Está en la página 1de 13

Subspace quantum dynamics and the quantum action principle

Richard F. W. Bader, Shalom Srebrenik, and T. Tung NguyenDang



Citation: J. Chem. Phys. 68, 3680 (1978); doi: 10.1063/1.436226
View online: http://dx.doi.org/10.1063/1.436226
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v68/i8
Published by the American Institute of Physics.

Additional information on J. Chem. Phys.
Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Subspace quantum dynamics and the quantum action
principle
Richard F. W. Bader, Shalom Srebrenik,a) and T. Tung Nguyen-Dang
Department of Chemistry. McMaster University. Hamilton. Ontario. Canada L8S 4Ml
(Received 31 October 1977)
Schwinger's quantum action principle is used to obtain a quantum mechanical description of a subspace
and its properties. The subspaces considered are those regions (11) of real space as defined by a property of
the system's charge distribution p(r), namely, that they be bounded by a surface S(r) through which the
flux in vp(r) is zero at every point on S(r). Through the variation of the action integral, expressed in
terms of the appropriately defined Lagrangian integral operator for a subspace, one obtains the quantum
equations of motion and an expression for the change in the subspace action integral operator, a'\\? (11).
This expression obeys a principle of stationary action, and thus, as for a system with boundaries at
infinity, il defines the generators of infinitesimal unitary transformations. The change in a subspace
property () (11) as induced by such an infinitesimal unitary transformation is investigated and related to the
corresponding change as described by the calcul us of variations. The latter consists of two contributions,
one from the variation over the domain of the subspace, the other from a variation of its surface. It is
found that the change in [) (11) caused by the domain variation is expressible in terms of the commutator
of!9 (11) and &(t), the generator of the infinitesimal unitary transformation. Through the definition of a
subspace projector in the coordinate representation this commutator is shown to contain the subspace
projection of the all space transformation and a term which corrects for the nonhermiticity of the
projected generator. The contribution to the change in 0(11) from the surface variation is not expressible
in terms of a commutator. However, this change is still quantitatively determined by the action of the
generator as a result of the subspace boundary being defined in terms of the observable charge
distribution. The Heisenberg equation of motion for a subspace property is a particular result obtained
from the general analysis of the change induced in a subspace
A
when the total system is subjected to a
canonical transformation. The effect of the temporal generator-Jell t on a subspace property is expressible
as a projection of the usual all space result and a term describing the flux in the vector current of the
property density across the boundary of the subspace (the domain variation). plus a contribution arising
from the change in the boundary with time (the surface variation). Schwinger's quantum action principle is
re-expressed as a sum of the changes in the action integral operator for each subspace in the system,
changes which assume their simplest physical form for the particular class of subspaces studied here. Since
the application of the zero-flux boundary condition to a molecular system partitions it into a collection of
chemically identifiable atomiclike fragments, the total change in the transformation function as given by
the action principle, may be expressed in terms of a sum over the change in action for each atom in a
molec;ule.
I. INTRODUCTION
The purpose of this paper is to obtain a general quan-
tum mechanical description of the properties of a sub-
space. The subspaces of particular interest are those
regions of real space bounded by a surface S(r) through
which the flux of the gradient in the charge density p(r)
is zero,
basis employing the correspondence principle and the
associated assumptions based on classical Hamiltonian
dynamics are replaced by a single dynamical prinCiple,
the principle of stationary action. This approach yields
the equations of motion and the commutation relation-
ships.
Vp(r)n(r)=O VrES(r). (1)
As shown earlier, 1 and in the preceding paper, 2 the
quantum mechanical relationships governing the varia-
tional properties of such subspaces exhibit maximum
correspondence with the expreSSions of all space quan-
tum mechanics, that is, with the expreSSions which de-
scribe the variational properties of a system with bound-
aries at infinity. The description of the subspace and
its properties are obtained through the application of
Schwinger's action principle. 3
A. Schwinger's action principle
This principle provides a formal axiomatic develop-
ment of quantum mechanics in which the conventional
apresent address: Department of Physical Chemistry, Hebrew
University, Jerusalem, Israel.
For a statement of the axiom upon which SchWinger's
approach is based we quote Roman
4
: "For every quan-
tized system there exists an action-integral operator
W constructed from the operators qj(t) and ql(t) in ex-
actly the same manner as the corresponding classical
action integral
'\\' = f2 (ql' qj, t) dt ,
t1
(2)
such that in performing an arbitrary general operator
variation, the ensuing change in the action operator
(3)
is the difference between the values at t2 and t1 of the
generator of a corresponding unitary transformation,
causing the change in the quantized system." This is a
statement of the operator principle of stationary action.
Formally, the change in the action integral as given by
Eq. (3) is the same in the classical and quantized cases
and thus Schwinger's action principle establishes the
3680
J. Chern. Phys. 68(8). 15 Apr. 1978 00219606178/68083680'$01.00 1978 American Institute of Physics
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
3681
correspondence between the generator of a claSSical
canonical transformation and the generator of a unitary
transformation of the equivalent quantized system.
The generator of an infinitesimal unitary transfor-
mation is defined in terms of the operator it,
(4)
The operator iL causes a unitary transformation up to
first-order in the infinitesimal t: if
<fit = i-it: & ,
that is, if if is Hermitian. For such a transformation,
the infinitesimal changes in the state vectors and opera-
tors are given by
(5)
and
6<1 = it:[if, &] .
(6)
B. Subspace properties and field theory
All subspace properties are determined by the appro-
priate averaging of a one-electron operator. This is
true even of properties which require a knowledge of the
two-electron probability, as for example, the energy of
a subspace.
1
,a Because of the Single-particle nature of
subspace properties, they are naturally stated and de-
fined in terms of the corresponding integral operators
of field theory. To quote Schwinger
5
; "The essence of
field theory is to provide a conceptually Simpler and
more fundamental description by depending on the par-
ticle as the basic entity." Thus, field theory, because
of its basic single-particle approach, appears to be the
natural language to use in the formal development of the
quantum mechanical description of a subspace. In addi-
tion, its use leads naturally to the Heisenberg picture,
a picture which is most appropriate in exhibiting the
analogies with classical mechanics which arise through
the use of the quantum action principle.
The subspace average of a single-particle property in
the Schrooinger picture,
M(U) = 1 drlf dr'ml(rt. r1hP*/fi
= L dr[p(r, r'),J!i(r, r')]I""
is equivalently defined as the field operator :JR:(U) in
terms of the field density /fit(r)/fi(r),
:m(n) = f dr[/fit(r');lft(r, r')/fi(r)Jr .. ' .
n
(7)
If M is a two-particle property described by a two-par-
ticle operator diagonal in the coordinate representation,
then the field operator is obtained by first averaging
ml(r
h
ra) over the field density l/l (rz)/fi(r
z
) over all space
to obtain the corresponding one-particle operator :m(r
l
)
where
:m(r
l
) = f dra/fi
t
(rzR (r
l
, rz)l/J(r
z
) .
