic in the variation of the field strength.. (3. B ij = −(F )ik G kj . Using this notation. (3. type 1 ≡ STr(δ2 F F1 F2 · · · Fn−1 ) = Tr(δ2 F F1 F2 · · · Fn−1 ). (3. (3. Recalling that the background field strength takes values in the CSA. (3. The Lagrangian reads then.b=1 where Eab is an N × N matrix unit. the commutator term has to be considered as a single Lie algebra element. (3. As an intermezzo. to. Later on. it is an N × N matrix which is zero except on the ath row and bth column where there is a 1. They are of the form. we now turn to the calculation of the symmetrized traces.13) a. 139  . we take the background field purely magnetic. (3. we can perform the trace in the type 2 term. we get two kinds of contributions: those linear and those quadratic in the field strength variation.5) δF ≡ δ1 F + δ2 F. Since all fields are in the fundamental representation of the group U (N) [9]. G 0i = Again.6) δ1 Fαβ ≡ Dα δAβ − Dβ δAα . / Nuclear Physics B 581 (2000) 135–155 1 + Bδ1F Bδ1F ) + (tr Bδ1 F )2 8 where we used. we will see that this term gives an additional contribution to the zero point energy.   1 1 −1/4 Lquad = − STr (det G) − B ij δ2 Fij − G ij δ1 F0i δ1 F0j 2 2  2 1 ij kl 1 ij kl 1 ij + G G δ1 Fik δ1 Fj l + B B δ1 Fil δ1 Fj k + B δ1 Fij . we have F= N X a=1 F a Eaa . we get that for the linear terms. when performing the symmetrized trace. .e.10) − Str(det G)−1/4 B ij B kl F[ij .   3i (3. δAβ ]. the symmetrized trace reduces to an ordinary trace. implying B 0i = 0.

are non-zero. 2. .14) + 2n!F1 F2 · · · F2n . Denef et al. Given some arbitrary even function of the backgroundfields H (F ). without loss of any generality.15) + 2n + 1 This suggests a simple way to implement the symmetric trace. (3. Defining F a = F 0 + F 3 and F b = F 0 − F 3 . σ2 and σ3 are the Pauli matrices. . (3. In the remainder of this section. The fluctuations are written as δF = 3 X a=0 δF (a)σa . (3. The background is of the form fi = fi0 σ0 + fi3 σ3 .b=1 b b b b b × 2n!F1 F2 · · · F2n + (2n − 1)!F1a F2b · · · F2n + (2n − 1)!F1b F2a F3b · · · F2n b a b + · · · + (2n − 1)!F1b F2b · · · F2n−1 F2n + 2!(2n − 2)!F1a F2a F3b · · · F2n + ···  a a a (3. (3.17) 0 These results allow for the calculation of the symmetrized trace through second order in the fluctuations in the presence of constant abelian background. we get in a U (2) subsector. n}. We consider the Born–Infeld theory on an even-dimensional torus T 2n . we get for fixed a and b.19) . we will assume this anyway.16) STr δ1 F¯ δ1 F H (F ) = δ1 F¯ ba δ1 F ab + δ1 F¯ ab δ1 F ba I(H ). In other words we work with a U (2) subgroup. From this expression we see that we can look at each sector with given a and b separately. . (3. we will use this to make the action Eq.   1 ba ab ab ba 0 0 ¯ ¯ + F13 F23 F30 · · · F2n type 2 = δ1 F δ1 F + δ1 F δ1 F F10 · · · F2n 3  0 0 3 3 + F13 F20 F33 F40 · · · F2n + · · · + F10 · · · F2n−2 F2n−1 F2n  1 0 + ··· + ··· + F13 F23 F33 F43 F50 · · · F2n 5  1 3 F13 · · · F2n . we will work with a U (2) theory.140 F. for i ∈ {1. where I(H ) ≡ 1 2 Z1   dα H F 0 + αF 3 + H F 0 − αF 3 . Furthermore. . This time an orthogonal transformation is not sufficient to bring the background into a form where only fi ≡ F2i−1 2i .   (3.18) where σ0 is the 2 × 2 unit matrix and σ1 . In order not to overload the formulae.9) as explicit as possible. / Nuclear Physics B 581 (2000) 135–155 type 2 = N X 1 δ1 F¯ ba δ1 F ab (2n + 1)! a.

(3. (3. 6 Note that we slightly abuse language as also part of the potential term will contribute to the zero-point energy.F. Denef et al. (3. . T-dualizing along one of the directions gives two D1-branes intersecting each other at a certain angle. Subsequently. As the expression for the Lagrangian is rather involved.10). 4.20) fj3 I fi fj t +2 (1 + fi2 )(1 + fj2 ) j 6=i In a similar way. Using gauge invariance.j (1 + fk2 ) . This picture allows for a string theoretic calculation of the spectrum. ( v uQ n u k6=i (1 + (fk0 + fk3 )2 ) X  (3) δF2i−1 2i −t fi0 + fi3 L0 = 0 3 2 1 + (fi + fi ) i=1 v uQ u k6=i (1 + (fk0 − fk3 )2 )  fi0 − fi3 +t 0 3 2 (1 + (fi − fi ) v Q !) u X u k6=i.j (1 + fk2 ) I t L2 = − 2 (1 + fi2 )(1 + fj2 ) i=1 j 6=i n 1 XX × 2 X 2 (a) δ1 F2i−1 2j −1 + 2 δ1 F2i(a)2j +2 2 (a) δ1 F2i−1 2j ! a=1 (v ) 2 uQ  u k6=i (1 + fk2 ) X 2  (a) t I F δ . A case study: D2-branes on T 2 The simplest case at hand are two D2-branes on T 2 . one can show that the rescaling in Eq.22) The total Born–Infeld Lagrangian is the sum of L0 . − 1 2i−1 2i (1 + fi2 )3 i=1 a=1 n X (3. this can be compared to the spectrum obtained from the non-abelian Born–Infeld theory.22). we get the zero-point energy term 6 . L1 = 2 1 + fi i=1 a=1 (3. we obtain from Eq.9) the kinetic term (v ) 2 uQ n   u k6=i (1 + fk2 ) X X (a) 2 (a) 2 t I δ1 F0 2i−1 + δ1 F0 2i .20) coincides with the last rescaling in Eq. (3. we turn in the next two sections to some specific examples which we then compare to string theory. L1 and L2 . (3. / Nuclear Physics B 581 (2000) 135–155 141 Combining the term proportional to δ2 F with Eq.21) and the potential ) (v u Q u k6=i.

in the presence of two D1-branes wrapped around T 2 and intersecting each other at an angle γ . The spectrum from string theory We consider IIB string theory on T 2 × M8 . Combining the string coordinates on the torus as Z ≡ X1 + iX2 . Denef et al.142 F. We take the two branes along the 1-axis and rotate one of them over an angle γ into the 12-plane. / Nuclear Physics B 581 (2000) 135–155 4. we get the boundary conditions for the end point of a string tied to the first D1-brane . We now proceed with the calculation of the masses low-lying states thereby closely following [15] (see also [16] and [17]).1. where M8 is the eight-dimensional Minkowski space.

Im Z .

σ =0 = 0. .

= 0. (4.1) Re ∂σ Z .

σ =0 and for the end point tied to the rotated brane .

Im eiγ Z .

.σ =π = 0.

= 0. Re ∂σ eiγ Z .

2 2 48 4 (4. (n − β) = − 2 24 4 n=1  ∞  X 1 1 1 1 − n − − β = − + β 2.7) 48 48 4 . we find that the vacuum energy in the Ramond sector vanishes. = 2+ (4. 2 24 4 (4.4) dx 1 − e2x 2x 12 6 we get the regularized expressions for the vacuum energy of a boson with a moding shifted by β.2) Implementation of the boundary conditions leads to the following mode expansion X  † ei(m+β)(τ −σ ) .6) n=1 For the configuration under study. (4. ∞ 1 1 1X − β(1 − β).3) σ =π m∈Z where β = γ /π . Using.  1   1 1 d eax −6(a − 1)a − 1 − (a − 1)a(2a − 1) x + O x 2 . am−β ei(m−β)(τ +σ ) + a−m−β Z= (4. = 2+ x dx 1 − e x 12 6    1 1 1 d e(1+a)x 1 − 3a 2 + a − a 3 x + O x 2 . while in the Neveu–Schwarz sector it is given by:      1 1 1 + 2 × − + β(1 − β) 2π 6 × − 24 24 4     1 2 1 1 +2× − + β +6× − = γ − π. ∞ 1 1 1X (n − β) = − + β(1 − β). (4.5) n=1 and for Ramond and Neveu–Schwarz fermions respectively.

20). This can easily be generalized.14) From this it follows that the spectrum of the non-abelian Born–Infeld theory is the same as that of the corresponding non-abelian Yang–Mills theory.a=1  X 2 1 f0 −f3 f0 +f3 (a) 2 − 3 p −p . δ1 F12 0 3 2 0 3 2 2f 1 + (f + f ) 1 + (f − f ) a=1 (4.   f0 +f3 f0 −f3 (3) −p δ2 F12 Lquad = − p 0 3 2 0 3 2 1 + (f + f ) 1 + (f − f ) + 2 arcsinh(f 0 + f 3 ) − arcsinh(f 0 − f 3 ) X (a) 2 δ1 F0i 3 2f i. (3. We consider the U (2) Born–Infeld Lagrangian on T 2 and choose the diagonal background f ≡ F12 = f 0 σ0 + f 3 σ3 . / Nuclear Physics B 581 (2000) 135–155 143 Now it becomes simple to calculate the masses of the low-lying states in the Neveu– Schwarz sector. to include the transverse scalars and the fermions as well. we get an expression for the part of the Lagrangian quadratic in the fluctuations. we obtain the final result.8) 2 with a mass given by M 2 = (2m − 1)γ .2. The spectrum from Born–Infeld theory We will restrict ourselves to the study of the off-diagonal gauge field fluctuations.F.9) and (a−β )m ψ 0 − 1 −β |0iNS (4.11) 4. without altering our results. but rescaled by a factor ε. (4. . (4. (4.   f0 +f3 f0 −f3 (3) +p δ2 F12 Lquad = − p 0 3 2 0 3 2 1 + (f + f ) 1 + (f − f ) ( ( ) 2 2 ) 2 XX X 1 1 (a) 2 (a) 2 +I p −I p . The relevant states are (a−β )m ψ− 1 +β |0iNS (4. δ1 F0i δ1 F12 2 2 3 1+f (1 + f ) a=1 a=1 i=1 (4. Denef et al.10) 2 with a mass given by M 2 = (2m + 3)γ .13) Performing the symmetrized trace integrals.12) From Eqs.21). (3.

. Some other cases 5. The low-lying NS states have masses M 2 = (2m1 − 1)γ1 + (2m2 + 1)γ2. D4 configurations on T 4 were investigated. Comparing results T-dualizing the Born–Infeld configuration. (4. However. n) = (0.1. M 2 = (2m1 + 1)γ1 + (2m2 − 1)γ2. (4. we get two D1-branes wrapped around T 2 and intersecting each other at an angle γ given by   2f 3 . The Taylor expansion of the Born–Infeld rescaling around f 0 = f 3 = 0 consists of terms of the form (f 0 )2m (f 3 )2n . We reexamine this case and show that for self-dual configurations. we provide strong indications that any other configuration will disagree at order F 6 and higher. n ∈ N.9) and (4. both results match. which probes the abelian sector. 5. Combining this with the previous.16) 4. Denef et al.17) γ = arctan 1 + (f 0 )2 − (f 3 )2 Combining Eqs. The mismatch appearing at order F 6 and higher is quite serious. (4. 1). (4. M 2 = 2(2m + 3)εf 3 . from which it follows that through order F 4 . shows that the spectrum calculated from string theory does not match the spectrum predicted by the non-abelian Born–Infeld theory.3. M 2 = (2m1 + 3)γ1 + (2m2 + 1)γ2. The spectrum was obtained in [10]. / Nuclear Physics B 581 (2000) 135–155  ε= p × f0 +f3 1 + (f 0 + f 3 )2 −p f0 −f3  1 + (f 0 − f 3 )2 1 .15) The spectrum of a U (2) Yang–Mills theory on T 2n was calculated in [14]. the string theory results agree with the non-abelian Born– Infeld predictions. Of course. D4 on T 4 In [10]. n) = (m.17) and comparing it to Eq. we get the Born–Infeld spectrum for the off-diagonal fluctuations of the gauge fields. the Born–Infeld action gives correct results. it is given by M 2 = 2(2m − 1)εf 3 .144 F. in the limit when f 3 vanishes. We start by taking two 2-branes on T 4 along the 14 plane and rotating one of them first over an angle γ1 in the 12 plane and subsequently over an angle γ2 in the 43 plane. m. arcsinh(f 0 + f 3 ) − arcsinh(f 0 − f 3 ) (4. (4. Agreement with the string theoretic results occurs only for (m.11) with Eq. 0) and (m. Such a term arises from the F 2m+2n+2 terms in the non-abelian Born–Infeld action.16).

We choose a simple background F12 = f1 σ3 . two integers. (5. Denef et al. one finds the spectrum of the U (2) Yang–Mills theory on T 4 . 2 tan Eqs. It is given by (5.20). (5. / Nuclear Physics B 581 (2000) 135–155 145 M 2 = (2m1 + 1)γ1 + (2m2 + 3)γ2. f 2f In [13]. (5. (5.3) 1 + f22 !) f1 f2 (3) δ2 F12 + 2f2 I q 1 + f12 2 2 (1 + f1 )(1 + f2 ) s !) ( 1 + f12 f1 f2 (3) δ2 F34 + −2f2 + 2f1 I q 1 + f22 2 2 (1 + f1 )(1 + f2 ) (s (s ) 2 2 ) 2 4 1 + f22 X X 1 + f12 X X (a) 2 (a) 2 +I F + I F δ δ 1 1 0i 0i 1 + f12 a=1 i=1 1 + f22 a=1 i=3 (s (s ) 2 ) 2 X X 1 + f22 1 + f12 (a) 2 (a) 2 −I F − I δ δ1 F34 1 12 2 2 3 3 (1 + f1 ) a=1 (1 + f2 ) a=1 ) ( 1 −I q 2 (1 + f1 )(1 + f22 ) Lquad = −2f1 × 2  X (a) δ1 F13 2 (a) + δ1 F24 2 (a) + δ1 F14 2 (a) + δ1 F23 2  .2) The background fields are related to the angles by tan γ1 = f1 . which corresponds to a BPS configuration [15]. 2 (5. In that case.1) with m1 and m2 .4) a=1 The simplest configuration is the self-dual one f ≡ f1 = f2 .j =1 So the spectrum of the off-diagonal fluctuations of the non-abelian Born–Infeld theory is that of the corresponding Yang–Mills theory.21) yield s ( γ2 = f2 .5) a=1 i.6) . ε= γ arctan(f ) = . (3. (3. F34 = f2 σ3 . but rescaled by a factor ε. the integrals are standard and we get XX arctan(f ) (3) (3)  (a) 2 f δ2 F12 + δ2 F34 + δ1 F0i f 2 Lquad = −2 4 a=1 i=1 − 1 arctan(f ) 2 f 2 X 4 X (a) 2 δ1 Fij .F.

It looks highly unlikely to us that the Born– Infeld spectrum will match the string spectrum for generic choices of backgrounds/angles. was not properly taken into account.2. Denef et al. except when all backgroundfields are equal in magnitude. (5. Such a configuration should be described by choosing constant magnetic backgrounds F2i−1 2i = fi with fi = tan(γi /2). shows that this is indeed the desired result up to this order.1).7).7) Rescaling this spectrum by ε and comparing it to the string spectrum. (5. but with fi replaced by fi − (fi )3 /3. M 2 = 2(2m1 + 1)f1 + 2(2m2 + 3)f2 . (5. shows that for self-dual configurations. In fact this can already be seen by putting the f2 background to zero.22) in terms of known functions. However we believe that it is highly unlikely that the Born–Infeld action would reproduce the string spectrum. the rescalings are such that the result cannot be brought into a Yang–Mills form. (5. 2. using Eq. This time the integrals are more involved and can only be expressed in terms of elliptic integrals of the first and second kind.1). we briefly comment on the situation on T 6 . . 5. Eq. f ≡ f1 = f2 or γ = γ1 = γ2 . M 2 = 2(2m1 + 3)f1 + 2(2m2 + 1)f2 . Keeping only those corrections induced by the F 2 and F 4 terms in the Born–Infeld action gives a spectrum as in Eq. Indeed. BPS states on T 2 correspond to abelian backgrounds and on T 4 they are either abelian or self-dual. (5. followed by another rotation over an angle γ2 in the 43-plane and finally we rotate it by an angle γ3 into the 65-plane. (3. 3}. This can be circumvented by Taylor expanding the coefficients in Eq. one could repeat the analysis of [14] which will be complicated by the fact that the mass operator will not be diagonal in the Lorentz indices. Once more. However. the calculation reduces to the one studied in the previous section where no agreement was found. Doing this one notices that once more. We now turn to the more complicated case. Comparing this to Eq.4) around f1 = f2 = 0. In order to calculate the spectrum. When all background fields are equal in magnitude. a more serious problem arises here as well: no linear coordinate transformation can bring Eq. where i ∈ {1.3). one would be tempted to conclude that Tseytlins proposal for the non-abelian Born–Infeld action does work for BPS states. However.146 F. / Nuclear Physics B 581 (2000) 135–155 M 2 = 2(2m1 − 1)f1 + 2(2m2 + 1)f2 . (5. Eq.10). there is perfect agreement! The apparent disagreement found in [10] was due to the fact that the contribution of the B 2 terms to the zero point energy. For this choice. the Born–Infeld spectrum can be calculated from the corresponding Yang–Mills spectrum. the Born–Infeld spectrum does not match the spectrum obtained from a string theoretic calculation. It seems impossible to express the integrals in Eqs. where f1 6= f2 . BPS configurations on T 6 From the previous examples. again we can perform the integrals by Taylor expanding all the coefficients till a certain order around fi = 0. Consider two D3-branes on T 6 aligned in the 146-hyperplane.4) into a pure Yang–Mills form. in this subsection. M 2 = 2(2m1 + 1)f1 + 2(2m2 − 1)f2 . However. (3. (5.20)–(3. We rotate one of them over an angle γ1 into the 12-plane.

. this would imply that the non-abelian Born–Infeld theory is not able to correctly describe BPS states on T 6 . To see this. where Λ = {(i1 L1 .e. (6. i. with arbitrary wrapping structure. ip Lp )|i1 . A sublattice Σ is specified by a basis {si }i with si = mij ej . Classification of brane wrappings Consider an arbitrary p-brane wrapped (without branch points) around the torus T p = Rp /Λ. the condition under which a configuration of branes at angles preserves some supersymmetry was derived. . .F. The classification of N -fold wrappings is thus equivalent to the classification of sublattices of rank N . but it might be useful to have in mind some concrete examples.. . Different possible sublattices correspond to different possible wrappings. discuss the phenomenon of fractional momentum quantization. like the class of D2-brane wrappings on T 2 studied in [10]. ip ∈ Z}. can be identified with Rp /Σ. this is f1 = f2 + f3 . In [15]. We will study the most general case here. considered as a covering of the torus. and explain how this can be understood physically in terms of the string picture of brane fluctuations. . An SL(2. Multiply wrapped branes In this section. . Bases that differ by an SL(p. This indicates it is indeed the trace prescription that needs to be reconsidered rather than the basic form of the non-abelian Born–Infeld action. where Σ is a sublattice of Λ (see Fig. that is. With the above configuration on T 6 this would correspond to D3-branes at angles satisfying γ1 = γ2 + γ3 . In this single brane case. . The D-brane.1. Z) transformation acts on mij as      a b m11 m12 m11 m12 → . Z) transformation are equivalent as they generate the same lattice Σ. Then N . If this is correct. 1 − f2 f3 (5. 6. . 1 as an example).1) m21 m22 m21 m22 c d . and furthermore the non-abelian Born–Infeld spectrum matches string theory expectations. We give a general construction of the fluctuation spectrum. we consider a single multiply wrapped D-brane on the torus T d . let us first consider the case p = 2. where {ei }i is the standard basis of Λ and mij is an integral matrix with determinant N . the number of elements in the group Λ/Σ. mij = 0 for i < j . This equivalence can be used to transform mij into a lower triangular matrix. In terms of the backgrounds. the number of times the brane covers the torus is equal to the rank of Σ in Λ. the Born–Infeld action scales correctly and does reproduce correct results and this for arbitrary choices of the background fields. Denef et al. the symmetrized trace always reduces to an ordinary trace. / Nuclear Physics B 581 (2000) 135–155 147 This is reinforced by the fact that through quartic order in the fieldstrength.8) 6.

for p > 2. one can also read off the winding numbers `i that appear further on in the momentum quantization condition: for example the off-diagonal mode generated by E16 has `1 = 2. m12 = 0. Similarly. the classification parameters mij introduced in the text are m11 = 3. . parametrized by p winding numbers mi and p(p − 1)/2 “mixing” numbers mij . Z) equivalence. A D2-brane wrapped 6 times around the torus R2 /Λ can be represented as a quotient R2 /Σ with Σ a rank 6 sublattice of Λ. 1. b = −m12 /r with r = gcd(m12 . With the indicated sheet labeling 1. Indeed. c ∈ Z (with the other mij unchanged). Now the residual SL(p. and N = i mi . so the equivalence is fixed by requiring 0 6 m21 < m11 . 6. with all diagonal elements equal to ±1. j with i > j . 7 Note that this “winding number” m is not really canonically defined.e. . the brane has positive orientation). m21 = 1. From this diagram. The latter can be taken to be +1 by putting all mii > 0 (we assume N > 0. The number of times the brane wraps Q the x i direction 7 is given by mi ≡ mii . in the sense that it depends on how one i chooses to gauge fix the SL(p. i > j . m22 = 2. e1 [3] = 1.2) i mi =N} inequivalent N -fold wrappings of T p by a single p-brane.N = {mi : Q (m1 )p−1 (m2 )p−2 · · · mp−1 (6. d such that da − cb = 1 we obtain the desired SL(2. m22 ). Z) transformations are also lower triangular. the closed path e1 acts as the permutation (123)(456) on the sheets (so e1 [1] = 2. Denef et al.148 F. while e2 acts as (143625). etc. So all in all we find X wp. For the lattice shown here. . e1 [2] = 3. / Nuclear Physics B 581 (2000) 135–155 Fig. so in particular m12 → am12 + bm22 . .. . Z) subgroups of SL(p. `2 = 1. the analogous statement can be deduced by repeated application of the p = 2 reasoning. mij can be put in lower triangular form by repeated application of elements of SL(2. The residual equivalence is fixed by requiring 0 6 mij < mjj for all i. and c. Choosing a = m22 /r. Z) transformation putting m12 = 0. i. Z).). for p = 2. For p > 2. the residual equivalence acts on mij as m21 → m21 + c m11 .

which in turn belongs to a certain sheet γ [a]. xp ) = Ω1 φ(0. Moving along a closed loop γ will move this to the point x + x(a) + x(γ ) mod Σ of R/Σ. . Implementation in worldvolume U (N ) gauge theory The low energy dynamics of a Dp-brane wrapped N times around T p is described by a U (N) (Born–Infeld) gauge theory on T p with nontrivial boundary conditions (’t Hooft twists [19]).F. . . x2 . . (6. x2 . Pn . xp ). . that is. . where the Ωi are the index permutation matrices corresponding to the wrapping structure. Suppose γ intersects the coordinate boundary surfaces at the points P1 . . Consider an arbitrary path γ between points Pi and Pf . Then parallel transport of U (N) fundamentals along γ is given by φ → U (γ )φ with " ZPn # " ZP1 # " ZPf # −1 −1 A Ω(Pn−1 ) · · · P exp i A . 0. . Indeed. 0). Denef et al. . The U (N) indices can be identified with the sheets of the covering Rp /Σ → Rp /Λ. identifying the basis element ei of Λ with a closed loop in the x i direction: (Ωi )a b = δei [a].2. Now in the case of zero background field A. . Lp ) = Ωp φ(x1 .b (6. starting on a sheet a and moving along a closed path γ in the torus. φ(x1 . x2 . the wrapping structure can be implemented by the following boundary conditions on fields in the fundamental of U (N): φ(L1 .5) . x2 . . In general. . depending on the structure of the wrapping. which as discussed above is given by a sublattice Σ of Λ. L2 . xp ). . / Nuclear Physics B 581 (2000) 135–155 149 6. . Conversely. . . . . . . xp ) = Ω2 φ(x1 . . one ends up on a different sheet γ [a]. . uniquely determined by Σ (and γ of course). a point on sheet a of the D-brane with coordinate x on Rp /Λ is represented by a unique point x + x(a) of the covering space Rp /Σ. In the case of vanishing background fields. . . . . the implementation of the wrapping structure in the gauge theory can be described in terms of U as U (γ )a b = δγ [a]. . .. Note that by construction U (γ2 ◦ γ1 ) = U (γ2 )U (γ1 ).b (6.4) U (γ ) ≡ P exp i A Ω(Pn ) P exp i Pn Pn−1 Pi where the path ordered exponentials are over the indicated parts of γ and the Ω(Pi ) are the appropriate U (N) transition functions at the intersection points. Σ is uniquely determined by the permutations a → γ [a]: it is the lattice of points in the covering space Rp reached by paths σ that start from x = 0 and for which σ [a] = a for all sheets a. The sheet permutation a → γ [a] is (up to sheet relabeling) completely determined by Σ. On objects X transforming in the adjoint. . U acts as X → U XU −1 .3) An equivalent and more invariant description is in terms of parallel transport. . φ(x1 .

changing (6.9) and for two closed paths γ1 . Note also that parallel transport of objects transforming in the adjoint is invariant under continuous path deformations. it is invariant (up to conjugation) under such transformations. In this paper we only consider backgrounds with diagonal and (covariant) constant field strength F .5) in U (γ )a b = δγ [a]. .b . / Nuclear Physics B 581 (2000) 135–155 for any closed path γ . we have U (σ )a b = δa.γ2 ) γ2 Since U (σ ) is diagonal for σ ∈ Σ (cf.1. (6.10) where S(γ1 . In particular for all σ ∈ Σ (that is paths starting at 0 and ending at a point of Σ when lifted to the covering space Rp ). (6. (6. Denef et al. parallel transport of a diagonal matrix is insensitive to the phase factors: U (γ ) diag(da )a U (γ )−1 = diag(dγ [a] )a .8) Note that in general there exist many different equivalent gauge theory descriptions of the same brane system. More explicitly.12) Fij dx σ1 σ2 Applying this to the basis elements si = mij ej of Σ introduced in Section 6.13) m−1 ik m−1 j l qkl Fij = Li Lj where the qkl = −qlk are integers. σ2 ∈ Σ we must have Z Z i dx j ∈ 2πZ. (6. Such transformations in general change both background fields and boundary conditions. γ2 with equal base points: U (γ2 )U (γ1 ) = e i R S(γ1 . (6.7)). γ2 ) is the surface swept out by contracting γ2 ◦ γ1 ◦ γ2−1 ◦ γ1−1 . related by gauge transformations g(x) (not necessarily satisfying the boundary conditions).10) implies that for σ1 . sweeping out the surface S when contracted to a point.11) γ1 S(γ1 . we have U (C) = ei R S F IN . Eq. position dependent).7) However.150 F.γ2 ) F U (γ1 )U (γ2 ). (6. one has Z Z Z i F = Fij dx dx j . yields the flux quantization condition   2π (6.6) When diagonal background fields are switched on. (6. unlike boundary conditions and background fields.b eiφb (γ ) . This together with (6. (6. We introduced the description in terms of U (γ ) here because. the relation (6. U (γ ) and the boundary conditions will in general pick up additional diagonal components (possibly path resp. Then for a closed contractible path C.8) and the fact that we are considering a single wrapped brane implies that F is proportional to the unit matrix IN : F = F IN .

so γab [a] = b. b = 1. runs from 0 lifted to sheet a to 0 lifted to sheet b. automatically satisfies condition (6.18) δA0 = lim 0 Λ0 →Λ |Λ0 | 0 0 λ ∈Λ denotes a subset of Λ of size |Λ0 |. hence the notation λ.19) λ ∈Λ with E an arbitrary matrix.F.14) are of the form δA(x) = U (γx ) δA0 U (γx )−1 ei R γx ki dx i . (6. Di Ai = 0.17) is a linear combination of the matrices hEab ik . we have . first note that (6. The fluctuation modes of the quadratic Lagrangian (3. Denef et al. (6. independent of specific background field and boundary condition choices. .7) implies we can write Eab = U (γab )Db with γab a closed path that. Conversely. which clearly shows the physical origin of the modifications to the momentum quantization condition. lifted to the covering space.13). The main difference is the modified quantization condition on the momenta. the derivation of the fluctuation spectrum of the non-abelian Born–Infeld theory in this background is quite analogous to the derivation in the abelian U (1) case discussed in Section 2. Below we give a general construction of the spectrum. in the gauge A0 = 0. a. .17) is satisfied. δA0 ≡ hEik ≡ lim Λ0 →Λ |Λ0 | 0 0 where Λ0 (6. and with Db the diagonal matrix defined by (Db )cc = e−iφb (γab ) δcb . as defined under Eq.16) gives indeed a well defined globally smooth solution. provided δA0 = U (λ) δA0 U (λ)−1 ei R λ ki dx i (6. / Nuclear Physics B 581 (2000) 135–155 151 6. (6. any δA0 defined by R 1 X 0 0 −1 i λ0 ki dx i U (λ ) E U (λ ) e .14) with energy M 2 = G ij ki kj (6. N . (3.4) as usual. (6. . Because U acting in the adjoint is invariant under continuous path deformations.18) under the shift λ0 → λ ◦ λ0 . Then for σ ∈ Σ.9). To see this. Now hEab ik is only nonzero for specific values of the momentum k. Note that if condition (6. as follows from U (λ)U (λ0 ) = U (λ ◦ λ0 ) and invariance up to a phase of (6.16) where γx is any path from 0 to x.17). Solutions to (6. this modification is nontrivial [10]. with the matrix basis Eab . .17) for any closed path λ based at 0.15) where G is given by (2. So the general solution to (6.3. Such paths can be identified with the points of the lattice Λ. one trivially has R 1 X 0 0 −1 i λ0 ki dx i U (λ ) δA U (λ ) e . are of the form δAi = ui δA with k i ui = 0 and δA a solution of Dj δA = ikj δA. Fluctuation spectrum Because the field strength is proportional to the unit matrix. When the brane carries flux.

152 F. So hEab ik can only be nonzero if the following momentum quantization condition is satisfied: Z Z i ki dx + F ∈ 2πZ (6.13).g. e.20) more explicit. and the quantization condition is shifted. and (6.σ ) for all σ ∈ Σ. see.σ ) # F S(γab . Li (6. So here the non-abelian Born–Infeld theory matches nicely with string theory expectations. Let us make (6. / Nuclear Physics B 581 (2000) 135–155 hEab ik = U (γab )Db k " Z # i = exp i ki dx U (σ )U (γab ) Db U (σ )−1 k σ " Z #* " = exp i ki dx i exp i σ " Z = exp i ki dx i + i σ + F U (γab )U (σ ) Db U (σ )−1 k S(γab . this additional term is easily understood: it is nothing but the usual coupling of a string running from sheet a to sheet b and going around a loop σ of the brane.. where the off-diagonal fluctuations correspond to open strings with endpoints on different sheets. as in [10]. Denef et al.15) can be understood directly in the string theory picture. . to the given gauge field background.σ ) Equation (6.20) σ S(γab .σ ) " Z = exp i ki dx i + i σ Z # Z # Z F U (γab ) Db k Eab k . However. qj l = −qlj are the flux quantum numbers from (6. Also the effective metric G appearing in (6. though we did not work out the details for the general case.10) was used in going from the second to the third line. off-diagonal excitations pick up an additional phase when going around a loop. For diagonal excitations. and in particular we expect the fluctuation spectrum to be reproduced by string theory on an appropriate T-dual system.11) and (6. the flux term is zero and the quantization condition is simply the usual one for a particle on a compact space — here the covering space Rp /Σ corresponding to the wrapped brane. S(γab . Using (6.8) together with σ [b] = b to go from the third line to the fourth. and taking σ equal to the basis elements si = mij ej introduced in Section 6. In the string picture of these excitations.21) where the nj are arbitrary integers.13). and `k is the winding number of γab in the x k direction: γab = lk ek .1. the quantization condition can be rewritten as ki =   2π −1   m ij nj + m−1 kl qj l `k . but the minimal momentum quantum will never be less than 1/N . [20]. The precise fractionalization of momentum depends on the details of the different integers involved. as could be expected.

It also shows that all modes constructed from hEab ik are nonzero. 7. the first two terms of the Born– Infeld action. because all terms in the finite sum are linearly independent matrices. As we demonstrated in this paper. Dβ }Fγ δ − [Fαβ . the vertices appearing are rather simple. Conclusions The calculation of the mass spectrum is probably the simplest question which can be addressed by means of the effective action. So the quantum numbers ni and `i appearing in (6. the non-abelian Born– Infeld action defined using the symmetrized trace prescription is not able to reproduce the spectrum as predicted by string theory. Keeping in mind that the non-abelian Born–Infeld action should reduce to the sum of abelian Born–Infeld actions for well separated D-branes. However at order F 6 and higher. are ignored in the effective action. As advocated by Tseytlin.F. the infinite sum in (6.. the calculation has to be pushed to an order which is probably unfeasible without the aid of a computer. Fγ δ ]. The present work clearly demonstrates the need to address this problem. One possible way would be to construct these terms using string field theory along the lines developed in [21]. The anti-symmetrized products however should be kept and will contribute to the mass spectrum! Precisely these terms are not captured by the proposal in [9]. give the same contribution.19) can be reduced to a finite sum over Λ/Σ: R 1 X i U (λ0 ) Eab U (λ0 )−1 ei λ0 ki dx . the same `i ) are actually equal up to a phase factor. Not surprisingly. Another way to get the F 6 terms would be by calculating a five loop beta function. under the assumption that the velocities vary slowly. give correct answers. This gives an in because paths principle straightforward way to compute the hEab ik explicitly for a given U (N) bundle. Despite of this. the F 2 and F 4 terms.1) Dα Dβ Fγ δ = {Dα . This corresponds to the fact that only the relative position (in the covering space) of the sheets is physical. as they are mapped to each other by a path shift λ0 → λ0 ◦ γ for a certain γ . However this is ambiguous in a non-abelian gauge theory as one has that 1 i (7. It only probes the effective action through second order in the fluctuations. with σ ∈ Σ. the calculation . At higher order one expects corrections going as derivatives acting on the fieldstrength [18] which. this can be understood as follows. However in order to make concrete statements about the F 6 terms.22) hEab ik = N 0 λ ∈Λ/Σ λ0 and λ0 ◦ σ .e. (6. things go wrong. Due to the fact that for the present purpose it is sufficient to consider trivial gravitational backgrounds.21) label the modes without degeneracy. additional corrections should involve commutators of the field strengths. Finally note that all hEab ik with the same γab (i. Denef et al. 2 2 It is clear that the symmetrized product of derivatives acting on a fieldstrength should be viewed as an acceleration term which can safely be neglected. / Nuclear Physics B 581 (2000) 135–155 153 Note that if the momentum quantization condition is satisfied.

We also know that the Born–Infeld spectrum should have the same form as the Yang– Mills spectrum. . such as the precise implications of T-duality transformations and D-brane configurations on more involved geometries. which deserve further investigation. requiring a consistent behaviour of the Born–Infeld action under T-duality further reduces the possibilities. focussing on T 6 we saw that at the level of the Yang– Mills theory the BPS relation is given by f1 − f2 − f3 = 0 while in the Born–Infeld theory it becomes f1 − f2 − f3 = (2πα 0 )2 f1 f2 f3 . The initial exploration of general wrappings of D-branes on tori. Perhaps the simplest way to proceed is by using the mass spectrum as a guideline. this puts strong additional constraints on the possible modifications. In [12].S. It might be that the requirement of invariance under κsymmetry favorizes certain ordenings. / Nuclear Physics B 581 (2000) 135–155 remains very involved. As became clear from Section 5. (7. The two-torus. In fact a closer analysis shows that there are only 15 possibilities. A. Additional hints are provided by the BPS configurations on T 4 for which Tseytlins prescription does provide the correct spectrum. only 5 inequivalent ordenings remain. as the mass spectrum automatically narrows the group to a U (2) factor. On T 6 and T 8 further insights are provided by the fact that in the Yang–Mills case BPS configurations involve linear relations between the backgrounds [14]. Yet another approach could consist of supersymmetrizing the non-abelian Born–Infeld action. Acknowledgements We would like to thank Walter Troost for several illuminating discussions and Alberto Santambrogio for a useful comment. ignoring the Lorentz indices all together. are supported in part by the FWO and by the European Commission TMR programme ERBFMRX-CT96-0045 in which both are associated to K.T. However. while for the Born–Infeld theory the backgrounds should satisfy non-linear relations [15].U. Now. is the best place to start the analysis. the number of truly inequivalent permutations is considerably smaller. A systematic study of modifications of the non-abelian Born–Infeld action satisfying all requirements will be presented in a separate publication [22]. and J. A priori we expect 5! different permutations of the Lie algebraic factors to be relevant. Finally. being a particularly simple case. many more examples are available. opens several interesting questions.2) We reinserted the factors of 2πα 0 in order to clearly demonstrate that this indeed relates different orders in F in the Born–Infeld theory. a first attempt was already made but it was clear that the T 4 case was not sufficient to unambiguously determine the F 6 term. We hope to return to this in the future. Indeed. one would have to make a systematic analysis of possible modifications of the F 6 terms in the non-abelian Born–Infeld action. As a first step. Denef et al.154 F. but with rescaled backgrounds. Furthermore. Leuven.

[13] P. S. Phys. W. Nucl. Taylor. Douglas. Commun. A. Phys. Constant field strengths on T 2n . 94 (1984) 397. Nucl. hep-th/9712160. B 280 (1987) 599. Symp. / Nuclear Physics B 581 (2000) 135–155 155 References [1] W. Nucl. [17] M. Tseytlin. Open Superstring in Non-Abelian Gauge Field. Troost. [4] E. Phys. Born–Infeld action. M. [15] M. Tseytlin. supersymmetry and string theory. [18] O. Phys. B 499 (1997) 241. J. p. Witten. R. hep-th/9908105. W. Taylor. Andreev. Phys.R. hep-th/9708089. Witten. hep-th/9801182. M. [2] J. [21] W. hep-th/9703217. hep-th/9606139. [3] A. [12] P. Phys.D.J. Costa. Troost. Humboldt-Universität zu Berlin. B 277 (1986) 1. Phys. TASI lectures on D-branes. Math. June 1997. 1989. Akademie der Wissenschaften der DDR. van Baal. 174. [20] N. B 524 (1998) 333. Nucl. B. Abouelsaood. Leigh. Phys. 85 (1982) 529. lectures presented at the Trieste summer school on particle physics and cosmology. [11] Z. B 460 (1996) 35. Ramgoolam. D-brane effective field theory from string field theory. Phys. hep-th/9702099. Untersuchung der WeylInvarianz im Verallgemeinter σ -Modell für Offene Strings. to appear in the Yuri Golfand memorial volume. Taylor. Bain. Polchinski. B 291 (1987) 876. M.A. Nucl. [7] D. B 520 (1998) 205. [9] A. Guralnik. Phys. Nappi. to appear in the proceedings of the Cargèse ’99 Summer School. hep-th/9909187. hep-th/9712026. 81 (1981) 267. [5] D. [16] A. Commun. Yost. E. Phys. [14] J. Nucl. Lectures on D-Branes. hep-th/9701125. Berkooz. of the XXIII Int. Seiberg. Troost. Nucl. hep-th/9801127. Nucl. Gauge Theory and M(atrices). Phys.A. Nucl. Nucl. [19] G. Phys. Nucl. Witten. work in progress..F. JHEP 9909 (1999) 032. Perry. Nucl. Tseytlin. 1990. Math. C. Phys. hep-th/0001201. . On the Non-Abelian Born–Infeld Action. Brecher. Gross. Math.A. Hashimoto. [8] K. B 311 (1988) 205. B 276 (1986) 391. Tseytlin. E. Callan. Nucl. B 521 (1998) 129.A. [10] A. S. in: Proc. [6] A. ’t Hooft. Perry. Commun. B 503 (1997) 193. hep-th/9908142. Ahrenshoop. B 501 (1997) 41. B 480 (1996) 265. Phys. hep/th9611050. Sevrin. Phys. PhD Thesis. to appear in Nucl. hep-th/9510135. Behrndt. hep-th/9909154. Nucl. C.J. Denef et al. [22] A. Phys. B 527 (1998) 121. Phys.

Canada Received 29 March 2000. which confine open strings to their world-volume. D3-branes. 100 Sussex Drive. obtained via F-theory compactified on Calabi–Yau fourfolds. Herman Verlinde a a Joseph Henry Laboratories. this proposal seems like a rather drastic departure from the more conventional Kaluza–Klein framework.4 . Introduction This note is intended to clarify the realization and interpretation of the Randall–Sundrum compactification scenario within string theory. A concrete theoretical realization of such world-branes are the D3-branes of IIB string theory.elsevier. accepted 3 May 2000 Abstract We give a short review of a large class of warped string geometries. Chan a . the physics of these more general scenarios is still largely unexplored.  2000 Elsevier Science B. A second important ingredient of the RS scenario is that part or all of the observable matter may be thought of as confined to a 4d submanifold of the higher-dimensional spacetime. Ottawa. the four uncompactified directions and the compact manifold are assumed to form a simple direct product. All rights reserved.Nuclear Physics B 581 (2000) 156–164 www. USA b National Research Council.V. In the model of [1]. in most works on string compactifications thus far. As a result. Paul b . PII: S 0 5 5 0 . Although it was realized for a long time that this basic KK set-up can be generalized to include the possibility of warped products. however. our 4d world is extended with an extra direction r to a 5d spacetime with the warped metric ds 2 = e2σ (r) ηµν dx µ dx ν + dr 2 (1) with σ (r) = −k|r|.3 2 1 3 ( 0 0 ) 0 0 2 6 7 . Princeton University. do not bind 4d gravity. All rights reserved.nl/locate/npe A note on warped string compactification Chang S. the warped form of the metric ensures that the effective volume of the extra direction is finite. that upon reduction to 5 dimensions give consistent supersymmetric realizations of the RS compactification scenario. Princeton.V. Ontario K1A OR6. Possible supersymmetric realizations of the Planck-brane. Percy L. Indeed. matter particles sufficiently close to the domain wall region near r = 0 will experience ordinary 4d gravity at long distances [1]. located 0550-3213/00/$ – see front matter  2000 Elsevier Science B. NJ 08544. 1. At first sight. Even while the range of r is infinite.

Chan et al. it is an elaborate task to derive the dimensionally reduced theory. etc. It will therefore be much more general — and also easier — to first (a) identify a general class of warped compactifications of the 10-dimensional theory. Since the scalar fields φ a arise as moduli of some internal 5d compact space K5 .. It is still an open problem to find a direct construction of these geometries via the other route. while it seems feasible to classify the possible types of supersymmetric solutions for each of these special dimensional reductions. Given the current interest in the subject. any domain wall solution in 5d gauged supergravity describes (upon lifting it back up to 10 dimensions) some specific warped compactification of the 10d theory. . Various attempts have been made to find smooth domain wall solutions of this type within 5d gauged supergravity. Instead of following the above procedure of (1) performing some special dimensional reduction to 5 dimensions and (2) looking for RS domain wall type solutions. / Nuclear Physics B 581 (2000) 156–164 157 Fig. such a class of warped IIB geometries indeed exists in the form of quite generic F-theory compactifications on Calabi–Yau four-folds.Ch.3]. However. There are several reasons for why this is indeed a hard problem. and then (b) perform the same type of dimensional reduction from 10 to 5 dimensions. there is no guarantee that they provide a general enough framework. around r = 0 in (1). 1. Thus far this has been done only for reductions over rather special symmetric 5-manifolds K5 such as S 5 or S 5 /Z2 . are therefore rather expected to be found in the form of domain wall type configurations. As will be described below. In the end. one can then hope to identify a class of 10d geometries for which the resulting dimensionally reduced solution has all the required properties. 1 These have been studied in some detail in the recent literature — a list of references include [5–9] — and indeed none of our equations will be new. Even for a given compactification from 10 dimensions. or possible stringy generalizations thereof. and/or for special theories with extended supersymmetry. 1 Another special realisation of an RS geometry in terms of a toroidal type IIB orientifold compactification has been described in [4]. To identify supersymmetric RS type geometries. it seems more practical to reverse the two steps. we will follow the route IIB → (a) → (b) → RS.S. but thus far without real success [2.

Chan et al. due to the special geometric properties of the Calabi–Yau four-fold K8 — vanishing first Chern class and SU(4) holonomy — the associated IIB background by construction will preserve 4d supersymmetry. rather it gives an economical way to characterize the compactification geometry as well as the expectation values of other fields. where the elliptic fiber degenerates. we will here only present the general form of the compactification geometry without any proof that it is really a supersymmetric solution to the 10d equations of motion. in particular from the very clear discussion by Becker and Becker [5]. It can however be taken to change its shape when moving along the base K6 . takes the form 2 = e2α(y)ηµν dx µ dx ν + e−2α(y) gMN (y) dy M dy N . and furthermore that they are in fact quite generic. Warped compactification in F-theory F-theory is a geometric language for describing compactifications of type IIB string theory. The four-fold K8 is not the actual compactification geometry. Compactifications of F-theory down to four dimensions are specified by means of a Calabi–Yau four-fold that admit an elliptic fibration with a section. We will nonetheless attempt to make this 5d perspective as transparent as possible. An explicit example of this Tduality transformation is discussed in [6]. as well as on other data such as the possible non-zero expectation values of other fields and the locations of the possible D-branes. by direct translation of the M-theory analysis of [5]. It can however be straightforwardly translated to the F-theory context by performing the T-duality transformation outlined in [7]. Since all derivations are contained in existing papers.S. it can have nontrivial monodromies around singular co-dimension 2 loci inside the K6 . 2. at least at the classical and perturbative level. It is therefore not easy to explicitly perform the dimensional reduction of these solutions to 5 dimensions. Moreover. in which the expectation values of the dilaton and axion fields are allowed to vary nontrivially along the compactification manifold [10].158 Ch. / Nuclear Physics B 581 (2000) 156–164 however. The warped geometry of this type of F-theory compactifications has been derived in [6]. it seems useful to collect some of the known facts about these compactifications. In particular we will show that they indeed give rise to a 5d metric of the generic form (1). This proof can however be quite directly extracted from the literature. The full solution for the 10-dimensional IIB string metric. In other words. dsIIB (2) where gMN denotes the metric on the 6-dimensional base-manifold K6 . The shape of the warp factor e2α will depend on the detailed geometry of the CY four-fold K8 . since it has not been generally appreciated that supersymmetric RS geometries indeed exist in string theory. Although the 10d perspective will allow us to identify a large class of RS-type compactification geometries. Their analysis was done in the context of M-theory compactifications on CY four-folds. . The two-torus will be taken to shrink to zero size. their geometrical structure is rather involved. these are 8-dimensional compact manifolds K8 that locally look like a product of a complex three-fold K6 times a two-torus T 2 . In particular. in the string frame.

These 4d submanifolds are associated with the locations of D7-branes. of which the remaining 3 + 1 dimensions span the uncompactified spacetime directions. H NS and H R . of the IIB supergravity into a single four-form field-strength G on K8 as follows [9]. the shape of the two-torus inside the K8 . (iii) the NSNS 3-form field strength H NS . the fields that can take nontrivial expectation values are the following: (i) the dilaton field φ. Z) monodromy τ→ aτ + b . it will be convenient to combine the NSNS and RR three-from field strengths. τ =φ (3) As mentioned above. however. The value of this charge is proportional to the Euler characteristic χ(K8 ) of the original Calabi–Yau four-fold K8 . the modulus field τ can pick up an SL(2. describes the variation along the 6d base-manifold K6 of the dilaton and e. an effective total D3-brane charge.Ch. The full non-perturbative string theory. For supersymmetric configurations. Here χ(K8 ) is defined via . but instead are multivalued with branch cut singularities at the locations of the D7-branes. a key feature of F-theory is that this modulus in general has nontrivial monodromies around 4d submanifolds inside K6 . The fieldstrength G is invariant under these transformations. the modulus τ of the elliptic fibration. An important aspect of F-theory compactifications is that they typically carry. (v) the RR 5-form field strength F (5) . (5) G= iτ2 H¯ = H R − τ¯ H NS . We thus notice that the dilaton and axion field are not smooth single-valued functions. (iv) the RR 3-form field strength H R. e (ii) the RR-scalar or axion field φ.S. For the following. but nonetheless via (3) amounts to a nontrivial duality transformation of the IIB string theory. H defines an integral harmonic (1. 2)-form on K6 satisfying H ∧ k = 0 with k the Kähler class of K6 [9]. cτ + d (4) which leaves the geometric shape of the two-torus fibre inside K8 invariant. Chan et al. via the identification axion fields. (6) H = H R − τ H NS . Then we can write π (H ∧ d z¯ − H¯ ∧ dz). φ and φ e + ie−φ . It transforms under the SL(2. In going around one of the D7-branes. In the type IIB theory. is expected to be well-behaved everywhere. via their nontrivial topology. Z) monodromy transformations (4) around the seven-branes as H → H /(cτ + d). The expectation values of all these fields can all be conveniently characterized in terms of the geometry of K8 . / Nuclear Physics B 581 (2000) 156–164 159 Besides the ten-dimensional spacetime metric. Let z and z¯ denote the coordinates along the T 2 fiber.

160 Ch. The Euler number χ(K8 ) can be non-zero only if K6 has a non-vanishing first Chern class. (8) 192 4 with R the curvature two-form on K8 . or conservation of the RR 5-form flux. mentioned in [8]. Shape of the warp factor Let us summarize. Alternatively. Starting from the elliptically fibered CY four-fold K8 we can extract a complete characterization of the warped compactification. (7) K8 where   1 1 tr R 4 − (tr R 2 )2 . and then solve the reduced equation to obtain the function e−4α directly on K6 . following [6]. Finally. The number of such D3-branes is therefore not free. The points y = yi where correspond to the location of the N D3-branes. (9) N = χ(K8 ) − 24 8π 2 I8 (R) = K8 Depending on the topology of K8 . provided the F-theory compactification makes use of a non-zero number of D7branes.6] ∆(8) Fµνλσ M = εµνλσ ∂M e−4α . In this limit. . there is also an nontrivial expectation value for the self-dual RR five-form field strength. that is. An example with N = 972. Here. equal to [5. but that via (9) the total charge adds up to zero. This global tadpole cancellation relation reads Z 1 1 G ∧ G. (11) We note that via (10) the D3-branes indeed form a source for this field strength. / Nuclear Physics B 581 (2000) 156–164 1 χ(K8 ) = 24 Z I8 (R). These other sources come from possible non-zero fluxes of the NSNS or RR two-form fields. since K8 ' K6 × T 2 . requires that this charge must be canceled by other sources. The equation of motion for the warp factor e2α obtained in [5] and [6] reads as follows ) ( N X 1 (8) −4α 2 (8) = ∗4π I8 (R) − G∧G− δ (y − yi ) . 3. Chan et al. or from the explicit insertion of N D3-brane world branes. but completely determined via charge conservation. via integration over the T 2 fiber. one may also first reduce the right-hand side to K6 . Global tadpole cancellation. First. using the analysis of [8]. D3-branes that span the (3 + 1)d uncompactified world but are localized as point-like objects inside the K6 . N can reach values of up to 103 or larger. that is.S. is provided by an elliptically fibered CY four-fold over P3 . (10) ∆ e 8π 2 i=1 denotes the Laplacian and ∗ the Hodge star on K8 . even though in F-theory the elliptic fiber T 2 inside K8 has been shrunk to zero size. we have written the equation on the full 8d manifold K8 . the solution for α obtained via (10) only depends on the 6 coordinates y M on K6 .

this geometry describes a one-sided RS domain wall solution.S. The term e−4α0 parametrizes the constant zero mode of e−4α . the warp function α(y) reduces to α(y) ' log |y − y0 | + const. We then deduce the form of the warp factor e2α from (10). The contours with constant warp factor e2σ define a particular slicing of the 6-dimensional compactification manifold K6 . Sethi for bringing this feature to our attention. / Nuclear Physics B 581 (2000) 156–164 161 Fig. 2.Ch. the rescaling by e−2α of gMN in (2) may have a drastic effect on the shape of the compactification manifold. which can be used to represent K6 as a one parameter flow along r of five-manifolds K5 . we obtain the actual compactification geometry. We may formally solve the equation (10) via Z √ e−4α(y) = e−4α0 + 4π 2 d 8 y 0 gG(y. This implies that the warped 6-geometry automatically has a minimal volume. In particular. it is possible that near the locations of the D3-branes one of the internal directions may become non-compact. Finally from (2). which is not fixed by Eq. say y = y0 . 2. the compactification geometry (2) still gives rise to a 4d Einstein action with a finite 4d Newton constant 1/`24 equal to 2 We thank S. as well as the dilaton and axion via (3). (See Fig. y 0 ) denotes the Green function for ∆(8) . since the second term on the rhs of (12) can become negative. (13) Via (2) this describes the familiar semi-infinite near-horizon geometry of N D3-branes: √ AdS5 × S5 with radius R = 4 4πNgs . Upon dimensional reduction to 5 dimension. (12) × I8 R(y ) − 8π 2 i=1 where G(y. 2. Note that for e−4α to be everywhere positive. which incorporates the complete backreaction due to the G-flux and D3-branes. Close to this point. (10). as indicated in Fig. Chan et al. this constant e−4α0 can not be arbitrarily small. we obtain the metric gMN on the base K6 . Note that. y 0 ) # " N X  1 0 (8) 0 G∧G− δ (y − yi ) . . 2 An interesting limiting case is when all D3-branes are concentrated in one point.) Although the radial AdS5 coordinate r ' −R log |y − y0 | runs over semi-infinite range. which indeed may look quite different from that of the original K6 .

such that the 10d metric takes the following form 2 = e2σ (r)ηµν dx µ dx ν + dr 2 + hmn (y. 4.S. [1]) 1 1 = 2 (`4 ) (`5 )3 Z0 dr e2σ (r). Reduction to 5 dimensions We would now like to show that these F-theory compactifications. Chan et al. dsRS (`10 )5 where (16) 2 dsRS (17) K5 2 is the metric in the 5d Einstein Here the prefactor in (17) is chosen such that dsRS frame (where we have taken the 5d Planck length `5 equal to the 10d one). In this way we indeed obtain a solution that from the 5d perspective looks just like an RStype warped geometry. as well as other fields such as the dilaton. reduce to supersymmetric RS domain wall solutions. r) dy m dy n . s. The 5d slices of constant r define 5d submanifolds K5 . Upon dimensional reduction. It therefore produces to a finite 4d Newton constant equal to (cf. which we can take to be r = 0. Eventually. this result coincides with the value (14) found via direct reduction from 10 dimensions. at which the transverse 5-manifold K5 shrinks to zero size. r). (15) where m now runs over 5 values. all reduce to scalar fields that provide the matter multiplets of the 5d . On the other end. upon performing a suitable dimensional reduction to five dimensions. near the ‘equator’ of the K6 . This fact that r takes a maximal value r = 0 implies that the AdS-space is indeed compactified. the metric hmn of the internal K5 . on which 2α(y) + φ(y)/2 = const. in the sense that. To this end we will look for a specific coordinate system y M = (y m . it has a finite volume. / Nuclear Physics B 581 (2000) 156–164 1 1 = 2 (`4 ) (`10 )8 Z √ 2φ−4α ge (14) K6 with `10 the 10d Planck length. this does not correspond to any singularity. somewhere outside the throat region of the AdS-tube. V5 = h. (18) −∞ It is easily verified that via (17) and (16).. the warp factor behaves like e2σ ' e−2|r|/R with R the AdS-radius of the N D3-brane solution.162 Ch. However. close to the D3-branes. 2. dsRS related to the original IIB string metric (2) via the rescaling Z √ 1 2 2 = eφ/2 (V5 )1/4 dsIIB . For large negative r. as is clear from Fig. there is a boundary value for the coordinate r. relative to the 4d spacetime. the warp factor e2σ (r) reaches some maximal value. this 2 just depends on r and not on correspondence guarantees that the warp-factor e2σ in dsRS the remaining y m. but just to a smooth cap closing off the 6d manifold K6 .

However. this RS localization of gravity is not a new phenomenon.S. An RS domain wall in between two AdS-type regions can be obtained by starting with a Z2 symmetric 6-manifold K6 . From the higher-dimensional viewpoint. the 4d graviton indeed looks like some bound state. 3. Due to the presence of the D7-branes in the F-theory geometry. Kaluza–Klein excitations. however. Our description of the solutions makes clear that the internal structure of the RS Planckbrane is not that of an actual brane. the holographic AdS/CFT correspondence [11] tells us that they are in fact localized excitations of the low energy gauge theory. where the warp factor e2σ is maximal. Finally. the solutions are not completely smooth: the dilaton and axion field have isolated branch cut singularities at the D7-brane loci around which τ is multivalued. and since this radial flow is supersymmetric. 4. While these particles naively look like new degrees of freedom arising from the presence of extra dimensions. because its wave-function is sharply peaked near the wall region. due to the rather complex geometrical structure of the typical F-theory compactification. The best way to understand this instability is via the RG language: in general there will exist relevant operators whose .1. we need to emphasize that the solutions as given here are generically unstable against small perturbations. in which the D3-branes are located at opposite image points under the Z2 . it should in principle be described as some gradient flow driven by some appropriate superpotential.Ch. Discussion In this note we have summarized the geometrical description of warped F-theory compactifications. the two usually separate stages of dimensional and low energy reduction should now be combined into one single procedure. as a closely related point. when localized far inside the AdS region can describe particles with masses much smaller than the inverse size of the original K6 . and shown they can be used to obtain geometries very analogous to the RS scenario. Consequently. the localized graviton zero mode of [1] is just the standard KK zero mode of the 10d metric. it unfortunately seems impossibly hard to find the explicit form of this potential. however. Hence via the holographic identification of the RG scale with a real extra dimension. The same holds for string excited modes in this region. is welldefined in this background. but rather that of a compactification geometry (which however also contains the D7-branes). is that there is no clear distinction between the low energy and extra-dimensional physics. Chan et al. The most interesting aspect of these warped string compactifications. The string theory. / Nuclear Physics B 581 (2000) 156–164 163 Fig. supergravity. All these fields vary with the radial coordinate r.

CFT’s from Calabi–Yau four-folds. hep-th/9606122. / Nuclear Physics B 581 (2000) 156–164 couplings. this means that the dual 4d field theory should be made approximately conformal invariant over a large range of scales.C. E. In RG language.164 Ch. We would like to thank M. M-theory on eight-manifolds. orientifolds and G-flux. Kallosh. Verlinde. hep-th/9602022. Vafa. will quickly grow and typically produce some singularity that effectively closes off the AdS-tube [12]. Lu. [11] J. R. C. . This requirement is special. 2 (1998) 231. Dijkgraaf. Nucl. Phys. Evidence for F-theory. once turned on in the UV region. separating the Planck scale from the scale set by the nontrivial gauge dynamics. the problem of realizing an RS geometry in string theory therefore is reduced back to the original problem it was designed to solve. and we should allow for deformations that may spoil this property. hep-th/9906064. Verlinde for helpful discussions. G. Supersymmetry and the brane world. Polchinski. hep-th/9711200. [3] M. [4] H. Gremm. Rev. hepth/9603150. namely how to generate a large gauge hierarchy. B 480 (1996) 213–224. Rajesh. Randall. [9] S. Lett. Pope. S. C. we did not immediately notice this instability. is supported in part by an NSF graduate research fellowship. [6] K. B 471 (1996) 195–216. 83 (1999) 3370. Math. hep-ph/9905221. H. Maldacena.N. Becker. The string dual of a confining four-dimensional gauge theory. hepth/0002054. S. Rev. Sethi. hep-th/9605053. Vafa. hepth/9908088. in searching for realistic string compactification scenarios. Phys. hep-th/9906070.S. [7] E. Witten. Or stated in more positive terms. Kallosh. Witten. M. Linde. A large mass hierarchy from a small extra dimension. A. Adv. An alternative to compactification. In our way of obtaining the solutions. Cvetic. Cvetic. Nucl. Randall. [8] S. [10] C. Lett. References [1] L. we either need to fine-tune the UV initial conditions or introduce a symmetry that eliminates these unstable modes. [2] R. Holography and compactification. [12] J. Strassler. because we required that the original K6 geometry and all other fields. Becker. Witten. Sethi and E. Gukov. hepth/0003136. In order to ensure that the perturbed geometry contains a substantial intermediate AdSlike region. Sethi. Vafa. Localised gravity in the singular domain wall background?. are smooth at the locations of the D3-branes. M. Chan et al. the observed gauge hierarchy can be viewed as an indication that warped geometries of this type deserve serious attention. however. hep-th/0001071.S. M. Via the AdS/CFT dictionary. Berkooz. Dasgupta. Constraints on low-dimensional string compactifications. L. B 477 (1996). Phys. R. Phys. Phys. 83 (1999) 4690. Phase transitions in M-theory and F-theory. R. E. Nucl. hep-th/9906182. Phys. C. M. C. M-theory. Theor. JHEP 9908 (1999). [5] K. except for the warp-factor and the 5-form flux. Sundrum. Acknowledgements This work is supported by NSF-grant 98-02484.

Brazil Received 17 February 2000. by many scalar fields) in addition to the usual symmetric rank-2 tensor has considerably revived in the recent years.0 . Introduction The assumption that gravity may be propagated by a scalar field (or.elsevier. PII: S 0 5 5 0 .V.1 . If gravity is essentially scalar-tensorial. Xavier Sigaud.2 a Centro Brasileiro de Pesquisas Físicas.br 2 Also at Grupo de Física Teórica. M. there will be direct implications for cosmology and experimental tests of the gravitational interaction ∗ crisnfer@cat. Indeed. Urca CEP: 22290-180. DF. Universidade Católica de Petrópolis. Rio de Janeiro.Nuclear Physics B 581 (2000) 165–178 www. Brazil b Universidade de Brasília. Departamento de Matemática. Brasilia. 98. We compare our results with those obtained in the framework of General Relativity.X.  2000 Elsevier Science B.50. All rights reserved. From the theoretical point of view.3 2 1 3 ( 0 0 ) 0 0 2 8 2 .br 0550-3213/00/$ – see front matter  2000 Elsevier Science B.A. PACS: 04. Campus Universitário – CEP: 70910-900.N. J. Guimarães b.cbpf. helayel@cat. These modifications are induced by an arbitrary coupling of a massless scalar field to the usual tensorial field in the gravitational Lagrangian.E.nl/locate/npe Current-carrying cosmic strings in scalar-tensor gravities C. a scalar-tensor gravity seems to be the most natural alternative to General Relativity.br 1 emilia@mat. more generally. Helayël-Neto a.∗ . most attempts to unify gravity with the other interactions predict the existence of one (or many) scalar(s) field(s) with gravitational-strength couplings. Rua Dr. RJ.80.cbpf. accepted 9 May 2000 Abstract We study the modifications on the metric of an isolated self-gravitating bosonic superconducting cosmic string in a scalar-tensor gravity in the weak-field approximation. We show that this metric depends on five parameters which are related to the string’s internal structure and to the solution of the scalar field.V.C Keywords: Topological defects.unb. All rights reserved. revised 2 May 2000. Scalar-tensor gravities. 150. Ferreira a. Cosmology 1. The metric is derived by means of a matching at the string radius with a most general static and cylindrically symmetric solution of the Einstein–Maxwell-scalar field equations.

In this paper. cosmic strings have been widely studied in cosmology in connection with structure formation. That is. In Section 2. bosonic cosmic string. numerical integrations of the fields equations (Einstein plus material fields) [16]. These theories are metric. charged vaccum condensates [12] and sources of ultrahigh-energy cosmic rays [13]. among others.20]. topological defects are expected to be formed during phase transitions in the early universe. In this context. On the other hand. These modifications come from an arbitrary coupling of a massless scalar field to the tensor field in the gravitational Lagrangian. R Sm denotes the action describing the general matter fields Ψm . at sufficiently high energy scales where gravity becomes scalar“constant” G tensor in nature [2]. some authors have studied solutions for cosmic strings and domain walls in Brans–Dicke [3–5]. In 1985.. we will solve the linearised (modified) Einstein’s equations using distribution’s functions while taking into account the scalar-tensor feature of gravity. For this purpose. we will make an adaptation of Linet’s method [18] to our model. Ferreira et al. we describe the configuration of a superconducting string in scalar- .e g ΦR − gµν . in dilaton theory [6] and in more general scalar-tensor couplings [7]. In General Relativity. we need to solve the modified Einstein’s equations having a current-carrying vortex as source of the spacetime. In particular. e e. Witten showed that in many field theories cosmic strings behave as superconducting tubes and they may generate enormous currents of order 1020 A or more [8]. the gravitational field of superconducting strings has been studied by many authors [15–20]. This fact has raised interest to current-carrying strings and their eventual explanations to many astrophysical phenomena. any gravitational phenomena will be affected by the variation of the gravitational e0 .166 C.g. such as origin of the primordial magnetic fields [9–11]. [14]. the following technics have been employed to derive the spacetime surrounding superconducting vortex: analytic integration of the Einstein’s equations over the string’s energy-momentum tensor [15]. This work is outlined as follows. (1) S= e 16π Φ e is the curvature scalar associated to it and where e gµν is the physical metric in this frame. specially those which were formed in the early universe. it seems worthwhile to analyse the behaviour of matter in the presence of a scalar-tensorial gravitatinal field. In particular. Among them. In Ref. Sen have considered solutions of a superconducting string in the Brans– Dicke theory.N. The action which describes these theories (in the Jordan–Fierz frame) is   Z e µν p   ω(Φ) 1 4 e e e e d x −e e g ∂µ Φ∂ν Φ + Sm Ψm . among others. matter couples minimally and universally to e gµν and not to Φ. The aim of this paper is to study the implications of a class of more general scalar-tensor gravities for a superconducting. such as cosmic strings. Hence. we will be interested on the modifications induced on the string metric and their possible observable consequences on the current carried by the string. / Nuclear Physics B 581 (2000) 165–178 (we refer the reader to Damour’s recent account on “Experimental Tests of GR” [1]). In particular. linearization of the Einstein’s equations using distribution’s functions [18. The main purpose of this paper is to study the influence of a scalar-tensorial coupling on the gravitational field of a current-carrying cosmic string described by Witten’s model [8].

q in much the same way as we proceed with ordinary (non-conducting) cosmic strings. In Section 3.C. we start by solving the equations for the exterior region. In the Subsection 3. We restrict ourselves to contemplate configurations of an isolated and static vortex in the z-axis. Superconducting string configuration in scalar-tensor gravities In what follows. / Nuclear Physics B 581 (2000) 165–178 167 tensor gravities. Then. z). In Subsection 3. (2) 16π 16π with Dµ ϕ ≡ (∂µ + iqCµ )ϕ. we match the exterior solution with the internal parameters. in Section 4. [18]. Cµ ) pair is spontaneously broken. Finally. |σ | = 4 4 2 (3) with positive η. associated to the other U 0 (1) symmetry of this model.N. λϕ parameters. we solve the linearised equations by applying Linet’s method. A vortex configuration arises when the U (1) symmetry associated to the (ϕ. introduced in Ref. Aµ ). λσ . respectively. In a cylindrical coordinate system (t. we will search for a regular solution of a self-gravitating superconducting vortex in the framework of a scalar-tensor gravity. the simplest bosonic vortex arises from the action of the Abelian–Higgs U (1) × U 0 (1) model containing two pairs of complex scalar and gauge fields  Z p 1 µν 1 µν g Dµ ϕDν ϕ ∗ − e g Dµ σ Dν σ ∗ g − e Sm = d4 x −e 2 2   1 µν αβ 1 µν αβ e g e e g e − g Hµα Hνβ − g Fµα Fνβ − V |ϕ|. Ferreira et al. r→∞ (5) . Dµ σ ≡ (∂µ + ieAµ )σ and Fµν and Hµν are the field-strengths associated to the electromagnetic Aµ and gauge Cµ fields. we derive the deficit angle associated to the metric found previously. and lim P (r) = 0. |σ | . f . We also require that these functions be regular everywhere and that they satisfy the usual boundary conditions for a vortex configuration [21] R(0) = 0 lim R(r) = η r→∞ and P (0) = 1. is spontaneously broken in the core of the vortex. θ. Hence. we make the choice  1 (4) ϕ = R(r) eiθ and Cµ = P (r) − 1 δµθ . We also compare our results with previous results obtained in the framework of General Relativity. The functions R. The superconducting feature of this vortex is produced when the pair (σ. V |ϕ|. The potential is “Higgs inspired” and contains appropriate ϕ–σ interactions so that there occurs a spontaneous symmetry breaking 2  λϕ λσ m2 |ϕ|2 − η2 + f |ϕ|2 |σ |2 + |σ |4 − |σ |2 .2. 2. such that r > 0 and 0 6 θ < 2π .3. P are functions of r only. we end with some conclusions and discussions. r.

Outside the string. In order to make our calculations as broad as possible. In the conformal frame. the Einstein equations are modified. R is the curvature scalar associated to it and Ω(φ) is an arbitrary function of the scalar field. for technical reasons. (10) .168 C. (9) and by a redefinition of the quantity e−1 GΩ 2 (φ) = Φ which makes evident the feature that any gravitational phenomena will be affected by the variation of the gravitation “constant” G in the scalar-tensorial gravity. = 2ω(Φ) α ≡ ∂φ which can be interpreted as the (field-dependent) coupling strength between matter and the scalar field. dr dr lim σ (r) = 0 and lim A(r) 6= 0. A straightforward calculus shows that the “Einstein” equations are Gµν = 2∂µ φ∂ν φ − gµν g αβ ∂α φ∂β φ + 8πGTµν . their configuration are taken in the form   ∂ψ z 1 δ . and by introducing a new parameter    ∂ ln Ω(φ) 2  2 e + 3 −1 . we choose not to specify the factors Ω(φ) and α(φ) (the field-dependent coupling strength between matter and the scalar field). Although action (1) shows explicitly this gravity’s scalar-tensorial character.N. and the Aµ is the gauge field which produces an external magnetic field. (8) S= 16πG where gµν is a pure rank-2 tensor in the Einstein frame. leaving them as arbitrary functions of the scalar field. the Aµ field has a nonvanishing component along the z-axis which indicates that there will be a non-vanishing energy-momentum tensor in the region exterior to the string. Ω 2 (φ)gµν . we can see that σ breaks electromagnetism inside the string and can form a charged scalar condensate in the string core. Action (8) is obtained from (1) by a conformal transformation e gµν = Ω 2 (φ)gµν . Ferreira et al. r→∞ r→∞ (7) With this choice. 2g φ = −4πGα(φ)T . / Nuclear Physics B 581 (2000) 165–178 The σ -field is responsible for the bosonic current along the string. Aµ ) is subjected to the following boundary conditions dA(0) dσ (0) = 0 and A(0) = = 0. we choose to work in the conformal (Einstein) frame in which the kinematic terms of the scalar and the tensor fields do not mix Z    √  1 d4 x −g R − 2g µν ∂µ φ∂ν φ + Sm Ψm . (6) σ = σ (r) eiψ(z) and Aµ = A(r) − e ∂z µ The pair (σ.

C. Ferreira et al. 4π e2 β2 4πq 2   e2Ψ 1 Trr = Ω 2 (φ) e2(Ψ −γ ) R 0 2 + σ 0 2 − 2 R 2 P 2 − e−2Ψ σ 2 A2 2 β   02    e2(2Ψ −γ ) P 0 2 A −2 2 + Ω −2 (φ) e−2γ (φ) (φ)V (R. we impose that the metric is static and cylindrically symmetric. We choose to work with a general cylindrically symmetric metric written in the form  (12) ds 2 = e2(γ −Ψ ) −dt 2 + dr 2 + β 2 e−2Ψ dθ 2 + e2Ψ dz2 . Ω −6 (φ)T µν and T Guided by the symmetry of the source. the metric functions satisfy the regularity conditions at the axis of symmetry r = 0 dΨ dβ dγ = 0. σ ) . = 0. (11) Tµν ≡ √ −g δgµν µ but in the conformal frame it is no longer conserved ∇µ Tν = α(φ)T ∇ν φ. + Ω − 2Ω 4π e2 β2 4πq 2   e2Ψ 1 Tθθ = − Ω 2 (φ) e2(Ψ −γ ) R 0 2 + σ 0 2 − 2 R 2 P 2 + e−2Ψ σ 2 A2 2 β   02    e2(2Ψ −γ ) P 0 2 A −2 −2γ −2 2 + Ω (φ) e − Ω (φ) + 2Ω (φ)V (R. β 0 γ 0 = β(Ψ 0 )2 − β(φ 0 )2 + 8πG e2(γ −Ψ ) Trr . / Nuclear Physics B 581 (2000) 165–178 169 We note that the last equation brings a new information and shows that the matter distribution behaves as a source for φ and gµν as well. (11)) are   e2Ψ 1 Ttt = − Ω 2 (φ) e2(Ψ −γ ) R 0 2 + σ 0 2 + 2 R 2 P 2 + e−2Ψ σ 2 A2 2 β   02    e2(2Ψ −γ ) P 0 2 A −2 −2γ −2 2 + Ω (φ) e + Ω (φ) + 2Ω (φ)V (R. σ ) .N. where the metric functions γ . (13) dr dr dr With metric given by expression (12) we are in a position to write the full equations of motion for the self-gravitating superconducting vortex in scalar-tensorial gravity. The energy-momentum tensor is defined as usual 2 δSm . In addition. γ = 0. (βφ 0 )0 = −4πGα(φ)β e2(γ −Ψ ) T . and β are functions of r only. these equations are   β 00 = 8πGβ e2(γ −Ψ ) Ttt + Trr . Ψ = 0. and = 0. In the conformal frame. Ψ. 4π e2 β2 4πq 2 . (14) where (0 ) denotes “derivative with respect to r”.   (βΨ 0 )0 = 4πGβ e2(γ −Ψ ) Ttt + Trr + Tθθ − Tzz . σ ) . It is clear from transformation (9) that we can relate quantities from both frames in such a way that Teµν = eνµ = Ω −4 (φ)Tνµ . The non-vanishing components of the energy-momentum tensor (computed using Eq.

the equation ∇µ Tνµ = α(φ)T ∇ν φ. 4π e2 β2 4πq 2 (15) As we said before. r0 has the same order of magnitude of the string radius. In the next section. Superconducting string solution in scalar-tensor gravities 3. Also.1. 3. we will attempt to solve the field equations (14). where all the fields drop away rapidly and the only survivor is the magnetic field. where all the string’s field contribute to the energy-momentum tensor. Ttt = −Trr = Tθθ = −Tzz = − e−2γ 2 4πe2 Tµµ = 0 and Tνα Tαµ = (17) Thus. 3 Just as a reminder. the electromagnetic field is the only field which contributes to the energy-momentum tensor. The exterior solution and the modified Rainich algebra In this region. That is. Therefore. For the purpose of these calculations. Ferreira et al.170 C. our problem is reduced to solve the modified Einstein’s equations with source given by (16). / Nuclear Physics B 581 (2000) 165–178   e2 Ψ 1 Tzz = − Ω 2 (φ) e2(Ψ −γ ) R 0 2 + σ 2 + 2 R 2 P 2 − e−2Ψ σ 2 A2 2 β   02    e2(2Ψ −γ ) P 0 2 A −2 −2γ −2 2 − Ω (Φ) e + Ω (φ) + 2Ω (φ)V (R. 4 which leads expression (15) to take a simple form  02  1 A . throughout this paper we will work in the conformal frame for the sake of simplicity. Conveniently. we e0 = GΩ 2 (φ0 ). where φ0 is match the exterior and the interior solutions (to first order in G a constant) providing a relationship between the internal parameters of the string and the spacetime geometry. we work in units such that h¯ = c = 1 and keep Newton’s “constant” G.N. r > r0 . where T is the trace of the energy-momentum tensor. Instead. Then. the energy-momentum tensor is not conserved in the conformal frame. the energy-momentum tensor has the form 3   1 1 T µν = F µα Fαν − g µν F αβ Fαβ (16) 4π 4 with the following algebraic properties  1 µ δν Tαβ T αβ . gives an additional relation between the scalar field φ and the source. for convenience. σ ) . . and an interior region r 6 r0 . we can divide the space into two regions: an exterior region r > r0 .

(19) In a scalar-tensor gravity. Rθθ = −Rzz and Rtt = −Rrr − 2(φ 0 )2 e2(Ψ −γ ) . The first and last equations in (18) can be solved straightforwardly: β(r) = Br. thus. we have now: R ≡ Rtt + Rrr + Rθθ + Rzz = 2(φ 0 )2 e2(Ψ −γ ) . Rθθ = Rzz and Rtt = −Rrr . (18) In General Relativity (i.e. The two equations above admit three sets of solutions: the magnetic case. Ψ (r) = n ln(r/r0 ) − ln (1 + κ) (23) . β 0 γ 0 = 8πGβ e2(γ −Ψ ) Trr + β(Ψ 0 )2 − β(φ 0 )2 . Instead of the algebraic conditions stated above. and Rtt 2 = Rrr 2 = Rθθ 2 2 = Rzz . / Nuclear Physics B 581 (2000) 165–178 171 β 00 = 0. (βφ 0 )0 = 0.   (r/r0 )2n + κ . (21) We are now in a position to solve the modified Einstein’s equations. in the absence of the dilaton field). A source of the form (16) leads to some algebraic conditions on the curvature scalar and the Ricci tensor.   (βΨ 0 )0 = 4πGβ e2(γ −Ψ ) Ttt − Tzz . the electric case and a third case which can correspond to either a static electric or a static magnetic field aligned along the z-axis [22].C. φ(r) = l ln(r/r0 ). (22) The second and third equations in (18) are solved with the help of the algebraic conditions (20) and (21): 1 0 γ = 0. r r We.. Ferreira et al. r 1 n2 Ψ 00 + Ψ 0 − Ψ 0 2 = − 2 . we notice however that the Rainich conditions are no longer valid because of the very nature of the modified Einstein’s equations (actually an Einstein– Maxwell-dilaton system). known as the Rainich conditions: R ≡ Rtt + Rrr + Rθθ + Rzz = 0. (20) and the analogous to the magnetic case in the scalar-tensor gravity is a solution of the form: Rtt = Rθθ .N. The superconducting string defined in Witten’s model (2) correponds to the magnetic case: Rtt = Rθθ . find the remaining metric functions: γ 00 + γ (r) = m2 ln(r/r0 ). these equations have been previously investigated by many authors [22].

The internal solution and matching We start by considering the full modified Einstein’s equations (14) with source (15) in the internal region defined by r 6 r0 . in the framework of General Relativity). In this region.N. (25) µ µ (0) in a harmonic coordinate system such that (hν − 12 δν h). metric (24) belongs to a class of metrics corresponding to the Case 1 in Sen’s paper [14]. +) is the Minskowski metric tensor and φ0 is a constant. Therefore. the electromagnetic field being as in Minkowski spacetime. 3. m will be fully determined after the introduction of the matter fields. e0 . (22). with an appropriate adjustment in the parameters. Ferreira et al. In what follows. we proceed with the junction between the internal and external solutions at r = r0 . the solution for the scalar field φ(r) in the exterior region is given by Eq. The integration constants B. n. we will consider the solution for the superconducting e0 . +. (26) Then. the exterior metric is given by:  −2n  2m2   r r W 2 (r) −dt 2 + dr 2 + B 2 r 2 dθ 2 ds 2 = r0 r0  2n 1 r dz2 . the modified Einstein’s equations (10) reduce to the usual linearised Einstein’s equations [23]. +. . φ = φ0 + φ(1) . Tµν is the string’s energye0 = GΩ 2 (φ0 ) (evaluated in flat space) and T (0) its momentum tensor to zeroth-order in G trace. Therefore. l. we also need to solve the linearised equation for the scalar field ∇ 2 φ(1) = −4πGΩ 2(φ0 )α(φ0 )T (0).2. we will briefly recall the method of linearization using distribution functions (presented in Linet’s paper [18] and applied later by Peter and Puy in [20]. (1 + κ) Besides. all fields contribute to the energymomentum tensor. In the particular case of Brans–Dicke. Besides. 4 To linear order in G e0 . we assume that the metric gµν and the scalar field string to linear order in G φ can be written as: gµν = ηµν + hµν .ν = 0. Thus. with both solutions evaluated to linear order in G While doing these calculations.172 C. our problem reduces to solve the linearised Einstein’s equations 4  (0) ∇ 2 hµν = −16πGΩ 2(φ0 ) Tµν − 12 ηµν T (0) . / Nuclear Physics B 581 (2000) 165–178 where the constant n is related to l and m through the expression n2 = l 2 + m2 . + r0 W 2 (r) (24) where W (r) ≡ (r/r0 )2n + κ . where ηµν = diag(−.

