Arciniega

INVARIANTS OF HYPERBOLIC 3-MANIFOLDS IN RELATIVE
GROUP HOMOLOGY
JOS
E ANTONIO ARCINIEGA-NEV
AREZ AND JOS
E LUIS CISNEROS-MOLINA
Abstract. Let M be a complete oriented hyperbolic 3manifold of nite vol-
ume. Using classifying spaces for families of subgroups we construct a class
P
(M) in the Hochschild relative homology group H
3
([PSL
2
(C) :

P]; Z),
where

P is the subgroup of parabolic transformations which x in the
Riemann sphere. We prove that the group H
3
([PSL
2
(C) :

P]; Z) and the
Takasu relative homology group H
3
(PSL
2
(C),

P; Z) are isomorphic and under
this isomorphism the class
P
(M) corresponds to Zickerts fundamental class.
This proves that Zickerts fundamental class is well-dened and independent of
the choice of decorations by horospheres. We also construct a homomorphism
from H
3
([PSL
2
(C) :

P]; Z) to the extended Bloch group

B(C) which is iso-
morphic to H
3
(PSL
2
(C); Z). The image of
P
(M) under this homomorphism
is the PSLfundamental class constructed by Neumann and Zickert.
Contents
1. Introduction 1
2. Classifying spaces for families of isotropy subgroups 3
3. Relative group homologies 6
4. The group SL
2
(C) and some of its subgroups 16
5. Invariants for nite volume hyperbolic 3manifolds 25
6. Relation with the extended Bloch group 28
7. Mappings via congurations in X
(P)
32
8. Computing
P
(M) and
B
(M) using an ideal triangulation of M 37
9. The invariant
P
(M) is Zickerts fundamental class 40
10. (G, H)representations 46
11. Complex volume 49
References 50
1. Introduction
Given a compact oriented hyperbolic 3manifold M there is a canonical repre-
sentation. :
1
(M) = PSL
2
(C). To this representation corresponds a map
B : M BPSL
2
(C) where BPSL
2
(C) is the classifying space of PSL
2
(C) con-
sidered as a discrete group. There is a well-known invariant [M]
PSL
of M in the
2000 Mathematics Subject Classication. Primary 57M27, 55R35, 18H10; Secondary 58J28.
Key words and phrases. Hyperbolic manifolds, classifying spaces for families of subgroups,
Hochschild relative group homology, Takasu relative group homology, extended Bloch group,
boundary-parabolic representation, complex volume.
1
a
r
X
i
v
:
1
3
0
3
.
2
9
8
6
v
1

[
m
a
t
h
.
G
T
]

1
2

M
a
r

2
0
1
3
2 JOS
AREZ AND JOS
group H
3
(PSL
2
(C); Z) given by the image of the fundamental class of M under
the homomorphism induced in homology by B .
In the case when M has cusps, a construction of [M]
PSL
was rst described by
Neumann in [13, Theorem 14.2] and a detailed proof is given by Zickert in [22].
Suppose that the complete oriented non-compact hyperbolic 3manifold of nite
volume M (i.e., with cusps) is triangulated. From the triangulation a class in the
(Takasu) relative homology group H
3
(PSL
2
(C),

P; Z) is constructed where

P is the
image in PSL
2
(C) of the subgroup P of SL
2
(C) of matrices of the form
_
1 b
0 1
_
.
Then a natural splitting of the map
H
3
(PSL
2
(C); Z) H
3
(PSL
2
(C),

P; Z)
is used to obtain [M]
PSL
. The class in H
3
(PSL
2
(C),

P; Z) a priori depends on
choices of horoballs at the cusps, but the image [M]
PSL
H
3
(PSL
2
(C); Z) is
independent of this choice.
In the present article we generalize the construction given in Cisneros-Molina
Jones [5] using classifying spaces for families of isotropy subgroups to give an alter-
native construction of the class [M]
PSL
for an hyperbolic 3manifold with cusps.
A family F of subgroups of a discrete group G is a set of subgroups of G
which is closed under conjugation and subgroups. A classifying space E
F
(G) for
the family F is a terminal object in the category of Gsets with isotropy sub-
groups in the family F. We consider the case G = PSL
2
(C) and the family
F(