The averaging of the latter operator over the subspace
(8)
again yields an operator which is the equivalent of the
corresponding subspace average of M
M(U)=[N(N-l)/2] In dr
l
rz)f" dr"/fi*/fi
r f (Z)
= J
o
dr
l
rz)p (rl , r
2
) .
To quote Schwinger: "Space itself is the Simplest
structure that one can conceive as a basis for carrying
various phySical properties. Thus we take seriously
the idea that one can meaningfully speak of the amount
of energy, momentum and other physical properties in-
side arbitrary small elementary volumes of three-di-
mensional space." These remarks as applied to a field
and its corresponding field denSity, parallel the concept
of using the charge denSity as the vehicle for the under-
standing of chemistry.
II. QUANTUM ACTION PRINCIPLE IN THE
DEFINITION OF A SUBSPACE
A. The all space action principle
The statement of the principle of stationary action is
obtained through a general variation of the action inte-
gral wexpressed in terms of the field variables /fi(r) and
/fit(r). Thus
i
lZ
A
W= V/fi, /fi, t)df ,
tl
where is the integral of the Lagrangian denSity L,
, = f L(/fi(r, t) ,V/fi(r, t), t), t)dr . (9)
The general variation of (Vas denoted by 1l:(I,', implies
a functional change in the field variable and a change in
the pOSitional and time coordinates:
t../fi =/fi'(r, t)- /fi(r, t) (10)
The first-order change in the action integral resulting
from this variation is
(11)
Applying the calculus of the variations to Eq. (11) with
the boundary condition that 6/fi = 6/fit = 0 when I r 1_ co,
yields to first-order
ilzf (aL aL a aL)
t..'\\'= --V.--- --.- 6/fidrdt+c c
tl a/fi aV/fi at e/fi . .
{f(
aL aL) {I ta
+ dr+.c(t)6t) II
(12)
Hamilton's principle corresponds to a variation of the
action integral with fixed time end points and with the
boundary condition that 6/fi vanish at the time end points.
In this case, 6W yields, through Eqs. (11) and (12), the
equations of motion from the stationary condition that
6W= O. This was exemplified in the previous paper
where Schrooinger's time dependent equations were ob-
tained using Hamilton's, principle. If Hamilton's prin-
ciple is modified in such a way as to relax the boundary
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
3682
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
condition that ol/J vanish at the time end pOints, then
oW yields in addition to the equations of motion the sec-
ond integral on the rhs of Eq. (12). This was the pro-
cedure used in the previous paper to obtain the subspace
variational principle. The final term in Eq. (12) ob-
tained from the general variation of W is the contribution
to the change in the action resulting from the change in
time at the time end points.
If the Lagrangian integral operator is taken to be
. = f - - (lf
2
/2m)Vl/Jt . VI/J
- V(r)l/Jtl/J} - f dr f dr'l/Jt(r)l/Jt(r')U(r, r')I/J(r')I/J(r) ,
(13)
then the variation of '\\, yields the following expression
(and its adjoint) for the equations of motion
- m(r, t) - (1i
2
/2m)V
2
1/J(r, t) + V(r )I/J(r, t)
+ f I/Jt(r', t)U(r, r' )I/J(r', t)dr' I/J(r, t) = 0 . (14)
This is the field equation which is assumed to govern
the matter field. It is Schrodinger's equation for a
single particle moving in an external potential V(r) to-
gether with a term U(r, r') representing the interaction
of the field with itself. The latter term is required if
the field is ultimately to describe the properties of a
system of particles with two-body interactions.
Using the Lagrangian . given in Eq. (13), assuming
6
the field equation, Eq. (14), the change in action is giv-
en by
.:l'\\'={ f(7rOI/J- 01TI/J)dr+.(t)6t} I:: ' (15)
Where, in correspondence with particle mechanics the
momentum 1T(r, t) conjugate to the field variable I/J(r, t)
is defined by
(16)
By defining the Hamiltonian of the matter field :JeW in
analogy with the classical expression for H(t)
(17)
to give
7
JC(t) = f - il/J jar - .(t) ,
one obtains
;;C(t) = f {(1i
2
/2m)Vl/Jt. VI/J + Vl/Jtl/J
+tl/Jt(r)f I/Jt(r')U(r, r')I/J(r')I/J(r)dr'}dr. (18)
USing these definitions the change in the quantum action
integral may be re-expressed as [using as well, Eq.
(10)],
(19)
By appealing to Schwinger's action prinCiple, we may,
from Eq. (19), identify the operator &(t),
(20)
as the generator of all possible infinitesimal unitary
transformations, both spatial and temporal. For a pure-
ly temporal change, .:l1/J = 0 and the generator of the in-
finitesimal unitary transformation caused by a displace-
ment of the time is -fc(f)ot. From Eq. (6), the change
in any operator a caused by such a infinitesimal unitary
transformation is
oa = - (i/n)[fc(t), a]Oi = - dot, (21)
which yields Heisenberg's equation of motion from the
action principle
a. = (i/n) [;;C(t), d] . (22)
For a purely spatial change, 01/J is arbitrary, ot= 0
and the generator in Eq. (20) is given by
s(t) = S (7rol/J- o7rl/J)dr. (23)
This generator of spatial changes is composed of two
independent operators, 3
8001= f 7rol/Jdr,
which transforms I/J into I/J + ol/J/2 and
SI>r =- f o7rl/Jdr,
which transforms 7r into 7r + 07r/2. The commutation re-
lationships are derived by conSidering the separate ac-
tions of 9
011
and \1tir on I/J and 7r, respectively. One has
and
1 ') [ , ) ]
2tlfOl/J(r ,t = I/J(r , t , So.
O=[I/Jt(r', f), SOoI]
for the action of SOil and
tinol/Jt(r', t) = [I/Jt(r', f), So.]
and
0= [S6r' I/J(r', f)]
(24)
(25)
for the action of Stir. Schwinger
3
has shown that ol/J and
07r commute with the field operators I/J and 7r except when
they deSignate a field with half-integral spin, in which
case they anticommute. Thus, from Eqs. (24) and (25),
one obtains, respectively,
i drfl/J(r', f), 7r(r, t)l.ol/J(r, t) =tinol/J(r', t) ,
o
L dr07r(r, t)[I/J(r, f), 7r(r', t)]. =tm07r(r', f) ,
where the - and + subscripts denote commutator and
anticommutators, respectively. As 01/J and 07r are arbi-
trary, one derives the commutation relations,
[I/J(r', f), 7r(r, t)J. = hno(r' - r) .