2. / Nuclear Physics B 581 (2000) 165–178 173 3. Let us define its components (to zeroth-order in G energy per unit length U : Zr0 U ≡ Mtt = −2π Ttt r dr. − σ 2 A2 − 2 4πq 2   0 2   0 2 2  1 R2 P 2 A P 2 2 − σ A − + 2V . 0 the tension per unit length τ : Zr0 τ ≡ Mzz = −2π Tzz r dr.C.1. 0 and the remaining transversal components as: Zr0 X ≡ Mrr = −2π Trr r dr. in terms of the cartesian components of the energy-momentum tensor we can define a quantity Z such that .N. y. 0 Zr0 Y ≡ Mθθ = −2π Tθθ r dr. Tt = − R + σ + 2 r2 4πe2 4πq 2     1 R2 P 2 A0 2 R0 2 + σ 0 2 − + Tx(0)x = cos2 θ − 2 r2 4πe2  02    P 1 + 2V . z). Ferreira et al. − σ 2 A2 − 2 4πq 2    0 2  1 R2 P 2 A (0)y R0 2 + σ 0 2 − + Ty = sin2 θ − 2 r2 4πe2   02   P 1 + 2V . Let Zr Mνµ (r) ≡ −2π Tνµ (r 0 )r 0 dr 0 0 e0 ) as follows. let us evaluate the superconducting string’s energy-momentum tensor to e0 in cartesian coordinates (t. 0 Now. x. The linearised field equations First of all. we can stablish some linear densities which will be very useful in our further analysis. (27) + Tz(0)z = − R 0 2 + σ 0 2 + 2 r2 4πe2 4πq 2 With the help of the source tensor defined by Thorne [24]. The non-vanishing components of zeroth-order in G the energy-momentum tensor can now be re-written under the form:  02   02    1 02 R2 P 2 A P (0)t 02 2 2 +σ A + + + 2V . The be the source tensor.

one can still write it in terms of distributions. 2πr 2  I2 2x i x j − r 2 δij .N. Though the energy-momentum tensor (29) expresses the string’s energy in the exterior region. = ∇ ln r0 r r0 r4 ij = Tem Therefore. Namely. If we assume that the string is (idealistically) infinetely thin. y. X + Y = 2Z = −r0 2e and. (29) 4 2πr where i. 4π r0 I2 r ∂i ∂j ln . j = x. µ Using the scalar property of Tµ we have that y Trr + Tθθ = Txx + Ty . T(0) 4π r0   ∂i ∂j ln rr0 ij 2 ij . we conclude that Z is proportional to the electric current I . (29) becomes tt zz = Tem = Tem ij =− Tem   I2 2 r 2 ∇ ln . T(0) = U δ(x)δ(y) + 4π r0   I2 2 r 2 zz ∇ ln = −τ δ(x)δ(y) + . (28) Equation (28) represents the string’s energy-momentum tensor with all quantities integrated in the internal region r 6 r0 . T µν = diag(U.174 C. (31) T(0) = I δ δ(x)δ(y) − 2π . We can find easily that: tt zz = Tem = Tem I2 . Ferreira et al. therefore. then its energy-momentum tensor may be described in terms of distribution functions. −Z. After intgrating in the transverse plane of the string we obtain that  0 2 2 A . 2π r0 (30) We are now in a position to calculate the linearised Einstein’s equations (25) with source identified by:   I2 2 r 2 tt ∇ ln . taking into account the relations     2 r r 2 (r 2 δ ij − 2x i x j ) 2 = 2 and ∂i ∂j ln . −τ )δ(x)δ(y). in the cartesian coordinate system. −Z. / Nuclear Physics B 581 (2000) 165–178 Z Z=− Z r dr dθ Txx =− y r dr dθ Ty . Let us now evaluate e0 in cartesian cothe electromagnetic energy-momentum tensor (16) to zeroth-order in G ordinates.

since the linearised (modified) Eisntein’s equations are the same as in General Relativity. Ferreira et al. we make a change of variable r → ρ. we can solve Eq. (35) gθθ = r 1 − 2G0 I − 4G0 U + τ + I ln r0 In order to preserve our previous assumption that gt t = −gρρ (corresponding to the particular case of a magnetic solution of the Einstein–Maxwell-dilaton equations). We return now to the original cylindrical coordinates system and obtain:     r 2 2 r 2 e + U − τ + I ln . ρ = r 1 + G0 4U + I − 4G0 U ln − 2G0 I ln r0 r0 and. (25):    e0 I 2 ln2 r + U − τ + I 2 ln r . Using expression (32) for the trace of the energymomentum tensor. thus. h00 = −4G r0 r0    r r e0 I 2 ln2 + U − τ − I 2 ln . (32) A straightforward calculus lead to the following solution of Eq. such that    r 2 2 2 r e e e . gzz = 1 − 4G r0 r0  r e0 I 2 − 4G e0 U + τ + I 2 ln . we re-obtained here the same solutions (33) as in Refs. / Nuclear Physics B 581 (2000) 165–178 175 and trace given by:  T(0) = − U + τ − I 2 δ(x)δ(y). hzz = −4G r0 r0  2   r r I 2 2 e r ∂i ∂j ln + U + τ + I δij ln .20]. with hµν given by (33). gt t = − 1 + 4G0 I ln r0 r0    e0 I 2 ln2 r + U − τ − I 2 ln r . e0 U − τ − I 2 ln ρ + 4G e0 U + I + 4G + ρ 2 1 − 8G 2 r0 r0 (36) . However. are identically satisfied. [18.ν = 0. we have      e0 I 2 ln2 ρ + U − τ + I 2 ln ρ −dt 2 + dρ 2 ds 2 = 1 + 4G r0 r0     ρ ρ e0 I 2 ln2 + U − τ − I 2 ln dz2 + 1 − 4G r0 r0    2  e0 I 2 ln2 ρ dθ 2 .N. (26) straightforwardly:  e0 α(φ0 ) U + τ − I 2 ln r . grr = 1 + 2G r0    r 2 2 2 e e . (34) φ(1) = 2G r0 As expected. the scalar-tensor coupling still brings a new information coming from solution (34).C. hij = −4G0 2 r0 r0 µ (33) µ One can easily verify that the harmonic conditions (hν − 12 δν h).

Expanding them in power series of the paramenters m and n. Bending of light rays A light ray coming from infinity in the transverse plane has its trajectory deflected. B 2 = 1 − 8G 2  e l = 2G0 α(φ0 ) U + τ − I 2 . gt t (ρmin ) which gives in turn: 2   p 1/(1−m ) ρmin = . 1 + κ r0 (1 + κ)2 r0 Making the identification of the coefficients of both linearised metrics.  e1/2 U − τ − I 2 . Comparison with the external solutions (22) and (24) requires a linearisation of these ones since they are exact solutions. ∆θ = 4π G (38) 3. the solutions of the scalar field and an isolated current-carrying string in the conformal frame. given by dρ/dθ = 0: gθθ (ρmin ) = −p2 . we finally obtain h(ρ) = 2n e0 I 2 .N. m2 = 4 G  2 e0 U + I . by an angle given by:  Z∞  2 −2 g p gθθ −1/2 dρ − θθ − − π. respectively. / Nuclear Physics B 581 (2000) 165–178 Expressions (34) and (36) represent. r0 Br0 .3. r0 1 .176 C. as long as the weak-field approximation is valid. we find ρ gρρ = −gt t = 1 + 2m2 ln + h(ρ). Ferreira et al. with 1 + κ2 2 ρ 1−κ ρ ln + 2n2 ln . ∆θ = 2π 1 − √ gρρ dρ we finally obtain  e0 U + τ + I 2 . κ =1+G 0 (37) Calculating now the deficit angle for metric (36)   d √ 1 gθθ . for an observer at infinity. gzz = 1 + h(ρ)   gθθ = B 2 ρ 2 1 + h(ρ) . ∆θ = 2 gρρ gt t gρρ ρmin where ρmin is the distance of closest approach.

in much the same way as Peter and Puy [20].N.D. we made an adaptation of Linet’s method which consists in linearising the Einstein’s and dilaton’s equations using distribution’s functions while taking into account the scalar-tensor feature of gravity. suggestions and a critical reading of this manuscript. an interesting investigation that opens up. is the analysis of the properties of the cosmic string generated by the action (2) in the supersymmetrized version. and we have already initiated to pursue. Acknowledgements The authors are grateful to Brandon Carter. the Departamento de Campos e Partículas) and to the Abdus Salam ICTP-Trieste for hospitality during the preparation of part of this work. Conclusion In this work we studied the modifications induced by a scalar-tensor gravity on the metric of a current-carrying string described by model with action given by Eq. if we compare our results with those obtained in General Relativity. Bernard Linet and Patrick Peter for many discussions. (2). We found that the metric depends on five parameters which are related to the string’s internal structure and to the scalar field (dilaton) solution. Performing an expansion to We can now evaluate the deficit angle to first order in G linear order in this factor. we see that expression (39) does not change substantially.C. Ferreira et al. it might be of relevance to analyse the possibility of gaugino and gravitino condensation in this scenario. The study of such a model raises the question of understanding the rôle played by the fermionic partners of the bosonic matter and by the gravitino in the configuration of a string. albeit the metric structure is indeed modified with respect to the one in General Relativity. CNF thanks to CNPq for a Ph. Using expressions (37). we have    e0 U + I 2 3 + ln ρ e0 I 2 . 2 with s ≡ (p/Br0 )(ρ/r0 )m −1 . . / Nuclear Physics B 581 (2000) 165–178 177 e0 . where it is implicit that the scalar-tensor degrees of freedom of the gravity sector are accomodated in a suitable supergravity multiplet. ∆θ = 4π G + 8ν G 2 r0 2 (39) 4. For this purpose. Concerning the deflection of light. Now. grant. 1 + m − m ∆θ = B(1 − m2 ) 2 Br0 where we have defined the quantity ν as Z1 ν ≡− √ 0 ln s 1 − s2 ds = π ln 2. Also. One of the authors (MEXG) thanks to the Centro Brasileiro de Pesquisas Físicas (in particular. we find:     p π 2 2 2 ln ν − π.

Nucl.S. Cambridge Univ. in: L. A 143 (1990) 438. Moss. Gundlach. Gravitation: An Introduction to Current Research. Laguna-Castillo.B.P. Damour. 36 (1990) 5800.R. Wiley. E. Witten. Phys. [17] A.N. P. Lett. M. D. Phys. Phys. Phys. Thorne. 1962.E. [7] M. C. Phys. [19] T. K. Noé Morales. Barros. [13] C. Revista Mexicana de Física 36 (1990) S199. Rev. Pimentel. P. B 209 (1988) 477. Phys. Linet. Peter. [3] C. Quantum Gravity 14 (1997) 435. [14] A.T. Rev. Sen. Schwarz.. [12] J. J.S. / Nuclear Physics B 581 (2000) 165–178 References [1] Th. Phys. Nielsen. Witten (Ed. D 42 (1990) 2521.X. Hill. Nordtverdt. Witten. Piran. Cho. Superstring Theory. Schramm. Nucl. Damour.H. T. [18] B. [16] P. Lett. [24] K. Quantum Gravity 6 (1989) 435.). [8] E. D 37 (1988) 877. Lett.M. D. Amsterdanski. Rev. D 57 (1998) 4629. D. hep-th/9902135. Walker. S. Rev. Press. D 48 (1993) 3436. Santos. J. Phys. [23] T. Phys. Class. Lett. [4] L. Rev. Class. 1987. Gregory. [22] L. Vilenkin. Olesen. Phys. I. B 293 (1992) 287. Green. to appear. Math. Phys. [10] R. Babul. D. Phys. Nascimento. Phys. D 60 (1999) 067501. B 249 (1995) 557.N. [5] A. Puy. [9] T. Dimopoulos. Helliwell. 138 (1965) 251. A. [21] H. Phys. Lett. New York. Vachaspati. Ferreira et al. [15] I.O. A. Cambridge.H.B. Nucl. Phys. Brandenberger et al. B 265 (1991) 258. D 56 (1997) 1194. Phys.A. . Rev. Guimarães. Phys. T. C. [11] K. Phys. Witten.E. Rev. Konkowski. Rev. [2] M. Spergel. Ortiz. [6] R. D 36 (1987) 1007. [20] P. B 61 (1973) 45. Romero. Rev. D 48 (1993) 5546. B.N. Poletti. gr-qc/9904057. B 199 (1987) 34.178 C.

For most of our results the explicit form of the matrix U connecting the d = 7 gravitino to the Killing spinor is not needed. The ansatz itself is nonpolynomial but the final d = 7 results are polynomial. USA Received 4 February 2000. PII: S 0 5 5 0 . but we derive the equation which U has to satisfy and present the general solution. we find the ansatz for the 4-form F .3 2 1 3 ( 0 0 ) 0 0 1 9 3 . Diana Vaman 2 . Consistent truncation 1.V.sunsyb.50. Requiring that the expression δF = dδA in d = 11 can be written as δd(fields in d = 7). accepted 23 March 2000 Abstract We discuss the full nonlinear Kaluza–Klein (KK) reduction of the original formulation of d = 11 supergravity on AdS7 × S4 to gauged maximal (N = 4) supergravity in 7 dimensions.physics.Mj Keywords: Supergravity.N. The corresponding ansatz for the 3-form A modifies the geometrical proposal by Freed et al. 04. The correct d = 7 scalar potential is obtained. Yang Institute for Theoretical Physics.A . 11.65.25.+e.edu 3 Research supported by NSF Grant 9722101.elsevier. Namely. All rights reserved. by including d = 7 scalar fields.+h. Therefore selfduality in odd dimensions originates from a first order formalism in higher dimensions. It satisfies the Bianchi identities. PACS: 04.10.nl/locate/npe Consistency of the AdS7 × S4 reduction and the origin of self-duality in odd dimensions Horatiu Nastase 1 .Nuclear Physics B 581 (2000) 179–239 www.sunysb. NY 11794-3840. Introduction The consistency of Kaluza–Klein (KK) truncation at the nonlinear level of a massless gravitational higher-dimensional field theory to a massless gravitational lower-dimensional field theory has been a fundamental problem for a long time.physics. A first order formulation for A in d = 11 is needed to obtain the d = 7 supersymmetry laws and the action for the nonabelian selfdual antisymmetric tensor field Sαβγ . KK compactification.sunysb. SUNY Stony Brook. All rights reserved. Peter van Nieuwenhuizen 3 C. E-mail: vannieu@insti.edu 0550-3213/00/$ – see front matter  2000 Elsevier Science B. In this article we present a complete solution for a particular model. 11.0 .physics.edu 2 dvaman@insti.  2000 Elsevier Science B. we consider the original formulation 1 hnastase@insti.Kk.V. We derive the full nonlinear embedding of the d = 7 fields in the d = 11 fields (“the ansatz”) and check the consistency of the ansatz by deriving the d = 7 supersymmetry laws from the d = 11 transformation laws in the various sectors.

However. η(x)). This article contains a detailed and self-contained derivation of our results. . (4. it was later realized that it was impossible to obtain in this way chiral fermions and/or a vanishing cosmological constant [15]. x)ηI (x) between the d = 11 susy parameter ε(y. x)8(y. A particular nonlinear combination of φ’s 0 and η’s enters as a 4 × 4 matrix U I I (y. The basic problem we must analyze is that in relations like δφ(y)Y (x) = ε(y. It is a square root of a Killing vector and is the spherical harmonic of the local supersymmetry parameters. while the contributions to U quadratic in scalars were computed in [7]. It appears in the ansatz for all fermionic 8’s 0 and in the relation ε(y. There exists in the literature also a modified version of d = 11 supergravity with a local SU(8) symmetry and for this theory de Wit and Nicolai [2–5] have shown that the KK reduction is consistent. and the compactification on AdS4 × S7 seemed promising ([13]. and write all components of all d = 11 fields. We shall analyze all bosonic variations δ8(φ(y). All spherical harmonics will be expressed in terms of (products of) Killing spinors ηI (x) with I = 1. and we present explicit solutions. (In an earlier study of the KK reduction of the original d = 11 supergravity to d = 4 they gave partial results on the matrix U [2. (An example of such a relation is Eq. in terms of which we express all spherical harmonics. Nastase et al. η(x)) up to terms with three d = 7 fermion fields. denoted by 8(y. we expect that consistency holds there as well. In fact. as nonlinear expressions in terms of the massless d = 7 fields φ(y) and spherical harmonics Y (x) on S4 . 4. is the Killing spinor. . The consistency of this ‘ansatz’ amounts to proving that the d = 11 supersymmetry variations of 8 produce the correct d = 7 supersymmetry variations of φ. This matrix U must satisfy a linear matrix equation.) This requires a large number of properties of spherical harmonics. without any string extensions. x) = 8(φ(y). but this is algebraically very complicated and given all small miracles we have encountered in our work. and all fermionic variations up to terms with two d = 7 fermion fields. x) and the d = 7 parameter (y).180 H. Initially.h. The spherical harmonics of scalars. x).s. but all results are to all orders in bosonic fields. x). . Other KK reductions to a subset of massless fields have also in some cases been shown to be consistent [8–11]. starting from scratch. see also [14]). . but then they abandoned this project. The basic object. and is an expanded version of a previous letter [12]. Recent developments in the AdS-CFT correspondence have renewed interest in KK .3]. This is the first time a complete nonlinear KK reduction of an original supergravity theory to all massless modes in a lower dimension has been given. spinors and other fields are expressed into Killing spinors. they did not try to determine it. x) one finds products of Y (x)’s on the r. Because in their version of the theory the matrix U only described gauge degrees of freedom.30). Hence one must prove relations of the form Y (x) = ΠY (x). For the reader who is not an expert in these matters we have added a special section (Section 3) in which we give a derivation of all identities we needed. / Nuclear Physics B 581 (2000) 179–239 of d = 11 supergravity [1]. Previous work in this area has also made the same restriction [3–5]. the matrix U was first introduced in an implicit form in [6]. In principle one could also study the variations containing higher orders in Fermi fields with our methods. Hence we present below explicit expressions for 8(y. x) ∼ I 0 (y)U I I (y. the aim of KK reductions of higher dimensional supergravities (sugras) was to find realistic field theories in d = 4 dimensions.

But even keeping the scalars in the theory is in general not enough to obtain consistency. 3) × SU(8) tangent space symmetry group. In the course of its study. In the Einstein equation. The study of KK truncations took a new direction with the advent of supergravity. They realized that the field equation for the scalar contains source terms depending on the gauge fields. and G0 acts transitively on Mk (i. and it was speculated that the truncation to a sugra multiplet is always consistent. In [34] it was claimed that the truncation to 4D N > 1 sugra multiplets with second order formulation for the fields is always consistent. more fields are present — for instance in 11 dimensions we have µ the antisymmetric tensor AΛΠΣ — and so the condition Vµa Vb = δba gets modified in a complicated way). on the other hand. This was already noticed by Kaluza and others who studied the reduction from 5 to 4 dimensions. depending on the 4D scalar fields. any point in Mk can be reached from any other point by a G0 gauge transformation). other criteria for consistent truncations were found in [9. this mechanism gives a group theoretical explanation for the consistency of a particular truncation consisting of massless gauged N = 3 supergravity coupled to some massive matter multiplets (this system is not contained in N = 8 sugra). For the Maxwell–Einstein system. hence setting the scalar to zero was inconsistent. the truncation to massless fields is consistent as far as the gauge symmetries are concerned [33]. In their pioneering work. whereas the rest of the terms have spherical harmonic 1. and dimensionally reduced this theory. This requirement implies in particular that Mk is a coset manifold. one can find such subsets of Killing vectors. In the case of compactifications on a group manifold and on spheres S 4n+3 . The most intensively studied model was the N = 8 sugra.H. In general. The dependence of this matrix on one of the scalars was even determined. which cannot be truncated to gravity plus gauge fields of the symmetry group K of the compact manifold MK . this gives an infinite set of fields (the untruncated set is of course always infinite). The study of the consistency of the KK reduction of d = 11 sugra at the nonlinear level was tackled in a series of papers by de Wit and Nicolai. / Nuclear Physics B 581 (2000) 179–239 181 compactifications on anti-de Sitter (AdS) spaces [16–31] (for a recent review see [32]). but they did not succeed in finding the dependence on all other scalars. the terms involving the µ gauge fields will have as spherical harmonics Vµa Vb . this is not in general so. However.. In [9] it was realized that for a field theory invariant under a group G. whereas on S 2n one can not (as shown in [8]). b ∈ K 0 ) [8] (in sugra. this yielded an ansatz for the fermions only up to a matrix U . a consistent truncation can be obtained by retaining only all fields which are singlets under a subgroup G0 ∈ G. While KK reduction (more precisely. A simple example of an inconsistent truncation is Einstein gravity. truncation to the set of massless fields) on tori is always guaranteed to be consistent. One was to construct directly an ansatz for the 11D metric as opposed to the vielbein.e. They then switched to another approach in which they reformulated 11D sugra as a theory with a local SO(1. In the case of the AdS4 × S7 compactification. Nastase et al.34]. which proved to be very challenging. the subset of fields is finite. This latter approach was used to prove consistency of the reduction . they followed two approaches. if one restricts the gauge group K to a subgroup K 0 such µ that the subset of Killing vectors satisfies Vµa Vb = δba . G/H = G0 /H 0 . then the KK reduction is consistent (a. If G is the isometry group of the KK compactification on a space Mk .

a rotation from A.A . Consistent truncations of bosonic subsets of fields to other dimensions were also considered in [37–40]. It was found that the second order field equation for the field Aαβγ . this mechanism of self-duality in odd dimensions leads to a consistent coupling of a multiplet of massive antisymmetric tensor fields to non-abelian gauge fields. d = 10 sugra which reproduced the bosonic action of N = 2. Lately. In Section 2. the authors of [10. It was first discovered in a study of the KK reduction of d = 11 sugra on S4 [42]. and found to give a formulation dual to Chern–Simons theory for the abelian case [43. they did not explicitly checked consistency. (1. but left the ansatz of the original 11-dimensional fields in terms of the 4D ones implicit. After that. In sugra. a model with 4 scalars in [36]). which one obtains from KK reduction of the field AΛΠΣ . one SU(2) gauge field and one antisymmetric tensor in [35].A . d = 7 gauged sugra. however. / Nuclear Physics B 581 (2000) 179–239 by reproducing the 4D susy transformations laws and equations of motion. In a previous letter [12]. a series of other papers looked at other truncations of 11D sugra on AdS7 × S4 (a model with one scalar.182 H. This paper is organized as follows. see (2.A : A = S + b. b yielded the final massless field Sαβγ .11] proposed nonlinear ansätze for the embedding in 11D sugra and 10D IIB sugra of a further truncation of the gauged sugras in 4. In Section 2.41). The authors of [41] also worked out an ansatz or an S3 truncation of N = 1. B to S.A with action B 2 . The linearized KK reduction is given in Section 2. and prove its local supersymmetry. The consistency of the truncations was proven by reproducing the lower dimensional equations of motion from the ones of d = 11 and d = 10 sugras (see next section). We have found that B is obtained by KK reduction of the auxiliary field BΛΠΣΩ in d = 11.2 we present d = 11 sugra with a first-order formulation for the 3-index photon Aαβγ . Nastase et al. 7 and 5 dimensions to a set of U (1) gauge fields and scalars (in the AdS7 case. Here we present the details of the construction.1 we discuss the issue of the consistency of KK truncations in general. In Section 3 we discuss Killing spinors and spherical harmonics. but not for the non-abelian case [45]. We begin in Section 3. B = b − O(S + b). we have given the full nonlinear ansätze for the KK reduction of 11D sugra on AdS7 × S4 .3. Although it was already given in [46]. suggested by the need to embed AdS4 and AdS7 black holes as solutions of 11D sugra (and AdS5 black holes as solutions of 10D IIB sugra).1) where O is an operator involving derivatives of the form D + m. The ansatz for FΛΠΣΩ was not explicitly given in the first approach. Adding in d = 7 an auxiliary field Bαβγ . was of the form 2A + m2 A + m∂A = 0. two U (1) gauge fields and two scalars). which is obtained by using a first-order formulation for the 3-index photon in d = 11.1 with a description of the . but the equations needed to construct FΛΠΣΩ from the second approach were highly involved. In the second approach no FΛΠΣΩ appears.A and a massive mode bαβγ .44]. we summarize it here since it forms the starting point for the nonlinear extension. Crucial for the proof of consistency is a proper understanding of the origin of the concept of self-duality in odd dimensions — seven in this case. We argue that it is sufficient to obtain the susy transformation laws of the dimensionally reduced theory from the susy transformation laws of the untruncated theory.

e.2 we give some details of the charge conjugation matrices in various dimensions and properties of modified Majorana spinors. with the gauge fields set to zero.1 we give our conventions and some useful relations for Dirac matrices.1 we present the full ansatz for the d = 11 vielbein and gravitino and we discuss the susy transformation laws of the vielbein.2 we derive an ansatz for the d = 11 3-index antisymmetric tensor field AΛΠΣ by a partial analysis of the susy laws for AΛΠΣ and by a preliminary study of the susy transformation laws of the gravitino. In Appendix A. S (D) ({8}). In Section 5 we derive for completeness the seven-dimensional bosonic equations of motion from the 11-dimensional bosonic equations of motion. δφ(n) = gn ({φ(0) })+ more. We also derive the seven dimensional bosonic action and compare our ansatz with other ansätze for consistent truncation to bosonic subsets of fields. More precisely. 2. if we truncate the fields in a Lagrangian by putting the ‘massive’ ones {φ(n) } to zero and keeping only the ‘massless’ ones {φ(0) }. δS/δφ(n) must not contain any term depending only on the massless fields (at the linear level this is true because φ(n) are eigenfunctions of the kinetic operator) because otherwise setting φ(n) = 0 would be inconsistent. If we can reproduce the correct d-dimensional susy laws from .1. including certain completeness relations. In Appendix A. In Section 4. / Nuclear Physics B 581 (2000) 179–239 183 spherical harmonics used for the massless multiplet and give various useful identities in Section 3. there exists a massive field which varies into a term involving only massless fields. In general. It leads to the d = 7 ‘self-duality in odd dimensions’.. Similarly. KK consistency Let us briefly discuss the issue of the consistency of the KK reduction. Nastase et al.3 we give the reduction of the rest of the d = 11 gravitino transformation laws and derive the transformation law of the 3-index d = 7 tensor Sαβγ . but we need to find the nonlinear ansatz which specifies how the final massless fields φ(0) are embedded in the higher-dimensional fields 8. We finish in Section 6 with conclusions and discussions. In the case of the KK reduction of a D-dimensional sugra action. in the transformation rules for the symmetries of the system. there should be no term linear in massive fields in the untruncated action R d d x φ(n) fn ({φ(0) }). For a consistent truncation. In Section 4. we usually know the Lagrangian in lower (d) dimensions S (d) ({φ(0)}) which we should obtain after truncation. First-order d = 11 supergravity and linearized KK reduction 2.H. because such a term would give an inconsistency of the equations of motion. In Section 4. Let us assume that both the equations of motion and the susy laws are inconsistent. i. the equations of motion of the massive fields. we have to check that the full equations of motion and transformation laws (for all the symmetries of the system) are consistent under this truncation.A . again because then we cannot put the massive fields to zero. Section 4 is the heart of this paper: it contains the proof of the consistency of KK reduction.2. the massive fields should not transform into a term with only massless ones.

{φ(0)} = δS (d) {φ(0)} + gn {φ(0)} fn {φ(0)} + O(φ(n) ) = 0. (For our conventions and notation. In [10.11] the consistency of the KK reduction from the original d = 11 supergravity to AdS7 × S4 to a small subset (two scalars and two U (1) gauge fields) of our fields was studied using the other criterion.. (2.1) Since the d-dimensional theory is invariant.2) 8 where FΛΠΣΩ is an independent field with field equation FΛΠΣΩ = FΛΠΣΩ . Nastase et al. the equations of motion and susy laws cannot both contain purely massless terms for a given massive field. / Nuclear Physics B 581 (2000) 179–239 the D-dimensional susy laws. [35] and a model with only 4 scalars [36]).Λ4 Σ 9Σ + 129 1 0 Λ2 Λ3 9 Λ4 24 F − (2. Hence.. the two inconsistencies are presumably always equivalent..184 H. and consider both fermionic and bosonic fields. After that.Λ4 .2. Supergravity and first-order formulations We start with the sugra Lagrangian in (10. For completeness we shall also consider consistency of the bosonic field equations.4) PN − (Λ ↔ Π). see the appendix).. a paper appeared which gave first-order formulations of supergravities. δS (d) ({φ0 }) vanishes. Only bosonic fields were considered. Therefore one criterion for a consistent truncation is to find a nonlinear ansatz which gives the susy transformation laws of S (d) from the susy transformation laws of S (D) .5) . the variation of the term in the action which is linear in massive fields gives an extra piece depending only on massless fields: Z     δS (D) {φ(n) }. Ω) − 9 Λ 0 ΛΠΣ DΠ 2k 2 2  E FΛΠΣΩ F ΛΠΣΩ − 48F ΛΠΣΩ ∂Λ AΠΣΩ + 48 √ 2 Λ0 . E 9 Π 0 ΠΛ1 . Ω) = RΛΠ MN (Ω)EM RΛΠ MN (Ω) = ∂Λ ΩΠ MN + ΩΛ M P ΩΠ (2.. for consistency gn fn = 0. one SU(2) gauge field and one antisymmetric Sαβγ . We do not restrict our attention to subsets..   b E Ω +Ω 1 9Σ L = − 2 ER(E.Λ10  ∂Λ0 AΛ1 Λ2 Λ3 ∂Λ4 AΛ5 Λ6 Λ7 AΛ8 Λ9 Λ10 −k 6 √  1  2k  Λ eΛ1 . 1) dimensions with a first-order formalism for the antisymmetric tensor field.e. (2. and FΛΠΣΩ = ∂Λ AΠΣΩ + 23 terms. R(E.. i. Since the commutator of the susy transformations gives the equations of motion. this work has been extended to other subsets of bosonic fields (a model with one scalar. (2. namely consistency of the field equations.3) M and Furthermore E = det EΛ Π Λ EN . 2. More recently. in particular a formulation of 11D sugra which is first-order in both the spin connection and the antisymmetric tensor field [47]. This formulation was presented in our previous letter [12].

12) δEΛ 2 √  1 2 Λ1 . but rather it is the solution of its field equation b term. Then the F field equation reads b and becomes supercovariant.. / Nuclear Physics B 581 (2000) 179–239 185    bΠ MN b 1 ΩΠ MN + Ω Ω +Ω 9Σ = ∂Π 9Σ + 0MN 9Σ . Ω is not independent. (2.....13) 24 E E √ 2 ε¯ 0[Λ1 Λ2 9Λ3 ] ..Λ4 ε 24 δ9Λ = DΛ (Ω)ε Λ − 8δΛ 0 k 12   1 Λ1 Λ2 Λ3 1 Λ1 .2) by F F b.Λ4 Λ1 Λ2 Λ3 Λ4  1 b b + 0 FΛ1 . (2.e.Λ5 9Λ5 = −0 DΠ 9Σ − R (9) = E δ9 Λ 4 √  bΛΠΣΩ 0ΠΣ 9Ω ..Λ4 − a0 √ BΛΛ1 Λ2 Λ3 ε. ΩΠ MN = Ω 8 (2.Λ4 F b is the supercovariant curl of A 4 and we use 1.. √ E Substituting this definition into the terms in the action involving F .9) It is useful to redefine: Q M ..5-order formalism for the spin Here F connection. − 3 2k 1 F Λ 24 (2... Nastase et al....6) 2 4 2    b F +F 1 = = 24 ∂[Λ1 AΛ2 Λ3 Λ4 ] + √ 9 [Λ1 0Λ2 Λ3 9Λ4 ] . Ω 4 b The relation between Ω and Ω is then given by bΠ MN − 1 9 Λ 0ΛΠ MN Σ 9 Σ .H. the terms ( 9009)F e get absorbed.8) (2. (2. (2. we obtain: FΛΠΣΩ = ∂Λ AΠΣΩ + 23 terms + 1 E(∂[Λ AΠΣΩ] + 23 terms)2 + − 48 2 BMNP Q 48 ..Λ4 1 b0Λ + √ BΛ1 . The (to which only the Ω term. . i..15) δBMNP Q = E ε¯ a0MNP EQ where RΛ (9) is the gravitino field equation: √  1 δL 2 1 b ΛΠΣ k 24 FΛ1 . (2.16) 4 By replacing 24F ∂A in (2.. (2.10) (2.11) 2 The supersymmetry transformation rules which leave the action with the BMNP Q term invariant read k M = ε¯ 0 M 9Λ .Λ4 0 ΛΛ1 . (2. We have not been able to absorb the remaining four-Fermi terms by using F =F our new first-order formulation.7) 2 16 2 Λ1 . in the gravitino action contributes). but not the Ω b supercovariant spin connection Ω is obtained from Ω by adding terms bilinear in the fermions such that there are no ∂ terms in its susy transformations: 2  bΠ MN = ΩΠ MN (E) + k 9 Π 0 M 9 N − 9 Π 0 M 9 n + 9 M 0Π ψ N ..E BMNP Q EΛ Ω .14) δAΛ1 Λ2 Λ3 = − 8 √  Λ RΛ (9) + b0MNP QΛ R Λ (9) ..Λ4  DΠ eΛ1 .

/ Nuclear Physics B 581 (2000) 179–239 bΛΠΣΩ denotes the usual supercovariantization of ∂Λ AΠΣΩ + 23 terms The expression F and a and b are free parameters. The auxiliary field would then be obtained as the totally antisymmetric part of the difference between the independent spin connection and the solution of its equation of motion. R=R (4) +R (7) = − 32 m2 . Of course δB is proportional to the gravitino field equation. hence kA is dimensionless and k9 has dimension 1/2.19) (2.23) First order formulations in sugra have been given before in d = 4. Could we have chosen another first-order formulation which would give us at the linearized level this auxiliary field? The other obvious choice is the first-order formulation for the spin connection (Palatini formalism). because B = 0 is a field equation and field equations (usually) transform into field equations.20) (2. This theory admits a background solution which satisfies the equations of motion and which describes a geometry of the type AdS7 × S4 .186 H. Nastase et al.24) BΛΠΣ = Ω[Λ MN − Ω .17) Fµνρσ = √ m det e˚µm (x) µνρσ . (2. a first-order formulation for the spin connection was given in [50]. They will be fixed by the requirement of consistency of the KK truncation on S4 . The maximally symmetric solution is given by   Rµν mn e˚ (4) = m2 e˚µm (x)e˚νn (x) − e˚νm (x)e˚µn (x) . The parameter m is the inverse radius of the sphere as will now be show. BΛΠΣΩ = 0. Soon after the N = 1 sugra was formulated with a second-order formulation for the spin connection [48]. The gravitational constant k has dimensions 11/2.. i. The reason we start with a first-order formulation for the antisymmetric tensor field in d = 11 is that at the linearized level one needs in 7 dimensions a 3-index auxiliary field to combine with the 3-index tensor to give the 7-dimensional supergravity field Sαβγ . The field equations for the background geometry read then  (2. a first-order formulation for the spin connection (with an independent field ωµmn and an expression for δωµmn ) was given [49].e.21) (2. The source is given by  3 (2. For d = 11 sugra. 2 where e˚µm is the vielbein on S4 and e˚αa the vielbein on AdS7 . this suggested to us that the auxiliary field in d = 11 gives by reduction the auxiliary field in d = 7.22) All other fields vanish in the background: M = 0. Since the 3-index auxiliary field can also be written as a 4-index auxiliary field by a duality transformation.18) Rµν − 12 g˚µν R = − 16 Fµ··· Fν ··· − 18 g˚ µν F 2 = − 94 g˚ µν m2 . Our results confirm this suggestion. Below we set k equal to unity. 9) EMΠ ENΣ] . 9Λ FαΛΠΣ = 0.A found in [51]. and Rαβ − 12 g˚αβ R = 2 1 48 g˚ αβ F = 94 g˚ αβ m2 .  b[Λ MN (E. (2. New in the transformation laws are the B terms in δ9Λ and the expression for δB. (2.   Rαβ ab e˚ (7) = − 14 m2 e˚αa (y)e˚βb (y) − e˚αb (y)e˚βa (y) .

27) hαβ (y. where the redefinition of the graviton is needed in order to diagonalize its kinetic term. In full generality. We shall suppress these spinor indices. . i. x) = SI J KL (y)ηµν (2. We reproduce it here for completeness.e. In the fluctuations we keep only ‘massless’ modes. x) = Bα. 2.6) and (3. x)g˚ µν (x) . but its symplectic trace and its totally antisymmetric part vanish (see below (3.3. J = 1. I. (2. x) produces the metric g M vielbein EΛ ΛΠ = EΛ EMΠ = g˚ ΛΠ + khΛΠ . The linearized terms in the expansion of the bosons have the generic form: X φα. 32 of an 11-dimensional spinor decomposes as A = a˜ ⊗ a˜˜ with a˜ spinor indices on AdS7 and a˜˜ spinor indices on S4 . For fermions one finds a similar decomposition: X ˜ 9Ia˜ (y)Y I. we obtain an auxiliary antisymmetric tensor field Bαβγ .. (The linearized kinetic term for the graviton is the Fierz–Pauli action for a massless spin 2 field and reads in any dimension L = 1/2h2µν. where hΛΠ are the fluctuations given by the following expressions in the various sectors: (i) In AdS7 space–time:  (2. the massless 7-dimensional fields may occur nonlinearly in the expansion of 11-dimensional fields. We will see later (in Section 3) that this approach does not produce the 7-dimensional auxiliary field we need.I J (y)VµI J (x). (2. The linearized ansatz has been given in [46].H.I (y)YµI (x). hµν (y.) (ii) In the mixed sector: hµα (y. x) = I where the spinor index A = 1.28) where VµI J (x) = VµJ I (x) are the ten Killing vectors on S4 . Nastase et al.a˜ . / Nuclear Physics B 581 (2000) 179–239 187 When we dimensionally reduce. (iii) In the S4 sector: I J KL (x).26) 9 A (y. 4-dimensional indices to the spherical harmonics Y of the internal space. 4.25) 8αµ (y. and thus a˜ and a˜˜ will not be used below. The M (y. (2. fields which in 7 dimensions belong to the maximal (N = 4) sugra multiplet. x) = I where φ are 7-dimensional fields and Y are spherical harmonics. x) = hαβ (y) − 15 g˚αβ (y) hµν (y. The 7-dimensional indices are attached to 7-dimensional fields.ρ + h2µ − hµ hµ + µ 1/2h2µ where hµ = ∂ ν hνµ and h = hµ .29) I J KL is where the scalars SI J KL are in the 14 representation of USp(4) and ηµν the corresponding spherical harmonic. Linearized Kaluza–Klein reduction on S4 Any 11-dimensional field is written as a sum of the corresponding background field and the fluctuations.7)). It has the symmetry of the Riemann tensor.

30) The λJ. see Appendix A. 12 2 (2. Nastase et al.KL (x) the corresponding spherical harmonic. The singlet has bigger mass. 4 are the fermions in the 16 representation of USp(4). and the lowest mass modes are the singlet and the 14.KL (y)γ5 ηµJ KL (x). . a fact which we know for instance from the case of toroidal compactification. (2. First.35) 3 3 √ √ 2 2 fαβγ δ = 4 ∂[α Aβγ δ]. . The spinors have as lowest modes a 4 and the 16. x) = 40 i (2. with ηµJ.1.40) where Aαβγ . x) = λJ.31) −1/2 = 1+i where ηI (x) is the Killing spinor.I J (y)φ5I J (x). hλλ ≡ hµν (y. (2. and is found to belong to another multiplet. x).KL with J.188 H. (2. By definition. (2.I J (y)η¯ I (x)γµν γ5 ηJ (x).I J ∂ν φ5I J (x). γ5 2 (iγ5 − 1). (2. and its totally antisymmetric part vanishes (and so all its √ symplectic traces are zero).33) 3 √ i 2 IJ I fµναβ = ∂[α Bβ] η¯ γµν γ5 ηJ . where the spin 1/2 fields are written as 1/2 9µ (y.I J φ5I J (x). (2.37) Aµνρ (y. K. L = 1.34) 3 3 √ √ 2 2 fναβγ = Aαβγ . . Aαβγ (y. γ5 = i−1 2 (1 + iγ5 ). We took the scalars to be in the 14 representation and the spinors in the 16 representation because of the following reason. The antisymmetric tensor FΛΠΣΩ = F˚ΛΠΣΩ + fΛΠΣΩ (where fΛΠΣΩ is the fluctuation and F˚ΛΠΣΩ the background solution) decomposes at the linearized level as follows: √ q 2 fµνρσ = det g˚µν µνρσ hλλ .36) 3 3 so that √ 2p g˚ µνρσ D σ hλλ . Second. (2. but again the 4 has higher mass and belongs to the same multiplet as the scalar singlet. (2. the complete mass spectrum on S4 was found in [55]. x) = ψαI (y)γ5 η (x) − 15 τα γ5 γ µ 9µ (y. . but we . The gravitini 9α also have to be redefined in order to diagonalize their kinetic term: ±1/2 I 9α (y. x) = 16 Aαβγ .KL is antisymmetric and symplectic traceless in KL. / Nuclear Physics B 581 (2000) 179–239 The gravitino 9Λ splits into 9µ and 9α . x) g˚ µν (x). and the total number of fermions 16.33) was determined in [46] by solving the Bianchi identities.32) 3 √ q 1 τ  2 fανρσ = det g˚µν νρσ τ 10 D Dα hλλ − hτα .I J is a set of antisymmetric symplectic-traceless tensors in the 5 representation of USp(4) and φ5I J = η¯ I γ5 ηJ = −η¯ J γ5 ηI the corresponding scalar spherical harmonic.38) Aαµν (y. (2. x) = √ Bα. Clearly.39) Aαβµ = 0. The factor 1/10 in (2. the total number of scalars has to be 14. λJ.

where c¯c is defined ωi Hi and the spin connection term 1/2ω(spin)a bc Jbc is 1/4c¯ab c ab 0 0 a b i d a aa by [Ka . On S4 we use µ = 1. the field Aαβγ . 3.I J has an action with 2 derivatives. The rest of the result reproduces (2. cab d = 0 and thus ωi Hi is the spin connection.1. see (3. 4 as flat coset indices. where these issues are discussed.I J + 16  αβγ δηζ Dδ Sηζ.3).I J is linear in derivatives.I J on Sαβγ . one for Sαβγ . one then obtains (d + ωi Hi )L−1 = −(em Km )L−1 .32) one needs 2g˚ µν ηµν µν ¯ µ η.I J and one for a massive field b. For a sphere S4 of radius m as chosen in (2. c = ±m/2. see [54.I J .I J = 15 Sαβγ . Spherical harmonics and Killing spinors 3. Nastase et al. Since S4 = SO(5)/SO(4) is a symmetric manifold. / Nuclear Physics B 581 (2000) 179–239 189 have here given the linearized KK formulas for the fields themselves. For a treatment of spherical harmonics on S4 and on general coset manifolds. in (2. and therefore can have the same spherical harmonic. This can be derived from the integrability condition .I J .H. c¯db = cdb + δ (ca 0 d δbb0 + b ↔ d) [53]. Note that if we want to understand B as coming from a KK reduction. where the scale parameter c is fixed in terms of the curvatures. In general two fields in lower dimensions with the same index structure cannot have the same spherical harmonic.1) ∂µ + 14 ωµmn (x)γmn ηI (x) = ceµm (x)(iγm )ηI (x). (The index α is a spinor index and runs from 1 to 4.20).55]. 4 instead of α as curved coset indices. That means that we need to introduce by hand an auxiliary field Bαβγ . Spherical harmonics for the massless multiplet The coset representatives of a reductive coset manifold L(x) = exp(−x α Kα ) satisfy the equation L−1 dL = ea Ka + ωi Hi .I J . see (3. γn ] and Km = −icγm we define ηI (x) ≡ L−1 ηI (0) where ηI (0) is a constant spinor with ηI (0)α = Ω αI . its spherical harmonic should be the same as for the field A.To show that (2. one would expect that Aαβγ . Then the spinors ηI (x) are Killing spinors as they satisfy  (3.I J reads  (2. Kb ] = cab Hi + cab Kd . the difference between c J b .33) then follows if one uses D˚ µ η = 2i γµ η and Vµ = ηγ When massive modes are put to zero. and rotate the A and B fields such that the sum of the action of A with 2 derivatives and the action of B with none gives two decoupled actions each linear in derivatives.37) I J KL = −10g˚ µν ηI J KL . the dependence of the auxiliary field Bαβγ . whereas the action of Sαβγ . However. and m = 1. Using the spinor representation of SO(5) with Hi = 14 [γm .I J becomes equal to the 7-dimensional supergravity field Sαβγ . while Ω is a constant antisymmetric 4 × 4 matrix. We suppress the spinor index α in most of the text.7).) In general.41) Bαβγ . From dL−1 = −(L−1 dL)L−1 with coset generators Kα and subgroup generators Hi . After setting the massive field b to zero. but in our case B and A have different mass dimensions.

φ5I J ΩI J = 0. D˚ ν ]η = R˚ µν mn 41 γmn η.2 we discuss the definition η¯ I = ηI. understanding that we should replace everywhere ∂µ by ∂µ /m to get the correct dimensions. (In fact D˚ µ CνA = D˚ (µ C A .1). when we substitute the background curvature.2) ηαI = exp βΩ . It is shown there that Ω I α = −(C −1 )I α . (3.7) 5 It follows from this integrability condition that in general Killing spinors exist only on Einstein manifolds. ηµν  I J KL (−10) = g˚µν φ5I J φ5KL + 14 CλI J C λKL .T C.24) that Vµ = µ µ Y [A ∂µ Y B] I J KL = ηI J KL . D˚ µ η¯ I = − η¯ I γµ . They satisfy: i i η¯ I ηJ = Ω I J . They satisfy D(µ Cν) = 4 gµν (D Cρ ). Nastase et al. ηµν I J KL . or CµA = i IJ 1 A A A ˚ A ˚ A ˚ ˚ρ A 4 Cµ (γ )I J = Dµ Y . (3. VµI J = VµJ I = η¯ I γµ ηJ or VµAB = −VµBA = − 8i VµI J (γ AB )I J . −10ηµν (3. with Young tableau in the form of a 2 × 2 box: • Symmetric tensors ηµν νµ I J KL I J KL I J KL = ηµν (−2) − 13 ηµν (−10). CµI J ΩI J = 0.) ν) • Killing vectors 10.6) (3. . On spheres there are two sets of Killing spinors η± satisfying D˚ µ η± = ±cγµ η± which are related by η+ = γ5 η− in even dimensions. In terms of Killing spinors one has: I J KL (−2) 2ηµν I J KL I J KL (−2) = C(µ Cν) − 14 g˚ µν CλI J C λKL .11). where the charge conjugation matrix (C4− )αβ on S4 is equal to the symplectic metric Ωαβ .3) D˚ µ ηI = γµ ηI . γ5 }I J by lowering the index according to the rule (A. CµI J = −CµJ I = η¯ I γµ γ5 ηJ .11). We show in (3. AB = 0 and V I J = iV AB (γ IJ AB satisfying D˚ (µ Vν) AB ) .) • Conformal Killing vectors 5. ηµν I J KL −2ηµν I J KL (−10) 2ηµν (3. The Killing spinors can be given in an explicit form: n  −i oα m βI x µδm γ (3. where Dµ Y = ∂µ Y . µ 2 where x µ are general coordinates in patches covering S4 . while the second one is a pure trace. All spherical harmonics can now be built from Killing spinors: • Scalars 5. 4 and γ A defined in (A. or Y A ≡ 14 φ5I J (γ A )I J with I.5) = and = The first harwhere monic is traceless. and substituting into (3. A direct consequence of the second equation is a completeness type of relation satisfied by the Killing spinors ηJα η¯ βJ = −δβα with ηJα ≡ ηαI ΩI J .14). while the second one is easy to check. φ5I J = −φ5J I = η¯ I γ5 ηJ .190 H. In the following we choose to work with Killing spinors for which c = m/2. 2 2 (In Appendix A. we shall also drop the factors of m.) The first relation follows from our earlier discussion. J = 1. / Nuclear Physics B 581 (2000) 179–239 on the Killing spinor: 5 [D˚ µ . expressing L in terms of a cosine and sine as in (3. (The matrices (γ A )I J are antisymmetric and obtained from {iγm γ5 .4) One can directly verify the Killing equation by constructing the vielbein em and spin connection ωi from L−1 dL.

one obtains the vectors Y A in terms of the coset parameters as   x) (3.2.13) = 1. If the unit sphere lies on top of the stereographic plane one obtains: x2 (Y A )2 2Y m .8) The first harmonic satisfies γ ν D˚ ν ηµJ KL (−2) = −2ηµJ KL(−2) and is gamma traceless. 4 + ξ2 2 .4)) and without a totally antisymmetric part (use Fierz rearrangements). 1 − Y5 In these ξ coordinates the vielbein has a very simple form: ξ µ = δm µ eµ m (ξ ) = δµ m 4 . (3. azimuthal angle of a point on the unit sphere with coordinates Y .  (3.7) are symplectictraceless (use (3.12) P = x µx ν δ (3. Nastase et al. / Nuclear Physics B 581 (2000) 179–239 191 (Use 2φ5I J = −4φ5I J and 2CµI J = −CµI J .2) into the definition of Y A in terms of the Killing spinors. Again these spherical harmonics can be expressed in terms of Killing spinors:  (3.9) ηµJ KL(−2) = 3 ηJ CµKL − 14 γµ γ ν ηJ CνKL . 4 + ξ2 (3. (3.5) and 1 in (3. • Vector-spinor ηµJ KL with Young tableau in the form of a gun: ηµJ KL = ηµJ KL(−2) + ηµJ KL(−6). Identities involving spherical harmonics Substituting (3. Clearly.8) are not fixed by properties of the spherical harmonics themselves.H.11) Y A = − 14 Tr γ A γ5 exp(−i/ or.17) 1 ± η± (ξ ) = m µ η (0).14) (3.6) and (3.10) ηµJ KL(−6) = γµ ηJ φ5KL − 14 γ ν ηJ CνKL . x Y 5 = − cos x. On a sphere one can also define stereographic coordinates ξ related to Y m by a conformal mapping. while the second one is a pure gamma trace and satisfies γ ν D˚ ν ηµJ KL(−6) = −6ηµJ KL(−6).16) ωµ mn = 1 + ξ2  4  i µ m ± γ ξ δ (3.) Both (3. 3. (3.15) Also the spin connection and Killing spinors are simple rational functions of the ξ ’s:  1 −δµm ξ n + δµn ξ m . and x has a geometrical meaning: −x is the where µν . but rather by consistency of the susy transformation rules and equations of motion at the linearized level [46]. The relative factors 1/3 in (3. in components: Y m = −δµ m xµ sin x.

20).s. is zero upon using (3.22).18). / Nuclear Physics B 581 (2000) 179–239 We have now 3 different coordinate systems on S4 : the coset coordinates x. we obtain: . where ηI α = Ω αI .19) and (3. We shall mostly use the Euclidean coordinates Y A . (3. (3. Acting with D˚ λ on both sides of (3. (3. and the l.s. the r.h.19) It is easy to check that D˚ µ D˚ ν Y A = −g˚µν Y A .21) Another useful identity is D˚ µ Y A D˚ µ Y B + Y A Y B = δ AB . (3. The scalars Y A parameterizing the sphere in a 5-dimensional space satisfy the identity Y A YA = 1. We also give another proof because it illustrates the techniques we will use. is then zero trivially.18) It follows that Y A D˚ µ YA = 0 or Y A CAµ = 0. is constant. the Euclidean coordinates Y A and the stereographic coordinates ξ .h.192 H. but note that the explicit expressions for the Killing spinors (which we do not use) are simplest in stereographic coordinates.s.22) A and Y A ≡ C A form an orthogonal 5 × 5 matrix It follows by using that ∂m Y A = Cm 5 according to (3. (3. Nastase et al. and at x = 0.20) Hence we obtain the completeness relation: D˚ µ Y A D˚ ν YA = g˚µν or CµA CAν = g˚µν .21). (3. So the l.h.

D˚ µ Y A .  e = −δµA .

x=0 = 14 Tr Ciγ5 γµ γ A Ω .

 e = −δ A . = 1 Tr Cγ5 γ A Ω (3.23) Y A.

gives Tr(Cγρσ γ5 γ A1 A2 Ω). one may show that the r.h.s.24) AB = 0. e. From (3. using (3. the l. Using the definition of VµAB .h. both results are equal. (3. The proportionality constant is fixed by taking x = 0 for a particular case. and the r.22) and (3. At x = 0.22). A = 5.s.24) we find also: .s. we may derive: p  (3. Similarly. D˚ (µ Vν) using (3.20) and the antisymmetry in AB.23). x=0 4 5 This proves (3.g. gives e Evaluating the trace one finds that 4 g˚ µνA1 A2 and the r. satisfies the Killing vector equation.h. by namely. To prove this. This proves (3.25).h.25) 4 g˚ µνρσ ∂ ρ Y [A1 ∂ σ Y A2 ] = +η¯ I γµν γ5 ηJ γ A1 A2 I J . the l. gives at x = 0 − 12 gµν .23) also − 12 gµν .h.s. we act with D˚ λ on the last term and get +i η¯ I γ5 γµνλ ηJ (γ A1 A2 )I J . In the same way one may prove that VµAB = Y [A D˚ µ Y B] . B = ν.s. gives with (3. We obtain ˚ the p same result if we act with Dλ on the first term.

. .30) S4 where V4 is the volume of S4 . has to be an invariant tensor of SO(5) with 2 indices because the Haar measure d 4 x g˚ on S is 4 SO(5) invariant.38) p 6 The l.A5 dY A1 ∧ · · · ∧ dY A4 Y A5 = g˚ µνρσ dx µ ∧ · · · ∧ dx σ .20).19) and (3. .A5 dY A1 = 4 g˚ µνρσ dx µ ∂ ν Y [A2 ∂ ρ Y A3 ∂ σ Y A4 Y A5 ] . 5 (3. when D˚ λ hits the ∂Y ’s or Y A5 gives zero by antisymmetry. See also Appendix in [31].. the r. . p 5 g˚ µνρσ δ µ[A1 .A5 dY ∧ dY ∧ dY = 2 g˚ µνρσ dx ∧ dx ∧ dx ∂ Y Y p = −2 g˚ µνρσ dx µ ∧ dx ν ∧ dx ρ V σ [A4 A5 ] ..h.. / Nuclear Physics B 581 (2000) 179–239 ν VµAB VAB = 12 δµν . A basic identity which we will use repeatedly is given by p (3. should be proportional to the l.18).h. The constant of proportionality is found by taking a trace with δ AB .. and because of (3. This allows only δ AB . By repeatedly using (3..A4 δ5A5 + A2 .s..A5 .s. p A1 A2 A3 µ ν ρ σ [A4 A5 ] A1 . At x = 0. Contracting these identities with Y A5 . Integrals of an odd number of Y A ’s will give zero.36) A1 .h... on the r..  1 AC  AB 1 C 2 Vµ YB Y − A ↔ C = 2 Vµ . (3. ηαI = Ω αI .H. leads to a further chain of identities: p A1 .h.. Being constant.s. A and C A form an orthogonal matrix. (3.A5 dY A1 ∧ dY A2 ∧ dY A3 Y A5 = g˚ µνρσ dx µ ∧ dx ν ∧ dx ρ ∂ σ Y A4 . 5·7 (3. and the proportionality constant is fixed by looking at x = 0. Another proof It also follows from the fact that Cm 5 ˚ is obtained by acting with Dλ on both sides.A5 dY A1 ∧ dY A2 = 3 g˚ µνρσ dx µ dx ν ∂ ρ Y [A3 ∂ σ Y A4 Y A5 ] ..37) (3.. (3.s. µ B] . so from (3.23) we get on the r..27) (3.28) (3..32) the following further identities: p A1 .32) A1 . gives zero.35) p A1 A4 µ σ A5 (3.s. one derives from (3.29) From SO(5) symmetry 6 we deduce: Z p V4 d 4 x g˚ Y A Y B = δ AB .. δ σ A4 δ 5A5 ] = A1 .34) A1 . (3. In the same way we find: Z p  V4 AB CD d 4 x gY ˚ AY B Y C Y D = δ δ + δ AC δ BD + δ AD δ BC .A1 δ5A4 = A1 . .. VµAB VCD = Y [A Y[C δD] VµAB YA (3. p A1 .s. ..31) S4 and integrals of any even number of Y A ’s will be proportional to symmetrized products of delta functions in a similar way..h..A5 dY ∧ · · · ∧ dY = g˚ µνρσ dx ∧ · · · ∧ dx Y ..26) = 12 D˚ µ Y B . ∂ σ Y A4 Y A5 ] .33) p (3. (3. 193 n VmI J VInJ = −4δm . The l.A5 = 5 g˚ µνρσ ∂ µ Y [A1 ..A5 δ5A1 + · · · + A5 ..h.21). Nastase et al.

A5 dY A1 ∧ dY A2 Y A5 = Y A5 A1 . They have the correct mass-terms which ensure ‘masslessness’ in d = 7 AdS-space [55. .A + 48 me−1  αβγ δηζ δ AB Sαβγ . we consider the point x = 0 and use (3. 4 are spinor indices. µ ν g˚ µνρσ dx ∂ Y µ [A1 g˚ µνρσ ∂ Y [A2 ρ ∂ Y σ . R) rigid symmetry of the action is lost and replaced by the SO(5)g gauge invariance. of A A A σ δ 5 on the r.. namely maximal gauged sugra in d = 7.s. J.. .A5 dY A1 = = p p p g˚ µνρσ dx µ dx ν ∂ ρ Y [A3 ∂ σ Y A4 ] ..56].40) (3.. 4.43) γµ ηI VµJ K − 12 γµν ηI η¯ J γµν ηK = 4η(K Ω J )I .h. = 1. . . we obtain:  (3. we get: γ5 ηI φ5J K − γµ γ5 ηI CµJ K = 4η[K Ω J ]I − ηI Ω J K . The model has the following 0 fields: the vielbein eα a .. / Nuclear Physics B 581 (2000) 179–239 A1 . ... 4 are spinor indices. ∂ Y A3 A4 ] σ ∂ Y A4 ] (3. we turn to the main problem of this article.41). As a check on the normalization of the identities (3. while I. .39) . R)/SO(5)c but in the gauged model the SL(5. Nastase et al.B  − 1 ψ¯ α τ αβγ 5β ψγ − 1 λ¯ i τ α 5α λi − 1 m 8T ij − T δ ij λ¯ i λj + + + 2 2 8 ij ¯ α β α i j 1 1 ¯ 1 ¯ αβ 2 mT λi γj τ ψα + 2 ψα τ τ γ λ Pβij + 8 mT ψα τ ψβ  αβγ δ AB 1 ¯ − 2g αβ g γ δ γij ψδ ΠA i ΠB j Fβγ 16 ψα τ βγ α AB AB 1 ¯ 1 ¯ j τ γi λj ΠA i ΠB j Fβγ + 32 λi γ γkl γ i τ αβ λj ΠA k ΠB l Fαβ 4 ψα τ  im A + √ ψ¯ α τ αβγ δ + 6g αβ τ γ g δ γ i ψ Π −1 i Sβγ δ. and c for composite.A and the spin 1/2 fields λIi (vector-spinors under 0 SO(5)c .194 H. .42) from which we retrieve (3. the SO(5)g vector BαAB = −BαBA .h. For example. For completeness. (3. The scalars ΠA i parameterize the coset SL(5. the scalars 0 ΠA i . the 4 gravitinos ψαI . the l. Fierzing η(K (η¯ J ) ηI ). finding the complete nonlinear Kaluza–Klein ansatz. The dimensional reduction of the 11-dimensional vielbein transformation law With the tools developed in the previous section. j. = 1.. J 0 . = 1. .A5 δµ 1 δν 2 δρ 3 which agrees with 2µνρσ δ[A 4 A5 ] By Fierz rearrangements. . the antisymmetric tensor Sαβγ . The indices i.22) after multiplying with η¯ L .41) The last identity follows from (3. Derivation of the complete nonlinear Kaluza–Klein reduction 4. The model has a local SO(5)g gauge group for which A. .. . In particular.33) to (3.A Fδηζ. (3.23).s.. The action reads:   AB 2 e−1 L = − 12 R + 14 m2 T 2 − 2Tij T ij − 12 Pαij P αij − 14 ΠA i ΠB j Fαβ 2 A 1 + 12 Π −1 i Sαβγ .A 8 3 .1.35) yields A1 . .A5 Y A5 A1 . 5 are vector indices. 5 are SO(5)c vector indices and I 0 . satisfying γ i λIi = 0). we give below the action and transformation laws of the dimensionally reduced theory. (3.42) will be heavily used. = 1. we get other identities: fierzing η[K (η¯ J ] ηI ). . The subscripts g stand for gauge. (3. B.32).

A 20 3  + 12 m Tij − 15 T δij γ j  + 12 τ α γ j Pαij .A 4 3 im A − √ λ¯ i τ αβγ γ j λi Π −1 j Sβγ δ. 1 αβ 16 τ (4. Further. so Fαβγ δ. + 8 16 The local supersymmetry transformation rules are given by: (4.1) δeαa = 12 ¯ τ a ψα . so the the field strength of the gauge field is defined as AB = ∂ B AB + 2B AC B CB − (β ↔ α).A = − ] 8m √  i 3 δij ΠA j D[α 2¯ τβ γ i ψγ ] + ¯ τβγ ] λi − 4m √  i 3 B δAB Π −1 i 3¯ τ[αβ γ i ψγ ] − ¯ ταβγ λi . Π −1 i δΠA j = 14 ¯ γi λj + γ (4.. g = 2m (or rather g = 2mk but k = 1) and (2. and Dα = ∂α + Qα. However.A = 4∇[α Sβγ δ]. respectively.4)) are the symmetric and antisymmetric parts of A (Π −1 )i (δA B ∂α + gBα A B )ΠB k δkj . The curl of Sαβγ .2) ΠA i ΠB j δBαAB = 14 ¯ γ ij ψα + 18 ¯ τα γ k γ ij λk .1) with Tij = δij . The tensor Pα ij and the connection Qα ij (appearing in the covariant derivatives as ∇α = ∂α + Qα. (4. / Nuclear Physics B 581 (2000) 179–239 195  im A − √ ψ¯ α τ αβγ δ − 3g αβ τ γ δ λi Π −1 i Sβγ δ.A .3) √  i 3 BC ΠA i 2¯ γij k ψ[α + τ ¯ [α γ l γij k λl ΠB j ΠC k Fβγ δSαβγ . Nastase et al.. + Pα. .7) 0 0 0 The contractions over indices I 0 and J 0 are always as in ¯I 0 τ a ψαI and ¯I 0 (γ i )I J 0 ψαJ .H.A has strength 4. + 12  A ¯ j λi . 10 3 δλi = (4.5)  AB 1 1 mT τα  − 40 τ α βγ − 8δαβ τ γ γij ΠA i ΠB j Fβγ δψα = 5α  − 20  im A + √ τα βγ δ − 92 δαβ τ γ δ γ i Π −1 i Sβγ δ.. We put m = 1 most of the time. note that the limit m → 0 is singular Fαβ α β α β due to the factors of m−1 in front of the Chern–Simons terms.G Fδ Fηζ 16 3 m−1 −1 m−1 −1 e Ω5 [B] − e Ω3 [B].21) is the AdS7 solution of the field equations of (4.6)  AB γkl γi − 15 γi γkl ΠA k ΠB l Fαβ im A j + √ τ αβγ γi j − 4δi Π −1 j Sαβγ .4) (4.A . A B Here Tij = Π −1 i Π −1 j δAB and Ω3 [B] and Ω5 [B] are the Chern–Simons forms AB for Bα (normalized to dΩ3 [B] = (Tr F 2 )2 and dΩ5 [B] = (Tr F 4 )). (4.A 8 3 ie−1 BC DE + √  αβγ δηζ ABCDE δ AG Sαβγ . in (4.

(4. and requiring that the action for ψα and ψm becomes diagonal we find that A = +1 and p = ± 12 . we need an ansatz for the vielbein rather than the metric as we M is constructed as follows.10) where BαAB (y) is the 7-dimensional SO(5) gauge field and V µAB the corresponding Killing µ vector. one has Eαa (y. x)(γ5 ) where ψa = eaα (y)ψα (y.T C (4) . x) = eαa (y)1−1/(d−2) (y. The freedom in the signs of p and q corresponds to 7 More generally. 1(y. x) = ¯I 0 (y)U I I η¯ I (x) where ¯I 0 = IT0 C (7) and η¯ I = ηI. (4. needed to obtain the usual Einstein action in d = 7 without extra powers of 1. we proceed analogously to the vielbein and begin by introducing fields 9a (y. (4. x) + A 15 τa γ5 γ m 9m (y. x). The gravitinos 9M ≡ EM Λ Λ Λ 9a ≡ Ea 9Λ and 9m ≡ Em 9Λ (We suppress the 32-dimensional spinor index. x) = Bαµ (y. x)ηI (x) ψm (y. q = ± 12 (four combinations). x) = 11/10(y. . In order to did in Section 2.8) This is a standard result in KK reduction of theories with gravity. 0 ¯ (y.12) ε=1 1/10 q (γ5 ) ψm . x)U L L (y. x)(γ5)p ε(y. x) = I 0 (y)U I I ηI (x).196 H. / Nuclear Physics B 581 (2000) 179–239 At the nonlinear level. we will find that only p = q works. The natural extension of the rescaling in (2. x) = eαa (y)1−1/5 (y.13) where A. The inverse relations. However.16) (4. −1/10 ¯ (γ5 ) −p . x) = 1−1/10(y. x). let’s analyze the fermions. The ansatz for the vielbein EΛ fix the off-diagonal part of the Lorentz group and remain with local SO(6. x)(γ5 )p eαa (y) 9a (y. x)Eµm (y. x)ηm 0 (y. together with our ansatz are given by:  ψα (y. x) and 9m (y. x) 0 0 0 J J KL (x). It contains the gauge bosons Eαm (y. p and q will be fixed.14) −q 9m (y. x) 0 = ψαI 0 (y)U I I (y.3. (4. −1/10 (γ5 ) −p ε¯ = 1 . (4. x) ≡ det Eµm det e˚µm .15) (4. when dimensionally reducing a D-dimensional supergravity theory to d dimensions. is: 7 Eαa (y. The factor of (−2) in the ansatz of Bα comes from g = 2m. 1) × SO(4) invariance. x) = −2BαAB (y)V µAB (x). and the minus sign is related to the sign convention in the definition of the covariant derivative Dα = ∂α + gBα . The ansatz for Eαm (y.17) Substituting this ansatz for 9a and 9m into the 11D action. (4.) In order to obtain diagonal kinetic terms for the fermions in d = 7.11) 9m = 1 (4. Nastase et al.9) Bαµ (y. (4. Λ 9 split into Before we do that. we impose the gauge choice Eµa = 0. x). by trying to match the 7-dimensional transformation laws. x) is also standard. x)U K K (y. x) as follows: 9a = 11/10(γ5 )−p ψa − A 15 τa γ5 γ m 11/10(γ5 )q ψm . The only nontrivial step is finding an ansatz for Eµm . = λJ 0 K 0 L0 (y)U J (y.27). x). x) = 1 −1/10 (y.

12) is the only one which diagonalizes the action for the seven-dimensional gravitini and spin 1/2 fermions. this proves that U is charge conjugation matrix (see Appendix A. If we try the more general rotation: 9a = A(γ5 )p ψa + Bτ a γ m (γ5 )q ψm . One can either rotate the label index I or the spinor index of the Killing spinors. / Nuclear Physics B 581 (2000) 179–239 197 a rescaling by γ5 . ψm and  is a local (x and y dependent) SO(5) matrix in the spinor representation depending on the scalar fields in 7 dimensions (ΠA i ). Since in [46] only the leading fermionic transformation laws were studied.22). Various authors [2. the d = 11 SU(8) invariance is fixed. −1/2 in ψα is needed to cancel the γ5 coming from 0 αβγ = In the field redefinitions.6. This freedom will be fixed in the following by the requirement of consistent truncation. (4.30)).1) and Ω . Therefore. 9m into ψa .H. Since δ9Λ = DΛ ε+ more. since the fermions have an SO(5) composite index. As we shall see in (4. (See for instance (4. We choose the former. 9m = C(γ5 )r ψm + Dγm (γ5 )s τ a ψa . We will find that only p = q = −1/2 works.11) and (4. Nastase et al. When one compactifies.18) by requiring a diagonal kinetic action for the gravitino and spin 1/2 fermions we recover (4. We also note that the rotation in (4. ψm is needed to cancel the 1/2 mixed terms 9α 0 αβm Dβ 9m .137) below. whereas the Killing spinors have naturally an SO(5) gauge index. only the 4D part remains unfixed. y). The U matrix in the expansions of ψα . (De Wit and Nicolai took the alternative rotation of spinor indices by U because they reformulated d = 11 sugra in a form with local SU(8) invariance parameterized by an arbitrary U (x. Since Ω is the e = −Ω −1 . 1−1/10 is needed to bring the gravitino action to the τ usual Rarita–Schwinger form with no extra powers of 1 (which would come from det E = det e(7)11−7/5 det e˚(4) ). γ5 αβγ ⊗ γ5 in the gravitino action. e (so that U ΩU (4.) Incidentally we note that in the linearized ansatz used in [46] the other sign was chosen. and we shall find the same need (see below (4. x) by a gauge factor. x) which acts on the Killing spinors. it did not matter at that point which rescaling by γ5 was used. and in d = 11 there are only spinorial indices are available. the gauge fixing produces also for us a given field-dependent matrix U (y. which is perhaps more natural. consistency of our results for the transformation rules requires the matrix U to satisfy the condition  0 eI 0 J 0 → Ω e · U T · Ω II 0 = −(U −1 )II 0 eI J U J 0J = Ω UI IΩ e T = Ω.11) and (4. e U T ΩU e = Ω).7] found it necessary to add a U matrix for consistency of the truncation of the susy transformation laws. There is a difference between the gauge U = 1 which gives the usual d = 11 sugra (in the triangular gauge Eµa = 0) and the gauge needed for S7 compactification. and this would modify the y dependence of U (y. the rotation of 9a . For the linearized case we reobtain (2.12).19) The need for this relation will occur in various other places as well. we need factors γ5 and 1−1/10 in ε. finally.) In d = 7 one can always make SO(5)c gauge transformations on the fermions.31).

19) appearing.198 H.29). (4. We need the compensating Lorentz transformation in (4. an USp(4) matrix as it is explained in Appendix A.25).21) We can now require that we get the correct graviton transformation law in d = 7. we have introduced the notation δ(d = 11) for the d = 11 susy transformations and δ 0 for the intermediate transformation.25) with −2VAB and equating the first term in (4. Using (2. The result is given in (4. But in 7 dimensions.3): A  B  ¯ α γ k γ ij λk .22) Since according to (3. for instance: µ m Ω a.24) 2 5 where we used (A.2). Adding a term 15 τ ab γ m 9m eαb inside the brackets. x) = δ(d = 11) Eαm Em     1−1/5 µ γ mγ n m mn Em −¯ γ ψα − i ¯ τα δ + =i γ5 ψn . we see the condition (4. we obtain: δeαa = 12 ¯ τ a ψα = 12 ¯I τ a ψαJ Ω I J = 12 ¯I τ a ψαI . .3) η¯ I ηJ equals Ω I J . Because this is not yet the final form of the vielbein law. we get from (4. (4.9) in the last step. (4.23) This is the correct 7-dimensional transformation law of the d = 7 vielbein. δBαAB (y) = Π −1 i Π −1 j 14 ¯ γ ij ψα + 18 τ (4. δEaµ = δSU SY Eaµ + Em etc. 1) µ Lorentz transformation since Bα has no Lorentz index. / Nuclear Physics B 581 (2000) 179–239 a SO(5) matrix in the spinor representation.9). (Namely. 8 The transformation law for the gauge fields follows from (4. (4.20) This Lorentz transformation acts on all vielbein components. We will determine the ansatz for Eµm by matching δBα[AB] with the expression in 7-dimensional gauged supergravity. but there is no SO(6. One finds:    µ µ + Eαm Eaµ Ω am + Eαa Ωam Em δBαµ (y. we have from (4.12). 1) Lorentz transformation characterized by an antisymmetric matrix Ωm a in order to stay in this gauge: δEµa = 0 = δSU SY Eµa + Eµm Ωm a ⇒ Ωm a = − 12 ε¯ 0 a 9m = −Ω am . we add a field dependent SO(6. In order to obtain agreement with (4.25) µ By multiplying (4.) Let us now return to Eµm . Excluding 3-fermion terms.24) and (4. one obtains: 8 The d = 7 susy results are thus a combination of d = 11 susy and d = 7 Lorentz symmetry. (4.8):    µ δ(d = 11)Eµm Eαa + Eαm Ωm a δ 0 eαa = 11/5 δ(d = 11)Eαa + 15 Em  −p  = 12 1−1/10 ¯ γ5 τ a γ5 9b eαb + 15 γ m 9m eαa 0 0 = 12 ¯I 0 τa ψαJ 0 U I I U J J η¯ I ηJ + · · · .20) on both vielbeins in (4. Nastase et al.2.9). the d = 7 Lorentz transformations do not show up anywhere else. Since we are imposing the gauge Eµa = 0. 1) rotation to the d = 11 transformation laws. we need a compensating SO(10.

of (4.H.8) and used expression of ηm + (4.h. of (4. 5 Vm Ω J KL in (3.26):  A  B µ i µ = 11/5 Π −1 i Π −1 j VAB Tr γ ij U Vm U T Ω . Contracting with VνAB and Eµm and using (3. 4 As a check on our results developed so far.24). We already know at this µ point Em .s of (4.28) Em 4 e implies that U T ΩU = Ω as follows from eliminating U T . However. the ψm term in the d = 11 transformation law of the gauge field is: 0 0 µ −1/5 0 1 ¯I τα λJ 0 K 0 L0 U I I U J J U K δBαµ |ψm term = 32 Em  × φ5I J (C m )KL − φ5KL (C m )I J . of (4.42). (4.29) Eνm = 1−1/5 Tr γij U V m U T Ω ΠA i ΠB j VνAB . We find:    B µ  I 0 J 0 −1 A 1 Π −1 j VAB Tr γ ij U Vm U T Ω U V m U T 4 Π i A  B µ I 0 J 0 = Π −1 i Π −1 j VAB γ ij . Thus. we note that substituting (4.h.27) one should get an identity.27). If one sets U = 1 in (4. we move the factors Π −1 and γ ij in the r. Nastase et al.26).26) Dropping a common factor 1/2¯I 0 ψαJ 0 .31) the Fierz relation where we have substituted the (3.30) and removes the factors Π −1 (y) one is left with an x-dependent equation for the Killing vectors which is incorrect.19) and (3.30) 0 The matrices U I I are really needed for the consistency of (4.h. = 2 Π Π i j 199 (4. When U is y-dependent.s.s.) e T =Ω (We used that U ΩU To obtain an explicit expression for Eµm . we find the equation: I 0 J 0 A  B µ I 0 J 0 µ = −11/5 Π −1 i Π −1 j VAB γ ij . For this purpose we work on the combination of spherical harmonics on the r. U V mU T iEm (4.h.24) is proportional to VµAB .31) will contribute because λ is symplectic traceless.28) into (4.I J = η¯ I γ m ηJ . hence first we should make sure that the r.25) of the gauge field. / Nuclear Physics B 581 (2000) 179–239 i1−1/5 0 µ I m J I0 −Em η¯ γ η U I U J J ¯I 0 ψαJ 0 2    0 0 −1 A −1 B µ ij I J 1 VAB γ ¯I 0 ψαJ 0 . one cannot factor off the Π −1 and the relation becomes a condition on U .27) to the left. 0 KU L0 L (4.45). We should now check if with our ansatz for the vielbein in (4.32) . They can be rewritten as follows   KL I J KL − Cm φ5 η¯ I δ mn + 15 γ m γ n γ5 ηnJ KL = 3 φ5I J Cm 24 I [L K]J − 65 VmI J Ω KL . (4. when projected on λJ 0 K 0 L0 only the first two terms on the r. It will follow from the fundamental condition on U which we obtain in (4.s.28) we also reproduce the second term in the transformation law (4. we find:  i (4.27) where V m. µ This equation can be solved for Em by multiplication with (U V n U T )I 0 J 0 and using (4.

. Next. .39) is the starting point for obtaining the nonlinear metric ansatz for the compact dimensions.34).32) and a complicated term of the form (ηγ ¯ mn γ5 η)KL CnI J .. .19). we square (4.µ4  µ1 .40) Π −1 j ¯ γ i λj + ¯ γ j λi YA YB . . ∂ν4 Y B4 TA−1 ... Substituting the 7D transformation law of the scalar fields in (4. it follows from (3. ..37) Π −1 i Π −1 j Π −1 i Π −1 j .ν4 ∂ν1 Y B1 . (4. Using Y A = 14 φ5I J (γ A )I J and CµB = µ ∂µ Y B .35) − 14 Π −1 i Π −1 j ¯ τα γ k γ ij λk VAB and this should agree with (4. . To remove the dependence on U . . The two terms are equal if we assume the following normalization of the seven-dimensional spin 1/2 fields: J 0 K 0 λI 0 J 0 K 0 .. (4.25). .36) λkI 0 = 3i γ k At this point we have learned that if (4. ∂µ4 Y A4  ν1 . we compute 1 = det(Eµm )/ det(e˚µm ). 2 Π i . 0 0 which however vanishes since λJ 0 K 0 L0 is antisymmetric in K L . 1−2/5 g µν = 2VAB VCD The matrix g µν is the inverse of the metric gµν = Gµν = EµM EνM = Eµm Eνm .39) we get:  A  B δ 1−6/5 = δT AB YA YB = 2δ Π −1 i · Π −1 j δ ij YA YB   B  −1 A = −1 (4.B5 YA5 YB5 TA−1 . Using (3.42) contracted with η¯ I γm .15) for ψm . TA−1 = gT ˚ AB YA YB 1 B1 5 B5 to conclude that 1−6/5 = Π −1  A i Π −1  j B ij δ YA YB ≡ T AB YA YB . gµ4 µ0 116/5  µ1 .. TA−1 1 B1 4 B4 = g1 ˚ 16/5 A1 .27) is correct then the transformation of the gauge field BαAB follows from Kaluza–Klein reduction.27) we arrive at δBαµ |ψm term = −3i Π −1  A i Π −1  j B γ ij J 0 K 0 0 0 µ Ω I L ¯I 0 τα λJ 0 K 0 L0 VAB ..39) Eq. but there are still the extra U matrices. / Nuclear Physics B 581 (2000) 179–239 where we have used the ansatz in (4.µ4 ∂µ1 1 4 × Y A1 . We get J 0 K 0 0 0 µ −1/5 0 1 ¯I τα λJ 0 K 0 L0 Ω I L U V m U T . Nastase et al. The 7-dimensional transformation law is equal to A  B µ (4.34) µ Since we have arrived at an expression proportional to VAB we can compare with (4. By starting with (3.36) and the fact that det T = 1 we can evaluate directly the determinant of gµν = 14/5 ∂µ Y A (T −1 )AB ∂ν Y B : −1 µ A ν B ∂ Y ∂ Y det gµν ≡ det TAB 0 0 ≡  µ1 . (4. .200 H... (4.µ4 gµ1 µ0 ..24) that this expression is proportional to the Killing vector VAB .27). we obtain: A  B C  D µ ν (4. It is µ µ µ given by g µν = Em E mν and differs from Gµν = EM E Mν by a term Ea E aν .A5  B1 .33) δBαµ |ψm t erm = −3Em Using (4.38) (4. we obtain the second line in (4. . Using (4.. (4. .

30):   B  I 0 J 0 1 −1 A Vm CD V m EF .39) we arrive at C  D δ Π −1 i · Π −1 j δ ij YC YD  0 0  0 0 e T I J U γ C ΩU e T KL. At this point we have come in 7 dimensions as far as de Wit and Nicolai in their first approach without a local SU(8).h.30).29) to sum over the Killing vectors. Splitting (U γ ED U −1 )I J into (U (γ E γ D −δ ED )U −1 )I J . Nastase et al.s. On the other hand. the U -matrices must satisfy:  0 0   0 0 e T I J = ±13/5 Π −1 A γ i I J YA . The proof that (4.19) and applying (4.30) follows from (4. so that (4. (4. a relation corresponding to our (4. (4. YA U γ A ΩU i (4.H.45). but they did find the one corresponding to (4.45) once we obtain:  B   I 0 J 0 . (Actually. of (4. but whose spherical harmonics are not yet manifestly the same. γ CD U −1 U γ ED U −1 In the product of gamma matrices only the term δ AC γ D survives due to symmetry 0 0 0 0 arguments.30) goes as follows.45) appears twice. we have found solutions to (4.44) In order that this result for δΠ −1 agrees with (4.45) implies (4.s. Secondly.45) will turn out to be the most important tool in our endeavor to find a consistent truncation of the 11-dimensional supergravity. / Nuclear Physics B 581 (2000) 179–239 A 201 B where we recall the definition TAB = (Π −1 )i (Π −1 )j δ ij and used (4. of (4.41) by using the Fierz relation (3. Substituting (4.43) δ 1−6/5 = −3i1−6/5YA YC ¯I 0 λJ 0 K 0 L0 U γ A ΩU There are four U matrices.36) to remove the spinors.AB Tr γ ij U γ CD U −1 U γ EF U −1 4 Π i where one substitutes (3. First of all. the r.h. because we want that our matrix U is equal to 1 on the background (and not to −1).42) η¯ I γm γ5 ηm Again. from 11D sugra we obtain  µ · Eµm = − 35 1−6/5 ε¯ 0 m 9m .47) − 14 1−3/5 Π −1 i Cµ B YA YC YE Tr γ i U γ A .46) Π −1 j Vµ. We rewrite the spherical harmonic on the r. we have been able to show that (4. we get: .45).) We have been able to find solutions for U without having to extend the d = 11 theory. (4.s.45) where we used the normalization relation (4.AB = Y [A CµB] and writing 2γ ij = γ i γ j − γ j γ i .42):  J KL = − 5φ5I J φ5KL + Ω I J Ω KL − 4Ω I [L Ω K]J .5). One begins with the l. Equation (4. eliminating the first U −1 by (4. and using (4.41) δ 1−6/5 = 65 1−6/5 δEm Again we find two expressions which should be equal.45) is not present in their work. We will choose the minus sign.h. (4. (4.40). three from ψm and one from . = −i 32 1−6/5YA YC ¯I 0 λJ 0 K 0 L0 U γ A ΩU (4. but now in the opposite direction. only the terms with φ5 φ5 contribute because λI 0 J 0 K 0 is symplectic-traceless and one finds (φ5 ∼ YA γ A ):   0 0  0 0 e T I J U γ C ΩU e T KL.45). Using Vµ.45) is the crucial equation. of (4.