P) of subgroups generated by the subgroup

P. Using general properties of
classifying spaces for families of isotropy subgroups and the canonical represen-
tation :
1
(M) PSL
2
(C) we construct a canonical map

P
:

M B
F(
P)
(G)
where

M is the end compactication of M and B
F(
P)
(G) is the orbit space of
E
F(
P)
(G). We have that H
3
(
M; Z)

= Z and the image of the generator under the
homomorphism (
P
)
: H
3
(
M; Z) H
3
(B
F(
P)
(G); Z) gives a well dened element
P
(M) H
3
(B
F(
P)
(G); Z). Then we notice that the group H
3
(B
F(
P)
(G); Z) is the
Hochschild relative homology group H
3
([PSL
2
(C) :

P]; Z) dened by Hochschild
[10, 4] which in general is dierent from the Takasu relative homology group
H
3
(PSL
2
(C),

P; Z) dened by Takasu in [18].
Then we construct a homomorphism from H
3
([PSL
2
(C) : P]; Z) to the extended
Bloch group

B(C) which is isomorphic to H
3
(PSL
2
(C); Z) by Neumann [13, Theo-
rem 2.6]. The image of
P
(M) under this homomorphism is the fundamental class
[M]
PSL
.
Using the relative homological algebra of [10] and results in [18] we prove that in
the case of PSL
2
(C) and the subgroup

P both relative homology groups coincide,
that is
H
3
([PSL
2
(C) :

P]; Z)

= H
3
(PSL
2
(C),

P; Z).
We also give an explicit expression for this isomorphism using the construction of
H
3
(PSL
2
(C),

P; Z) given by Zickert in [22] using a complex of truncated simplices.
Under such isomorphism we prove that the class in H
3
(PSL
2
(C), P; Z) dened by
Zickert coincides with the class
P
(M) proving that Zickerts class is independent
of the choice of horoballs.
This construction of [M]
PSL
has some advantages:
(1) It does not require a triangulation of M.
INVARIANTS OF HYPERBOLIC 3-MANIFOLDS IN RELATIVE GROUP HOMOLOGY 3
(2) It shows that the class
P
(M) obtained in the (Hochschild) relative homol-
ogy group H
3
([SL
2
(C) : P]; Z) is a well-dened invariant of M which lifts
the classical Bloch invariant (M).
(3) Choosing an ideal triangulation of M one can give a explicit formula for
the class
P
(M). Computationally is seems to be more ecient to use
the group H
3
([PSL
2
(C) : P]; Z) rather than the group H
3
(PSL
2
(C), P; Z)
since it necessary less information to describe elements in the former group
than in the later one.
In [13, Proposition 2.5, Theorem 2.6] Neumann dened a homomorphism
L:

T(C)
C/
2
Z and proved that it corresponds to the CheegerChernSimons class of the
universal at PSL
2
(C) over the classifying space BPSL
2
(C), where PSL
2
(C) is
considered as a discrete group. Hence the image of the PSLfundamental class
under

L gives the complex volume of M (times i), ie
L
_
[M]
PSL
_
= i
_
Vol(M) +i CS(M)
_
C/i
2
Z,
where Vol(M) is the volume of M and CS(M) is (a multiple) of the ChernSimons
invariant.
Our construction also works in the general context of (G, H)representations
of tame nmanifolds considered in [22]. In Section 10 we prove that any G
representation mapping boundary curves to conjugates of a xed subgroup H gives
a well-dened class in the Hochschild relative homology group H
n
([G : H]; Z). In
this general context the Hochschild and Takasu relative homology groups do not
necessarily coincide, the class in H
n
(G, H; Z) constructed by Zickert in [22, 5]
depends of a choice of decoration and dierent classes given by dierent choices
are all mapped to the class in H
n
([G : H]; Z) by a canonical homomorphism
H
n
(G, H; Z) H
n
([G : H]; Z). So in this general context it is more appropriate
to use Hochschild relative group homology than Takasu relative group homology
because we obtain classes independent of choice.
In the case of boundary-parabolic PSL
2
(C)representations of tame 3manifolds
the construction for the geometric representation generalizes and we obtain a well-
dened
P
() class in H
3
(PSL
2
(C),