(26)
J. Chem. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
3683
Thus, from a sing' e variational principle, one obtains
the equations of motion, the canonical equations and the
commutation relations. We now address ourselves to the
problem of applying the same principle to a subspace.
B. The subspace Lagrangian
To determine the change in action over a subspace (0)
of the total system, one must construct a corresponding
subspace Lagrangian (0), the variation of which yields
the correct field equation, Eq. (14), and which reduces
to the all space Lagrangian when the boundaries of (0)
are at infinity. This is accomplished by restricting the
integration of the coordinate r in the all space Lagrang-
ian, Eq. (13), to the region (0) and by the imposition of
constraints on the variation. The constraints are an ex-
pression of the fact that the definition of a subspace re-
quires a partitioning of the potential energy of interac-
tion between the subspace and the remainder of the sys-
tem into contributions belonging to each component.
Consider the following subspace Lagrangian expressed
in terms of the trial density cpt(r)cp(r),
(CP, 0, t) = f drHiJf(cptq> - 1> tcp) - (1f2/2m)Vcpt. Vcp}- f dr Vcptcp
o 0
-{L drcpt(r) f dr'lj/(r')U(r, r')IjJ(r')cp(r) - Lo drljJt(r) f dr'ljJt(r')(- r' V'U(r, r'IjJ(r')IjJ(r)} . (27)
In Eq. (27), IjJt(r) and ljJ(r) denote field variables which
satisfy the field equation. They are not to be varied in
the determination of Their appearance in Eq.
(27) indicates the presence of constraints in the varia-
tion of and on the value of the potential energy ariSing
from the se If- interaction potential U.
In the variation of the all space t" Eq. (13), both
ljJ(r) and ljJ(r') must be varied in the self- interaction
term. The factor of t multiplying this term in will
yield the correct contributions from the self-interac-
tion term to both the field equation, Eq. (14) and to the
field Hamiltonian, Eq. (18).8 Limiting the integration
over r in this term to the region (0) in the case of a
subspace, while yielding the required results for the
field Hamiltonian, does not yield, upon variation, the
result required for the field equation. This difficulty
may be overcome by taking cognizance of the fact that in
defining the energy of a subspace one must partition an
energy of interaction of the subspace with the remainder
of the system into separate contributions for both re-
gions. This partitioning is accomplished through the use
of the vi rial sharing operator
1
,2 - r' . V' which appears
in the expression for t,(0) as given in Eq. (27). One
may view the self-interaction of the field as one between
the field to be varied, cpt(r)cp(r) and another IjJt(r')IjJ(r'),
fixed by the field equation. The first of the self-inter-
action term s in Eq. (27) states that cpt (r) and cp (r) are
to be varied in the presence of a fixed repulsive poten-
tial energy at r generated by
f dr' IjJt(r')U(r, r')IjJ(r')
This variation yields the result required for obtaining
the field equation from the variation of Because
of the identity
(-r V-r' V')U=U (28)
which holds when U is derived from an inverse square
force, one may interpret the operator - r' . V' U, appear-
ing in the final term of Eq. (27) as one which determines
the portion of the interaction U which belongs to the
primed system. Thus this final term is a number which
fixes the fraction of the self-interaction energy belong-
ing to the primed field. Its value is determined by de-
manding that
L drljJt(r) (r V) J dr'/fi t(r')U(r, r' )1jJ(r')1/J(r)
= i drljJt(r)f dr'1/Jt(r')(r'. V'U(r,r'))IjJ(r')1/J(r), (29a)
o
or equivalently
L drf(r)(r. V) fa. dr'/fit(r')U(r, r')IjJ(r')IjJ(r)
= f dr/fit(r)i dr'ljJt(r')(r'. V'U(r, r'1/J(r')IjJ(r) (29b)
Cl 0'
where 0 and 0' denote the subspace and the remaining
space of the total system, respectively. The condition
given in Eq. (29) is a statement of the physical require-
ment that the energy of interaction between identical
particles, (or fields) one in 0 and the other in 0', be
equally shared between them. Only if the restraint giv-
en in Eq. (29) is satisfied do the separate portions of
the same field in 0 and 0' exhibit identical average be-
havior with respect to their interaction. 1
Because of the identity given in Eq. (28), the self-in-
teraction terms in (0) reduce, at the point of variation,
to the lhs of Eq. (30)
L dr IjJt(r)(_ r V) f dr' /fit(r')U(r, r')1/J(r')1/J(r)
drljJt(r)f dr'1/Jt(r')U1/J(r')/fi(r). (30)
Finally, because of the constraint governing the sharing
of the self-interaction energy between the subspaces 0
and 0' as given in Eq. (29), one obtains the identity giv-
en in Eq. (30). When 0 refers to all space the rhs
yields the correct expreSSion for the self-interaction
energy as required for the field Hamiltonian. Thus
(CP, 0) yields the field equation when varied, and .(CP,O)
reduces to the all space Lagrangian integral when the
boundaries of 0 occur at infinity.
In defining the energy of a subspace, one cannot make
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
3684
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
reference to an external potential. One must instead
determine what fraction of the interaction of the source
V with the system belongs to the system and what frac-
tion belongs to the source. While such a partitioning of
the contribution of V to (Ij;, n) is not required to obtain
the field equations, (as it is in the case of the self-in-
teraction energy), one may add the appropriate virial
restraint to (Ij;, n) to obtain a field Hamiltonian which
yields an expression for the energy of just the system
rather than of the system and the source.
For an external potential V derived from an inverse
square force, one may write
2
(- r 1 V 1 + <p.) V = V , (31 )
where - r
1
V
1
and cP. are the virial sharing operators
1
for the field and the source respectively. In a molecular
system, within the Born-Oppenheimer approximation
for example, the nuclei are the source of V and
tVe=-LRa' Va
a
Thus the quantity 1J.(n) where
(32)
is a number giving the share of the total interaction en-
ergy which belongs to the source. In a molecular sys-
tem, the term - Va Vlj;t (r)1j; (r) expresses the force ex-
erted on nucleus a by the field density at r and for some
fixed position of the nucleus at some time f, this force
and its corresponding virial have fixed values as deter-
mined by Ij;t(r)lj;(r). The addition of 1J
e
(n) to n)
will not alter the variational result but will, at the point
of variation, reduce the contribution of V to the energy
to the term
r (_ r. V VJlj;t (r )I/!(r)dr = i (V - cP
e
V)lj;t(r )Ij;(r )dr ,
JQ Q
which is the definition of the portion of the external po-
tential energy which belongs to the system.
C. Variation of the subspace action integral
The subspace action integral operator '\\'(n) is defined
as
A J.
t
2 A
'\\"(n) = .(n, f) dt
t1
(33)
where .(n, t) is the integral operator defined in Eq.
(27). The variation of involves a variation of
(n, t) over the domain of the subspace (denoted by
and a variation of the surface bounding the subspace
(denoted by s).