55) implies the equations: NA (YA − vA ) = 0. then from (4.56) automatically. (4. (4.30) (with γ ij written again as γ i γ j − δ ij ). / Nuclear Physics B 581 (2000) 179–239 1 4 Π −1 − 1 4  A i Π Π −1  −1 A i  eI 0 J 0 Cµ A YE 4δ ij Ω  E  I 0 J 0 . 1+v·Y 2(1 + v · Y ) (4. Π −1 j Cµ A YE Tr γ i U γ D U −1 γ j U γD U −1 E j (4. If we take N 6= 0. (4. U −1 = −ΩU (4. Finally. 2 2 (4. says that e T Ω = N + NA γA − NAB γAB . of (4.53) On the other hand.h. From (4. 1−v·Y (4. γA . (4.57) we get that N= 2NAB vA YB . (4.51) NA NAB = 0. (4. Eq.52) 2NNA = ABCDE NBC NDE . N 2 + NA2 + 2NAB (4. the condition that it is an SO(5) matrix in the spinor representation.57) N[A (Y + v)B] = 12 ABCDE (YC − vC )NDE . we look at the solutions of equation (4. (4. γAB as follows: U = N + NA γA + NAB γAB .45) rewritten as U Y/ = v/U. so we will disregard the solutions which are equal to Y/ on the background.59) whereas by multiplying (4.50) Unitarity of U implies that N .49) Then. A vi = Π −1 i YA 13/5 .45) for the matrix U . The fact that U U −1 = U −1 U = 1 gives then the conditions: 2 = 1.58) with YB + vB we get NA = ABCDE vB YC NDE YA + vA + N0 .s.56) N(YA − vA ) = −2NAB (YB + vB ).52) we obtain after contraction with YB : . (4.202 H. (4. NA and NAB are real. This satisfies (4.12)) in order to get directly the r. But we want solutions which are equal to 1 on the background.19). Nastase et al. The 4 × 4 matrix U can be expanded into the complete basis 1.60) where at this point N 0 is arbitrary.58) The first observation is that if U is an SO(5) solution of U Y/ = v/ U then also U Y/ is a solution.48) What is left to do is the sum over γ D (use (A.54) v = Y = 1.

besides YA and vA . Z2A .62) (Notice that if N = 0 one divides in (4. SA = √ 2(1 + v · Y ) YA − vA DA = √ 2(1 − v · Y ) (4. another 3 independent vectors Z1 .66) and (4.71) D[C 0 ND 0 E 0 ] + 4SD ND[E 0 SC 0 DD 0 ] = 0.64) (4.70) satisfying S 2 = D 2 = 1. The SO(5) conditions become now constraints on NAB :   YG (YA0 + vA0 )NA0 F = 0.65) where the last condition fixes the normalization. also orthogonal to Y and v. (4. 1−v·Y (Y + v)[F 1 A]BCDE vB YC NDE = AF CDE (YC − vC )NDE . = −2(YA + vA ) NA00 B 00 vA00 YB 00  ABCDE NDE 2 = 1. Z2 . Z1A . / Nuclear Physics B 581 (2000) 179–239 2 ABCDE vB YC NDE NAF YF . Z3A }. (4.69) (4.61) by zero.67) The most general solution of these remaining five equations can be obtained by noting that in 5D space we have. v A . We can then make out of them an orthonormal set.67) with AF C 0 D 0 E 0 ): (S · N · D)DA + NAB SB = 0. S · D = 0.68) m6=n=1 where {XiA } = {Y A . so the case N = 0 should be treated separately.66) (4. 1+v·Y 2 (4.72) .. The U Y/ = v/ U conditions yield further constraints on NAB : NBC vB YC (YA − vA ) + NAB (YB + vB ) = 0. have N 6= 0) are parameterized by an antisymmetric matrix NAB .67) yield (after multiplying (4. N 2 + NA2 + 2NAB (4. NAB vA YB   FBCDE vB YC NDE (YA + vA )NF G YG δAF − NA = .e.) The solutions which are equal to 1 on the background (i.61) (4. We define YA + vA . 1+v·Y NA0 B 0 vA0 YB 0 N0 = − 203 (4. Nastase et al. Z3 .63) ABCDE vB YC NDE NAG δF G − NA00 B 00 vA00 YB 00  1 − (v · Y )2 NBC − 2vB YC 2(NA00 B 00 vA00 YB 00 ) FBCDE vB YC NDE NF G YG .H. (4. (4. Then (4. Then the most general solution for NAB will be written as NAB = 5 X [A B] amn Xm Xn .

 u= Y −v m   m 2 −1/2 p (Y m − v m )2 arctan − . because ABCDE v A Y B D C XD = 0. Nastase et al. We will now look at particular cases. now we can take X to be a noncovariant vector.65) fixes the normalization. Z3 out of 3 suitably chosen (z) Xi ’s (z 6= 0. so that N = a(1 + v · Y ) and NA = 0.75) Thus the most general solution contains 3 parameters. XA = aSA + bi ZA (4. (4. 0. Y 5 + v5 (4. SA .74) The normalization condition (4.63) and (4. (Y5 + v5 )2 + (Ym − vm )2 m = 1. v) plane we reproduce U(1) .64) are satisfied trivially. The normalization condition (4. Then (4. (4. / Nuclear Physics B 581 (2000) 179–239 Substituting (4. and any component of XA along DA will drop out of NAB . Then γ5Y/ + v/γ5 Unoncov = √ √ 2 1 − v · Y + 2Y5 v5 Y5 + v5 − iγm (Ym − vm ) =p . However. The unique covariant solution built out of only YA and vA then reads: U(1) = 1 + v · Y + vA YB γAB 1 + v/Y/ √ =√ . Then NAB = av[A YB] . We can easily check that if X (Y.77) It clearly satisfies U Y/ = v/ U and is and USp(4) matrix.76) Xi = Y ·Tz ·Y i . 0.79) which can be written as an exponent    U = exp − iγm Y m − v m u . then the X(z) ’s will generate the whole sphere (they will not where z ∈ Z and if ΠA i 6= δA lie in a lower dimensional hyperplane). 1). which we can take to be X = (0. 1). 0.78) U(X) = ± √ √ 2 1 − v · Y + 2X · Y X · v 2 1 − v · Y + 2X · Y X · v E is in the where X now is still arbitrary.72) implies D[C 0 ND 0 E 0 ] = 0. Z2A . We note that the general solution (4. whereas (4. so we can build Z1 . Z1A .73) with XA an arbitrary vector. The simplest possibility is to build U only out of Y A and v A . Z2 . but has unit norm. 2(1 + v · Y ) 2(1 + v · Y ) (4.80) . We notice that DA .65) gives X bi2 + a 2 v · Y = 1. (4.73) can be rewritten as X · (Y + v) + XA (Y B − v B )γAB X / Y/ + v/X / =√ √ . of which the most general solution is NAB = D[A XB] (4.71) into (4. Z3A make an orthonormal set.51) fixes then a = ±(2(1 + v · Y )−1/2 . We note that the covariant vectors at our disposal are of the form: z A [ Π −1 ]i YA (z) . 4.204 H. so that we have i . (4.

Gαµ dx α dx µ = 214/5TAB while the metric of the 7-dimensional space–time is (4.83) = 1 + δvA YB γAB + Y m Cm 2 2 with α1 parameterizing the deviation from covariance. We note that taking α˜ to infinity. Since (Π −1 )i = A A (z) (z) (e−δπ )i . one easily checks that Gµν g νρ = gµν g νρ = ρ δµ . we get the maximally noncovariant solution Unoncov . =√ p U(X) 2 1 + v · Y + 2α(Y ˜ 5 + v5 ) + α˜ 2 (1 − v · Y + 2Y5 v5 ) (4. And so the only vectors at our disposal are Y A and δv A .81) to obtain 1 + v/Y/ + α(γ ˜ 5Y/ + v/γ5 ) noncov . where δvA = YA Y · δπ · Y − δπAB YB . Squaring this result would then lead to an expression for the metric gµν . because of the linearity of the equations we can add the arbitrary noncovariant piece coming from the U Y/ solution. the vectors Xi = [(Π −1 )z ]i YA /Y · T z · Y become Xi ' Y A − zδvA . gµν = 14/5Cµ A Cν B TAB g µν =1 2/5 µ CA CBν T AB YC YD T CD  µ − CA YB T AB CCν YD T CD .H. Actually the most general linearized A δv B γ Y solution is found by adding also a term βY m Cm AB / . / Nuclear Physics B 581 (2000) 179–239 205 We can also generate the most general noncovariant solution made out of only Y A and by choosing vA X= p E 1) Y + βv + α(0. we see that the metric gµν = Gµν describes an ellipsoid with a conformal factor 14/5 .82) where α˜ = α/(β + 1) is an arbitrary parameter.84) (4. It is independent of U and has simple geometrical properties as we shall see.28) and (4.37): U '1+ −1 . Recalling that CµA = Dµ Y A = ∂µ Y A . Nastase et al. −1 C BC Y B dY A . 1 + β 2 + α 2 + 2αβv · Y + 2αY5 + 2αβv5 (4. the metric of the internal space reads: −1 ∂µ Y A ∂ν Y B dx µ dx ν ds42 ≡ Gµν dx µ dx ν = 14/5TAB −1 = 14/5TAB dY A dY B .85) Since we are in the gauge where Eµa = 0. From (4.45) one can in principle obtain an expression for the vielbein. time are determined by the eigenvalues of TAB To summarize. so the most general solution for U which is equal to 1 on the background is obtained from noncov : U(X) 1 (δvA YB γAB + α 0 δvA γA5 ) 2(1 + αY ˜ 5) α1 1 A δvB γAB (4. It is easier to obtain the result for gµν by directly verifying that it is the inverse of g µν in (4. We end this section with a discussion of the metric.86) . A Maybe it is interesting to analyze what happens at the linearized level. whose axes at a specific point y in the d = 7 space −1 . (4.

(4. viz. √    p 1 2 T Fµνρσ = µνρσ det g˚ 1 + − 5 3m 3 YA YB T AB   2 YA (T 2 )AB YB − 1 (4. Then we can write the 11-dimensional metric in a concise form: 2 ≡ GΛΠ dx Λ dx Π ds11   −1 = 1−2/5 gαβ dy α dy β + 14/5TAB dY A + 2B AC YC dY B + 2B BD YD . We begin by presenting the final form for our ansatz for the KK reduction on S4 of the field strength FΛΩΠΣ . the gauge covariant derivative is dY + 2B · Y ≡ DY . the gauge fields and the scalars.91) E µm = 14 Π −1 i CA CBm 1−2/5 Tr U −1 γ i U γ B . 4.111). For the derivation of this result one may use (4. we rewrite the vielbein and its inverse in a simpler form. The ansatz for the 11-dimensional antisymmetric tensor and auxiliary field and self-duality in odd dimensions At this point we have obtained the complete nonlinear ansatz for the metric and the gravitino and we checked the transformation rules of the d = 7 bosonic fields which are embedded in the d = 11 vielbein.93) + g˚ µνρσ C α D 3 YA T AB YB (YA T AB YB )2 √   EF Y  2 2 F AB C D T Fµναβ = ∂[α ABCDE Bβ] Cµ Cν 3m 3 Y ·T ·Y  EG Y  G AF BC D T (4. since after compactification the gauge coupling is g = 2m.2. Nastase et al. / Nuclear Physics B 581 (2000) 179–239 −1 BA ds72 ≡ Gαβ dy α dy β = 1−2/5gαβ dy α dy β + 414/5YA TBD B YA B DC YC . Thus the metric is manifestly gauge invariant: −1 2 = 1−2/5 gαβ dy α dy β + 14/5 (DY )A TAB (DY )B . ds11 (4. the graviton. (4. and we set m = 1.206 H. We now turn to the antisymmetric tensor AΛΠΣ .90) Eµm = 14 12/5 ΠA i CµA C mB Tr U −1 γ i U γB . A µ  (4.88) Note that.92) − 3 (YA T AB YB )2 √  EF Y  2 F AB C D T Fµνρα = ∂[µ ABCDE Bα Cν Cρ] 3m Y ·T ·Y   p  CAσ ∂α T AB YB T AB YB CD − ∂ T Y Y (4.89) For later use for the gravitino transformation law. a discussion of how we arrived at this result will be given afterwards.94) + 2∂[µ ABCDE B[α YF Bβ] Cν] Y ·T ·Y .87) where GΛΠ denotes the 11D metric. involving only the conformal Killing vectors:  (4.

α8 (fµνρσ Fα1 .97) where k is an arbitrary scale factor which will be fixed below (see below (5..H. (4.32). fµνρα = µνρτ D τ Dα 21 f and will not contribute in the 11-dimensional Chern–Simons action F FA..98) which vanishes if we partially integrate the Dα1 in fµνρα1 . In principle. This eαβγ is needed to convert the second-order field equation obeyed by mixing of Aαβγ with B Aαβγ into the first-order one obeyed by the antisymmetric tensor field of 7-dimensional gauged sugra.96) The independent fluctuation.α8 − 32fµνρα1 Fσ α2 α3 α4 Fα5 . Aαβγ .. Since they contain the same spherical harmonic Y A they will mix.A Y . It is the same as the geometrical ansatz of [57. since the ansatz proposed by the authors of [57.α8 ) (4.58] if the scalars are set to zero. eαβγ will be written as Aαβγ ∼ Sαβγ .. It must reproduce the linearized term in (2. if we work for simplicity in 12 dimensions (where the Chern–Simons form becomes F F F ).95) + ∂µ ABCDE ∂[α BβAB + 43 B[α ] YE √  2 T EH YH DG Fαβγ δ = 4∂[α Aβγ δ] + 4∂[α ABCDE 43 BβAB BγCF YF Bδ] YG 3m Y ·T ·Y   CD E AB 4 AF FB + ∂β Bγ + 3 Bβ Bγ Bδ] Y .A (y)Y A (x) and Bαβγ ∼ Both Aαβγ and B AB B T Sαβγ .. Sαβγ . The precise expression of the ansatz in the Fµνρσ sector is highly constrained. use the Bianchi identity. All the other components of BΛΠΣΩ are set to zero because in d = 11 they vanish on-shell. However.. must be of the form fµνρσ = µνρσ f . they would have to be proportional to d = 7 field equations.A .α4 Fα5 . Fµνρα sector. We note that any closed form f(4) which appears only in the Fµνρσ . as we shall show. The two terms in Fµνρσ and Fµνρα which only depend on scalar fields are components of a separately closed form. in order that they also vanish on d = 7 on-shell. the other components of BΛΠΣ could still depend on the field equation for Sαβγ because that one is linear in derivatives. then f(4) contributes only to a term of the type  µνρσ  α1 . This is so because. so that.31). mixes with the auxiliary field Bαβγ δ = kαβγ δ ηζ B˜ηζ . The ansatz for FΛΩΣΠ is obtained by requiring consistency of susy laws. One would expect that this should be the case. this produces incorrect answers for the susy transformation rules of the fermions. and it must yield the correct scalar potential in . Nastase et al. / Nuclear Physics B 581 (2000) 179–239 207 √   4 2 T EH YH AB CF DG Fµαβγ = ∂µ Aαβγ + ∂µ ABCDE B[α Bβ YF Bγ ] YG 3m 3 Y ·T ·Y   EG YG D T − 2∂[α ABCDE BβAB BγCF ] YF Cµ Y ·T ·Y   AF FB Bβ BγCD (4. Field equations of the form ∂ 2 φ are ruled out because they would lead to (∂ 2 φ)2 terms in the action. and then partially integrate back the resulting Dσ . (4..58] was constructed such that the Chern–Simons terms in the 7-dimensional action are obtained after integrating the Chern–Simons term of the 11-dimensional action on S4 ..

99).39) to convert 1 into action. let us consider a term of the 9 We used a symbolic manipulation program to obtain (4.) The function S must be homogeneous of degree zero in T . In fact. To be more explicit.100) T AB . Furthermore. In order to perform this integral to which both the Einstein action and the kinetic action of the 3-index photon contribute we start with the d = 4 scalar curvature associated with the conformal metric g˜ µν = 1−4/5 gµν . where √3 (det g(x)) ˚ Fµνρσ µνρσ (1 + S) = Fµνρσ . Nastase et al. a linear combination of T 2 and Tr(T 2 ): (−31/70) Tr(T 2 ) + (23/70)T 2 . (YA YB T AB )2 and its relation with the d = 4 scalar curvature  e(4) − 6g µν Dµ Dν ln 1−2/5 R (4) = 1−4/5 R   − 6g µν Dµ ln 1−2/5 Dν ln 1−2/5 .208 H. the most general expression S can have is     T Y T 2Y − 5 . the integral over the compact space of the Einstein action. = (4. when setting the gauge fields to zero. namely. S A depends only on the scalar fluctuations in (Π −1 )i .104) − 1 + b S =a (Y T Y )2 YTY because higher powers of T appearing in the ratios would yield terms proportional to Tr(T −1 ) in the 7-dimensional action.) At this point. namely: 9 e(4) = g˜ µν R λ R λµν −2YA YB (T 3 )AB + 2T YA YB (T 2 )AB + YA YB T AB (Tr(T 2 ) − T 2 ) .101) (YA YB T AB )2 We further use that Z p YA YB 1 ˚ = T −1 Vol(S4 ). because in d = 7 the potential is proportional to T 2 − 2Tij T ij = T 2 − 2TAB T AB and the leading term (YE YF T EF )2 has already two T’s. and adding the other contributions from the Einstein Using (4. and disregarding terms with d = 7 space–time derivatives yields: Z p −1 ˚ d 4 x det g(x) 2   YA YB (T 3 )AB YC YD T CD − (YA YB (T 2 )AB )2 −6/5 e(4) R +2 × 1 . / Nuclear Physics B 581 (2000) 179–239 d = 7 after integrating over the compact space. TAB TCD + TAC d 4 x det g(x) EF 2 5·7 (YE YF T ) (4. . the integrated Einstein action contribution is of the desired form.102) d 4 x det g(x) YC YD T CD 5 AB Z p YA YB YC YD Vol(S4 ) −1 −1 −1 −1 −1 −1  ˚ = TBD + TAD TBC . On the other hand. the integrated kinetic action of the 3 index photon has the form 2 1/2  ∼ (YE YF T EF )2 (1 + S)2 .99) (4. (Use that 2 (det g µν )(det g˚µν ) = 1−2 and substitute (4.39). (4. (4. (4.103) (This becomes clear after diagonalizing T CD .

being integrated over S4 . containing the d = 11 nonlocal term D˚ σ 21˚ Dα (1 + S) would lead to the correct d = 7 action. It will turn out that the consistency of the gravitino transformation law excludes the first solution (a = 0).92) has no linearized contributions. (4. (4. and requires the second (a = −2/3).92) yields the results of [46]. while the coefficient of the second term gets fixed to 1/3 by requiring agreement with [46]. we keep only the gravitino terms.102). the ansatz for Fµνρα was found to be truly nonlocal already in a perturbative expansion. in fact.92). we get: √ . With our present ansatz (a = −2/3). ˚ 2 where 1 + S contains the ansatz for Fµνρσ in (4. One requires then that the linearized limit of (4. b = 1/3 while a is still a free parameter. Then. our original expression for S contained no term with (Y · T · Y )−2 (it corresponded to the solution a = 0). We begin with the Fµνρα sector. and that when adding the Einstein action contribution we recover upon integration the d = 7 scalar potential. a has to satisfy a quadratic equation. we made the ansatz p 1 (4. after substituting the 11-dimensional susy variation in terms of our fermionic ansätze. and though the factor 21˚ by itself may not be fatal. The last term in (4. Initially.104) and requiring that after adding the Einstein action contribution one obtains (T 2 − 2 Tr(T 2 ))/4. This ansatz satisfies the Bianchi identity: 4D˚ [σ Fµνρ]α − ∂α Fµνρσ = 0.103) and (4. such a term also generates a integral of the form d 4 x (Y T 3 Y )2 /(Y T Y )4 which clearly will produce unwanted Tr(T −1 )’s (apply (4.103)). However. the fact that infinitely many spherical harmonics would enter could lead to inconsistencies in the KK reduction. when integrating S 2 .93). Thus. and for simplicity. Nastase et al. Then. It seems very unlikely that the first solution. / Nuclear Physics B 581 (2000) 179–239 209 form Y T 3RY /(Y T Y )3 . and for this choice of S. The gauge field dependence of the FΛΣΠΩ is dictated by the 11-dimensional susy laws. The result for Fµνρα is rather surprising. the apparently nonlocal expression is.106) However.105) Fµνρα = g˚ µnuρσ D˚ σ ∂α (1 + S) + gauge field dependent terms. Hence.H. namely a(a + 2/3) = 0. since its presence requires that we expand 21˚ Dα (1 + S) into an infinite series of spherical harmonics. a gets also fixed after evaluating the integral over S4 of the square of (1 + S) using (4. local and given by (4.

2 3! ∂[µ ε¯ 0νρ] 9α − ∂α ε¯ 0[νρ 9µ] − 2∂[µ 0αρ 9ν] .

110) 2 Y ·T ·Y .109) = 4 Y ·T ·Y √  EF Y  3 2 F D T ∂[µ δBαAB CνC Cρ] ABCDE = 2 Y ·T ·Y  √   EF Y  3 2 F AB C D T ∂[µ Bα Cν Cρ] =δ ABCDE .107) δFµνρα |ψ = − 8 √ 3 2 0 0 p ∂[µ 1−1/5 ¯I 0 ψα.108) =− √4    YE 3 2 B ∂[µ ¯ γ ij k ψα CνA Cρ] ΠA i ΠB j Π −1 k E (4.J 0 η¯ I γnp γ5 ηJ U I I U J J Eνn Eρ] (4. (4.ψ (4.

) The total susy variation δ could be pulled out in (4.94). γk ] ∼ δj k γi − δik γj .93)).109) we used (3. look for simplicity only at gravitino dependent terms. varying the r. replace 2γ CD (4.108) to (4. Eµm 1−2/5U I I U J J Cm Vm ∼ γ CD YC ∂ m YD .s. Nastase et al.s.29) that (4. So. and rewrite (after substituting the various fermionic and vielbein ansätze) the 11-dimensional susy variation in terms of a total 7-dimensional susy variation of 7-dimensional bosonic (gauge and scalar) fields. (4.93) reproduces the BαAB -dependent terms which we get from the 11-dimensional susy variation of Fµνρα . The Π −1 from (4. of (4. Another nontrivial check of this ansatz is that the susy variation of the scalars in the l.110) allows us to read off the gauge field dependence of Fµνρα (see Eq. / Nuclear Physics B 581 (2000) 179–239 where to go from (4. and scalars vary into spin 1/2 fields.111) γ Cγ D − γ Dγ C by and use (Substitute into (4.45) cancels then one of the Π ’s in Eµm . Fµναβ will be fixed again by requiring the consistency of susy laws.h. one gets: √ 2 (2!)2 2∂[µ ε¯ 0ν][α 9β] − ∂[µ ε¯ 0βα 9ν] − ∂[α ε¯ 0νµ 9β δFµναβ |ψ = − 8 .210 H. namely. The complete answer is given in (4. One obtains then the commutator [γij .110) because we are keeping only gravitino dependent terms.45) to eliminate γ C . The procedure to determine the ansatz remains the same: vary Fµναβ under 11-dimensional susy laws. Hence (4.4) and the identity 0 0 0 0 IJ = −iΠA i (γi )I J CµA .h. The ansatz in the next sector.

+ 2∂[β ε¯ 0α][ν 9µ] .

113) we again made use of (3.111).4) and (4. With the use of the Schouten identity  CF D T EG YG AB Y[F ABCDE] = 0 Bα] Cν δ B[β Y ·T ·Y which yields  CF D T EG YG AB YF ABCDE Bα] Cν δ B[β Y ·T ·Y   A BC D T EG YG  CD E 1 AB + δ B[α Bβ] Cν ABCDE .113) the first term vanishes trivially. = δ(B[α ) · Y Bβ] Cν Y ·T ·Y 2 (4.112) + 2  EG Y  √ G A AB CF D T ΠA i ¯ γ i τ[α ψβ] − 4ABCDE δB[β Bα] YF Cν] = 2 ∂[µ Cν] Y ·T ·Y   √ T EF YF . in (4.112) to (4.113) − 2 ∂[α ABCDE δBβAB CνC CµD Y ·T ·Y To go from (4. Furthermore. and the last one can be written as a total 7-dimensional susy variation.ψ √ 0 0 = − 2 ∂µ −i ¯I 0 τα ψβ J 0 1−2/5 η¯ I γn γ5 ηJ U I I U J J Eνn 0 0 ρ  − 2¯I 0 ψβ J 0 1−1/5 η¯ I γnp γ5 ηJ U I I U J J Eνn Eρp BαAB VAB √  2 0 0 ∂[α ¯I 0 ψβ] J 0 U I I U J J 1−1/5 η¯ I γnm γ5 ηJ Eνn Eµm (4.114) (4.115) . (4.

of (4.116) After acting with a ∂[µ on (4. We follow the procedure outlined previously.116). To derive the ansatz of the next sector Fµαβγ is a bit more laborious. / Nuclear Physics B 581 (2000) 179–239 211 we get that  CF D T EG YG AB YF ABCDE Bα] Cν δ B[β Y ·T ·Y  A BC D T EG YG  1 T EG YG AB + δ(B[α ) · Y Bβ] = Cν (Bα] · Y )C CνD δ B[β 2 Y ·T ·Y Y ·T ·Y   AB CD E Cν ABCDE . Nastase et al. Then the 7-dimensional susy variation of the 3 gauge field dependent part of Fµαβγ will be the difference between (4.118) + 6 2∂[α ABCDE δ B[γ Y ·T ·Y To get the correct gauge-field dependence in the Fµαβγ sector we take into account the susy variation of the independent fluctuation Aαβγ whose ansatz in terms of 7-dimensional fields is (see below) − √2i Sαβγ . Bβ] + 12 δ B[α (4.H.117) and with the same tricks as we used before we arrive at √ √  3 2 −1  A A 3 2 i D[α ¯ τβ γ i ψγ ] ΠA i CµA ¯ [αβ γ ψγ ] + Π δFµαβγ |ψ = i Cµ τ 4 2   √  T EF YF AB + 6 2∂µ ABCDE δ B[γ (Bα · Y )C (Bβ] · Y )D Y ·T ·Y   EF √  T YF AB (Bβ] · Y )C CµD . We are ready now to extract the susy variation of the gauge field dependent piece of Fµαβγ : . Then we substitute back into (4.117) Then we substitute our ansätze in the l. Note that the first two terms in (4. and vary under susy the r.118) and the 7-dimensional susy variation of ∂√µ Aαβγ .h. of (4.h. (4. (4. the remaining terms can be cast into a total susy variation.95): √ − 2 3!(∂µ ε¯ 0[αβ 9γ ] − 2∂[α ε¯ 0µβ 9γ ] − ∂[α ε¯ 0βγ ] 9µ )|ψ δFµαβγ |ψ = 8 √ −3 2  −1/10 √ m −1/5 ∂µ ¯ 1 = γ5 τ[αβ 1−2/5 + 2γm τβ γ5 E[α 1 4   √ m n + γmn E[α Eβ 1−1/10 γ5 ψγ ] √ 3 2  −1/10√ ∂[α ¯ 1 γ5 2γm τβ γ5 Eµm 1−1/5 + 2   √ + γmn Eµm Eβn 1−1/10 γ5 ψγ ] .94).s.118) are already parts of the susy variation of −i 6Sαβγ .113) and we recognize that the 11-dimensional susy variation of Fαβµν is the same as the 7-dimensional susy variation of the ansatz given in (4.s.A (y)Y A (x).

δ(Fµαβγ )|B terms only .

ψ .

Nastase et al. First.119): .119) Fβγ ] Cµ . then we use a Schouten identity to release a susy gauge field variation from its contraction with the spherical harmonic:    T EF YF AB (Bα · Y )C (Bβ] · Y )D ∂µ ABCDE δ B[γ Y ·T ·Y  EF Y  F AB C D T = δ∂µ ABCDE B[γ (Bα · Y ) (Bβ] · Y ) Y ·T ·Y   T EF YF AB δ(Bα · Y )C (Bβ] · Y )D − 2∂µ ABCDE B[γ Y ·T ·Y   EF T YF 1 AB (Bα · Y )C (Bβ] · Y )D = δ∂µ ABCDE B[γ 3 Y ·T ·Y   1 AB δ BαCD (Bβ] · Y )E .119) we get:  EF Y   F AB C D T ∂[α ABCDE δ Bγ (Bβ] · Y ) Cµ Y ·T ·Y    EF Y   E 1 1 F AB C D T AB CD + ∂[α Bγ δ Bβ] Cµ .212 H. + ∂µ ABCDE B[γ (4. + 2 We are going to rewrite the first two terms as total 7-dimensional susy variations. / Nuclear Physics B 581 (2000) 179–239    √  T EF YF AB = 6 2 ∂µ ABCDE δ B[γ (Bα · Y )C (Bβ] · Y )D Y ·T ·Y   EF  T YF + ∂[α ABCDE δ BγAB (Bβ] · Y )C CµD Y ·T ·Y √  CD E 3 2 AB ABCDE δ B[α (4.120) and (4.120) 3 Schoutenizing the second term of (4. Then we shall “peel off” a δ from both sides of (4. = ABCDE δ∂[α Bγ (Bβ] · Y ) Cµ 2 Y ·T ·Y 2 (4. We begin by working out the first term. we shall pull outside a total δ susy variation.121) into (4.119) and read off the ansatz for (Fαβγ µ )|B terms only .121) The last step is to substitute (4.

δ(Fµαβγ )|B terms only .

ψ    EF Y  √ 2 AB F C D T = ABCDE 3 2δ ∂µ B[γ (Bα · Y ) (Bβ] · Y ) 3 Y ·T ·Y  EF Y  F AB A D T + ∂[α Bγ (Bβ] · Y ) Cµ Y ·T ·Y √ √  E   CD AB δ BαCD (Bβ] · Y )E Cµ + 2 2 ∂µ B[γ + 3 2∂[α BγAB δ Bβ]  √   E AB CD CD Cµ + 3 2 δ B[α ∂β Bγ ] + 2(Bβ · Bγ ] )   EF Y  √ F AB C D T = ABCDE δ 2 2∂µ B[γ (Bα · Y ) (Bβ] · Y ) Y ·T ·Y .

Aµνρ = − √ ABCDE CµA CνB CρC Y D ˚ Y ·T ·Y 6 2 2 2 2 E 1 (T · Y ) .124) are trivially satisfied in the F sectors which we derived so far.123) We conclude that.H.s.92)–(4.125) which yields   √ (T · Y )E 0 = ∂µ Fαβγ δ − 4ABCDE ∂[α 2 2BδAB (Bβ · Y )C (Bγ ] · Y )D Y ·T ·Y √   CD E 3 2 ∂β BγAB + 43 (Bβ · Bγ )AB Bδ] Y (4. from the requirement that the susy laws are consistent we obtain the ansatz for Fµαβγ given in (4.122) 2 To achieve the last expression on the l.95). + √ ∂[α Bβ + 3 (B[α · Bβ ) 2 2 (4.127) . + (4.126) + 2 to solve for Fαβγ δ . and satisfies trivially ∂[ Fαβγ δ] = 0.96) suggests that we can also give the ansatz for the 11-dimensional 3-index field.122) we need to perform one more trick. At this moment. Thus we use the last Bianchi identity ∂µ Fαβγ δ = 4∂[α Fµβγ δ] (4. indeed. while the purely scalar sectors are components of a separately closed form. we can (again for simplicity of the argument) solve Fαβγ δ from the Bianchi identities. Aνρα = √ ABCDE BαAB CνC CρD Y ·T ·Y 6 2 E 1 CD D (T · Y ) . Cν Aναβ = √ ABCDE (B[α · Y )A Bβ] Y ·T ·Y 3 2 √ i 6 Sαβγ . A brief inspection of the ansatz given in (4. namely to use the Schouten identity to rewrite:  EF F AB δ BαCD Bβ] Cµ ABCDE B[γ    AB = ABCDE δ B[γ (Bα · Bβ] )CD CµE − 3δ(B[γ )AB (Bα · Bβ] )CD CµE . because the gauge field dependent terms F can be easily cast into an exact form. of (4. The Bianchi identities ∂[Λ FΠΣΩ1] = 0 (4. The result. / Nuclear Physics B 581 (2000) 179–239 213  EF Y  √ F D T − 3 2∂[α BβAB (Bγ ] · Y )C BγBC C µ ] Y ·T ·Y √   CD E 3 2 AB AB 4 ∂µ ∂[α Bβ + 3 (B[α · Bβ ) Bγ ] Y . (4.A Y A Aαβγ = − 6 √ 2 AB (T · Y )E B[α (Bβ · Y )C (Bγ ] · Y )D + ABCDE 3 Y ·T ·Y   CD E 1 AB AB 4 Bγ ] Y .96) is compatible with the susy laws.h. which is given in (4. Nastase et al. AΛΠΣ : p  1 D˚ σ 1 (T · Y )E − √ µνρσ g˚ .

B=0 = m εFαβγ δ + 288 288 (4. This should already display the self-duality mechanism. B=0 = 288 5  − 24i τ αβγ  + 92 δ[α τ βγ ] γ µ Fµαβγ . eαβγ = Bαβγ . / Nuclear Physics B 581 (2000) 179–239 Also. where F(2) e . B (4. Consider the linearized gravitino transformation law.. we find:  b 1 b0 αβγ δ Bαβγ δ − a0 αβγ Bαβγ  + 15 τ γ5 γ m 0 αβγ δ m Bαβγ δ  δψ |lin. + (4.58] we present the expression of FΛΠΣΩ in form language.41) into the transformation law of ψ . F term. √  2 (T · Y )E 1 F(4) = ABCDE − DY A DY B DY C DY D 3 3 Y ·T ·Y  D (T · Y ) (T · Y )E 4 AB Y E + 2F(2) DY C DY D + DY A DY B DY C D 3 Y ·T ·Y Y ·T ·Y  AB CD E F(2) Y + F(2) + d(A).131) We introduce a normalization constant c for Sαβγ .A Y 24 5 5 .. B=0 = ζ − 8δζ 0 288 √  2 40 µαβγ ζ + 3 · 8δζα 0 µβγ εFµαβγ .A ∂µ Y A . (4.132) and a corresponding relation for the auxiliary field. we require that we obtain all the Sαβγ terms in the d = 7 To find the ansatz for A and B gravitino susy transformation law from KK reduction. (4.96)) and the geometrical ansatz of [57..A by embedding the 7D field S into the 11D curvature F as follows: Fµαβγ |B=0 = cSαβγ .92)–(4. to ease the comparison between the ansatz necessary to achieve a consistent truncation (given in (4.A Y A . It is gauge invariant.130) The result for δψ can be worked out further by decomposing the 11-dimensional gamma √ matrices and using the commutation relations of γ5 : √  2 γ5  3 3τ αβγ δ  − 8δ[α τ βγ δ] Fαβγ δ δψ |lin.. and also will fix the free parameters a and b in the d = 11 susy transformation law of the auxiliary field B. F term. (4. Nastase et al.A in (2.214 H.A Y A . Fαβγ δ |B=0 = 4c∂[α Sβγ δ].133) Substituting the linearized ansatz for Bαβγ . B term = 24 24     16 k A 9 [α βγ δ] ηζ αβγ bτ − a − b δ τ γ5  = αβγ δ Bηζ.129) 288 √ √  2 αβγ δ 2 µ αβγ 0 0 40 µαβγ m − 8em εFµαβγ . δ9m |lin.128) AB = 2(dB AB + 2(B · B)AB )) and DY A = dY A + 2(B · Y )A . F term. The FΛΠΣΩ term in δ9Λ reads: √ 2 αβγ δ [α βγ δ]  0 εFαβγ δ δ9ζ |lin.

That means that at this moment the sign is fixed relation of γ5 by requiring a consistent truncation.16) instead of p = q = −1/2 we would get (iγ µ Dµ + m) in (4. m Hence √    K 2 9 [α βγ ] A αβγ m τ − 2 δ τ γ K η Sαβγ .A Y A −m 24 5    k 16 − a − b 6τ αβγ 4D[ Sαβγ ].H. δψ |lin.138) M E N = 0. (4.136) (4. we get by contracting with (γ A )J K and using 1/4φ5J K (γ A )J K = Y A . Now.1:   1 m + iγ µ Dµ Y A γ5 ηI I = −I γ A I K ηK . let us analyze the nonlinear level. we need a 4-index tensor to mix with the independent fluctuation Sαβγ .132) in the 11D action for FΛΠΣΩ . one can see why a first-order formulation for gravity instead of the antisymmetric tensor field does not work.A Y A + 6τ 4Dα Sβγ δ. where the same. Then we obtain: √ 2  αβγ 9 [α βγ ]  − 2 δ τ τ δψ |lin. In conclusion.134) 24 5 In order to reproduce the 7-dimensional sugra result (and have a consistent truncation of 11-dimensional sugra to the massless 7-dimensional fields). 11D 9α is at the linearized Σ] 1 βγ level is 4 Bαβγ τ ε.135) 60 Using (3.135). as done in [50]. But this has no γ5 with respect to (4. we fix c = −i 6. / Nuclear Physics B 581 (2000) 179–239 = 215  γ5 9 kb αβγ δ 9 kb 4!3! [α βγ ] δ τ Sαβγ . we cannot introduce by hand the missing γ5 . (4. we must reproduce the I gauged sugra √ result for δψ√ . Thus the B and 1Ω 0 [Λ MN EΠ αβγ term we get in δsusy.A γ5 (−iγ µ Dµ + m)Y A . according to Section 3.131).A Y A . S terms = −c Sαβγ .A Y A −m 5 5 24 2 5 24   16 k a − b 3!τ αβγ Sαβγ . with DΛ (Ω)ε bΛ MN (E. S terms = c 60 (4. which does not have any DS terms. we note that here if we would use the p = q = +1/2 solution in (4. Therefore it will be impossible to cancel the F terms in (4.6). as we did in (4. so we must require kb = −5 2c/72. Consider the spin connection Ω as an independent field. ka = −5 2c/9. √ In order to match this result with the S term in δψ in (4..137) Incidentally.134). 9) + BΛΠΣ E MΠ E NΣ + 1Ω 0 Λ MN ΩΛ MN = Ω (4.. The nonlinear KK reduction gives for the terms involving S:  −1/2 0 a  Λ Λ e Ea δ9Λ |S term + 15 τa γ5 γ m Em δ9Λ |S term δψ |S term = 1−1/10 γ5 . Nastase et al. and since the susy law is fixed by the definition of B.A .A .135) (because the commutation ±1/2 past γµ gives a ± sign).42). We can then check this value of c by putting (4. We obtain the correct normalization for the 7D action of Sαβγ . At this moment. then the 11-dimensional gravitino transformation law is b replaced by DΛ (Ω)ε.

24). / Nuclear Physics B 581 (2000) 179–239 √   2 −1/10 −1/2 0 a 1 = γ5 e 0 αβγ δ a − 8δa[α 0 βγ δ] εFαβγ δ |S term 288  Bαβγ δ 12 + √ b0 αβγ δ a − aδa[α 0 βγ δ] ε √ 2 E  µαβγ [α µβγ ] +4 0 εFµαβγ |S term a + 8δa 0    1 12b Bαβγ δ m αβγ δ + τa γ 5 γ 0 √ m ε Fαβγ δ + √ 5 2 E   µαβγ µ αβγ εFµαβγ . µ × 1−1/5 Em − iγ m Fµαβγ +B (4.139) In the same way as it happened at the linear level.42).A Y A − αβγ δ ηζB = 5 5 (4. whereas the Sαβγ terms in Bαβγ δ and Fµαβγ add. and moreover the term µ BαAB Sβγ δB in Bαβγ cancel the Bδ Fµαβγ term. and doing a bit of algebra. we have P Q VnP Q CnNI = 4φ5P [I Ω N]Q − Ω P Q φ5NI + Ω NI φ5 . the ∂α Sβγ δ terms in Bαβγ δ will cancel the Fαβγ δ terms.C Cm A  B 0 (4.141) αβγ We further work out the Fµαβγ term in δψ by substituting our ansätze for various fields. Nastase et al. b.22) and (3. Then. where in Fµαβγ we consider only the term with S: Fµαβγ |B=0 = cSαβγ .143) × Π −1 i Π −1 j V m AB C n NI ηN U I I I 0 . In the following.144) and substituting this relation for the two products Vn Cn in (4. (4. (4. 5[α Sβγ δ].216 H.A or Bα Sβγ δ.140) e contains only S terms (no ∂ S AB where B α βγ δ. c take the values which we determined previously. provided we choose the ansatz (neglecting “massive” fields which are put to zero)  k Bαβγ δ 6k µ e eηζ = Fαβγ δ + 4Bδ Fµαβγ − αβγ δ ηζB √ 5 5 E 24kc k e eηζ . Then we obtain: √ 2 αβγ 9 [α βγ ]  τ − 2 δ τ γ5 13/5 δψ |S term = − 60    e . using (3.142) µ So.28) and introducing the β β identity as η¯ αN ηN = δα . using the ansatz for Em in (4.143). the parameters a.A ∂µ Y A . by substituting for Fµαβγ . we get: . we get:  µ γ5 −iγ m 13/5 Fµαβγ  1−1/5 Em  C Tr γ ij U V n U T Ω = − 14 13/5 cSαβγ . +4 0 m − 2δm 0 (4.B ) and we have used (3.

After finding the correct ansatz for Bαβγ δ . we can easily find the nonlinear terms in δλi : Λ δ9Λ δψm |S term = 1−1/10γ5 Em √   2 −1/10 1/2 αβγ δ µαβγ µ αβγ 1 = γ5 0 Fµαβγ m Fαβγ δ + 4 0 m − 2Em 0 288  Bαβγ δ 5 − 0 αβγ δ m √ ε. 217 (4. (4.145) But now we can use the condition we got on U from matching the scalar transformation law (4. Using that k = 5(6 2) (as it will be determined below (5.148) The necessity of adding the gauge field terms in the ansatz for the auxiliary field BMNP Q can be seen if we try to derive the 7-dimensional action from the 11-dimensional one.B     0 × 12 φ5P I U T γ ij U P N − φ5P N U T γ ij U P I ηN U I I I 0 . and when integrating out AB )4 which are not present in the d = 7 (Fαβγ δ )2 on S4 we get as a result terms with (Fαβ action. the ansatz of the auxiliary field BMNP Q is √ √ Bαβγ δ = −24 3i ∇[α Sβγ δ]. and we are left with . if we choose the ansatz for Bαβγ to be: e eαβγ = cT AB Sαβγ .A Y A + 3i αβγ δ ηζ T AB Sηζ.B YA √ E BC DE A + 9ABCDE F[αβ Fγ δ] Y + 2 − Fermi terms.A term of δψ in [51]). We know that BMNP Q = 0 is a solution of the 11D equations of motion.31)).B YA γ5 ηI U I I I 0 . B (4. as in δψ .45) and get  A I0 0 0 (4.140) the necessary bilinears in fermions and gauge field strength to complete the 7-dimensional equation of motion for Sαβγ . / Nuclear Physics B 581 (2000) 179–239 A  B 13/5 Π −1 i Π −1 j Y A cSαβγ . they are precisely canceled by the gauge field strengths which enter in the ansatz of BMNP Q .B YA . The second observation is that no other bosonic equation of motion can appear in BMNP Q . If we now use the ansatz for µ Bαβγ / E. The ansatz for Fαβγ δ is already bilinear in the gauge field strength.A Π −1 i γ j L0 U L N ηN I 0 − 13/5T AB cSαβγ . whereas the second term gets canceled by e e the B αβγ term.149) 6k E 1/2 where √ we have substituted the value of b found before.146) cSαβγ .and 4-Fermi terms). and so we would get terms with 4 derivatives in the 7D action. since all other bosonic equations have 2 derivatives. However. Then the first remark is that we have to add to (4.H. after integrating out (BMNP Q )2 . Nastase et al. We now see that the first term is exactly what we want to have in 7-dimensional supergravity (the Sαβγ .147) However. the Fαβγ δ and Bδ Fµαβγ terms cancel again. So we can say that the ansatz for the various components of BMNP Q has to be such that the various components of BMNP Q = 0 correspond to various 7D equations of motion. (4. let us pause and comment on the ansatz for BMNP Q .A (remember that we always √ dropped in the susy transformation rules 3.

we get (7D) =− δψm  A I 0 1 √ τ αβγ Sαβγ .45). we need a few identities: by multiplying the Fierzing relation (3. Then we write E µ in terms of scalars via (4.151) Π −1 i YC Tr U −1 γ i U γ D CnD = 0 which we easily get by using (4. For that. (4.150) as follows: (11D) = δψm  B  1 n √ τ αβγ Sαβγ .154) Using the U matrix relation (4. we find that for ηm (7D) = δψm   B 1 √ τ αβγ Sαβγ . Next we want to put in a similar form the expression which we get for δψm from the 7-dimensional result for δλiJ 0 . We next substitute for Fµαβγ and B αβγ n Using (3. L ηm (4.152) − 413/5T AB YB γm ηI I 0 U I I . we get I −iγp ηI CpA + Y A ηI = YB γ A γ B J ηJ (4.42) with γ5 (γA )J K .B Π −1 j 60 3   I 0 0 × i(γmp − 2δmp )C pA + γm Y A ηJ γj U γA U −1 J 0 U J J I 0 . ηm (4.158) .157) γ5 ηI = −YA γ A J ηJ .150) e .15) and the normalization in (4.21) for the contraction of Cµ ’s which we get.B i Tr U γ A U −1 γi Π −1 i (γmp − 2δmp )C pA 48 3  + 413/5T AB YA γm   B 1 − √ τ αβγ Sαβγ .218 H. we can rewrite (4.A Π −1 j γ ij + 4δ ij J 0 240 3 0 × (γi )K 0 L0 I 0 U J J U K 0 KU L0 J KL . and the identity C  (4.91). × 1 En γm n − 2δm (−iFµαβγ ) − γmB (4.156) Now we work on the last term (the last two lines).45).153) J KL as Then we write the spherical harmonic ηm   J KL = i(γmp − 2δmp )C pA + γm Y A ηJ (γA )KL .36). Using this relation and the expression derived J KL . Nastase et al.B −i Π −1 i CnD Tr(U −1 γi U γ D ) γm n − 2δm 48 3  0 (4.42) with YA (γA )J K we deduce that I (4.155) which will be heavily used in the following. / Nuclear Physics B 581 (2000) 179–239 √ i 2 αβγ 3/5 4τ δψm |S term = 1 288  −1/5 µ    e n eαβγ . From the ansatz in (4. we find the identity  A Tr U γA U −1 γi YA = 413/5 Π −1 i YA (4. If we multiply the same relation (3. and substituting the 7-dimensional transformation law of δλiJ 0 .

156) is zero. for completeness of our work we give a separate check on Sαβγ susy transformation law. The gravitino and Sαβγ .161) because the BαAB terms in Fµαβγ appear at least quadratically. and the terms involving Sαβγ in the transformation laws of the gravitini ψαI and fermions λiI . one with ∂α acting on ψγ I 0 and λiI 0 .164) − 4 .A transformation laws Let us recapitulate what we have done so far. that A B Y (γA γB )I J ηJ .159) and (4. so that we are left with the correct 11-dimensional result. The analysis of the independent fluctuation Sαβγ . γm γ p CpA ηJ (U γA )I J = 4Cm (4. (so δBαAB will always be multiplied by BβCD . corresponding to the terms in the 7-dimensional transformation law.159) From (4.14) and the fact that  1/2 1/2 (4.162) 9α |B=0 = 1−1/10 γ5 ψα + 15 τα γ m γ5 ψm .H. Nastase et al.A |B=0 ∂µ Y A √ = −i 3∂[µ ε¯ 0αβ 9γ ] √  h  i i 3 i ∂µ ¯ 1−3/5 τ[αβ γ5 ψγ ] + ταβγ γ m ψm =− 4 5   n −2/5  ¯ − 2iγn γ5 τβ ψγ ] + 15 τβγ ] (−2γnm + 3δnm )ψ m . Using the transformation law for AΛΠΣ in (2.159) we also deduce A J η (U γA )I J .3.A was implicitly done when we studied the Fαβγ µ sector. We have checked that using our ansätze for the 11-dimensional fields we reproduce the 7-dimensional transformation laws for the graviton eαa . We begin by considering the case when BαAB = 0 in the Sαβγ . in 7 dimensions we have: √ B  i 3 δAB Π −1 i 3¯ τ[αβ γ i ψγ ] − ¯ ταβγ λi δSαβγ .A |B=0 = 12 √  i 3 δij ΠA j 5[α 2¯ τβ γ i ψγ ] + ¯ τβγ ] λi .160) Using (4.A from i √ δFµαβγ |B=0 = δSαβγ . by multiplying with CAn .A transformation law and we find δSαβγ .A |B=0 ∂µ Y A 6 (4. the scalars ΠA i .158) and (4. However. Indeed. We have also checked the gauge (11D) field dependence of FΛΠΣΩ by matching terms in δsusy FΛΠΣΩ with the variation of all the fields in the ansatz for FΛΠΣΩ . one with ∂α acting on ΠA i . and one with no ∂α . γm ηI = iCm (4. the gauge field BµAB .163) We split this contribution into 3 terms. 4. we get δSαβγ .160) it follows that the last term in (4. − ∂[α Eµ 1 (4. which we put to zero). / Nuclear Physics B 581 (2000) 179–239 219 from which we also derive. (4.

163) is given by √    i i 3 −3/5  3/5 1 ¯ αβγ γm ψ m .166) ¯I 0 τ[αβ γ i ψγ ]J 0 Π −1 i ∂µ YA .45). we consider the terms with ∂α acting on ψαI 0 and λiI 0 : √ i 3 −3/5 1/5 n  0 1 1 Eµ (−∂[α ¯ 2iγn γ5 τβ ψγ ]I 0 )U I I ηI 4  J0 K0 L0 J KL + 15 (∂[α (−2γ ¯ . Substituting (4. Π (Ω) (4.170) we find for (4. .)j = ∂α (.169) + 15 ¯I 0 τ[βγ λJ 0 K 0 L0 η¯ I − 2γ nm + 3δ nm ∂α] U I I U J J U K K U L L ηm The spherical harmonic of the spin 1/2 fields.s. . / Nuclear Physics B 581 (2000) 179–239 A where 5α (.45) (because YA γ A ∼ φ5 ).173) . . .167) J i 4·3 We compare the sum of these two contributions with the 7-dimensional result: √  i 3 −1  A Π ¯ 3τ[αβ γ i ψγ − ταβγ λi ∂µ Y A (4. Together with (4. Nastase et al. . contracted with λJ 0 K 0 L0 .165) 1 ∂µ 1−3/5 + ∂µ ¯ τ[αβ γ5 ψγ ] + τ − 4 5 To evaluate the contribution with the the gravitino ψγ . (If BαAB is not zero. − 4 This agrees with the result in 7-dimensional gauged supergravity: √ B  i 3 ΠA i Π −1 i ∂[α ΠB j (2τ ¯ β γj ψγ ] + τβγ ] λj )∂µ Y A . (4.)j .169): √  i 3 0 0 0 0 ∂[α ΠA j −¯I 0 2τβ γjI J ψγ ]J 0 + ¯I 0 τβγ ] λj J 0 Ω I J ∂µ Y A .170) (4.)i + (Π −1 )[i (∂α ΠA j ] )(. From ¯ ταβ γ5 ψγ we obtain a factor U ηγ combination which appears on the l. .220 H.h. 4 Similarly.168) i 12 and we can conclude that we find agreement. nm + 3δnm )(τβγ ] λJ 0 K 0 L0 )U J U K U L ηm In the last term we use (4.172) (4. . The term with ∂α acting on the scalars is given by √   i 3h ∂[α Eµn 1−2/5 −2i ¯ γn γ5 τβ ψγ ] + 15 ¯ τβγ ] (−2γnm + 3δnm )ψ m 4 0 0  + 1−2/5 Eµn −2iCnI J ¯I 0 τ[β ψγ J 0 ∂α] U I I U J J i  0 0 0 0  J KL  .165) reads √ i 3 −1  A e I 0 J 0 ¯I 0 ταβγ λi 0 ∂µ YA . . .)i acts as ∂α (. − 4 Finally. we make use of the equations ¯ 5 ηU T ∼ U φ5 U T . the contribution of the spin 1/2 fermions in (4. . the ∂µ derivatives of 13/5 cancel and we are left with √ A −i 3 I 0J 0 (4. of (4. (4.)i + Qα i j (.171) (4. This is the (4. Qαij will contain also a B term.170) and get (4. yields 0 0 1 I J KL 0 0 0 J 0 λJ K L U J U K K U L L 5 η¯ (−2γn γm + 5δnm )ηm 0 0 0 = 3Ω I J CnKL λJ 0 K 0 L0 U J J U K K U L L .) The term with no ∂α in (4.39) and (4.45).

(4.176) −i 4 This was the last term to be checked in the transformation law of the 3-index tensor field Sαβγ .175) −i This agrees with the d = 7 result √  3 ΠA j ∂[α 2¯ τβ γ i ψγ ] + ¯ τβγ ] λi ∂µ Y A δij . (4. Nastase et al.H. Hence we keep from the 11-dimensional susy variation of Sαβγ only the terms AB dependent terms in the d = 7 susy with potential contribution to the spin 1/2 and Fαβ law: √ . Since the dependence on the gravitini and gauge fields was already checked (see the analysis performed for the susy law in the Fαβγ µ sector). we will only check the Fαβ λ terms in the Sαβγ 7-dimensional susy law. 2 4 (4. Now.174) 4 e we obtain e T = Ω) Using (4. we let the gauge fields to take non-zero values. which therefore is in complete agreement with the 11-dimensional transformation laws. / Nuclear Physics B 581 (2000) 179–239 221 √ 3 0 0 2i1−2/5Eµn (∂[α ¯I 0 τβ ψγ ] )U I I U J J CnI J 4 √ i 3 −2/5 n I J KL I 0 J 0 K 0 0 31 + Eµ Ω Cn U I U J U K U L L (∂[α ¯I 0 τβγ ] λJ 0 K 0 L0 ).A with BαAB set to zero.111) (and U ΩU √ √ i 3 −i 3 0J 0 0 0 j j I A ΠA ∂[α ¯I 0 (γj ) τβ ψγ ]J 0 ∂µ Y − ΠA j ∂[α ¯I 0 τβγ ] λJ 0 Ω I J ∂µ Y A .

6 (T · Y )E AB .

we need to evaluate the spherical harmonic of  η¯ I 15 γmn γ p ψp + 2γ[m ψn] which is  KL .178) . (4. δ BγCF Y CD −i 6δSαβγ . (4.177) is √ 9i 2 −1/5 0 0 0 0 AB ν 1 ¯I 0 τ[γ λI 0 J 0 K 0 U I I U J J U K K U L L Fαβ] VAB Eν[n Eµm] − 2  × η¯ I γmn ηJ φ5KL + 2η¯ I γ[m ηJ CnKL √  3 2 DE F[αβ ¯ τγ ] γ [ij λk] ΠA i ΠB j − Π −1 k C = 2Y · T · Y   + 2ΠA i ΠB k Π −1 j C VνDE CµA CνB Y C . 3 η¯ I γmn ηJ φ5KL + 2η¯ I γ[m ηJ Cn] The contribution of the last two terms in (4.177) After we substitute the ansätze of various 11-dimensional fields.A Y A |F λ terms = √ ABCDE F[αβ ] λ terms F µ Y · T · Y 2 √ 3i 2 −1/5 AB ν 1 ¯ γmn τ[γ γ p ψm Fαβ] VAB Eνn Eµm − 10 √ AB ν + 3i 21−1/5 ¯ τ[γ γ[n ψm] Fαβ] VAB Eνn Eµm .

178): √ DB   ¯ γ ] γ ij kl λl ΠA i ΠB j Π −1 k C CµA YC YD ¯ τγ ] γ l γ ij k λl + τ − 2F[αβ  D (T · Y )E 6 . Since the susy variation of a gauge invariant object must be gauge invariant.180). we already know from the analysis of the sector Fαβγ µ where we derived the gravitino dependent terms in (4.179) to rewrite (4.181) Π −1 i Π −1 i τα I 0 . = √ ABCDE (F[αβ · Y )A δ BγBC ] Cµ Y ·T ·Y 2 (4. (F[αβ · Y )A δBγBC ] CµD Y E ABCDE will give the 7-dimensional susy variation δ(d = 7) Sαβγ .180) we replaced γ ij k by −1/2 ij kmn γmn and we also used the property of the scalar fields ΠA m of having determinant one. we have: A   0 δψαI 0 |S=B=0 = Dα I 0 + 14 Π −1 i ∂α ΠA j γ i j I 0 J J 0 A A 1 − 20 (4.176) is in fact.177) can be written as the sum of two terms upon using the Schouten identity to release the variation δ(BγCF ) from its contraction with YF . B=S=0      3 1 2 YA (T 2 )AB YB T −1 4!1 τa ε 1 + = −5 − −1 288 3 YA YB T AB 3 (YA T AB YB )2  4 · 3! −1 α 1 (τ a + 2eaα )γq γ5 ε 11/5Eσq − 288 .222 H.  n − E nρ Eµp ∂ν Epρ Ωµ MN (E)0MN |B=0 = γmn E mν 2 ∂[µ Eν]    + τ α γm γ5 11/5 ∂α Eµm + 11/5 Eµn E mρ ∂α Enρ . The first term in (4. In 7 dimensions. we find: EaΛ δ9Λ |Fµνρσ term.183) For the terms involving Fµνρσ and Fαµνρ . / Nuclear Physics B 581 (2000) 179–239 Use now the identity γ [ij λk] = 1 3 γ l γ ij k λl − γ lij k λl  (4. (4. and we obtained in fact a gauge covariant derivative.180) where to reach the final result in (4. But in 11 dimensions we have got δ9Λ |B=S=0 = ∂Λ ε + 14 ΩΛ MN (E)0MN |B=0 ε + F terms + 3-Fermi terms.176). Nastase et al.182) By direct substitutions one obtains: Ωα MN 0MN |B=0 = ωα ab τab + γmn E mµ ∂α Eµn   − 2τα γm γ5 1−1/5 E mµ 11/5 ∂µ 1−1/5  − 2τ β α 11/5 ∂β 1−1/5 .) Let us now turn back to δψα . (That this is the case. The same argument implies that the partial derivative ∂α which we obtained in (4.A |λF terms Y A . (4. we will not show explicitly the cancellation of terms produced by the 11-dimensional susy variation which are proportional with bare BαAB gauge fields. One of these terms cancels precisely the contribution of (4. a gauge covariant Dα derivative. and the other.

184) Λ δ9Λ |Fµνρσ term. B=S=0 Em      3 1 2 YA (T 2 )AB YB T −1 3! · 81 γm γ5 ε 1 + =− −5 − −1 288 3 YA YB T AB 3 (YA T AB YB )2 − 14 11/5(Eσ m − 2Eσr γmr )τ α    T AB YB ∂α T AB YB CD × CAσ − ∂ T Y Y . / Nuclear Physics B 581 (2000) 179–239  × CAσ   T AB YB ∂α T AB YB CD − ∂ T Y Y .H. C α D YA T AB YB (YA T AB YB )2 223 (4. C α D YA T AB YB (YA T AB YB )2 We also have that . Nastase et al.

−1/2 a  Λ Λ eα Ea δ9Λ + 15 τa γ5 γm Em δ9Λ .

terms with τα .B=S=0 .187). we prove that   (4.45) to get rid of the resulting traces we obtain. That means that the last two terms.187) × CAσ − ∂ T Y Y C α D YA T AB YB (YA T AB YB )2 to be compared to A   0 0 (4.91) to rewrite the vielbeins. terms with τ β α .187) should give zero.186) Putting everything together.191) . δψα |S=B=0 = 1−1/10 γ5 (4. Using that 1−6/5 = Y · T · Y and the identity 16/5 ∂α (Y · T · Y ) T AD YD − ∂α T AD YD = 16/5 ∂α (Y · T · Y ) T BD YD δ AB − YA YB  − ∂α T BD YD δ AB − Y A Y B . (4.187). whereas the terms without any τα ’s simplify to   3i 1 0  γmn E mµ ∂α Eµn  + I 0 ∂α U I I ηI + γm 11/5Eσm 1−6/5 4 4    T AB YB ∂α T AB YB CD (4.  (4. and the second term gives the correct 7-dimensional result in (4.188) − 14 Π −1 i ∂α ΠA j γ i j J I 0 J 0 U I I ηI .185) (4. We will first of all work on the first term in (4. After some algebra.90) and (4. Using (4.190) The first term cancels against the second term in (4. after some algebra. Using the Fierzing relation (3. besides the usual term ∂α I 0 + ωα ab τab I 0 .189) C mB C nD γmn ηI = (γDB )I J + YB (γD )I J γ5 − YD (γB )I J γ5 ηJ . the previous relation to get rid of γmn and (A.12) and (4. as they should. 1 mµ ∂α Eµn γmn  4E  A 0 0 0 = −∂α U I I ηI I 0 − 14 ΠA i ∂α Π −1 j (γij )I J 0 U J I ηI I 0  A 0 0 − 12 YA ∂α 13/5 Π −1 i (γi )I J 0 U J I γ5 ηI I 0  A  B I 0 0 + 12 ΠA [i ∂α Π −1 j ] Π −1 j YB 13/5 γ j J 0 U J I γ5 ηI I 0 . we find. together with the last term in (4.42) twice for each γ matrix acting on ηI . and terms without any τα ’s. the terms with τ β α cancel.188).

194) − τα 1−1/5 E mµ E nρ D˚ µ Eρq (γmnq − 2δmq γn )  20 to be compared with  A −1 A 1 (4. the ansatz for the scalar field dependence of Fµνρσ is fixed by the consistency of the gravitino transformation law..187). substituting for relation (A.42). Bosonic equations of motion and action In this section we will do some further checks on our ansatze by looking at the bosonic action and equations of motion. Now taking the last term in (4.. but with the opposite sign.190) as    ∂α (Y · T · Y ) BD 1 −3/5 AB A B BD T YD − ∂α T YD δ −Y Y 1 4Y · T · Y Y ·T ·Y × ΠA i (γi )I 0 J0U J0 I γ5 η I I 0 .188). So all the extra contributions cancel and we are left with the 7-dimensional result (4. (5.192) Eσm from (4... Nastase et al. they seem to involve the explicit expression of the matrix U and we prefer to devote a separate paper to this issue.3) BΛΠΣΩ = 0.2) ∂Λ EF ΛΛ1 Λ2 Λ3 = √ 2(24)2 (5.Λ7 FΛ8 . and reproducing some of the corresponding equations of motion and terms in the action in seven dimensions.104). / Nuclear Physics B 581 (2000) 179–239 we rewrite the last two terms in (4. using the summation (4. we give the 11-dimensional bosonic equations of motion following from the action in (2. 20 Π i This last problem turns out to be surprisingly complicated.Λ11 .2):  1 GΛΠ F 2 . as promised..12) and the Fierzing relation   C mB γm ηI = −i (γB )I J γ5 ηJ + YB ηI (4. Note that at this moment if we chose the a = 0 solution for (4.195) Π −1 i τα . We have only partial results. So. The hardest term in δψα is the τα term.90). In seven dimensions we have the bosonic action . First. we get the same result as in (4. which becomes (after some algebra)     i 1 i 0  1−1 ∂µ 1  + γ˚µ  − I 0 ∂µ U I I ηI τα 1−1/5 E mµ γm 50 10 5      T 2 YA (T 2 )AB YB 1 3 −6/5 τα 1 + −5 − −1  − 1 20 3 YA YB T AB 3 (YA T AB YB )2   i (4. The same holds for the scalar dependent terms in the susy law of the 7-dimensional spin 1/2 fields.193) which we can prove by using (3. it will not work. as we should.1) RΛΠ = − 16 FΛΛ1 Λ2 Λ3 FΠ Λ1 Λ2 Λ3 − 12  k  Λ1 .224 H.. (5. 5.Λ11 FΛ4 .192).

so in the action we need to keep only terms at most linear in BαAB which we can rewrite as follows:  2 A e−1 L07d = − 12 R + 14 m2 T 2 − 2Tij T ij − 12 Pαij P αij + 12 Π −1 i Sαβγ .B − − 12 T AB Sαβγ .B] = 0.3) was already discussed.7) (5. (5. Nastase et al. = 0. Let us look now at (5.10) together with (5. / Nuclear Physics B 581 (2000) 179–239 225   AB 2 e−1 L07d = − 12 R + 14 m2 T 2 − 2Tij T ij − 12 Pαij P αij − 14 ΠA i ΠB j Fαβ 2 A 1 + 12 Π −1 i Sαβγ .9) We note here that in (5.A S α β γ B  − 14 Tr ∂α T −1 ∂β T − 3T AB Sαγ δ.B . First. 48 me  (5. setting the gauge field to zero.A Fδηζ.  0 0 0  2 m2 T 2 − 2TAB − 4T AB Sα 0 β 0 γ 0 .8)  (5. so we obtain the following bosonic equations of motion for TAB .A and the metric gαβ . 2 m T − 2 Tr T 1 me−1  αβγ δηζ Sβγ δ.6) = 0. From (5.A S αβγ B + + −1 αβγ δηζ AB 1 δ Sαβγ .A S αβγ B 2 m (T δAB − 2TAB ) + 4 T  1   −1  1 2 2 2 − 15 TAB + 4 ∂α Tr ∂α T T −1 + 12 Sαβγ .A Fδηζ.4) For the equations of motion.A −1 αβγ δηζ AB 1 δ Sαβγ .A Sηζ. which is just the equation of motion of the antisymmetric tensor.B + 1 2 T ∂α T −1 αβγ δηζ 1 Fδηζ.10) ∂α ∂α T T −1 [AB] = 0 and from (5. (7) − Rαβ 1 10 gαβ (5.7) we also get that  αβγ δηζ δ AB Sαβγ . BαAB and Sαβγ . Sαβγ . + √  αβγ δηζ ABCDE δ AG Sαβγ .A + 48 me−1  αβγ δηζ δ AB Sαβγ .A .5) We notice that the action is now expressed in terms of TAB instead of ΠA i . Substituting the ansatze for FΛΠΣΩ and gΛΠ the {νρσ } component of (5.G Fδ 16 3 + (5. The only nontrivial component is the {αβγ δ}.11) Now we want to see that we reproduce these equations from the 11-dimensional equations of motion.6) we have used Lagrange multipliers for the condition det TAB = 1.A 24 me   −1 [AB] (5.A S αβγ B T AB = 0. we will put the gauge field BαAB to zero.B m−1 −1 m−1 −1 e Ω5 [B] − e Ω3 [B] 8 16 ie−1 BC DE Fηζ . we notice that (5. respectively:  −1 1 2 1 ∂α ∂α T T −1 AB + 12 Sαβγ .[A ∂δ Sηζ.2) becomes .A S βγ δ B = 0.A Fδηζ.6) we obtain also  (5.B 48 me    2 1 1 2 = − 2 R + 4 m T − 2TAB T AB + 18 Tr ∂α T −1 ∂ α T  − 12 BαAB T ∂α T −1 AB + 12 T AB Sαβγ .H.2).

226

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239   

Y(A ∂µ YB) TCB 2(T δAC − 2TAC ) + T −1 ∂ α ∂α T T −1 AC + 2Sαβγ ,A S αβγ ,C 

− Y(A ∂µ YB) Sαβγ ,A Sαβγ ,C TCB + 16 me−1  αβγ δηζ ∂δ Sηζ,B
− 13 Y [A ∂µ Y B] me−1  αβγ δηζ Sαβγ ,A ∂δ Sηζ,B = 0.

(5.12)

We notice that the first line is the scalar equation of motion (5.6), the second line is
the antisymmetric tensor equation of motion in (5.8) and the third is (5.11) which is a
combination of the scalar and gauge field equations of motion.
Similarly, by substituting the ansatze for FΛΠΣΩ and gΛΠ , the {νρα} component of
(5.2) becomes:   

(5.13)
∂[µ Y A ∂σ ] Y B T −1 ∂α T BA + 16  α1 ...α6 α Sα1 α2 α3 ,A Sα4 α5 α6 ,B = 0,
which is just the equation of motion for the gauge field, (5.7). The {βγ δ} component of
(5.2) is   

Y A 4∂ δ ∂[δ Sαβγ ],A + 16  αβγ α1 ...α4 ∂α1 T AB Sα2 α3 α4 ,B − Sαβγ ,A T 2AB YB 

YA 
−4∂[δ Sαβγ ],A Y · ∂δ · Y + 16  αβγ α1 ...α4 Sα2 α3 α4 ,B T AB (Y · ∂α1 T · Y )
Y ·T ·Y    

Y ·T2 ·Y
− T −2
Y A 16  αβγ α1 ...α4 ∂α1 Sα2 α3 α4 ,A + Sαβγ ,B T AB = 0, (5.14)
Y ·T ·Y

+

where all three lines are now zero due to the antisymmetric tensor equation of motion,
(5.8). The {αβµ} component of (5.2) gives  

T AD YD T CE YE
−1 
A
∂γ Sαβγ ,B δ BC + Sαβγ ,B (∂γ TBD )TDC
Cµ TAC −
Y ·T ·Y 

− Sαβγ ,C T BC + 16 ∂α2 Sα3 α4 α5 ,B  αβα1 ...α5  

Y · ∂γ T · Y
1
AD
AD
∂γ T YD YB − T YD YB
.
(5.15)
×
Y ·T ·Y
Y ·T ·Y
In the first line, the second bracket is zero upon using the equation of motion for the 7D
−1
, whereas the
antisymmetric tensor, (5.8), to convert both Sαβγ ,B ’s into ∂α1 Sα2 α3 α4 ,C TBC
second line gives directly the antisymmetric tensor equation.
The Einstein’s equations (5.1) are considerably more involved. We have computed all
the terms for zero gauge field, but the task of reconstructing the seven-dimensional field
equations is quite laborious, and we have not completed it, but we can see that we get
nontrivial combinations of these seven-dimensional field equations. The fact that already
from the antisymmetric field equation (5.2) we get all the seven-dimensional field equations
(except Einstein’s equation) is already quite nontrivial. Moreover, the eleven-dimensional
Einstein’s equation for ΛΠ = αβ contains the seven-dimensional Einstein’s equation,
(5.9), together with many more terms. So we can say that the seven-dimensional bosonic
equations of motion solve the eleven-dimensional equations of motion in all the sectors
we have checked. We leave the complete check of the Einstein’s equations to the diligent
reader.
Let us now turn to the bosonic action in (5.4). Part of the action was already calculated.
We have also reproduced the scalar field potential, and this was how we fixed the

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

227

dependence of FΛΠΣΩ on the scalars ΠA i . The condition that the Einstein action in
11 dimensions gives us the Einstein action in 7 dimensions gave us the ansatz for the
11-dimensional vielbein component Eαa . Incidentally, we note also that we found the
0
kinetic terms for the gravitini ψαI and spin 1/2 fermions λiI 0 (without the Q-connection
piece). By requiring that we reproduce them we have fixed the ansatz for the components
of the 11-dimensional gravitino. So we need to reproduce the scalar field and gauge field
AB )2 , and also the S
kinetic terms, − 12 Pαij P αij − 14 (ΠA i ΠB j Fαβ
αβγ ,A terms. The Sαβγ ,A
terms give us the mechanism for ’self-duality in odd dimensions’ at the level of the action
(we have deduced this mechanism from the susy laws).
Let us first look at the Pαij P αij term. It is equal to   

2 
−1 AB CD
−1 α
1
∂ T − 12 BαAB T −1 ∂α T AB − 14 TAC TBD
Bα Bα − BαAB . (5.16)
8 Tr ∂α T
We will look only at the first two terms, the last two being inferred as being the gauge
invariant completion (already the term linear in B should be dictated by gauge invariance,
so this will be a nontrivial check on the algebra). The contributions to this term come from
2
.
the 11-dimensional Einstein action and the Maxwell term, −1/48FΛΠΣΩ
αij
2
from −1/48FΛΠΣΩ come from the components
The contributions to Pαij P 

0
0
0
0
1
× 4E Fαµνρ Fα 0 µ0 ν 0 ρ 0 Gαα Gµµ Gνν Gρρ
− 48
0

0

0

+ 2Fαµνρ Fσ 0 µ0 ν 0 ρ 0 Gασ Gµµ Gνν Gρρ

0

0
0
0
0
+ 14 Fµνρσ Fµ0 ν 0 ρ 0 σ 0 Gµν Gνν Gρρ Gσ σ .

(5.17)

We note that the metric Gµν gives also a gauge field dependence:
Gµν = g µν + 12/5Bαµ Bαν 

µ
µ
= 12/5 CA CBν T AB YC YD T CD − CA YB T AB CCν YD T CD + Bαµ Bαν ,

(5.18)

but since it already has two gauge fields it can contribute only in the last term in (5.17).
The first term contributes (after substituting the ansatze for the various fields)    

2 12 AB −1 

1
V4 1
−1 α
−1
Tr ∂α T ∂ T +
Tr T ∂α T
+ Bα T ∂α T AB ,
(5.19)
4 7
7·5
5
whereas the second term contributes 

4 AB −1
Bα T ∂α T AB .
+
5·7

(5.20)

2 , so we will neglect it.
Finally, the third term in (5.17) contributes only to the B 2 term in Pαij
We ‘see’ that the two contributions add up to the combination required by gauge invariance: 

 
2
1
V4 
Tr ∂α T −1 ∂ α T ) − 8BαAB T −1 ∂α T AB +
Tr T −1 ∂α T ,
(5.21)
4·7
7·5·4
where the last term is gauge invariant by itself (we can freely replace ∂α with 5α ).
0
0
0
0
One would think that we need to add also the term Fαµνρ Fα 0 µ0 ν 0 ρ 0 Gαµ Gµα Gνν Gρρ
which has also two gauge fields and scalars, however it has both two gauge fields
and two derivatives on scalars, so it should cancel with similar terms coming from the

228

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

11-dimensional Einstein action. For the same reason, the gauge field contribution of Gµν
to the first term in (5.17)
R has not been taken into account.
The Einstein action − 12 R, gives for the term with two derivatives on scalars and no
gauge fields 
2 

5
1
Tr T −1 ∂α T +
Tr ∂α T ∂ α T −1 .
(5.22)

5·7·4
7·8
R
But we know that the metric is gauge invariant, and so we expect that R is a gauge
invariant object too, that means that
R complete the previous terms in a gauge
R we can
invariant way. Then the sum of the F 2 and R contributions reproduces (5.16).
AB )2 . It
Next we try to reproduce the kinetic term for the gauge field, − 14 (ΠA i ΠB j Fαβ
has also contributions from the Einstein action and from the antisymmetric tensor kinetic
term, more precisely the component
0

0

0

1
EFµναβ Fµ0 ν 0 α 0 β 0 Gµµ Gνν Gαα Gββ
− 48

0

(5.23)

which gives
3V4 −1 −1 AB CD
T T F F .
(5.24)
20 AC BD αβ αβ
We note that if we did not have any scalar field dependence in Fµναβ , we would get a term
of the type 
−1 −1 AB CD
−1
−1 
AC CB
Fαβ TAB
− Tr T −1 TAB
.
(5.25)
4TAC TBD Fαβ Fαβ + Fαβ

From the Einstein action, if we look only at (∂B)2 terms, we get
Z
V4
−1 −1
CD
TBD ∂α BβAB ∂[α Bβ]
.
(5.26)
− 4 TAC
10
In the same manner as before, we assume that we can complete this term in a gauge
invariant manner to an F 2 term. When we add the two contributions, we get the correct
term,
V4 −1 −1 AB CD
TBD Fαβ Fαβ .
(5.27)
− TAC
4
Finally, let us come to the S terms in the bosonic action. They come from the F 2 , F FA
AB , and
and B 2 terms in the 11-dimensional action. We will treat first the terms with no Fαβ
BC
AB
αβγ
δηζ
ABCDE
DE 

Sαβγ ,A Fδ Fηζ .
afterwards the term with Fαβ , namely, 
2
AB
The F terms contributing to the part with no Fαβ are: 

0
0
0
0
0
0
0
0
1
E 4Fµαβγ Fµ0 α 0 β 0 γ 0 Gµµ Gαα Gββ Gγ γ − 3Gµα Gµ α Gββ Gγ γ
− 48
0
0
0
0
0
0
0
0
+ Fαβγ δ Fα 0 β 0 γ 0 δ 0 Gαα Gββ Gγ γ Gδδ + Fµαβγ Fδ 0 α 0 β 0 γ 0 Gµδ Gαα Gββ Gγ γ . (5.28)
When we substitute the ansatz for the Sαβγ ,A dependence of F and the ansatz for the metric
(Gµν from (5.17)), we get  

−1
2
αβγ
[α βγ δ]
(5.29)
B TAB + TAB 5[α Sβγ δ],A 5 S
B .
5 Sαβγ ,A S
The B 2 terms give

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239 


1 6k 2 
−1
5[α Sβγ δ],A 5[α S βγ δ] B
Sαβγ ,A S αβγ B TAB − 4TAB
5 5 

+ 13  αβγ δηζ Sηζ,A 5α S βγ δ,A .

229

(5.30)

Finally, the F FA term contributes
1 αβγ δηζ
Sηζ,A
20 

5α S βγ δ,A .

αβγ
1
2 Sαβγ ,A SB TAB

+

(5.31)


The condition of cancellation of the terms with 2 derivatives gives k = 5/(6 2), and we
are left with
1 αβγ δηζ
Sηζ,A
12 

5α S βγ δ,A

(5.32)

exactly as we have in the seven-dimensional action. So we have still to recover only the 
SF F term. It comes from two contributions:

B2
3 2 µνρσ
αβγ δηζ
(5.33) 

F 

A
F
+

µνρσ
αβγ
δηζ
48
6(24)2
and they sum up to
i
BC DE
Fηζ ,
√  αβγ δηζ Sαβγ ,A ABCDE Fδ
16 3

(5.34)

as it should. This finishes the analysis of the bosonic action.
In the last part of this section we will make some comments on other ansatze for bosonic
fields which appeared in the literature. Most notably, we will try to see the relation with
the truncation considered in [35] to the bosonic sector of N = 1 gauged supergravity with
gauge group SU(2). A series of papers [10,11,36] also looked at other bosonic truncations
of 11D sugra to 7 dimensions with abelian gauge groups. Consistent truncations to other
dimensions were considered in [38–40].
The SU(2) truncation has the metric, one scalar X, an SU(2) gauge field Ai(1) and a 3form field A(3) . The reduction ansatz in [35], written in form language, is
2
˜ 1/3 ds72 + 2g −2 X3 1
˜ 1/3 dξ 2
=1
ds11
X 
2
˜ −2/3 X−1 cos2 ξ
+ 12 g −2 1
σ i − gAi(1) ,

(5.35)

i

1 −3 −8 2
b(4) = − √
g X sin ξ − 2X2 cos2 ξ
F
2 2 
−2
˜ cos3 ξ dξ ∧ (3)
+ 3X−3 cos2 ξ − 4X−3 1
5
˜ −2 X−4 sin ξ cos4 ξ dX ∧ (3) + sin ξ F(4)
− √ g −3 1
2 2

1
i
+ 2g −1 cos ξ X4 ∗ F(4) ∧ dξ − √ g −2 cos ξ F(2)
∧ dξ ∧ hi
2
1
˜ −1 sin ξ cos2 ξ F i ∧ hj ∧ hk ij k ,
− √ g −2 X−4 1
(2)
4 2

(5.36)

230

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

where hi = σ i − gAi(1) , (3) = h1 ∧ h2 ∧ h3 , σ i are the three left-invariant forms on S3 ,
and
˜ = X−4 sin2 ξ + X cos2 ξ
1

(5.37)

and where the self-duality of F(4) is imposed by hand
1
1
X4 ∗ F(4) = − √ gA(3) + ω(3) ,
2
2
j

i
− 16 gij k Ai(1) ∧ A(1) ∧ Ak(1) .
ω(3) = Ai(1) ∧ F(2)

(5.38)

The corresponding seven-dimensional action is
8
√3 φ
− √2 φ 
1 √10 φ
− 2e 10 − 2e 10 ∗ 1
4e
4
√2 φ
i
i
1 − √10 φ
∗ F(4) ∧ F(4) − 12 e 10 ∗ F(2)
∧ F(2)
2e

L=R∗1−

1
2

∗ dφ ∧ dφ − g 2

1 i
1
i
+ F(2)
∧ F(2)
∧ A(3) − √ gF(4) ∧ A(3),
2
2 2
where X = e

(5.39)

− √φ

. On the other hand our bosonic action is  

e−1 L07d = − 12 R + 14 m2 T 2 − 2TAB T AB + 18 Tr ∂α T −1 ∂ α T 

αβγ
− 12 BαAB T −1 ∂α T AB + 12 T AB Sαβγ ,A SB
10

+

−1 αβγ δηζ AB
1
δ Sαβγ ,A Fδηζ,B
48 me  

2  m−1 −1
1
m−1 −1
−1 AB CD
TAC TBD
e Ω5 [B] −
e Ω3 [B]
Bα Bα − BαAB
+
4
8
16
i
BC DE
Fηζ ,
(5.40)
+ √  αβγ δηζ Sαβγ ,A ABCDE Fδ
16 3

and the metric and field strength in form language are given in (4.88) and (4.128),
˜ =
respectively. A comparison of the seven-dimensional part of the metric tells us that 1
−6/5
−4
and suggests the ansatz for embedding X into TAB : TAB = diag{X, X, X, X, X },
1
and breaking the SO(5) invariance by writing the 4-sphere in terms of 3-spheres as Y A =
eµˆ 2 = 1. Indeed, then we reproduce the form of 1,
˜ and if we also
eµˆ , sin ξ }, with Y
{cos ξ Y
µ5
ˆ
choose B = 0, we get 

2
ˆ ν eνˆ 2
˜ 1/3 ds72 + 1
˜ 1/3X3 dξ 2 + 1
˜ −2/3 X−1 cos2 ξ d Y
eµˆ + 2B µˆ
=1
(5.41)
ds11
Y
and we also reproduce the ansatz for the 11-dimensional antisymmetric tensor at zero
gauge field. The gauge field dependence looks somewhat different. We are thankful to
C.N. Pope, H. Lü and A. Sadrzadeh for pointing to us that the ansatz in (5.35) is in fact
contained in our general ansatz and is the same as the one given in (5.41). We know that
SO(4) ' SU(2) × SU(2). One can restrict then the set of six gauge fields B µˆ νˆ to one of the
two sets of SU(2) gauge fields by imposing a (anti)self-duality condition
ˆ ν ρˆ σˆ
Bρˆ σˆ .
B µˆ νˆ = − 12  µˆ

(5.42)

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

231

As a consequence, with this constraint imposed on our initial set of gauge fields, we are
ˆˆ
left with only three independent gauge fields, namely B i 4 = −1/2Ai(1), with i = 1, 2, 3.
Also, (5.42) implies that
B µˆ νˆ B ρˆ νˆ = 14 δ µˆ ρˆ B 2 = 14 δ µˆ ρˆ B σˆ τˆ B σˆ τˆ .