P; Z)

= H
3
([PSL
2
(C) :

P]; Z) and therefore a
well-dened PSLfundamental class []
PSL
H
3
(PSL
2
(C); Z). Then we can dene
as in Zickert [22, 6] the complex volume of a boundary-parabolic representation
by
i
_
Vol() +i CS()
_
=

L
_
[]
PSL
_
.
Using an ideal triangulation of M we obtain an explicit formula for the complex
volume of .
2. Classifying spaces for families of isotropy subgroups
In this section we dene the classifying spaces for families of subgroups and we
give their main properties. We recommend the survey article by L uck [12] and the
bibliography there for more details.
Let G be a discrete group. Let X be a Gspace. For each subgroup H of G,
we dene the set X
H
= x X [ h x = x for all h H of xed points of H. We
denote by G
x
= g G [ g x = x the isotropy subgroup xing x X. More
generally, let Y X be a subspace, then G
Y
=
yY
G
y
is the isotropy subgroup
(pointwise) xing Y . We also denote by G
(Y )
= g G [ g Y = Y the subgroup
leaving Y invariant. Note that in general G
Y
G
(Y )
.
4 JOS
AREZ AND JOS
Since G is discrete, a GCWcomplex is an ordinary CWcomplex X together
with a continuous action of G such that,
(1) for each g G and any open cell of X, the translation g is again an
open cell of X.
(2) if g = , then the induced map given by the translation x g x
is the identity, i. e., if a cell is xed by an element of G, it is xed pointwise,
in other words G
()
= G
.
See for instance tom Dieck [19, Proposition II.1.15].
Remark 2.1. Notice that in a GCWcomplex X for each open cell of X one
has G
= G
x
for every x . Hence
G
[ is a cell of X = G
x
[ x X,
ie the set of isotropy subgroups of the points of X is the same as the set of isotropy
subgroups of the cells of X.
A family F of subgroups of G is a set of subgroups of G which is closed under
conjugation and taking subgroups. Let H
i
iI
be a set of subgroups of G, we
denote by F(H
i
) the family consisting of all the subgroups of the H
i
iI
and all
their conjugates by elements of G.
Let F be a family of subgroups of G. A model for the classifying space for
the family F of subgroups is a GCWcomplex E
F
(G) which has the following
properties:
(1) All isotropy groups of E
F
(G) belong to F.
(2) For any GCWcomplex Y , whose isotropy groups belong to F, there is up
to Ghomotopy a unique Gmap Y E
F
(G).
In other words, E
F
(G) is a terminal object in the Ghomotopy category of GCW
complexes, whose isotropy groups belong to F. In particular two models for E
F
(G)
are Ghomotopy equivalent and for two families F
1
F
2
there is up to Ghomotopy
precisely one Gmap E
F1
(G) E
F2
(G).
Remark 2.2. There is another version for the classifying space for the family F in
the category of Fnumerable Gspaces (see Luck [12, Denition 2.1] or tom Dieck
[20, page 47] for the denition), but both versions are Ghomotopy equivalent when
G is a discrete group [12, Theorem. 3.7]. Moreover, any GCWcomplex with all
its isotropy groups in the family F is Fnumerable [12, Lemma 2.2], thus we can
work with any of the two versions.
There is a homotopy characterization of E
F
(G) which allows us to determine
whether or not a given GCWcomplex is a model for E
F
(G).
Theorem 2.3 ([12, Theorem. 1.9]). A GCWcomplex X is a model for E
F
(G)
if and only if all its isotropy groups belong to F and the Hxed point set X
H
is
weakly contractible for each H F and is empty otherwise. In particular, E
F
(G) is
contractible.
There are dierent ways of constructing models for E
F
(G). We consider the
following.
2.1. Simplicial Construction. The following proposition gives a simplicial con-