,
where is given by the expression for in Eq.
(11) with the all space replaced by (n) and
We consider first the domain variation of W(n).
From a comparison of the expressions for .e(n, t) and
(t), Eq. (13), and from the above discussion concern-
ing the variation of the self-interaction term, it is clear
that the result of the domain variation of w(n) will con-
tain a set of terms which are simply the subspace ana-
logues of those obtained in the variation of the all space
action integral W, as given in Eq. (12). In addition one
obtains two surface terms. At the point of variation
where the field equation is satisfied, the expression for
is [compare with Eqs. (15) and (19) for
'\\7 (n) = {L - - ic(n, t)/it} \::
- f dSVlj;t. nQl/!+hnfdS </itOIj;}dt+ c. c. ,
(34)
where in analogy with the all space definition, the sub-
space field Hamiltonian is defined as
ic(n, t) = i - n-Ij;)dr - (n, t)
Q
Since (n, f) is a functional of VIj; and of $, the use of
Green's theorem to transform the integrals resulting
from the variation of .c(n, t) with respect to oVI/! and
[as detailed in Eqs. (38a) and (38b) of the preceding pa-
per2] yields the surface terms in Eq. (34). These same
terms vanish in the all space variation as olj; and olj;t
vanish on the surface at infinity.
We now consider the contribution to the change in the
subspace action integral which arises from a variation
of the (finite) surface of the subspace. The surface
S(r) is assumed to be a function of the (real) observable
charge denSity p(r) of the system, which in turn is de-
termined by the state function >II(r, f). Thus an infinites-
imal change in >II causes an infinitesimal change in S(r).
Subspaces with particular properties are variationally
defined when the variations in >II are restri cted to those
which satisfy the constraint, 1,2
(35)
a condition which is necessarily satisfied by ,demanding
that the surface of the subspace always be defined by the
zero flux condition
Vp(r;>II*,>II)' n(r)=O "rES(r).
(36)
The variational constraint, Eq. (35), and the condition
of zero flux through the surface, Eq. (36), restated in
terms of the field density operator p(r) = Ij;t(r )Ij;(r) are
(37)
and
(>IIlvp(r). n(r)I>J!)=0
(38)
The variational constraint in Eq. (37) may be interpreted
either as one which restricts the variation in the state
functions such that
or as one which restricts the variation in the field oper-
ator such that
J. Chem. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II 3685
(>It I 6 L VZp(r)drl>It) =0
(39b)
for any variation of the field variables 1j/(r) and 1/J(r).
That is, the only admisSible variations are those which
satisfy Eq. (39). Therefore, in the variation of the field
theoretic action integral, the variational constraint cor-
responding to the zero flux condition on the gradient of
p(r), may be restated as
6 L VZp(r)dr = 0 . (40)
It is to be understood that Eq., (40) has phYSical meaning
only when averaged over the state function >It(r, t) which
determines the charge density and hence the surface.
The contribution of the surface variation to the change
in action is given by the variation of the surface of
,c(0, t),
(41 )
where L is the Lagrangian density obtained by the sur-
face variation: The surface of the subspace integral
labelled by an 0
0
, appearing in the constraint term of
(0, t), Eq. (27), is not varied as it is determined by
the correct field variables. Thus the Lagrangian den-
sity appearing in Eq. (41), is expressible as
V<t>- f dr'1/Jt(r')
xU(r, r')1/J(r')<t>(r)]}+c. c. - (h
z
/4m)V
z
(<t>t<t. (42)
This result is in accord with the corresponding result,
Eq. (44), obtained in the treatment of the surface term
in 6
s
(0) in the preceding paper. 2 Thepresenttreatment
of6
s
(O, t) parallels this previous case2, given in the
SchrBdinger picture. At the point of variation where the
field equation applies, Eq. (42) reduces to
(43)
and the surface integral obtained from the variation of
w(O) may be expressed as
=- (1/z/4m) fZ fdS6SVZp(r)dt. (44)
11
The constraint, Eq. (40), imposed on the variation of
w(O) leads to the identity
6{Ia VZp(r)dr} = L 6 (vzp(rdr+ fdS6SVZl>(r)=0 , (45)
and hence Eq. (44) may be re-expressed as
= (1/z/4m) fZ i 6 (vzp(rdr dt .
11 0
(46)
Evaluation of the variation of the divergence term in Eq.
(46), which contributes only a surface term, 2 and its
addition to the result for Eq. (34), yields
={ fo - - X(0)6t}I::
1 lIZ A }
-2" {s , (61/J, Vp n=0)+CP'(61/J, Vp. n=O) dt+c. c.
11
(47)
for the change in action for a subspace bounded by a
surface satisfying the zero flux condition, Eq. (1). The
surface integrals S' and <P' in Eq. (47) are analogous ex-
pressions to those obtained in the variation of Wand
appearing in the subspace variational expreSSion for
6(0). Z The integral operator S' is the flux in the vec-
tor current across the surface of the subspace generated
by 61/J,
S '(61/J, V p. n = 0) = (1/z/2m) fdS{(V1/Jt)61/J - 1/J
t
V(61/J)}. n ,
(48)
while <P' is the change in the value of generator h1/1/J
t
61/J
ariSing from a change in the surface with time
(49)
With the exception of the contribution - X(O)6t resulting
from the variation of the time end pOints, (and thus with
= 6), the result given in Eq. (47) is in complete cor-
respondence with the result given for the variation of
6 W(O), Eq. (46), of the preceding paper. Z
In field theory, the operator X(O), conjugate to tin
the expression for the change in action is the (4,4) com-
ponent of the stress energy tensor, and by definition is
the energy of the subfield, 9 i.e., the field contained in
the subspace. If the variation of the subspace action
integral is carried out as above, but with the addition
of the constraint which fixes the share of the external
interaction energy which belongs to the source, Eq.
(32), then the only change in the result for (0) is in
the expression for the field Hamiltonian. The new ex-
pression is
Jet (0) = (1/z/2m) ( V1/J
t
V1/J dr + i (- r VV)1/Jt1/J dr
Jo 0
+i L dr1/J
t
(r) f 1/Jt(r')U(r, r')1/J(r')1/J(r) ,
which differs from X(O) only in that the value of the ex-
ternal potential energy is now that of the subfield alone,
rather than that of the subfield and source. The averag-
ing of X'(O) over a particular state function yields the
energy of a subspace E(O) as previously defined
1
,z,10
E(O) =( >It I Jet(O)1 >It) .