(5.43)

ˆ ν Y νˆ in (5.41), we get:
Thus, when squaring B µˆ
ˆ ν νˆ µˆ ρˆ ρˆ
Y B Y = δµˆ ρˆ Y µˆ Y ρˆ B 2 =
4B µˆ

3
X 

2
Ai(1) .

(5.44)

i=1

Using the connection between Euclidean coordinates on S3 and Euler angles
θ
θ
ϕ+ψ
ϕ+ψ
cos ,
Y2ˆ = sin
cos ,
2
2
2
2
θ
θ
ψ −ϕ
ψ −ϕ
sin ,
Y4ˆ = sin
sin ,
(5.45)
Y3ˆ = cos
2
2
2
2
we can now easily check that in (5.41) another term in the 4-dimensional line element,
P
(dYµˆ )2 corresponds to the term 1/4(dθ 2 + dϕ 2 + dψ 2 + 2 dϕ dψ cos θ ) =
namely 4µ=1
P3 ˆ 2
1/4 i=1 dσi in (5.35). What remains to be shown to is that the cross terms, linear in the
gauge field also coincide in both relations. This is indeed the case, as we can deduce by
using the set of relations (5.45):
ˆˆ
ˆ ˆ
ˆ ˆ
ˆ ˆ
ˆ ˆ
ˆˆ
ˆˆ
ˆν
dY µˆ Y νˆ = 4B 14 dY 1 Y 4 − dY 4 Y 1 + dY 3 Y 2 − dY 2 Y 3 + B 24 (. . .) + B 34 (. . .)
4B µˆ
Y1ˆ = cos

=−

3
X

Ai(1) σi .

(5.46)

i=1

In fact, we can understand the previous result by organizing the four Yµˆ in a 2 × 2 SU(2)
matrix called G. Then we obtain the SU(2) invariant one-forms σi by considering the
product G −1 dG.
Our ansatz for the field strength of the 11-dimensional 3-index antisymmetric tensor up
to one gauge field and no S(3) reduces in the truncated case to 
−2
1
˜
− √ g −3 X−8 sin2 ξ − 2X2 cos2 ξ + 3X−3 cos2 ξ − 4X−3 1
2 

× A1 ...A5 dY A1 . . . dY A4 Y A5 + dY A1 dY A2 dY A3 B A4 A5
5 ˜ −2 −4
X sin ξ cos ξ dXA1 ...A5 dY A1 dY A2 dY A3 M A4 Y A5 ,
(5.47)
−√ 1
2
eµˆ ) and we have used here a certain rewriting of Fµνρα , namely:
where M A4 ≡ (cos ξ, sin ξ Y
√   

EF 

1
2
T YF
A B C
D
DE
Fµνρα = ABCDE Cµ Cν Cρ Y ∂α
− Bα
3
3
Y ·T ·Y 
A BC 
Y T YC
2
.
(5.48)
+ µνρσ BαAB ∂ σ
3
Y ·T ·Y
Working along the same lines as before, it can be shown that this ansatz coincides with the
one proposed by [35] (for a complete discussion see [61]). The field equations obtained

232

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

from our action (5.40) also coincide with the field equations and the constraints of (5.39).
Hence, as observed by C.N. Pope, H. Lü and A. Sadrzadeh, our general KK ansatz [12]
for the maximal sugra in d = 7 contains as a special case the N = 2 model in [35]. In
conclusion, there is one consistent embedding which contains all (known) subcases in
d = 7 [61].

6. Discussion and conclusions
In this paper we discuss the consistency of the KK reduction of the original formulation
of 11D sugra [1] on AdS7 × S4 [46] using the nonlinear ansatz presented in a previous
letter [12] for the embedding of d = 7 fields in d = 11. Our ansatz for the metric factorizes
into a rescaled 7-dimensional metric and a gauge invariant two form which depends on the
A
B
composite tensor T AB = Π −1 i Π −1 i , where ΠA i are the coset elements for the scalar
fields:  

−1
TAB
2
(DY )B ,
= 1−2/5 gαβ dy α dy β + (DY )A
(6.1)
ds11
Y ·T ·Y
DY A = dY A + 2B AB YB .

(6.2)

The scale factor satisfies 1−6/5 = Y · T · Y . We note that the full internal metric gµν =
−1
∂ν Y B has the geometrical meaning of a metric on an ellipsoid multiplied
14/5 ∂µ Y A TAB
by a conformal factor, namely, gµν = 14/5 ∂µ Z A ∂ν ZA where Z A is constrained to lie on
an ellipsoid. Therefore the overall effect of all scalar fluctuations on the internal metric is
to deform the background sphere into a conformally rescaled ellipsoid.
The 4-form field strength is given by
√ 

2
(T · Y )A5
1
F(4) = A1 ...A5 − (DY )A1 . . . (DY )A4
3
3
Y ·T ·Y  

(T · Y )A4
4
Y A5
+ (DY )A1 (DY )A2 (DY )A3 D
3
Y ·T ·Y 

(T · Y )A5
A A
A A
A A
+ F(2)1 2 F(2)3 4 Y A5
+ 2F(2)1 2 (DY )A3 (DY )A4
Y ·T ·Y 

B
(6.3)
+ d S(3)B Y ,
where S(3)Bαβγ = − √8i Sαβγ ,B is real. It is again gauge invariant but differs from the
3
geometric proposal of Freed, Harvey, Minasian and Moore [57,58] by the dependence
on T . Setting T = I we recover their result but our expression follows from consistency
of the KK program and still satisfies DF(4) = 0. To prove DF(4) = 0, one may use the
Schouten identity [A1 ...A5 YB] = 0. This fixes all relative coefficients in F(4) . Because in
this process we can at most convert one factor T · Y/(Y · T · Y ) into a Y , there do not seem
possible deformations with two factors T · Y/(Y · T · Y ).
A confirmation of our ansatz for F (4) is obtained by evaluating the term F FA in the 11dimensional action. We begin with F F F in d = 12. The terms without bare B’s contain

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

233

F(2) , Y , ∂µ Y and T ’s. Using (3.41) in reverse order and the orthogonality relations in
(4.103), integration over S4 produces 2 Tr F 4 − (Tr F 2 )2 , which is indeed the exterior
derivative in the d = 7 Chern–Simons term. The B terms should not affect this result
because both F (4) and the final result are gauge invariant.
The 4-index antisymmetric auxiliary field B has only a 7-dimensional part:
Bαβγ δ
i
δS (7) A
= √ αβγ δηζ

Y .
δSηζ,A
E
2 3

(6.4)

On d = 7 shell, Bαβγ δ should vanish and since it should contain at most one derivative
to exclude higher derivative terms in the d = 7 action, it can only be proportional to the
field equation of Sαβγ ,A . Since it should mix with Aαβγ to produce self-duality in odd
dimensions, it should have the same spherical harmonic as Aαβγ , and this explains the
factor Y A in (6.4) and rules out an alternative (T · Y )A /(Y · T · Y ).
Our ansatz for the fermions (4.14)–(4.16) is the standard one except for the following
aspects:
(i) There are U matrices connecting the SO(5)c spinor indices of the d = 7 fermions to
the SO(5)g indices of the Killing spinors;

(ii) Factors of γ5 . They are fixed by requiring that ψα (y, x) varies only into the Killing
0
spinor (δ(d = 11)ψα (y, x) = δ(d = 7)ψαI 0 (y)U I I (y, x)ηI (x)).
The consistency of the truncation is proven by obtaining the correct d = 7 susy
transformation laws from d = 11 ones. We checked all transformation rules except: (i) the
variation with more than one fermion field and (ii) some terms in the variation of the
fermions which depend only on scalars. No authors have ever determined the higher order
fermionic terms, but we would like to come back to them in the future. Presumably, they
fix the remaining freedom in U . As to the scalar terms in δψα and δλ, many consistency
checks are satisfied, but this is a complicated problem which deserves a separate study.
Also here we expect that the remaining freedom in U will get fixed.
We have followed de Wit and Nicolai [2,3] by introducing a matrix U (y, x) which
connects the d = 7 fermions to the Killing spinors in the ansatz for the gravitino. 10 This
A
matrix must satisfy equation (4.45), U Y/ = v/ U with vi = Π −1 i YA 13/5 , in order that the
susy transformation laws δBαAB for the gauge fields and δΠA i for the scalars come out
correctly. We have found general solutions to this equation, but for most of our results, the
explicit form of U is not needed.
The ansatz for the antisymmetric tensor field strength FΛΠΣΩ was determined in three
steps: first, the dependence on the gauge fields was uniquely fixed by requiring that the
11-dimensional susy variation of FΛΠΣΩ matches the 7-dimensional susy variation of our
ansatz for FΛΠΣΩ . Next, the ansatz for the embedding of the independent d = 7 fluctuation
Sαβγ ,A was derived together with the ansatz for the d = 11 auxiliary field BMNP Q . Finally,
the dependence of FΛΠΣΩ on the d = 7 scalar fields was fixed by requiring that one
obtains the correct scalar field potential in d = 7 and the correct dependence of the d = 7
10 Actually, in their work, the matrix U interpolates between spinor indices, whereas in our approach U connects
the SO(5)g indices I of the Killing spinors with the SO(5)c indices I 0 of the d = 7 fermions.

234

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239
A

B

sugra transformation rules on the tensor Tij = (Π −1 )i (Π −1 )j δAB where ΠA i is the
group vielbein for the scalars.
For completeness of our results, we have derived the 7-dimensional bosonic action
and bosonic equations of motion at zero gauge field from the corresponding objects in
11 dimensions. Since in the bosonic sector all criteria for consistency are satisfied (the
consistency of the transformation rules, equations of motion and even the action agrees) we
infer that our results also prove the consistency of the bosonic truncation. We have shown
the relation to other ansätze found in the literature for consistent truncation to subsets of
bosonic fields.
We have also explained the origin of self-duality in odd dimensions. One has to use a
first-order action for FΛΠΣΩ in d = 11 to obtain a self-dual action for Sαβγ ,A in d = 7
which is linear in derivatives. The d = 7 fluctuation Sαβγ ,A not only appears in the d = 11
curvatures Fµαβγ and Fαβγ δ , but also in the d = 11 auxiliary field B in (6.4), namely, in
the form B ∼ S + DS, and substituting the ansatz for B and F (S) into d = 11 action, we
recovered the self-dual action in d = 7. We believe that in all cases, self-dual actions can
be obtained from KK reduction of a first-order formalism for the corresponding field.
It may be useful to point out why in our opinion the original d = 11 sugra theory is to be
preferred for purposes of compactification over the SU(8) version of de Wit and Nicolai.
(i) Their theory has a local SO(3, 1) × SU(8) group in tangent space. The SO(3, 1)
becomes the Lorentz group in d = (3, 1) dimensions, and thus it becomes
immediately clear that for compactifications to other dimensions we cannot use
this theory. (For compactification to d = (6, 1) we would need an 11D theory with
a SO(6, 1) × SO(5) tangent group. For compactifications to other dimensions one
would need other versions of 11D sugra than the SU(8) version, namely, versions
with SO(2, 1) × SO(16), SO(4, 1) × U Sp(8) and SO(5, 1) × SO(5) × SO(5) tangent
group. These were found in [63,64].) On the other hand, the standard version of
d = 11 sugra can be used for compactifications to any dimension.
(ii) If we gauge fix in the SU(8) theory the group SU(8) to SO(7) and compare to the
standard d = 11 sugra with SO(10, 1) gauged fixed to SO(3, 1) × SO(7) then in the
SU(8) theory the equations of motion correspond to both equations of motion and
Bianchi identities in the usual d = 11 sugra. This means that the two gauged fixed
theories are only equivalent on-shell. Of course, also in Cremmer and Julia’s theory
for N = 8 sugra in d = 4, one needs to go temporarily on-shell to dualize certain
fields. Note that in our paper we never need to dualize and always stay off-shell.
(iii) In fact, in the work of de Wit and Nicolai there is no action for their SU(8) theory,
only field equations. To quote Ref. [4], page 389: “. . . the d = 11 Lagrangian
depends explicitly on the antisymmetric tensor field AMNP for which no expression
exists in terms of the SU(8) covariant expressions used in this paper.” In the field
equations only the field strength appears, but in the action bare A’s appear.
(iv) To lift solutions of d = 4 gauged sugra to solutions of the usual d = 11 sugra such
as in [10], one would need the analog of our results for d = 4. The ansatz of de Wit
and Nicolai could be used to lift such solutions to their SU(8) theory. However, all
recent work on the consistency of truncations of the KK reductions to d = (3, 1)

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

235

has used the standard d = 11 theory [62]. In particular, all work on M-theory and
membranes is based on the standard sugra theory.

Acknowledgements
We would like to thank I. Park for collaboration at the early stages of this work, B. de Wit
and H. Nicolai for useful discussions on their work and C.N. Pope, H. Lü and A. Sadrzadeh
for pointing out that our bosonic ansätze reduce to theirs [35] when we further truncate
N = 4, d = 7 to N = 2, d = 7 gauged sugra.
Appendix A
A.1. Conventions and gamma matrix algebra
We denote in 7-dimensional Minkowski space the coordinates by y, flat vector indices
by a, b, c . . . and curved vector indices by α, β, γ . . . Similarly, we denote in 4-dimensional
Euclidean space the coordinates by x, flat indices by m, n, p, and curved vector indices by
µ, ν . . . Eleven-dimensional vector indices are denoted by M, N, P for flat indices and
Λ, Π, Σ . . . for curved indices. The SO(5) ' Usp(4) gauge group is denoted by SO(5)g
and A, B . . . = 1, 5 are SO(5)g vector indices and I, J . . . = 1, 4 are Usp(4) indices (or
SO(5)g spinor indices). The scalars form a coset Sl(5, R)/SO(5) and to avoid confusion
we denote the composite subgroup by SO(5)c , with vector indices i, j . . . and spinor indices
I 0 , J 0 , . . . = 1, 4.
The d = 11 metric gΛΠ has signature mostly plus, (− + · · · +). The Clifford algebra in
d = 11 reads
0M 0N + 0N 0M = 2ηMN ,

M, N = 0, . . . , 10.

(A.1)

We introduce Dirac matrices in d = 7 and d = 4:
τa τb + τb τa = 2ηab ,

a, b = 0, . . . , 6,

(A.2)

γm γn + γn γm = 2δmn ,

m, n = 1, . . . , 4.

(A.3)

They are used to construct the d = 11 Dirac matrices:
0a = τa ⊗ γ5

for a = 0, 6 and 06+m = 1 ⊗ γm

for m = 1, 4,

(A.4)

where γ5 = γ1 . . . γ4 hence γ52 = 1. We choose τ0 = τ1 . . . τ6 , hence 00 = 01 . . . 010 . We
normalize the  symbols as  0...10 =  0...7 =  1...4 = +1, so that in d = 7
(−1)[k/2+1]
,
(A.5)
(7 − k)!
where all antisymmetrizations are with strength one. A similar duality relation between 0
matrices holds in d = 11:
τ [α1 . . . τ α7 ] ≡ τ α1 ...α7 =  α1 ...α7 → τ α1 ...αk =  α1 ...α7 ταk+1 ...α7

0 Λ1 ...Λk =

(−1)[k/2+1] Λ1 Λ2 ...Λ11 

0Λk+1 ...Λ11 .
(11 − k)!

(A.6)

236

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

Also, for general d,
α ]

[α1
. . . δβkk] . 
α1 ...αk αk+1 ...αd β1 ...βk αk+1 ...αd = −k!(d − k)! δ[β
1

(A.7)


A definition of γ5 is obtained from eiαγ5 = cos α + i sin α γ5 . Choosing α = π/2 we
π
get γ5 = −iei 2 γ5 , and thus
1/2

γ5

=−

1−i
(1 + iγ5 ),
2

−1/2

γ5

=−

1+i
(1 − iγ5 ).
2

(A.8)

Then we also have:
1/2 T

Cγ5 C −1 = γ5
1/2

−1/2

,

1/2

γµ γ5

−1/2

= −iγ5

γµ ,

(A.9)

= γ5 γ5 and Cγµ C −1 = −γµT , see Appendix A.2. Some useful formulas for
where γ5
gamma matrices in d dimensions are:
1/2

0a 0 b1 ...bn 0 a = (−)n (d − 2n)0 b1 ...bn ,
0a 0 b1 ...bn = 0a b1 ...bn + nδa[b1 0 b2 ...bn ] ,
[b1
0a1 ] b2 ...bn ] + n(n − 1)δa[b21 δab12 0 b3 ...bn ] ,
0a1 a2 0 b1 ...bn = 0a1 a2 b1 ...bn + nδ[a
2
[b1
[b1 b2
0a2 a3 b2 ...bn ] + 2n(n − 1)δ[a
δ 0a1 ] b3 ...bn ]
0a1 ...a3 0 b1 ...bn = 0a1 a2 a3 b1 ...bn + 2nδ[a
1
3 a2

+ n(n − 1)(n − 2)δa[b31 δab22 δab13 0 b4 ...bn ] .
As representation for the SO(5) Clifford algebra we take 

γ A = iγ µ γ5 , γ5 .
In 5 dimensions,  

Cγ A I J Cγ A KL = +2(ΩI K ΩJ L − ΩJ K ΩI L ) − ΩI J ΩKL,  

Cγ AB I J Cγ AB KL = 4(ΩI K ΩJ L + ΩI L ΩJ K ),

(A.10)

(A.11)

(A.12)
(A.13)

because these tensors are USp(4) invariant tensors with definite symmetry properties.
Hence they should be constructed from the symplectic metric CI J .We choose a gamma
matrix representation such that CI J = ΩI J . (This matrix CI J is thus the charge
(5)
(4)
in 5 dimensions, see below. It equals C−
.)
conjugation matrix C (5) = C+
We lower USp(4) indices with ΩI J as follows:
λI = λJ ΩJ I ,

(A.14)

eI J as
and raise them with Ω
eI J λJ ,
λI = Ω

(A.15)

eI J ΩI K = δ J . Of course, Ω
e is −Ω −1 and can be obtained from
eI J is defined by Ω
where Ω
K
eI J = Ω I J . We have found it convenient
ΩI J by raising indices with Ω˜ I J . The result is Ω
I
J
I
J
e
to use a different symbol for Ω , namely, Ω .

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

237

A.2. Charge conjugation matrices and modified Majorana spinors
In even dimensions there are two charge conjugation matrices, C(+) and C(−) , satisfying
−1
) = ±γ µ,T . In odd dimensions there is only one charge conjugation matrix,
C(pm) γ µ C(±
either C(+) or C(−) . They are either symmetric or antisymmetric, and all their properties
are independent of the representation chosen. For a general discussion of Majorana and
modified Majorana spinors in Minkowski or Euclidean space and charge conjugation
matrices, see [59].
In 11 Minkowski dimensions, there is only one C matrix. It satisfies
T
,
C (11) 0M C (11)−1 = −0M

(A.16)

(11)
. Furthermore, C (11),T = −C (11) .
hence C (11) = C(−)
(7)
, but is symmetric:
In 7 Minkowski dimensions, C (7) = C(−)

C (7) τa C (7)−1 = −τaT ,
(4)
In d = 4 both C(+)
(4)
.
C (4) = C(−)

and

(4)
C(−)

C (7),T = C (7) .

(A.17)

exist. Both are antisymmetric, but C (11)

= C (7) ⊗ C (4) , where

Then Cγµ is symmetric and Cγ5 antisymmetric (since C T = −C). It also follows that
for the 5-dimensional gamma matrices, Cγ A in (A.12) are antisymmetric and the matrices
Cγ [AB] in (A.13) are symmetric. This means that (Cγ A )[I J ] and (Cγ [AB] )(I J ) give the
isomorphisms between the 5, respectively the 10 representations of SO(5) and Usp(4)
(SO(5) ' Usp(4)).
The Majorana condition in 11 dimensions is
9 T C (11) = i9 † 0 0 .

(A.18)

In 7 Minkowski dimensions we can only define a modified Majorana condition,
eJ I = i(λJ )† τ 0 ≡ λ¯ J .
(λI )T C (7) Ω

(A.19)

In the text we always define λ¯ I ≡ λTI C(7) . Using (A.19), the space–time spinors satisfy
λ¯ I = (λJ )† iτ 0 ΩJ I .

(A.20)

To determine which Majorana condition we need in d = 4 Euclidean space, we
decompose the anticommuting 11-dimensional Majorana spinors 9 into 4 anticommuting
7-dimensional modified Majorana spinors times corresponding commuting 4-dimensional
spinors ηI . The spinors η must satisfy the following modified Majorana condition 
T (4) 

(A.21)
ηK C(−) ΩKJ = − ηJ γ5 .
(4)

(4)

Since in 4 Euclidean dimensions there are 2 choices for C (4) , we can use C(+) = C(−) γ5 ,
(4)
(4)−1
γm C(+)
= γmT , in terms of which (A.21) takes the same form as (A.19):
satisfying C(+) 
T (4) 

(A.22)
ηK C(+) ΩKJ = ηK .
(4)

(4)

Note that both (A.22) as well as the same condition with C(−) instead of C(+) satisfy the
consistency condition obtained by taking its complex conjugate and applying the relation
twice [59].

238

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

(4)
Also, in the text we always define η¯ I ≡ (ηI )T C(−)
. Using (A.21), the Killing spinors
satisfy 

eI J .
(A.23)
η¯ I = − ηJ γ5 Ω
0

In the fermionic ansätze we decompose the d = 11 spinors into λI 0 (y)U I I ηI (x). Using
the Majorana conditions in the (10, 1), (6, 1) and (4, 0)-dimensional spaces one obtains
the condition
0 ∗
0
eJ I .
(A.24)
U I I = ΩI 0 J 0 U J J Ω
e derived in the text, we deduce that U is unitary.
e T =Ω
Combined with the relation U ΩU
Hence the matrices U are matrices of the group USp(4), namely the intersection of U (4)
and Sp(4, C).
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]

E. Cremmer, B. Julia, J. Sherk, Phys. Lett. B 76 (1978) 409.
B. de Wit, H. Nicolai, Nucl. Phys. B 243 (1984) 91.
B. de Wit, H. Nicolai, N.P. Warner, Nucl. Phys. B 255 (1985) 29.
B. de Wit, H. Nicolai, Nucl. Phys. B 274 (1986) 363.
B. de Wit, H. Nicolai, Nucl. Phys. B 281 (1987) 211.
M.A. Awada, B.E.W. Nilsson, C.N. Pope, Phys. Rev. D 29 (1984) 334.
B.E.W. Nilsson, Phys. Lett. B 155 (1985) 54.
M.J. Duff, B.E.W. Nillsson, C.N. Pope, N.P. Warner, Phys. Lett. B 149 (1984) 90.
M.J. Duff, C.N. Pope, Nucl. Phys. B 255 (1985) 355.
M. Cvetic, M.J. Duff, P. Hoxha, J.T. Liu, H. Lü, J.X. Lu, R. Martinez-Acosta, C.N. Pope, H. Sati,
T.A. Tran, hep-th/9903214.
M. Cvetic, J.T. Liu, H. Lü, C.N. Pope, hep-th/9905096.
H. Nastase, D. Vaman, P. van Nieuwenhuizen, hep-th/9905075.
B. de Wit, H. Nicolai, Phys. Lett. B 108 (1981) 285; Nucl. Phys. B 208 (1982) 323.
B. de Wit, P. Fayet, P. van Nieuwenhuizen (Eds.), Supersymmetry and Supergravity ’84, Proc.
Trieste School, April 1984.
E. Witten, in: Shelter Island Conference II, 1983, p. 227.
J. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231, hep-th/9711200.
S. Gubser, I.R. Klebanov, A.M. Polyakov, Phys. Lett. B 428 (1998) 105, hep-th/9802109.
E. Witten, Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150 and private communication.
W. Muck, K.S. Viswanathan, Phys. Rev. D 58 (1998) 041901.
D.Z. Freedman, S.D. Mathur, A. Matusis, L. Rastelli, Nucl. Phys. B 546 (1999) 96.
H. Liu, A.A. Tseytlin, Nucl. Phys. B 533 (1998) 88.
G. Chalmers, H. Nastase, K. Schalm, R. Siebelink, Nucl. Phys. B 540 (1999) 247.
G. Arutyunov, S. Frolov, Phys. Rev D 60 (1999) 026004.
H. Liu, A.A. Tseytlin, Phys. Rev. D (1999) 086002.
D.Z. Freedman, S.D. Mathur, A. Matusis, L. Rastelli, Phys. Lett. B 452 (1999) 61.
G. Chalmers, K. Schalm, Nucl. Phys. B 554 (1999) 215, hep-th/9901144.
E. D’Hoker, D.Z. Freedman, Nucl. Phys. B 544 (1999) 612; Nucl. Phys. B 550 (1999) 261.
E. D’Hoker, D. Freedman, S. Mathur, A. Matusis, L. Rastelli, hep-th/9903196.
R. Corrado, B. Florea, R. McNees, Phys. Rev. D 60 (1999) 085011.
F. Bastianelli, R. Zucchini, Class. Quant. Grav. 16 (1999) 3673, Phys. Lett. B 467 (1999) 61,
hep-th/9909179.

H. Nastase et al. / Nuclear Physics B 581 (2000) 179–239

[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]

239

S. Lee, S. Minwalla, M. Rangamani, N. Seiberg, Adv. Theor. Math. Phys. 2 (1998) 697.
O. Aharony, S.S. Gubser, J. Maldacena, H. Ooguri, hep-th/9905111.
V. Chattorpadhyay, A. Karlhede, Phys. Lett. B 139 (1984) 279.
C.N. Pope, K.S. Stelle, Phys. Lett. B 198 (1987) 151.
H. Lü, C.N. Pope, hep-th/9906168.
M. Cvetic, S.S. Gubser, H. Lü, C.N. Pope, hep-th/9905096.
H. Lü, C.N. Pope, T.A. Tran, hep-th/9909203.
E. Cremmer, B. Julia, H. Lü, C.N. Pope, hep-th/9909099.
M. Volkov, hep-th/9910116.
A.H. Chamseddine, M.S. Volkov, Phys. Rev. D 57 (1998) 6242.
A.H. Chamseddine, W.A. Sabra, hep-th/991180.
P.K. Townsend, K. Pilch, P. van Nieuwenhuizen, Phys. Lett. B 136 (1984) 38.
S. Deser, R. Jackiw, Phys. Lett. B 139 (1984) 371.
K. Pilch, P.K. Townsend, P. van Nieuwenhuizen, Phys. Lett. B 137 (1984) 443.
A. Karlhede, U. Lindström, M. Roˇcek, P. van Nieuwenhuizen, Phys. Lett. B 186 (1987) 96;
Class. Quant. Grav. 4 (1987) 549.
K. Pilch, P. van Nieuwenhuizen, P.K. Townsend, Nucl. Phys. B 242 (1984) 377.
B. Julia, S. Silva, hep-th/9911035.
D.Z. Freedman, P. van Nieuwenhuizen, S. Ferrara, Phys. Rev. D 13 (1976) 3214.
S. Deser, B. Zumino, Phys. Lett. B 62 (1976) 335.
L. Castellani, P. Fre, F. Giani, K. Pilch, P. van Nieuwenhuizen, Ann. Phys. 146 (1983) 35.
M. Pernici, K. Pilch, P. van Nieuwenhuizen, Phys. Lett. B 143 (1984) 103.
P. van Nieuwenhuizen, Class. Quant. Grav. 1 (1984) 339.
A. Eastaugh, P. van Nieuwenhuizen, in: Kyoto Summer Institute, 1985, p. 0001, Preprint ITPSB-85-43.
M. Gunaydin, N.P. Warner, P. van Nieuwenhuizen, Nucl. Phys. B 255 (1985) 63.
P. van Nieuwenhuizen, Class. Quant. Grav. 2 (1985) 1.
L. Mezincescu, P.K. Townsend, P. van Nieuwenhuizen, Phys. Lett. B 143 (1984) 103.
J.A. Harvey, R. Minasian, G. Moore, hep-th/9808060.
D. Freed, J.A. Harvey, R. Minasian, G. Moore, hep-th/9803205.
P. van Nieuwenhuizen, in: Proc. Relativity, Groups and Topology II, Les Houches, 1983, p. 823,
for an earlier but less detailed acount, see [60].
P. van Nieuwenhuizen, Phys. Rep. 68 (1981) 189.
H. Nastase, D. Vaman, in preparation.
M. Cvetic, H. Lü, C.N. Pope, hep-th/9910252.
H. Nicolai, Phys. Lett. B 187 (1986) 316.
B. Drabant, M. Töx, H. Nicolai, Class. Quant. Grav. 4 (1989) 255.

Nuclear Physics B 581 (2000) 240–256
www.elsevier.nl/locate/npe

Noncommutative field theory and spontaneous
symmetry breaking
Bruce A. Campbell, Kirk Kaminsky
Department of Physics, University of Alberta, Edmonton, Alberta, Canada T6G 2J1
Received 6 April 2000; accepted 3 May 2000

Abstract
We investigate the noncommutative analogue of the spontaneously broken linear sigma model
at the one-loop quantum level. In the commutative case, renormalization of a theory with a
spontaneously broken continuous global symmetry depends on cancellations that enable the limited
set of counterterms consistent with that symmetry to remove the divergences even after its
spontaneous breaking, while preserving the masslessness of the associated Goldstone modes. In
the noncommutative case, we find that these cancellations are violated, and the renormalized oneloop correction to the inverse pion propagator explicitly yields a mass shift which depends on the
ultraviolet cutoff. Thus, we cannot naively take the ultraviolet cutoff to infinity first, and then take
the external momentum to zero to verify Nambu–Goldstone symmetry realization. However, from the
Wilsonian perspective where the cutoff is fixed and physical, the zero external momentum limit of
the inverse pion propagator still vanishes, and implies the masslessness of the pion fields at oneloop. This is another demonstration of the failure of ultraviolet and infrared limits to commute
in noncommutative field theories, and signals the incompatibility of Nambu–Goldstone symmetry
realization with the continuum renormalization of these theories.  2000 Elsevier Science B.V. All
rights reserved.

1. Introduction
In this paper we undertake a perturbative analysis of quantum field theories on
noncommutative Rn which exhibit the spontaneous breaking of a continuous global
symmetry. The perturbative expansion of noncommutative field theories has been the
subject of much recent investigation (see, for example, references [1–23]). In the original
work of Filk [1], it was suggested that despite being (infinitely) nonlocal, these theories
exhibit the same divergence structure as their commutative counterparts (due to nonplanar
oscillatory damping). More recently, Minwalla, Van Raamsdonk, and Seiberg [13] have
shown that the effective cutoff of nonplanar graphs at one-loop in scalar theories replaces
0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 2 6 9 - 8

B.A. Campbell, K. Kaminsky / Nuclear Physics B 581 (2000) 240–256

241

a UV divergence (which would ordinarily be cancelled with a counterterm) with an IR
divergence in the external momenta:
Λ2eff =

1
1
,

p◦p
1/Λ2 + p ◦ p

Λ2 → ∞

(1)

They find that UV and IR limits do not commute, and suggest that UV divergences that
persist at higher orders can be interpreted as IR divergences.
We continue these investigations, and examine global spontaneous symmetry breaking
in the simplest case of the (noncommutative) linear sigma model. In particular, we wish
to investigate the status of Goldstone’s theorem at one-loop for the noncommutative
case. As is well known, the renormalizability of spontaneously broken theories is more
subtle because the number of counterterm vertices exceeds the number of renormalization
parameters (eight and three respectively in the linear sigma model). In the standard
commutative case, the renormalizability of the theory, and the persistence of Goldstone’s
theorem ensuring the masslessness of the classical pions at the quantum level, involve
a delicate cancellation between the relevant graphs and the pion propagator counterterm
at zero external momentum [24–27]. The pion propagator counterterm is however fixed
by the sigma tadpole counterterm, which in turn is fixed by the usual renormalization
condition imposed on tadpoles: namely that they vanish so that the vev of the sigma is not
renormalized. Thus, the persistence of masslessness of the pions (Goldstone’s theorem) at
the quantum level is a genuine prediction of the quantum field theory. We now generalize
this standard textbook calculation to the noncommutative case.
We will find that the cancellation in the calculation of the pion mass renormalization
is violated already at one loop order, and the 1PI renormalized (i.e., after adding the
counterterm fixed by the tadpole renormalization condition) effective action depends
explicitly on the UV cutoff Λ, so the naive continuum limit does not exist; there is no more
counterterm freedom to subtract off these divergences. Put another way, turning on the
noncommutativity parameter induces UV cutoff dependence in the one-loop corrections to
the renormalized pion propagator, and renders renormalization inconsistent with Goldstone
symmetry realization. Viewed however as a Wilsonian effective theory where Λ is fixed,
the limit of the mass correction as the external momentum is taken to zero still vanishes,
so that Goldstone’s theorem is satisfied for the Wilsonian action. The difference between
the Wilsonian effective theory, and the putative continuum renormalized theory is a direct
consequence of the noncommutativity of the UV and IR limits in noncommutative field
theories.
The basic idea behind this rather elementary calculation is that the noncommutativity
does not affect the tadpole calculation as no external momentum flows into a trilinear
tadpole vertex. Insofar as the sigma tadpole measures quantum corrections to the sigma vev,
this suggests that the order parameter for spontaneous symmetry breaking is insensitive
to the underlying noncommutativity. Furthermore, since the tadpole counterterm, and
the pion propagator counterterm are (essentially) the same (modulo the momentum
dependent wavefunction renormalization parameter), the latter is therefore unmodified,
and fixed with respect to renormalization. Now however, the re-weighting of planar

242

B.A. Campbell, K. Kaminsky / Nuclear Physics B 581 (2000) 240–256

graphs (due to noncommutativity) with respect to commutative graphs, and the distinct
behaviour of new nonplanar graphs occurring in the one-loop contributions to the pion
propagator, lead to an inexact UV cutoff cancellation with the associated counterterm. We
emphasize that this cannot be evaded by simply imposing that the propagator counterterm
cancel the cutoff dependence of the one-loop planar and nonplanar diagrams (in effect
changing the renormalization scheme), because now the sigma tadpole corrections will
unavoidably diverge. This is a direct consequence of the relationship between the
counterterm vertices in a spontaneously broken quantum field theory. In the following we
calculate the one-loop renormalization of the inverse pion propagator in the linear sigma
model of noncommutative field theory, and demonstrate the incompatibility of continuum
renormalization with spontaneous symmetry breaking.

2. Background and formalism
In this section we introduce the noncommutative geometry and linear sigma model
formalism we will need for the calculations of the next section in a reasonably selfcontained way. The underlying noncommutative Rn is labelled by n coordinates satisfying 
µ ν
(2)
xˆ , xˆ = i θ µν , θ µν ∈ C
so that the commutators are C-numbers. Let Ax be the associative algebraic structure
naturally defined by (2). We are naturally interested in functions defined over the
noncommutative spacetime. Following Weyl [28] define
Z
1
µ
(3)
dn k eikµ xˆ f˜(k),
O(f ) =
(2π)n/2
where f˜ is the Fourier transform of f :
Z
1
µ
˜
dn x e−ikµ x f (x).
f (k) =
n/2
(2π)

(4)

This uniquely associates an operator in O(f ) ∈ Ax with a function of classical variables:
replacing the commuting variables x with operators xˆ in a symmetric fashion. The product
of two such operators is then defined in the obvious way
Z
1
µ
ν
˜
dn k dn p eikµ xˆ eipν xˆ f˜(k)g(p)
O(f ) · O(g) =
n
(2π)
Z
1
µ i
µν
˜
(5)
dn k dn p ei(kµ +pµ )xˆ − 2 kµ θ pν f˜(k)g(p),
=
n
(2π)
where on the second line we have used the Baker–Campbell–Hausdorff lemma
eA eB = eA+B−1/2[A,B]+··· ,
and the fact that for the canonical structure (2), the higher commutators vanishes. This
allows us to establish a homomorphism, O(f ) · O(g) = O(f ∗ g), between this operator
product (·) and the Moyal [29] product (∗) of ordinary functions:

A.B. K. Campbell. Kaminsky / Nuclear Physics B 581 (2000) 240–256 (f ∗ g)(x) = 1 (2π)n =e Z + 2i θ µν ∂y∂µ dn k dn p ei(kµ +pµ )x e− 2 kµ θ µ ∂ ∂zν .

f (y)g(z).

the noncommutative generalization of this symmetry will manifestly pose no problem with respect to the Moyal product. Renormalizability (of the commutative case) in four dimensions implies the Lagrangian: 2 λ  i 2 2 1 1 φ (7) L = ∂µ φ i ∂ µ φ i + µ2 φ i − 2 2 4 with implicit sums over the internal i index. . the classical potential is minimized by a constant field φ0i configuration such that 2 µ2 (9) λ and the O(N) symmetry is spontaneously broken. we can study field theories defined by R classical actions of the usual form S = dn x L[φ] but with the ∗-product of fields. Note the rescaling of the coupling λ to avoid awkward factors of 1/6. . our sigma-model conventions will be essentially those of Peskin and Schroeder [30]. . The linear sigma-model Lagrangian (7) rewritten in terms of these fields. for example. N − 1. . The global O(N) symmetry acts as φ i → R ij φ j . which explicitly degenerates into the pointwise product if one of the factors is a spacetime constant. Since this determines only the length of the vector φ i . σ ). the rotational invariance allows us to choose coordinates so that µ (10) φ0i = (0. v ≡ √ . . (8) where R is a spacetime-constant N ×N orthogonal matrix.y. . [22]). Because the symmetry is global (so R is constant). At this point we specialize to four dimensions for the remainder of this paper. Furthermore. For µ2 > 0. . The commutative linear sigma model involves a set of N interacting real scalar fields φ i (x) with a continuous O(N) internal symmetry. (11) with hσ i = 0. 0. In particular. k = 1. . v). i µν p ν 243 f˜(k)g(p) ˜ (6) This induced homomorphism allows us to view the algebra of functions on noncommutative Rn as the algebra of ordinary functions on commutative Rn with the Moyal ∗-product instead of the usual pointwise product.z→x . λ φ0i = Then define σ = φ0n − v such that φ i = (π k . 0. and Wick rotated to Euclidean space (via x 0 = −ixE0 ) yields the following (commutative) action:  Z  1 1 1 λ  k 2 2 4 π SE = − d x ∂µ π k ∂ µ π k + ∂µ σ ∂ µ σ + 2µ2 σ 2 + 2 2 2 4 . In particular we will not need to worry about the intricacies of the noncommutative generalization of gauge invariance (see. Now consider the spontaneously broken linear sigma model.

Campbell. and π k π k σ σ . The renormalization counterterm structure determined from the symmetric theory (if the O(N) symmetry is to hold quantum mechanically) is given by: 2 δλ  i 2 2 δZ 1 ∂µ φ i ∂ µ φ i + δµ φ i + φ 2 2 4  2 δZ  2 δZ 1 ∂µ π k + (∂µ σ )2 + δµ + δλ v 2 π k = 2 2 2   2 1 2 2 + δµ + 3δλ v σ + δµ v + δλ v 3 σ + δλ vσ π k + δλ vσ 3 2 δλ  k 2 2 δλ 2 k 2 δλ 4 π + + σ π + σ . so we will include both orderings for each interaction in the model with unit total weighting. However. As discussed above.A. from (6). Z Z (15) d4 x ∂φ ∗ ∂φ = d4 x ∂φ · ∂φ. Thus the most general noncommutative.244 B. Interactions in higher powers of the fields are modified. K. 4 2 (12) which reveals explicitly the masslessness of the N − 1 pions of Goldstone’s theorem at tree level. −LE. by assuming appropriate asymptotic conditions on the fields. (13) 4 2 4 Now consider the noncommutative generalization of this theory. Thus dropping total derivatives. under the spacetime integral.nc = − d x ∂µ π k ∂ µ π k + ∂µ σ ∂ µ σ + 2µ2 σ 2 2 2 2     λ λ + f π k ∗ π k ∗ π l ∗ π l + (1 − f ) π k ∗ π l ∗ π k ∗ π l 4 4  λ  λ k + σ ∗ σ ∗ σ ∗ σ + vλσ ∗ π ∗ π k + f π k ∗ π k ∗ σ ∗ σ 4 2 . spontaneously broken linear sigma model action in Euclidean space:  Z  1 1 1 4 SE. as we will see in a moment. we have: Z Z d4 x φ ∗ φ = d4 x φ · φ. and yield nontrivial phase factors in the momentum space Feynman rules. For example.ct = i (φ1 ∗ φ2 )(x) = φ1 (x)φ2 (x) + θ µν ∂µ φ1 (x)∂ν φ2 (x) + · · · 2    i = φ1 (x)φ2 (x) + θ µν ∂µ φ1 (x)∂ν φ2 (x) − φ1 (x)∂µ ∂ν φ2 (x) + · · · 2 (14) = φ1 (x)φ2 (x) + total derivative + · · · . where · · · represent the higher terms in θ that behave identically with respect to this calculation. There are two possible orderings for the noncommutative generalizations of the quartic terms π k π k π l π l . Kaminsky / Nuclear Physics B 581 (2000) 240–256  2 λ 2 λ + σ 4 + vλσ π k + π k σ 2 + λvσ 3 . an elementary but important result is that the quadratic part of the action is identical with commutative theory because θ µν is antisymmetric. the effect at the Lagrangian level of the noncommutativity is to replace the (implicit) pointwise product of fields with the Moyal ∗-product.

Kaminsky / Nuclear Physics B 581 (2000) 240–256    λ + (1 − f ) π k ∗ σ ∗ π k ∗ σ + λvσ ∗ σ ∗ σ . .A. We have both three and four point vertices in the spontaneously broken linear sigma model. Throughout solid lines refer to the sigma. = −2vλδ cos 2 ij (20) . we will need the Feynman rules. To summarize we write the symmetrized Feynman rules for the spontaneously broken noncommutative linear sigma model using (16) and (18). Campbell. Note that the sum contains n(n − 1)/2 terms. We will prove this in the appendix inductively. + 2µ2 (19) The symmetrized vertices (all momenta flow into the vertex) of the theory are now given by:   p1 × p2 . and dotted lines refer to the pions. but not under arbitrary permutations of the momenta. = i=1 (17) i=1 where pi × pj ≡ piµ θ µν pj ν . . 2 245 (16) and similarly for the counterterm structure. Since we wish to examine the theory at the quantum level. In momentum space the Moyal product (6) of n fields yields Z d4 x φ1 (x) ∗ · · · ∗ φn (x) ! ! Z Y n n X i P d4 pi δ pi e− 2 i<j pi ×pj φ˜1 (p1 ) · · · φ˜ n (pn ). p2 = p2 1 . a fact that will be crucial for our analysis. . Note that the global O(N) symmetry of the model implies that only a single f can occur. the quadratic terms are not modified. Otherwise the Feynman rules will be identical to the linear sigma model. As discussed above. K. . Thus the only modification to the Feynman rules is due to the phase factor i V (p1 . so the propagators are not modified from the commutative case: = δ ij . pn ) = e− 2 P i<j pi ×pj (18) at each vertex of n fields with pi the momentum into the vertex from the ith field. and by momentum conservation is invariant under cyclic permutations.B.

+ (1 − f ) cos 2 2      p3 × p4 p1 × p2 cos = −2λ cos 2 2     p2 × p4 p1 × p3 cos + cos 2 2     p2 × p3 p1 × p4 cos . In the calculations of the next section we will only be interested in one and two point amplitudes. . (21) 2 2 2      p3 × p4 p1 × p2 = −2λδ ij f cos cos 2 2   p1 × p3 p2 × p4 + . + (1 − f ) cos 2 2 (22) (23) (24) Working with the totally symmetrized vertices allows us to capture both planar and nonplanar terms at once. Campbell.A. so the counterterms from (13) we will need are given by  = −δ ij δµ + δλ v 2 − δz p2 . K. Kaminsky / Nuclear Physics B 581 (2000) 240–256        p1 × p2 p1 × p3 p2 × p3 = −2vλ cos + cos + cos . + cos 2 2       p3 × p4 p1 × p2 ij kl = −2λ δ δ f cos cos 2 2   p1 × p3 p2 × p4 + + (1 − f ) cos 2 2      p2 × p4 p1 × p3 cos + δ ik δ j l f cos 2 2   p1 × p2 p3 × p4 + + (1 − f ) cos 2 2      p2 × p3 p1 × p4 il j k cos f cos +δ δ 2 2   p1 × p2 p4 × p3 + . though to make this explicit we will a posteriori extract planar and nonplanar parts below.246 B.

including the masses of both σ and the π ’s. At one loop the sigma tadpole counterterm is fixed by the requirement that + + = 0. but since our explicit interest is in Goldstone’s theorem at the quantum level for the noncommutative case. Kaminsky / Nuclear Physics B 581 (2000) 240–256 247  = − δµ v + δλ v 3 . we will not do this calculation explicitly: for a demonstration using dimensional regularization. and the vanishing of the one-point amplitude or vacuum expectation value renormalization of the σ . In the renormalized perturbation theory sense. Of course. our results will not depend on this particular renormalization scheme. and in this scheme the masslessness of the pions is a prediction of the quantum field theory. the 4-σ scattering amplitude at threshold. we will employ it in the following. In the following we will examine the one-loop quantum structure of the noncommutative theory.  = − δµ + 3δλ v 2 − δz p2 . as inspection of (25) reveals.A. are predictions of the quantum field theory. One-loop quantum level calculations The linear sigma model contains three counterterm parameters. the reader is referred to [30]. . and yield finite predictions for everything else then the theory is perturbatively renormalizable. (25) 3. These are conventionally taken to be conditions specifying the field strength of σ . We will consider the general case later. and so three renormalization conditions are needed. Everything else.B. but completely fixed (at zero external momentum) by the renormalization condition on the sigma tadpole. Campbell. (27) 1 Actually we will directly use only the renormalization condition on the sigma one-point amplitude: namely that it vanishes. K. 1 In the ordinary commutative calculation of pion mass renormalization (see [30] for example) the masslessness of the pions at one-loop (Goldstone’s theorem) follows from a cancellation of the following graphs at zero external momentum (p = 0): + + + (26) Because the full noncommutative calculation we are about to explore subsumes this as the special case where θ → 0. We do point out that the counterterm displayed is not arbitrary. if the counterterms can be adjusted order by order to maintain the renormalization conditions.

momentum conservation dictates that no external momentum flows into the cubic vertices. K. Nonetheless we need to explicitly calculate it from (27).248 B. Next. (32) Λ + ξ 16π 2 ξ2 Λ where ξ is a small infrared mass for the pion. Kaminsky / Nuclear Physics B 581 (2000) 240–256 Now consider the noncommutative case. This is discussed and derived in the appendix.A.32]). for example. or equivalently the arguments of the cosine terms in the symmetrized vertices always vanish. α2 (29) 0 Thus high momentum divergences are replaced with small α divergences which are 2 regulated by inserting a factor of e−1/(αΛ ) under the α integral. The basic integral we will encounter repeatedly is then of the form: Z∞  m2 dα −αm2 − x 22 αm = e + m2 log x 2 + m2 (2γ − 1) + O(x) (30) 2 2 α x 0 in the limit of small x(= m/Λ). coupled with a C ∞ smooth UV momentum cutoff regularization throughout. we will henceforth use them. Since loop integrals containing phases like (18) are most easily evaluated using Schwinger parameters (see. fact is that noncommutativity does not alter the tadpole graphs at one-loop. The first. [31. where Λ is the (fundamental) UV momentum cutoff. Thus: Z d4 k 1 1 = (−6vλ) 4 2 2 (2π ) k + m2σ   2   3λv Λ mσ 2 2 2 . so that phase factor (18) always degenerates to 1. Z δ ij d4 k (2π 4 ) k 2 + ξ 2  2    (N − 1)λv 2 Λ ξ 2 2 =− − ξ log (2γ − 1) + O . to be taken to zero later. This follows from the elementary observation that while the one-loop tadpole graphs above contain cubic vertices (and so yield nontrivial noncommutative phase terms). Thus (27) fixes the tadpole counterterm to be: 1 = (−2vλδ ij ) 2 . perhaps surprising. (31) =− Λ − mσ log + mσ (2γ − 1) + O Λ 16π 2 m2σ where m2σ ≡ 2µ2 is the (tree-level) mass-squared for the sigma. Campbell. The basic Schwinger parameter representation is: Z∞ 1 2 2 = dα e−α(k +m ) (28) k 2 + m2 0 whence Z d4 k = k 2 + m2 Z Z Z∞ dΩ4 dα 3 −α(k 2 +m2 ) dk k e 0 Z∞ =π 2 dα −α 2 m2 e . Thus the sigma tadpole counterterm will not be modified in the noncommutative case.

(36) = whereas writing cos(k × p) = 12 (eik×p + eip×k ) and using the fact that the exponentials are the same under the integral over all momenta. K. Campbell. + (2γ − 1) 3m2σ + ξ 2 (N − 1) + O Λ 249 (33) As discussed this fixes the pion propagator counterterm at zero momentum (and at one loop).B. In order to consider separately the (noncommuting) [13] UV (Λ → ∞).A. (24) yields a phase factor of:      2 p×k + N − 1 + (1 − f ) 2 + (N − 1) cos(p × k) f 2 cos 2  = f N + 2(1 − f ) + f + (1 − f )(N − 1) cos(p × k). Denoting the loop momentum by k. Kaminsky / Nuclear Physics B 581 (2000) 240–256 = −(δµ v + δλ v 3 )  2  2  Λ Λ λv 2 2 2 = Λ (N + 2) − 3mσ log − ξ (N − 1) log 2 16π 2 m2σ ξ     1 . the nonplanar contribution is:  1 = −2λδ ij (f + (1 − f )(N − 1)) 2 Z λδ ij f + (1 − f )(N − 1)] 16π 2  2   2 × Λeff − ξ log Λ2eff /ξ 2 + ξ 2 (2γ − 1) + O(ξ/Λeff ) . we will not (2) for an arbitrary external set p = 0 a priori. d4 k eik×p (2π)4 k 2 + ξ 2 =− (37) . Thus. including the momentum-dependent wave function renormalization term we have:  2  2  λδ ij Λ Λ 2 2 2 = Λ (N + 2) − 3mσ log − ξ (N − 1) log 2 2 2 16π mσ ξ     2 1 + δ ij δz p2 . (35) The constant term is the planar contribution to the noncommutative analogue of the first graph in (26). and is explicitly: Z  1 d4 k 1 −2δ ij λ(f N + 2(1 − f )) 2 (2π)4 k 2 + ξ 2 λδ ij [f N + 2(1 − f )] =− 16π 2  2   × Λ − ξ 2 log Λ2 /ξ 2 + ξ 2 (2γ − 1) + O(ξ/Λ) . The first graph in (26) has a single quartic vertex (24). but will compute the ππ amplitude Γ1PI momentum p. (34) + (2γ − 1) 3mσ + ξ 2 (N − 1) + O Λ Now consider the one-loop contributions to the pion propagator. and IR (p → 0) limits.

so that its square is diagonal. (40) 0 where lµ = kµ − i/(2α)θµν pν is a linear change of variables with respect to the k integral. (43) The third diagram of (26) will yield momentum dependent corrections to the propagator. Note also that it does not depend on f . Note the (irrelevant) 1/4 factor in the definition of p ◦ q which does not appear in [13].A. Kaminsky / Nuclear Physics B 581 (2000) 240–256 where Λ2eff = 1/Λ2 1 +p◦p (38) and p◦q ≡− 2 q pµ θµν ν (39) 4 in a basis where θ µν is skew-symmetric (the coordinates form pairs of noncommuting coordinates).250 B. as the external momentum circulates in the loop. The phase factor from (22) is now: f + (1 − f ) cos(k × p). (41) Thus the planar contribution to the noncommutative analogue of the second graph in (26) is: Z 1 d4 k 1 = (−2λδ ij f ) 2 (2π)4 k 2 + m2σ λδ ij f  2 =− Λ − m2σ log(Λ2 /m2σ ) 16π 2 + m2σ (2γ − 1) + O(mσ /Λ)  (42) and the nonplanar contribution is: Z = −λδ ij (1 − f ) =− d4 k eik×p (2π)4 k 2 + m2σ  λδ ij (1 − f )  2 Λeff − m2σ log Λ2eff /m2σ 2 16π  + m2σ (2γ − 1) + O(mσ /Λeff ) . The second diagram in (26) is handled similarly. Campbell. This effective cutoff arises because completing the square in the Schwinger integral now yields an inhomogeneous term that is the same form as the fundamental UV cutoff regulator: eik×p 2 k + m2 Z∞ = dα e −α(k 2 +m2 )+ik×p Z∞ = 0 2 +m2 ) 1 e− 4α pρ p νθ µν θ µρ 0 Z∞ →reg dα e−α(l dα e −α(l 2 +m2 )− 1 αΛ2 eff . K.

since we have two cubic vertices. The former yields 1 = 4v 2 λ2 δ ij 2 Z Z1 dx " Z1 # = 2v λ δ 0 −v 2 λ2 δ ij 8π 2 1 1 d4 k 4 2 2 (2π) k + ξ (k + p)2 + m2σ d4 l 1 v 2 λ2 δ ij → reg (2π)4 (l 2 + ∆2 )2 8π 2 2 2 ij = Z  log ∆2 /Λ2 dx + 2γ + O(∆/Λ) . Now. (49) (47) = 16π 2 Λ2 p m2σ Noting that l = k + (1 − x)p implies that k × p = l × p. and writing as usual cos(k × p) = cos(l × p) in exponential form. the vertex (20) yields for the third diagram of (26) the phase:      −k × −p 1 k×p cos = 1 + cos(k × p) . we will combine the propagators using a Feynman parameter. (l 2 + ∆2 )2 (48) To proceed. e (45) α which makes explicit the logarithmic (as opposed to quadratic) divergence of this type of diagram. Kaminsky / Nuclear Physics B 581 (2000) 240–256 251 since there are no quartic vertices. (44) (l 2 + ∆2 )2 d(l ) 0 0 Then we will need (see appendix) the small x expansion: Z dα −α∆2 − x 22 α∆ = − log(x 2 ) − 2γ + O(x). Instead of using two Schwinger parameters for the two internal propagators. and then use a Schwinger parameter via the identity: " Z∞ # Z∞ d 1 2 2 2 2 =− 2 dα e−α(l +∆ ) = dα α e−α(l +∆ ) . Campbell. and where in the second line we introduced the Feynman parametrization: 1 = (k 2 + ξ 2 )[(k + p)2 + m2σ ] Z1 dx 0 1 . and noting the symmetry of the terms under integration over all l. (46) cos 2 2 2 Thus planar and nonplanar contributions are weighted equally.B. we can take the IR regulator ξ to zero without any difficulty. so that (recalling v 2 = µ2 /λ)  2   2    p + m2σ p + m2σ m2σ −λm2σ δ ij log + 2 log + 2(γ − 1) + O(∆/Λ) . the nonplanar contribution to the third graph in (26) is: . we pick up two phase factors due to noncommutativity. Denoting by k the pion momentum in the loop. K. l = k + (1 − x)p. to simplify the Feynman parameter integral. Z1 Z∞ dx 0 dα −α∆2 − 1 2 αΛ e α 0 (47) 0 where ∆2 = x(1 − x)p2 + xξ 2 + (1 − x)m2σ .A.

the result is unavoidably UV cutoff dependent (the freedom to choose δZ clearly does not help us because the divergences are not proportional to p2 ).252 B. Thus the mass shift of the pions due to oneloop quantum corrections vanishes. However. Discussion This result is rather remarkable. Campbell. (50). (50) Λeff p2 m2σ Λ2eff Thus the effect of noncommutativity is to re-weight the planar and nonplanar graphs with respect to the commutative graphs (where there is no distinction between planarity and nonplanarity) in the cases where nonplanar graphs are generated. the explicit dependence on Λ cancels. and we are left with the usual finite commutative result:  2   X λδ ij 2 p + m2σ p2 + m2σ = m 1− log (53) + δ ij δZ p2 . (47). This sum is the noncommutative equivalent of (26). + 2m2σ 1 − p2 m2σ or eliminating Λeff in favour of Λ. (42). and p:    X  2 λδ ij  1 = N(1 − f ) + f Λ 1 − 16π 2 1 + Λ2 (p ◦ p) 1-loop  − (2 − f )m2σ log 1 + Λ2 (p ◦ p) + 2m2σ  2   p + m2σ p2 + m2σ log 1− + δ ij δZ p2 . θ . The one-loop correction to the (inverse) propagator is the sum of the six graphs (36). and the counterterm (34). (37). for a strictly nonzero θ . First note that if we take the noncommutativity θ to zero. p2 m2σ (52) 4. and therefore naively . comm which furthermore vanishes in the p → 0 limit. (43). Explicitly it is equal to:  X    λδ ij  = N(1 − f ) + f Λ2 − Λ2eff + (2 − f )m2σ log Λ2eff /Λ2 2 16π 1-loop  2   p + m2σ p2 + m2σ log (51) + δ ij δZ p2 . 8π 2 σ p2 m2σ 1-loop. Kaminsky / Nuclear Physics B 581 (2000) 240–256 Z = 2v 2 λ2 δ ij Z Z1 = 2v λ δ 2 2 ij dx 0 = −λm2σ δ ij 16π 2  1 eik×p d4 k (2π)4 k 2 + ξ 2 (k + p)2 + m2σ d4 l eil×p 4 2 (2π) (l + ∆2 )2  2  2     p + m2σ p + m2σ m2σ ∆ log log + + 2(γ − 1) + O .A. K. and some comments are in order. and replace the Λ in the planar graphs with Λeff in the nonplanar graphs. this is just Goldstone’s theorem.

The Feynman rule for noncommutative phases Proceeding inductively. because no external momentum flows into a tadpole vertex. and D. spontaneous symmetry breaking is blind to the noncommutativity at this level. The conclusion is that not only do UV and IR limits not commute. we conclude that the Nambu–Goldstone realization of the O(N) symmetry of this model is not perturbatively compatible with the continuum renormalization of the noncommutative theory. Shaw for help with the Feynmf package. and hence the sigma tadpole renormalization counterterm is not modified. this means that the pion propagator counterterm is not modified (modulo wave function renormalization). K. On the other hand. and (52) becomes zero: again yielding Goldstone’s theorem in this ordering of the limits. so UV and IR limits do not commute. viewed as a Wilsonian action.A. Allahverdi for useful discussions. Acknowledgements This work was supported in part by the Natural Sciences and Engineering Research Council of Canada. since clearly if we somehow arrange the ππ amplitude to contain no net fundamental Λ dependence. or N . that this occurs after the renormalization programme has been carried out. K. noncommutativity does not affect tadpole graphs. for which Λ is a fixed. To summarize. This cannot be evaded by any choice of f ∈ [0.1. In sum. we have used up the counterterm freedom in fixing the sigma vev. Appendix A A. This is not an artifact of the renormalization prescription (of which we have only used the condition imposing nonrenormalization of the sigma vev). 1]. the graphs that contribute to the one-loop quantum corrections of the pion propagator are modified. Kaminsky / Nuclear Physics B 581 (2000) 240–256 253 divergent (both quadratically and logarithmically) in the continuum (Λ → ∞) limit. the limit p → 0 exists. This is a direct consequence of the restrictive counterterm structure present in a spontaneously broken theory. However. that cutoff dependence is restored.B. The interpretation and implications of this result will be explored elsewhere. would like to thank R. physical parameter. Since tadpole terms reflect shifts in vacuum expectation values of fields. However. since the counterterm structure is fixed. We emphasize. Campbell.K. postrenormalization Λ dependence will be shifted into the sigma tadpole. and since the sigma vev measures the spontaneous symmetry breaking of the theory. and the usual cancellation that ensures the pion masses are finite (and zero) after quantum corrections is violated in such a way. colloquially. but the continuum limit of the renormalized theory is inconsistent with Nambu–Goldstone realization of the global symmetry. we have:  (φ1 ∗ · · · ∗ φn ) ∗ φn+1 (x) .

pn )µ (pn+1 )ν i(p1+···+pn+1 )·x e 2 e (2π)2 i × e− 2 i Pn i<j pi ×pj φ˜ 1 (p1 ) · · · φ˜ n+1 (pn+1 ) Z Y 4 d pi i(p1 +···+pn+1 )·x − 2i Pn+1 i<j pi ×pj φ ˜ 1 (p1 ) · · · φ˜ n+1 (pn+1 ). and in the cases n = 1. We split the integration region into [0. Thus.A. Since we are really only interested in the small x (large Λ) behaviour. Kaminsky / Nuclear Physics B 581 (2000) 240–256 = Z Y 4  i µν y z d pi lim e 2 θ ∂µ ∂ν ei(p1 +···+pn )·y eipn+1 ·z 2 (2π) y.2. K. we encountered integrals of the form Z∞ In (x) ≡ dα −αm2 − x 22 αm .z→x i i × e− 2 = Pn i<j pi ×pj φ˜ 1 (p1 ) · · · φ˜ n+1 (pn+1 ) Z Y 4 d pi − i θ µν (p1 +..254 B. . (55) A. Campbell. and then expand the appropriate exponential on each region. ∞). i V (p1 . . after changing variables α → (1/α)(x 2 /m4 ) m4 I2 (x) = 2 x Z∞ e −αm2 − x2 αm2 dα 0 = m4 x2 2 " x/m Z  1 − αm2 +  2 α 2 m4 − x + O(α 3 ) e m2 α dα 2 0   Z∞  x2 2 1 − 2 + O(x 4 ) e−αm dα + m α x/m2   x 2 −x 2 3 −x 2 4 − x Ei (x) + O(x ) + e − x Ei (x) + O(x ) e 1 1 x2 2 2   m = 2 1 − x 2 + 2x 2 log(x) + 2x 2 γ + O(x 3 ) x =  m2 x− . A useful integral In evaluating the loop integrals in this paper. . e e = (2π)2 (54) i where the induction hypothesis is used in the second line. For n = 2. 2. . an elementary treatment suffices to determine dominant behaviour. x/m2]. e αn (56) 0 which can be taken as definitions (modulo constants) of the K type modified Bessel functions.. and [x/m2. pn ) = e− 2 P i<j pi ×pj .

A. S. E.W. H. Phys. Wulkenhaar. hep-th/9912238. hep-th/9912140. E. M. Belov. JHEP 06 (1999) 015.A. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] T. L. Phys. K. I. Chepelev. C. Phys. P. Campbell. J. J. Seiberg. Vazquez-Mozo. S. Paban. J. A 14 (1999) 1305. Sheikh-Jabbari. B 9 (1969) 649. Presnajder. hep-th/9903107.C. Susskind. S. Iso. Van Raamsdonk. hep-th/0002067. (60) It can be shown that C = −γ . P. Gorbatov. J. Chaichian. Aref’eva. Mod. Martin. hep-th/9910004. Varilly. 83 (1999) 476–479. hep-th/9908052. but since it must cancel out of all physical amplitudes. Benaoum. Wess. hep-th/9812180. H. H. J. Bigatti. Kawai. Zachos. Pouliot. Madore. M. J.S. R. S. JHEP 02 (2000) 020. Filk. hep-th/0002075. discovered through:  d  Ei1 (x) = − dx Z∞ e−xt dt = − e−x x 1 ∞ =− 1 X (−x)n−1 + . A. D. hep-th/9912167. Hinterding. hep-th/0001182. I. Kitazawa. Seiberg. Kashani-Poor. hep-th/9908056. x2 where we have used the exponential integral = Z∞ Ei1 (x) = e−xt dt t 255 (57) (58) 1 and its asymptotic expansion near x = 0. hep-th/0001203. M. Kitazawa. M. Phys. G. Krajewski. hep-th/9904132. hep-th/9912036. Kawai. Minwalla. hep-th/9911098. H. Cho. Nucl. R. P. Ishibashi. Hawkins.B. Iso. Lett. N. W. Steinacker. hep-th/0002186. Schupp. A. L. N. x n! (59) n=1 whence Ei1 (x) = − log(x) + C + O(x). hep-th/9903239. Fischler. A. C. Y. hep-th/0001027. D. we heed it no further. hep-th/9804001. N. Schraml. Roiban. . N. S. Rev. Grosse. Susskind. Koshelev. Van Raamsdonk.M. Y. hep-th/0001215. Krajewski. Matusis. M. Gomis. Wulkenhaar. D. J.M. Madore. Demichev. Lee. B.P. H. Kaminsky / Nuclear Physics B 581 (2000) 240–256 m2 + m2 log(x 2 ) + m2 (2γ − 1) + O(x). Arcioni. R. The n = 1 case is handled identically. T. S. Gracia-Bondia. hep-th/9912072. hep-th/9903077. JHEP 0001 (2000) 028. hep-th/9912094. hep-th/9903187. Int. M. and we omit it here. B 376 (1996) 53.Y. Sanchez-Ruiz. R. Lett.A. Hayakawa. Toumbas. T.

Phys. Soc. Quantum Field Theory and Critical Phenomena.A. Campbell. 1995. Commun. 1980. An Introduction to Quantum Field Theory.-L. Kaminsky / Nuclear Physics B 581 (2000) 240–256 J. Symanzik. Math. B. K. Zinn-Justin. 45 (1949) 99. Cambridge Philos.W. Phys B 12 (1969) 627. Nuovo Cimento 1 (1969) 10. K.E. Physik 46 (1927) 1.256 [25] [26] [27] [28] [29] [30] [31] [32] B. M. Lett. Lee. C. Addison-Wesley. Z. McGraw-Hill. Itzykson. Weyl. J. Moyal. Symanzik. Zuber. J. Oxford University Press. Gervais. D. Proc. . Schroeder. 16 (1970) 48. H. Peskin. J. K. 1996. Nucl. Quantum Field Theory.

For example. which get a mass via Yukawa couplings to a scalar which gets a vev.  2000 Elsevier Science B. 9500 Gilman Drive. NJ 08855-0849. 0) field theories Kenneth Intriligator UCSD Physics Department. necessarily includes a new kind of Wess–Zumino term. in the context of a single M5 brane probe in the background of N distant M5 branes.3 2 1 3 ( 0 0 ) 0 0 1 4 8 . a Wess–Zumino term. generated by integrating out massive fields.V. All rights reserved. 2000. All rights reserved. so it plays a role in the low energy theory even when the masses are taken to be infinite (via large Yukawas or scalar vevs). the probe must have the Hopf–WZ term with coefficient proportional to N(N + 1). USA 1 Received 31 January 2000.g. with a particular coefficient.elsevier. which do not decouple even when the masses of these fields is taken to be infinite.. La Jolla. can be needed to reproduce the contribution to the ’t Hooft anomalies of integrated out fermions. when away from the origin of the moduli space of vacua.6 . Winter. Various related checks and observations are made. Piscataway.Nuclear Physics B 581 (2000) 257–273 www. Rutgers University. and propose that they are the W -boson strings.V. A related effect is the Wess–Zumino term associated with integrating out massive fields. USA Department of Physics. 1. accepted 6 March 2000 Abstract We point out that the low energy theory of 6d N = (2. For example.nl/locate/npe Anomaly matching and a Hopf–Wess–Zumino term in six-dimensional. The size of the WZ term cannot depend on any parameters (e. The GW current does not depend on the masses of the integrated out fermions (as long as they are nonzero). PII: S 0 5 5 0 . CA 92093. N = (2. which is generated by integrating out fermions which get a mass via Yukawa coupling to a scalar which gets an expectation value. which is often needed in the low energy theory on symmetry grounds. Yukawa or gauge 1 Visiting address. We also point out that there are skyrmionic strings. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. 0) field theories. The classic example is the Goldstone–Wilczek current [1]. The form of this term is related to the Hopf invariant associated with π7 (S 4 ). The coefficient of the Wess– Zumino term is fixed by an anomaly matching relation for a global flavor symmetry. Introduction Low energy effective field theories can have effects.

despite [7]) two-form gauge fields. P a a and the string with charges α i has tension |α i Φi | (here |Φi | ≡ a Φi Φi . . (And G = U (1) for the free N = (2. or E6. which is hoped to be in the same universality class. with separations Φia in the 11d bulk. these strings come from M2 branes which stretch between the M5 branes. remains mysterious. the massless spectrum is that of r(G) free. with an SO(5)R flavor symmetry. Much about these theories. non-supersymmetric. 0) tensor multiplets. QCD [12]. or the RG scale: since its coefficient must be quantized [2]. the W -bosons of G come from these strings wrapped on S 1 [8]. where r(G) = rank(G) and WG is the Weyl group of G. which can be made arbitrarily large by taking the scalar expectation values Φia to be huge. in the highest possible dimension [10. there is an ordinary gauge symmetry with gauge group G. r. .. pure glue. 0) field theories. . there is a correspondence with nonAbelian groups G. any effects associated with degrees of freedom which were massless in the interacting theory. Naively. The 6d N = (2. 0) theories are interesting. In the realization via M5 branes. On the other hand. String theory indicates that G can be an arbitrary ADE group: SU(N). M theory on AdS7 × S 4 [9]. The existence of these field theories.) Upon compactification to lower dimensions. with real scalar expectation value coordinates given by Φia . . . there is a moduli space of supersymmetric vacua M = (R5 )r(G) /WG . In 6d. Although the gauge fields are two-forms rather than one-forms. These charge vectorspspan the root lattice of G. SO(2N). IIB string theory on a ALE singularity [8].). has come from string theory. including how to properly formulate them as field theories. They have the exotic property of having. Z) electric–magnetic duality manifest. rather than ordinary gauge fields. . . more generally. where they occur in various related contexts: the IR limit of the M5 or IIA NS-5 brane world-volume theory. 0) tensor multiplet. where a = 1. They are the maximally supersymmetric conformal field theories. The 6d N = (2. For example. as field theories. it cannot be renormalized. etc. Intriligator / Nuclear Physics B 581 (2000) 257–273 couplings. See. vevs.8 . the length of the SO(5) vector). there are effects associated with the interacting theory at the origin which cannot decouple from . in their own right. e. compactifying on a T 2 gives 4d N = 4 theories and makes SL(2. . etc. both because of these connections to string theory and duality and. but become massive for the generic vacuum in the bulk of M.7. with self-dual three-form field strengths.258 K. . However. as the geometric symmetry of the complex structure of T 2 .11]. . for the generic vacuum in the bulk of M. where M is singular. which couple to the r two-form gauge fields with charges α i . r(G). One such degree of freedom are BPS strings. 0) theories are chiral. Upon S 1 compactification. as well as all of their known properties. at the boundaries of M.g. would decouple at energy scales much less than their mass. . 5 is an index in the 5 of SO(5)R and i = 1. . The theory is interacting at the origin and. interacting (somehow!. i = 1. This paper will be concerned with flavor anomaly matching and Wess–Zumino terms in the 6d N = (2. 6d N = (2. and worthy of further study. ordinary. but more tractable than. [1–6] and references cited therein. Instead compactifying on a T 2 with supersymmetry breaking boundary conditions leads to the theory known as MQCD. no matter how far the vacuum is from the origin of the moduli space. and other interesting theories can be obtained by compactification and RG flow. which can be made arbitrarily large by going to large vevs in M.

In Section 4.K. perhaps the story of this section is a fantasy. In this case. with the 5d the uncompactified directions and the strings coming from M5 branes wrapped on a 4-cycle of the Calabi–Yau. 0) field theory has been obtained from string theory. Intriligator / Nuclear Physics B 581 (2000) 257–273 259 the low energy theory. The generalization for the G = SO(2N) and G = E6. this would give a derivation of the anomaly of the interacting theory at the origin. everything which is presently known about the interacting. In particular. we discuss how the needed WZ term of the N = (2. the coefficient of this term is the difference between the anomaly c(G) of the interacting theory at the origin and that of the low energy theory away from the origin. which probes N distant M5 branes. the WZ term can be derived by a standard [1. Skyrmionic strings associated with π4 (S 4 ) are also discussed.. consequently. directly in the context of some sort of quantum field theory. and the WZ term thus required in the low-energy theory when away from the origin. This argument shows that the anomaly for the N = (2.) In particular.8 cases has not yet appeared in the literature.3] for integrating out some massive degrees of freedom. This anomaly differs from that of the r tensor multiplets comprising the massless spectrum away from the origin. the result of [13] could be re-derived and checked directly in the context of field . we discuss anomaly matching and a Hopf–WZ term in the 2d N = (0. As will be discussed. 0) theory indeed arises in the world-volume of a M5 brane. it would be possible to derive directly the Wess–Zumino term.g. In Section 3. X4 ) = b4 )2 .7. As mentioned above. (A hope is that it will eventually be understood how to properly formulate these theories. finding the anomaly (in the large N limit) c(N. 16]. The reason is that the interacting theory at the origin generally has a nontrivial ’t Hooft anomaly associated with the global SO(5)R symmetry. in the context of field theory. The interesting anomaly coefficient for the G = SU(N) case was found [13] to be c(SU(N)) = N 3 − N . 0) theory associated N 3 /vol(X with G = SO(2N) is c(SO(2N)) = 4N 3 + terms lower order in large N . In Section 5. we review how the N 3 dependence of the entropy and Weyl anomaly.3] 1-loop calculation [14]. the nontrivial SO(5)R ’t Hooft anomaly mentioned above was found in [13] in the context of 11d M-theory. which is related by supersymmetry to the SO(5)R anomaly c(SU(N)). 4) CFT which arises in the world-volume of strings in 5d. Turning around our anomaly matching discussion. We generalize this argument to M-theory on AdS7 × X4 for general Einstein space X4 . since there is no moduli space in 2d. and it is proposed that they are the W -boson strings. we briefly discuss 4d N = 4 theories. ’t Hooft anomaly matching. and recover the properties predicted via string theory. realized as N parallel M5 branes. We also review some math facts concerning the Hopf invariant and map. In the next section. In Section 6. Nontrivial maps in π7 (S 4 ) imply a nontrivial quantization condition on the WZ term (which is related to the Hopf invariant of the map) and. by some analog of the 1-loop computation of [1. One might expect that. This occurs via M-theory on a Calabi–Yau three-fold. on the anomaly: c(G) ∈ 6Z. 6d N = (2. was originally found [15. via M-theory on AdS7 × S 4 . We will argue that a Wess–Zumino term must be present in the low energy theory to account for what would otherwise be a deficit in the ’t Hooft anomaly. e. which gave the anomaly for the case G = SU(N). we discuss anomaly matching and the Hopf–Wess–Zumino term which it requires.

2. which gives the anomaly by the descent formalism: I8 = (0) (0) (1) (1) dI7 . The argument for ’t Hooft anomaly matching is same as the original argument of ’t Hooft in 4d [19]: we could imagine adding spectator 2 fields..g. the N 3 behavior of c(G = SU(N)) was first discovered in these two ways. E.2) p1 (F ) = 2 2π 8 2π (Writing the Chern roots of F /2π as λ1 and λ2 . 0) theories: c(G) = |G|C2 (G). δI7 = dI6 . also depends on G. without having to invoke M-theory. compactifying the 6d theory and flowing in the IR to 4d N = 4. 0) supermultiplet [17. and thus the ’t Hooft anomaly c(G) also enters in a term in the Weyl anomaly.1). 0) theory at finite temperature is also proportional to c(G). and in a general gravitational background: I8 (G) = r(G)I8 (1) + c(G)p2 (F )/24. Indeed. 0) theory associated with arbitrary group G is expected to have an anomaly of the following general form when coupled to a background SO(5)R gauge field 1-form A.1) gives the anomaly at the origin.1) must be reproduced everywhere on the moduli space of vacua. which remain 2 In the context of M5 branes. p1 = λ21 + λ22 and p2 = λ21 λ22 . Viewing c(G) as a c-function. D. 0) effective field theory The N = (2.260 K. The SO(5)R current is in the same supermultiplet as the stress tensor.18]: I8 (1) = (p2 (F ) − p2 (R) + 14 (p1 (R) − p1 (F ))2 )/48. E type N = (2. The anomaly (2. (2. which we refer to as the ’t Hooft anomaly of the SO(5)R flavor symmetry. Nevertheless. appear only in I8 (1). where the SO(5)R global symmetry is unbroken. the role of these “spectators” is played by contributions from the 11d bulk: the anomaly inflow and the Chern–Simons term contributions of [13. associated with any nontrivial curvature R. Intriligator / Nuclear Physics B 581 (2000) 257–273 theory.16]. (2. The gravitational anomalies. for the case G = SU(N). SO(5)R is spontaneously broken. In Section 7. p2 (F ) = tr F 2 ∧ tr F 2 − 2 tr F 4 . with the result that c(SU(N)) = N 3 − N . we speculate on deriving the WZ term via integrating out tensionful strings and on a possible formula for the ’t Hooft anomaly for general G = A. r(G) is the rank of the group G associated with the N = (2. The analog for for other G has not yet appeared in the literature. Six-dimensional N = (2. . A hope is that these issues could lead to a better understanding of the interacting 6d N = (2.        1 i 2 1 i 4 2 tr F . this suggests that in all cases cUV = c(G) > cIR = |G| ≡ dim(G). 0) field theory. in the context of N M5 branes in 11d SUGRA [15. N = (2.20]. In (2.1) where pi are the Pontryagin classes for the background SO(5)R field strength F . I8 (1) is the anomaly polynomial for a single.) I8 is the anomaly polynomial 8-form. we argue that the ’t Hooft anomaly of (2. with I6 the anomalous variation of the Lagrangian under a gauge variation δ. Away from the origin. it should decrease in RG flows to the IR. 0) theory and the quantity c(G). free. The result (2.2) was found via M-theory in [13]. The entropy of the N = (2.

Ω3 (φ. a = 1. . the global SO(5)R symmetry is broken to SO(4)R and the configuration space. so we can always choose this to be the case by our choice of boundary conditions at infinity. b ∈ 5 of SO(5)R (Fijab is its field strength). 0) multiplet associated with the U (1) is free. 0) and (0. Massless 8 fermions of chiralities ( 21 . it is a modulus labeling superselection sectors.. 4 Although there is no potential which requires hφ a i 6= 0. The needed Wess–Zumino term is given by the following term in the action Z  1 ˆ A) ∧ dΩ3 (φ. ˆ A) + · · · . 0) multiplet associated with the U (1). Consider the 4-form be AdS7 ). (2.  1  1 ˆ A ≡ e4Σ ≡ ˆ a1 (Di2 φ) ˆ a2 (Di3 φ) ˆ a3 (Di4 φ) ˆ a4 a1 ···a5 (Di1 φ) η4 φ. Away from the origin. conjugate group representations contribute to the anomaly polynomial I with the same sign.4) SWZ = c(G) − c(H ) 6 Σ7 where (· · ·) are terms related by supersymmetry. I8 (G) − I8 H × U (1) = 24 For hφ a i 6= 0. The massless spectrum for hφ a i 6= 0 is that of the N = (2. The Ward identities of the symmetry must then always be satisfied. . ∂Σ7 = W6 (e. 0) field theory.5) b a with (Di φ)a ≡ ∂i φ a − Aab i φ the covariant derivative of φ . along with a single additional N = (2. For simplicity. is needed to ensure that this is the case.K. Naively. (2. The φ a are the scalars in the N = (2. Ω3 (φ. we consider the case that the scalar vacuum expectation values are chosen to be Φia = φ a (T )ii . The 3 In 6d. Σ7 could ˆ A) is a 3-form which is defined as follows. to cancel 3 the anomalies. with a. .1) between the G theory at the origin and the massless H × U (1) N = (2. allowing the global symmetry to be weakly gauged. the H and U (1) theories are decoupled and the N = (2. whose boundary is the 6d spacetime W6 of the N = (2. . is Mc = SO(5)/SO(4) = S 4 . 0) supermultiplet √ associated with this U (1). 0) theory for |φ| 6= 0:   1 (2.g. involving the background ab ba SO(5)R gauge field Ai = −Ai . The requirement that |φ| be fixed is not essential: we only need Mc ∼ = S 4 topologically. with specific coefficient. the U (1) multiplet of the theory on the Coulomb branch is actually never really free: it must always include a WZ interaction term. Σ7 is a 7-dimensional space. for energy E  |φ| ≡ (φ a φ a )1/4 . 0) CFT associated with H . 5. for fixed non-zero 4 |φ|. Thus. including the scalar expectation values. subtracting the constant contribution of the spectators. 2 2 64π a a ˆ a3 (Di4 φ) ˆ a4 − 2Fi11i2 2 (Di3 φ) 1 a a  + Fia11i2a2 Fi33i4 4 φˆ a5 dx i ∧ · · · ∧ dx i4 . with coordinates φˆ a ≡ φ a /|φ|. where (T )ii are the diagonal components of a generator of the Cartan of G whose little group is H × U (1) ⊂ G. However. . a Wess–Zumino term. and requiring hφ a i 6= 0 is enough. the anomalous Ward identities of the original theory must be independent of any deformations.3) c(G) − c(H ) p2 (F ). Intriligator / Nuclear Physics B 581 (2000) 257–273 261 decoupled from the rest of the dynamics. 12 ) under the SO(4) ∼ = SU(2) × SU(2) little group contribute with opposite signs. The WZ term is needed to compensate for what would otherwise be a difference in the ’t Hooft anomaly (2.