Because of the zero flux boundary condition imposed on
the subspace, the average value of its kinetic energy as
determined by the averaging of the field Hamiltonian
Jet(O) is the same as that obtained in the usual Schro-
dinger representation. Thus, the definition of the ener-
gy of a subspace of a total system, where the potential
energy of the subspace is defined in terms of the virials
of the forces exerted on it, may be obtained from the
canonical formalism of field theory. 11
D. The subspace stationary action principle
The expression for the change in action over a sub-
space, Eq. (47), is best interpreted using a four-dimen-
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
3686
Bader, Srebrenik, and Nguyen-Dang: Subspace Quantum dynamics_ II
y
FIG.!. Pictorial representation of the space-time develop-
ment of a 2-D subspace (l(t) of a total system. The spacelike
surfaces are denoted by (l(t), each being bounded by a surface
S(r;t). The collection of the latter constitute a timelike sur-
face, which, together with (l(t!) and (l(t
2
), defines the integra-
tion volume in the subspace action integral w(fl). The prin-
ciple of stationary action then states that the total change in
wW) has only contributions from the space-time surface
bounding that volume.
sional space-time reference system. ConSider, for
simplicity, a system with two spatial dimensions. Take
the time axis to be perpendicular to the spatial plane.
The change of the system with time then generates a
cylinder along the time axis (Fig. 1). The cylinder is
capped by two "spacelike" surfaces, surfaces in which
all points are defined by the single time tl or t
2
The
wall of the cylinder is a "time like" surface, the position
of the wall being determined at each til11e t by the spatial
boundary of the system.
For a system with boundaries at infinity, Schwinger's
action principle, Eq. (3) or Eq. (15), states that the
total change in action is equal to the difference in the
values of the generator of a corresponding unitary tranS-
formation in the two space like surfaces, that is, to the
difference in the values of the spatial integral eval-
uated at tl and t
2
The corresponding space like contri-
butions appear in the expression for but the in-
tegrations which determine the values of the generator
are limited to the domain of the subspace in each of the
space like surfaces. For a system with finite boundaries
the total change in action includes a contribution from the
timelike surface as well. That is, in the general case
is determined by an integration over the total sur-
face area of the cylinder swept out by the motion of the
system through time. The integral over the timelike
surface determines the change in action resulting from
the infinitesimal transformations both in and on the spa-
tial boundary of the system which occur between the
times tl and t
2
.
In accordance with Schwinger's action principle, we
identify the operator attached to the spacelike surfaces
in the expression for (0) as the generator ;J (t) of the
infinitesimal unitary transformation which causes the
change in the quantized system. Thus, one obtains
(50)
which is identical to the all space result. Since the
operator ff (t) generates an infinitesimal unitary trans-
formation over the total system, the limits On the inte-
grals appearing in Eq. (50) are not restricted to the re-
gion (0) (which would imply incorrectly, that one per-
forms a unitary transformation over the subspace
alone). The integrations are so limited in the expres-
sion for .6.,\\'(0), as the change in action is determined
by the difference in the values of the generator for the
system in question, namely one extending over (0), at
the times tl and t
2
Thus, the expression for the change
in action is used, as it is for a system extending over
all space, to define the generator of infinitesimal unitary
transformations. The statement of the principle of sta-
tionary action for a subspace bounded by a surface of
zero flux in the gradient of p(r), Eq. (1), is as follows;
the change in the subspace quantum action integral is the
difference between the values of the generator of a uni-
tary transformation as determined within those regions
of the two space like surfaces defined by the boundary of
the subspace at times tl and t
2
, plus an integral over the
time like surface corresponding to the time integral of
the change in the value of the generator as determined
by the flux in its vector current across the spatial
boundary of the subspace and by the change in the bound-
ary with time.
E. Unitary transformations and subspace properties
Throughout this work, we have applied the calculus of
variations in the evaluation of the changes in the sub-
space energy or action
2
caused by infinitesimal varia-
tions in either the state fUnctions >J.!(r, t) and >J.!*(r, t), or
in the field variables 1/i(r, t) and 1/i
t
(r, t), which COnserve
the zero flux condition. We noW focus our attention on
changes in any subspace property induced by infinitesi-
mal unitary transformations, changes which we shall
ultimately require to equal those obtained through the
use of the calculus of variations. The relationship be-
tween the change in a subspace operator and the action
of the generator of an infinitesimal unitary transforma-
tion is determined by deriving the subspace analogues
of the commutator expressions given in Eqs. (6) and
(22).
This discussion will make use of the coordinate rep-
resentation which is expressed in terms of the complete
set of local state vectors
I q> = I r
1
, r
2
, ", rN>
and their duals, with the operators 0 and ff: referring to
observables and both expressible in terms of their ma-
trix elements as
(ql&lq'>=o(q'7 (51a)
( q I ff: I q' > = f (q, 7 v.) a (q - q') = ( q' I ff: I q> *
(51b)
J. Chem. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
3687
The operators o(q, (li/i)V
q
) and f(q, (1i/i)V
q
) are con-
structed fz:om the position and momentum operators
q ={r
t
, r
2
, ", r
N
}
(Ii/ i)V q = V h V 2, , V N}
l t l
acting on the delta function
o(q - q') = o(r
t
- r{)O(r
2
- ... o(r
N
-
so as to ensure the hermiticity of tJ and:t, respective-
ly.
The changes induced in a state vector I t;) and an oper-
ator 0, Eqs. (5) and (6), by an infinitesimal unitary
transformation with generator :t as defined in Eq. (4)
ensure that
In terms of the above representation, the identity stated
in Eq. (52) can be rewritten in terms of averages over
the wavefunctions t;(q) and t;*(q) as
f dq{ot;*(q)ot;(q) + t;*(q)o o (q)} = iE: f dq{t;*(q)[o, il t;(q)} ,
(53)
where
and
ot;(q) = id t;(q) , ot;*(q) = - iE:;*t;*(q) .
When one coordinate among the set of N is selected for
integration over a finite region 0, we shall write
I q) == I r
t
, r') .
We are speCifically interested in finite regions which
are bounded by a zero flux surface as defined in Eq. (1).
The subspace average of a property J
(6)0= L drtf' dr''lto'lt (54)
can be written in terms of a projection operator n(o) as
(55)
where ll(O) is defined as the sum of the local one-parti-
cle operators rr(r) = Ir)(rl over 0:
1l(0) = i drtf' dr'lrhr')(rtor'l ={ drlr)(rl
o 0
(56)
and where
(57)
By virtue of the identity in Eq. (52), one then has
(o( 'ltl )1l(O)e I 'It) =( 'ltl Il(O)O (0 I 'It))
=-iE:('ltI[:t,ii(o)o]I'lt), (58)
or, explicitly in the coordinate representation
10 drt r dr'{o'lt*O\J!' + 'It*8c5'lt}
= - iE: So dr1 r - 'It*8i'lt} . (59)
From the calculus of variations as exemplified by the
variation of V.'(O), the infinitesimal change in a sub-
space operator 0(0) caused by a variation of the field
variables IjJt(r) and ljJ(r) consists of the variation of its
functional denSity 0(1jJt, 1jJ) over the domain of the sub-
space and of the variation of the surface of the subspace.
That is,
(60)
which when averaged over the state in question yields
the corresponding expression for the variation in a sub-
space property
o( 6)0 = L dr
1
f' dr'{o'lt*a'lt + 'It*oo'lt}
+ fdS(r1)f' dr'oS(r1){'lt*6'lt} . (61)
From a comparison of Eqs. (59) and (61) it is clear that
the commutator relationship
(62)
describes only the domain variation of 0(0) and of the
subspace property (6) 0' This is understandable, as the
transformation in Eq. (62) changes only the operator,
leaving the state vector unaltered, and thus the boundary
of the subspace is not varied.
Expanding the commutator in Eq. (62) yields
OD0(0) = iE:n(O) [if, el + iE:[iF, IT (0)] 0. (63)
The separate averaging of the commutators in Eq. (63)
demonstrates in a precise manner the important proper-
ty of the domain variation of a subspace operator, name-
ly that it consists of the simple projection of the sub-
space contribution from the total transformation over
all space,
('It I rr(o)[iF, ell 'It) = L dr
1
f dr''lt*[r, a]'lt (64)
plus a contribution which arises from the fact that f does
not possess the property of hermiticity when averaged
over a subspace,
('lti[iF, n(0)]61'lt) = fodrlfdr'{f*'lt*O'lt-'lt*ia'lt}. (65)
Because of the partitioning of oD6(0) given in Eq. (63),
it is clear that if the property 0 is conserved over all
space under the action of the generator then the
whole of 0D(t)O arises from the noncommutativity of if
with n(O). For example, the number operator in the
coordinate representation is NIl, where
ll=ll(O=R
S
)=l,
and since the total number of particles is conserved
over all space under any unitary transformation, one
obtains from Eq. (63)
ODNIl(O) = iE:[&, Il(O)]N= .
Thus the domain variation in a subspace population is
given by
OD( it(O = - N( 'It I iE:[:t, n(o)] I 'It)
= N L dr1 J' dr'{o'lt*'lt + 'It*o'lt} .
J. Chem. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
3688
Bader, Srebrenik, and Nguyen-Oang: Subspace quantum dynamics. II
In some specific cases the result of averaging the com-
mutator of and n(O), the rhs of Eq. (65), may be
transformed into an integral over the surface of the
subspace. For example, when
one has
\
[
A A A \ 1f2 i S'
- ('k JC, II(U)] t) 'k) = 2m rdS(rl) dr'
X{('Vl'k*)O'k- 'k*'v't(u'k)}. (66)
The integral on the rhs of Eq. (66) is the surface inte-
gral S'(6'k,Vp. n=O) as previously defined, 1,2 a quantity
which represents the flux in the vector current density
of the property () through the surface of the subspace.
Since -:fcot is the generator of an infinitesimal temporal
,.
change, one finds that the contribution to (0(U) from
the temporal variation over the domain of the subspace
is given by
(i/If)('k/rr(u)[:k,8}j'k)-S'(6'k,Vp. n=O). (67)
If the results of the above transformations are to be
identical with those obtained through the calculus of var-
iations as given in Eqs. (60) and (61), we must infer the
existence of a surface variation term in both the average
value and the operator expressions. That is,
= i( ('k/ [0(0), ] / 'k) +( 'k/ osIT(U) 0 / 'k) ,
and [compare with Eq. (60)],
o8(U)=iE:[,0(u)1-osn(u)8.
(68)
(69)
The symbol osIT(O)O is to denote the surface variation
resulting from the change in the domain of integration
as caused by the unitary transformation. The quantity
osll(U) denotes the variation in the projection operator
Il(U) resulting from the change in the state vector l'k)
and its dual as caused by the unitary transformation.
We now proceed to derive its analytical expression for
the case when the surface of U is one of zero flux in V p.
As noted above, the average of .Nii(0) in the state
I 'k) gives the population of the subspace U, and the
commutator iE:[, .Nii(U)] averages to the domain varia-
tion of the population. Thus one must have
('k\ 0 sIT (0) \ 'k> = os( &(U) = dS(r
1
) r dr' oS(rl)'k*'k ,
(70)
or equivalently
osll(U) = tdS(rl) f dr' oS(r
l
) / rl> r') (rl> r' \
= tdS(r)os(r)\r)(r\ . (71)
Indeed, osr1(U) can be expressed as
OsIT(U)=IT(U')-fl:(U)= (Ir)dr(r\-i /r)dr(r/, (72)
JOI 0
where U' is the volume bounded by the surface S such
that
'tirES, Vp'(r)n(r)=O, (73)
with
p' (r) = ( 'k' / r) ( r / 'k') , (74)
where
IntrodUCing
I r') = {ell r)
and its dual
( r' / = ( r I <U ,
one has from Eq. (74),
p'(r)=p(r') . (75)
Equation (75) states that the transformed density equals,
at any point r, the value of the original denSity measured
at the point r' provided that
I r') = i-ll r> or I r) =<u1 r') .
Bearing in mind the infinitesimal nature of the trans-
formation, we can assume the existence of an infinitesi-
mal shift or, such that
r =r' + or
and (to first-order)
n(r) = n(r') .
The zero-flux condition Eq. (73) on the transformed sur-
face S, can be written as
Vp(r'). n(r') = 0, r' = r - or, 'tIr E S' , (76)
and thus r' E S. This yields the interesting result that
the transformed surface S' can be obtained from the
original surface S by shifting any vectors r' E S by the
amount or such that
I r' + or> = 'lil r') .
We may now obtain an explicit formulation of the rhs
of Eq. (72), that is, of the variation of the surface in-
duced by a unitary infinitesimal transformation on I >It).
Performing the change in variable
r = r' + or
in the integral over U' and using
Ir'+or)(r'+or/ =V(!r)(rl)or+lr')(r'\,
we obtain to first-order, and with oS(r) = or n(r) the
result
osIT(U) = f dS(r)oS(r)I r) (r \ ,
(77)
which is identical to the result given in Eq. (71), the
form required to yield agreement with the calculus of
variations. Therefore, the change induced in a surface
of zero flux by an infinitesimal unitary transformation
is of the precise form required by the calculus of vari-
ations.
The utility afforded by the above formalism in the
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
3689
treatment of subspace properties is illustrated by a
derivation of the subspace expression for the stationary
state variational principle,
IlS'(a'1/!, O)=H([:if, a'])o +c. c.}
discussed in the preceding and derived original-
ly by the calculus of variations. 1 The energy functional
S' (1/!, 0) may be expressed in terms of the proj ection
operator IT(O) as [compare with Eq. (8) of the preceding
paper ],
S'(1/! +c.c.} +tfo
v2
p(r)dr .