R). as Ω3 ∧ dΩ3 6= φˆ ∗ (ω7 ).7). 5 Note that to construct the WZ term requires extending φˆ : W → S 4 to φˆ : Σ → S 4 .R By a general analysis 6 [22.4) is not exact. which sets it apart from the general analysis of [22. even for A = 0. (2. the pullback of the S unit4 volume form. The form (2. This defines the Ω3 appearing in (2. δp2(0) = dp2(1) . ˆ A .4) is not of this form.5) is closed and. 6 I thank E. for zero background SO(5)R field.4) has the desired nontrivial gauge variation under SO(5)R gauge transformations. when in the background of N other M5 branes. so the ambiguity in the choice of Σ7 is nontrivial.g..23] WZ terms are generally of the form Σd+1 φˆ ∗ (ωd+1 ).23]. The η4 in (2. D’Hoker for pointing this out to me and for related correspondences. the dependence in Ω3 (φ. of the global.9) 6 S7 This can be nontrivial. (2.9) gives the Hopf number of the map φˆ a : S 7 → S 4 .262 K. An aspect of the present case. Note that the φ ˆ A) has dropped out in the gauge variation (2. The 7-form Ω3 ∧ dΩ3 in (2.4) is Z   1 ˆ A ∧ dΩ3 φ.4) indeed compensates for the deficit (2.6) implies that Z Z Z 1 1 δp2(0) (A) = p2(1) (A). c(G) − c(H ) Ω3 φ. Indeed. Using (2. SO(5)R gauge variation of the WZ term (2. which can have an 7 6 obstruction if the original φˆ is in the nontrivial component of π6 (S 4 ) = Z2 . both with boundary W6 . Ω3 φ. because we take Σ7 such that H (Σ7 ) is trivial. Euler class 4-form e4 which also entered in [20]. As an example.6) dΩ3 ∧ dΩ3 ≡ e4Σ ∧ e4Σ = p2 (F ) + dχ.4).3). To see this. A = 0) = φ (ω4 ). and obviously H 7 (S 4 . ˆ A) = dΩ3 (φ. which can be an arbitrary integer. angular. 4 4 where χ is invariant under SO(5)R gauge transformations. The difference between choosing Σ7 and Σ70 . the Wess–Zumino term (2. Indeed. Using the result of [13] that c(G) = (N + 1)3 − (N + 1) and c(H ) = N 3 − N . with ωd+1 ∈ H d+1(Mc . η4 (φ. S4 ω4 = 1. and thus for a M5 brane in AdS7 × S 4 . (2. consider G = SU(N + 1) and H = SU(N). is the integral over Σ7 − Σ70 ∼ = S7 Z    1 ˆ A ∧ dΩ3 φ. ˆ A . the SO(5)R gauge variation of (2. our WZ term (2.7) δ Ω3 ∧ dΩ3 = 4 4 Σ7 Σ7 W6 where p2(1) (A) is the anomaly 6-form found by descent. However. it must be exact. e4Σ ≡ φˆ ∗ (e4 ) is the pullback to Σ7 .7) in [21] implies that 1 1 (2. A = 0. In (2.4) is non-zero.5) 4 ˆ ˆ∗ is R normalized so that η4 (φ. Intriligator / Nuclear Physics B 581 (2000) 257–273 x i are the coordinates on Σ7 . here Mc = SO(5)/SO(4) ∼ = S 4 .5). Writing the left hand side as (0) d(Ω3 ∧ dΩ3 ) and p2 (F ) = dp2 (A). R) = 0.8) 2 N(N + 1) Σ7 This Wess–Zumino term must be present in the world-volume of a M5 brane. e. via 5 φˆ : Σ7 → S 4 . is mentioned at the end of this section. The Wess–Zumino term (2.7). the integral in (2. ˆ A). (2. . we note that Eq. (2. nevertheless.

defined as in (2. d ∗ H3 = Jelec = αe η4 (2. a p-brane skyrmion is a field configuration φˆ a (Xt ) which only depends on the d − p − 1 space coordinates of Xt . π4 (S 4 ) = Z means that there are nontrivial p = 1 branes in d = 6. Here Q ∼ Ns . W6 W6 Given that the skyrmionic string is charged under H3 as outlined above.2] to the SO(5)R flavor current. We also note that there are topologically stable. The WZ term means that there is a Goldstone–Wilczek contribution [1.12) means there is an interaction Z Z  ˆ A . so αe and αm need not be equal.e. which can give the skyrmions SO(5)R charges. For each Ns charge.9) thus must be quantized in order for e2πiS to be well-defined and invariant under the choice of Σ7 :  1 c(G) − c(H ) ∈ Z. The argument of [24] for the magnetic relation in (2.e . including [24.12) and the coupling (2. Happily.5): the string number is Ns = Xt η4 ..11) is indeed satisfied by c(G = SU(N)) = N 3 − N . 0) theories. In d spacetime dimensions. φˆ a must approach a constant value when the coordinates of Xt are taken to infinity.11) for all ADE groups G.) The topological chargeR density for the skyrmionic strings is η4 .13) have already appeared in [24]. with Ns R the π4 (S 4 ) topological string number.K.10) 6 To have (2. corresponding to the G = SU(N) N = (2.12) involves . Ns = Xt η. have briefly considered solitonic strings in the M5 brane theory. (There are also Z2 particles since π5 (S 4 ) = Z2 .12) for some non-zero constants αm. there is a BPS field configuration φˆ a which minimizes the energy. In the particular context of N M5 branes. The H3 in (2. and the tension of such a string satisfies T > |Qφ|. It is these BPS skyrmionic string solitons which should be identified with the BPS W -boson strings. there are skyrmionic strings. 0) multiplet: dH3 = Jmag = αm η4 . (Other works. (2. In order for this to be a finite-energy configuration. though the interpretation of these relations in terms of the π4 (S 4 ) skyrmionic strings was not discussed there. it follows from completely general considerations that the supersymmetry algebra has central term Z = |Qφ|.13) Ssky = αe B2 ∧ η4 = −αe dB2 ∧ Ω3 φ. the space transverse to the p-brane world-volume. satisfying T = |Qφ|. the skyrmionic string density η4 should act as electric and magnetic flux sources for the H3 in the U (1) N = (2. rather it is related to a self-dual tensor h3 by a non-linear transformation [26]. but not specifically the π4 (S 4 ) skyrmionic solitons. solitonic “skyrmion” field configurations in the theory with non-zero |φ|.25]. i. Such field configurations are thus topologically classified by πd−p−1 (Mc ). We propose that these skyrmionic strings are actually the “W -boson” BPS strings mentioned in the introduction. the magnetic relation in (2.10) hold for arbitrary ADE groups G and subgroups H requires 1 c(G) ∈ Z 6 (2. The coefficient of (2. (2. The electric relation in (2.) In line with this proposal. Intriligator / Nuclear Physics B 581 (2000) 257–273 263 corresponding to π7 (S 4 ) = Z + Z12 . (2.12) is not self-dual. In the present case.

and thus also Ω3 . which immediately generalizes to general N = (2. again with ∂Σ7 = W6 . as inherited from 6d where H3 is not simply self-dual. (2.12) for H3 as  ˆ A . the term (2.15) gives the coupling (2.14) H3 = dB2 + αm Ω3 φ. However. Solving (2. and d ∗ F2 = αe η4 (now ∗ acts in the uncompactified 5d). and that the π4 (Mc = S 4 ) skyrmions are electrically charged particles in 5d which. 0) G → H × U (1) Coulomb branch Higgsing.13) with αe = 14 nW . this gives the action Z αe (dB2 + αm Ω3 ) ∧ dΩ3 .5) shows that η4 .8) the coefficient is proportional to N(N + 1) rather than just nW ∼ N .12) is more difficult to determine. In the M5 .27]): consider S 1 reducing to 5d. using (2.5). we get (2. It could have been anticipated that the 6d WZ term (2. using H3 → H3 + F2 ∧ dx6 and η4 → η4 . The 5d U (1) gauge field in the Higgsing G → H × U (1) arises from Bµ6 in 6d.4): in direct analogy with the discussion in [2]. The coupling (2. this is also related to the analysis and results of [13. the term (2.. E. since αe = 14 nW . (2. The argument of [24] for the coupling (2. Naively. We emphasize that.20]. We can also consider S 1 -dimensional reduction of the relations (2. In this light.3]. Note that. one might identify H3 as ∗F2 . in (2.4) would be difficult to obtain by dimensional reduction because (unlike (2. in Σ7 .4) actually vanishes (unless the Kaluza–Klein S 1 momentum modes are included) since. it does not seem possible (at least in any obvious way) to get our 6d Hopf–Wess–Zumino term (2. dF2 = 0. upon S 1 -dimensional reduction. the 5d H3 and ∗F2 are not simply equal. Here is a possible insight into the origin of the 6d WZ term (2.g.15)) provided that 16 (c(G) − c(H )) = αe αm = 14 nW αm .4) by a direct calculation in the dimensionally reduced 5d gauge theory. (2.13). The constant αm in (2. have no dx6 component. is as follows (see also [14. with the right coefficient (assuming that the WZ term indeed arises entirely from (2. and is further discussed in Section 5.15) Σ7 The first term in (2. making the H3 equations repeats of the F2 equations and suggesting that αm be identified with αe .4). whose ends on ∂Σ7 = W6 are the skyrmionic strings of W6 discussed above. This again makes αm more difficult to determine. η4 becomes a density for skyrmionic membranes.12) gives dH3 = αm η4 . the action of an electric current in a magnetic background is R Σ7 H3 ∧ Jelec . where the theory is ordinary Yang–Mills and is IR free. The F2 equations show that there are no magnetic charges in 5d.g. where now.13)) its coefficient is not simply nW . with η4 = dΩ3 . Indeed.264 K. can be identified with the nW electrically charge W -bosons. Intriligator / Nuclear Physics B 581 (2000) 257–273 accounting for the fact that M5-branes act as G4 sources.13) then arises by a standard type of 1-loop calculation.12). d ∗ H3 = 0.13) is generated with coefficient 12 N . η4 has no dx6 component because we take φˆ to be independent of x6 ∈ S 1 in dimensional reduction.10) is simply Dirac quantization. with the nW = |G| − |H | − 1 massive gauginos running in the loop. and plugging in Jelec from (2. (2. much as in [1.13).13) and the second term gives the WZ term (2. Taking care with the normalization. Then (2. for G = SU(N + 1) and H = SU(N). for φˆ independent of x6 . e.12).. unlike the term (2.

1. only defines Ω3 up to exact forms. for energies E  |φ|. The remaining Alternatively.g. −1 H . (2.15).) If the WZ term arises from (2.16). one might be tempted to forget about the effects of the nW = |G| − |H | − 1. 7 Noting that the physical quantity H3 must also be well-defined. with dΛ2 |W6 = 0. via η4 = dΩ3 . we can simply use (2.16) SWZ → SWZ − c(G) − c(H ) 6 W6 this freedom must somehow be fixed in order for the effective action to be well-defined.. an aspect which sets it apart from the general analysis of [22. Under such a change. Here a ∈ 6 of SU(4)R and taking hφ a i 6= 0 breaks the gauge symmetry G → H × U (1) and the flavor √ symmetry SU(4) → SO(5). the nW gauginos in these multiplets contributed to the anomaly (3. our WZ term (2.e. Consider now moving away from the origin of the moduli space via Φ a = φ a T . Ω3 → Ω3 + dΛ2 . (In the M5-brane realization. In short.4) changes by Z  1 dΛ2 ∧ Ω3 . without 7 I thank E.1). in Σ7 = AdS7 . with little group H × U (1) ⊂ G. (2. Miscellaneous notes 3.. to be discussed in Section 5. the configuration space is Mc = SU(4)/SO(5) ∼ = S 5 .1) I6 (G) = 6 2π with |G| the dimension of the gauge group G. with ’t Hooft anomaly tr SU(4)3R = |G|. The WZ term of 4d N = 4 theories A completely analogous relation between ’t Hooft anomaly matching and a WZ term holds in the 4d N = 4 theory.14) shows that the change Ω3 → Ω3 + dΛ2 requires a compensating shift B2 → B2 − αm Λ2 . with the change in (2. Intriligator / Nuclear Physics B 581 (2000) 257–273 265 brane realization. is harmless in (2. (3.K.13) under B2 → B2 − αm Λ2 cancelling (2. (2. there is an anomaly determined via descent from   |G| i 3 tr FB3 . with T a generator of the Cartan of G. This anomaly comes from the |G| gauginos in the 4 of SU(4)R and is not renormalized. we tie up a loose end: our definition of Ω3 . Lastly. Witten for correspondences.4) needs B2 to be well-defined. G/H × U (1) gauge field multiplets. it is unchanged by this combined shift of Ω3 and B2 .16). these are like skyrmionic M2 branes living. The N = 4 theory has a global SU(4)R ∼ = SO(6)R flavor symmetry. with field strength FB . i. e. which stressed the need to fix this issue and suggested the following discussion.23]. which measures membrane winding number in Σ7 . However. The WZ R term is proportional to Σ7 Ω3 ∧ dΩ3 . 3.14) to define Ω3 on W6 as αm 3 ambiguity on Σ7 of taking Ω3 → Ω3 + dΛ2 . . this is the freedom of C3 gauge transformations. in a background SU(4)R gauge field AB . The massless spectrum is that of the N = 4 theory with decoupled groups H × U (1) and. For fixed |φ| ≡ φ a φ a . ultra-massive.

) Corresponding to (3. in the 4 of SU(4).g. the dimensional reduction of the 6d WZ term vanishes. The quantization condition is thus 1 2 nW ≡ 12 (|G| − |H | − 1) ∈ Z. there is an induced Goldstone–Wilczek current 1 1 µνρσ  a1 a2 ···a6 ∂ν φˆ a3 ∂ρ φˆ a4 ∂σ φˆ a5 φˆ a6 . The 4d WZ term (3. with e5 the S 5 global. with subgroup H × U (1) obtained via adjoint Higgsing.. [14].2.13).2) in order to have e2πiS be invariant under Σ5 → Σ50 with ∂Σ50 = ∂Σ5 = W4 . 3. The WZ term thus required in the low-energy theory is   1 ˆ (3. (3. now there are no p-brane skyrmions. The fact that integrating out the nW massive fermions actually does generate precisely the WZ term (3.5) Fortunately. normalized so that 12 e5 (A = 0) = ω5 . As shown in [4]. The difference involves the 5-form of (3. in particular. Intriligator / Nuclear Physics B 581 (2000) 257–273 i 3 them there is a deficit in (3. where Γ [φ.2). in [28]. as in both cases there is a SU(4) flavor symmetry with order parameter in the 6 of ˆ AB ] contributes to the constant magnitude. See. which got a mass via Yukawa couplings to φ in the Higgsing G → H × U (1).4). not the 6d WZ term (2. which is an arbitrary integer associated with π5 (Mc = S ) = Z. on its coefficient (3. Some math notes on the Hopf invariant We now summarize some facts which can be found.2) 2 nW Γ φ. Γ [φ. there is no obstruction to extending φˆ : W4 → S 5 to φˆ : Σ5 → S 5 . the SU(4) variation of Γ [φ. Z   1 a1 . there is a quantization condition Because the WZ 5-form term in Γ [φ. (3.4) jaµ1 a2 = nW 2 24π 2 ˆ AB ] appeared in [4] in the context of N = 1 SUSY QCD with Nf = The same Γ [φ.1) of I6 (G) − I6 (H ) − I6 (U (1)) = 16 nW ( 2π ) tr FB3 . this is indeed satisfied for arbitrary group G.3) integrated over Σ50 − Σ5 ∼ = 5 5 S . This deficit must be accounted for by a Wess–Zumino term in the low energy theory.3) Γ φˆ = 240π 2 Σ5 ˆ A] = where ∂Σ5 = W4 . it properly accounts for the contribution to the ’t Hooft anomaly of the nW gauginos. R 1 2 Σ5 φˆ ∗ (e5 ). (3.266 K. Since all π3−p (Mc = S 5 ) = 0.. e..2) follows from the standard 1-loop calculation of the type appearing in [1..3]. The Hopf invariant H (f ) of a mapping f : S 2n−1 → S n is an integer which can be defined as the winding coefficient of curves f ∗ (a) and f ∗ (b) in S 2n−1 for distinct a and b in S n .a6 ∂i1 φˆ a1 · · · ∂i5 φˆ a5 φˆ a6 dx i1 ∧ · · · ∧ dx i5 . f ∗ (a) is the . Euler class form in the appendix of [13]. angular. Again. ˆ AB ] = Γ [φ] ˆ + Z[φ. ˆ AB ]/48π 2 with ˆ AB ] is conventionally [2] written as Γ [φ. the unit S 5 volume form.2) is related to the dimensional reduction of (2.2). SU(4)3 flavor ’t Hooft anomalies the same as with two fermions in the 4 of SU(4). AB + superpartners. Thus. (Since π4 (S 5 ) = 0. with the coefficient of the WZ term as in (3.. ˆ AB ] is not exact. Nc = 2.

with lP the 11d Planck length. The basic map S 3 → S 2 with Hopf number 1 is given by writing S 3 as (z1 . c = vol X lP9 (4. (3. which is the horizon size of AdS7 (related to the size of the negative cosmological constant). and the powers of L. z2 ). It thus follows that L5 . This term is related by supersymmetry to the coefficient of the 7d Einstein–Hilbert action in AdS7 . H (f ) ∈ Z. z2 ]. z2 ) → [z1 .2) leads to the general result c= N3 + lower order in N.. H (f ) = 0 for n odd. [z1 .1) c= G7 where G7 is the 7d Newtons constant. where X4 is a general. λz2 ] for arbitrary λ ∈ C∗ .K. compact. Getting the N 3 via gravity and the G = SO(2N ) case via an orbifold We now review how c ≈ N 3 appears via 11d SUGRA. with the simple replacement that zi and λ now take values in the quaternionic rather than the complex numbers.e. π3 (S 2 ) = Z and π4 (S 7 ) = Z ⊕ Z12 . G4 = L3 vol X (4. where ωn is the unit volume form of S n . 4. S n ωn = 1. The entropy [15] and Weyl anomaly [16] are also proportional to (4. By the dimensional reduction from 11d SUGRA or M-theory on compact space X4 . The form f ∗ (ωn ) is closed and. must be exact. Einstein space. and writing S 2 as CP 1 . with zi ∈ C and |z1 |2 + |z2 |2 = 1. E. i. Intriligator / Nuclear Physics B 581 (2000) 257–273 267 (n − 1)-dimensional curve in S 2n−1 which is mapped by f to a point a ∈ S n . b4 ))2 (vol(X (4.1) as  9 b4 L . (4.g. we will everywhere drop universal constants (factors of 2 and π ).4) . z2 ] ∼ [λz1 . f ∗ (ωn ) = dθn−1 .. The Hopf invariant can be written as Z θn−1 ∧ dθn−1 . for simplicity. 9 G−1 7 = vol(X4 )/ lP . For n = 2k even.3) X4 so (4. The anomaly coefficient c arises as the coefficient of a Chern–Simons term in AdS7 . Thus π4k−1 (S 2k ) is at least Z.6) H (f ) = S 2n−1 Clearly. generalizing to M-theory on AdS7 × X4 . We thus write (4. are determined by dimensional analysis.1). H (f ) can be written as an integral overRS 2n−1 as follows: consider the pullback f ∗ (ωn ). as H n (S 2n−1 ) is trivial. The G4 flux quantization condition gives Z  b4 = NlP3 . The map with Hopf number 1 for S 7 → S 4 is exactly the same as that above. z2 ] with zi ∈ C∗ and [z1 . taking all integer values for various maps f .2) b4 ) is the dimensionless volume of X4 measured in units of L (normalized so where vol(X b that vol(X4 ) = 1 for X4 = S 4 of radius L). The map is then simply f : (z1 .

2) is invariant under the gauge invariance δC3 = dΛ2 . A) + fluctuations in C3 . we argue that writing the “Wess–Zumino” term of [32] as the integral over Σ7 of a 7-form. with Ωp+2 not exact. This argument is analogous to that of [29] for IIB on b5 ) = N 2 |Γ |. actually gives the Hopf–Wess–Zumino term which we want. δB2 = −ΛW 2 and satisfies a generalized self-duality condition (it is only self-dual at linear order. (4. there is a field transform to a 3-form h which is exactly self-dual [26]). AdS5 × X5 .g. thus . 0) theory with G = SU(N). but our WZ term is the integral over Σ of a form which is not exact. Of course. for Dp-branes. Following [13].1) on Wp+1 . with η4 the 4-form (2. they could be written as an integral over Σ of an exact form. E.2) arises for a D3 brane probing other D3 branes.4) gives c = N 3 |Γ |2 + lower order in N. (5. for large N . Similarly.6) c G = SO(2N) = 4N 3 + lower order in N. which is naively exact.18]) H3 = dB2 + C3W .5) implies that the anomaly is  (4. these could not be exactly the Wess–Zumino terms of the type we have argued for. naively.. there is a G4 background. (4. the WZ term generally cannot be written as the local term R (5.31] s Z ˆ T)  A(R C ∧ tr exp i(F − B)/2π ∧ . despite the fact that.2) with B2 the two-form gauge field and C3W the pull-back of the 11d C3 field to the M5 brane world-volume W6 . The point is that the naively exact 7-form actually is not exact upon properly taking into account the fact that 5-branes act as a nontrivial source for G4 in M-theory. As written. It must be written as Σp+2 Ωp+2 . with Ωp+1 the form R in (5. e. Nevertheless. Naively F5 = dC4 and there is no difference.1) SWZ = ˆ N) A(F Wp+1 and the presence of some similar terms for the M5 brane is well-known [32].268 K.1) R contains W4 C4 . which should really be written as Σ5 F5 . for Dp branes it is usually written as [30.g. in the presence of other D3 brane sources.5) in the normalization where c = N 3 (plus lower order) for the N = (2. for orbifolds X4 = S 4 /Γ . the N = (2. (5.5). however. this should be equivalent to a M5 brane in AdS7 × S 4 . We consider a probe brane in the background of the N others. for a D3 brane (5. Intriligator / Nuclear Physics B 581 (2000) 257–273 In particular. Ωp+2 = dΩp+1 . 5. corresponding to X4 = S 4 . with ˆ pullback GW 4 = Nη4 (φ. This is how (3. This H3 (5. The WZ term via the M5-brane world-volume action Branes in string or M-theory always have some sort of “Wess–Zumino” terms. The M5 brane world-volume theory depends on (with sign conventions of [13. as they are written as local integrals over the world-volume W and not over a higher-dimensional space Σ with ∂Σ = W .. writing RP4 = S 4 /Γ with Γ = Z2 . which gave c = N 2 /vol(X In particular. 0) theory with group G = SO(2N) arises from M-theory on AdS7 × RP4 and. (4. F5 is not exact.1).

This would spontaneously break the SO(3)R global symmetry of the probe string world-volume theory to an SO(2)R subgroup. . whose expectation values gives the positions of the strings in the 3 transverse directions.13) with αe = 14 nW = 12 N (exact). The SO(3)R symmetry is that of the normal bundle of the three transverse directions of these strings in 5d. Intriligator / Nuclear Physics B 581 (2000) 257–273 ˆ A) + fluctuations.15). Reduction on a Calabi–Yau 3-fold Following the discussion in [13. so (5. 4) CFT.5) also contains the coupling (2. S 1 -dimensional reduction suggests that we get the term (2. Indeed.. (5. with M5 branes wrapping a four-cycle to yield strings.8).8). given by (5. As discussed in Section 2. Perhaps this discussion is relevant in some sort of Born–Oppenheimer approximation. 269 (5.20].4) 4 . We re-write the WZ term of [32] as an integral over some Σ7 with ∂Σ7 = W6 : Z   ∗ G4 + 12 dB2 + C3Σ ∧ GΣ (5.2) becomes (2. with the correct leading order in large N coefficient of 12 N 2 . which has a SO(3)R global symmetry. we can consider the situation of separating one string from N others in these three transverse directions.g. 3. where the spreading of the wave-function is initially neglected or suppressed. If the WZ term indeed arises entirely as in (2. as the scalars Φ a have a wavefunction which spreads over all values.8) should arise from correcting αm ≈ N to αm = N + 1. this cannot really happen: there is no spontaneous symmetry breaking in 2d [33]. However. 6.5) is of exactly the form (2.2) and (5.13) needed for the π4 (S 4 ) solitonic strings to couple electrically to the B2 field as the nW = 2N “W-boson” strings. ignoring the ∗G4 term. with coefficient αe αm . C3W = NΩ3 (φ. The fact that dH3 ≈ Nη4 . a = 1. E.4) will be order N .15) with αe ≈ 12 N and αm ≈ N . similar to the C3 ∧ I8inf term of 11d SUGRA. but ∗G4 needs to be properly interpreted to see if it also contributes to the Hopf– Wess–Zumino term (naively it’s just a contribution to the AdS7 vacuum energy). Classically. We should also get the term proportional to N in (2. is needed to get the order N term in (2. has already been suggested (with ASO(5) = 0) in [24]. These strings live in the 5 uncompactified dimensions of M-theory on X and their world-volume theory is a 2d N = (0. the order N term in (2. (5.14) with αm ≈ N . There is no moduli space. 2.3) extended to Σ7 . we get Z   ∗ G4 + 12 N dB2 + NΩ3 ∧ dΩ3 . which follows from (5. We now discuss how the story with anomalies and the WZ term would go if we ignore the fact that there is actually no moduli space in 2d. in the 3 of SO(3)R . there are 2d world-volume scalars Φ a . The term ∗G4 in (5.3) and (5. Perhaps a new term. we now consider M-theory on a Calabi–Yau 3-fold X. Σ7 Σ Plugging in C3Σ and GΣ 4 = dC3 .3).8).K. Or perhaps this section is just a fairy tale.5) Σ7 This indeed contains the Hopf–WZ term (2.

1) I4 (G. X) − I4 (U (1).2) k U (N). Euler-class form e2 appearing in [20]. corresponding to ˆ is the Hopf number of the map φˆ : S 3 → S 2 .3) must be exact. . For G = U (N) [13.) We expect that G can be U (1) or an ADE group. X))H [φ]. We take the string world-volume to be W2 = ∂Σ3 . X) − I4 (H. There must be a Wess–Zumino term on the probe string world-volume to compensate for the deficit I4 (G. with background field strength F ab .2) and χ is invariant under SO(3) gauge transformations. Intriligator / Nuclear Physics B 581 (2000) 257–273 The SO(3)R symmetry is an affine Lie algebra. corresponding to the ADE classification of SU(2)k modular invariant partition functions. The coefficient of the WZ term must be quantized and properly normalized in order for e2πiΓ to be invariant under changing Σ → Σ 0 . 4 4 where p1 (F ) is as in (2.34]  (6. angular. X) − k(H.4) compensates for the deficit in the anomaly (6.4) contributes to the anomaly matching. X). 2 8π where the covariant derivatives include a background SO(3) gauge field Aab = −Aba . H [φ] where H [φ] 8 There can be an obstruction to extending φˆ from W to Σ if φˆ is in the nontrivial component of 2 3 π2 (Mc = S 2 ) = Z. the classical configuration space is Mc = SO(3)/SO(2) = S 2 and G is broken to H × U (1).2). with ∂Σ = ∂Σ 0 = W2 . 4 with F the SO(3) background field and p1 (F ) as in (2. Consider the two-form. Z    ˆ A ∧ dΩ1 φ.5) implies Z Z Z 1 1 (0) δp1 = p1(1) . with δp1(0) = dp1(1) .4) integrated over S 3 ∼ = Σ − Σ 0 .3) η2 φ. which is the SO(3)R ’t Hooft anomaly in the 2d anomaly polynomial 1 (6. ˆ ∈ Z. with level k. (6. Because H 2 (Σ3 ) is ˆ A). (We ignore gravitational contributions to I4 since they do not require a WZ term. X) = k(G. where D0 and c2 · P0 are determined in terms of the geometry of the 3-fold X and the 4-cycle of X on which the N M5 branes wrap. the form (6. (6. Writing p1 = dp1(0) . (6. X) − k(H.1) in the low energy theory. The Hopf–Wess–Zumino term is trivial. which gives (k(G. (6. η2 = dΩ1 (φ. e2Σ is the pullback (via 8 φˆ : Σ3 → S 2 ) to Σ3 of the global. and φˆ a = φ a /|φ|. X) − k(U (1).6) δ Ω1 ∧ dΩ1 = 4 4 Σ3 Σ3 W2 so (6. we note that [21]   1 ˆ A ∧ e2Σ φ. ˆ A .270 K. X)p1 (F ). ˆ A = 1 p1 (F ) + dχ.5) dΩ1 ∧ dΩ1 ≡ e2Σ φ.  1  1 ˆ A ≡ e2Σ = abc Di φˆ a Dj φˆ b − Fijab φˆ c dx i ∧ dx j . X = N 3 D0 + Nc2 · P0 /12. Γ = k(G. X) Σ3 To see that (6. √ In the fairy tale where we can consider fixed non-zero |φ| = φ a φ a . X) − k(U (1).4) Ω1 φ. The difference is ˆ (6. (6.

c(E6 ) = (78)(12) = 912. To have the functional integral be well defined under Σ → Σ 0 thus requires all (k(G. e. these were the nW = |G| − |H | − 1 gauginos. Perhaps. so we should be getting c(G) − c(H ) ≈ 3N 3 . the guessed formula incorrectly gives 3(nW /2)(nW /2 − 1) ≈ 34 N 4 since nW = N(N − 1). 7.4) of the N = (2. the explicit H dependence is in αm . then.8) case. since it is related to ’t Hooft anomalies.. X)). is c(G) = |G|C2 (G).2) where C2 (G) is the dual Coxeter number.4). via integrating out tensionful strings. 0) theory is related to an anomaly. E. would then be the general guess c(G) − c(H ) = 3(nW /2)(nW /2 + 1). 0) field theories. Perhaps.1) is correct. we might try a function only of nW which. Speculations Because the WZ term (2. and thus all k(G. and the strings are not fundamental but. rather. (7. this guess does not work for the case G = SO(2N) and H = SU(N) in the large N limit: using (4. then. running in the loop. we have c(G) ≈ 4N 3 and c(H ) ≈ N 3 . 0) theory are the BPS strings. these are not the correct degrees of freedom to be integrating out in deriving the WZ term. This suggests a WZ term proportional to nW . interacting. it seems that the coefficient of the WZ term must contain some explicit dependence on the massless. is actually possible.6). The analog in the 6d N = (2.2) is that it . 0) theory is really a field theory. On the other hand. In the 4d N = 4 theory. the skyrmionic strings of Section 2.8) that it cannot be exactly just nW . it is natural to expect that it can be found exactly by a 1-loop calculation. This gives c(SU(N)) = N 3 − N . (7.. with αm = N + 1 for the case (2. In any case.15). X) − k(H. We have not yet demonstrated that such a derivation of the WZ term (2.K. it is possible to derive the WZ term directly by a 1-loop string calculation. Intriligator / Nuclear Physics B 581 (2000) 257–273 271 π3 (S 2 ) = Z. and c(E8 ) = (248)(30) = 7440.g.g. though we know from (2. However. some kind of solitonic objects.6).1) 6 4 e. based on the SU(N) case and (4. with these nW strings.. In analogy with ordinary QFT. one might suppose that the coefficient of the WZ term is some function of only those degrees of freedom which become massive when G → H × U (1). A check of (7. coupling to φ. c(SO(2N)) = 2N(N − 1)(2N − 1). c(E7 ) = (133)(18) = 2394.g.8). H degrees of freedom. following (2. X) − k(U (1). with some fields which become massive due to hφi 6= 0 running in the loop. One might object that the N = (2. Indeed. So then the challenge is to get the factor of αm . normalized to be N for SU(N). it is hoped that reproducing the answers for the WZ terms presented here could lead to a better understanding of the 6d N = (2. which get a tension 1/α 0 (φ) ∼ φ. which get a mass via Yukawa couplings to hφi 6= 0. to always be an integer. If (7. Based on this failure. on the other hand. perhaps the distinction between fundamental vs composite degrees of freedom is irrelevant for deriving the WZ term. X). we speculated that the WZ coefficient is  1 1 c(G) − c(H ) = αe αm = nW αm . based on the (2. Our conjecture for c(G) for general G.

Theor. the conjectured formula (7. Decomposing the adjoint of G as ad(G) → W + ad(H ) for some representations W (which is the rep of the massive W -bosons. (Eds. S. hep-th/9805144. Adv. D. S. Seiberg. Note that this expression depends explicitly on H . H degrees of freedom. E. Sethi for discussions. A. Since (2. Rubakov. this could be just via αm . hep-th/9711200. which would have explicit H dependence as given by (7. Lett. Sevrin. Gorbunov. the magnetic charge of the skyrmionic strings in W6 (or membranes in Σ7 ). Dubovsky. hep-th/9903155. which here arise [8] from D3 branes wrapped on the blown-up two-cycle. This suggests that an eventual derivation of the WZ term must include effects which couple the massive degrees of freedom. Phys.V.). interacting. hep-th/9712074. Rev. 2 (1998) 231. Nahm. Bekaert. by IAS grant NSF PHY-9513835 and Rutgers grant DOE DEFG02-96ER40959. Nucl. of Strings 95.M. it is given as suggested by the above discussion. Goldstone.3).2) gives for the coefficient of the WZ term:  (7. M. hep-th/9909094. B 248 (1984) 59. Math. D’Hoker. while I was a visitor. Phys. and thus check (7. via |H | and C2 (H ). V. F. Intriligator / Nuclear Physics B 581 (2000) 257–273 is a multiple of 6. A.S. Nucl. E. It would be interesting to directly determine αm and see if. B 135 (1978) 149. Farhi. This work was supported in part by UCSD grant DOE-FG03-97ER40546 and. B 223 (1983) 422. along with a singlet = ad(U (1))) of G. J. Assuming (7. It would be interesting to derive the Hopf–Wess–Zumino term (2. A. Phys. Phys. W. Lett. G.4) depends ˆ the size |φ| of the blown-up two cycle can be arbitrarily large. Moore. B 468 (1999) 228. hep-th/9507121. Bars et al. Manohar. and how. hep-th/9812128. B 223 (1983) 433. to the massless. and S. and not only on the massive reps in W . Nucl. in: Proc. Wilczek. I would also like to thank E.10). Phys. Libanov. Rubakov. 2 (1998) 781.V. Math. Minwalla. Theor.2).3) c(G) − c(H ) = |H |C2(W ) + |W | C2 (H ) + C2 (W ) . Maldacena. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] J. 47 (1981) 986. M. in the context of IIB string theory on a C2 /ΓG ALE space.A. Manohar. V. which are integrated out. X. Henneaux. Phys.1). Phys. Witten. The π4 (S 4 ) skyrmionic strings should again be identified with the W -boson strings. Kapustin. Lett. satisfying (2.11) and thus (2. I would like to thank the IAS and Rutgers theory groups for their support and hospitality. Rev.L. D’Hoker and E. 81 (1998) 1558.4). C4 ∧ H3 ∧ H3∗ interaction of the 10d IIB string looks promising for leading to WZ terms. The only on the angular φ. Witten. Acknowledgements I would like to thank A.A. .272 K. where φ a are the periods of the 3 Kahler forms and two B fields on a blown-up two-cycle. Witten for helpful email correspondences. Phys. E. It also satisfies the c-function condition c(G) > |G| in all cases. Adv. for all ADE groups G. E. N. Nucl. Phys.

and references cited therein. Zarembo.K. Nucl. Sezgin. G. JHEP 12 (1997) 2. A. Y. Intriligator / Nuclear Physics B 581 (2000) 257–273 [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] 273 E. Weinberg. Phys. Phys. L. L. 22 (1997) 103. D. t’Hooft. D. Gauntlett. Aharony. I. ’t Hooft et al. hep-th/9706109. Minasian. S. S. Strominger. E. . Townsend.K. IJMP A 13 (1998) 5409. Phys. Moore. M. E. Freed. Phys. Skenderis. Alvarez-Gaume’. B 451 (1995) 725. hep-th/9702008. B 476 (1996) 470. hep-th/9605033. 1988. Quant. Phys. Boulahouache. Nucl. hep-th/9610234. Plenum Press. Witten. Gubser. JHEP 9807 (1998) 023. Rev. J. hep-th/9803108. Bott. C. Rev.). JHEP 9805 (1998) 009. D 60 (1999) 045004. hep-th/9710206. Comm. G.R.A. Geom. D’Hoker. Harvey. Mateos. S. R. JHEP 9809 (1998) 004.S. Klebanov. P. West. New York. E. Motl. hep-th/9604089. G. hep-th/9911246. Grav. Ganor. M.C. J.S. E. Recent Developments in Gauge Theories. E. D 50 (1994) 605. Moore. Phys. B 399 (1997) 49. G. R. B 475 (1996) 164. 31 (1973) 259. D’Hoker. hep-th/9604103. Howe. Hazewinkel (Ed. Nucl. J. Diff. Encyclopaedia of Mathematics. Yin. J.E. J. B 234 (1983) 269. 14 (1997) 47. ATMP 2 (1998) 601. E. hep-th/9711053. Math. Geom. D 59 (1999) 025006. hep-th/9801083. Tseytlin. Class. Moore. M. Witten. (Eds. K.M. dg-ga/9710001. B 507 (1997) 658. hep-th/9807164. Minasian. Witten. Harvey. O. Witten. Cattaneo. Zamaklar. Tseytlin. A. Phys. A. A.K. Z. G. M. P. Green. Phys. hep-ph/9409402.A. Maldacena. Henningson. Nucl. Nucl. 48 (1998) 91. Phys. Lett. Kluwer Academic. O. hep-th/9903156.). Phys. hep-th/9808060. J. 1980. Phys. Koehl. hep-th/9806087. R.S. Cheung. Thompson. hep-th/9803205. Rev. Coleman.B. C. Harvey. J. K. hep-th/9502162. in: G. P.

3 2 1 3 ( 0 0 ) 0 0 2 7 4 . 11. University of Silesia. can be easily obtained 1 caffo@bo.b. The expansion at pseudothreshold of the amplitudes is constructed and some lowest terms are explicitly presented.nl/locate/npe The pseudothreshold expansion of the 2-loop sunrise selfmass master amplitudes M.b.2 .V.Kk.infn.it 0550-3213/00/$ – see front matter  2000 Elsevier Science B.3 a INFN. i = 1. as a consequence of tensorial reduction. accepted 1 May 2000 Abstract The values at pseudothreshold of two loop sunrise master amplitudes with arbitrary masses are obtained by solving a system of differential equations.elsevier. This paper is devoted to the analytic evaluation of the coefficients of the expansion of the sunrise amplitudes in p2 . Sezione di Bologna.1 .  2000 Elsevier Science B. For a proper understanding of their behaviour as well as for a check of the numerical calculations. PACS: 11. E.10. Italy b Dipartimento di Fisica.pl 3 remiddi@bo.V. PL-40007 Katowice. All rights reserved. it is convenient to know the amplitudes off-shell. Due to the presence of heavy quarks. Università di Bologna.-z. Poland Received 5 January 2000. and also around some particular values of p2 . −(m1 − m2 − m3 )2 . Remiddi a.Nuclear Physics B 581 (2000) 274–294 www. if pµ is the external momentum (in n-dimensional Euclidean space). 2. and the external scalar variable p2 . Introduction The sunrise graph (also known as sunset or London transport diagram) appears naturally. 11. p2 = ∞. 3. revised 20 March 2000. in several higher order calculations in gauge theories.Bt 1. All rights reserved. Caffo a.infn. at the threshold and at the pseudothresholds. vector bosons and Higgs particles all the internal lines may carry a different mass. at the pseudothreshold value p2 = −(m1 + m2 − m3 )2 (the other pseudothreshold values −(m1 − m2 + m3 )2 .10. PII: S 0 5 5 0 . Italy c Institute of Physics. so that sunrise amplitudes depend in the general case on three internal masses mi . Czy˙z c. I-40126 Bologna.edu. I-40126 Bologna.15.1 . H.it 2 czyz@us. such as p2 = 0.

say s0 . for arbitrary value of the dimension n. It turns out that the formula. We expand the equations in the masses in the dimension n around n = 4 and solve them explicitly up to the finite part in (n − 4). the expansion around threshold was recently obtained in [7]..e. Caffo et al. further. ∞. once the initial conditions. threshold and pseudothresholds) correspond to the Fuchsian points of the differential equations. Fortunately. When the expansions are inserted in the differential equations. expanded in (n − 4) and solved explicitly up to the finite part in (n − 4). the obtained algebraic equations can then be solved recursively. while in [4] the values of the amplitudes at threshold and pseudothreshold were obtained. with different masses. involves the coefficient 1/(n − 4). the values of the sunrise amplitudes at the considered pseudothreshold. the formulae expressing the higher order coefficients of the (p2 − s0 ) expansion involve coefficients like 1/(n − 5). in Section 5 the finite part in (n − 4) of the first term in the p2 expansion is given. In Section 2 the system of differential equations for the sunrise master amplitudes is recalled. With the method established in [5] and [6]. Rather than evaluating those first order terms in (n − 4). The sixth section is a short summary. while the Appendix A contains a list of various functions (mainly polynomials) used in the text. we find in our approach a system of linear differential equations in the masses satisfied by them.. This result for n = 4 and p2 = s0 is in agreement with the literature [4]. see. etc. so that they can be expanded as a single power series in (p2 − s0 ) around that point [8]. and can be used without further problems for evaluating those higher order terms. expressing the first order coefficients of the (p2 − s0 ) expansion in terms of the zeroth order values. we prefer to evaluate directly the required finite part in n of the first order terms of the (p2 − s0 ) expansion by means of a suitably subtracted dispersion relation in (p2 − s0 ) at n = 4. i. We then look at the recursive solution of the algebraic equations for the coefficients of the expansion in (p2 − s0 ). The sunrise diagram. which is known to be satisfied by the sunrise amplitudes themselves [1]. The sunrise amplitudes are regular at any of the pseudothresholds. / Nuclear Physics B 581 (2000) 274–294 275 by permutation of the masses). Therefore the finite part at (n − 4) of the first order terms in the (p2 − s0 ) expansion involve the first order terms in (n − 4) of the zeroth order terms in (p2 − s0 ). which therefore emerge as a natural tool for their discussion. are given.M. has been investigated in [3]. not yet discussed in the literature for the pseudothreshold (our results are given in Sections 4 and 5). 1/(n − 6). the approach relies on the exploitation of the information contained in the linear system of first order differential equations in p2 . in Section 3 the resulting differential equations in the masses for the values at the pseudothreshold are derived. for example. the equations become a set of algebraic equations in the coefficients of the expansions. in Section 4 the expansion in p2 at the pseudothreshold is discussed. As those initial values are in turn functions of the masses. which are finite at n = 4. The analytic properties of Feynman diagrams at threshold and pseudothresholds are well known. [2]. . The plan of the paper is as follows. It is to be noted that all the above points (p2 = 0.

m21 . p2 ) ∂p2 3  X  m2i Fi n. being the combinations of other known polynomials times the products of two one-denominator. as in [1]. the three masses and pµ the external n-momentum. T (n. m21 . mi . m21 . m21 . where. p2 Z Z 1 dn k2 dn k1  2  . m23 . . Caffo et al. ∂mi (3) but the 4 amplitudes are otherwise independent (i. m23 . p2 Fα n. can be found in [1]). m22 . / Nuclear Physics B 581 (2000) 274–294 2. m23 . none of them can be expressed as a linear combinations of the others times rational functions of the arguments). m22 . 3. p2 = − 2 F0 n. p2 . which are referred to. m23 . m22 . m23 . m23 . and the functions Bi (n. m2 . m21 . p2 ). m2i )T (n. p2 = 2 ∂p D m1 . The equations It is known that the two-loop sunrise self-mass graph with arbitrary masses m1 . m23 . α = 0. m21 . p2 + i=1  1 ∂  p2 2 Fi n. m21 . m22 . are also known. as in [1]. m21 . 2. 1. m22 . i = 1. m22 .α (n. one-loop vacuum amplitudes of two different masses. T (n. by  (1) Fα n. which is essentially irrelevant in practice. m21 . m21 . m23 . m23 . p2 . p2 " 3 X   Ai. m22 . m21 . m23 . m22 .276 M. m21 . p2  # . (n − 2)(n − 4) (5) and C(n) is a coefficient depending on n which at n = 4 takes the value C(4) = 1 (its explicit form. m21 . constituting the non homogeneous part of the equations. m23 . m22 . m2j ). m22 . 3. p2 ) are known polynomials of the arguments. m2 ) = C(n) mn−2 . m22 . m23 . m22 . m22 . 2. m23 . m3 has four independent master amplitudes [9]. p2 .α n. m23 .e. = (n − 3)F0 n. p2 × α=0 + Bi n. In [1] it was shown how to write for them a system of first order linear differential equations of the form p2 ∂ F0 (n. 3   ∂ Fi n. m23 . (2) = n−2 n−2 2 2 2 (2π) (2π) k1 + m1 k2 + m2 (p − k1 − k2 )2 + m23 while for i = 1. p2 ) is the scalar amplitude  F0 n. 2. (4) where the Ai.. m22 . m21 . where n is the continuous number of dimensions. F0 (n. m22 .

m1 . m23 . p2 = −(m1 + m2 − m3 )2 = H (0. (3). (4). p2 ) and the Bi (n.h. we can expand F0 (n. α = 0. for the following. m22 . p2 ) have the same structure as the Ai. m22 . G0 (n. such as  1 ∂ Fi n. m22 . m23 . m23 . m1 . −(m1 − m2 − m3 )2 .0)(n. m23 . and at the physical threshold p2 = −(m1 + m2 + m3 )2 . 2. 3. m3 ) p2 + (m1 + m2 − m3 )2 2 + H (α. The pseudothreshold values of the master amplitudes As the master amplitudes are regular at the pseudothresholds.1)(n. m3 ) ≡ F0 n. from which we can write at once. m22 . it is sufficient to give the explicit expression of the polynomial D(m21 . m1 . m22 .0)(n. The recurrence relations solving the integration by part identities of [10. (4) must be canceled by a corresponding zero in the numerator. m23 . m23 . m1 . p2 = p2 + (m1 + m2 + m3 )2 p2 + (m1 + m2 − m3 )2    (6) × p2 + (m1 − m2 + m3 )2 p2 + (m1 − m2 − m3 )2 . p2 ) of Eq. m23 . m21 . m22 .s. therefore.11] for the sunrise amplitudes can be also written (see [9] or Appendix of [1]) in a form similar to Eq. m21 . of Eq. (4) is given in [1] and is not repeated here for the sake of brevity.α (n. m23 . which reads     D m21 . m23 . m22 .M. m22 . 2. the Ri. It is known that the Fi (n. m22 . m21 . as in Eq. −(m1 − m2 + m3 )2 . p2 ) in p2 around that point as  Fα n. p2 = H (α. m22 . m1 . m2 . m21 . m23 . m23 . p  2 # . p2 ). / Nuclear Physics B 581 (2000) 274–294 277 The explicit form of the Ai. appearing in Eq. m22 . 3. m22 . p2 Fα n. m2 . m2 .α (n. m22 . m23 . m21 . m3 ) p2 + (m1 + m2 − m3 )2 + · · · . (4). (7) where i = 1. m21 . the apparent pseudothreshold singularity of Eq. The above polynomial vanish when p2 takes one of the three pseudothreshold values −(m1 + m2 − m3 )2 . p2 × α=0 + Si n. m21 .α n. m21 . considering for definiteness the pseudothreshold at p2 = −(m1 +m2 −m3 )2 . (4).2)(n. Caffo et al. m3 ). p2 ) " 3 X   Ri. m22 . 1. m22 . p2 ). the apparent pseudothreshold pole in the r. m3 )  + H (α. p2 ) and their p2 derivatives are regular at the pseudothresholds. m21 . m21 .α (n. m23 . p2 = 2 ∂m3 D(m21 . p2 ). m21 . m23 . m23 . p2 ) and the Si (n. m21 . by using the definitions Eq. m22 . m2 . (8)  . m22 . Bi (n. 3. m23 . m21 . m2 . (7) must be canceled by a corresponding zero of the numerator. m23 . m21 .

m1 . m2 . m1 . m3 ). m3 ) ∂ = − 2 H (0. which can be written as (n − 3)(3n − 8) G0 (n. α = 0. m1 . m1 . m22 . m1 . m3 ) (11) and. m3 ) + 4m2 m3   (n − 2)2 T n. 3. m2 . m3 ) ∂m2 m1 + m2 − m3 (0. m23 + (n − 3)m2 (2m2 − m3 + m1 )G2 (n. m1 . m1 . and their derivatives in respect to m3 . Caffo et al. m2 . m2 . m1 . m3 ) ∂ = − 2 H (0. m2 . / Nuclear Physics B 581 (2000) 274–294 G1 (n. m3 ) ∂m3 m1 + m2 − m3 (0. p2 = −(m1 + m2 − m3 )2 = H (2. m2 . 2. m21 T n. m1 . m2 . m23 . m23 . m2 . m3 ) − 2   (n − 2)2 T n. m2 .1) H (n. m2 . m3 ) ≡ F3 n. (m3 − m1 − m2 )(m3 − m1 )(m3 − m2 ) ∂ Gi (n. m1 . m1 . m2 . m21 T n. m3 ) + 4m1 m3   (n − 2)2 T n. m22 T n.1) H (n. m3 ) 2 ∂m3 = (n − 3)G0 (n. for i = 1. p2 = −(m1 + m2 − m3 )2 = H (3. m1 . m1 . m2 . m1 . m2 . m3 ) ∂m1 m1 + m2 − m3 (0. + (n − 3)m3 (2m3 − m1 − m2 )G3 (n. m1 . (7) for i = 1. m21 . m2 . 2. m1 . m2 . we find only five algebraically independent equations for Gα (n. m22 . m2 . m3 ). m3 ) + m21 G1 (n. m3 ) + m22 G2 (n. − m1  G2 (n. (9) + m3 By inserting the above expansions in the four equations Eq.0)(n. m1 . 3. − m2  G3 (n. m2 . m2 . m3 ) ∂m3 . m1 . m3 ) ≡ F1 n. m1 . m22 . 2.0)(n.0)(n. m23 = (n − 3)m1 (2m1 − m3 + m2 )G1 (n.0)(n. m2 . m1 . m1 . m21 . (4) and in the three equations Eq. m2 .0)(n. m3 ) ∂ = − 2 H (0. m21 . m2 . m3 ) − 4m1 m2 (10) m 3 − m1 − m2 ∂ G0 (n.0)(n. m3 ) ≡ F2 n. 3. m23 . m1 .1) H (n. p2 = −(m1 + m2 − m3 )2  = H (1.278 M. m3 ). m2 . m3 ). m3 ) +m3 (m1 + m2 )G3 (n. m1 . m2 . 1. m22 .

m2 .1 (n. m22 . while the rational functions Pi. m3 ) and Qi. m1 . m3 2   m2 1 P3. m1 . m3 ) = 3 2 3 1 . m2 . m1 . Q1. m3 )   + Qi.j (n. P2. 2 4m22 m23 Q1. m2 . m1 . m2 ) is defined in Eq. m3 ) = − (13) . m2 . m2 . m3 ) = − 2   1 P2. m3 ) = −(n − 3) (m3 − m2 ) m3 − m1 − m2 .2 (n. m23   + Qi. 2   m1 1 P3. m1 .j (n. (12) where T (n. P2. m22 T n.2 (n. m2 . m2 .3 (n.M. m1 .1 (n. m2 . m3 ) + Pi. m21 T n. m2 . m1 .2 (n. m3 )T n.2 (n. m1 . m1 . m3 )T n. m3 ) = −(n − 3) m3 (m3 − m1 − m2 ) + m1 (m1 + m2 ) .1 (n. m2 . m1 . m3 )G2 (n. 8m21 m2   (n − 2)2 1 m (m − m ) m − Q2. m3 ) = −(n − 3) m3 (m3 − m1 − m2 ) + m2 (m1 + m2 ) . m2 . m3 ) = 2   1 P1.2 (n. m1 . m1 .3 (n. m1 . m3 ) (which differ from the functions of the same name.3 (n.1 (n. m1 .3 (n. m1 . m2 . m3 ) = − and (n − 2)2 (m3 − m2 ). m1 .1 (n. m23   + Qi. m3 ) = − (n − 2)2 (m2 − m1 ). / Nuclear Physics B 581 (2000) 274–294 279 = Pi. m3 ) = m3 n−3 4(m3 − m1 )(m3 − m2 )(m3 − m1 − m2 ) + 2m3  + m1 m2 (m1 + m2 − 2m3 ) P1. m1 . m21 T n.3 (n. m3 ) = 2 n−3 m1 (m3 − m1 ) − (m3 − m1 )(m3 − m2 ). P1. 8m2 m23   (n − 2)2 1 (m3 − m1 ) m3 − m2 . m2 . m2 . m2 . m1 . m3 )T n. m2 . m2 .1 (n. m2 . (5). 2 n−3 m2 (m3 − m1 ). m2 . appearing in [1]) are defined as n−3 m2 (m3 − m2 ) − (m3 − m1 )(m3 − m2 ). m1 . Caffo et al. m1 . but different arguments. m3 2 1 (m3 − m1 )(m3 − m2 )(m1 + m2 − 2m3 ) P3. m3 ) = −(n − 3) (m3 − m1 ) m3 − m1 − m2 .2 (n. m1 . m2 . m2 . m1 . m1 .3 (n.1 (n. m3 ) = 2 4m21 m23 Q1. m3 )G3 (n. m2 . m2 . m2 . m1 . 2 n−3 m1 (m3 − m2 ). m3 )G1 (n. m3 ) + Pi.

that means that only three of the four master amplitudes are independent at the pseudothreshold and. m ) + O(n − 4) . according to the notation of [1] we write Gα (n. As any system of k linear first order equations is equivalent to a single kth order equation. again from [1]. m2 . m1 . we transform the system Eq. m3 )  1 1 G(−1) (m1 . m2 . m2 .1 (n. α = 0. m3 ) and its derivatives with respect to m3 . we have  1 2 m + m22 + m23 . m2 . m3 ) = .2 (n. m1 . m3 ) G(−2) (m1 . as we show later. which we choose to be G0 (n. m3 ) = − 4m1 m2 m3 2 2 Q2. m1 . m3 ). m2 . 8 1  1 3 m21 + m22 + m23 (m1 . 8m1 m23 (n − 2)2 (m1 − m2 ). where the singular parts of Fα (n. and not G0 (n.2 (n. 3. further. m1 . m2 . 1. / Nuclear Physics B 581 (2000) 274–294 Q2. m3 ) in terms of the Gi (n. m1 . 8m1 m22   (n − 2)2 1 (m3 − m2 ) m3 − m1 . the Gi (n. m2 . m2 . we use Eqs. Q3. m1 . m21 . m1 . 8 1 (−2) Gi (m1 . m2 . m3 ) + = C 2 (n) α 2 (n − 4) (n − 4) α  (m . is solved. i = 1. m2 .280 M. m3 ). m3 ) = − (14) Three equations (12) form a system of equations involving only the master amplitudes Gi (n. Before continuing. m3 ) = − G(−2) 0 . (11) is implied by the other four equations and is not an independent equation. m3 ). 2. m3 ). (12) into a single third order differential equation for a single unknown function. m3 ) can be easily recovered. (10) and (11) for expressing the Gi (n. 2. G0 (n. i. m1 . m2 . as a consequence. m2 . m1 . m1 . m1 . m1 .3 (n. m22 .e. m3 ) = 2 4m2 m33   2 (n − 2) 1 (m3 − m1 ) m3 − m2 . m2 . m2 . m2 . m3 ). m . m2 .3 (n. + G(0) 1 2 3 α (15) The function C(n) has been already introduced above. m1 . m2 . m23 . m3 ) is known explicitly. m2 . Q3. m1 . Q3. m3 ) = 2 4m1 m33   (n − 2)2 1 1 m3 − m1 − m2 . 3. m1 . m2 . (10) can be regarded as an identity expressing G0 (n. m3 ) = − (m1 + m2 − m3 )2 + G(−1) 0 32 16    1 2 2 2 − m1 log m1 + m2 log m22 + m23 log m23 . m3 ) in terms of G0 (n. m2 . p2 ) are given for arbitrary values of p2 . 8 (m1 . Caffo et al. while Eq. we expand all the functions in n around n = 4. m1 .. m2 . m3 ) = − (n − 2)2 (m3 − m1 ). Once the equation for G0 (n. Eq.

1)(m1 . m3 ) = −2x 4 + 2(m1 + m2 )x 3 + (3m21 + 8m1 m2 + 3m22 )x 2 + 4m1 m2 (m1 + m2 )x. m2 . m2 . m3 )a(m1 . which is written in the Appendix A. (17). (21) is that G0(0. m3 ) in the form of a rational function in m3 and m1 . . (19) can be given the explicit form G(0) 0 (m1 .e. 3. m2 .   1 g2 (m1 . i. m2 . m3 ). m3 ) ∂2 a(m1 . m3 ) by imposing that in the resulting differential equation for G0(0. m2 . m3 ) ∂m3 0 (17) +g(m1 . m2 . m3 ) = −6 m3 − (m1 + m2 ) m23 − m3 (m1 + m2 ) + m1 m2 . by imposing g3 (m1 . m3 ) 3 ∂m3 + g1 (m1 . m3 ) ∂m23 ∂ a(m1 .. m2 . Caffo et al. m3 ) then reads g3 (m1 . m3 ) does not appear anymore in the equation). m2 . m3 )G0 (m1 . m2 . m2 . m2 . m2 . are the following polynomials g3 (m1 . In order to solve Eq. +g1 (m1 . m2 . (19) with the idea of fixing the otherwise undetermined function a(m1 . 2(m3 − m1 − m2 )2 0 (21) With the above substitution Eq. m3 ) = − 2m3 − m1 − m2 G(0. m3 ) ∂3 a(m1. m2 . m3 ) + g0 (m1 . m3 ) = (m3 − m1 )(m3 − m2 ) 5m3 + (m1 + m2 ) x 2 . m2 . m2 . m3 ) = −  1 1 + log m2i . while the function g(m1 . 3. m3 ) ∂3 G(0)(m1 . m2 . m3 ). is a combination of similar polynomials and of first and second powers of the logarithms of the masses generated in the expansion in (n − 4) of the T (n. m3 ). m2 . m2 . m2 . m2 . 281 (16) The third order differential equation for G(0) 0 (m1 .    1 1 g0 (m1 . m3 ) were the gα (m1 . By iteration of the method. m3 ) = m3 (m3 − m1 )(m3 − m2 )x 3 . m2 . m2 .1)(m1 .M. we make the ansatz (0) (0. (18) 2 3 with x = m3 − m1 − m2 . 2. the following chain of substitutions is obtained . m2 . m3 ) + g0 (m1 . We further look for a(m1 . m2 . m3 ) = a(m1 . m3 ) ∂m23 0 ∂ (0) (0) G (m1 . by trial and error we find that Eq. m2 . m3 ) + g2 (m1 . 16 8 i = 1. m2 .1) (m1 . m2 . m3 ). m2 ). m3 ) + g2 (m1 . m2 . m2 . m3 ) = 0. m3 ) 3 0 ∂m3 ∂ 2 (0) G (m1 . m3 )G0 (m1 . . (17). m3 ) the coefficient of the term without derivatives vanishes. 2 g1 (m1 . α = 0. m2 . . m2 . m2 . m2 .1) G0 (m1 . m2 . m3 ) = 0. / Nuclear Physics B 581 (2000) 274–294 (−1) Gi (m1 . (17) has taken the form of a second order differential equation in the first m3 -derivative (the essential property of the transformation Eq. (20) ∂m3 which of course is nothing but the homogeneous version of Eq. .

(m3 − m1 )2 (m3 − m2 )2 m23 − (m1 + m2 )m3 − 13 m1 m2 a1 = (0. / Nuclear Physics B 581 (2000) 274–294 ∂ G(0.4)(m1 .3) × G0 (m1 . m2 .2)(m1 . where the master integrals of the sunrise graph are known in closed analytic form (0) [1].0m33 + q3. m2 . m2 . m2 .1) (m1 . ∂m3 0   1 a1 (0.1 m3 log(m3 ) = m3 (m3 − m1 − m2 ) −3/2  2 2 + q−1. m2 .2 log (m3 ) + q1.2m−1 3 log (m3 ) + q0.5)(m1 . From Eqs. m2 . m3 ). but in fact it is a very simple differential equation. indeed.0 m−1 + m3 (m3 − m1 − m2 ) 3 + q−1. satisfies the following first order differential equation ∂ G(0. Eq. m2 . which for that value of m3 corresponds to p2 = 0.4)(m1 . m3 ).5)(m1 . we are considering the pseudothreshold value p2 = −(m1 + m2 − m3 )2 .3) (m1 . m2 . (24) where the coefficients qα . finally.2 m3 log2 (m3 ) + q2.1 log(m3 ) + q0. m2 . ∂m3 0 1 G0(0.0 + q0. 6 1 (m1 + 3m2 )(3m1 + m2 )(m1 − m2 )2 .0m3 + q1. m3 ) = a2 G0(0. 12 √ (2m3 − m1 − m2 ) m3 (m3 − m1 − m2 ) a2 =  2 . m3 ) =− 2 8(m3 − m1 − m2 ) . 2 (22) where a0 = −  1 3m21 + 2m1 m2 + 3m22 . m3 ).1m43 log(m3 ) + q5.2 m33 log2 (m3 )  + q4. m2 . (24) is lengthy.0m43 + q4.282 M. m3 ) = (2m3 − m1 − m2 ) + a0 + 4 (2m3 − m1 − m2 )2 (0. Once the integration constants are fixed. m3 ) = G0(0. m3 ) ∂m3 0  q0 + q1 log(m3 ) + q2 log2 (m3 ) −1/2  −1 q−1. m2 . ∂ G(0. m3 ) = G0(0. m2 . m3 ).0 m53 + q5.2) 2 G0 (m1 . which can immediately be solved by quadrature.1m3 log(m3 ) + q1. m3 )  3 1 m1 (m1 + 2m2 )L(m1 . m3 ).2 m23 log2 (m3 ) + q3.1m33 log(m3 ) + q3. m3 ) reads (0) G0 (m1 . Caffo et al.5) G0 (23) (m1 . qα. m2 .1 m53 log(m3 ) . (22) and (24) is obvious that we need (0) to integrate three times in m3 to get G0 (m1 .β are functions of m1 and m2 only and are reported in the Appendix A. the explicit form of G0 (m1 .1m23 log(m3 ) + q2.0m23 + q2. m3 ). m2 . The integrations are elementary and the integration constants can be fixed at m3 = m1 + m2 .

as done already in Eq. m3 ) + m1 (m3 − m2 )G1 (n. m2 . m1 . m3 ). m2 . m2 . m1 . m2 . we can express Gi (n. (25) − 128 64 where         m3 m1 m3 m1 − Li2 − + log log . and we get after some algebra the following two additional relations ∂ G0 (n. m2 . m3 ) 8   1  2 2 2 m1 log m1 + m22 log2 m22 + m23 log2 m23 − 32       + m21 + m22 − m23 log m21 log m22 + m21 − m22 + m23 log m21 log m23     + −m21 + m22 + m23 log m22 log m23 − m1 (7m1 + 2m3 ) log m21    − m2 (7m2 + 2m3 ) log m22 + 2m21 + 2m1 m2 + 2m22 − 5m23 log m23  13 11 m21 + m22 + m23 − (−m1 m2 + m1 m3 + m2 m3 ). m3 ) and G3 (n. m2 . m1 . 2. m1 . m2 . m3 ) only. which allows to express master amplitudes Gα (n. m1 . m2 . 2. m1 . as a function of first derivatives of G0 (n. To obtain such relations. and other amplitudes G1 (n. m3 ) = Li2 1 − m2 m2 m1 + m2 m2 The above result is in agreement with [4]. m3 ) + m23 G3 (n. ∂ G0 (n. m3 ) is known in closed analytic form. m2 . as a function of G0 (n. (26) L(m1 . (9) of [1]) and the n-dependent overall coefficients of [4]). 3. / Nuclear Physics B 581 (2000) 274–294 283  +m32 (2m1 + m2 )L(m2 . m1 . m2 . m1 . m3 ) ∂m1  −2 (n − 3)G0 (n. i = 1. As G0 (n. m2 . m3 ). m3 ) and two of . m2 . To check it we have simply to expand in n around n = 4 our overall coefficient C(n) (Eq. Caffo et al. we differentiate G0 (n. m1 . m2 . As already mentioned. m3 ) + m22 L(m2 . (10) and (27) form a system of equations. m3 ) = m3 − m1 − m2  + m22 G2 (n. m3 ) − 4(m3 − m1 − m2 )       m3 m3 1 + m2 log + m21 + m22 + m1 m2 m1 log 2 m1 m2   1 + m23 − m21 − m22 L(m1 . m3 ) + m23 G3 (n. m3 )   1 (m1 + m2 ) m21 L(m1 . m1 . m3 ) . m3 ) = m3 − m1 − m2  (27) + m2 (m3 − m1 )G2 (n. m2 . m1 . m1 . m1 . m3 ) is already known. m2 . α = 0. m3 ). m3 ) can be easily obtained through simple relations with them.M. m1 . m2 . m1 . m2 . G2 (n. m3 ) + m21 G1 (n. 1. m2 . m2 . m1 . The first of the Eqs (12). m3 ) . m2 . once G0 (n. m1 . m3 ) + L(m2 . (11) for m3 . m1 . m3 ) in its integral representation with respect to m1 and m2 . m2 . m1 . m1 . m3 ) ∂m2  −2 (n − 3)G0 (n. its derivatives are also known. 3. m2 . m2 . m1 . m1 . m1 .

3. m3 ) + L(m2 . m3 ) = 2 8(m3 − m1 − m2 ) 1 + 4(m3 − m1 − m2 )       1 m3 m3 m1 log + m2 log × (m1 + m2 )L(m1 . m1 . 3. m3 ) m3 − m1 − m2      1 (n − 2)2 1 T n. m1 . α = 0. (16). m1 . m21 T n. m1 . m22 T n. m1 . m2 . m21 T n. m23 − 2(n − 3) m2 m1    m3 T n. automatically satisfy the Eq. i = 1. m2 . m23 + T n. m2 . m21 T n. m1 . 2. m3 ) (m1 − 3m3 ) = 8m1 m3 ∂m1 ∂ G0 (n. m3 ) can be easily found from G1 (n. − m1 m2 G3 (n. m3 ) + (m2 + m3 ) ∂m2  1 + (2(n − 3)(m1 + m2 ) − (n − 4)m3 )G0 (n. m3 )   1 m1 (m1 + 2m2 )L(m1 . m1 . m3 ) 8   1  2 2 log m1 + log m21 log m22 + 32 . m2 → m1 ) G1 (n. m2 . already explicitly given in Eq. m22 . (28) − m1 m2 (−2) (−1) Expanding around n = 4 we find that Gα (m1 . while for G(0) i (m1 . m3 ). m2 . m3 ). m23 + T n. m22 T n. m21 T n. m2 . m22 . 2. m3 ) + (3m2 − m3 ) ∂m2   1 (n − 3)(6m1 + 6m2 − 7m3 ) − m3 G0 (n.284 M. m3 )  ∂ 1 G0 (n. m1 . m3 )  ∂ 1 G0 (n. m2 . m2 . m2 . The solution reads (the expression for G2 (n. Caffo et al. m3 ) and Gα (m1 . / Nuclear Physics B 581 (2000) 274–294 its first derivatives (we choose derivatives in respect to m1 and m2 ). m2 . m2 . m2 . m2 . m3 ) (3m1 − m3 ) = 2 ∂m1 8m3 ∂ G0 (n. m2 . m2 . m1 . m3 ) by the substitution m1 → m2 . m3 ) + m3 − m1 − m2      1 (n − 2)2 1 T n. we find the explicit expressions (0) G1 (m1 . m3 ) − m22 L(m2 . m3 ) + 2 m1 m2  1 + L(m1 . m1 . m1 . (28). m23 + 2(n − 3) m2 m1    m3 T n. m2 . 1. m2 . m1 .

we have H (0. m2 . m2 . m2 . m3 ) + O(n − 4) . m1 . = C (n) 32 (n − 4) (31) with H (0. m3 ) + m22 L(m2 . m3 ) + m22 L(m2 . Putting expansions Eq. m3 )   1 2 (0. m3 ) is defined by Eq. m3 ) . m3 )   1 + −2(m1 + m2 ) m21 L(m1 . Eq. m2 . m2 . (8) into the system of equations Eq. m3 ) =− 4 8(m3 − m1 − m2 )  + m32 (2m1 + m2 )L(m2 .1) +H (n = 4. m3 ) + m23 G3 (n. m2 . m3 )  1 (n − 3)G0 (n. m3 ) = 8(m3 − m1 − m2 )2 1      m3 m3 1 m1 log + + m2 log 8(m3 − m1 − m2 ) m1 m2  1 − L(m1 .M. (29) is obtained 4. m1 .1)(n. m1 . m2 . m1 . m1 . m2 . The expansion at the pseudothreshold As recalled in [1]. (4) we get (comparing the lowest terms in expansion) from the first equation H (0. m2 . m1 . m3 ) 8   1  2 2 log m3 − log m21 log m22 + 32       + log m21 log m23 + log m22 log m23 − 2 log m23 − 1 . m2 . / Nuclear Physics B 581 (2000) 274–294 285        + log m21 log m23 − log m22 log m23 − 4 log m21 + 2 log m23 − 1 . m1 . m3 ). m1 . m1 . (26). m2 . m1 . m2 . m2 → m1 in the expression of G1 (m1 . In this section we obtain the expansions in p2 at the pseudothreshold of all the master integrals Eq.1)(n. (1). m2 . m1 . m3 )   2 1 m L(m1 . Substituting for the Gα (n. it is easy to find the coefficients of the expansion of a function in any variable once a system of differential equations in that variable is given. G(0) 3 (m1 . m2 . The function G2 (m1 . m3 ) (0) by substituting m1 → m2 . (0) where L(m1 . m3 ) + m21 G1 (n. m2 . m3 ) =− 2 (m3 − m1 − m2 )  (30) + m22 G2 (n. m1 . Caffo et al. m3 )  3 1 m1 (m1 + 2m2 )L(m1 .1)(n = 4. (8). m3 ) their expressions expanded around n = 4. m2 . m3 ) + L(m2 . m2 . m3 ) 3 8(m3 − m1 − m2 )       m1 m2 m1 m2 3 3 log log + m + m + m1 m2 (m1 + m2 ) log 1 2 m3 m3 m23 . m1 .

m2 . (a similar problem was found in [7] for the expansion at the threshold). m1 . m3 ) ∂m21 m3 − m1 − m2 (0. 16 We do not put here the rest of the expression as it is not relevant for the subsequent discussion. m1 . m2 . so we decided to get H (i. m1 . m3 ).286 M. m3 ) and H (3. m2 . m3 ) + · · · . i = 1. m1 . As it is evident from the Eq. m2 . m3 ) 16   1 (33) + (2m3 − m1 − m2 )G3 (n. m2 . The solution for H (1. (33). m1 . m2 .1)(n. m3 ) H (3. 3. in a simpler way. m3 ) = − ∂m22 m3 − m1 − m2 (0. m3 ) = − (34) .1)(n. m3 ) H (2.2) +2 H (n. m2 . Similar formulae are found for H (2. m2 . m3 ) reads H (1. m2 . m3 ) we need terms ∼ (n − 4) of the Gi (n. m3 ) 8(m3 − m1 − m2 )2 1  1 − m21 + m1 m2 + m22 2         3 3 m1 m2 − m2 m1 + m2 log − m1 m1 + m2 log 2 m3 2 m3 5 1 m 1 + m2 + log(m3 ) − . Caffo et al. m1 . The integrals obtained that way contain up to trilogarithmic functions (which are however expected to cancel out in the combination appearing in the r. The terms ∼ (n − 4) are not yet known and could in principle be investigated with the method used in the previous section. m3 ).1)(n. m2 . m3 ). m1 . m3 H (1. m1 ∂H (0. 2. + 16(m3 − m1 − m2 ) 16 128 − (32) The other three equations form a system of linear equations. m3 ). m2 .1)(n.1)(n. which can be easily solved algebraically. 2. m1 .2) −2 H (n. m1 . m2 . m1 . m2 . m3 ).2) +2 H (n.1)(n.1)(n. m1 . m2 . to get terms ∼ (n − 4)0 in the (n − 4) expansion of H (1. (33) — see below). m3 ). m2 . / Nuclear Physics B 581 (2000) 274–294  1 m2 L(m1 .s. m3 ) =− m1 (m3 − m1 − m2 )2 n − 4 16 1 + (m3 − m1 − 2m2 )G2 (n. m2 . (3) we get the following relations ∂H (0.1) (n. m1 . i = 1.1)(n. m3 ) = − ∂m23 m3 − m1 − m2 (0.1)(n. m2 . m1 . m2 . m1 . 3. m1 .1)(n. m3 )   1 1 1 (m3 − 2m1 − m2 )G1 (n.1)(n. m2 .h. m2 ∂H (0. m1 . m1 . m1 . m3 ) + m22 L(m2 . m1 . By using Eq. m2 . of Eq. m1 .

(35) m3 Therefore.1)(n = 4.1) (n. 2. m1 .2)(n. The explicit expressions of the solutions are easily obtained with an algebraic program. m2 . m2 . m3 ) H (2.1)(n = 4.1)(n = 4. m3 ) H (3. m3 ). i = 1. m2 . m1 . Let us also observe that for the practical purposes of solving the system of differential equations numerically only the values at pseudothresholds and first derivatives are needed. m1 . Eq. / Nuclear Physics B 581 (2000) 274–294 287 As the pole terms in (n − 4) expansion of H (i.2) +2 H (n = 4. m2 . . m3 )(−4m1 − 8m2 + 4m3 ) (0) + G3 (m1 . m2 . 3. m3 ) . m2 .2)(n = 4. m1 . The reason is that the troublesome denominator 1/(n − 4) appears only when expressing the first order coefficients in terms of the zeroth order coefficients. m3 ) = C 2 (n) − ∂m21  m3 − m1 − m2 (0.1)(n = 4. m3 )(−8m1 − 4m2 + 4m3 ) + G(0) 2 (m1 . m2 . m3 ) . m2 . and to have the explicit form of H (0. m3 ) . (33) have to cancel) and the relation between them. 1/(n − 6). i = 1. It is worth to note that the functions G0i (m1 . m2 . m1 . . m2 . regular at n = 4.1)(n = 4. (34) takes the form for n = 4  ∂H (0. (31) we deduce that H (0. m3 ) = C 2 (n) − ∂m22  m3 − m1 − m2 (0. 3. (34) and with the result in Eq. 2. m2 . m3 )(−4m1 − 4m2 + 8m3 )  1 + (−m1 − m2 + m3 ) + log m21 (−m2 + m3 ) 2    + log m22 (−m1 + m3 ) − log m23 (m1 + m2 ) . m1 . m3 ). m1 .M. by solving iteratively the system of linear equations. (29). m3 ) H (1. There is no further obstacle in obtaining higher order terms in the p2 expansion at the pseudothreshold. m1 . i = 1. (32). .2) −2 H (n = 4. it is enough to insert the derivatives in respect to the masses. m1 . m2  ∂H (0. m2 . . 2. but they are somewhat lengthy and we do not present them here for the sake of brevity. m1 . m2 . m2 . m3 ).1)(n = 4. 3 are not independent (the pole terms in Eq. using Eq.2) +2 H (n = 4. reads (0) 0 = G1 (m1 . m2 . m3 ) has no pole at n = 4. m3 ). m1 . from Eq. the denominators appearing at higher orders are 1/(n − 5). m2 . easily obtained from Eq. (15) and Eq.1)(n = 4. m1 . to find H (i. m3 ) = C 2 (n) − ∂m23  m3 − m1 − m2 (0. obtaining the coefficients of any order in terms of the coefficients of lower orders. We evaluate it in the next section using dispersion relation method. Caffo et al. m2 . as can be verified with an explicit algebraic calculation. are known to vanish [1]. m1 . m1  ∂H (0.

m21 . 0   e p2 . b. m22 . (40) The function H (0. m3 . m1 .2)(n = 4. m2 . 0 + F (37) e(p2 ) reads The subtracted dispersion relation for F ep F 2  = p  2 2 Z∞ (m1 +m2 +m3 )2 du u2 (u + p2 )  1 Im F0 n = 4. m23  u u − (m3 − m1 − m2 )2 3 . m23 . u . m22 . m21 . (29). m1. m1 . z) = x 2 + y 2 + z2 − 2xy − 2xz − 2yz. p2 = F0 n = 4. m21 . m1 . + (36) This relation is of course fulfilled by the already known solutions Eq. m22 . m21.2)(n = 4. 5. m21 . + p2 F00 n = 4. m22 1 db = 16 b (m1 +m2 )2 Z∞ du √ ( b+m3 )2 R u. m21 . db u b (m1 +m2 (39) )2 with the usual definition of the function R q R(x. m2 . m2 .2)(n = 4. m23 R b. m1 . m22 . (38) reads  1 e00 2 p = −(m3 − m1 − m2 )2 H (0. m2 . m23 . (41) 16 u − (m3 − m1 − m2 )2 π 2 (m1 +m2 +m3 ) By using Eq. m23 . Caffo et al.2)(n = 4. m3 ) e(p2 ) through the equation Let us define the function F   F0 n = 4. u = π 16 √ 2 ( u−m Z 3)   R u. (37). Eq. u . m22 . b. m3 ) which we want to calculate in this section in the notation of Eq. m23 . π (38) where  1 1 Im F0 n = 4. m3 )  Z∞ R b. / Nuclear Physics B 581 (2000) 274–294  1  2 2 log m1 (2m1 + m2 − m3 ) + log2 m22 (m1 + 2m2 − m3 ) 8   + log2 m23 (m1 + m2 − 2m3 )     1 + log m21 log m22 (m1 + m2 ) + log m21 log m23 (m1 − m3 ) 4    + log m22 log m23 (m2 − m3 ) . The value of H (0. m23 . m3 ) = F 2 Z∞  du 1 1 2 2 2 = 3 Im F0 n = 4. m2 . y. (42) .288 M. m21 . m22 . (39) and exchanging the order of integration we get H (0.

−m3 ) + m2 + m23 − m1 m2 + m1 m3 = 4 5 32 2(m3 − m1 − m2 ) (m3 − m1 − m2 )  2m2 log(m2 /m3 ) 2 m1 + m23 − m1 m2 + m2 m3 + 5 (m3 − m1 − m2 )  2I0 2 2 2 2 2 2 m − m m − m m 2m + 1 2 1 3 2 3 (m3 − m1 − m2 )6  2m23 I1 m21 − m22 − 6 (m3 − m1 − m2 )  m1 m2 m3 I2  . to perform the integration in b of the derivative so obtained (the integration in b being now elementary) and then to integrate again in the mass. (m1 − m3 )2 . m1 . m3 )   2m1 log(m1 /m3 ) 2 1 R 2 (m1 . i = 0. (m1 + m2 − m3 )2 . (m2 − m3 )2 R b. are defined in the following way √   Z∞ log b/m3 log(yS ) + db  I0 = R b. m23  (44) (45) and yS = b + m23 − (m1 + m2 − m3 )2 − R(b) . m21 . / Nuclear Physics B 581 (2000) 274–294 289 The integration in u is elementary and gives H (0. m22 (m1 +m2 )2 . √ 2 b m3 (46) To perform the last integration. it is convenient to differentiate with respect to one of the masses Eq. m1 . m3 ) = 32  R b. m22 e(b).2)(n = 4. (47) − (m3 − m1 − m2 )3 R (m1 + m2 )2 . 1. m2 . + 4 2 2 (m3 − m1 − m2 ) (m3 − m1 − m2 ) R (b) R (b) with R(b) = R b. m2 . m2 . (m2 − m3 )2 where the integrals Ii .2) Z∞ 1 (n = 4. H db b (43) (m1 +m2 )2 where e(b) = H log(y) R(b)(m3 − m1 − m2 )2 2    2 m23 − b 2 b + m23 4bm23 × − + + (m3 − m1 − m2 )4 (m3 − m1 − m2 )2 R2 (b)   −b + m23 b + log m23 (m3 − m1 − m2 )6 + b + m23 1 2 1 − 2 . (m1 − m3 )2 . m21 . As a final result we find H (0. 2. Caffo et al. (43) in its integral form.M.

a system of four algebraic linear equations. Summary In this paper we have presented the expansion of the 2-loop sunrise selfmass master amplitudes at the pseudothreshold p2 = −(m1 + m2 − m3 )2 . (m2 − m3 )2  2       2  t (1 + t12 ) 1 − t12 t22 t2 (1 − t1 )2 1 − t1 t2 log − log log = −5ζ2 + log 2 1 − t2 (1 − t2 )2 1 + t22 1 + t22         t2 (1 − t1 ) t1 − t2 1 + t1 t2 t1 + t2 − 2Li2 + 2Li2 − 2Li2 − 2Li2 t2 − 1 1 − t2 1 + t2 1 + t2  2     2   2  2 2 2 t2 (1 + t1 ) 1 − t2 t2 − t1 t2 − 1 + Li2 + Li2 − Li2 . (31). − 2Li2 − m2 m3 m3 m21 − m22 (49) and Z∞ I2 = (m1 +m2 )2  R (m1 + m2 )2 . at each order. m23 . We define the expansion in Eq. m . 32 and (35). 6. while the second order term of F0 (n = 4. the values of the amplitudes at the pseudothreshold are given in Eqs. (50) + Li2 1 + t22 1 + t22 1 + t22 1 + t22 where √ √ m1 + m2 − m3 − m3 t1 = √ √ m1 + m2 − m3 + m3 √ √ m1 − m2 and t2 = √ √ . (8). m1 m3 m3 m3 (48)  db (m1 − m2 )(m1 + m2 − 2m3 )  log(yS ) b R b. (m1 − m3 )2 . (25) and (29). m   1 2  m1 m2 m3 log = −ζ2 + log m2 (m1 + m2 )2       m1 m1 m2 + Li2 1 − − Li2 1 − . (m2 − m3 )2 db  log(yS ) b − (m1 + m2 )2 R b. The first order terms in the pseudothreshold expansion of the master amplitudes at (n = 4) are presented in Eqs. m1 + m2 (51) The last expression was found assuming m1 > m2 > m3 . m22 . m21 .290 M. / Nuclear Physics B 581 (2000) 274–294  = ζ2 + log Z∞ I1 = (m1 +m2 )2        m3 m2 m1 m2 log − Li2 1 − − Li2 1 − . The expansion at the pseudothreshold . (m2 − m3 )2   √  b  log − 2 2 m3 R b. (16). but the analytic continuation to other regions is straightforward. (m1 − m3 )2 . p2 = −(m1 + m2 − m3 )2 ) is presented in Eq. Caffo et al. which are not given explicitly here. (m1 − m3 )2 . (15). The other pseudothresholds can be easily found by the permutation of the masses. can be easily found by solving recursively. (47). The higher order terms.

) wants to thank the Alexander von Humboldt Foundation for supporting his stay at the Theoretische TeilchenPhysik Institut of the Karlsruhe University.2 log2 m22    (52) + r2. 16  9 2 1 3 r1 = m1 x + (m1 + m2 )x 2 2 8    9 2 5 5 3 m1 + m1 m2 + m22 x + m1 m2 (m1 + m2 ) .1 log2 m21      + r1. (4) can be expanded as a single power series. where a substantial part of this work was carried out. m3 ) used in Eq.3 log m21 log m23 + r2.M. (17)     g(m1 . + 16 4 16 16 r0 = − . while at the threshold the sunrise amplitudes develop a branch point and its expansion is indeed the sum of two series [8].3 log m22 log m23 + r3. is grateful for the support and the kind hospitality to the Bologna Section of INFN and to the Department of Physics of the Bologna University. Caffo et al.R. H. Function g(m1 .2 log m21 log m22 + r1. where   11 5 147 485 2 211 485 2 3 x − (m1 + m2 )x 4 − m1 + m1 m2 + m2 x 16 32 64 16 64   81 3 839 2 839 81 m + m m2 + m1 m22 + m32 x 2 − 16 1 64 1 64 16   21 4 13 21 m1 + 5m31 m2 + m21 m22 + 5m1 m32 + m42 x − 16 2 16  7 − m1 m2 m31 + 2m21 m2 + 2m1 m22 + m32 . so that the solution of the Eq.C. Acknowledgements One of the authors (E. even if at first sight they seem to be connected by the change of sign m3 → −m3 . In fact the analytic properties of the amplitudes are different at the two points: at the pseudothreshold the sunrise amplitudes are regular.3 log2 m23 . / Nuclear Physics B 581 (2000) 274–294 291 cannot be simply deduced from the known expansion at the threshold [7] and vice versa. where the work was completed. m3 ) = r0 + r1 log m21 + r2 log m22 + r3 log m23 + r1. m2 . + 16 4 16 16  9 1 r2 = m22 x 3 + (m1 + m2 )x 2 2 8 !  5 2 5 9 2 3 m1 + m1 m2 + m2 x + m1 m2 (m1 + m2 ) . Appendix A In this appendix are given the definitions of the functions used in the text. m2 .

The definition of the coefficients used in Eq. 16 2 3   1 2 2 1 r1. 8 24 24 1 2 2 q2 = m1 m2 (m1 + m2 )3 .2 = − m2 m3 − (m1 + m2 ) m23 − m3 (m1 + m2 ) + m1 m2 .3 = − 16 2    2 × 3 m1 + m22 x 2 + 3 m31 + m21 m2 + m1 m22 + m32 x + m1 m2 (m1 + m2 )2 .292 M.3 = − m2 m3 − (m1 + m2 ) 3x 2 + 3(m1 + m2 )x + m1 (m1 + m2 ) . 8 2    1 2 1 r1. Caffo et al. 16    3 1 1 r1. 8 2   1 1 m3 − (m1 + m2 ) r3. (24):    7 1 m21 + m1 m2 + m22 − m21 log m21 q0 = m21 m22 (m1 + m2 ) 48 16    2 1 1 m1 log m21 − m2 log m22 . + m1 m2 m1 log m1 − m2 log m2 48     7 1 2 2 2 2 m1 log m1 + m2 log m2 . 16 2 3    1 2 1 r2.1 = − m21 m3 − (m1 + m2 ) m23 − m3 (m1 + m2 ) + m1 m2 .1 = m1 m2 (m1 + m2 ) − (m1 + m2 ) + 24 12 . 24   1 5m41 + 29m31 m2 + 104m21m22 + 29m1m32 + 5m42 q−1. − m22 log m22 + 16 96     1 1 1 q1 = m21 m22 (m1 + m2 )2 − (m1 + m2 ) + m1 log m21 + m2 log m22 . 8 2    3 2 1 1 r2.3 = − m1 m3 − (m1 + m2 ) 3x 2 + 3(m1 + m2 )x + m2 (m1 + m2 ) . q−1.0 = m1 m2 − 192     1 1 + m21 m21 + 5m22 log m21 + m22 5m21 + m22 log m22 48 48    1 2 2 2 . r3 = (53) and x = m3 − m1 − m2 . / Nuclear Physics B 581 (2000) 274–294   121 2 67 15 5 71 121 2 3 x + (m1 + m2 )x 4 + m1 + m1 m2 + m2 x 4 8 16 4 16   171 171 m2 m2 + m1 m22 + 3m32 x 2 + 3m31 + 16 1 16   9 4 11 3 35 11 9 + m1 + m1 m2 + m21 m22 + m1 m32 + m42 x 16 4 8 4 16 3 + m1 m2 (m1 + m2 )3 .2 = m1 m2 m3 − (m1 + m2 ) .

1 = m1 m2 − m31 + m21 m2 + m1 m22 − m32 6 24 24 6    1 2 2 + m1 m2 (m1 + m2 ) log m1 + log m2 .2 = − 3m41 + 6m31 m2 + 10m21 m22 + 6m1 m32 + 3m42 .M. 16  13 13 1 1 q0. Caffo et al. 12 1 2 5 21 67 67 21 5 q0.1 = (m1 + m2 ) − m1 − 8 24 8    9 2 9 + m1 log m21 + m22 log m22 .0 = m41 + 64  192 1 96 1 2  192 64  5 7 3 2 m1 + m1 m2 + m22 log m21 − m21 16 12 16    5 3 7 m21 + m1 m2 + m22 log m22 − m22 16 12 16    3 − (m1 + m2 )2 m21 log2 m21 + m22 log2 m22 32   1 2 2 + m1 m2 log m21 log m22 .1 = m1 + 24 3 24    3 5 2 2 3 2 − m1 m1 + m1 m2 + m2 log m21 8 4 8    3 3 2 2 5 2 − m2 m1 + m1 m2 + m2 log m22 . 8 137 3 25 137 4 m1 m2 + m21 m22 + m1 m32 + m42 q1. 4 65 1031 3 857 2 2 1031 65 m m2 + m m + m1 m32 + m42 q1. 8  193 2 1775 193 2 m1 + m1 m2 + m q2.0 = − m51 − m41 m2 − m31 m22 − m21 m32 − m1 m42 − m52 64  32 48 48 64 32  5 2 11 5 3 2 1 3 2 2 m + m m2 + m1 m2 + m2 log m1 + m1 16 1 24 1 24 16    5 1 5 3 11 2 m1 + m1 m2 + m1 m22 + m32 log m22 + m22 16 24 24 16   1 − m21 m22 (m1 + m2 ) log m21 log m22 .2 = 293 . 8 8 q−1. 8 1 q0.2 = m21 m22 (m1 + m2 ). 8 8 32 32  63 2 433 63 2 m1 m2 − m2 q2.0 = −(m1 + m2 ) 64 192 64 2   3 2  3 2 9 2 2 2 9 2 2 2 2 2 + m1 log m1 + m2 log m2 − m1 log m1 − m2 log m2 . / Nuclear Physics B 581 (2000) 274–294 1 2 2 m m (m1 + m2 )2 . 8 4 8  1 q1.

A. (54) q5. q4.A. 8 251 2 241 251 2 m1 + m1 m2 + m q3. Böhm. Nucl. see particularly pp. Remiddi.G. M. and references therein. Smirnov. p. Tkachov. Lett. R. Quantum Field Theory. / Nuclear Physics B 581 (2000) 274–294  9 q2.2 = − m21 + m22 . B 426 (1998) 95. Phys. hep-ph/9903328.A. Phys. [10] F. H. 134 (1990) 109. Phys.0 = − (m1 + m2 ). New York. Davydychev. [3] F. 1980. Smirnov. Lett.V. Berends. 4 37 79 q4. Zuber.1 = 8 1 8 2 4 1 4  3 q3. New York. C 63 (1994) 227. hep-ph/9712209. Phys. B 192 (1981) 159. [9] O. [4] F. Tarasov. Itzykson.0 = − 16 2 References [1] M. S. 284 and 365. Caffo. Commun. B 100 (1981) 65.B. [7] A. Caffo et al. Scharf. Z. Ordinary Differential Equations. Beneke.0 = 64 24 64 2  1    1 3 3 + m21 log m21 + m22 log m22 − m21 log2 m21 − m22 log2 m22 . M.1 = . 2 2 16 16   153 2 153 2 3 2 3 m + 39m1 m2 + m − m log m21 − m22 log m22 . E. Ussyukina. Czy˙z. N. hep-ph/9703319. Section 6-3. [11] K. Math. q3. V.2 = (m1 + m2 ) m21 + m22 . Davydychev. [8] E. Nucl. [2] C. Tkachov. [5] V.L.1 = − (m1 + m2 ). Nuovo Cimento A 111 (1998) 365.I. B 502 (1997) 455. 1956. [6] M. Buza. V. Phys. A. Chetyrkin. Berends.I. hep-ph/9711391. Ince. B 554 (1999) 391.294 M. hep-th/9805118.V. Phys. B 522 (1998) 321. Phys. Dover Publications. Nucl. F. J. Smirnov.A.A. Nucl. Laporta. 32 4 11 15 q5. McGraw-Hill. . 301.I. Phys.V.

kek.  2000 Elsevier Science B. Japan Received 28 January 2000. 11. We also consider the description of D1-branes as local excitations of gauge theory on D5-branes. accepted 13 March 2000 Abstract We consider IIB matrix model with D1–D5-brane backgrounds.b.V. Especially the spectrum of the zero modes in the off-diagonal parts is examined.3 2 1 3 ( 0 0 ) 0 0 1 7 1 . It is a large N reduced model [7–11] of ten-dimensional supersymmetric Yang–Mills theory. Tokyo Institute of Technology. It is crucial that this model has N = 2 supersymmetry because it guarantees the existence of gravitons. Japan b Department of Physics. Notable point is that supersymmetric gauge theory can be obtained on their world-volume as their low energy effective theory. Tokyo 152.nl/locate/npe IIB matrix model with D1–D5 backgrounds Yusuke Kimura a. noncommutative Yang–Mills theory has been studied in many situations. PII: S 0 5 5 0 . Using the fact that noncommutative gauge theory on the D-branes can be obtained as twisted reduced model in IIB matrix model. Oh-okayama. Recently.kek. IIB matrix model is one of these proposals [2].elsevier.2 a High Energy Accelerator Research Organization (KEK). Yoshihisa Kitazawa a.25.6]. It first appeared within the framework of toroidal compactification of matrix theory [12].1 .Lm.jp 0550-3213/00/$ – see front matter  2000 Elsevier Science B. These proposals are based on the developments of D-branes. Introduction Several kinds of matrix model have been proposed [1–4] to study the nonperturbative aspects of string theory or M-theory. PACS: 11.V. Tsukuba.phys.-w.jp or kimura@th.-q. It is postulated that it gives the constructive definition of type IIB superstring theory.10. we study two-dimensional gauge theory on D1-branes and D5-branes. 11. All rights reserved. Matrix model. Ibaraki 305-0801. 11. Meguro-ku.Pb Keywords: String theory. The idea of matrix models is that supersymmetric gauge theory can describe string or M-theory.15.titech. Noncommutative gauge theory. Relation to supergravity solution is also discussed.ac. D1–D5 system 1. The world-volume theory on D-branes with NS–NS two-form background is described 1 kimuray@ccthmail. All rights reserved.1 .30.jp 2 kitazawa@post. They have been shown to play a fundamental role in nonperturbative string theory [5.Nuclear Physics B 581 (2000) 295–308 www.

we consider D1–D5 backgrounds. Kitazawa / Nuclear Physics B 581 (2000) 295–308 by noncommutative Yang–Mills theory [13]. Section 5 is devoted to conclusions and discussions.296 Y. we study IIB matrix model with D1–D5 backgrounds. B-field background effect changes the behaviors of gauge theory. It is well known that gauge theory is realized in the world-volume of Dbranes as their low energy effective theory. . For their description. The relation between supergravity and gauge theory is formulated as AdS/CFT correspondence [18]. We examine the spectrum of this matrix model. In IIB matrix model.16]. The organization of this paper is as follows. In this paper. are described by the cluster property in matrix model. Kimura. This description gives the description of D1-branes bounded to D5-branes. The D1–D5 background is an interesting configuration since it is used by Strominger and Vafa [17] in order to study the microscopic interpretation of the black hole entropy. The counting of the zero modes is also discussed. flat spacetime. Comparison with string theory results is mentioned. Further researches about noncommutative Yang–Mills were done in [16] using the property of twisted reduced model. In Section 3. It was shown [14] that in matrix model picture noncommutative Yang–Mills theory is equivalent to twisted reduced model [15]. AdS5 × S5 background is investigated in [14. It is further studied by Maldacena [18] as AdS/CFT correspondence. The remarkable point is that there is a relation between the coordinate space and the momentum space. The upper left part and the lower right part represent D1-branes and D5-branes. respectively. gauge theory is realized as twisted reduced model. Twisted reduced model is defined by the expansion around a D-brane-like background in IIB matrix model. In string theory picture. The spectrum of open strings connecting D1-branes and D5-branes in string theory is discussed in [13]. 20]. Interactions in very far region from branes. It was pointed out that there was a crossover at noncommutative scale. In near horizon region (so-called AdS region) gauge theory description can be used. Y. off-diagonal parts represent the states appearing as open strings connecting D1-branes and D5-branes. In Section 4. We also consider another description of D1-branes as local excitations of gauge theory on the world-volume of D5-branes. In Section 2. we divide matrices into four parts. A noncommutative background is a D-brane-like background which is a solution of equation of motion and preserves a part of supersymmetry. The remaining parts represent the interactions of them. Noncommutative Yang–Mills is equivalent to twisted reduced model. while supergravity solutions are also changed [19. The behaviors of Wilson loop operators in the large momentum scale region and the small momentum scale region are studied. twisted reduced model is obtained as expanding the model around noncommutative backgrounds. We consider AdS3 × S3 × M4 background in matrix model. In IIB matrix model. we mention the relation to supergravity solutions. In small momentum scale region Wilson loops are the ones in ordinary gauge theory while in large momentum scale region Wilson loops become open string like objects. we review the relation between noncommutative Yang–Mills and IIB matrix model.

Aν ] = 0. Aν ]Γ µν ε. Y. half of the supersymmetry is preserved in this background. Noncommutative Yang–Mills in IIB matrix model In this section. We expand the theory around the following classical solution. The volume of the phase space is Vp = n(2π) . pˆ = iB µν . Aν ] = c-number is an interesting solution because it is a BPS background.21] and its relation to noncommutative Yang–Mills which is equal to twisted reduced model [14. (7) we can obtain N = 2 supersymmetry algebra:  (j ) (j ) δ˜ε(i) δ˜ξ − δ˜ξ δ˜ε(i) ψ = 0. This model is the large N reduced model of ten-dimensional N = 1 supersymmetric Yang–Mills theory. (4) Aµ = 0. we can regard the above symmetry as N = 2 supersymmetry. and Aµ and ψ are N × N hermitian matrices. IIB matrix model is defined by the following action   1 1 1 ¯ µ [Aµ . This corresponds to a D-brane background. pˆ satisfy the canonical commutation relations and ˜ √ d/2 det B. That is.  δ˜(2) = i δ (1) − δ (2) . (8) The equation of motion of (1) with ψ = 0 is   Aµ . [Aµ . This is based on the observation that a large N gauge theory is equivalent to its reduced model [7–11]. dimensional phase space. Aν ][Aµ .16].  (j ) (j ) δ˜ε(i) δ˜ξ − δ˜ξ δ˜ε(i) Aµ = 2i ε¯ Γ µ ξ δij . Aν ] + ψΓ (1) S = − 2 Tr 4 2 g Here ψ is a ten-dimensional Majorana–Weyl spinor field. (5) (2) and translation symmetry: δAµ = cµ 1. This model has the following symmetries: δ (1) ψ = 2i [Aµ . Kitazawa / Nuclear Physics B 581 (2000) 295–308 297 2. (6) If we interpret the eigenvalues of Aµ as the spacetime coordinates. µ (2) (3) and δ δ (2) ψ = ξ.  µ ν (10) pˆ . δ (1) Aµ = i ε¯ Γ ψ. This is a solution of (9) and is a ˜ BPS background. we review IIB matrix model [2. ψ] .Y. [Aµ . where Bµν is anti-symmetric tensor and c-number. Kimura. We take a linear combination of δ (1) and δ (2) as δ˜(1) = δ (1) + δ (2) . We assume the rank of Bµν to be d˜ and define its inverse C µν in dµ ˜ span the ddimensional subspace. (9) The solution in which [Aµ .

k µ is quantized in the unit of kµmin = Λ/n2/d = λ/n1/d . The range of kµ is ˜ ˜ restricted as −n1/d λ/2 6 kµ 6 n1/d λ/2. we obtain the following map. Y. ˜ ˜ The volume of one quantum in this phase space is Λd /n = λd where λ is the spacing of the quanta. The hermiticity requires that a˜ µ∗ (k) = a˜ µ (−k) and ψ˜ µ∗ (k) = ψ˜ µ (−k). noncommutative scale.298 Y. Kitazawa / Nuclear Physics B 581 (2000) 295–308 We expand Aµ = pˆ µ + aˆ µ and Fourier-decompose aˆ µ and ψˆ as X  a˜ µ (k) exp iC µν kµ pˆ ν . where ? is the star product defined as follows. (11) (12) k exp(iC µν kµ pˆ ν ) is the eigenstate of Pµ = [pˆ µ . ·] with eigenvalue kµ . . Kimura. say. Let Λ be the extension of each pˆ µ . aˆ µ = k ψˆ = X  ˜ ψ(k) exp iC µν kµ pˆ ν . is related to λ ˜ ˜ as B = λ2 /2π . aˆ µ → aµ (x) = (13) k Under this map. aˆ bˆ → a(x) ? b(x). Consider the map from a matrix to a function as X  a˜ µ (k) exp ikµ x µ . B. which is the component of Bµν .

  µν .

a(x + ξ )b(x + η). ∂2 iC .

.

23]. the adjoint operator of pˆ µ + aˆ µ is mapped to the covariant derivative:   (17) pˆ µ + aˆ µ . Noncommutative Yang–Mills action can be obtained by applying the above rules. This relation says that the coordinate space is related to the momentum space. D1–D5-brane solutions are constructed by preparing the following matrices [2. oˆ → 1i ∂µ o(x) + aµ (x) ? o(x) − o(x) ? aµ (x) ≡ [Dµ . We use the coordinate space description instead of the dual momentum space. 3. we consider IIB matrix model with D1– D5 backgrounds. the coordinate space is also embedded in the matrices of twisted reduced model through the relation xˆ µ = C µν pˆ ν .22. Although we have discussed the momentum space. Tr[a] ˆ = det B 2π (14) (15) (16) Using these rules. The gauge theory of D1–D5 system in IIB matrix model Following the rules of the previous section. This relation reminds us T-duality. a(x) ? b(x) ≡ exp µ ν 2 ∂ξ ∂η ξ =η=0 Tr over matrices can be mapped on the integration over functions as  d/2 ˜ Z √ 1 ˜ d d x a(x). . o(x)].

Kimura. D1-brane and D5-brane are constituted of 6 2 and nD5 (= n1 n2 n3 ) quanta of the volume lNC . 1. µ = 2. ˆ p] ˆ = i. 2πn3 299 (18) where [q. 0 0 ! aˆ µ cˆµ . We will use the indices α. Y. 9. The remaining parts of matrices represent the interactions of them. bb and cˆµ with √ and bα (x) can (1/ B )aa (x). where ˆ aα (x) √ be interpreted as√gauge fields. β as the indices running over the direction parallel to the branes and a. 3. µ = 6. We can consider multiple branes by replacing xˆ → xˆ ⊗ 1Q1 or xˆ ⊗ 1Q5 . Q1 and Q5 are the number of D1-branes and D5-branes. 8. We replace aˆ α and bˆα with C αβ aβ (x) and C αβ bβ (x) in our correspondence (13). . (20) We align a D1-brane along the 0. 5. [qˆ 00 . In IIB matrix where we set C 01 = C 23 = C 45 = 2π/B = lNC model. b as the indices over the direction transverse to the branes. 4. uµ are the positions of D1-branes in the 6. 7. l = 2π/λ is the respectively. We call this system D1–D5 3 This is the configuration considered in [13]. 1. xˆ µ satisfy the commutation relation such as  µ ν (19) xˆ . pˆ 0 ] = i.   0 xˆµ       0 0 cl (21) . 5 directions:   xˆ µ 0   . Aµ = Acl µ + Xµ . 9 directions. respectively. They are equivalent to the branes in the presence of a constant NS–NS two-form background bµν . nD1 (= n1 ) quanta of the volume lNC 2 and T L L L L L = n l 6 . 2 = (2π/λ)2 for simplicity. where T L1 = nD1 lNC 1 2 3 4 5 D5 NC NC noncommutative length scale. 2πn3 L2 xˆ 2 = √ qˆ 0 . Kitazawa / Nuclear Physics B 581 (2000) 295–308 T q. and the fluctuation is denoted by Xµ . xˆ = −iC µν . 3. We also replace aˆ a . pˆ 00 ] = i. The upper left part and the lower right part of matrices represent a D1-brane and a D5-brane. (1/ B )bb (x) and (1/ B )cµ (x). 2πn2 L5 xˆ 5 = √ pˆ 00 . (22) Xµ = cˆµ† bˆµ These Acl µ satisfy the equation of motion (9). ˆ 2πn1 L3 pˆ 0 . 3 We expand the model around the classical background. 4.Y. 1 directions and a D5-brane along the 0. xˆ 3 = √ 2πn2 xˆ 0 = √ L1 xˆ 1 = √ p. ˆ 2πn1 L4 xˆ 4 = √ qˆ 00 . µ = 0. 2. 7. [qˆ 0 . the D-brane is interpreted as not pure one but one with nonvanishing gauge field strength Bµν [24]. respectively. 8. Aµ =  0 xˆµ         uµ 0   . Acl µ .

D + − x tr D d 3 2 3 2 16π g 2π g B    e e (25) + 2B Dα ba Dα ba + B 2 [ba . 3. 5. cµ are Q1 × Q5 matrices and transform as the product of the fundamental representation of U (Q1 ) and the anti-fundamental representation of U (Q5 ). Other cases can be done in the same way. The bosonic part action with cˆµ = 0 can be easily calculated and become Z 1 T LQ1 − d 2 x tr [Dα . where Ωµ are unitary matrices and Rµ are compactification radii. aµ are Q1 × Q1 hermitian matrices and transform as adjoints of U (Q1 ). D eβ D eβ eα . 4. we can obtain the case that D1-branes are separated in x2 direction. bb ] . We examine the condition whether this background is supersymmetric: δ (1) ψ = 2i [Aµ . cµ† are the complex conjugate of cµ . Ωµ do not commute for the noncommutative torus [12]. (23) We find that the following conditions are needed for this configuration to be supersymmetric: C 01 Γ 01 ε = ξ. ? where Dα is the covariant derivative constructed by the gauge fields aα (α = 0.300 Y. 4) supersymmetry in two dimensions. ab ][aa . tr is D taken over the color indices. ? means that products are not usual ones but the ones defined by (15). Y. 1) while eα is the covariant derivative constructed by the gauge fields bα (α = 0. Kimura. Kitazawa / Nuclear Physics B 581 (2000) 295–308 system. 4 To describe the compact space. This makes aµ field massive. 2. 5). We have just considered the situation that D1-branes are overlapping and also that (i) D5-branes are overlapping. 1. cµ and cµ† are two-dimensional fields while bµ are six-dimensional fields. . Dβ ][Dα . 3. We now consider twisted reduced model action with the D1–D5 backgrounds. Aν ]Γ µν ε   01 01 0 C Γ ε. 4. This system is described by two-dimensional 4 supersymmetric U (Q1 ) × U (Q5 ) gauge theory in low energy limit. µ. By replacing left upper part of Acl 2 with w2 ⊗ 1Q1 (i = color index) which is the positions of D1-branes in the x2 direction. This theory has N = (4. aµ . = 0 C 01 Γ 01 + C 23 Γ 23 + C 45 Γ 45 δ (2) ψ = ξ. ab ] ? Z    3BT L5 Q5 1 6 eα . we impose the following condition for the matrices: A + 2π R δ µ µ µν = Ων Aµ Ων† . Dβ ] S= 4πg 2 B 2 2πB 3 g 2  + 2B(Dα aa )(Dα aa ) + B 2 [aa . Γ 2 Γ 3 Γ 4 Γ 5 ε = ε. (24) Therefore one fourth of the supersymmetry are preserved in this configuration. bb ][ba . ν = 2. bµ are Q5 ×Q5 hermitian matrices and transform as adjoints of U (Q5 ).

φ2 = √1 (c2 2 − ic3 ). 5) string is the open string with one end on a D1-brane and the other end on a D5-brane.Y. Therefore the energy eigenvalue of xˆ22 + xˆ32 is given by the Landau level. the kinetic and mass terms are given by Z 9 1 X X  1 2 x tr ∂α cµ ∂α cµ† + ∂α cµ† ∂α cµ ? d S=− 2 2 8πB g − 1 8πg 2 +i Z d 2 x tr 1 8πBg 2 P9 Z α=0 µ=0 9 n 5 XX o   cµ xˆα2 cµ† + U 2 cµ cµ† + cµ† cµ xˆα2 + U 2 cµ† cµ ? α=2 µ=0  d 2 x tr c2 c3† − c2† c3 − c3 c2† + c3† c2 + c4 c5† − c4† c5 − c5 c4† + c5† c4 ? . We ignore the interaction terms. 3. xˆ22 + xˆ 32 is the Hamiltonian of the charged particle moving on the two-plane in a uniform magnetic field perpendicular to the plane (remember (19)).) and similarly for xˆ42 + xˆ52 .4. Kimura. . Y. 2(N + 12 )/B (N = 0. ? where φ1 = √1 (c2 2 + ic3 ). 4. We divide matrices into four parts in the same way as the bosonic part. We can say that only φ2 and φ4 can be massless scalar at the ground state (M = 0 or N = 0) and U 2 = 0 while φ1 . . (26) where U 2 ≡ µ=6 u2µ . 1. ? Z h   1 † 2 2 d x tr (2N + 2)/B + U φ3 φ3 + φ3† φ3 + 2 2 8πg B i + (2N/B + U 2 ) φ4 φ4† + φ4† φ4 . We next examine the fermionic part. ? Z X h  1 2 (2M + 1)/B + (2N + 1)/B + U 2 d x tr − 2 8πg µ6=2. It corresponds to the states appearing in (1. . 1) string 5 in string theory picture. ! ψˆ 1 ψˆ 3 ˆ . After the gauge fixing [2]. (28) ψ= ψˆ † ψˆ 2 3 5 (1. Now we can rewrite mass terms in (26) as Z h   1 2 x tr (2M + 2)/B + U 2 φ1 φ1† + φ1† φ1 d − 2 8πg i + (2M/B + U 2 ) φ2 φ2† + φ2† φ2 . φ3 = √1 (c4 2 + ic5 ) and φ4 = U2 √1 (c4 2 − ic5 ).5 i (27) × cµ cµ† + cµ† cµ . . is the distance between D1-branes and D5-branes. Only four real degrees of freedom are massless degrees of freedom. 5) are always massive because of the zero point energy. Fermionic background is set to zero. 5) and (5.3. 2. Kitazawa / Nuclear Physics B 581 (2000) 295–308 301 We next consider cµ and cµ† fields.φ3 and cµ (µ 6= 2. These mass terms mean that a field appearing in the string connecting two branes has a mass proportional to the distance between two branes in string picture. Mass terms are given in second and third lines in (26).

We consider the massless conditions for ψ3 and ψ3† . Kitazawa / Nuclear Physics B 581 (2000) 295–308 Kinetic and mass terms are calculated as follows. 7. 5. β = 0. we can make annihilation–creation operators: q q   † ≡ B2 xˆ 3 − i xˆ 2 . a23 (31) (32) Annihilation–creation operators for xˆ 4 and xˆ 5 also can be made in the same way. 9. 3. γ = 2. 16π 3 g 2 (29) where Γeβ = Γδ εδβ . 2. 4. Z  1 d 2 x tr ψ¯ 1 Γeα ∂α ψ1 ? S=− 2 4πg Z  1 ¯ 2 − x tr d ψ¯ 3 Γeα ∂α ψ3† + ψ3† Γeα ∂α ψ3 4πg 2    ¯ + B ψ¯ 3 Γγ xˆ γ − uτ Γτ ψ3† + B ψ3† Γγ xˆ γ + uτ Γτ ψ3 ? Z  B2 6 ¯ e − d x tr ψ2 Γβ ∂β ψ2 ? . . 1. 1. From the commutation relation (19). So the following conditions are needed for ψ3 and ψ3† to be massless: a23 |ψ30 i = 0. and α = 0. 8. a23 a23 ≡ B2 xˆ 3 + i xˆ 2 . Mass terms are given by the fourth and fifth terms of (29). Y.  †  = 1. and τ = 6. We rewrite Γγ xˆ γ = 0 as   (Γ2 + iΓ3 ) xˆ 2 − i xˆ 3 + (Γ2 − iΓ3 ) xˆ 2 + i xˆ 3   (30) + (Γ4 + iΓ5 ) xˆ 4 − i xˆ 5 + (Γ4 − iΓ5 ) xˆ 4 + i xˆ 5 = 0. 4. a23 . 5. 3. Therefore the massless conditions for ψ3 and ψ3† are Γγ xˆ γ ψ3 = 0 and Γγ xˆ γ ψ3† = 0 and uτ = 0. Kimura.302 Y.

(Γ2 + iΓ3). † ψ30 = 0.

.a23 a45 |ψ30 i = 0.

(Γ4 + iΓ5 ). † ψ30 = 0.

There are also the contribution from (5..a45 (33) Last two conditions say that one fourth of ψ3 components survive. In the R-sector. There are four periodic world-sheet fermions in the NN and DD directions (0. the zero point energy is zero. Both cµ and ψ3 transform as the product of the fundamental of U (Q1 ) and anti-fundamental of U (Q5 ). . We compare these results with string theory results. two of the four ground states survive. These zero modes generate 24/2 = 4 ground states. they are same with the massless degrees of freedom of cµ . 5)-string in the presence of bµν field by the Ramond–Neveu–Schwarz formalism. because world-sheet bosons and world-sheet fermions have the same moding. We examine the spectrum of (1. these are super-partners. Superstring theory is usually formulated in the Ramond–Neveu–Schwarz formalism. In other words. while matrix model is formulated in the Green– Schwarz formalism. The massless spectrum is not affected by a bµν field. We have seen that massless mode of cµ has real four degrees of freedom. 1) string. So. there are four zero modes. 1 directions are not counted). i. After imposing the GSO projection. Off-diagonal elements of matrices correspond to (1. Since the massless physical degrees of freedom of ψ3 are four.e. 1) string. 5) or (5. the massless mode of ψ3 fermion has eight components.

Y. Y. 2. The equation of motion in IIB matrix model is given by (9). the massless spectrum is affected by the bµν field. D eν = 0. a D5-brane is regarded as to be constituted of infinitely D1-branes or anti-D-branes. 5. we conclude that the number of massless modes for the off-diagonal elements in matrix model matches the number of massless modes for (1. 4. appear from the NS-sector after imposing the GSO projection. µ = 0. From these facts. We now look at the Chern–Simon coupling in D5-brane action: Z tr C2 ∧ F2 ∧ F2 . Kitazawa / Nuclear Physics B 581 (2000) 295–308 303 Therefore there appear four massless modes. the spectrums are different in these cases. 5. Same discussions are applied to the D1–D5 system. tachyon state appears. The zero point energy has the dependence of the value of the bµν field. For large b23 and b45 . F ∧ F plays the role of D1-brane charge. 5) and (5. So this system is unstable and not supersymmetric. We are interested in D1–D5 bound states because these states are associated with the dynamics of the black hole [17]. Kimura. 66 where C2 is the R–R two form field and F2 is the gauge field strength. Therefore Q1 D1-branes embedded in D5-brane are constructed as Q1 instantons in U (Q5 ) gauge theory [25. 9. 1. total four massless modes. 3. (From the open string stretching between D-string and anti-D-string. 8. lower-dimensional Dbranes embedded in higher-dimensional D-branes are expressed as the magnetic flux on higher-dimensional D-branes [27]. In the NS-sector. (35) Xµ = (3)† bˆµ bˆµ(2) The size of the right lower part of matrices are 2 × 2. In general. 3. In this coupling. 0 xˆµ (34) Acl µ=  0. We now consider the description of D1-branes embedded in D5-branes. 6. 2. from the R-sector. The size of the left upper part of matrices are (Q5 − 2) × (Q5 − 2). After the replacements from matrices to functions. In this case. According to [13]. 4. D D µ. It can be shown that the instantons of U (Q5 ) gauge theory on the D5-branes carry R–R two-form charge and D1-branes considered in the previous part of this section correspond to the zero size limit of the instantons.  ˆ (1) ˆ (3)  bµ bµ . which correspond to the massless modes of ψ3 . ν = 0. which correspond to the massless modes of φ2 and φ4 .26]. 1) string. The spectrum is discussed in [13] about the case of the D0–D4 system. 7. µ = 6. the equation of motion of this gauge theory becomes    eµ . eµ . 1. We consider U (Q5 ) gauge theory on D5-branes with a self-dual gauge field strength on T 4 in matrix model:    xˆ µ 0 .) The case which is considered in this paper is the former case in which this system is supersymmetric. (36) .

3. we embed an (anti-)instanton into SU (2) part (right lower part). (2) eρα F = 0. respectively. supersymmetry transformations (2) and (4) are given as follows: i δ (1) ψ = [Aµ . Therefore we also need the following condition to obtain pure D1–D5 bound states. no D3-branes: Z 1 e(2) = 0. σ = 0. 1. In other words. e(2) . ∓ correspond to self-dual and anti-self-dual configurations. 3. Therefore (2(Q5 − 2)Q1 + 4Q1 )/2 = Q1 Q5 chiral 6 Instantons on non-commutative R4 space are constructed in [28]. The difference of the chiral fermions is given by αV Q1 [29]. tr F (39) 2π T4 In these configurations. we set the following condition: Z 1 e(2) . τ run over 2. 4. (40) where repeated indices α. D 0 C µν + iC µτ C νρ D 2 1 = C µν Γ µν ε + 2 0 0 0 αγ βδ iC C [Dγ . where αV is the Dynkin index depending on the representation and Q1 is defined in (38). αV is 1 and 4. In these cases. 5 and ρ. Aν ]Γ µν ε 2  1 C µν Γ µν 0   = eρ(2) Γ µν ε eτ(2) . Therefore to express Q1 D1-branes on D5branes. (41) We study the zero mode. β = 2. (37) where α. the winding number of instanton represents the D1-brane charge. Y.304 Y. It is known that the second Chern number is equal to the difference in the number of the chiral fermion zero modes whose chiralities are positive and negative (the index theorem). Γ 2 Γ 3 Γ 4 Γ 5 ε = ±ε. We expect that chiral fermions appear from the off-diagonal parts and the right lower part. i. Following constraints are needed for this configuration to be supersymmetric: 1 µν µν ε 2C Γ = ξ. Kimura. 5.e. F (2) constructed by the gauge field b . Self-dual and anti-self-dual configurations are referred to instantons and anti-instantons.. γ . e(2) = ± ∗ F F αβ αβ (2) eρσ F = 0. e(2) ∧ F tr (38) F Q1 = 8π 2 T4 Also. respectively. respectively. D5-branes carry D3-branes charge. The star product e(2) is the gauge field strength which is defined in (15) is used in these equations. These parts transform as the fundamental and adjoints of SU(2). β. . In D-brane picture. ∗ means the dual tensor on T 4 . 4. Kitazawa / Nuclear Physics B 581 (2000) 295–308 We analyze the situation that only gauge fields bµ(2) have nontrivial classical solutions. 1∓Γ 2 Γ 3 Γ 4 Γ 5 2 !  Γ αβ ε. Dδ ] δ (2) ψ = ξ. The instanton solution 6 can be constructed as follows. respectively.

Kimura. In this T-duality transformation. b45 = tan θf1 f5−1 h3 . . the zero mode of ψ3 satisfies Γ2 Γ3 Γ4 Γ5 ψ30 = −ψ30 . x4 . The relevant supergravity solution to matrix model is supergravity solution with bµν field since the D-brane in matrix model is not pure one but one with bµν background. we would like to study the supergravity solution in IIB matrix model. This gives D0–D4 bound states. x3 and x4 . Relation to supergravity solution In this section. This implies that ψ30 has negative chirality as a SO(4) (the rotation of x2 . we obtain the following solution: 7   −1/2 −1/2 1/2 1/2 ds 2 = f1 f5 h1 dx02 + dx12 + f1 f5 dr 2 + r 2 dΩ32 (42) h  i 1/2 −1/2 2 2 2 2 h2 dx2 + dx3 + h3 dx4 + dx5 . x1 )-plane with a rotation angle θ .30. b23 = tan θf1 f5−1 h2 . This gives black string solution if Q1 and Q5 are large. f1. x3 . Then we rotate the (x0 . x3 . f5 2 α 0 R1. We notice that this chirality is opposite to the spinor which satisfy the condition of unbroken supersymmetry (see (41)). The supergravity solution of D1–D5 system with the bµν fields which have nonzero components along all the directions parallel to the branes can be obtained likewise as in [20. This solution is considered in string frame. x5 . +f1 f5 f1 h1 h2 h3 . Now we consider making the T-duality transformation in the x1 (this is a rotated coordinate) direction. x1 directions.5 .31]. Y. 7 The notation is referred to [20]. We apply T-duality transformation in the x1 direction in a usual D1–D5 system. θ is a parameter which represents noncommutativity. The degrees of freedom of these zero modes is associated with the dynamics of the black hole entropy [17]. Kitazawa / Nuclear Physics B 581 (2000) 295–308 305 fermion zero modes appear from the index theorem. directions are compactified on a torus T 4 . x5 ) Weyl spinor. −1 −1 2 2 h−1 2 = h3 = f1 f5 sin θ + cos θ. b01 = tan θ is induced along the x0 . x1 directions. This gives the mixed boundary conditions in the x0 .5 = 1 + r b01 = tan θf1−1 f5−1 h1 .Y. x4 . From the condition (33). The reason is that these zero modes are Goldstone modes. 4. The procedure is as follows. e2φ = g 2 where −1 −1 2 2 h−1 1 = f1 f5 sin θ + cos θ. The alignment of D1-branes and D5-branes is same with the previous section. x5 directions. x2 . By applying the same procedure in the x2 . We will work out in Euclidean space.

In IIB matrix model.1 = b˜ . The string coupling decrease as the energy scale grows. This is the same behavior with one in AdS5 × S5 which behaves like the twisted reduced model [14]. b˜ r = α 0 R 2 u. Instanton configuration in D5-brane is discussed at the last part in the previous section.4. . In the spirit of this relation. the interactions are described as the interactions between diagonal blocks using the cluster property [2.4. We can ignore the bµν effect (i.4. that is. Kimura. The behavior of the dilaton is 1 1 (45) eφ ∼ gˆ 2 2 = gˆ ˜bR 2 u2 a u at large u. dilatons and axions can be seen similar to D3–D(−1) case [14].5 = (α 0 )1/2x˜2.e. u and g˜ stay fixed and R 2 = R1 R5 . We consider the instanton and (anti-)instanton pair. The string coupling constant depends on the energy scale. √ (46) hρ(x)i ∼ g e−φ(x). This is in accordance with the fact that the relevant noncommutative Yang–Mills theory is two-dimensional [32]. In the IR region. Kitazawa / Nuclear Physics B 581 (2000) 295–308 We consider the region which is very far from branes.3.5. (44) ˜ 1 /R5 )2 is the string coupling in the IR region. g = α 0 g.. α 0 → 0. noncommutativity) in this region.3.5 . The spacetime become the flat spacetime. In such a region. Y. and the situation that their distance is very far. ˜ ˜ x˜0. r is very large. α0 tan θ = x˜0.306 Y.1 . α0 (43) x2.1 . graviton exchange can be seen in the instanton and anti-instanton case [14]. Solution dual to two-dimensional gauge theory on D1–D5-branes with bµν field is obtained by taking the near horizon limit in the solution (42).14] in matrix model. From the perspective of where gˆ = g˜ b(R gauge theory. . x0. The exchange of gravitons. where b.    dU 2 R1 ds 2 2 2ˆ 2 2 2 = R u h d x˜ 0 + d x˜ 1 + 2 + dΩ3 + d x˜ n d x˜n . α0 U R5 1 ˜ 2. x˜2. We embed D1–D5 bound states in the left upper part of matrices and the other D1–D5 bound states in the left upper part of matrices. radial coordinate u is interpreted as energy scale.3. It is conjectured by Maldacena [18] that two-dimensional CFT describing the Higgs branch of D1–D5 system on M4 is dual to IIB string on AdS3 × S3 × M4 . each is in D5-brane. On the other hand. a 2 = bR hˆ = 1 + a 4 u4 ˆ e2φ = gˆ 2 h. curved spacetime is thought to be given by assuming the following eigenvalue distribution [33]. this metric represents AdS3 × S3 × M4 . The interaction potential between instantons vanish because this system is BPS saturated. supergravity solution dual to noncommutative Yang–Mills is studied by Maldacena and Russo [20]. Hashimoto and Itzhaki [19].

We also considered the anti-instanton solution which is interpreted as and an anti-D1-brane embedded in a D5-brane. 5/2 −1 1/2 5/2 −1 √ −φ(x) ge = α 0 R1 R5 gˆ uhˆ = α 0 R1 R5 gˆ  u2 1 + a 4 u4 1/2 . We can check it as follows. This fact coincides with string theory result. It was also shown that D-instantons couple to gravity by considering the interactions between diagonal blocks. so this system can be stable as the bound state. . We obtain −C 01 σ3 Γ 01 ε = 12×2 ξ instead of the first condition in (24). Kimura. in six-dimensional gauge theory on D5-branes. we can calculate as follows. Y. We showed that only four degrees of freedom can be massless modes. Discussions In this paper. we considered D1-brane solutions as local excitations of gauge theory on D5-branes. instantons. This is invariant under u → 1/a 2 u as expected.. boundary). Twisted reduced model is obtained by the expansion around the noncommutative background. 5) or (5. The topological charge of the instanton is related to the existence of the chiral fermion zero modes by the index theorem. D1-branes embedded in D5-branes are expressed as nontrivial solutions. In matrix model. They correspond to the state which appear in (1. D-instantons were constructed as local excitations of gauge theory on D3-branes. In Section 4. 1) open string. Using the equivalence between twisted reduced model and noncommutative Yang–Mills. Therefore the interactions between D1-branes and (anti-)D1-branes can be viewed as the interactions between infinitely many D-instantons and (anti-)D-instantons. (48) in only AdS3 part. (47) i In this case. we studied the two-dimensional gauge theory on D1–D5 system. We also considered the zero modes of the instanton configuration. In this paper. we have considered D1–D5 system in IIB matrix model. However. It is impossible to find the spinors which satisfy this condition. Kitazawa / Nuclear Physics B 581 (2000) 295–308 where ρ(x) = X 307 δ (10)(x − x i ). We can understand that antiD1-brane and D5-brane system is unstable because this system is not supersymmetric. These degrees of freedom play an important role in the black hole physics. The anti-D1-brane background is constructed by replacing xˆ1 with −xˆ1 in (21). an anti-D1brane can dissolve in a D5-brane and it is described as anti-instanton in the gauge theory on D5-brane. we discussed the supergravity solution which is dual to the gauge theory of the D1–D5 system in matrix model. We discussed the spectrum of the fields. especially the fields which appear in the off-diagonal parts of matrices. 5. We have shown that this anti-instanton solution is supersymmetric.e. This eigenvalue distribution has a peak at u = 1/a while ordinary AdS has a peak at u = ∞ (i.Y. D1–D5 background is supersymmetric and one fourth of the supersymmetry survive. It was shown that there were Q1 Q5 SO(4) Weyl fermions as the zero modes. D1-brane is expressed by infinitely many D-instantons. In [14].

Phys. L.G. Theor. hep-th/9610043. Prog. 113B (1982) 47. A. Costa. D 27 (1983) 2397. K. Y. [24] I. Int. B 379 (1996) 99. Mod. Phys. Michaud. [3] R. A 12 (1997) 1447. [14] H. hep-th/9705120. Phys. Lett. Ishibashi. hep-th/9611050. Phys. . [20] J. 99 (1998) 129. 64. [23] A. Bhanot. S. M. Myers. [29] A. Phys. 223. Phys. JHEP 9909 (1999) 025. Phys. Kawai. Fayyazuddin. N. [22] I. [9] D. Nucl. Kitazawa. S. [16] N. Theor. hepth/9711200. Nucl. B 490 (1997) 121. Y. hep-th/9910004. Schwarz. J. Phys. 1998. hep-th/9701168.J. Kawai. [18] J. Kitazawa. Vafa.S. Das. 48 (1982) 1063.R. Verlinde. Hashimoto. Tsuchiya. B 400 (1997) 43. Kawai. Commun. JHEP 9811 (1998) 007. Kitazawa. Ishibashi. Suppl.308 Y. [21] H. Nucl. Lett. Phys. Raamsdonk. [2] N. Phys. hep-th/9708123. A. D. Phys. Polchinski. Iso.A. Math.G. H. hep-th/9908038. R. D 55 (1997) 6438. H. [8] G. String Theory. A. B 511 (1998) 629. hep-th/9512077. Gross. 198 (1998) 689. G. [11] S. [15] A. J. hep-th/9908142. II. H. 2 (1998) 231. Phys. Lett. hep-th/9912072. Breckenridge. A.A. H. Rev. 75 (1999) 4724. hep-th/9807185. Kitazawa. Rev. Maldacena. B 500 (1997) 43. Nucl. hepth/9612115. Connes. Wadia. Iso. [33] S. Itoyama. hep-th/9611174. C. Aoki. Witten. [27] W. Phys. Kitazawa. Y. Schwarz. hep-th/9802068. Maldacena. Lett. Okawa. Kawai. [5] J. [32] S. [17] A. N. JHEP 9909 (1999) 032. Y. H. Dijkgraaf. Kawai. M. Cambridge. hep-th/9510017. B 465 (1999) 142. Theor. A. Iso.R. E. Russo. B 498 (1997) 467. Smith. Prog. Aoki. Phys. hep-th/9611047. 117B (1982) 228. 38 (1999) 1113. hep-th/9701151. Phys. 134 (1999) 47. Tokura. hep-th/9801182. A. Zarembo. Phys. Chepelev. Seiberg. Tada. E. Verlinde. B 206 (1982) 440. 112B (1982) 463. Banks. Eguchi. [7] T. N. Schwarz. Kitazawa / Nuclear Physics B 581 (2000) 295–308 Acknowledgements We would like to thank N. [30] J. Lett. Parisi. Neuberger. H. Tseytlin. [19] A. Strominger. [31] J. U.V. Y. Phys. JHEP 9802 (1998) 003. 172. Nekrasov. References [1] T. B 67. Lett. Lett. Y. Seiberg. [12] A. Theor. Shenker. hep-th/9711162. Phys. hep-th/9903217. Tsuchiya. hep-th/9907166. [4] H. Rev. Gonzales-Arroya. [13] N. Fischler. Y. Phys. Phys. Phys. T. Kimura. Math. Phys. Phys. Ishibashi. Vol. Phys. Adv. Tseytlin. Heller. Susskind. Math. S. Lett. Kawai. hep-th/9908134. T. Itzhaki. Iso. Makeenko.S. Rev. Chepelev. Rev. [25] M.C. S. Lett. Minwalla. [26] M. Taylor. hep-th/9703030. [6] J. Nucl.C. Douglas. Douglas. [28] N. Tada. Commun. W. D 55 (1997) 5112. M. Ishibashi and the members of the KEK theory group for useful discussions. A. Russo. H. Polchinski. S. Lett. hep-th/9601029. hep-th/9908141. [10] G. H.

elsevier.Nuclear Physics B 581 (2000) 309–338 www.3 2 1 3 ( 0 0 ) 0 0 2 7 1 .edu 0550-3213/00/$ – see front matter  2000 Elsevier Science B. for example [3–8]. intersecting brane configurations and configurations involving negative tension branes. Los Alamos National Laboratory. Swansea. with and without scalar fields. We show that under certain mild assumptions there is a universal equation for the gravitational fluctuations.2 . We find that three-brane metrics in five dimensions can arise from a single scalar field source.princeton.4 a Theory Division T-8. 1. These generalizations include three-branes in higher dimensional spaces which are not necessarily anti-de Sitter far from the branes. PII: S 0 5 5 0 . USA b Department of Physics. We also comment on the possible contribution of resonant modes.lanl. Timothy J. Joshua Erlich a. Hollowood a. E-mail: csaki@lanl.6 .nl/locate/npe Universal aspects of gravity localized on thick branes Csaba Csáki a. We study both the graviton ground state and the continuum of Kaluza–Klein modes and we find that the four-dimensional gravitational mode is localized precisely when the effects of the continuum modes decouple at distances larger than the fundamental Planck scale.  2000 Elsevier Science B. Yuri Shirman c. for previous relevant work. Robert Oppenheimer Fellow.gov 2 erlich@lanl. All rights reserved.1 .gov 3 pyth@skye. SA2 8PP.V. revised 16 March 2000. Princeton. We also show that a single scalar field cannot produce conformally-flat locally intersecting brane configurations or a p-brane in greater than (p + 2)-dimensions.2] to localize gravity in the vicinity of a brane with non-vanishing tension in anti-de Sitter (AdS) space has recently attracted enormous attention (see. The decoupling is contingent only on the long-range behaviour of the metric from the brane and we find a universal form for the corrections to Newton’s law. and we rederive the BPS type conditions without any a priori assumptions regarding the form of the scalar potential. NM 87545.V. Introduction The proposal of Randall and Sundrum (RS) [1. UK c Department of Physics. NJ 08544. USA Received 24 January 2000. we find general classes of metrics which maintain localized four-dimensional gravity. Los Alamos. Given this. for 1 J. [9–43].b.3 . accepted 2 May 2000 Abstract We study gravity in backgrounds that are smooth generalizations of the Randall–Sundrum model.gov 4 yuri@feynman. Princeton University. University of Wales Swansea. All rights reserved.

/ Nuclear Physics B 581 (2000) 309–338 more recent generalizations. In the first part of the paper we consider thick branes in one extra dimension and then generalize to an arbitrary number of extra dimensions. and also to include brane junctions [10. there is a single normalizable graviton bound-state with zero energy. which is a consequence of the unbroken fourdimensional Poincaré invariance. For a generic study. RS found that in a setup with a single brane. In a more restricted set of generalizations of the RS solution we calculate this suppression more rigorously. no explicit solutions to these equations are known yet. and find qualitative agreement with the WKB result. Similar BPS equations for intersecting domain walls in more than one extra dimension were found in [12. which hints that when the metric falls off slowly enough from the brane the soft KK modes are suppressed sufficiently inside the brane so that their corrections to the Newton’s law are negligible. their contribution to the Newton potential is highly suppressed. Such domain wall solutions are obtained if the scalar potential originates from a superpotential (although as recently discussed in [60. [44–51]. as in [47. RS found that due to the suppression of the wavefunctions of the continuum modes close to the brane.51]). a negative bulk cosmological constant and a single large extra dimension (with the cosmological constant and brane tension tuned such that the effective four-dimensional cosmological constant vanishes) the solution to Einstein’s equation results in a single graviton zero mode. We examine the spectrum of graviton modes and find necessary and sufficient conditions for such backgrounds to localize gravity on the branes. and therefore a realistic model with uncompactified extra dimensions could be built.13]. one can find solutions to Einstein’s equation coupled to a single scalar field. just like for the case of the infinitely thin branes of RS. The branes in the RS setup and its generalizations mentioned above are included as static point-like external sources in the extra dimensions. These domain walls were first found in [44. however. we “smear” the RS solution and its generalizations in such a way that the non-dynamical source terms correspond to a smeared (thick) brane in the background of a slowly varying negative bulk cosmological constant. This model has been generalized in [9] to models with intersecting branes with more than one uncompactified extra dimension. It has been shown in [46–48] that. for embeddings in string theory and supergravity. we also examine the behavior of the continuum modes.61] this theory cannot be embedded in any known fivedimensional supergravity theory). In this paper we study generic properties of localized gravity on thick branes. Normally the presence of these continuum modes would render a setup like this unrealistic due to the large deviation from Newton’s law the low energy continuum modes tend to induce. However. As was done in [46–50. Instead of starting with a coupled gravity-scalar system. for the general relativity aspects and finally [76–92].11]. where the scalar creates a domain wall — a “thick brane” — while the metric away from the brane asymptotes to a slice of AdS5 . for cosmological and phenomenological aspects). Csáki et al. with no dynamics to produce them. In this case the solutions found in [46–48] originate from a BPS equation.48]. and a continuum of Kaluza–Klein (KK) modes. [62–75]. Besides general arguments about the ground state (some of which appear in [47. for work on smooth brane scenarios.310 C. [52–61].45]. A particularly nice example of this sort has been recently worked out in detail in [51]. . we use the WKB approximation for the “volcano-type potential”.54].

In the latter case. . three-branes in more than five dimensions or higher-dimensional intersecting branes with a four-dimensional intersection. . 5 We will assume that A(0) = 0. We find that the stress-tensor source terms for a general smearing of the RS solution can be obtained from a single scalar field. n. We also comment on the possible contributions of “quasi-bound-states” — resonant modes in the continuum spectrum whose wavefunctions are not suppressed at the location of the brane — and study their significance in a toy model. for a = 0. −. for i = 1. are the coordinates on the n = (d − 4)-dimensional transverse space. Csáki et al. . / Nuclear Physics B 581 (2000) 309–338 311 This leads to necessary conditions of the quantum mechanical potential. in order for the KK modes to decouple. 2. 5 In our conventions the metric g µν has signature (+. and also emphasizes that this is the most general solution with a single scalar field. · · ·). Backgrounds with four-dimensional Poincaré invariance In general. but is the case most relevant for phenomenology): ds 2 ≡ gµν dx µ dx ν = e−A(z) ηab dx a dx b + gij (z) dzi dzj . In the case of branes in higher dimensions the situation is more complicated. for example. on a more restricted set of metrics which are conformally flat.2) ds 2 = e−A(z) ηab dx a dx b − dzi dzi . that is of the form  (2. These scenarios could describe. and again we find conditions on the long-distance behavior of the background metric in order for there to be localized gravity on the brane intersection. and hence the background metric. −. We find that the potential at large distances must fall off no faster than in the asymptotically AdS case in order for the KK modes to make a small contribution to Newton’s law. This provides a particularly simple derivation of the BPS equations without an a priori assumption about the form of the scalar potential. . for the most part. we find that it is impossible to generate the desired background metric or sources of the stress tensor with a single scalar field. In this present work we will concentrate. −.1) Here. and we rederive the same BPS-type equations for this scalar field as [46–48]. the properties of the graviton in such backgrounds are studied in the same way as for the case of one extra dimension. We also study the relevance of background fields that create the branes. where x a . we are interested in d-dimensional backgrounds which have a fourdimensional Poincaré symmetry (the restriction to four dimensions is unnecessary. x µ = (x a . . . so that the four-dimensional metric at the origin in the transverse space is canonically normalized. the thick brane background could be given by an appropriate smearing of the intersecting brane solution of [9]. zi ). . (2. . 3. We next show how to generalize our results to situations in more than five dimensions. .C. are the usual coordinates of fourdimensional Minkowski space and zi = x i+3 . This requirement is equivalent to demanding normalizability of the ground state graviton wavefunction. Nevertheless. Contrary to the case of one extra dimension.

This metric approaches the AdS form asymptotically for |z|  1/k. and we will do so in most of our examples. the dynamics of gravity in such backgrounds does not depend on the particular realization of those backgrounds (as long as a particular condition is met on the form of the stress tensor). we study the action Z √  √  (2. Here g = |det gµν |. This action gives rise to a stress tensor  1 (2. although for practical purposes it is more convenient to smooth in the z-basis: 1 (2. where M∗ is the fundamental Plank scale in five dimensions. where A(z) depends on only one scale k. µ  k. However. µ where the substitution is made only in the warp factor and not in dz2 so that it is not simply a coordinate change. g(4) = |det gab | is the determinant of the induced metric on the domain wall and κ 2 = M∗−3 .312 C. We will show later that these results apply to dynamically generated backgrounds. Consider the five-dimensional case where the domain wall is generated by an explicit position dependent term in the gravity action. or equivalently A(z) = log(k 2 z2 + 1) [51].3) with A(r) = 2k|r|. The function V (r) will approximate a delta function which generates the domain wall. One simple way to introduce thick brane is by smoothing out RS ansatz. We first point out that smoothing of the RS solution can be performed with nondynamical background sources. On the other hand. much in the spirit of the original Randall–Sundrum scenario. The metric is usually written in the form ds 2 = e−A(r) ηab dx a dx b − dr 2 (2. Csáki et al. We will be interested. as well.5) S = − d5 x g κ −2 R + Λ(r) + g(4) V (r) .5). In the RS limit.6) Tµν = Λ(r) gµν + V (r) gab δµa δνb . Hence. Here µ is an independent parameter which determines the thickness of the brane. we expect that any additional fields will be localized near z = 0. / Nuclear Physics B 581 (2000) 309–338 with a suitable choice of coordinates. for simplicity it is convenient to study the behaviour of the gravitational modes in a different limit µ ∼ k. For example. We .4) |z| → log cosh(µz) . but can be written in conformally-flat form with A(z) ≡ A(r(z)) = 2 log(k|z| + 1). amongst other things.2) is perfectly general. Notice that when d = 5. In this case the equivalent smoothing in conformally-flat coordinates can be written as A(r) = 2 log cosh(kr). and Λ(r) is roughly constant away from the domain wall and corresponds to the bulk cosmological constant. and therefore is derivable from an action of the form (2.1) is conformally flat. 2 We note that any stress tensor which is four dimensional Lorentz covariant can be decomposed in this way. In the absence of fields other than gravity. which will be assumed in much of the following discussion (we will comment on subtleties associated with smearing of the matter fields over the thickness of the brane later). in smooth versions of the RS metric in five dimensions discussed in [1]. one can make a substitution |r| → µ−1 log cosh(µr) in A(r). an arbitrary metric of the form (2. we will also consider metrics which are more general and not necessarily asymptotic to an AdS space. so in that case (2.

V (r) = 3κ −2 A00 (r). For example. where it is understood that r is periodic over an interval 2rc . this reproduces the original Randall-Sundrum system [1]. Csáki et al. (2. 2 where µ is. in which the transverse direction is compactified on an orbifold S 1 /Z2 .8) A(r) = n∈Z where θ (r) is the usual step function. For example. which has delta function singularities with positive coefficient for r = 0 mod 2rc and negative coefficient for r = rc mod 2rc . for example.7) For A(r) = 2k|r|. By choice of a static background metric (2. . we will consider smooth versions of the multi-dimensional patched AdS space with metric [9]  1 a b i i (2.3) the four-dimensional effective cosmological constant must vanish. Pn 1 + k i=1 |zi | This metric represents n = d − 4 intersecting (d − 1)-branes.3) with A(r) = 2k|r|. as described previously. In the RS scenario the metric has the form (2. The “bulk cosmological constant” and “brane tension” are then determined to be Λ(r) = −3κ −2 A0 (r)2 .C. we can study a multiple cover of the circle. −rc < r < rc .  1 (2.5) vanishes for the background metric (2.7). / Nuclear Physics B 581 (2000) 309–338 313 are interested in actions which admit solutions of the form (2. one can integrate out the extra dimension and check that the action (2. one of which has positive tension and the other negative tension. We should comment that because this analysis does not depend on what type of field creates the domain walls. which is equivalent to vanishing of the four-dimensional cosmological constant.3). and the fine tuning of cosmological constant and brane tension is replaced by (2. the question for these kinds of backgrounds is whether four-dimensional gravity is localized on the intersection. we can study smooth versions of the RS solution to the hierarchy problem [2]. The brane tension is related to A00 (r). In the latter cases. with X    k rc + (−1)n 2r − (2n + 1)rc θ (r − nrc ) θ (n + 1)rc − r . which mutually intersect in four dimensions. we can study solutions in which the domain walls have negative tension. (2. Indeed. An alternative smoothing procedure which may be useful for numerical calculation is a truncation of the Fourier expansion of the sawtooth:  1 rc 4rc X cos π(2n + 1)r/rc .3). Branes are stuck at the two fixed planes of the orbifold action.9) θ (x) → θ˜ (x) ≡ tanh(µx) + 1 .10) A(r) ≈ − 2 2 π (2n + 1)2 06n<N We will also consider backgrounds that can be interpreted as three-branes embedded in spaces of dimension d > 5 and also backgrounds which can be interpreted as intersections of higher dimensional branes with a four-dimensional intersection.11) ds 2 = 2 ηab dx dx − dz dz . a parameter which characterizes the thickness of the brane. (2. In order to make the periodicity explicit. As before the solution can be smoothed by an appropriate smearing of the θ -functions.

δTµν = Tµκ hκν . It will be convenient to define hµν to be a fluctuation whose only non-zero components are hab . (3. / Nuclear Physics B 581 (2000) 309–338 3.  eµν = ∂ ρ ∂(ν hµ)ρ − 1 ∂ ρ ∂ρ hµν − 1 ∂µ ∂ν hρρ − 1 ηµν ∂ ρ ∂ κ hρκ − ∂ ρ ∂ρ hκκ . Csáki et al. .1) ds 2 = e−A(z) ηab + hab (x. The other terms either vanish due to the gauge conditions or they vanish because hµν only has non-zero components hab and. The next question concerns the variation of the stress tensor. (3. Gravitational fluctuations In this section we derive a universal equation for the effective four-dimensional gravitational fluctuations in the conformally-flat backgrounds described in the last section.. but we will not comment on such modes here. Since the metric (2.2) of the form    (3. [93]):  eµν + d − 2 1 ∇ eν A + ∇ eµ ∇ eµ A ∇ eν A Gµν = G 2 2   d −3 eρ A .  1 1 1 δGµν = ∂ ρ ∂(ν hµ)ρ − ∂ ρ ∂ρ hµν − ∂µ ∂ν hρρ − ηµν ∂ ρ ∂ κ hρκ − ∂ ρ ∂ρ hκκ 2 2  2  d − 2 1 ρκ η ∂µ hνρ + ∂ν hµρ − ∂ρ hµν ∂κ A − 2 2  + hµν ∂ρ ∂ ρ A + ηµν −∂ρ hρκ ∂κ A + 12 ∂ρ hκκ ∂ ρ A − hρκ ∂ρ ∂κ A   d −3 hµν ∂ρ A ∂ ρ A − ηµν hρκ ∂ρ A ∂κ A . We will use the transverse traceless gauge for these fluctuations. For any action of the form (2.1. This means that we study fluctuations of the metric (2. for example. eρ A ∇ eρ A − eρ ∇ (3.314 C. Later we will show that this transformation property remains valid when the background is generated by scalar fields. it is convenient to use the the general form of the Einstein tensor for metrics of the form gµν = e−A g˜µν (see. A is a function of the {zi } only.5) which is automatically symmetric.1) will be discussed separately in Section 3. (3. Using the form of the Einstein tensor for linear perturbations about flat spacetime [93].e.2) ∇ − g˜ µν ∇ 4 where indices are raised and lowered with g˜µν in this context. Universal aspects 3.4) − 4 Only the terms which are underlined are actually non-zero.5). moreover. ∂µ hµν = 0 and µ hµ = 0. i. z) dx a dx b − dzi dzi . The more demanding case of the general background (2.2) becomes. We should note that there may be additional fluctuations not of the form (3.1) in transverse traceless gauge.4.3) δG 2 2 2 the linearized perturbation of the Einstein tensor (3.2) is manifestly conformally flat.

The Einstein tensor derived from the pure gravity part of the action transforms differently under metric fluctuations. the stress tensor transforms under metric perturbations according to (3. We will later show that the stress tensor for scalar fields transforms the same way.6) −∂ρ ∂ ρ A + 2 4 Again only the underlined terms survive due to either the gauge conditions or the particular choice of background. From (3. hµν = e(d−2)A/4 h˜ µν . Given the action functional of the theory. where x = −ηab ∂a ∂b and ∇z2 = ∂i2 . (3. z) = hˇ ab (x)ψ(z) with x hˇ ab (x) = m2 hˇ ab (x).8) − ∂ ∂ρ hµν + 2 32 8 We should emphasize that indices are raised and lowered here using the flat metric ηµν . + (3. For the case of a nondynamical background specified by the action (2. the graviton equations-ofmotion (3..9) −∇z2 ψ(z) + 16 4 This has the form of a Schrödinger equation for the “wavefunction” ψ(z). Csáki et al. The fact that when expressed in terms of the variable h˜ µν . Hence. given the action. i. the stress tensor is immediately determined in terms of the matter fields and metric. (3. where m is the four-dimensional Kaluza–Klein mass of the fluctuation.5).7) − ∂ ρ ∂ρ hµν + 2 4 If we redefine the metric perturbation so that its kinetic term has the canonical normalization. “energy” m2 and potential d −2 2 (d − 2)2 ∇z A(z) · ∇z A(z) − ∇z A(z). Then since A = A(z) only.9) is the generalization of the fluctuation equation for the RS case A(z) = (1 + k|z|)−2 given in [2] to arbitrary conformally flat metrics.5).10) 16 4 The fluctuation equation (3. / Nuclear Physics B 581 (2000) 309–338 315 We pause here to review the logic of this procedure. and look for solutions of the form h˜ ab (x. Now we use the fact that ∂ ρ ∂ρ = −x − ∇z2 . (3. In addition. leaving d −2 ρ 1 ∂ A ∂ρ hµν = 0. Both the surviving terms of δTµν cancel with two terms of δGµν in the graviton equation of motion δGµν − κ 2 δTµν = 0. (3. only the hab components of the fluctuation hµν are non-vanishing. we have   (d − 2)2 d −2 2 ∇z A · ∇z A − ∇z A ψ(z) = m2 ψ(z). which is why the linearized Einstein equations which follow are nontrivial. and the unperturbed Einstein equation.C.8) have no single derivative terms is equivalent to the fact that the action for these fluctuations has the form of a canonical kinetic term: V (z) = . then the term linear in derivatives is removed:   (d − 2)2 ρ d −2 ρ 1 ρ ˜ ∂ A∂ρ A − ∂ ∂ρ A h˜ µν = 0.5).e. the form of fluctuations of the stress tensor under metric perturbations is also determined. we derive   d − 2 −2 1 κ ∂µ A ∂ κ A + ∂µ ∂ κ A hκν δTµν = 2 2   d − 2 −2 d −3 κ ∂ρ A ∂ ρ A hµν .

12) d z ψ(z) · d4 x ∂c hˇ ab (x) ∂ c hˇ ab (x).316 C. (3. and R (4) is the four-dimensional curvature scalar created by those matter perturbations. it is convenient to decompose the action into a four dimensional part and higher dimensional parts. This can be seen by expanding the scalar curvature about the background (2. / Nuclear Physics B 581 (2000) 309–338 Z S∼ dd x ∂ρ h˜ µν ∂ ρ h˜ µν . Csáki et al. (3. the action takes the form. and determines 6 the four-dimensional gravitational coupling GN ∼ M4−2 .15) M42 = M∗d−2 dn z e−(d−2)A(z)/2. (Notice that this differs from the inner-product employed in [47]. which is related to the coupling κ via κ 2 = M∗2−d . then the kinetic terms in the x and z directions would have had different conformal factors multiplying them. generalizing the 6 Note that the above relation relies on our choice A(0) = 0. Z Z q S ∼ M∗d−2 dn z e−(d−2)A/2 · d4 x gˆ(4) R (4) + · · · .14) where gˆ(4) is the determinant of the four-dimensional metric for matter perturbations about flat spacetime. we notice. . including the fundamental (d-dimensional) Planck scale M∗ .11).) The redefinition of the metric hµν = e(d−2)A(z)/4 h˜ µν then absorbs the conformal factor multiplying the kinetic terms and puts the action in the canonical form (3. for solutions in the form hˇ ab (x)ψ(z). from which we deduce that the appropriate inner-product for the “wavefunctions” ψ(z) is the conventional quantum mechanical one. As in [1.9].9).11) where indices are contracted with the flat metric ηµν . (3. In particular. In particular.9) admits a normalizable zero-energy ground state.11) includes the term Z Z n 2 (3. To find this zero-energy state. 3. Z (3.13) (3.) In order to calculate the strength of the four-dimensional gravitational coupling. This allows us to identify the four-dimensional Planck scale M4 via.2.2) and keeping track of powers of the conformal factor in the metric. (If we had chosen coordinates in which the metric is not explicitly in the conformally-flat form. before rescaling the graviton. in order to have an effective four-dimensional theory of gravity we require that (3. The four-dimensional graviton The question of whether there is localized (four-dimensional) gravity supported in the vicinity of the brane now becomes contingent on properties of the quantum mechanical system described by the Schrödinger equation (3.

ψ 4 Q = ∇z + (3.C. Rather than deal with this complication. that the Schrödinger equation (3.17) (3. the matter fields in the four-dimensional theory on the brane would be smeared over the width of the brane in the transverse space. we also require that the other solutions of the Schrödinger equation (3. The situation where V (z) = 0 as |z| → ∞. the zero-energy wavefunction. b0 b0 |ψ ψ (3. b0 (z) is always normalizable. only. we will.20) This requires that A(z) → ∞ sufficiently fast as |z| → ∞. then ψ b0 (z) is not normalizable and therefore is of no interest to us since it cannot |z| → ∞.9).3. consider the gravitational potential . b0 (z) is normalizThe condition for having localized four-dimensional gravity is that ψ able. there are no normalizable negative energy graviton modes. ∇z A. On the contrary.7) always admits the solution where hab = hab (x). Normalizability is intimately connected with the asymptotic behaviour of the potential of the Schrödinger equation (3. (3. we make the obvious remark that in any scenario where the transverse b0 (z) is nonspace is asymptotically flat Euclidean space (A(z) → constant. then ψ describe localized four-dimensional gravity. Corrections to Newton’s law In order to have localized four-dimensional gravity. (3. the most interesting and we will focus on that case in most of the remainder of this paper. as required for stability of the gravitational background. do not lead to unacceptably large corrections to Newton’s law in the four-dimensional theory. 3. Q† = −∇z + 4 4 Hence. Furthermore. is   d −2 b0 (z) = exp − A(z) .15)) be non-vanishing. as |z| → ∞) ψ normalizable and gravity cannot be localized. is. Csáki et al.9) can be rewritten as a supersymmetric quantum mechanics problem of the form.19) dn z exp − 2 Notice that normalizability of the ground state wavefunction is equivalent to the condition that the four-dimensional gravitational coupling (via (3. indeed GN ∼ b0 (0)2 M∗2−d ψ . since the “Hamiltonian” Q† · Q is a positive definite Hermitian operator.18) Notice that such a wavefunction always exists since (3. At this point. in other words   Z d −2 A(z) < ∞. / Nuclear Physics B 581 (2000) 309–338 317 observation of [47. and x hab (x) = 0. if V (z) < 0 as If V (z) > 0 as |z| → ∞. annihilated by Q. perhaps. Q† · Q ψ(z) = m2 ψ(z). In any realistic brane scenario.55–57] to higher dimensions.16) where the n-vector of supersymmetry charge is d −2 d −2 ∇z A .9). the KK modes. for simplicity.

For states which form a continuum in n-dimensions starting at m0 the correction to Newton’s law is Z∞ M1 M2 M1 M2 e−mr 2−d + M∗ ψm (0)2 . Notice that in flat db0 (z) and dimensional space there would be no normalizable zero-energy wavefunction ψ the continuum would extend down to m = 0 and be unsuppressed: ψm (0) = 1. This assumption is justified in cases when the thickness of the brane is small compared with the bulk curvature. Csáki et al. dm mn−1 (3. 7 As long as m is large enough. we note that a discrete eigenfunction (these are not present in the RS case) of the Schrödinger equation ψm (z) of energy m2 acts in four-dimensions like a field of mass m and consequently contributes a Yukawa-like correction to the four-dimensional gravitational potential between two masses M1 and M2 : M1 M2 M1 M2 e−mr + M∗2−d ψm (0)2 . A more complete analysis would involve the effects of the overlap of the gravitational modes with the matter modes. The fact that ψm (0) appears in (3.22) with n replaced by p. to unity over a period at |z| → ∞.318 C.e. zi = 0. there are continuum modes which are localized in a subset of the transverse dimensions. in the transverse space [1. the most interesting case is where the potential goes to zero at infinity.22) U (r) ∼ GN r r m0 where the wavefunctions ψm (z) of the continuum are normalized as plane waves.. as expected for the gravitational potential in d-dimensions. the wavefunctions behave as plane waves only in p < n of the transverse dimensions. We will not have more to say about such corrections here.21). In other words. 7 Note that in our convention the zero-energy state ψ b0 (z) is not unit normalized. The correction from any continuum states ψm (z) is obtained by integrating over these states with the relevant density-of-states measure. For the case when V (z) > 0 as |z| → ∞. A(0) = 0 we have ψ . We expect that our conclusions will be at least qualitatively correct in a general case and present some supporting arguments in Section 6. Hence corrections to Newton’s law are exponentially suppressed as in (3. As we have already mentioned. however. since we have chosen b0 (0) = 1. (3.55–57]. In such a case U (r) ∼ M1 M2 M∗2−d r −n−1 = M1 M2 M∗2−d r 3−d . The factor of mn−1 is just the n-dimensional plane wave continuum density of states (up to a constant angular factor). and would correct the factors of ψm (0). i. In this case there is a continuum of scattering states ψm (z) with eigenvalues m2 > 0 and so the behaviour of the soft modes at zi = 0 is crucial for determining whether the corrections are small. In the case of intersecting branes.21) U (r) ∼ GN r r R n where the wavefunction ψm (z) is normalized d z ψm (z)2 = 1. In this case the corrections from this part of the continuum spectrum are of the form (3. / Nuclear Physics B 581 (2000) 309–338 between two point-like sources of mass M1 and M2 located at the origin.21) is due to our simplifying assumption that the sources are point-like and located at zi = 0.9. the excited states are clearly separated by a gap from the ground-state. this will be a small correction. In order to evaluate the correction to Newton’s law.

g (3.25) − g˜ −1/2 ∂i g˜ g˜ ij ∂j A + 2 4 Assuming that the variation of the stress-tensor is given by (3.29) − √ ∂i g g ij ∂j ϕ(z) = m2 g 00 ϕ(z). Notice that when m = 0 (3. − ∂ c ∂c hµν − g˜ −1/2 ∂i g˜ g˜ ij ∂j hµν + 2 2 4 which can be rewritten in terms of the original metric as  1 √ √ ∂ρ g g ρκ ∂κ hµν = 0.29) always admits the solution ϕ(z) = constant. we follow essentially the same steps as for the conformally-flat case detailed in Section 3. we can then write down the variation of the original Einstein tensor: p  d −2 1 1 ∂i A∂ i hµν δGµν = − ∂ c ∂c hµν − g˜ −1/2 ∂i g˜ g˜ ij ∂j hµν + 2 2 4   p  d −3 d −2 ∂i A∂ i A hµν . until further notice.2) in order to find the variation of the Einstein tensor Gµν .1. For a conformally-flat background (3. Csáki et al. eµν = − 1 ∇ (3. First of all. + (3. δTµν = 2 4 Hence. indices are raised and lowered with g˜µν . For a fluctuation of the form hab (x. we will make immediate use of the following µ four facts: (i) ∂µ hµν = 0. In other words.23) ds 2 = e−A(z) ηab + hab (x.7).4. This has been noted in the case of one transverse dimension in [55] which also discusses its significance within the AdS/CFT correspondence. we have  1 √ (3. Rather than writing down all of the terms.1. By brute force one can show that the eµν is variation of the Einstein tensor G p  eρ ∇ eρ hµν = − 1 ∂ c ∂c hµν − 1 g˜ −1/2 ∂i g˜ g˜ ij ∂j hµν .24) δG 2 2 2 where. z) dx a dx b − gij (z) dzi dzj . we briefly indicate how the preceding analysis of gravitational fluctuations extends to cases where the background is of the general non-conformally flat form (2. Extension to non-conformally-flat backgrounds In this section. and (iv) the only nonvanishing components of the variation hµν are hab .28) reduces to (3. g This describes the z-dependence of a mode with four-dimensional Kaluza–Klein mass m.27) (3. z) = ϕ(z)hˇ ab (x). the general equation for the fluctuations is simply the covariant scalar wave-equation.5). we have p  d − 3 −2  d − 2 −2 κ −g˜ −1/2 ∂i g˜ g˜ ij ∂j A + κ ∂i A∂ i A hµν . In order to derive an equation for metric fluctuations  (3. Einstein’s equation gives p  d −2 1 1 ∂i A∂ i hµν = 0. it is convenient to define the metric g˜µν (z) = eA(z) gµν (z) and apply the relation (3.26) (3. / Nuclear Physics B 581 (2000) 309–338 319 3.1). as we did in Section 3.2).28) where now indices are raised and lowered with gµν .C. (ii) hµ = 0. Using (3. this will lead to . with x hˇ ab (x) = m2 hˇ ab (x). (iii) gµν only depends on z.

/ Nuclear Physics B 581 (2000) 309–338 b0 (z) in the conformally-flat case. The normalizability condition is consequently Z p (3.9):   ψ(z) 1 √ ij gg ∂j 00 √ 1/2 = m2 ψ(z). Localization and decoupling in d = 5 We begin by discussing the case in d = 5. in addition. ψ(z) ≡ ϕ(z) g 00 (z) g(z) which satisfies a generalization of the Schrödinger equation (3. In other words we shall consider metrics of the form  (4.1) is  (4.1) ds 2 = e−A(%) ηab dx a dx b − d%2 − %2 dΩ 2 .1. We will.2). The fluctuation the analogue of the zero-energy solution ψ equation (3. we deduce the generalized expression b0 (z): for the zero-energy “wavefunction” ψ  00 p 1/2 b0 (z) = g (z) g(z) . (3. We will briefly consider the more general radially symmetric background which is not necessarily conformally flat at the end of Section 4. restrict ourselves mainly to a conformally-flat background which is. The “wavefunction” is consequently p 1/2  .32) dn z g 00 (z) g(z) < ∞. The function A(%). inserting ϕ(z) = constant. as per Section (2.30) From the z-integral above.34) 4. (3. 4. + − dz2 16 4 (4. as indicated. − 00 √ 1/2 ∂i [g g] [g g] (3. when the transverse space is one-dimensional. Gravity localized on thick three-branes In this section we consider in some detail the case when the 3-brane is embedded in a space with dimension d > 5.5.18) in the conformally-flat case.320 C.33) (3. where % is the radial coordinate in the transverse directions and Ω are the angular coordinates on S d−5 . corresponding to the RS scenario. The Schrödinger equation (3.3) .28) is derivable from the action Z √ S ∼ dd x g ∂ρ hµν ∂ ρ hµν Z Z p = dn z g 00 (z) g(z) ϕ(z)2 · d4 x ∂c hˇ ab (x)∂ c hˇ ab (x) + · · · . only depends on the radial variable. radially symmetric in the space transverse to the brane.31) ψ which reduces to (3. Csáki et al.9) is:   9 0 2 3 00 d2 ψ(z) A (z) − A (z) ψ(z) = m2 ψ(z).2) ds 2 = e−A(z) ηab dx a dx b − dz2 . In this case. the metric (4.

if we assume that the potential falls b0 (z) falls off as e−c/|z|β−1 . 1. As we explained in b0 (z) is normalizable is intimately connected with the Section 3. the question of whether ψ b0 (z) is always asymptotic behaviour of the potential V (z).2. In this case the potential V (z) falls off as α(α + 1)/z2 . and includes smoothed versions of the AdS scenario of [1]. V (z) → 0 as |z| → ∞.. If V (z) > 0 as |z| → ∞. then ψ0 (z) is not normalizable and therefore is no interest to us since it cannot describe localized four-dimensional gravity. Fig.e. Consequently. the bound-state of lowest energy. / Nuclear Physics B 581 (2000) 309–338 321 Fig. then ψ b normalizable. includes the AdS scenario. then normalizability requires α > 1/2.2. The zero-energy state (3. the excited states are clearly separated by a gap from the ground-state. is the most interesting case. For the case when V (z) > 0. (4. for |z| → ∞. Csáki et al. 1 shows the potential V (z) for a particular smoothing of the AdS case with A(z) = log(k 2 z2 + 1).10) for the case A(z) = log(k 2 z2 + 1). On the contrary. The relevant effects of these modes on Newton’s law were established in Section 3. Conversely. The fact that there are no bound-states with negative energy follows from the factorization (3.C. if β < 1. The Schrödinger potential (3. exp[− 32 A(z)] must fall off faster than 1/z.4) For this to be normalizable. We now have to consider the question of whether or not the other modes in our problem decouple. dz 4 b0 (z) = Hence the zero-energy state (4. Hence corrections to Newton’s law will be . i. which satisfies the supersymmetric condition Qψ 0. where the potential falls off just slowly enough to give rise to a normalizable bound-state. For example.16):    3 0 3 0 d d + A (z) + A (z) ψ(z) = m2 ψ(z). The borderline case. It is interesting to note that the borderline case. V (z) ∼ |z|−2 .5) − dz 4 dz 4 which has the form of supersymmetric quantum mechanics Q† Qψ(z) = m2 ψ(z).4).3. 4 (4. then the off as |z|−2β the wavefunction ψ wavefunction is not normalizable.18) is [48.51]   3 b ψ0 (z) = exp − A(z) . with Q ≡ 3 d + A0 (z). discussed in Section 4. when the wavefunction (3. if V (z) < 0 as |z| → ∞.18) has a power-law fall-off ψ(z) ∼ |z|−α . is the ground state.