, 2 (1/! III (U) I1/! > ( 1/!, 1/! > 0
Because of the variational constraint on the volume 0,
Eq. (39a) and the zero flux surface condition, the varia-
tion of (1/!, 1/! >0 S' is equal to the variation of the average
value of :iC(0) = Hrr(o)fc+:JCIT(O)), [compare with Eq. (11)
of the preceding paper]. Thus, with ei' = iE:a, consid-
ered as the generator of an infinitesimal unitary trans-
formation, one obtains, through the use of Eq. (68),
s' (a1/!, O){iE: (1/!1 [rr(O), a]l1/! >
+(1/!lll
s
IT(O)I1/!>}+( 1/!1 IT(O)I1/!> IlS'(cl1/!, 0)
=i( 1/!1[:JC(O),cl] I1/!) +H(1/!lll
s
IT(o)i;I1/!) +c. c.} .
Expanding the commutator on the rhs of the above equa-
and recalling that 9' = E when 1/! is an eigenfunction
of JC, one obtains
(1/!1 IT(o)I1/!) Il S'(cl1/!, 0) = HiE:( 1/!1 IT (O)[JC, a]I1/!) + c. c.}
or equivalently,
Il S' (<i1jJ, 0) = HiE:( [:if, ei])o/( l/i, 1/!)0 + c. c.} .
Thus the variation of S'(1/!, 0) is given by the average
of the subspace projection of the commutator of :JC and
the generator el' of the infinitesimal unitary transforma-
tion.
The expression detailing the change of a subspace op-
erator under an infinitesimal unitary transformation in
the coordinate representation may be restated in field
theoretic language by using the following formal corre-
spondences between two sets of definitions: for a sub-
space operator the correspondence is,
0(0) = fi(O)o - = fa drO (1/!t, 1/!) ,
and a matrix element in the coordinate representation
is replaced by the corresponding functior,al denSity of
the property
(qI8Iq)=S(q,
The change in a subspace operator caused by an in-
finitesimal unitary transformation is then given by
Il 8(0)" zg, 0(0)] + fdS(r)IlS(r)8 (1/!t, 1/!) , (78)
where the infinitesimal generator f:1F has been replaced
by 9 = f 9 (l/it, 1/!)dr, where ?;(1/!t, 1/!) is the functional denSity
of the generator. The two terms on the rhs of Eq. (78)
correspond respectively to the domain variation of 0(0),
IlDO(O) and its surface variation, Ils0(0).13 For a purely
temporal change, the Heisenberg equation of motion for
a subspace operator becomes
0(0) = (i/Ii)[:JC(t), 0(0)] - (i/Ii)&>' (e, V p' n = 0) , (79)
where from Eq. (50), :JC(t) is the field Hamiltonian as de-
fined in Eq. (18) and where the surface integral cP',
is the same as the corresponding integral obtained for
the change in action over a subspace and defined in Eq.
(49) with the functional density for 6 substituted for the
density 1/!
t
o1/!. In general, the commutators in Eqs. (78)
and (79), which in both cases correspond to the domain
variation of the operator, will yield the same result as
for the all space operator 6, plus in some cases, an
additional surface term.
Consider as an example the time rate of change of
the subspace number operator
(81)
From the general discussion of the commutator relations
for subspace operators, one anticipates [see Eq. (66)]
that the commutator of Je(t) and should yield only a
surface term S'. In the particular language of field the-
ory one indeed findS that the evaluation of the commuta-
tor, using either the boson or fermion commutation re-
lations for the field operators, Eq. (26), yields the ex-
pected surface integral,
(82)
In this instance S', as defined in Eq. (48) with o1/! re-
placed with 1/!, reduces to the flux of the vector current
through tre surface of the subspace. Thus, the expres-
sion for may be written as
(83)
where the current J is
Equation (83) generalizes the result obtained from the
usual equation of continuity by accounting for the change
in number of particles resulting from a change of the
surface with time. The result given in Eq,. (83) is
equivalent to the expression obtained for N(O) in the
Schr'odinger representation in the preceding paper.
2
Corresponding expreSSions for other subspace proper-
ties, such as the equation for may be derived in
Similar fashion.
F. Change in action over a subspace and the quantum
action principle
From the prinCiple of stationary action as expressed
in Eq. (3) one may obtain the fundamental relationships
of quantum mechanics. 3 This principle is itself the re-
sult of a more fundamental dynamical prinCiple, the
quantum action principle. Denoting an arbitrary state
of a system by I a, t> where a denotes the eigenvalues
of some complete set of commuting Hermitian operators,
the quantum action principle is a differential expression
for the transformation fUnction ( a', t1 I a", t
2
), which is
like a probability amplitude for a change in the represen-
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
3690
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
tation of the state at t1> through a unitary transforma-
tion, to a new representation at tao The principle as
given by Schwinger is
3
6( a', tll a", t
2
) = (i/1i) ( a', tll a", f
2
) , (84)
which states that changes in the transformation function
are all derived from variations of one operator, name-
ly the action operator. It was also shown
3
that the ac-
tion operators and their variations obey an additive law
of composition with respect to the defining times:
W13 =W
12
+W
23
and
6W13 = 6-\'\'12 + 6W
23

Exploiting the above additive property, one may con-
struct the action operatorW
12
as the time integral of a
Lagrangian (t).
A J.
t
2 A
'W
12
= df(t) .
tl
Equivalently, the Lagrangian (f) can be defined through
Wt, t+dt = (t)dt . (85)
On the other hand, the possibility of a well-defined
partitioning scheme {Ol (t); i = 1,2, ... , n} at any time t
by the zero-flux condition, ensures the spatially additive
property of the action:
and its variation
where
A J.
t2
A
\\'(0
1
)12 = df(OI(t)) ,
tl
(0
1
(t being the proj ection of (t) on the subspace
0
1
(f).
(86)
Thus the action principle may be expressed as a sum
over the change in the action operator for each sub-
space in the total system,
or
tll a", t
2
) = L: A(OI; a'; t
1
; a", t
2
) , (88)
I
where A(OI) is the contribution to the total change in the
transformation function as determined by the change in
action for a particular region 0
1
of real space.
Consider the transformation function linking two states
which are separated by an infinitesimal time interval.
The contribution to the total change in the transforma-
tion function from a particular subspace 0 is
A(O; a't', a", t+dt) =( a', tl a", t+dt) , (89)
which, by virtue of Eq. (85) may be expressed (for an
infinitesimal time increment) as
A(O; a', t; a", f+ df) =( a', tl 6(0). dtl a", t+ dt) (90)
Equating the two operators in Eqs. (89) and (90) and
using the expression for (0) in Eq. (47) one obtains
6.(0) = tf(O, Vp n=O)
+ 6>' (61j!, Vp n=O)+c. c.}. (91 )
In Eq. (91), &(0, f) denotes the time derivative of the
generator evaluated over the subspace. For a purely
spatial change with 6 t = 0, the generator reduces to
S(t), Eq. (23) and its time derivative evaluated over the
subspace is
(92)
To obtain the equivalent of the commutator expression
for 6(0) as given in the preceding paper2
6(0)=-H([x, al>n+c.c.}, (93)
one must substitute for the generator an operator corre-
sponding to some particular property, multiplied by an
infinitesimal,
(l/iIi)S(t)=a(t)=hS (Ij!tfilj!+fitlj!tlj!)dr (94)
As in the classical case, 2 when the property a has no
explicit time dependence its time derivative can be
equated to its commutator with:!C. With these restric-
tions the term t) in Eq. (91) can be replaced by
the time derivative of the subspace property a(o, f),
(95)
and Eq. (91) becomes
6(0)=-[:iC, a(o)]-HS'(nlj!, Vp. n=O)+c.c.}. (96)
Recalling [see Eq. (67)] that the commutator expression
for the temporal variation over the domain of a subspace
implicity includes the surface term - Hs' + C. c.}, Eq.
(96) is the analogue of the subspace variational principle
for (0),2 Eq. (93).
III. SUMMARY AND CONCLUSIONS
It is possible to use the quantum action principle and
the concepts of field theory to both define and deter-
mine the properties of a subspace of a total system.
The construction of the subspace Lagrangian which
yields the quantum field equations from the principle of
stationary action and which is applicable as well in the
limiting situation of all space, requires that one define
the energy of the subspace. Such a definition in turn re-
quires a partitioning of the energy of interaction of the
subspace with the remainder of the system into a part
belonging to the subspace and a part belonging to the
remainder of the system. As previously discussed
2
this
is accomplished through the definition of a single-par-
ticle potential energy, a concept which finds a natural
expression in the language of field theory.
The restraint ariSing from the zero flux surface con-
dition defines a subspace for which the principle of sta-
tionary action assumes its simplest physical form. The
change in action for a subspace must necessarily include
a contribution from the timelike surface, which is phys-
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Bader, Srebrenik, and Nguyen-Dang: Subspace quantum dynamics. II
3691
ically interpretable as resulting from the infinitesimal
flux in the current of the generator across the spatial
boundary and from changes in the boundary of the sys-
tem with time. Without the zero flux restraint there is
an additional contribution to the change in action from
the time like surface, namely,
(1I2/4m) f2\S i V'2(ljitlji)dr}dt.
t1 l 0
This quantity is of arbitrary value and appears to have
no physical meaning other than the one obtained by the
requirement that it vanishes as it does in all space phys-
ics. Finally, we note that the set of surfaces as defined
in Eq. (1) by a property of the charge denSity, partitions
a molecule into a collection of chemically identifiable
atomiclike fragments. 2,14 These are the subspaces which
are summed over in Eq. (87). Thus, the quantum action
principle is expressible as a sum of the separate changes
in action over each atom in a molecule.
ACKNOWLEDGMENTS
We wish to thank Professor Y. Nogami and Professor
J. P. Carbotte for a number of valuable discussions
concerning this work.
IS. Srebrenik and R. F. W. Bader, J. Chern. Phys. 63, 3945
(1975).
2S. Srebrenik, R. F. W. Bader, and T. T. Nguyen-Dang, J
Chern. Phys. 68, 3667 (1978), preceding paper.
3J Schwinger, Phys. Rev. 82, 914 (1951).
4p. Roman, Advanced Quantum Theory (Addison-Wesley, Read-
ing, Mass., 1965), p. 37, p. 81.
5J Schwinger, Brandeis Summer Inst. Theoret. Phys. 2, 157
(1964).
sIn deriving the principle of stationary action, 8chwinger
3
was
able to demonstrate from general considerations of the prop-
erties of infinitesimal unitary transformations that the change
in action is given solely by an expression of the form of Eq.
(3) or Eq. (15). A comparison of Eqs. (12) and (15) for
then shows that the equations of motion must be obtained as a
consequence of the principle of stationary action.
TThere are two independent fields employed in the definition of
.c , Eq. (13). They are 1/i and 1/i t The momentum conjugate to
1/i is, by Eq. (16), rr=!ifi/2)1/it. Correspondingly, the momen-
tum conjugate to 1/i
t
is, again using Eq. (16) but this time dif-
ferentiating with respect to rrt = - (ifi/2)1j:. Thus, the quan-
tum analogue of pq for the two field case is -;1/i. Alterna-
tively, one may consider simply adding the divergence term
to a Lagrangian density defined for a single field,
i. e., without the term This yields the Lagrangian de-
fined in Eq. (13) without altering the equations of motion ob-
tained from the variational principle.
8Roman4 defined a Lagrangian density for a system with inter-
actions without the factor of in the interaction term, claim-
ing that it will yield the correct field equation by virtue of the
Euler-Lagrange equations. The factor of is then (rather
arbitrarily) added to the interaction term in the definition of
:!CW. However, in the interaction case, the Lagrangian den-
sity is afunctional of 1/;. Thus, in using the Euler-Lagrange
equations both 1/;(r) and 1/;(r') must be varied. Thus the factor
of must appear in the Lagrangian itself.
9
p
M. Morse and H. Feshbach, Methods of Mathematical
PhYSics (McGraw-Hill, New York, 1953), Vol. I.
IOR.F. W. Bader and G. R. Runtz, Mol. Phys. 30, 117 (1975).
lilt is assumed that the origin chosen to fix the value of the ex-
ternal constraint in Eq. (32) is the same as that required to
satisfy the value of the constraint placed on the self-interac-
tion energy, Eq. (29). In molecular systems described by the
Born-Oppenheimer approximation this assumption is numeri-
cally justified for Hartree-Fock wavefunctions if the subspace
is bounded by a zero-flux surface. 1,12
128 8rebrenik and R. F. W. Bader, J. Chern. Phys. 61, 2536
(1974).
13The surface variation Os e(m arises from the dependence of
the surface S(r) on the real charge distribution per) as de-
s'cribed in Eqs. (36) and (38). A change in the state vector
I >It), or in the operator per) induced by S, causes the shift
OS = 0 r' n, in the surface coordinates where, as before,
Ir+or) = (l+iS') Ir) =U I r) vr ES(r)
We also note that the relationship between the transformed
and the original surfaces, Eq. (76), remains formally un-
changed when stated in terms of the operators per) and {I (r)
=Up(r)U-
I

I4R. F. W. Bader, Acc. Chern. Res. 8, 34 (1975).
J. Chern. Phys., Vol. 68, No.8, 15 April 1978
Downloaded 28 Apr 2013 to 137.99.31.134. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

También podría gustarte