The most interesting case is where the potential goes to zero at infinity.21). Since the bottom of the continuum is at m = 0 it is clear that b0 (z) decoupling is a delicate issue. This suggests that there is a natural decoupling of the normalizability of the state ψ b0 (z) is normalizable. As we have discussed above. in the WKB approximation. 8 " # zZ 1 (m) p 2 T (m) ∼ exp −2 dz V (z) − m . in other words (4. (2) 1/k  z  1/m. the central region extends over a scale 1/k. From (3. In order for the soft continuum states to decouple we would need T (0) = 0. 9 Recall from Section 3. we must solve the Schrödinger equation with the tail of the potential (4. We shall not assume any particular form for the potential V (z) except. as in (3. In regions (2). consider a potential of the form α(α + 1) . Before we attempt some rigorous analysis. (4. In order to get a feeling for what might be required.7): V (z) ∼ 8 Of course. what we need to calculate in order to investigate the decoupling of the continuum modes is the limiting behaviour of ψm (0) at small m.7) is valid for |z|  1/k. it is instructive to consider the WKB approximation for this tunneling probability. so that. Csáki et al. the WKB approximation is not valid in the central region of the potential for states of small energy. In this case there is a continuum of scattering states ψm (z) with eigenvalues m2 > 0. let us consider the problem with some rather crude apparatus.22). To this is lead to small corrections to Newton’s law) precisely when ψ end. Later we shall present a more rigorous analysis. / Nuclear Physics B 581 (2000) 309–338 exponentially suppressed. for simplicity.6) has a divergence for large z. This matches precisely the condition on b0 (z). in other words. that it only depends on a single dimensionful scale k. if V (z) falls off at least as slowly as 1/z2 .7) z2 for large |z|. 9 Let us consider four different regions in z (we will always be considering modes with energies m  k): (1) z  1/k. this can be achieved if the integral (4. (4. (3) 1/k  z ∼ 1/m and (4) 1/m  z. The transition probability is. m . Consider a continuum mode with energy m2 incident upon the potential from the right.6) z0 (m) where 0 < z0 (m) < z1 (m) are the two points in the rightmost barrier region where V (z) = m2 . decoupling would require that 0 dm e the bottom limit of integration. Since z1 (0) = ∞. We shall find that the crossover from localization to de-localization occurs for some critical value of α.322 C. If the continuum modes are to decouple we want the probability for continuum modes to tunnel into the central region of the potential to be vanishingly small for m → 0.3 that the wavefunctions ψ (z) must be normalized as plane waves for |z| → ∞. the continuum states precisely when the ground state ψ So the hypothesis that we want to establish is that the continuum modes decouple (that b0 (z) is normalizable. (3) and (4). for instance. given that gravity is localized so that ψ R∞ −mr 2 |ψm (0)| has no singularity at is normalizable.

which matches the first term in (4. To leading order in m.12) with z  1/k. The value of the wavefunction at z = 0 is then extracted from (4. this would require a non-generic behaviour of the potential. / Nuclear Physics B 581 (2000) 309–338 d2 ψm (z) α(α + 1) + ψm (z) = m2 ψm (z).11).11) then match the next term in the expansion (4. 11 Using this matching procedure. we must multiple it by an overall factor of mα−1 . Csáki et al. the dominant term in region (4) comes from the second term in (4. coming from the expansion of Yα+1/2 (mz) are more complicated since they depend on whether α + 1/2 is an integer.11) if am ∼ mα+1/2 . (4. coming from the expansion of J α+1/2 (mz) for small mz. The first correction where ψ φ(z) satisfies the inhomogeneous equation   d2 b0 (z). Now we show how to match regions (1) and (2).14) ψm (0) ∼ k 10 The corrections in the second square bracket.  α−1 m . (4. ψm (z) = ψ (4.C. In these regions mz  1 and it is meaningful to solve the Schödinger equation as a series in m2 . (4.12).13) − 2 + V (z) φ(z) = ψ dz Now we can match (4. In this region b0 (z) ∼ z−α .10) where the coefficient of the cosine goes like bm m−1/2 ∼ m−α+1 . the Bessel functions become plane waves: r r   2 2 π π sin mz − 2 α − 2 + bm cos mz − π2 α − π2 . However. are a power series in (mz)2 . however. and the exact form of the Bessel function then determines the behavior in the intermediate region (3).9) In region (4).8) ψm (z) = am z1/2 Yα+1/2 (mz) + bm z1/2 Jα+1/2 (mz) .12) b0 (z) is the suitably normalized zero energy solution (3.12). dz2 z2 The solution is given by a linear combination of Bessel functions 323 − (4. . the two terms indicated are the dominant terms for small mz. The first two terms are b0 (z) + m2 φ(z) + · · · . could match terms higher in the expansion (4. where mz  1.12) as long as bm ∼ m−α+3/2 . or not. The second and ψ third terms in (4. (4. 11 A caveat is that the third term in (4. (4.18).11) + 0(α + 3/2) 2 In other words. Since m is small. where mz  1 (but kz  1). we have determined the m-dependence of am and bm . coming from J α+1/2 (mz). The corrections in the first square bracket. Hence to normalize the wavefunction ψm (z) as a plane wave.11).10) ψm (z) = am πm πm In region (2). with the series in region (2) (4. we can match the wavefunction in regions (2) and (4) by using the asymptotic behavior of the Bessel functions. the Bessel functions can be expanded in mz giving 10     2  am z1/2 0(α + 1/2) 2 α+1/2 mz 1 ψm (z) = − + ··· 1+ π mz α − 1/2 2  α+1/2 1/2 mz bm z [1 + · · ·].

/ Nuclear Physics B 581 (2000) 309–338 where the factor of k is dictated by dimensional analysis.17) 2 2 2 2 2 (k z + 1) (k z + 1)α/2 In other words. If the matter is smeared over the distances 1/k one needs to obtain more information about the wave-functions of the excited modes to study the corrections to the Newton’s law. i. we find that the integral over the continuum modes in (3. It is important to note. b0 (z) is normalizable. since it only depends on one parameter.14) without specifying details of the potential near z = 0.14) establishes the asymptotic behaviour of the continuum modes at z  1/k. (4.14) remain valid.16) A(z) = 3 for some constants α. yet the arguments leading to (4. V (z) = k 2 α . however. The case when α = 3/2 is particularly interesting because in this case we can easily transform back to the r coordinate in which case A(r) = 2 log cosh(kr). The example above has the advantage of being simple. the potential in the Schrödinger equation and the ground-state corresponding to (4. (4. To begin with. For general α.324 C. the thickness of the brane is comparable to the bulk curvature. in agreement with the exactly solved delta-function potential of [1]. when α > 1/2. In this case the corrections are   C M1 M2 M1 M2 M1 M2 C0 + 3 2α−2 2α = GN 1+ . We emphasize that the AdS case of [1] corresponds to taking α = 3/2. Notice that in the AdS scenario of [1]. so that GN ∼ kM∗−3 . (4. in which case we find ψm (0)2 ∼ m/k. 1. however. recently discussed in [51]. Examples At this point it is probably worthwhile to consider some examples. this class of examples has precisely the asymptotic form of the potential discussed in the last section. ψ . Thus. From (4. we can easily introduce a potential such that matter is localized near z = 0.16) are. 4. Csáki et al. consider the class of conformally-flat five dimensional backgrounds for which  2α log k 2 z2 + 1 . which agrees with the exact calculation of the correction presented in [1].7) there are power-law corrections to Newton’s law with a universal exponent α determined simply by the long-range fall-off of the potential.. Notice that when the potential falls off as (4. that we obtained (4. Equation (4.2. k.14). that Now we see that it is precisely when ψ the continuum states give a correction to Newton’s law which is suppressed relative to the leading term. where k is characteristic decay width of the potential V (z).e. (α + 1)k 2 z2 − 1 1 b0 (z) = . α = 3/2.16) is asymptotically AdS and therefore represents a smoothing of the AdS example of [1]. respectively. The shape of the potential for α = 3/2 appears in Fig.22) is only nonsingular at the bottom limit of integration if α > 1/2.15) U (r) ∼ GN r M∗ k r r (kr)2α−1 where C and C 0 are dimensionless numbers. So when α = 3/2 the space given by (4. We are assuming that the metric only depends on one dimensionful parameter k.

In this limit for general α.18) A(z) = 3 from which we can easily calculate the potential V (z) = αkµ2 k(α + 1)(cosh(2µz) − 1) − 2µ − 2k log cosh(µz) .C.21) So in the limit µ  k we would expect that matter fields are localized near z = 0. (4. / Nuclear Physics B 581 (2000) 309–338 325 Fig. Csáki et al. V2 = max[V (z)] and z1 . however. For example. The generic shape of this potential is illustrated in Fig. one expects that the matter fields living on the brane will be smeared over the distance 1/k and such smearing may affect the estimates for the corrections to the Newton’s law. defined by V (z1 ) = 0. Without any further assumptions. for α = 3/2. 2µ (α + 1)k µ  k.20) so the central well becomes more like a delta function as µ  k. 2 2 cosh2 (µz) k log cosh(µz) + µ (4. (k|z| + 1)2 |z|  µ−1 . (4. 2. in the RS scenario. while z1 is asymptotically z1 = 4µ 1 log . In this limit k controls the long-range fall-off of the potential. The depth of the central well is easily found to be V1 = αkµ. which both depend on a thickness parameter µ. It is consequently more convenient to start in the z-coordinate basis with A(z) = 4α 3 log(k|z| + 1) (where the RS scenario has α = 3/2) and smooth |z| with either z tanh(µz) or µ−1 log cosh(µz).19). in which case we can neglect the overlap of those fields with the gravity modes as corrections to . In order to localize the matter fields near z = 0 we can introduce a parameter characterizing thickness of the brane (as observed by the matter fields) in addition to the overall scale k. via the asymptotic form V (z) = α(α + 1)k 2 . V2 ∼ k 2 . we could replace |r| in A(r) = 2k|r| by either r tanh(µr) or µ−1 log cosh(µr). the potential approaches that of Randall and Sundrum [1]. (4.19) which has the asymptotic form (4. in these cases we cannot calculate z = z(r) and hence V (z) in closed form. 2 where we have introduced the parameters V1 = −V (0).7) for |z|  max(µ−1 . In the limit µ  k the brane is very thin and. k −1 ). The shape of the potential (4. independent of µ. In the latter case  4α log kµ−1 log cosh(µz) + 1 .

The volcano box potential.3. In this limit. Because the potential is exactly zero beyond the barrier the can be no bound-state at zero energy. z1 6 |z| 6 z2 . both V1 and V2 scale like kµ. and a continuum for m2 > 0. Csáki et al. is the possible existence of resonances. Newton’s law and the formulae of Section 3. but the depth and width of the well can be easily arranged such that there is a single bound-state. where k1 = p m2 + V 1 . 3). as in [2]) is:   cos k1 z. (4. k3 = p m2 . / Nuclear Physics B 581 (2000) 309–338 Fig. The solution for a symmetric continuum wavefunction (orbifold boundary conditions. is also interesting.23) ψ(z) = a ek2 z + b e−k2 z . This possibility was also noticed in [51]. |z| > z2 .326 C. µ  k. while z1 asymptotes to r 2 (kµ)−1/2 .22) z1 = 2α + 1 In this limit we expect that the smearing of the matter fields over the transverse direction will become important and that the analysis of the corrections to Newton’s law as described in Section 3. A useful toy model in which the Kaluza–Klein modes can be calculated exactly is the volcano box potential (Fig. It can happen that for some particular energies. 4. Resonant modes The final issue that we mention regarding decoupling. 3.24) . (4. |z| 6 z1 . with a vanishing small energy m2 < 0. k2 = and the coefficients are given by p V 2 − m2 . The other limit µ  k. incident plane waves can resonate with the potential V (z) and consequently have a large value of ψm (0).3 will be valid. where the brane is much thicker than k.  c cos k3 x + d sin k3 x.3 will require some modification. (4.

then we find that there are no resonant modes in the volcano box approximation. . V (z) = 15k 2 − 3kδ(z). of the well part of the potential.25a) b= 2 2  e−k2 (z1 +z2 )  k2 cos k1 z1 e2k2 z1 + e2k2 z2 k3 cos k3 z2 c= 2k2k3    + e2k2 z1 − e2k2 z2 k2 sin k3 z2 + k1 sin k1 z1 e2k2 z1 − e2k2 z2 k3 cos k3 z2   + e2k2 z1 + e2k2 z2 k2 sin k3 z2 . (4. then it is m2 ∼ V2 − (V1 z1 )2 ∼ k 2 .25b) a= d=  e−k2 (z1 +z2 )  k2 sin k1 z1 e2k2 z1 + e2k2 z2 k2 cos k3 z2 2k2k3  − (e2k2 z1 − e2k2 z2 )k3 sin k3 z2  + k2 cos k1 z1 − e2k2 z1 − e2k2 z2 k2 cos k3 z2   + e2k2 z1 + e2k2 z2 k3 sin k3 z2 .25c) Resonant modes occur when the coefficient a of the growing exponential in the region z1 6 |z| 6 z2 . k2 (4. the width 2z1 . or equivalently the depth V1 . (4.C. the contribution of the resonance is negligible for r  z2 V2 − m2 /m2 . for instance.27) 1+ −p 1+ ∼1− p 1− 2V1 V1 V1 V 2 − m2 V 2 − m2 If there is a solution. In order to make contact with smooth versions of the RS model we set V1 z1 = k and V2 = k 2 . up to numerical coefficients. / Nuclear Physics B 581 (2000) 309–338 327   e−k2 z1 ek2 z1 cos k1 z1 − kk12 sin k1 z1 . cos k1 z1 + kk12 sin k1 z1 . Hence. (4. respectively. we find that if the well is deep. cos k1 z1 − k1 sin k1 z1 = 0. However. Notice √ that by (4.e. i. The contribution of the narrow resonance to Newton’s law would be of the form √ 2 e−mr+z2 V2 −m .26) We assume that m2 < V2  V1 . In the RS case the potential is [1].26) in powers of (V1 z1 )2 /V1 and m2 /V1 to obtain. The spacing between resonances would be of order π 2 /z12 . (4. Csáki et al. The smoothings can in general introduce other dimensionful quantities besides k. so if there is a resonance below the barrier height V2 there is only one (for narrow wells).28) U (r) ∼ r p Hence.29) 2 If we naıvely read off the coefficients 3k and 15 4 k of the delta function and barrier height.     m2 V1 z1 m2 V1 z1 (V1 z1 )2 = 0. 4(k|z| + 1)2 (4. vanishes. (4. If we in addition take √ (V1 z1 )2  V1 (or x1 V2  1) then we can expand (4.27) if V1 z1 > V2 there will not be a resonance at all. this matching of the RS case to the volcano box approximation is too glib and a more careful analysis is needed.

The Lykken–Randall scenario One issue that is worth commenting on is the fate of the Lykken–Randall scenario for the solution of the hierarchy problem [17].012 as a result of the weakening effect of the potential barrier at higher energies. We find a sharp resonance near m2 = V2 − (V1 z1 )2 = 9. where T increases to 0. In any case. the barrier height.7). of width (1m2 )/m2 ∼ 10−8 . Away from the resonance T is smaller than O(10−4 ). In order to create the necessary hierarchy of scales we need z0 such that 2 −2A(z0 ) e ∼ TeV2 . this requires that z0 is much larger than k (since k ∼ MPl ).31) .2) but now associate the matter fields that describe our world with another brane — the “TeV brane” — located at a point z0 in the transverse space.4. MPl (4. but in the absence of explicit solutions for the wavefunctions it is difficult to make definite predictions. b0 b0 |ψ ψ (4. the TeV brane sits at a place where we may approximate the potential V (z) by its asymptotic form (4. but below. 4 we plot the “transmission coefficient” T = 1/(c2 + d 2 ) (which is not restricted to T 6 1) into the well versus the energy of the KK mode m2 . z1 = 1/V1 and z2 = 10. there is at most a single resonance at the order of. the effective Newton’s constant on the brane is now given by GN ∼ b0 (z0 )2 M∗2−d ψ . In this example. In other words. In the context of the backgrounds discussed in Section 4. because the resonances give rise to Yukawa type contributions to the gravitational potential. they are irrelevant below a certain scale which for smooth versions of the RS case is expected to be of order the fundamental Planck scale M∗ . in the context of thick branes. 4. as expected. We expect that this will be the case for general smoothings of the RS scenario. except near V2 . in units where the fundamental Planck scale M∗ = 1. we take V1 = 106 . Csáki et al. V2 = 10. For example.328 C. / Nuclear Physics B 581 (2000) 309–338 Fig. In this scenario we consider again the general five-dimensional backgrounds (4. 4.30) where MPl is the Planck mass. Resonance of transmission T into well with Kaluza–Klein mass m. in Fig. In particular. Since the relevant dynamics now takes place at z = z0 we have to re-assess the effects of the KK modes.

As z0 increases then one might worry that the ratio ψm (z0 )/ψ and the effect of the KK modes is less suppressed and there would be large corrections to Newton’s law. also suggests that in a more realistic scenario. it is convenient to use polar coordinates for z = (%. where the matter fields are located at z = z0 but are still pointlike in the fifth dimension.36) ψ0 (%) = exp − 4 .9) to assess the magnitude of the corrections. where the matter fields are smeared out in the fifth dimension. The discussion above. (4. For small enough m the first term in (4.12) (taking into account the correct normalizations on the wavefunctions) ψm (z0 ) ψm (0) ' ∝ mα−1 . 4. we find that the radial function R(%) then satisfies  d −5 0 (d − 2)2 0 2 d − 2 00 R (%) + A (%) − A (%) −R 00 (%) − % 16 4  l(l + d − 6) (d − 2)(d − 5) 0 A (%) + R(%) = m2 R(%). as explained in [17].5. (4.9) dominates.22) are modified to [17] " 2 #  Z∞ M1 M2 −mr ψm (z0 ) 1 + dm e .35) The zero-energy state which potentially describes a four-dimensional graviton is from (3.9). Small enough m in this context means that we can approximate ψm (z0 ) by the first term in (4. we consider the generalization to the cases when the transverse space is more than five dimensional. with l(l + d − 6) Yl (Ω). Actually.12). in the case when the continuum starts at m = 0. the zero-energy wavefunction ψ b0 (z0 ) becomes larger 1. Csáki et al. this is not the case. where Yl (Ω) is an n-dimensional spherical harmonic. the corrections from the KK modes (3. / Nuclear Physics B 581 (2000) 309–338 329 This modifies the effects of the KK modes. Since z0 lies in the tail of the potential (4.32) U (r) ∼ GN b0 (z0 ) r ψ 0 b0 |ψ b0 i = b0 (z) must be normalized to unity hψ In the above. (4.33) b0 (z0 ) b0 (0) ψ ψ and so the suppression is unchanged from the situation where matter fields are located at z = 0. the corrections to Newton’s law will also be similarly suppressed and our previous insistence that the matter sources were point-like and located at z = 0 was not overly simplistic. we can use the exact form of the solution in terms of Bessel functions (4. Localization and decoupling in d > 5 In this section. (4. − 4% %2 (4.18)   d −2 b A(%) . In this case.34) ∇z2 Yl (Ω) = − %2 From (3.7). Ω) and write ψ(z) = R(%)Yl (Ω).C. and one finds using the expansion (4.

where the coordinates {%.4. Asymptotically for large %. it is a simple matter. for l = 0.330 C. On the contrary. just as in the case in d = 5. when ψ In this case the potential in (4.39) ψm (0) ∼ k Plugging this into the correction to Newton’s law. (4.35) falls off as ∼ %−2 . using the formulae of Section 3. / Nuclear Physics B 581 (2000) 309–338 b0 (z) becomes We now want to argue that. It is clear that only the s-wave modes (l = 0) are non-vanishing at the origin. (4. As in d = 5. Although we have only presented a detailed analysis for the conformally-flat backgrounds. we find that the integral over the sb0 is wave KK modes has no singularity if α > (d − 4)/2. the wavefunction ψ normalizable precisely when the KK modes decouple. the most delicate case is b0 (z) has a power-law fall-off ∼ %−α .41) ds 2 = e−A(%) ηab dx a dx b − e−B(%) d%2 + %2 dΩ 2 .1.40) where C and C 0 are some dimensionless numbers and we have assumed that the metric depends upon a single dimensionful parameter k so that GN ∼ M∗2−d k d−4 . Normalizability requires that α > (d − 4)/2. Ω}. In the general case there is no canonical choice for the {zi } coordinates. . Ω} are polar coordinates in n = (d − 4) dimensions.38) We now match the solution to that in region (1) using the same procedure that we followed in Section 4. In this case. Furthermore. to generalize to the arbitrary radially symmetric background having the form  (4. (3) and (4). Finally. matching the condition that ψ normalizable. % %2 (4.  α−d+4 m . we find after normalizing the wavefunctions as plane waves at % → ∞.37) As previously. we will choose the {zi } to 12 We thank Martin Gremm for raising the issue of the l > 0 modes. −R 00 (%) − d −5 0 α(α + 6 − d) R (%) + R(%) = m2 R(%). even for fields transforming nontrivially under SO(n) the sum over l presumably leads to a convergent series. in regions (2). 12 For this reason we shall only consider the l = 0 modes here. We now follow exactly same steps as in Section 4. It turns out that in order to analyze the localization of gravity it is not judicious to choose the {zi } to be the Cartesian coordinates associated to {%. r (kr)2α−d+4 U (r) ∼ GN (4. the radial function satisfies the equation. we can solve this equation in terms of Bessel functions: for l = 0 R(%) = am %3−d/2 Yα+3−d/2(m%) + bm %3−d/2 Jα+3−d/2(m%). and the correction is of the form C M1 M2 M1 M2 + d−2 r M∗ k 2α−2d+8 r 2α−d+5   C0 M1 M2 1+ = GN . we do not expect the matter fields of interest to us to transform under the SO(n) global symmetry and hence they would only couple at tree level to the s-wave modes. Csáki et al.1.

the normalizability condition is Z Z    n b 2 d−5 ˆ ˜ < ∞. the total action of gravity plus the scalar takes the form Z  √  (4. %˜ n−1 d%˜ exp − d−2 d z ψ0 (z) = Vol S 2 A(%) (4. the same replacement A(%) → A( ˜ Consequently.31)     ˜ %) e ˜ = exp − d−2 A( ˆ %) b0 (z) = exp − 3 A( ˜ − d−5 ˜ .42) is conformally-flat and the previous equation for the radial fluctuations for the conformally-flat case (4. In general. ψ s-wave continuum modes are decoupled.43) is identical to that in the conformally-flat case with energy wavefunction ψ ˆ %). / Nuclear Physics B 581 (2000) 309–338 331 be the Cartesian coordinates associated to polar coordinates {%. we investigate whether the three-brane scenario that we have discussed in previous sections can actually be generated by gravity coupled to a single real scalar field. − 4%˜ %˜ 2 e %) ˜ %) When B( ˜ = A( ˜ the metric (4. which leads to a generalization of the radial equation (4. ˜ A( ˜ = d −2 d −2 To be completely explicit.44) (4. Our equation (4. 4. when the conformally-flat case to argue that gravity is localized. Csáki et al.5) for the behaviour of the stress tensor under gravitational fluctuations is. valid for a scalar field.46) matches that derived in [37] for the case when the transverse space is two dimensional. to (4.45) The equation for the fluctuations (3.e. i.43) ψ 4 4 B(%) 4 where we have defined d −5e 3 ˜ ˆ %) A(%) ˜ + B(%). ˜ Ω}. In the new set of coordinates {zi } the wavefunction is (3.42) ηab dx a dx b − d%˜ 2 − e−B(%) ds 2 = e−A(%) ˜ %) ˜ %) where the functions A( ˜ and B( ˜ are related to A(%) and B(%) by the coordinate transformation on the radial coordinate.35) is recovered.46) is identical ˆ %).35) with the replacement A(%) → A( ˜ Moreover the expression for the zerob0 (z) (4. ˜ for which the metric (4.34) can be simplified by separating the variables: ψ(z) = R(%)Y ˜ l (Ω).6. The first issue that we must verify is that our ansatz (3.41) has the form  ˜ ˜ e ˜ 2 %˜ dΩ 2.46) A (%) R(%) ˜ = m2 R(%).C.35):  d −5 0 (d − 2)2 ˆ 0 2 d − 2 ˆ 00 00 R (%) A (%) A (%) ˜ − ˜ + ˜ − ˜ −R (%) %˜ 16 4 e ˜ A( ˜ %) ˜  l(l + d − 6)eB(%)− (d − 2)(d − 5) ˆ 0 ˜ + ˜ (4. we can use the same analysis as in the b0 (z) is normalized. (4.47) S = dd x g − κ −2 R + 12 ∂µ φ∂ µ φ − V(φ) . . Notice that for s-waves the equation for the radial function R(%) ˜ (4. The fact that it is valid follows from the dependence of the action of the scalar field on the metric. involving a new radial coordinate %˜ = %(%). in fact. (4. Thick three-branes from a scalar field In this section.

332 C. This ansatz.47) is  ∂V(φ) 1 √ = 0. −φ 00 (r) + 2 A0 (r) φ 0 (r) + ∂V(φ) = 0.51a) (4. Later. / Nuclear Physics B 581 (2000) 309–338 The stress-tensor for the scalar field is   1 1 1 αβ Tµν = ∂µ φ ∂ν φ − gµν ∂α φ ∂β φ g − V(φ) . ∂φ (4.50a) κ 2 φ 0 (r)2 + 2κ 2 V φ(r) + 6A0 (r)2 − 6A00 (r) = 0.1 and consequently all our previous conclusions regarding the localization of gravity are still valid. since φ = φ(z) only. However.48). completely determines the form of the scalar potential V(φ) and the solution φ(r).52) or. as we will see. Plugging in the ansatz (2. with Tµν as in (4. The graviton equation of motion is. 54].   2 0 2 2 0 2 (4. (4.50b) κ φ (r) − 2κ V φ(r) − 6A (r) = 0. Gµν = κ 2 Tµν . 2 κ 2 φ 0 (r)2 = 3A00 (r).   3 κ 2 V φ(r) = −3A0 (r)2 + A00 (r). The scalar field equation following from the action (4.49) From this and the form of the stress tensor (4. there are two independent components of Einstein’s equation:   (4. (4. 2 2 2 (4. in Section 5. we will will prove that a single real scalar field cannot produce a three-brane in d > 5 dimensions.5) follows immediately. In the coupled system. which itself implies the scalar field equation due to the general covariance of the scalar . the variation of the stress tensor under metric fluctuations (3.3) and φ = φ(r).3) and take φ = φ(r) only. To see this we note that Einstein’s equation implies ∇µ T µν = 0. the analysis of fluctuations is naturally more involved. as usual. It immediately follows that.48) For metric fluctuations hµν which only have a non-vanishing component hab . √ ∂µ g g µν ∂ν φ(r) + g ∂φ (4. we have hµν ∂ν φ = 0.48). The system of scalar fields coupled to gravity was studied in the same context in [47–50. so for the rest of this section we will take d = 5 and take the metric in the form (2.51b) Note that a solution only exists for φ(r) if A00 (r) > 0. Next we consider whether the three-branes can be formed from a single real scalar field.53) One can easily show that the scalar field equation is solved automatically by a solution of Einstein’s equation.2. the fluctuations of the scalar are completely decoupled from the transverse traceless gravitational fluctuations as described in Section 3. Csáki et al. with our ansatz.

Fig.3) is a solution to the gravity and scalar equations-of-motion if the scalar field has the form (4.56). we have in mind smoothings of the multi-dimensional AdS metric in (2.C.51a). 13 Hence. we could take |z| → z tanh(10z). / Nuclear Physics B 581 (2000) 309–338 333 field action.55) (4. (5. For this particular smoothing in d = 6 with k = 1.2) with d > 5. Hence. that we can interpret as a four-dimensional intersection of d − 4 (d − 1)-branes. Gravity localized on thick intersecting branes In this section. For each i. if φ(r) is a strictly monotonic function of r then we can we can implicitly define   (4.   9 ∂W(φ) 2 − 3W(φ)2 .9). we consider the possibility that gravity can be localized on the fourdimensional intersection of higher dimensional branes. Alternatively. 5 shows the potential V (z1 . any solution to Einstein’s equation will automatically be a solution to the scalar field equations. φ 0 (r) = 3κ −1 ∂φ(r) Then we can write the scalar potential as. hence it follows from (4. 5. V[φ] = κ −2 2 ∂φ (4. we can immediately draw on the results of Section 4 to 13 We are grateful to Shanta de Alwis for pointing this out. ψ (i) (zi ) satisfies a one-dimensional Schrödinger equation identical to (4.54) W φ(r) ≡ κ −1 A0 (r). For example. we search for conformally-flat backgrounds (2. z2 ) appearing in the equation for the gravitational fluctuations (3. given the scalar potential V(r). In this case the solution represents n intersecting (d − 2)-branes where the intersection is four-dimensional. however.9) in the case of intersecting branes is fully ndimensional and consequently rather complicated.1) i=1 it is separable by writing ψ(z) = ψ (1) (z1 ) × · · · × ψ (n) (zn ). The ansatz (2.1. any metric of the form (2.11). in the case when A(z) = n X A(i) (zi ).3) with A(z) → A(i) (zi ). Csáki et al.3) for the metric determines that the scalar potential can be written in the supersymmetric form (4. A solvable example In general.51b) and the scalar potential is given by (4.56) We therefore see that the supersymmetric form of the scalar potential and BPS equations introduced in [47. For instance. .48] appear naturally in this approach. the Schrödinger equation (3.51b) that ∂W[φ(r)] . Furthermore. 5. In other words.

6. 2 (1) 1 (2) 2 ψm1 (z )ψm2 (z ). depending on whether each ψ (i) (zi ) is the normalizable wavefunction ψ0(i) (zi ) or a continuum wavefunction (i) ψm (zi ). the spectrum consists of 2n different sectors. when there are 2 transverse directions there will be 4 sectors in the spectrum spanned by (1) (2) (1) ψ0 (z1 )ψ0 (z2 ). In general. so that the zero-energy state (1) (n) ψ0 (z) = ψ0 (z1 ) × · · · × ψ0 (zn ). The set of continuum states (4) contributes to Newton’s law as in (3. Fig. Close-up of the potential near the brane intersection. establish the conditions under which gravity is localized on the intersection. .334 C. The Schrödinger potential V (z1 . (3) ψm 1 (2) (4) (1) (2) 2 ψ0 (z1 )ψm (z ).22). for each i. So for example. / Nuclear Physics B 581 (2000) 309–338 Fig. The first state (1) corresponds to the four-dimensional graviton. (5.11) with k = 1 and |z| → z tanh(10z). 5. Basically we   require that exp − 32 A(i) (zi ) falls off faster than 1/zi as |zi | → ∞. z2 ) for the metric (2. (1) 1 (z )ψ0(2) (z2 ).2) is localized in all the transverse directions and therefore represents the four-dimensional graviton. Csáki et al.

/ Nuclear Physics B 581 (2000) 309–338 Z∞ δ4 U (r) ∼ M∗2−d dm1 dm2 M1 M2 e q − m21 +m22 r r (1) (2) ψm (0)2 ψm (0)2 .7) so the particular combination Gii + G00 = κ 2 (Tii + T00 ) of components of the Einstein equation relates derivatives of the scalar field to derivatives of the metric. we consider whether the conformally-flat intersecting branes background can be generated from a single real scalar field. 2 Tii = −T00 + ∂i φ(z) . r (5. (5.2). (5.     1 1 1 ρ ∂ φ(z) ∂ρ φ(z) − V φ(z) .3) 0 The set of continuum states (2) or (3) are localized in one direction and contribute as an integral over the single eigenvalue m2 and m1 . for instance.9) . (5. 2 (5. independent of the form of the scalar potential V(φ):   2 d − 2 1 1 2 2 κ ∂i φ(z) = ∂i A(z)∂i A(z) + ∂i A(z) . we will also see. for components transverse to the intersection.4) 0 5.   d −2 1 2 ∂i A(z)∂j A(z) + ∂i ∂j A(z) Gij − κ Tij = 2 2 1 − κ 2 ∂i φ(z)∂j φ(z) = 0. In addition. and there is a single scalar field which generates the energy–momentum tensor. We assume the metric has the form (2.C. (5.2). that a single real scalar field cannot produce a conformally-flat three-brane embedded in d > 5.8) 2 2 2 The off-diagonal components of the Einstein equation are.   2 d −2 1 ∂i A(z) + ∂i2 A(z) . Z∞ δ2+3 U (r) ∼ M∗2−d k dm1 M1 M2 e−m1 r (1) 2 ψm1 (0) r 0 Z∞ + M∗2−d k dm2 M1 M2 e−m2 r (2) 2 ψm2 (0) . as promised in Section 4. respectively.2.6. We argue that a single scalar field cannot produce a general intersecting brane background. Csáki et al. but this leaves open the possibility that such backgrounds can be generated from more than one scalar field. 1 2 335 (5. Intersecting branes from scalar fields In this section. a single complex scalar as in [12].6) Gii = −G00 + 2 2 The diagonal components of the stress tensor satisfy a similarly simple relation. and satisfies.5) Tµν = ∂µ φ(z)∂ν φ(z) − gµν 2 2 2 The Einstein tensor Gµν can be read off of (3.

/ Nuclear Physics B 581 (2000) 309–338 or. such as appear in the RS solution to the hierarchy problem.. is an Oppenheimer Fellow at the Los Alamos National Laboratory. and the Kaluza–Klein modes become relevant at long distances. By a redefinition of coordinates. the solution of (5. Generically in that case the effective gravitational coupling on the domain wall vanishes. J. C. Dan Freedman. In particular. It turns out that it is possible in generalizations of the Randall-Sundrum scenario for the graviton zero mode to be nonnormalizable. Asad Naqvi. Salman Habib. Juan Maldacena.10) is.8). Michael Graesser. and in which the domain walls are created by non-dynamical sources in the absence of fields other than the graviton. ! n X i ai z . Conclusions We have studied general features of gravity in domain wall backgrounds. using (5.H. and argued that they are unimportant in smooth versions of the RS scenario. Therefore. We also thank Tanmoy Bhattacharya. C. For general domain wall type metrics we have found the conditions for there to be localized gravity on the domain wall or on the intersection of domain walls. Csáki et al.10) This constrains the type of metric which can be obtained by a gravitating scalar field. is supported by NSF grant PHY-9802484. (5. Michael Nieto and Erich Poppitz for useful discussions.E. and Shanta de Alwis and Christophe Grojean for comments and corrections.C. are supported by the US Department of energy under contract W7405-ENG-36. Martin Gremm. In addition. and T. Fred Cooper. this can always be recast in the form A(z) = A(z1 ). 2     ∂i ∂j eA(z)/2 = ∂i2 eA(z)/2 ∂j2 eA(z)/2 .11) A(z) = A i=1 for some constant coefficients ai .J. Negative tension branes. We have seen several illuminating examples in which smooth domain wall backgrounds are created by scalar fields. . can be studied in this formalism.336 C. We have studied intersecting domain walls in the absence of fields other than gravity.S. We commented on resonant modes in the continuum of Kaluza–Klein modes. Acknowledgements We thank Martin Gremm for sharing an early version of [51] with us prior to publication. Work of Y. which preserves a (d − 1)-dimensional Lorentz symmetry and hence can create only a single (d − 2)-brane in d dimensions.C. we conclude that a single real scalar field cannot produce a p-brane in d > p + 2. 6. and we argued that a single scalar field is not sufficient to produce intersecting domain walls. (5. it seems to be the case that additional fields are required to create intersecting branes in the presence of gravity.

hep-th/9909130. Karch. Chamblin. Phys. H. hep-th/9911237. Kim. C. hep-th/9910132. 381 (1992) 301. hep-th/9909199.G.W. H. Goldberger. / Nuclear Physics B 581 (2000) 309–338 337 References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] L. Smith. S. M. Lett. H. Arkani-Hamed. Phys. hep-th/9201007. D.W. Rev. hep-ph/9907447. Okada. Phys. Grojean.J. M.Z. Kaplan. hep-th/9911104. H. Davoudiasl. Mod. Freedman. hep-th/9909134. Phys. 83 (1999) 4922. Hewett. Rizzo. N. hep-ph/9912552. N.B. hep-th/9906064. Lett. Shirman. Randall. L. Phys. hep-ph/9909523. hep-ph/9909494. Cohen. Lett. A. Trodden. Chaichian. N. Rubakov. hep-th/9909076. Squires. M. hep-th/9912125. Lykken. H. hep-ph/9909496. E. Nucl. D. M. Okada. hep-ph/9912498. JHEP 9911 (1999) 022. Shaposhnikov. S. N. S.C.B. B 125 (1983) 139. J. Sundrum. W. Rev. Mouslopoulos. Arkani-Hamed. Nihei.B. Phys. Phys. Randall. hep-th/9909103. Behrndt. K. A. Wise. Santiago. M. L. Weiner. C. Yamaguchi. A. E. S. . I. Hall. hep-ph/9904383. Cvetic. Yamaguchi. hep-th/9908076. Kraus. Nakano. Rept. hep-ph/9905487. L. hep-th/9910174. hep-th/9909053.D. G. Y. Chang. A 14 (1999) 2393. J. Oda. Phys. hep-ph/9907218. DeWolfe. Ross. hep-th/9909001. Wise. Lett. B 159 (1985) 22. Z. G. Cline. hep-th/9910081. M. T. Dvali. M. Rev. Phys. hep-ph/9912453. A. A. S. B. N. C. R. Lett. N. Halyo. E. M. A. hep-ph/9911262. G. Tachibana. Cline. Takenaga. Kehagias. hep-th/9908186. Chang. Kiritsis. W. hepph/0001017. Rey. Csáki. I. Chodos. Nam. hep-th/9909127. D. Chacko. R. Gubser. D 61 (2000) 024008. Nelson. L. Kaloper. Oda.D. hep-th/9905210. gr-qc/9906106.G. hep-th/9912182. Poppitz.L. B 465 (1999) 81. Wiltshire. Lett. U. J. Nucl. hep-ph/9909555. hep-ph/9908347. 83 (1999) 4690. Rev. Lett. N. J. K. Cvetic. hep-th/9909048. Sakamoto. Nam.M. M. Soleng. Hellerman. M. Goldberger. Carroll. B 167 (1986) 286.E. S.D. S.L. V. Gibbons. Hatanaka. Kogan. J. M. hep-th/9910149. Csáki et al. Cvetic. G. Hisano. JHEP 9912 (1999) 011.B. hep-th/9912232. I. Phys. A. B 287 (1987) 717. Kim. Randall. Phys. S. D.B. Nelson. P. S.A. Hisano. hep-ph/9812365. T. Gibbons. M. Lett. S. M. Ellwanger. O. S. hep-th/9912186. Servant.E. hep-th/9908174. G. Grojean. D 60 (1999) 123506.S. Sundrum. hep-ph/9905221. Mersini. Phys. D 60 (1999) 107505. N. 83 (1999) 3370. H. M. Servant. Phys. Phys.E. hep-ph/9812296. E. Rev. 282 (1997) 159. T. Papazoglou. hep-th/9907209. Kobakhidze. Griffies. hep-th/9909058.G.H. hep-th/9604090. hep-th/9908104. Dimopoulos. Gogberashvili. G. Visser. Gabadadze. Kaloper.I. A. hep-th/9912193. Bajc. G. Kaloper. Li. Gogberashvili. M. Rev. A. J. hep-th/9911083. J.

H. D 61 (2000) 023511. Flanagan. Bakas. C.W. I. Phys.J. Wise. D.K. Cvetic. Langlois. Sfetsos. hep-ph/9908530. hep-th/9912175. Freese.N. E. J.L. hep-th/9912118. Kallosh. Shiromizu. Madea. Olive. Garriga. hep-th/9912160. K. Emparan. Cvetic. Kehagias. P. Dudas.Y. Gubser. D.338 [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] C. E. J. hep-th/0001159. Binetruy. Garriga. Cline. hep-th/9909205. Gherghetta. Sfetsos. Kim. S. Terning. hep-ph/9911406. G. Chamblin. Skenderis. Servant. H. T. T. M. I.E. hep-ph/0001017. Dienes. K. hep-th/9912135. Phys. I. Shiromizu. C. A. Mukohyama. Lett. C. Maeda. Rev. G. hep-th/9911043. Lee. Terning. D. Kang. Lett. Hewett. Linde. Wald. hep-ph/9906513. A. Chamblin. Charmousis. Davoudiasl. C. C. hep-th/9912242. hep-ph/9912344. hep-th/9912132. S. B 469 (1999) 123. S. Goldberger. Kanti. hep-th/9911165. M. Neronov. Mersini. hep-ph/9910235. . Vollick. hep-th/9912001. Freese. Deffayet. N. G. hep-th/9911055. hep-ph/9911457. M. hep-ph/9909481. Reall. Graesser. D.R. hep-th/9912230.E. Randall. R. Wang. Kyae. M. Verlinde. Graesser. R. Deffayet.S. Ivanov.M. T.N. Gremm. hep-th/0001018. 83 (1999) 4245. Dick. D.-H.G. / Nuclear Physics B 581 (2000) 309–338 M. Rubakov. hep-th/9911181. R. K. A. M. K.K. B 468 (1999) 31.I.S. A. S. B. J. Aref’eva. Tanaka. Phys. Pospelov. Brandhuber. K. Leontaris. B 462 (1999) 34. gr-qc/9910076. hep-th/9906182. L. R. Grojean. K. hep-th/9912287. Volovich. hep-th/0001013. C. hep-th/9909070. D. R. 1984. W. University of Chicago Press. Ellwanger. D. Eardley. Sasaki. Phys. Langlois. gr-qc/9911122.V. hep-ph/9909255. J. hep-th/9908116. Csáki. Bakas. L. Sfetsos. T. hep-ph/9906523. T. Lett. Pope.G. H.Y. Rev. J. hep-th/9911218. J. Ida. Mikulovicz. Sasaki. Cvetic. M.M. Youm. G. U. I. P. Lett. M. Myers. Rizzo. A. Horowitz. P. Kolda. R. Behrndt. hep-ph/9906542. J. Csáki. S. hep-th/9912233. hep-th/9909041. C. Chung. Phys. Tye. Phys.J. hep-th/9912060. Hawking. K. Chung. Wasserman. D. M. hep-th/9910269. Kogan. hep-th/9905012. Myung. P. K. M. Lu. K.M.H. Mukohyama. gr-qc/9912002. I. M. A. Townsend. Gregory. hep-th/0001002. hep-th/9910219. K. Csáki et al. hep-th/9912166. hep-th/9912187. A. Brandhuber. I. D. J. JHEP 9910 (1999) 013. hep-th/9906204. K. hep-ph/9910498. hep-th/0001071. H. General Relativity. Lett. Mavromatos. S. Binetruy. hep-ph/9912266. hep-th/0001003. C. B 468 (1999) 46.E. V. H.

uni-leipzig. to decompose bilocal light-ray operators into (harmonic) operators of definite twist is applied to the case of arbitrary 2nd rank tensors.de 0550-3213/00/$ – see front matter  2000 Elsevier Science B. like the Shuryak–Vainshtein. PII: S 0 5 5 0 . Leipzig University. Introduction Any careful quantum field theoretical analysis of the experimental data for the different light-cone dominated QCD-processes (for recent experiments and analyses.  2000 Elsevier Science B. D-04109 Leipzig. All rights reserved. see [1–15]) necessarily requires a rigorous twist decomposition of the various nonlocal light-ray operators appearing in the theoretical set up of these processes.de 2 lazar@itp.3 . As a generic example the bilocal gluon operator is considered which gets contributions of twist-2 up to twist-6 from four different symmetry classes characterized by corresponding Young tableaux. we extend these results to various trilocal light-ray operators.Nuclear Physics B 581 (2000) 341–390 www.uni-leipzig.V. Nonlocal light-cone operators. the three-gluon and the four-quark operators. 02. In addition.22 Keywords: Twist decomposition. Germany Received 13 March 2000. which are required for the consideration of higher-twist distribution amplitudes. accepted 11 April 2000 Abstract A group theoretical procedure. also the twist decomposition of the related vector and scalar operators is considered. All rights reserved.88. The present results rely on the knowledge of harmonic tensor polynomials of any order n which have been determined up to the case of 2nd rank symmetric tensor for arbitrary space-time dimension. The matrix elements of these special operators are related to the distribution amplitudes (or the hadronic wave functions) being necessary for the phenomenological description of the above mentioned 1 geyer@itp. PACS: 12.21]. Tensorial harmonic polynomials 1.nl/locate/npe Twist decomposition of nonlocal light-cone operators II: general tensors of 2nd rank Bodo Geyer 1 . introduced earlier in [20. Augustusplatz 10. Markus Lazar 2 Center for Theoretical Studies and Institute of Theoretical Physics. 13.3 2 1 3 ( 0 0 ) 0 0 2 2 7 .V.38.elsevier.

being given as (infinite) towers of local tensor operators of growing rank. furthermore. it is process-dependent because it is only applicable to matrix elements of the corresponding LC-operators and not to the operators itself. not immediately related to geometric twist. bilocal quark operators up to (antisymmetric) second rank have been considered. if d is taken as scale dimension of the operators it is not related to the generators of the Lorentz group but to those of the subgroup SO(2h − 1. In Section 4 we extend these results with minor modifications to the trilocal light-ray operators mentioned above. 1) ⊗ R+ . Appendix A contains some material about the tensorial harmonic functions. bilocal gluon operators. and by the (anti)symmetrization defined by corresponding Young operators. has been introduced [18.g. For that reason another notion of (dynamical) twist. Thereby. Lazar / Nuclear Physics B 581 (2000) 341–390 processes.. In the case of nonlocal tensor operators. specific Young tableaux. Any of these nonlocal operators consists of an infinite tower of local operators which are characterized by equal twist. M.. their twist content ranges from 2 up to 6. we continue this study for general tensor operators of 2nd rank. Young frames. trilocal quark-gluon correlation operators (related to so-called Shuryak–Vainshtein operators) and four-fermion operator. 3 The twist decomposition of the local operators of some well-defined tensor structure corresponds directly to their decomposition into irreducible tensor representations of the Lorentz group. being related to the LC-quantization in the infinite momentum frame [17]. therefore. their anomalous dimensions determine the Q2 -evolution of the distribution amplitudes. The notion of twist has been originally introduced for local operators by Gross and Treiman [16] as a geometric quantity. The paper is organized as follows. which is directly related to the irreducible representations of the Lorentz group. generalized harmonic polynomials being tensors up to rank 2 are introduced and the interior derivative on the light-cone is used. thereafter cited as I. Geyer. . this different notion of twist has serious theoretical disadvantages: It is not Lorentz invariant and. These representations are uniquely (up to equivalence) determined by their symmetry classes. twist τ = (canonical) dimension d-spin j . In Section 2 we shortly review the general method introduced in Part I. In Section 3 we give the complete twist decomposition of the various gluonic bilocal light-ray tensor operators of second rank being specified by different symmetry classes. 1) ⊗ R+ of the conformal group in 2h-dimensions. We carry out the complete twist decomposition of all relevant LC-operators. the definition of (geometric) twist is more subtle.e. There. as well as multilocal quark and gluon operators. However. Here. e. is quite advantageous. i. Moreover.21]. 3 Obviously.e. i.19] by counting powers of 1/Q which.. Thus a local operator with definite twist is a irreducible finite dimensional representation of the group SO(2h − 1.342 B. In order to be able to disentangle the twist content of nonlocal LC-operators which are relevant for various hard QCD processes a systematic group theoretical study has been started in an earlier papers [20. from a phenomenological point of view.

˜ κ2 x) ˜ = Fα (κ1 x)U ˜ (κ1 x. Geyer. κ2 x) = Φ 0 (κ1 x)Γ U (κ1 x. respectively. σµν . Dirac spinors (d = h − 1/2) as well as gauge field strength (d = h). eventually multiplied by appropriate tensors built up from x: ˜ ˜ κ2 x) ˜ = Fαρ (κ1 x)U ˜ (κ1 x. 2 x) with the path ordered phase factor ( Zκ2 ˜ κ2 x) ˜ = P exp ig U (κ1 x. Φ generically denotes the various local fields. Furthermore. µ (2. Aµ is the gauge potential having dimension d = h − 1.1) ) dτ x˜ Aµ (τ x) ˜ . The unrenormalized nonlocal (flavour singlet) operators at y = 0 are given by ˜ κ2 x) ˜ = ψ(κ1 x)Γ ˜ U (κ1 x.4) eαβ = 1 αβµν F µν are the gluon field strength and its dual. κ )RT OΓ (y + κ1 x. Those nonlocal operators which are relevant for the virtual Compton scattering [22] are obtained by the nonlocal LC expansion [23–25] of the (renormalized) time-ordered product of two electromagnetic hadronic currents:  RT J µ (y + x2 )J ν (y − x2 )S Z  µν ˜ y + κ2 x)S ˜ + ···. [20. the twist decomposition of operators (2. ˜ Gαβ (κ1 x. under renormalization these operators mix with appropriate chiral-even (or chiral-odd) bilocal gluonic tensor operators. γµ γ5 . Lazar / Nuclear Physics B 581 (2000) 341–390 343 2. x 2 →0 where Γ = {1.21. where Fαβ and F 2 Here.24. introduced in Part I.2) κ1 ensuring gauge invariance.g. if necessary. σµν γ5 } indicates the tensor structure of the nonlocal quark operators and x˜ is a light-like vector related to x. suppressing any indices indicating the group representations.5) consists of the following steps: (1) Taylor expansion of the nonlocal operators for arbitrary values of x at the point y = 0 into an infinite series of local tensor operators having definite rank n and canonical dimension d: . e. Let us denote such operators for arbitrary values of x as follows: OΓ (κ1 x. Now. of how to decompose arbitrary bilocal LC-operators into harmonic operators of definite twist. the twist decomposition of an arbitrary bilocal light-ray operator may be formulated for any space-time dimension D = 2h. the phase factors are to be taken in the adjoint representation.. eαβ (κ1 x. ˜ κ2 x) ˜ F ˜ G (2. Γ labels the tensor structure as well as additional quantum numbers. κ2 x)Φ(κ2 x) (2. ˜ κ2 x)F ˜ βρ (κ2 x). γµ . ρ eβρ (κ2 x).26]. As is well-known.B. scalars (d = h − 1). cf. In general. ≈ d2 κ CΓ (x 2 . General procedure Let us now shortly review the general procedure.5) where.23.3) (2. M. γ5. (2. ˜ κ2 x)ψ(κ ˜ ˜ OΓ (κ1 x.

← (2. . thus simplifying the explicit computations below. Lazar / Nuclear Physics B 581 (2000) 341–390 OΓ (κ1 x. (iv) j = m − 2. Geyer. . denote symmetrization and antisymmetrization with respect to that pattern. .7) ← y D µ = ∂µ − igAµ (y). reducible higher twist operators. the orthogonal group SO(2h). . in general.344 B. These higher twist operators again have to be decomposed according to the considered symmetry classes. which because of the tracelessness are restricted by `1 + `2 6 2h (`: length of columns of [m]). then 4 The corresponding representations are related through analytic continuation. . At the end of the computations the general values of κi may be reinstalled. m − 2. [20. 4 These representations are built up by traceless tensors of rank m whose symmetry class is determined by some (normalized) Young operators Y[m] = (f[m] /m!)QP. .6) with the generalized covariant derivatives ← ↔ → D µ (κ1 . are for a fixed value of m: (i) j = m. . where [m] = P (m1 .21. .. κ2 x) = ∞ X 1 µ1 x · · · x µn n! n=0   ↔ ↔ × Φ 0 (y)Γ D µ1 (κ1 . the lower spins in the above series correspond to the various trace terms which define. (Here. M. . κ2 ) · · · D µn (κ1 . mr ) with m1 > m2 > · · · > mr and ri=1 mi = m denotes the corresponding Young pattern.6) into irreducible components is independent of the special values of κ1 . m − 3. cf. we may choose κ1 = 0. . (ii) j = m − 1. (2) Decomposition of local operators with respect to irreducible tensor representations of the Lorentz group SO(2h − 1.6). Now. P and Q. . κ2 )Φ(y) y=0 (2. At first. . (iii) j = m − 2. κ2 . m − 3. m2 . . m − 3. The allowed Young patterns for SO(2h). if these irreducible local tensor operators are multiplied by x µ1 · · · x µn .28] . . equivalently. . m − 4. (2. . two remarks are in order. as usual. . this decomposition of the local operators appearing in (2. Furthermore.) Loosely speaking. we depicted only those patterns which appear for h = 2 where `1 + `2 6 4. (2. . m − 4. → → y D µ = ∂µ + igAµ (y).6). as it is necessary according to Eq. m − 4. κ2 = κ and factor out κ n for every term in Eq. Therefore. κ2 ) ≡ κ1 D µ + κ2 D µ . m − 2. 1) or.27.

Twist decomposition of a general 2nd rank tensor operator This section is devoted to the twist decomposition of a general 2nd rank tensor operator. For simplicity. ˜ leads to the required lightcone operator with well defined geometric twist.8) Obviously. to the usual ones: . these nonlocal operators of definite twist are Lorentz covariant tensors (contrary to the phenomenological concept of twist). That infinite tower of local operators creates a nonlocal operator of definite twist. i. the twist decomposition is independent from the dimension 2h of space-time. Furthermore. The nonlocal operators OΓτ (0. we get the (infinite) twist decomposition of the nonlocal operator we started with: OΓ (0.9) (4) Finally. this twist decomposition is unique. since the harmonic tensor polynomials essentially depend on (infinite) sums of powers of x 2 and  as well as some specific differential operators in front of it. τmax M τ =τmin OΓτ (κ1 x.10) The resulting light-ray operators of definite twist. which have been rewritten for general arguments.B. including the higher twist contributions resulting from the trace terms. in that limit only a finite number of terms survive: ˜ κ2 x) ˜ = OΓ (κ1 x. Let us remark that step (3) and (4) can be interchanged without changing the result. OΓτ (0. we make the special choice κ1 = 0. ˜ (2. κx) are harmonic tensor functions. For a detailed exposition of these objects. 3. Furthermore.. ˜ κ2 x). κx) = ∞ M κn n=0 n! OΓτ n (x).3). (3) Resummation of the local operators OΓτ n (x) belonging to the same symmetry class (for any n) and having equal twist τ . Geyer. the projection onto the light-cone. we restrict ourselves to four-dimensional space-time (h = 2). (2.e. (2. In addition. x).7). from a group theoretical point of view. Eq. Lazar / Nuclear Physics B 581 (2000) 341–390 345 harmonic (tensorial) polynomials of order n appear whose remaining tensor structure results from Γ . Second. However. κ2 = κ. x) = ∞ M τ =τmin OΓτ (0. The advantages of our method are: First. putting together all contributions. (2. M. are tensor functions on the light-cone which fulfill another kind of tracelessness conditions to be formulated with the help of the interior (on the cone) derivative. see Appendix A. thereby also simplifying the generalized covariant derivatives. Now. process. since the twist decomposition is independent from the chirality we demonstrate it only for the chiral-even gluon operator (2. x → x. this nonlocal operator is built up as direct sum of irreducible local tensor operators.and model-independent.

Geyer. κx) = Fαρ (0)U (0. Lazar / Nuclear Physics B 581 (2000) 341–390 Gαβ (0.346 B. κx)Fβρ (κx) = ∞ X κn n=0 n! x µ1 · · · x µn Gαβ(µ1 ···µn ) . M.

y y with Gαβ(µ1 ···µn ) = Fαρ (y)D(µ1 · · · Dµn ) Fβρ (y).

. ˜ this infinite series terminates at least at τ = 6 since higher order terms are proportional to x 2 . κx) ≡ Gα• (0. In fact.3) such representations are listed down which correspond to higher twist contributions contained in the trace terms of symmetry classes (i)–(iv). n−2 2 2 . j2 ) of the Lorentz group is      1 1 1 1 n n n−2 n−2 ⊕ ··· 2.1) here the symbol (· · ·) denotes symmetrization of the enclosed indices.e. after projection onto the light-cone.y=0 .6) . according to the spin content and the rank of the corresponding local tensor operators. having scale dimension d. of the group SL(2. as well as for later use. (3. 2).4) (3.2 ⊕ 2 .2) is n + 4.2) with the (nontrivial) normalization coefficients f[m] /m! of the Young operators given by αn+1 = 2(n + 1)/(n + 2) for [m] = (n + 1. the local operators with well-defined twist are irreducible tensors of the Lorentz group or. κx) ± Gβα (0. The canonical (or scale) dimension d of the local operator (3. the twist decomposition of these bilocal light-ray operators reads: 5 1 1 5 We use the notation A (αβ) ≡ 2 (Aαβ + Aβα ) and A[αβ] ≡ 2 (Aαβ − Aβα ).5) (3. βn = 3n/(n + 2) for [m] = (n.1) for arbitrary x contains contributions of twist ranging from τ = 2 until τ = ∞. In the last line of Eq. κx). In order to be more explicit. (3.3) ⊕ n2 . 2 ⊗ 2. 2    n n−2 n−2 ⊕ n−2 ⊕ ···. 2 the corresponding tensor spaces will be denoted by T(j1 .. κx). 2 β ± (0. the nonlocal operator (3. κx) . κx) := x G± G± α αβ (0. 2 ⊕ 2 .3) ranges from n + 2 up to 1 or 0 if n is even or odd. C) ⊗ R+ . κx) := G(0. 2   n−2 n+2 ⊕ n−2 ⊕ n+2 2 . m2 > 3): (i) (ii) Gαβ(µ1 ···µn ) = Gαβ(µ1 ···µn ) + αn+1 Gαβ(µ1 ···µn ) (iv) + βn G(iii) αβ(µ1 ···µn ) + γn Gαβ(µ1 ···µn ) + · · · . 1. (3. j2 ). (3. 2 ⊗ 2 . 2 ⊕ 2. κx) := x x α β G+ αβ (0. respectively. and the spin of the various contributions in (3. for m1 + m2 = n + 2. The corresponding Clebsch–Gordan series in terms of representations (j1 . x → x. G± αβ (0. However.. The local tensor operators Gαβ(µ1 .µn ) decompose according to the Young patterns (i) to (iv) and possible higher ones (i. 1) and γn = 4(n − 1)/(n + 1) for [m] = (n. 2     n+2 n n n+2 ⊕ n+2 ⊕ n2 . we introduce the (anti)symmetrization with respect to α and β and we define the following nonlocal operators:  1 Gαβ (0. equivalently. In general. 1). 2 2 . 2 2 . therefore. n2 = n+2 2 . (3.

κ x) [α•] (0.1. κ x). κ x) (α•) (0. κ x) (αβ) (0. κ x) ˜ =G (3. G Gn+2 (x) ≡ x µ x ν Fµρ (0)(xD)n Fνρ (0). κ x).1.1. ˜ G(0. κ x) (α•) (0. (A. κ x) [αβ] (0.11) The various twist contributions individually decompose further according to the symmetry classes which may contribute.8) ˜ = Gtw2 ˜ + Gtw3 ˜ + Gtw4 ˜ G+ α (0. M. κ x). x | Gn+2 (x) ≡ (n + 2)!k! 4 n+2 ˚ n+2 (x) = H (4) G n+2 k=0 . 2 ).10) (3. (3. κ x).9) (0. which have twist τ = 2 and are contained in the tensor n+2 space T( n+2 2 . κ x). Let us postpone the determination of the trace terms and make the resummation to the corresponding nonlocal operator in advance. κ x) [α•] (0. Lazar / Nuclear Physics B 581 (2000) 341–390 347 G+ ˜ = Gtw2 ˜ + Gtw3 ˜ + Gtw4 ˜ (αβ) (0. κ x) [αβ] (0. ˚ n+2 (x) is a harmonic polynomial of order n + 2. tw2 (3. and Young patterns (ii) and (iii) contribute to the antisymmetric tensor operators and related vector operators. where it will be shown that Young patterns (i) and (ii) as well as (iv) contribute to the symmetric tensor operators and related vector and scalar operators. ˜ = Gtw3 ˜ + Gtw4 ˜ G− α (0. 3. αβ (0. They have symmetry class (i) and are uniquely characterized by the following standard tableau (with normalizing factor 1): α β µ1 ··· ∧ µn = ρ S αβµ1 ···µn Fα (0)(Dµ1 · · · Dµn Fβρ )(0) − trace terms. cf. Special tensor operators are considered in Section 4. Geyer.7) ˜ = Gtw3 ˜ + Gtw4 ˜ + Gtw5 ˜ G− [αβ] (0. at first. κ x) αβ (0. κ x) (αβ) (0. This is obtained. κ x) (3. Eq. ∂α ∂β G (n + 2)(n + 1) with the denotation ˚ n+2 (x) = Gn+2 (x) − trace terms. The explicit expressions for generic tensor operators are given in the following subsections. κ x) ˜ + Gtw6 ˜ + Gtw5 (αβ) (0. κ x) (α•) (0. by contracting with x µ1 · · · x µn and rewriting in the following form: µ1 µn tw2(i) Gtw2(i) (αβ)n (x) = x · · · x G(αβµ1 ···µn ) = 1 ˚ n+2 (x). and their contractions with x.5). Appendix A: Here. Tensor operators of symmetry class (i) 3. G [ 2 ] X (n + 2 − k)!  −x 2 k  2 k Gn+2 (x). Construction of nonlocal symmetric class-(i) operators of definite twist Let us start with the simplest case of the totally symmetric traceless tensors.B. κ x) (αβ) (0. The irreducible local twist-2 operator reads ρ Gtw2(i) (αβµ1 ···µn ) = F(α (0)Dµ1 · · · Dµn Fβ)ρ (0) − trace terms.

κx) + ∞ Z X k=1 0 1  −x 2 dt t 4 k   1 − t k−1 k G(0. because of the appearance of the derivatives. 6 Here.17) we may recognize that in the case of symmetry class (i) the tensor and vector operators are obtained from the scalar operator by very simple operations. vector operator. we omit the indication of the symmetry class of the nonlocal twist-2 operators because only totally symmetric tensors contribute. (3.17) which.2. (3.15) and (3. κx) = 0. Gtw2 (0. we observe how in the case of the scalar operator the trace terms — being proportional to x 2 — are to be subtracted from G(0. ∂ α Gtw2 (α•) (0. κx) in order to make that operator traceless.13) k!(k − 1)! t The operator Gtw2 (αβ) (0. using ((n + 2)(n + 1))−1 = 0 dλ (1 − λ)λn and the integral representation of Euler’s beta function we obtain the nonlocal twist-2 tensor operator: 6 Gtw2 (αβ) (0. and in the following. (3.1. κx) = G(0. the trace terms being of higher twist must be classified according to their symmetry type.16) Multiplying with x α x β gives the twist-2 scalar operator. Comparing Eqs. κλx) gives the twist-2 x β and observing the equality (x∂x )G(0. Geyer.348 B.12) 0 with ˚ G(0. However. In the case of vector and tensor operators such subtraction. Furthermore. M.30]. satisfies the condition Gtw2(0.12) by ˚ ˚ κλx) = (λ∂λ + 2)G(0. (3. by definition. (3. (3. κx) = 0. Multiplying Eq.14) 3. (3. κtx). Lazar / Nuclear Physics B 581 (2000) 341–390 R1 Then. (3. κx) = 0. is more complicated. ∂ α Gtw2 (αβ) (0. Gtw2 (αβ) (0. (3. Reduction to vector and scalar operators Now.12).18) The expression (3. κx) = 0. (3.13) for the scalar operator has already been given by Balitsky and Braun [29. κx) = G(0. κx) = ∞ X κn n=0 n! Gtw2(i) (αβ)n (x) = ∂α ∂β Z1 ˚ dλ (1 − λ)G(0. . ˚ κx). κx) = 0. κx) = 0. we specify to the nonlocal vector and scalar operators. Z1 Gtw2 (α•) (0. κx) = ∂α ˚ dλ λG(0. κλx).15) 0 which satisfies the conditions  Gtw2 (α•)(0. κx) satisfies the conditions of a harmonic tensor function: g αβ Gtw2 (αβ) (0. κλx).

also extend to the case of general values (κ1 . The final expression for the symmetric twist-2 light-cone tensor operator is given by Z1 ˜ κ2 x) ˜ = ∂α ∂β Gtw2 (αβ) (κ1 x. (3. Geyer.1. .13) contribute. M. 2 in Eq. appearing in Eq. Lazar / Nuclear Physics B 581 (2000) 341–390 349 3. κ2 ). Because of the derivatives ∂α and ∂β . only the terms with k = 1.3. Projection onto the light-cone Let us now project onto the light-cone and.B. (3. at the same time.12).

κ2 λx). dλ (1 − λ)G(κ1 λx.

19) 0 ˜ κ2 x) ˜ = G>(i) (αβ) (κ1 x.    1 dλ (1 − λ + λ ln λ) gαβ + x(α ∂β)  2 0  .x=x˜ − G>(i) ˜ κ2 x). ˜ (αβ) (κ1 x. (3.

 1 + 2(1 − λ) + (1 + λ) ln λ xα xβ 2 G(κ1 λx. κ2 λx).

19). respectively. it is necessary to subtract its own trace terms which are of twist τ = 6. respectively. (3. To ensure that the vector operator ∂β G(κ1 x. G(κ1 x. κ2 x)|x=x˜ and 2 G(κ1 x. κ2 x)|x=x˜ also obtains well-defined twist. like ∂β G(κ1 x. Here. (3. an operator being contained in the trace terms is called a scalar and vector part. like G(κ1 x. (3.21) with ˜ κ2 x) ˜ = Gtw4(i)a (αβ) (κ1 x. κ2 x)|x=x˜ . κ2 x). be subtracted from the first term. occurring in Eq. 7 The scalar operators. (3. κ2 x). ˜ κ2 x). tw6(i)a tw6(i)b ˜ κ2 x) ˜ + G(αβ) (κ1 x. and a vector. ˜ κ2 x) ˜ G(αβ) (κ1 x. κ2 x) or ∂β ∂ µ Gµ (κ1 x.20) 4 Z1 ˜ κ2 x) ˜ contains the higher twist contributions which have to The operator G>(i) (αβ) (κ1 x.20) already have well-defined twist τ = 4 and τ = 6. The higher twist operators contained in the trace terms of the twist-2 tensor operator are the following: >(i) tw4(i)a tw4(i)b ˜ κ2 x) ˜ = G(αβ) (κ1 x. if the expression multiplied by gαβ . xα or xβ is a scalar. Eq. κ2 x) or ∂ µ Gµ (κ1 x. ˜ + G(αβ) (κ1 x.20) consists of scalar and vector parts.x=x˜ . In order to disentangle the different terms of well-defined twist we observe that the tensor operator (3. ˜ κ2 x) ˜ + G(αβ) (κ1 x. 1 gαβ  2 Z1 . in order to make the whole expression traceless. respectively.

κ2 λx). dλ (1 − λ + λ ln λ)G(κ1 λx.

Z1 .x=x˜ . 0 tw4(i)b ˜ κ2 x) ˜ = x(α ∂β)  G(αβ) (κ1 x.

dλ (1 − λ + λ ln λ)G(κ1 λx. κ2 λx).

x=x˜ 0 tw6(i)b − G(αβ) (κ1 x. x and x as well as g α β αβ which are necessary for the dimension and tensor structure of the whole expression do not belong to the higher twist operator itself. ˜ 7 Strictly speaking. . ˜ κ2 x).

1 xα xβ 2 4 Z1 . Lazar / Nuclear Physics B 581 (2000) 341–390 and tw6(i)a ˜ κ2 x) ˜ = G(αβ) (κ1 x. M. Geyer.350 B.

κ2 λx).  dλ 2(1 − λ) + (1 + λ) ln λ G(κ1 λx.

4  (1 − λ)2 1 − λ2 − − λ ln λ λ 2λ .x=x˜ . 0 1 tw6(i)b ˜ κ2 x) ˜ = xα xβ 2 G(αβ) (κ1 x.

κ2 λx). × G(κ1 λx.

n2 ) and T( n−2 2 .x=x˜ . the local twist-4 n−2 and twist-6 operators carry the representation T( n2 . Finally. 2 ).20) ˜ κ2 x) ˜ = Gtw4(i)a ˜ κ2 x) ˜ + Gtw4(i)b ˜ κ2 x) ˜ Gtw4(i) (αβ) (κ1 x. (αβ) (κ1 x. Z1 = 0    1 dλ (1 − λ + λ ln λ) gαβ + x(α ∂β)  2    1 (1 − λ)2 1 − λ2 2 − − λ ln λ xα xβ  − 4 λ 2λ . by inspection of the spin content of the symmetry type (i).  Z1 dλ 0 Obviously. we write down the symmetric twist-4 light-cone tensor operator contained in the expression (3. respectively. (αβ) (κ1 x. completing the twist decomposition of the symmetric nonlocal tensor operator.

× G(κ1 λx, κ2 λx)

x=x˜ ,

(3.22)

whereas the symmetric twist-6 light-cone operator contained in (3.20) reads
tw6(i)

tw6(i)a

tw6(i)b

˜ κ2 x)
˜ = G(αβ) (κ1 x,
˜ κ2 x)
˜ + G(αβ) (κ1 x,
˜ κ2 x)
˜
G(αβ) (κ1 x,
1
= xα xβ 2
4 

Z1

(3.23) 

1 − λ2
+ ln λ G(κ1 λx, κ2 λx)

x=x˜ .

0

3.1.4. Vector and scalar light-ray operators
Contracting Eq. (3.19) with x˜ β , making use of formula (x∂x )f (λx) = λ∂λ f (λx) and
performing the partial integrations we obtain the final version of the twist-2 light-cone
vector operator,
Z1
˜ κ2 x)
˜ =
Gtw2
(α•) (κ1 x,  

1
dλ λ ∂α + (ln λ)xα  G(κ1 λx, κ2 λx)

x=x˜ ,
2

(3.24)

0

which has already been used in [22], and the twist-4 light-cone vector operator,
1
tw4(i)
˜ κ2 x)
˜ = − xα 
G(α•) (κ1 x,
2

Z1
0

dλ λ(ln λ)G(κ1 λx, κ2 λx)

x=x˜ .

(3.25)

B. Geyer, M. Lazar / Nuclear Physics B 581 (2000) 341–390

351

In order to obtain the scalar twist-2 light-ray operator we multiply (3.24) by x˜ α . Then,
the twist-4 part vanishes and the remaining twist-2 operator restores the scalar operator
(compare Eq. (3.11)),
˜ κ2 x)
˜ = G(κ1 x,
˜ κ2 x).
˜
Gtw2 (κ1 x,

(3.26)

˜ κ2 x),
˜ is a
Let us point to the fact that the trace of the original gluon tensor, g αβ Gαβ (κ1 x,
tw4(i)a
˜ κ2 x)
˜ as well as similar expressions
twist-4 scalar operator. It is contained in G(αβ) (κ1 x,
occurring below, cf. also Eq. (3.83).
3.1.5. Condition of tracelessness on the cone
Finally, it should be remarked, that the conditions (3.14), (3.16) and (3.18), if translated
into the corresponding ones containing derivatives with respect to x,
˜ no longer hold for
the light-cone operators. This is clear because, by projecting onto the light-cone, part
of the original structure of the operators has been lost. Nevertheless, the conditions of
tracelessness of the light-cone operators may be formulated by using the interior derivative
on the light-cone [31,32] which has been extensively used for the construction of local
conformal operators [33–36]. In four dimensions it is given by 

1
dα ≡ 1 + x˜ ∂˜ ∂˜α − x˜ α ∂˜ 2
2

with


∂˜α ≡ α ,
∂ x˜

and has the following properties
d2 = 0, 

[dα , dβ ] = 0 and dα x˜ 2 = x˜ 2 dα + 2∂˜α .

Then, the conditions of tracelessness simplify, namely, they read:
˜ κ2 x)
˜ = 0,
g αβ Gtw2
(αβ) (κ1 x,

dα Gtw2
˜ κ2 x)
˜ = 0,
(αβ) (κ1 x,

as well as
˜ κ2 x)
˜ = 0.
dα Gtw2
(α•) (κ1 x,
Analogous conditions hold for the light-cone operators of definite twist in the case of
symmetry classes (ii)–(iv) obtained below.
In passing we remark that the interior derivative may be used in defining more directly
totally symmetric local light-cone operators [37]. For example it holds
1
dα dβ Gn+2 (x),
˜
(n + 2)2 (n + 1)2
1
˜ =
dα Gn+2 (x).
˜
Gtw2(i)
(α•)n (x)
(n + 2)2
tw2(i)

˜ =
G(αβ)n (x)

However, in case of symmetry types (ii)–(iv) the corresponding expressions should be more
involved.

352

B. Geyer, M. Lazar / Nuclear Physics B 581 (2000) 341–390

3.2. Tensor operators of symmetry class (ii)
3.2.1. Construction of nonlocal (anti)symmetric class-(ii) operators of definite twist:
Young tableau A
Now we consider tensor operators, and their contractions with x, having symmetry class
n
n n+2
(ii) and whose local twist-3 parts are contained in T( n+2
2 , 2 ) ⊕ T( 2 , 2 ). Contrary to
the totally symmetric case we have different possibilities to put the tensor indices into the
corresponding Young pattern. Without presupposing any symmetry of indices α and β we
should start with the following Young tableau:
µ1 µ2
α

···

∧ 2(n + 1)
µn β =
A
n + 2 αµ1

S

βµ1 ···µn

Fα ρ (0)

× Dµ1 · · · Dµn Fβρ (0) − trace terms,
with normalizing factor αn+1 . (The tableau with α ↔ β will be considered thereafter.)
Denoting this symmetry behaviour by (iiA) we may write the local twist-3 tensor operator
as follows:
(
1
tw3(iiA)
F ρ (0)D(µ1 · · · Dµn ) Fβρ (0) + Fαρ (0)Dβ D(µ2 · · · Dµn Fµ1 )ρ (0)
Gαβµ1 ···µn =
n+2 α
+

n
X
l=2

Fα ρ (0)D(µ1 · · · Dµl−1 D|β| Dµl+1 · · · Dµn Fµl )ρ (0)
)

− (α ↔ µ1 ) − trace terms.
Proceeding in the same manner as in the last subsection we multiply by x µ1 · · · x µn and
obtain: 

1
˚ µ•|n+1 (x)
(x) =
δ µ (x∂) − x µ ∂α ∂β G
Gtw3(iiA)
αβn
n(n + 2) α
with
˚ µν|n (x)
˚ µ•|n+1 (x) ≡ x ν G
G    

1
1 2
(4)
α
α
x

(x∂)



= δµα −
x
Hn+1 x 2 | Gα•|n+1 (x).
µ
µ
2
(n + 2)
2
˚ µ•|n+1 (x) is the harmonic vector polynomial of order n + 1 (see Eq. (A.7),
Here, G
Appendix A) and the symmetry behaviour of classR(ii) is obtained through the differential
1
operator in front of it. Now, using ((n + 2)n)−1 = 0 dλ (1 − λ2 )λn /(2λ) and Euler’s beta
function we sum up to obtain the following nonlocal twist-3 tensor operator:
(0, κx) =
Gtw3(iiA)
αβ

Z1
dλ 

1 − λ2 µ
˚ µ• (0, κλx),
δα (x∂) − x µ ∂α ∂β G

0

with the nonlocal traceless (vector) operator

(3.27)

B. Geyer, M. Lazar / Nuclear Physics B 581 (2000) 341–390

353    

−x 2 k
1 − t k−1 
k
dt
Gα• (0, κtx)
4
k!(k − 1)!
t
k=1 0  

1 2
µ
µ
− xα ∂ (x∂) − x ∂α ∂
2  

2 k k 
Z1
∞ Z1
X
1−t k
−x 

dτ τ dt t
Gµ• (0, κτ tx).
(3.28)
×
4
k!k!
t

˚ α• (0, κx) = Gα• (0, κx) +
G

k=0 0

∞ Z
X

1

0

The following decomposition into symmetric and antisymmetric part is useful for the
further calculations
tw3(iiA)

Gαβ

tw3(iiA)

(0, κx) = G[αβ]

tw3(iiA)

(0, κx) + G(αβ)

(0, κx),

(3.29)

with
(0, κx) =
Gtw3(iiA)
[αβ]

Z1

µ
˚ µ• (0, κλx),
dλ λδ[α ∂β] G

(3.30)

0

(0, κx) =
Gtw3(iiA)
(αβ)

dλ  

1 µ
1 − λ2
˚ µ• (0, κλx)
δ(α ∂β) −
∂α ∂β x µ G
λ
2

(3.31)

dλ 

1 − λ2 µ
˚ µ• (0, κλx).
δ(α (x∂) − x µ ∂(α ∂β) G

(3.32)

Z1
0

Z1
=
0

Again, these operators are harmonic tensor functions. The conditions of tracelessness
for the tensor operator are
tw3(iiA)

g αβ Gαβ

tw3(iiA)

∂ α Gαβ

(0, κx) = 0,

(0, κx) = 0,

tw3(iiA) 

Gαβ

(0, κx) = 0,

tw3(iiA)

∂ β Gαβ

(0, κx) = 0.

(3.33)

3.2.2. Reduction to vector operators
The corresponding twist-3 vector operator is obtained from Eq. (3.27) by multiplication
with x β :
Z1
(0, κx) =
Gtw3(iiA)
α• 

˚ µ• (0, κλx).
dλ λ δαµ (x∂) − x µ ∂α G

(3.34)

0

Obviously, a corresponding scalar operator does not exist.
This vector operator fulfills the following conditions of tracelessness
(0, κx) = 0, 
Gtw3(iiA)
α•

∂ α Gtw3(iiA)
(0, κx) = 0.
α•

(3.35)

354

B. Geyer, M. Lazar / Nuclear Physics B 581 (2000) 341–390

3.2.3. Projection onto the light-cone
(κ1 x,
˜ κ2 x)
˜ on the light(a) The calculation of the antisymmetric tensor operator Gtw3(iiA)
[αβ]
tw2
˜ κ2 x)
˜ in Part I; for the details we refer to it. The
cone is similar to that of M[αβ] (κ1 x,
resulting expression is:
(κ1 x,
˜ κ2 x)
˜ =
Gtw3(iiA)
[αβ]

Z1

µ
dλ λδ[α ∂β] Gµ• (κ1 λx,
˜ κ2 λx)
˜

x=x˜

0
>(iiA)

− G[αβ] (κ1 x,
˜ κ2 x),
˜
˜ κ2 x)
˜ =
G>(iiA)
[αβ] (κ1 x,

1
2

Z1

(3.36)


µ 
dλ(1 − λ) 2x[α ∂β] ∂ µ − x[α δβ]  Gµ• (κ1 λx, κ2 λx)

x=x˜ 0 .

κ2 λx). − (1 − λ + λ ln λ)x[α ∂β] G(κ1 λx.

obtain >(iiA) tw4(iiA)a ˜ κ2 x) ˜ = G[αβ] G[αβ] (κ1 x. For their twist decomposition one has to take into account Young pattern (i) as well as (ii). ˜ κ2 x) ˜ + G[αβ] (κ1 x.37) contains two vector operators multiplied by xα and xβ . not appear due to x[α xβ] = 0. The higher twist operator (3. ˜ κ2 x) ˜ + Gtw5(iiA)a [αβ] (3. (3. The procedure is analogous to the decomposition ˜ κ2 x) ˜ made in Part I. ˜ κ2 x) ˜ (κ1 x. After a straightforward calculation we of the vector operator Oα (κ1 x. ˜ κ2 x) ˜ = G[αβ] 1 x[α ∂β] ∂ µ 2 Z1 dλ . tw4(iiA)b (κ1 x.38) with tw4(iiA)a (κ1 x. respectively.37) >(iiA) ˜ κ2 x) ˜ contains twist-4 and twist-5 contributions.x=x˜ . but twist-6 contributions do G[αβ] (κ1 x.

κ2 λx). 1 − λ2 Gµ• (κ1 λx.

x=x˜ . ˜ κ2 x) ˜ = − x[α ∂β]  Gtw4(iiA)b [αβ] Z1 2  . λ 0 1 (κ1 x.

κ2 λx). 1 − λ2 + λ ln λ G(κ1 λx.

dλ 2λ  0 tw5(iiA)a G[αβ] (κ1 x.x=x˜ . ˜ κ2 x) ˜  µ = −x[α δβ] (x∂) − x µ ∂β]  Z1 dλ .

κ2 λx). (1 − λ)2 Gµ• (κ1 λx.

(3. after some lengthy but straightforward calculation (taking into account only the relevant terms of the k-summation and performing some partial integrations). we determine the symmetric tensor operator Gtw3(iiA) (αβ) cone.31). 4λ 0 (κ1 x. Putting Eq.28) into (3. we get the following result: . ˜ κ2 x) ˜ on the light(b) Now.x=x˜ .

M. Geyer. ˜ κ2 x) ˜ = Z1 dλ 355 .B. Lazar / Nuclear Physics B 581 (2000) 341–390 tw3(iiA) G(αβ) (κ1 x.

 1 − λ2 µ δ(α (x∂) − x µ ∂(α ∂β) Gµ• (λx. κ2 λx).

 Z1 = dλ 1 − λ2 1−λ µ gαβ ∂ µ + δ xβ)  + (1 − λ)x(α ∂β) ∂ µ 2λ 2λ (α 0  (1 − λ)2 µ xα xβ ∂  Gµ• (κ1 λx.x=x˜ 2λ 0 >(iiA) − G(αβ) (κ1 x. ˜ κ2 x). − ˜ κ2 λx) ˜ 4λ      1 − λ2 1 1 1 − λ2 + λ ln λ gαβ  + (1 − λ) + + λ ln λ x(α ∂β)  − 2 2λ 2 4λ  . ˜ (3.39) with the higher twist contributions of the trace terms >(iiA) ˜ κ2 x) ˜ G(αβ) (κ1 x.

  2 2 .

(1 − λ) 1 1−λ − + λ ln λ xα xβ 2 G(κ1 λx. κ2 λx) .

.

˜ κ2 x) ˜ + G(αβ) tw6(iiA)a (κ1 x. ˜ κ2 x) ˜ tw6(iiA)d3 tw6(iiA)d4 (κ1 x. ˜ κ2 x) ˜ (κ1 x.40) + 4 2λ λ x=x˜ It is obvious that Eq. twist-5 and twist-6 operators.40) contains scalar and vector operators.41) with 1 tw4(iiA)a (κ1 x. ˜ κ2 x) ˜ + G(αβ) (κ1 x. (3. This procedure is ˜ κ2 x) ˜ in Part I. we can decompose the vector part appearing in Eq. . (3. ˜ κ2 x) ˜ G>(iiA) (αβ) (κ1 x. ˜ κ2 x) ˜ + Gtw6(iiA)c (κ1 x. using Young patterns (i) and (ii) and subtracting the trace terms. Again. analogous to the twist decomposition of the (vector) quark operator Oα (κ1 x. In that manner we get the following decomposition: ˜ κ2 x) ˜ = Gtw4(iiA)a (κ1 x. ˜ κ2 x) ˜ + Gtw4(iiA)b (κ1 x. ˜ κ2 x). ˜ + Gtw6(iiA)e (αβ) (3. ˜ κ2 x) ˜ + G(αβ) (κ1 x. (3. ˜ κ2 x) ˜ (κ1 x. From the twist-4 and twist-5 vector part we have to subtract their trace terms being of twist-6 and add it to the other twist-6 scalar operator. ˜ κ2 x) ˜ = − gαβ  G(αβ) 2 Z1 0  . (αβ) (αβ) (κ1 x. ˜ κ2 x) ˜ + G(αβ) (κ1 x. Let us recall that the scalar twist-4 and twist-6 operators are already traceless on the cone. ˜ κ2 x) ˜ + Gtw4(iiA)c (αβ) (αβ) tw4(iiA)d2 + G(αβ) tw5(iiA)d + G(αβ) tw4(iiA)e (κ1 x.40) into twist-4. ˜ κ2 x) ˜ + Gtw6(iiA)b (αβ) (αβ) tw6(iiA)d1 + G(αβ) tw6(iiA)d2 (κ1 x. ˜ κ2 x) ˜ + Gtw4(iiA)d1 (κ1 x. ˜ κ2 x) ˜ + G(αβ) (κ1 x.

κ2 λx). 1 − λ2 + λ ln λ G(κ1 λx.

dλ 2λ  .x=x˜ .

˜ κ2 x) ˜ = Z1 1 − x(α ∂β)  2   1 − λ2 + 2λ ln λ dλ (1 − λ) + 2λ 0 . M.356 B. Geyer. Lazar / Nuclear Physics B 581 (2000) 341–390 tw4(iiA)b G(αβ) (κ1 x.

tw6(iiA)b × G(κ1 λx. κ2 λx).

˜ κ2 x) ˜ = − x(α ∂β)  Gtw4(iiA)d1 (αβ) 2  Z1 dλ  .x=x˜ − G(αβ) (κ1 x. ˜ κ2 x). ˜ 1 (κ1 x.

κ2 λx). 1 − λ ln λ + G(κ1 λx.

˜ κ2 x) ˜ = G(αβ) 1 xα xβ 2 4 Z1  (1 − λ)2 1 − λ2 − λ ln λ λ 2λ . ˜ and tw6(iiA)a (κ1 x.x=x˜ λ λ 0 tw6(iiA)d1 − G(αβ) (κ1 x. ˜ κ2 x).

× G(κ1 λx. κ2 λx).

x=x˜ .  dλ 0  Z1 (1 − λ)2 3 1 − λ2 − 2λ 2 2λ 0  .

ln λ G(κ1 λx. κ2 λx).

x=x˜ . ˜ κ2 x) ˜ = xα xβ 2 dλ G(αβ) 4 λ λ 2λ 0 . − λ ln λ − 2λ 1   Z 1 1 − λ ln λ ln2 λ tw6(iiA)d1 + + (κ1 x.

κ2 λx). × G(κ1 λx.

˜ κ2 x) ˜ = − xα xβ 2 Gtw6(iiA)d3 (αβ) 2 dλ Z1  1 − λ ln λ ln2 λ + + λ λ 2λ . 1 (κ1 x. ˜ κ2 x) ˜ = − xα xβ 2 Gtw6(iiA)b (αβ) 4 1 (κ1 x.x=x˜ .

× G(κ1 λx. κ2 λx).

˜ κ2 x) ˜ = Gtw4(iiA)c (αβ) 1 gαβ ∂ µ 2 Z1 dλ .  dλ 0 ˜ κ2 x). ˜ as well as being related to the scalar operator G(κ1 x.x=x˜ . (κ1 x.

1 − λ2 Gµ• (κ1 λx. κ2 λx).

λ 0 (κ1 x. ˜ κ2 x) ˜ = x(α ∂β) ∂ µ Gtw4(iiA)d2 (αβ) Z1  .x=x˜ .

κ2 λx). 1 − λ ln λ + Gµ• (κ1 λx.

˜ κ2 x) ˜ = x(α ∂β) ∂ µ G(αβ) Z1 0 . ˜ κ2 x).x=x˜ dλ λ λ  0 tw6(iiA)d2 − G(αβ) (κ1 x. ˜ tw4(iiA)e (κ1 x.

dλ (1 − λ)Gµ• (κ1 λx. κ2 λx).

x=x˜ .

Geyer. ˜ (αβ) and  1 µ (κ1 x.B. M. Lazar / Nuclear Physics B 581 (2000) 341–390 357 − Gtw6(iiA)e (κ1 x. ˜ κ2 x). ˜ κ2 x) ˜ = − x(α δβ) (x∂) − x µ ∂β)  Gtw5(iiA) (αβ) 2  Z1  .

1 − λ ln λ + Gµ• (κ1 λx. κ2 λx).

˜ κ2 x) ˜ − Gtw6(iiA)d4 (κ1 x. × dλ λ λ 0 (κ1 x.x=x˜ . ˜ κ2 x) ˜ = − xα xβ ∂ µ G(αβ) 4 Z1 dλ . ˜ − Gtw6(iiA)d3 (αβ) (αβ) and 1 tw6(iiA)c (κ1 x. ˜ κ2 x).

κ2 λx). (1 − λ)2 Gµ• (κ1 λx.

λ 0 1 (κ1 x. ˜ κ2 x) ˜ = − xα xβ ∂ µ Gtw6(iiA)d2 (αβ) 2 Z1 0 1 (κ1 x. ˜ κ2 x) ˜ = − xα xβ ∂ µ Gtw6(iiA)d4 (αβ) 2  1 − λ ln λ ln2 λ + + λ λ 2λ .x=x˜ .

× Gµ• (κ1 λx. κ2 λx).

 dλ  Z1 dλ  .x=x˜ .

1 − λ ln λ + Gµ• (κ1 λx. κ2 λx).

λ λ 0 1 (κ1 x. ˜ κ2 x) ˜ = xα xβ ∂ µ Gtw6(iiA)e (αβ) 4 Z1 dλ .x=x˜ .

(1 − λ)2 Gµ• (κ1 λx. κ2 λx).

42) 0 Its decomposition into symmetric and antisymmetric part yields (0. δβ (x∂) − x µ ∂β ∂α G 2λ (3.2. Contributions of Young tableau B Let us now consider the other Young tableau which is obtained from the former one by exchanging α and β: µ1 µ2 β ··· µn α ∧ 2(n + 1) = A S Fα ρ (0)Dµ1 · · · Dµn Fβρ (0) n + 2 βµ1 αµ1 ···µn − trace terms.x=x˜ . The corresponding nonlocal twist-3 operator which we obtain after analogous calculations is (with self-explaining denotations): (0. λ 0 ˜ κ2 x).43) . κλx). This symmetry behaviour will be denoted by (iiB). κx). κx) + Gtw3(iiB) (0. Gtw3(iiB) αβ [αβ] (αβ) with (3. κx) = Gtw3(iiB) (0.4. ˜ being related to the vector operator Gµ• (κ1 x. κx) = Gtw3(iiB) αβ Z1 dλ  1 − λ2 µ ˚ •µ (0. 3.

Lazar / Nuclear Physics B 581 (2000) 341–390 tw3(iiB) G[αβ] (0.43) resulting from the Young patterns (iiA) and (iiB). Since no further Young pattern contributes to twist-3 operators we omit the index (ii). dλ λδ[β ∂α] G (3.5.45) 0 The projection onto the light-cone and the calculation of the higher twist operators contained in the trace terms is completely analogous to case (A) and should be omitted here. After projection onto the light-cone the final result is ˜ κ2 x) ˜ = Gtw3 ˜ κ2 x) ˜ + Gtw3 ˜ κ2 x).2. κλx).46) with the antisymmetric twist-3 light-cone tensor operator 8 Z1 Gtw3 ˜ κ2 x) ˜ =2 [αβ] (κ1 x. M.29) and (3. 3. (3. ˜ Gtw3 αβ (κ1 x. Geyer. Construction of the complete twist-3 tensor operators on the light-cone In order to obtain the complete twist-3 operator it is necessary to add both twist3 operators (3. (αβ) (κ1 x.44) 0 tw3(iiB) G(αβ) Z1 (0. κx) = dλ  1 − λ2 µ ˚ •µ (0. δ(α (x∂) − x µ ∂(α ∂β) G 2λ (3. . κλx). κx) = Z1 µ ˚ •µ (0.358 B. [αβ] (κ1 x.

.

Z1 . tw5(ii) ˜ κ2 x) ˜ = −x[α G[αβ] (κ1 x. [αβ] (κ1 x. ˜ [αβ] (κ1 x. (3. κ2 λx) x=x˜ 0 − Gtw4(ii) ˜ κ2 x) ˜ − Gtw5(ii) ˜ κ2 x). dλ λ∂[α G− β] (κ1 λx.47) where ˜ κ2 x) ˜ = x[α ∂β] ∂ µ Gtw4(ii) [αβ] (κ1 x.

1 − λ2 − Gµ (κ1 λx. κ2 λx).

48) 0  µ δβ] (x∂) − x µ ∂β] Z1 × dλ dλ . λ (3.x=x˜ .

(1 − λ)2 − Gµ (κ1 λx. κ2 λx).

x=x˜ . 2λ (3. . we used the abbreviation G− ≡ x ν G [µν] instead of G[µ•] in order not to come into conflict with the µ antisymmetrization of the indices α and β.49) 0 The local twist-4 operators are contained in the tensor space T( n2 . n−2 2 ) ⊕ T( 2 . ˜ κ2 x) ˜ is the In turn. n2 ) and the local twist-5 n−2 n operators are contained in the tensor space T( n2 . let us remark that the bilocal light-ray operator Gtw3 [αβ] (κ1 x. same as the antisymmetric tensor operator which obtains from the Young tableau β µ1 µ2 α ··· µn 8 Here. 2 ).

Geyer. ˜ κ2 x) ˜ are let us note that the local operators contained in M[αβ] (κ1 x.B. M. That tableau has been used in Part I for the computation of M[αβ] ˜ Furthermore. tw2 tw3 ˜ κ2 x) ˜ and in G[αβ] (κ1 x. dλ . 1 ˜ κ2 x). However.36] by means of the interior derivative. the symmetric twist-3 light-cone tensor operator is much more involved.46) could be obtained much easier. From the above point of view our result for the antisymmetric part of twist-3 operator (3. in complete agreement with the local operators on the light-cone determined by Dobrev and Ganchev [35. It is given by 9 Z1 Gtw3 ˜ κ2 x) ˜ = (αβ) (κ1 x. Lazar / Nuclear Physics B 581 (2000) 341–390 359 tw2 (κ x.

 1 − λ2 µ .

˜ κ2 x) ˜ includes all the trace terms having higher twist. κ2 λx) x=x˜ 2λ 0 >(ii) − G(αβ) (κ1 x. δ(α (x∂) − x µ ∂(α ∂β) G+ µ (κ1 λx. these operators are where G(αβ) given by tw4(ii)a ˜ κ2 x) ˜ = −gαβ  G(αβ) (κ1 x.50) >(ii) (κ1 x. ˜ (3.  Z1 dλ  . ˜ κ2 x).

κ2 λx). 1 − λ2 + λ ln λ G(κ1 λx.

  1 − λ2 + 2λ ln λ dλ (1 − λ) + 2λ Z1 0 . 2λ 0 ˜ κ2 x) ˜ = −x(α ∂β)  Gtw4(ii)b (αβ) (κ1 x.x=x˜ .

tw6(ii)b ˜ κ2 x). ˜ × G(κ1 λx. κ2 λx).

x=x˜ − G(αβ) (κ1 x.  Z1 dλ  . tw4(ii)d1 ˜ κ2 x) ˜ = −x(α ∂β)  G(αβ) (κ1 x.

1 − λ ln λ + G(κ1 λx. κ2 λx).

1 xα xβ 2 2 Z1 0 tw6(ii)b G(αβ) 1 (κ1 x.x=x˜ λ λ 0 tw6(ii)d1 − G(αβ) (κ1 x. ˜ κ2 x) ˜ = − xα xβ 2 2  (1 − λ)2 1 − λ2 − − λ ln λ dλ λ 2λ . ˜ κ2 x). ˜ and ˜ κ2 x) ˜ = Gtw6(ii)a (αβ) (κ1 x.

κ2 λx). × G(κ1 λx.

µ (1 − λ)2 3 1 − λ2 − 2λ 2 2λ  .   Z1 dλ 0 9 Similarly.x=x˜ . we use the notation G+ instead of G (µ•) .

ln λ G(κ1 λx. κ2 λx).

− λ ln λ − 2λ .x=x˜ .

Geyer. ˜ κ2 x) ˜ = 1 xα xβ 2 2   .360 B. M. Lazar / Nuclear Physics B 581 (2000) 341–390 tw6(ii)d1 G(αβ) (κ1 x.

1 − λ ln λ ln2 λ + + dλ G(κ1 λx. κ2 λx).

λ λ 2λ Z1 0 tw6(ii)d3 ˜ κ2 x) ˜ = −xα xβ 2 G(αβ) (κ1 x.x=x˜ .   .

κ2 λx). 1 − λ ln λ ln2 λ + + dλ G(κ1 λx.

λ λ 2λ Z1 0 being related to the scalar operator G(κ1 x.x=x˜ . Z1 dλ . κ2 x). as well as tw4(ii)c ˜ κ2 x) ˜ = gαβ ∂ µ G(αβ) (κ1 x.

κ2 λx). 1 − λ2 + Gµ (κ1 λx.

 Z1 dλ  .x=x˜ . λ 0 tw4(ii)d2 ˜ κ2 x) ˜ = 2x(α ∂β) ∂ µ G(αβ) (κ1 x.

1 − λ ln λ .

κ2 λx) x=x˜ λ λ 0 tw6(ii)d2 − G(αβ) (κ1 x. ˜ ˜ κ2 x) ˜ = 2x(α ∂β) ∂ µ Gtw4(ii)e (αβ) (κ1 x. + G+ µ (κ1 λx. Z1 . ˜ κ2 x).

.

dλ (1 − λ)G+ µ (κ1 λx. ˜ and  µ ˜ κ2 x) ˜ = −x(α δβ) (x∂) − x µ ∂β)  Gtw5(ii) (αβ) (κ1 x. ˜ κ2 x). κ2 λx) x=x˜ 0 tw6(ii)e − G(αβ) (κ1 x.  Z1 × dλ  .

1 − λ ln λ .

2 Z1 dλ . λ λ 0 tw6(ii)d3 − G(αβ) tw6(ii)d4 (κ1 x. + G+ µ (κ1 λx. ˜ κ2 x) ˜ − G(αβ) (κ1 x. κ2 λx) x=x˜ . ˜ and 1 ˜ κ2 x) ˜ = − xα xβ ∂ µ Gtw6(ii)c (αβ) (κ1 x. ˜ κ2 x).

(1 − λ)2 + Gµ (κ1 λx. κ2 λx).

x=x˜ . λ 0 (κ1 x.  1 − λ ln λ ln2 λ + + λ λ 2λ . ˜ κ2 x) ˜ = −xα xβ ∂ µ Gtw6(ii)d2 (αβ) Z1 0 tw6(ii)d4 ˜ κ2 x) ˜ = −xα xβ ∂ µ G(αβ) (κ1 x.

.

 dλ Z1 0   . × G+ µ (κ1 λx. κ2 λx) x=x˜ .

1 − λ ln λ .

κ2 λx) x=x˜ . λ λ . + dλ G+ µ (κ1 λx.

Geyer. ˜ κ2 x) ˜ = 1 xα xβ ∂ µ 2 Z1 dλ 361 . Lazar / Nuclear Physics B 581 (2000) 341–390 tw6(ii)e G(αβ) (κ1 x.B. M.

κ2 λx). (1 − λ)2 + Gµ (κ1 λx.

51) where Z1 ˜ 2 x) ˜ Gtw3 (α•) (κ1 x. Determination of complete light-cone vector operators Now we are able to determine the twist-3 vector operator resulting from the tensor operator (3. Because neither symmetry type (iii) nor (iv) may contribute to the vector operator this will be the final expression.46) by multiplying with x β .x=x˜ . κ2 x). (3. ˜ Gtw3 α (κ1 x.2.6. [α•] (κ1 x.κ .κ =   dλ λ δαµ (x∂) − x µ ∂α − xα ∂ µ + ln λx µ 0 Z1 ˜ 2 x) ˜ = Gtw3 [α•] (κ1 x. (α•) (κ1 x. 3. The twist-3 light-cone vector operator consists of two parts. λ 0 being related to the (symmetric) vector operator G+ µ (κ1 x. one originating from the symmetric and the other from the antisymmetric tensor operator. ˜ κ2 x) ˜ = Gtw3 ˜ κ2 x) ˜ + Gtw3 ˜ κ2 x).

.

× G+ µ (κ1 λx.κ2 λx) x=x˜ . .

  .

dλ λ δαµ (x∂) − x µ ∂α − xα ∂ µ G− µ (κ1 λx. (α•) (κ1 x. κ2 λx) x=x˜ . Z1 .54) where ˜ κ2 x) ˜ = xα Gtw4(ii) (α•) (κ1 x. ˜ Gtw4(ii) α [α•] (κ1 x. ˜ κ2 x) ˜ = Gtw4(ii) ˜ κ2 x) ˜ + Gtw4 ˜ κ2 x).52) (3. (3. (3.53) 0 and the twist-4 vector operator which is contained in the trace terms of that twist-3 vector operator is given by (κ1 x.

  .

κ2 λx) x=x˜ .55) . (3. dλ λ ∂ µ + (ln λ)x µ G+ µ (κ1 λx.

.

(α•) (κ1 x. dλ λ ∂ µ G− µ (κ1 λx. 0 The antisymmetric part of the twist-4 vector operator is already complete. (3.56) 0 Z1 ˜ κ2 x) ˜ = xα Gtw4 [α•] (κ1 x. [α•] (κ1 x. κ2 λx) x=x˜ . Z1 = xα   . The complete twist-4 light-cone vector operator is given by ˜ κ2 x) ˜ = Gtw4 ˜ κ2 x) ˜ + Gtw4 ˜ κ2 x) ˜ Gtw4 α (κ1 x.

κ2 λx). ln λ µ x Gµ (κ1 λx.

57) 0 where the symmetric twist-4 vector operator obtains by adding together expressions (3.54). dλ λ ∂ µ + 2 (3. .x=x˜ .25) and (3.

Z1 = xα (3. (α•) (κ1 x.362 B. Lazar / Nuclear Physics B 581 (2000) 341–390 Gtw4 ˜ κ2 x) ˜ = Gtw4(i) ˜ κ2 x) ˜ + Gtw4(ii) ˜ κ2 x) ˜ (α•) (κ1 x. Geyer.58)   . M. (α•) (κ1 x.

κ2 λx). ln λ µ + µ dλ λ ∂ + x Gµ (κ1 λx.

n2 ) is given by: tw4(iii) Gαβµ1 ···µn = n F[α| ρ (0)D(µ1 · · · Dµn ) F|β]ρ (0) n+2 − F[α| ρ (0)D(β µ2 · · · Dµn ) F|µ1 ]ρ (0)  − F[µ1 | ρ (0)D(α Dµ2 · · · Dµn ) F|β]ρ (0) − trace terms.61) . (3.9) and (3.52). (3. ˜ G[α•] (κ1 x. (3. dλ λδ[α (x∂)δβ] − 2x ν ∂β] G (3.56).x=x˜ . κx) = Z1 0  µ ν ˚ [µν] (0. (3. κλx). [α•] (κ1 x. (α•) (κ1 x.58) and (3. [α•] (κ1 x. we obtain for the complete decomposition of the light-cone vector operator the following results anticipated by Eqs. (3.59) ˜ κ2 x) ˜ = Gtw3 ˜ κ2 x) ˜ + Gtw4(ii) ˜ κ2 x).1. respectively: ˜ κ2 x) ˜ = Gtw2 ˜ κ2 x) ˜ + Gtw3 ˜ κ2 x) ˜ + Gtw4 ˜ κ2 x). ˜ G(α•)(κ1 x. (α•) (κ1 x. putting together expressions (3. The corresponding traceless local operator having twist τ = 4 and being contained in the tensor space T( n2 . contracting this expression with x µ1 · · · x µn we obtain: µ1 µn tw4 Gtw4(iii) [αβ]n (x) = x · · · x G[αβ]µ1 ···µn  µ 1 ν ˚ [µν]n (x) (x∂)δ[β − 2x ν ∂[β δα] G n+2 with the (antisymmetric) tensorial harmonic polynomials of order n (see Eq. Construction of the nonlocal antisymmetric class-(iii) operators of definite twist Now we consider the twist-4 and twist-5 contributions originating from the (unique) Young tableau related to the symmetry class (iii) (with the normalizing factor βn ): ··· µ1 µ2 µ3 µn β α ∧ = 3n A S Fα ρ (0)Dµ1 · · · Dµn Fβρ (0) n + 2 αβµ1 µ1 ···µn − trace terms.60) 3.10). Now. (α•) (κ1 x.3.3.8)):   1 [µ [µ ˚ [αβ]n (x) = δαµ δ ν + 2x[α δβ] ∂ ν] (x∂) − x 2 ∂[α δβ] ∂ ν] G β (n + 1)n   2 x[α ∂β] x [µ ∂ ν] Hn(4) x 2 | G[µν]n (x).53). (A. Tensor operators of symmetry class (iii) 3. 2 0 Furthermore. respectively. (3. − (n + 2)(n + 1)n = Resumming these local terms gives the nonlocal twist-4 operator as follows tw4(iii) G[αβ] (0.24).

∂ α Gtw4(iii) [αβ] (0.2. Eq. κx) (3. by construction.64) 3. it is a harmonic tensor operator: Gtw4(iii) [αβ] (0. κx) G     ∞ Z X 1 − t k−1 dt −x 2 k k G[αβ] (0. (3.62). As is easily seen by partial integration it fulfills the following relation tw3 ˚ Gtw4(iii) [αβ] (0. (3. we introduced: ˚ [αβ] (0. because of Eq. After the same calculations as has been carried tw3 (0. (3. κx) − G[αβ] (0. M. κx) = 0. κx) = G[αβ] (0.B. κx) + t 4 k!(k − 1)! t 1 k=1 0 Z1 + dτ τ   [µ [µ 2x[α δβ] ∂ ν] (x∂) − x 2 ∂[α δβ] ∂ ν] ∞ Z X  1 dt k=0 0 0 [µ ν] − 2x[α ∂β] x ∂ ∞ Z X 1 k=0 0  −x 2 dt t 4 k −x 2 4 k   k 1 − t k k!k! t    1 − t k+1 k G[µν] (0. Projection onto the light-cone Let us now project onto the light-cone. κx) = 0. using the integral representation of the additional factor 1/n (the remaining factors of the denominator are taken together with 1/n! to get 1/(n + 2)! and of the beta function.61) defines a nonlocal operator of twist-4. Z1 . (3.3. κ x) ˜ in Part I we obtain out for the operator M[αβ] tw4(iii) ˜ κ2 x) ˜ = G[αβ] (κ1 x.33) and the properties of (3. Lazar / Nuclear Physics B 581 (2000) 341–390 363 where. κτ tx). Geyer. κtx) = G[αβ] (0.63) and.62) (k + 1)!k! t Then.

 µ ν dλ λδ[α (x∂)δβ] − 2x ν ∂β] G[µν] (κ1 λx. κ2 λx).

˜ κ2 x). ˜ (3. namely tw5(iii) ˜ κ2 x) ˜ =− G[αβ] (κ1 x.65) where the twist-5 part is determined by the trace terms. Z1 dλ 0  1 − λ2  [µ [µ x[α δβ] (x∂) − x [µ ∂β] ∂ ν] − x[α δβ] x ν]  λ .x=x˜ 0 tw5(iii) −G[αβ] (κ1 x.

(3.66) × G[µν] (κ1 λx. κ2 λx).

The complete antisymmetric twist-4 light-cone tensor operator obtains from Eq. Determination of the complete antisymmetric light-cone tensor operator This finishes the computation of twist contributions of the antisymmetric light-cone tensor operator.x=x˜ . 3.3. (3. Eq.48): .65) and the twist-4 trace terms of the twist-3 operator. (3.3. Obviously. there exist no vector (and scalar) operators of symmetry type (iii).

Lazar / Nuclear Physics B 581 (2000) 341–390 Gtw4 ˜ κ2 x) ˜ = Gtw4(ii) ˜ κ2 x) ˜ + Gtw4(iii) ˜ κ2 x) ˜ [αβ] (κ1 x. M. [αβ] (κ1 x. [αβ] (κ1 x. Geyer. Z1 =   µ ν dλ λ (x∂)δ[β − 2x ν ∂[β δα] 0   ν] 1 − λ2 [µ [µ ν]  [µ ν] [µ x[α ∂β] x ∂ + x[α δβ] (x∂) − x ∂β] ∂ − x[α δβ] x  − λ .364 B.

κ2 λx).67) × G[µν] (κ1 λx. (3.

49) and (3. [αβ] (κ1 x. Z1 = dλ 0  1 − λ [µ [µ x[α δβ] x ν]  − 2x[α δβ] (x∂) − x [µ ∂β] ∂ ν] λ . [αβ] (κ1 x.66) as follows: ˜ κ2 x) ˜ = Gtw5(ii) ˜ κ2 x) ˜ + Gtw5(iii) ˜ κ2 x) ˜ Gtw5 [αβ] (κ1 x. (3. the complete antisymmetric twist-5 light-cone tensor operator obtains from Eqs.x=x˜ . Furthermore.

κ2 λx). × G[µν] (κ1 λx.

4. These local tensor operators have twist-4 and transform according to the representation n−2 n−2 n+2 T( n+2 2 . Construction of the nonlocal symmetric class-(iv) operator of definite twist Now let us investigate the symmetric tensor operators of the symmetry class (iv) which are given by the following Young tableau (with normalizing factor γn ): ··· µ1 µ2 µ3 α β µn ∧ = 4(n − 1) A A S S n + 1 αµ1 βµ2 αβ µ1 ···µn × Fα ρ (0)Dµ1 · · · Dµn Fβρ (0) − traces.µn ) − trace terms. Eq.4.69) 3.68) Together with the twist-3 part. Tensor operators of symmetry class (iv) 3. ∂β) + x µ x ν ∂α ∂β G δαµ δβν x ρ (x∂)∂ρ − 2x µ (x∂)δ(α (n + 1)n . 2 ) ⊕ T( 2 .1. They are given by: Gtw4(iv) αβµ1 ···µn = 1 G(αβ)(µ1 ···µn ) − G(αµ2 )(µ1 βµ3 ···µn ) − G(µ1 β)(αµ2 ···µn ) (n + 1)n  + G(µ1 µ2 )(αβµ3 .47). 2 ). Multiplying with x µ1 · · · x µn we obtain: µ1 µn tw4(iv) Gtw4(iv) (αβ)n (x) = x · · · x Gαβµ1 ···µn =  1 ν ˚ (µν)n (x). (3. ˜ G[αβ] (κ1 x.8)): ˜ κ2 x) ˜ = Gtw3 ˜ κ2 x) ˜ + Gtw4 ˜ κ2 x) ˜ + Gtw5 ˜ κ2 x). (3. [αβ] (κ1 x. with G(αβ)(µ1 ···µn ) ≡ F(α| ρ (0)D(µ1 · · · Dµn ) F|β) ρ (0). [αβ] (κ1 x.x=x˜ ... we finally obtain the complete decomposition of the antisymmetric light-cone tensor operator (compare Eq. [αβ] (κ1 x. (3. (3.

κx) + G(αβ) (0. G Appendix A):  n ˚ (αβ)n (x) = δαµ δ ν + gαβ x (ν ∂ µ) G β (n + 2)2  1 (ν µ) (ν 2nx(α δβ) − ∂ − x 2 δ(α ∂β) ∂ ν) (n + 2)(n + 1)  1 2nx(α ∂β) x (ν ∂ µ) − x 2 ∂α ∂β x (ν ∂ µ) − (n + 2)2 (n + 1)   1 2 (n + 3)n µ ν x x − g x + α β αβ ∂ ∂ (n + 2)2 (n + 1)2 2    1 1 4 2 µ ν ˘ x − ∂ x + ∂ ∂ ∂ 2x ∂ G(µν)n (x). Eq.B. G (3. κtx) k!(k + 1)! t    − 2x(α ∂β) x (µ ∂ ν) (x∂) − gαβ x (µ ∂ ν) (x∂)(x∂ + 1) − x 2 ∂α ∂β x (µ ∂ ν)   x2 − xα xβ − gαβ ∂ µ ∂ ν (x∂)(x∂ + 3) 2    Z1  x4 2 µ ν dλ (1 − λ) − 2x ∂(α xβ) + ∂α ∂β ∂ ∂ 4 Z1 dλ λ 0 0 × ∞ Z1 X k=0 0  −x dt t 4 2 k  1−t k (k + 1)!k! t k+1 ˘ (µν) (0. α β (α β) 4 (n + 2)2 (n + 1)2 Resumming all local R 1 operators of symmetry class (iv) and using the integral representation ((n + 1)n)−1 = 0 dλ λn (1 − λ)/λ gives  µ ν ρ µ ν µ ν Gtw4(iv) (αβ) (0. κx) is given by where G ˚ (αβ) (0. Here. κλtx).71) . κx) = δα δβ x (x∂)∂ρ − 2x (x∂)δ(β ∂α) + x x ∂α ∂β Z1 × dλ 1−λ ˚ G(µν) (0. λ (3. κtx) =G t 4 k!(k − 1)! t 1 k=1 0  (µ (µ − 2x(α δβ) ∂ ν (x∂) − x 2 ∂(α δβ) ∂ ν) × ∞ Z X 1 k=0 0  −x 2 dt t 4 k    k 1 − t k+1 ˘ G(µν) (0. (A. Lazar / Nuclear Physics B 581 (2000) 341–390 365 ˚ (µν)n (x) is the harmonic symmetric tensor polynomial of order n (cf. κλx). Geyer. κx) G     ∞ Z X 1 − t k−1 ˘ dt −x 2 k k ˘ (αβ) (0.9).70) 0 ˚ (µν) (0. M.

κλx) + dλ ∂(α G β) λ 0 Z1 dλ 1−λ ˚ ∂α ∂β G(0.70) we may be rewritten as follows ˚ Gtw4(iv) (αβ) (0. κλx). not being proportional to gαβ or xα and xβ . i. Contrary to any of the formerly obtained symmetric twist-4 tensor operators.366 B.72) 3. κx) = 0. g αβ Gtw4(iv) (αβ) (0. Projection onto the light-cone The symmetric twist-4 tensor operator of symmetry type (iv) on the light-cone is obtained as follows  ν ˜ κ2 x) ˜ = δαµ δβν x ρ (x∂)∂ρ − 2x µ (x∂)δ(β ∂α) + x µ x ν ∂α ∂β Gtw4(iv) (αβ) (κ1 x. κx) = 0. Lazar / Nuclear Physics B 581 (2000) 341–390 Of course. κx) = G(αβ) (0. λ 0 which will be used in the further considerations. Obviously. Z1 × dλ . this operator is the “true tensor part”. κx) − 2 Z1 1 ˚ + (0.e.. Eq. ∂ α Gtw4(iv) (αβ) (0.4. κx).2. M. Geyer. this twist-4 operator satisfies the conditions of tracelessness Gtw4(iv) (αβ) (0. (3. (3.

κ2 λx). 1−λ >(iv) G(µν) (κ1 λx.

 Z1 = dλ 0  ν µ 2λx(α δβ) ∂ − (1 − λ)xα xβ ∂ µ ∂ ν G(µν) (κ1 λx.x=x˜ − G(αβ) (κ1 x. κ2 λx) + 2λ 2λ 2 2λ  . κ2 λx) λ λ 2λ     (1 − λ)2 1 1 − λ2 1−λ gαβ  + x(α ∂β)  − + ln λ xα xβ 2 G(κ1 λx. ˜ λ (3. ˜ κ2 x).73) 0 where the trace terms of higher twist are given by >(iv) ˜ κ2 x) ˜ G(αβ) (κ1 x. κ2 λx)  1 − λ2 (1 − λ)2 1 µ µ µ gαβ ∂ + x(α δβ)  − xα xβ ∂ G+ − µ (κ1 λx.

 .

κ2 λx) . λ ln λ 1 gαβ (x∂) − λx(α ∂β) − xα xβ  Gρ ρ (κ1 λx.

.

˜ κ2 x) ˜ (κ1 x. ˜ κ2 x) ˜ + G(αβ) tw5(iv)e (κ1 x. ˜ κ2 x) ˜ + Gtw4(iv)b (κ1 x. . (κ1 x. ˜ κ2 x) ˜ + Gtw6(iv)c (κ1 x. ˜ κ2 x) ˜ + Gtw4(iv)d2 (κ1 x. ˜ κ2 x) ˜ + Gtw4(iv)e1 (κ1 x. ˜ κ2 x) ˜ + Gtw6(iv)d1 (κ1 x. ˜ κ2 x) ˜ + G(αβ) (κ1 x. ˜ κ2 x) ˜ (κ1 x. ˜ κ2 x) ˜ + G(αβ) tw6(iv)a (κ1 x. ˜ κ2 x) ˜ = Gtw4(iv)a (αβ) (αβ) (αβ) (κ1 x. ˜ κ2 x) ˜ + Gtw6(iv)b (αβ) (αβ) (αβ) . (3. ˜ κ2 x) ˜ + Gtw4(iv)d1 (αβ) (αβ) (αβ) tw4(iv)e2 + G(αβ) tw5(iv)d + G(αβ) tw4(iv)f (κ1 x.74) + 2λ 2 x=x˜  The explicit twist decomposition of this operator by means of the Young tableaux (i) and (ii) gives >(iv) ˜ κ2 x) ˜ G(αβ) (κ1 x. ˜ κ2 x) ˜ + G(αβ) tw4(iv)h (κ1 x. ˜ κ2 x) ˜ + Gtw4(iv)c (κ1 x.

˜ κ2 x) ˜ + Gtw6(iv)d4 (κ1 x. ˜ κ2 x) ˜ + G(αβ) tw6(iv)k ˜ κ2 x). Z1 dλ . Geyer. Lazar / Nuclear Physics B 581 (2000) 341–390 367 + Gtw6(iv)d2 (κ1 x. ˜ κ2 x) ˜ + G(αβ) (κ1 x. ˜ κ2 x) ˜ + G(αβ) tw6(iv)e3 (κ1 x. ˜ κ2 x) ˜ + G(αβ) (κ1 x. M. ˜ κ2 x) ˜ (αβ) (αβ) (αβ) tw6(iv)e1 + G(αβ) tw6(iv)e2 (κ1 x. ˜ κ2 x) ˜ tw6(iv)e4 tw6(iv)f tw6(iv)h (κ1 x. ˜ +G(αβ) (κ1 x. with the following operators of well-defined twist tw4(iv)a ˜ κ2 x) ˜ = gαβ  G(αβ) (κ1 x.B. ˜ κ2 x) ˜ + G(αβ) (κ1 x. ˜ κ2 x) ˜ + Gtw6(iv)d3 (κ1 x.

κ2 λx). 1−λ G(κ1 λx.

2λ 0 tw4(iv)b ˜ κ2 x) ˜ = x(α ∂β)  G(αβ) (κ1 x. Z1 dλ .x=x˜ .

κ2 λx). (1 − λ)2 G(κ1 λx.

˜ κ2 x). ˜ κ2 x) ˜ = x(α ∂β)  Gtw4(iv)d1 (αβ)  Z1 dλ  . ˜ (κ1 x.x=x˜ 2λ 0 tw6(iv)b − G(αβ) (κ1 x.

1 − λ2 ln λ + G(κ1 λx. κ2 λx).

˜ κ2 x) ˜ = − xα xβ 2 Gtw6(iv)a (αβ) 2 Z1  . ˜ κ2 x). ˜ and 1 (κ1 x.x=x˜ 2λ λ 0 tw6(iv)d1 − G(αβ) (κ1 x.

κ2 λx). 1 − λ2 + ln λ G(κ1 λx.

dλ 2λ  0 (κ1 x. ˜ κ2 x) ˜ Gtw6(iv)b (αβ) 1 = − xα xβ 2 2 Z1  .x=x˜ .

1 − λ 1 − λ2 ln λ − + G(κ1 λx. κ2 λx).

dλ λ 4λ 2λ  0 (κ1 x. ˜ κ2 x) ˜ Gtw6(iv)d1 (αβ) 1 = − xα xβ 2 4 Z1   .x=x˜ .

1 − λ2 ln λ ln2 λ + + dλ G(κ1 λx. κ2 λx).

x=x˜ . 2λ λ λ 0 tw6(iv)d3 G(αβ) (κ1 x. ˜ κ2 x) ˜ 1 = xα xβ 2 2  Z1 dλ  .

1 − λ2 ln λ ln2 λ + + G(κ1 λx. κ2 λx).

as well as (3.x=x˜ .75) . 2λ λ λ 0 being related to the scalar operator.

Lazar / Nuclear Physics B 581 (2000) 341–390 tw4(iv)c G(αβ) (κ1 x. M. Geyer. ˜ κ2 x) ˜ = −gαβ ∂ µ Z1 .368 B.

1 .

κ2 λx) x=x˜ . dλ G+ µ (κ1 λx. ˜ κ2 x) ˜ = −2x(α ∂β) ∂ µ G(αβ)   . λ 0 Z1 tw4(iv)d2 (κ1 x.

1 − λ2 ln λ .

˜ κ2 x) ˜ = x(α ∂β) ∂ µ Gtw4(iv)e1 (αβ) Z1 dλ . + dλ G+ µ (κ1 λx. κ2 λx) x=x˜ 2λ λ 0 tw6(iv)d2 − G(αβ) (κ1 x. ˜ (κ1 x. ˜ κ2 x).

κ2 λx). 1 − λ2 + Gµ (κ1 λx.

x=x˜ λ 0 tw6(iv)e1 − G(αβ) (κ1 x. ˜ κ2 x). ˜ and (κ1 x. ˜ κ2 x) ˜ Gtw5(iv)d (αβ)  µ = x(α δβ) (x∂) − x µ ∂β)  Z1   .

1 − λ2 ln λ .

˜ κ2 x) ˜ − G(αβ) (κ1 x. 1 xα xβ ∂ µ 2 Z1 dλ .76) and tw6(iv)c ˜ κ2 x) ˜ = G(αβ) (κ1 x. ˜ κ2 x). + dλ G+ µ (κ1 λx. ˜ (3. κ2 λx) x=x˜ 2λ λ 0 tw6(iv)d3 tw6(iv)d4 − G(αβ) (κ1 x.

(1 − λ)2 + Gµ (κ1 λx. κ2 λx).

x=x˜ . λ 0 tw6(iv)d2 G(αβ) (κ1 x. ˜ κ2 x) ˜ 1 = xα xβ ∂ µ 2 Z1  .

1 − λ2 ln λ ln2 λ .

κ2 λx) x=x˜ . ˜ κ2 x) ˜ = xα xβ ∂ µ G(αβ)  Z1 dλ  . + + G+ dλ µ (κ1 λx. 2λ λ λ  0 tw6(iv)d4 (κ1 x.

1 − λ2 ln λ .

+ G+ µ (κ1 λx. κ2 λx) x=x˜ . 2λ λ 0 1 (κ1 x. ˜ κ2 x) ˜ = − xα xβ ∂ µ Gtw6(iv)e1 (αβ) 2 Z1   .

1 − λ2 ln λ .

+ dλ G+ µ (κ1 λx. ˜ κ2 x) ˜ = xα xβ ∂ µ Gtw6(iv)e3 (αβ) Z1  . 2λ λ 0 (κ1 x. κ2 λx) x=x˜ .

1 − λ2 ln λ .

. 2λ λ  0 being related to the symmetric vector operator. κ2 λx) x=x˜ . + G+ dλ µ (κ1 λx.

Lazar / Nuclear Physics B 581 (2000) 341–390 369 Gtw5(iv)e (κ1 x.B. M. ˜ κ2 x) ˜ (αβ) ν (x∂) − x ν ∂β) = x(α δβ) tw6(iv)e3 − G(αβ)  Z1 ∂ µ dλ . Geyer.

1 − λ2 G(µν) (κ1 λx. κ2 λx).

x=x˜ λ 0 tw6(iv)e4 (κ1 x. ˜ κ2 x). ˜ κ2 x) ˜ = xα xβ ∂ µ ∂ ν G(αβ) . ˜ κ2 x)G ˜ (αβ) (κ1 x. ˜ Z1 tw6(iv)e4 (κ1 x.

1 − λ2 G(µν) (κ1 λx. κ2 λx).

Z1 .x=x˜ . λ dλ 0 tw6(iv)f ˜ κ2 x) ˜ = −xα xβ ∂ µ ∂ ν G(αβ) (κ1 x.

κ2 λx). dλ (1 − λ)G(µν) (κ1 λx.

furthermore. 0 being related to the symmetric tensor operator and. tw4(iv)f ˜ κ2 x) ˜ = gαβ (x∂) G(αβ) (κ1 x.x=x˜ . Z1 dλ .

κ2 λx). 1 ρ G ρ (κ1 λx.

2λ 0 tw4(iv)h ˜ κ2 x) ˜ = −x(α ∂β) G(αβ) (κ1 x. Z1 .x=x˜ .

tw6(iv)h dλ λGρ ρ (κ1 λx. κ2 λx).

˜ κ2 x) ˜ = −x(α ∂β) Gtw4(iv)e2 (αβ) Z1 dλ .x=x˜ − G(αβ) (κ1 x. ˜ κ2 x). ˜ 0 (κ1 x.

κ2 λx). 1 − λ2 ρ G ρ (κ1 λx.

˜ and 1 (κ1 x. ˜ κ2 x).x=x˜ λ 0 tw6(iv)e2 − G(αβ) (κ1 x. ˜ κ2 x) ˜ = − xα xβ  Gtw6(iv)h (αβ) 4 Z1 dλ .

κ2 λx). 1 − λ2 ρ G ρ (κ1 λx.

λ 0 (κ1 x.x=x˜ . ˜ κ2 x) ˜ = Gtw6(iv)e2 (αβ) 1 xα xβ  2 Z1  .

κ2 λx). 1 − λ2 ln λ + Gρ ρ (κ1 λx.

2 Z1 .x=x˜ . dλ 2λ λ  0 1 tw6(iv)k ˜ κ2 x) ˜ = − xα xβ  G(αβ) (κ1 x.

κ2 λx). dλ λ(ln λ)Gρ ρ (κ1 λx.

x=x˜ . . 0 being related to the trace of the symmetric tensor operator.

4. M.74) contain operators having the same twist. κ2 λx) − (x∂) dλ G (κ1 λx.370 B. it should be remarked that. the local operators of the twist-4 scalar and vector operators are totally symmetric traceless tensors and. (ii) as well as (iv). let us sum up the remaining symmetric tensor operators of twist greater than 3. κ2 λx) = −gαβ 2 2λ ρ 0 Z1 + ∂µ ). to complete twist-4. n2 ). 3. τ = 4. 2 ). contrary to the former cases (i)–(iii). which appear in the trace terms of the symmetric twist operators with Young symmetry (i). contained in the tensor space T( n2 . having finished the decomposition of a general 2nd rank tensor operator. The “scalar part” of the twist-4 light-ray tensor operator is given by ˜ κ2 x) ˜ Gtw4. these operators being multiplied by either gαβ or xα and xβ are scalar or vector operators corresponding to symmetry type (i).s (αβ) (κ1 x. 2 ) ⊕ T( 2 . However. the trace terms (3. therefore. Lazar / Nuclear Physics B 581 (2000) 341–390 Here.3. as the primary operator of symmetry class (iv). Determination of the complete symmetric light-cone tensor operators Now. ( Z1 Z1 1 ρ 1  dλ λ(ln λ)G(κ1 λx. twist-5 and twist-6 operators. Geyer. Whereas the local operators resulting from the twist-4 tensor operator of symmetry class (iv) are contained in the tensor n−2 n−2 n+2 space T( n+2 2 .

.

.

+ dλ λGµ (κ1 λx. κ2 λx) .

.

( Z1      ln λ 1 − λ2 1 1 − λ2 + λ ln λ ∂β) + xβ) + λ ln λ +  dλ = −x˜(α 2λ 4 λ λ 0 Z1 + × G(κ1 λx.77) x=x˜ and the “vector part” of the twist-4 light-ray tensor operator reads ˜ κ2 x) ˜ Gtw4.v (αβ) (κ1 x. κ2 λx) λ 2 0 Z1 − 0     1 dλ 2 2 2 1 − λ ∂β) + xβ) 1 − λ + 2 ln λ  λ 2 ). 0 0 . κ2 λx)   1 dλ ∂β) + xβ) (ln λ) Gρ ρ (κ1 λx. (3.

.

.

κ2 λx) . µ + × ∂ Gµ (κ1 λx.

. .

the contributions to the complete “vector part” of the twist-5 light-ray tensor operator are constructed by means of the Young tableau (ii). and their local operators are . (3.78) x=x˜ Furthermore.

κ2 λx) λ 4λ 2λ    Z1   1 − λ2 ν 1 − λ2 ln λ ν ν µ µ ν δβ) (x∂) − x ∂β) − xβ) ∂ ∂ − + xβ) ∂ x − dλ λ 2λ λ 0 ). ( Z1   (1 − λ)2 ν δβ) (x∂) − x ν ∂β) − xβ) ∂ ν  dλ = −x˜(α 2λ 0   1 1 − λ 1 − λ2 2 ν − + ln λ xβ)  x G+ − ν (κ1 λx.B. 2 ).v (αβ) (κ1 x. Geyer. This complete twist-5 tensor operator is given by ˜ κ2 x) ˜ Gtw5. n−2 2 ) ⊕ T( 2 . M. Lazar / Nuclear Physics B 581 (2000) 341–390 371 n−2 n antisymmetric 10 traceless tensors contained in the space T( n2 .

.

.

(3. κ2 λx) .79) × G(µν) (κ1 λx.

. .

κ2 λx) 2λ . 2 ).  x=x˜ The “scalar part” of the twist-6 light-ray tensor operator has Young symmetry (i). κ2 λx) +  dλ 2λ 0 Z1 dλ (1 − λ)2 µ + ∂ Gµ (κ1 λx. the local twist-6 operators are totally symmetric traceless tensors lying in the space n−2 T( n−2 2 . κ2 λx) dλ = −x˜α x˜β 2 λ 2λ 0 Z1 − 1 − λ2 ρ G ρ (κ1 λx. This nonlocal twist-6 operator is given by ˜ κ2 x) ˜ Gtw6.s (αβ) (κ1 x.80) 0 Z1 − ∂ µ∂ ν ). Thus. (  Z1  1−λ 1+λ 1 2  + ln λ G(κ1 λx. (3.

.

1−λ .

κ2 λx) . G(µν) (κ1 λx.

dλ .

˜ κ2 x) ˜ + G(αβ) (κ1 x. (3. (3.81) Let us remark that by construction the terms of the first line are traceless.73). (3. (αβ) (κ1 x.50). we may pick up all the contributions to the symmetric (traceless) tensor operator. ˜ + Gtw5. Eq. whereas the traces of the third line vanish because of antisymmetry and x˜ 2 = 0 on the light-cone. Eq. ˜ κ2 x) ˜ + Gtw6. 10 Note that antisymmetry is not in α and β but results from the differential operator! . (3.7)): tw2(i) tw3(ii) tw4(iv) ˜ κ2 x) ˜ = G(αβ) (κ1 x.v (αβ) (κ1 x.v ˜ κ2 x) ˜ + Gtw4. Together with the twist-2 part.s (αβ) (κ1 x. we finally obtain the following complete decomposition of the symmetric lightcone tensor operator (compare Eq. (αβ) (κ1 x.19). λ 0 x=x˜ Now. and the twist-4 part. ˜ κ2 x) ˜ + Gtw4. the twist-3 part. ˜ κ2 x) ˜ G(αβ) (κ1 x.s ˜ κ2 x). ˜ κ2 x) ˜ + G(αβ) (κ1 x. (3. Eq.

77) and (3. (3. M.78). using (x∂) ∼ = λ∂λ . Geyer. ˜ g αβ Gαβ (κ1 x. (3. the traces of the second line do not vanish but restore the trace of the tensor operator. performing partial integrations and observing the equality .s ˜ κ2 x) ˜ + g αβ Gtw4.372 B.v ˜ κ2 x). However.82) This may be proven explicitly by taking the trace of Eqs. Lazar / Nuclear Physics B 581 (2000) 341–390 respectively. ˜ κ2 x) ˜ = g αβ Gtw4. (αβ) (κ1 x. (αβ) (κ1 x.

 (3. κ2 λx).83) g αβ − 2∂ α x β + (1/2)x α x β G(αβ) (κ1 λx.

The local versions of such higher twist operators have already been considered in the early days of QCD in systematic studies of deep inelastic scattering [38. F . Here. too. Various studies of local twist-4 operators determined their anomalous dimension matrices and the behaviour of their structure functions [46– 51]. γ5 σαβ }. Fµν = {Fµν . If written nonlocal all these operators are necessarily multilocal. γ5 γα }. nonlocal higher twist operators have been extensively used only in the study of the structure function g2 [29. Extension to trilocal tensor operators The procedure reviewed in Section 2 for bilocal operators may be extended to trilocal operators. General trilocal tensor operators: quark–antiquark–gluon. our main interest is not to give an exhaustive study of the various scattering processes but to present the twist decomposition of some characteristic multilocal operators.56. i = {σαβ . However. In the following we consider various trilocal operators of minimal twist-3 and twist-4. For compact notations.λ=0 = 0. These operators occur in higher twist contributions to light-cone dominated processes and.57] and of the photon and vector meson wave functions [58–61]. Γαβ i eµν }. the twist decomposition of a generic bilocal 2nd rank light-ray tensor operator is completed. 4.and multilocal light-ray operators. Herewith. we will introduce the following abbreviations: Γ i = {1. γ5 }. In the next section we show how this general formalism may be applied also to tri.1. We call them unconstrained because their truncation with x α x β does not vanish identically. Γαi = {γα . The relevance of local twist-3 operators for the structure function g2 of polarized deep inelastic scattering is also well-known [52–55]. 4. additionally. as counterparts in the renormalization of such higher twist operators which have been obtained in Section 3.39] and on behalf of giving some parton interpretation for the distribution amplitudes [40–45].30. four-quark and four-gluon operators This subsection is devoted to the consideration of trilocal operators which share the same twist decomposition as the gluon operators Gαβ (or their (anti)symmetric parts).

respectively. we consider the following quark–antiquark–gluon operators. Lazar / Nuclear Physics B 581 (2000) 341–390 373 4. respectively. Vαβ expand its three local fields around y = 0: 11 (κ1 x. they have the same symmetry as the gluon tensor operators Gαβ and G[αβ] . τ x)σα ρ Fβρ (τ x)U (τ x.1. the first of the above quark–gluon operators. and W[αβ] (κ1 x. 11 . τ x. 11 ij j (4.2) Vαβ (κ1 x. in some detail and Taylor for example. Quark–antiquark–gluon operators First.1. κ2x)ψ(κ2 x). Let us study.B. κ2x) Vαβ = ψ(κ1 x)U (κ1 x. 0)U (0. M. In order to be able to apply the general procedure of Section 2 as well as the results of Section 3 we have to verify the local structure of these trilocal operators. whose minimal twist is τmin = 3 and 4. τ x)Γαi ρ Fβρ (τ x)U (τ x. κ2x)ψ(κ2 x) ∞ X  → n . κ2 x)ψ(κ2 x). 0)U (0. κ2x) = ψ(κ1 x)U (κ1 x. Geyer. τ x. τ x)Γ i Fαβ (τ x)U (τ x. τ x.1) ij j (4. κ2 x) = ψ(κ1 x)U (κ1 x.

 ← n  κ1n1 τ m κ2n2  ψ(y) D y x 1 σα ρ (xDz )m Fβρ (z) x D y 2 ψ(y) .

Vαβµ 1 ···µN (4. κ2x) = Vαβ ∞ X 1 µ1 11 x · · · x µN Vαβµ (κ1 . (4.7).3) This leads to the following expansion into local tensor operators 11 (κ1 x. τ. τ. κ2 ) ≡ ψ(0)σα ρ D[βρ]µ1 ···µN (κ1 . where the left and right derivatives are given by Eqs.y=z=0 = n1 !m!n2 ! n1 . τ. `! 11 For simplicity we restrict to the nonsinglet case and suppress flavour matrices.n2 =0 = n    ∞ N X X X ← n−` 1 N n ψ(0) σα ρ κ1 D x N! n ` N=0 n=0 `=0 !   → ` N−n × (τ xD) Fβρ (0) κ2 x D ψ(0). (2. κ2 ) = N! n N X X ← κ1n−` ← τ N−n D µ1 · · · D µn−` (n − `)! (N − n)! n=0 `=0  κ` →  → × Dµn+1 · · · DµN Fβρ (0) 2 D µn−`+1 · · · D µn . Fαβ ].5) and the field strength-dependent generalized covariant derivative of N th order is given by D[βρ]µ1 ···µN (κ1 . 1 ···µN N! (4. κ2 )ψ(0). κ2 ). τ x. τ. (4.6) . and Dµ Fαβ = ∂µ Fαβ + ig[Aµ .m.4) N=0 where 11 (κ1 .

The only difference consists in a shift of any twist by one unit. despite of their arbitrary symmetry with respect to α and β they vanish identically if multiplied by x β and x α or x β . which are given as a sum of 12 (N + 1)(N + 2) terms. κ2x) = x ρ x σ ψ(κ1 x)U (κ1 x. κ2x)x ρ x σ I =2Qαρσβ (κ1 x.4) — their decomposition into terms of definite twist leads to the same expressions as we obtained in Section 3 but with Gαβ and G[αβ] ij ij exchanged by Vαβ and Wαβ . I counts the number of external x’s of the operators): I =0 ij Qαβ (κ1 x. In contrast to the operators V and W the quark–antiquark–gluon operators ij j Oαβ (κ1 x. τ → τ + 1. κ2 ) · · · D µN (κ1 . Because there are no restrictions concerning the free indices α and β the Young patterns (i)–(iv) are involved. τ x.8) are special. The explicit form of the local operators will be given below. because of the linearity of that procedure we are not required to do this for any term explicitly. ij ij Due to the same general local structure of Vαβ and Wαβ — which is governed by Taylor expansions completely analogous to Eq. e. respectively (here. have to be decomposed according to their geometric twist. they show some peculiarities which will be treated in Section 4. Furthermore. Lazar / Nuclear Physics B 581 (2000) 341–390 Obviously.. Geyer. in the various inputs of Tables 1 and 2 in the conclusions (Section 5). however. this has to be done term-byterm.63]. Their local versions have already been considered by various authors. [62. as in the foregoing cases.7) and i Fβσ (τ x) O0 αβ (κ1 x. τ x)ψ(τ x) ψ(τ x)U (τ x. if the field strength were not present this operation would reduce to the product ↔ ↔ D µ1 (κ1 . see. Anyway. We denote them by IQαβ with I = 0 and I = 2. M.5) of rank N + 2 with canonical dimension d = N + 5. (4. namely. In principle. The twist decomposition may be performed along the lines of Section 3 and. τ x.374 B. Four-quark and four-gluon operators Let us now consider trilocal operators which are built up from four quark or four gluon ij fields. κ2 x)   = x ρ x σ Fαµ (κ1 x)U (κ1 x.1. (4. the local tensor operators (4. κ2x)Fβν (κ2 x) .2.2. τ x)Γαρ j ij × U (τ x. cannot be reduced to the N th power of some extended derivative because F is equipped with some matrix structure. τ x)Γαi Fβρ (τ x)U (τ x. But. 4.10) Both sets of operators have minimal twist τmin = 4. κ2 ) of generalized derivatives (2.9) as well as ij I =2 ij Qαβ (κ1 x. for τ = 0 we are left with a more simple expression which. respectively. τ x. κ2x)   j = ψ(κ1 x)Γαi U (κ1 x. κ2 x)ψ(κ2 x) (4. κ2 x) = x ρ ψ(κ1 x)U (κ1 x. Therefore. respectively.g. τ x)Fρiµ (τ x) Fσj ν (τ x)U (τ x. the outcome . κ2 x)ψ(κ2 x) (4.7) introduced in Section 2. τ x. κ2x)Γβ ψ(κ2 x) (4. τ x.

κ2x)γβ ψ(κ2 x) = ∞ X n1 . up to now. (4. is the structure of the local operators.n2 =0 = ← n  κ1n1 τ m κ2n2  ψ(y) D y x 1 γα n1 !m!n2 !    → n .11) Open. I =0Q11 αβ . whose Taylor expansion may be written in two equivalent ways: I =0 11 Qαβ (κ1 x. relative to the gluon operators shown in Tables 1 and 2 of the Conclusions. τ x. because of the change in the external x-factors (not being accompanied by λ) the measures of the λ-integrations have to be changed according to (see also Section 4. Again. M. the difference is that now the value of twist raises by two units. 0)U (0. 0)U (0.2) dλ −→ dλ λI . τ x)ψ(τ x) ψ(τ x)U (τ x. Lazar / Nuclear Physics B 581 (2000) 341–390 375 will be almost the same as for Gαβ .m. κ2x)   = ψ(κ1 x)γα U (κ1 x. In addition. τ → τ + 2. Geyer.B. Let us study in detail the first of the four-quark operators.

× (xDz )m ψ(z)ψ (z) γβ x D y 2 ψ(y) .

y=z=0 ∞ X n .

↔ 1 ψ(y) x D y (κ1 . τ ) 1 γα ψ(y) .

y=0 n1 ! n1 =0 × ∞ X n .

κ2 ) 2 γβ ψ(z) . ↔ 1 ψ(z) x D z (τ.

τ x. τ. (2. κ2 ) = D µ1 · · · D µn ψ(0) × ψ(0) D µn+1 · · · D µN n n=0 = N X n=0 n X ← ← → κ n−` τ ` → N! D µ1 · · · D µn−` 1 D µn−`+1 · · · D µn ψ(0) (n − `)!`! `=0 . another form of the derivative obtains:     Dµ ψ(0)ψ (0) = ∂µ ψ(0)ψ(0) + ig Aµ . κ2 ) ≡ ψ(0)γα Dµ1 ···µN (κ1 . κ2x) = ∞ X 1 µ1 ij x · · · x µN I =0Qαβµ1 ···µN (κ1 . (4. The left. n2 ! n2 =0 where the two possibilities also use different generalized covariant derivatives.13) and the quark field-dependent generalized covariant derivative of N th order is given by N   X ↔ ↔ ↔ N ↔ Dµ1 ···µN (κ1 . κ2 ). τ. τ. ψ(0)ψ(0)  → ←  = D µ ψ(0) ψ(0) + ψ(0) ψ(0) D µ . right and left–right derivatives are given by Eqs. treating ψ(0)ψ (0) as a matrix in the group algebra. This leads to the following expansion into local tensor operators I =0 ij Qαβ (κ1 x. and.7). N! (4. τ.z=0 . κ2 )γβ ψ(0).12) N=0 where I =0 ij Qαβµ1 ···µN (κ1 .

Constrained trilocal operators: Shuryak–Vainshtein and three-gluon operators ij Now we consider the special quark–antiquark–gluon operator Oαβ . cf. . Lazar / Nuclear Physics B 581 (2000) 341–390 × N−n X ← ← ψ(0) D µn+1 · · · D µn+k k → τ k κ2N−n−k → D µn+k+1 · · · D µN .τ x. respectively. Any of these operators decompose in the same manner into irreducible representations of the Lorentz group. leading to a general expression of the following form: I Rαβ (κi x) = ∞ X 1 µ1 x · · · x µN x ρ · · · x σ IRαβρ···σ µ1 ···µN (κi ). M. The only difference will be that instead of the quark fields corresponding gluon field strengths appear in (4. . (4. The latter result is immediately related to the fact that all indices µ1 · · · µN are to be symmetrized.τ x. . τ x)Γαi Fβρ (τ x)U (τ x. with any field being located at a different point κi x. N! (4. i. 4.64] ( ψU ψ)( ψU ψ).7). . and that the derivatives are to be taken in the adjoint representation.. These operators have the general structure.14) Analogous generalized covariant derivatives occur for the four-gluon operators.e. Despite having a complicated structure this does not matter in the twist decomposition of the ij trilinear operators IQαβ (κ1 x. eventually modified by the power λI which is related to the number of external x’s in IR. . i = 1. (4. lying in the first row of any relevant Young tableau! Different symmetry classifications solely depend on the distribution of the remaining indices αβρ · · · σ — the others being somehow truncated — to the various Young tableaux. These operators are related to the following generalizations of the so-called Shuryak–Vainshtein operators [29. κ2x)ψ(κ2 x). k![N − n − k)! (4.376 B. the twist of the various components may be shifted by some (equal) amount. The expansion into local tensors is obtained in the same way as above. κ2x). κ x) with I = 0 and I = 1. τ x. Let us finish this subsection with a short remark concerning so-called four-particle operators which also have been considered in the literature for local operators. and related three-gluon operators both having minimal twist-3.κ2 x) = x ρ I =0Oαβρ (κ1 x. As long as only up to two free indices αβ are relevant the twist decomposition takes place according to the results of Section 3.16) where again some possible flavour structure has been suppressed. [40–45.2. ( ψU F U F U ψ) or (F U F U F U F ).κ2 x) j = x ρ ψ(κ1 x)U (κ1 x. Geyer.55]. they will be denoted by IO ij (κ x. The quark–antiquark–gluon 2 αβ 1 operators are given by ij I =0 ij Oαβ (κ1 x. In addition.14). 4. 30. Eq. τ x.15) N=0 where the terms with equal value of N are given as well-defined sums of local tensor operators.

κ2 x) a i bc (κ1 x)U ab (κ1 x. The generalization arbitrary values i and j is obvious. τ x. ij Because of (4.6). κ2 )ψ(0).20) where the field strength-dependent derivative has already been given by (4. which. because of the antisymmetry of the gluon field strength. τ x. I = 1. κ2 x) = ∞ X 1 µ1 11 x · · · x µN x ρ I =0Oαβρµ (κ1 . κ2 )Fσ (0). Again. κ2x) = x α I =1 11 22 Oαβ (κ1 x. τ. κ2 ) ≡ ψ(0)γα Dab [βρ]µ1 ···µN (κ1 . τ. They only differ in the rank of the local tensor operators which is due to the external powers of x. τ x. Lazar / Nuclear Physics B 581 (2000) 341–390 Sβ± (κ1 x.22) with the same field strength dependent generalized covariant derivative (4. From now on we omit the indices i and j . κ2 x)Fσdν (κ2 x).6). As it became obvious by the above considerations all the nonlocal three-particle ij operators IOαβ have the same general local structure and. κ2 x) vanish identically. that because of notational simplicity in the explicit expressions below we shifted the variable I → I − 1. ˜ τ x.19) N=0 with I =0 11 Oαβρµ1 ···µN (κ1 . 11 (κ x. τ x. τ x.B.21) N=0 where I =1 11 a bν Oαβρσ µ1 ···µN (κ1 . this must be taken into account by a change of the integration measure according to 12 dλ −→ dλ λI +1 . have to be determined by means of the Young tableaux (ii)–(iv). decompose according to the same symmetry patterns. The Taylor expansion of the related three-gluon operator obtains as follows: I =1 11 Oαβ (κ1 x. τ x. κ2 ). = x ρ x σ Fαν (4. M. (4. κ x) reads The Taylor expansion of the operator I =0Oαβ 1 2 I =0 11 Oαβ (κ1 x. therefore. κ2 x) = ∞ X 1 µ1 x · · · x µN x ρ x σ N! I =1 A Oαβρσ µ1 ···µN (κ1 .17) ab c (x) ≡ f acb Fαβ (x) and the phase factors are taken in the adjoint representation. (4. 1 ···µN N! (4. κ2 x) = x ρ x σ I =1Oαβρσ (κ1 x.18) the scalar operators x α x β IOαβ (κ1 x. but now in the adjoint representation. κ2 x) 377  α x . . τ. τ. in the flavour singlet case. τ. κ2 ) (4. κ2 x) ± i I =1Oαβ (κ1 x. Geyer. τ. τ x. where Fαβ Because of their construction these special trilocal operators have the property ij x β IOαβ (κ1 x.18) They contain twist-3 up to twist-6 contributions which. τ x. mix with the following three-gluon operators: I =1 i i Oαβ (κ1 x. ˜ κ2 x) ˜ for I = 0 and In order to determine the twist-3 light-cone operators IOαβ (κ1 x. τ x)Fβρ (τ x)U cd (τ x. τ x. τ x. κ2x) ≡ 0. κ2 ) ≡ Fαν (0)Dab [βρ]µ1 ···µN (κ1 . let us consider the following Young tableaux of symmetry pattern (iiB) 12 Observe. (4.

Let us remark that. (5. As in the case of the gluon tensor only the first terms. ˜ κ2 λx) ˜ 0 >(ii) ˜ τ x. let us to project onto the light-cone. ˜ κ2 x) ˜ = 1 2 Z1 µ dλ λ1+I (1 − λ) 2x[α ∂β] ∂ µ − x[α δβ]  0  .κ2 λx) are to be taken into account. because of here. 2.18). after partial integrations leads to the above expression (4. I (4. ˜ τ x.30]). κ2 x).τ λx. only the above mentioned Young tableaux contribute and. ˜ κ2 x). twist-5 as well as twist-6 operators resulting from the trace terms. τ x. first. τ x. Lazar / Nuclear Physics B 581 (2000) 341–390 ··· ρ µ1 µN α ρ σ µ1 β and β ··· µN α respectively: I tw3 Oαβ (κ1 x. Eq. their operator lacks to be really of twist-3 since it is not traceless. therefore it contains also twist-4.κ2 λx) 0 tw3 tw3 O[αβ] (κ1 x. κ2 x) + IO(αβ) (κ1 x.23) x ρ Oρµ .τ λx. ˜ − IO[αβ] (κ1 x.23). Geyer. For the antisymmetric part of the twist-3 operator one gets Z1 I tw3 O[αβ] (κ1 x.14) of [29. In the case I = 0 this operator is in agreement with the expression given by Balitsky and Braun (cf. we write Oµ ≡ property (4.23) and (3. τ x.24) where I >(ii) O[αβ] (κ1 x. −x µ ∂α ∂β IO˚ µ (κ1 x. namely the truncation by x ρ in the second term of the integrand.42) result also from that property. 1. k = 0. However. ˜ τ λx. κ2 x) = µ 2δ(α ∂β) IO˚ µ (κ1 x. M. κ2 x) 1 = 2 = Z1   µ   dλ λI 1 + λ2 δβ ∂α + 1 − λ2 δαµ ∂β IO˚ µ (κ1 λx. τ x. in the expansion of IO˚ µ (κ1 λx. κ2 x).378 B. (4. ˜ τ x. Now. τ x. second. the difference between the expressions (4. and in the following. ˜ κ2 x) ˜ = dλ λ2+I ∂[α IOβ] (κ1 λx.

τ λx. × IOµ (κ1 λx. κ2 λx).

(4. ˜ I 1 tw4(ii) O[αβ] (κ1 x. I >(ii) tw4(ii) tw5(ii) O[αβ] (κ1 x.25) >(ii) ˜ τ x.x=x˜ . ˜ κ2 x) ˜ = IO[αβ] (κ1 x.26) with Z1 dλ λI 1 − λ2  0 . ˜ τ x. ˜ τ x. ˜ κ2 x) ˜ arises as The decomposition of IO[αβ] (κ1 x. ˜ τ x. ˜ τ x. ˜ κ2 x) ˜ + IO[αβ] (κ1 x. ˜ κ2 x) ˜ = x[α ∂β] ∂ µ 2 (4. ˜ κ2 x).

× Oµ (κ1 λx. τ λx. κ2 λx).

x=x˜ . I (4.27) .

M. Geyer. ˜ τ x.B. Lazar / Nuclear Physics B 581 (2000) 341–390 I  1 µ tw5(ii) O[αβ] (κ1 x. ˜ κ2 x) ˜ = − x[α δβ] (x∂) − x µ ∂β]  4 Z1 .

× dλ λI (1 − λ)2 IOµ (κ1 λx, τ λx, κ2 λx)

x=x˜ .

379

(4.28)

0

The symmetric part reads
I

tw3
O(αβ)
(κ1 x,
˜ τ x,
˜ κ2 x)
˜

Z1

>(ii)

dλ λI ∂(α IOβ) (κ1 λx,
˜ τ λx,
˜ κ2 λx)
˜ − IO(αβ) (κ1 x,
˜ τ x,
˜ κ2 x),
˜

=

(4.29)

0

where
I

>(ii)
O(αβ)
(κ1 x,
˜ τ x,
˜ κ2 x)
˜ 

Z1
=

dλ λI
0 

1
1
µ
1 − λ2 gαβ ∂ µ + (1 − λ)δ(α xβ)  + λ(1 − λ)x(α ∂β) ∂ µ
2
2 

1
− (1 − λ)2 xα xβ ∂ µ  IOµ (κ1 λx, τ λx, κ2 λx)

x=x˜ .
4

(4.30)

Let us remark that the conditions of tracelessness for the light-cone operators are:
tw3
(κ1 x,
˜ τ x,
˜ κ2 x)
˜ = 0,
g αβ IO(αβ)

tw3
dα IO(αβ)
(κ1 x,
˜ τ x,
˜ κ2 x)
˜ = 0,
tw3
(κ1 x,
˜ τ x,
˜ κ2 x)
˜ = 0.
dα IO[αβ]

Now we can calculate the twist-3 vector operator from the twist-3 tensor operator. The
twist-3 light-cone vector operator reads
I

Oαtw3 (κ1 x,
˜ τ x,
˜ κ2 x)
˜
Z1
=

 

dλ λ2+I δαµ (x∂) − x µ ∂α − xα ∂ µ IOµ (κ1 λx, τ λx, κ2 λx)

x=x˜

0

Z1
= Oα (κ1 x,
˜ τ x,
˜ κ2 x)
˜ − x˜α
I

dλ λ2+I ∂ µ IOµ (κ1 λx, τ λx, κ2 λx)

x=x˜ ,

(4.31)

0

and the twist-4 vector operator which is contained in the trace terms of the twist-3 vector
operator
Z1
I

Oαtw4(ii)(κ1 x,
˜ τ x,
˜ κ2 x)
˜ = x˜α

dλ λ2+I ∂ µ IOµ (κ1 λx, τ λx, κ2 λx)

x=x˜ .

(4.32)

0

The determination of the antisymmetric twist-4 operator having symmetry type (iii)
obtains from Young tableaux

380

B. Geyer, M. Lazar / Nuclear Physics B 581 (2000) 341–390

...

ρ µ1

µN

ρ σ µ1
β
α

and

β
α

...

µN

respectively. Again, making use of property (4.18), the result is
I

tw4(iii)
O[αβ]
(κ1 x,
˜ τ x,
˜ κ2 x)
˜

Z1
=


µ
ν
dλ λ2+I δ[α (x∂)δβ]
− 2x ν ∂β] IO[µν] (κ1 λx,τ λx,κ2 λx)

x=x˜

0
tw5(iii)

−IO[αβ]

(κ1 x,
˜ τ x,
˜ κ2 x),
˜

(4.33)

twist-5 part is determined by the trace namely with
I

tw5(iii)
O[αβ] (κ1 x,
˜ τ x,
˜ κ x)
˜ =−

Z1  


dλ λI 1 − λ2 x[α δβ] (x∂) − x [µ ∂β] ∂ ν]

0


− x[α δβ] x ν]  IO[µν] (κ1 λx, τ λx, κ2 λx)

x=x˜ .

(4.34)

The antisymmetric twist-5 tensor operator is given by:
I

tw5
O[αβ]
(κ1 x,
˜ τ x,
˜ κ x)
˜
tw5(ii)

tw5(iii)

˜ τ x,
˜ κ x)
˜ + IO[αβ]
= IO[αβ] (κ1 x, 

1
µ
= − x[α δβ] (x∂) − x µ ∂β] 
4

Z1

(κ1 x,
˜ τ x,
˜ κ x)
˜

κ2 λx). 2 dλ λI 1 − λ IOµ (κ1 λx. τ λx.

x=x˜ 0 Z1 − 0   [µ [µ dλ λI 1 − λ2 x[α δβ] (x∂) − x [µ ∂β] ∂ ν] − x[α δβ] x ν]  .

τ λx. κ2 λx). × IO[µν] (κ1 λx.

x=x˜ .35) Finally. ˜ κ2 x) ˜ + IO[αβ] (κ1 x. ˜ + IO[αβ] (4. Making use of property (4. ˜ κ2 x) ˜ tw4(ii) tw5 (κ1 x. we obtain the complete decomposition of the antisymmetric tensor operator: I tw3(ii) tw4(iii) O[αβ] (κ1 x. ˜ τ x.36) The determination of the symmetric twist-4 operator having symmetry type (iv) obtains from Young tableaux ρ µ1 µ2 β α ··· µN and ρ σ µ1 β α respectively.18) the result is ··· µN . ˜ τ x. ˜ κ2 x) ˜ + IO[αβ] (κ1 x. ˜ τ x. (4. ˜ κ2 x). ˜ τ x. ˜ κ2 x) ˜ = IO[αβ] (κ1 x. ˜ τ x.

B. ˜ κ2 x) ˜ I Z1 = . ˜ τ x. Lazar / Nuclear Physics B 581 (2000) 341–390 381 tw4(iv) O(αβ) (κ1 x. M. Geyer.

κ2 λx). τ λx.  ν dλ λI (1 − λ) δαµ δβν x ρ (x∂)∂ρ − 2x µ (x∂)δ(β ∂α) IO(µν) (κ1 λx.

κ2 λx)   1 I 1 I µ µ I 2 2 µ λ gαβ ∂ + λ (1 − λ )x(α δβ)  − λ (1 − λ) xα xβ ∂ − 2 2 × IOµ (κ1 λx.x=x˜ 0 >(iv) − IO(αβ) (κ1 x. ˜ τ x. τ λx. ˜ (4.37) where I >(iv) O(αβ) (κ1 x. κ2 λx)   1 2+I 1 I 2+I λ gαβ (x∂) − λ x(α ∂β) − λ (ln λ)xα xβ  + 2 2 . τ λx. ˜ τ x. ˜ κ2 x). ˜ κ2 x) ˜  Z1 = dλ 0  ν µ 2λ2+I x(α δβ) ∂ − λ1+I (1 − λ)xα xβ ∂ µ ∂ ν IO(µν) (κ1 λx.

.

× IOρ ρ (κ1 λx. κ2 λx) . τ λx.

.

˜ κ2 x) ˜ = I ( Z1 1 gαβ (x∂) dλ λI IOρρ (κ1 λx. to a complete twist-4. which appear in the trace terms of the symmetric twist operators with Young symmetry (ii) and (iv). we use the property (4.18) and sum up the symmetric higher twist operators. The complete twist-4 operators are constructed by means of the Young tableau (i). ˜ τ x.s O(αβ) (κ1 x.38) x=x˜ Now. . (4. The “scalar part” of the twist-4 tensor operator is given by tw4. τ λx. κ2 λx) 2 0 Z1 −∂ µ ). twist-5 and twist-6 operator.

.

.

κ2 λx) . Oµ (κ1 λx. τ λx.

.

v O(αβ) (κ1 x. τ λx.39) x=x˜ and the twist-4 “vector part” reads ( Z1 I tw4. κ2 λx) 2 I 0    1 1 − λ2 ∂β) + xβ) 1 − λ2 + 2 ln λ  ∂ µ 4 0 ). ˜ κ2 x) ˜ = −x˜(α   1 dλ λ ∂β) + xβ) (ln λ) IOρ ρ (κ1 λx. ˜ τ x. 2+I I dλ λ 0 (4.

.

.

κ2 λx) . I × Oµ (κ1 λx. τ λx.

. (4.40) .

 Z1 − dλ λI x=x˜ .

˜ κ2 x) ˜ ( Z1 2 ν  1 dλ λI 1 − λ δβ) (x∂) − x ν ∂β) − xβ) ∂ ν  IOν (κ1 λx. the complete twist-5 vector operator is again constructed by means of the Young tableau (ii). The “vector part” of the twist-5 tensor operator is given by tw5.382 B. ˜ τ x. τ λx. κ2 λx) = −x˜(α 4 I 0  Z1 − dλ λI 0  ν  1 − λ2 δβ) (x∂) − x ν ∂β) − xβ) ∂ ν ∂ µ     1 2 µ ν 1 − λ + ln λ xβ) ∂ x − 2 ). Lazar / Nuclear Physics B 581 (2000) 341–390 Furthermore.v O(αβ) (κ1 x. M. Geyer.

.

.

× IO(µν) (κ1 λx. τ λx. κ2 λx) .

. .

τ λx. κ2 λx) = x˜α x˜β 2 I 0 1 − ∂ µ 4 Z1 0 Z1 +∂ ∂ µ ν 2 dλ λI 1 − λ IOµ (κ1 λx. The “scalar part” of the twist-6 tensor operator is given by tw6. ˜ κ2 x) ˜ ( Z1  1  dλ λI 1 − λ2 ln(λ) IOρ ρ (κ1 λx. ˜ τ x. κ2 λx) ). τ λx. (4.s O(αβ) (κ1 x.41) x=x˜ The full twist-6 scalar operator has Young symmetry (i).

.

.

τ λx. dλ λ (1 − λ) O(µν) (κ1 λx. κ2 λx) .

.

29). Eq. ˜ κ2 x) ˜ + IO(αβ) tw5. (4. ˜ τ x. ˜ κ2 x) ˜ + IO(αβ) (κ1 x. ˜ + IO(αβ) (4. Making use of the tensorial harmonic polynomials up to rank 2 we determined for the first time and in a systematic way the various contributions of definite twist for a generic bilocal 2nd rank tensor operator Gαβ (κ1 x.42) x=x˜ Thus. we finally obtain the complete decomposition of the symmetric tensor operator: I tw3(ii) tw4(iv) O(αβ) (κ1 x. ˜ τ x. Conclusion Let us shortly summarize our results. ˜ κ2 x) ˜ tw4. . κ2 x). κ2 x) and G[αβ] (κ1 x.s (κ1 x.37). I I 0 . ˜ τ x. ˜ κ2 x) ˜ = IO(αβ) (κ1 x. ˜ κ2 x).v tw6. κ2 x) as well as its symmetric and antisymmetric parts G(αβ) (κ1 x.43) 5.v (κ1 x. ˜ κ2 x) ˜ + IO(αβ) (κ1 x. ˜ τ x. (4. ˜ τ x. ˜ τ x. Eq. and the twist-4 part. (4. together with the twist-3 part.s tw4. ˜ κ2 x) ˜ + IO(αβ) (κ1 x. ˜ τ x.

65) (iii) – τ =4 (3. depending on the number I of external x-factors. the corresponding equations of Section 3.B.50) (ii) τ =3 – Eq. ˜ κ2 x) ˜ by their symmetry type (i) up to (iv) and the number of G(αβ) (κ1 x.80) (i) τ =6 – Table 2 Twist decomposition of general vector operators Gα = G(α•) + G[α•] YT G(α•) (κ1 x. ˜ κ2 x) ˜ (3. There.52) (ii) τ =3 – (3. These expressions also apply to generic trilocal light-ray operators.53) (ii) – τ =3 (3. Geyer.78) (i) τ =4 – (3. by a change in the integration measure. ˜ κ2 x) ˜ G[α•] (κ1 x. By projection onto the light-cone we obtained the decomposition of a generic bilocal light-ray tensor operator of 2nd rank — and its reductions to (anti)symmetric tensors as well as the related vector and scalar operators — into operators of definite twist. ˜ κ2 x) ˜ + G[αβ] (κ1 x. ˜ κ2 x) ˜ (3. in the second half of the tables the contribu