struction of a model for E
F
(G), compare with FarrellJones [9, Theorem A.3].
Let H
i
iI
be a set of subgroups of G such that every group in F is conjugate
to a subgroup of an H
i
, that is, F = F(H
i
). Consider the disjoint union
F
=
iI
G/H
i
, or more generally, let X
F
be any Gset such that F is precisely the set
of subgroups of G which x at least one point of X
F
. Notice that
F
is an example
of such a Gset.
Proposition 2.4. Let X
F
be as above. A model for E
F
(G) is the geometric re-
alization Y of the simplicial set whose nsimplices are the ordered (n + 1)tuples
(x
0
, . . . , x
n
) of elements of X
F
. The face operators are given by
d
i
(x
0
, . . . , x
n
) = (x
0
, . . . , x
i
, . . . , x
n
),
where x
i
means omitting the element x
i
. The degeneracy operators are dened by
s
i
(x
0
, . . . , x
n
) = (x
0
, . . . , x
i
, x
i
, . . . , x
n
).
The action of g G on an nsimplex (x
0
, . . . , x
n
) of Y gives the simplex (gx
0
, . . . , gx
n
).
Proof. Let = (x
0
, . . . , x
n
) be an nsimplex of Y . By the denition of the action
of G and since F is closed with respect to subgroups we have that
G
=
n
i=0
G
xi
F.
Hence by Remark 2.1 we have that all the isotropy subgroups of points of Y belong
to the family F. Let H F, then its set of xed points is given by
Y
H
=
_
(x
0
, . . . , x
n
)
H
n
i=1
G
xi
, n = 0, 1, . . .
_
Notice that if is a cell in Y
H
then d
i
Y
H
and s
i
Y
H
, and therefore Y
H
is
a simplicial subset of Y . Let x X
F
such that h x = x for every h H, that is,
x is xed by H, in other words, H G
x
. Such an x exists by denition of X
F
and
since H F. We shall see that Y
H
is contractible dening a contracting homotopy
c of Y
H
to the vertex (x). Let = (x
0
, . . . , x
n
) be an arbitrary cell in Y
H
, dene
c() = (x, x
0
, . . . , x
n
).
It is straightforward to see that d
0
c = Id and d
i+1
c = c d
i
for i > 1.
Therefore c indeed denes a contracting homotopy. This shows Y
H
is contractible
and, therefore by Theorem 2.3 Y is a model for E
F
(G), see also [9, Theorem A.2].
Remark 2.5. Let H be a subgroup of G and consider the family F(H) which
consists of all the subgroups of H and their conjugates by elements of G. In this
case we can take X
F(H)
= G/H.
Remark 2.6. When F = 1, the above construction corresponds to the universal
bundle EG of G, i. e., it corresponds to the standard Gresolution of Z.
The Gorbit space of EG is the classical classifying space BG of G. In analogy
with BG, we denote by B
F
(G) the Gorbit space of E
F
(G). Thus when F = 1,
we have that B
{1}
(G) = BG.
6 JOS
AREZ AND JOS
Let H be a subgroup of G. For a Gspace X, let res
G
H
X be the Hspace obtained
by restricting the group action. If F is a family of subgroups of G, let F/H =
L H [ L F be the induced family of subgroups of H.
Proposition 2.7 ([19, Proposition 7.2.4], [9, Proposition A.5]).
res
G
H
E
F
(G) = E
F/H
(H).
3. Relative group homologies
In this section we dene Hochschild and Takasu group homologies and we give
a condition where both homologies coincide.
Since a Gset X is equivalent to a homomorphism from the group G into the
group of permutations of X, in the literature the pair (G, X) is sometimes called a
permutation representation. For any Gset X we construct a complex (C
(X),
)
of abelian groups by letting C
n
(X) be the free abelian group generated by the
ordered (n + 1)tuples of elements of X. Dene the ith face homomorphism
d
i
: C
n
(X) C
n1
(X) by
d
i
(x
0
, . . . , x
n
) = (x
0
, . . . , x
i
, . . . , x
n
),
where x
i
denotes deletion, and the boundary homomorphism
n
: C
n
(X) C
n1
(X)
by
n
=
n
i=0
(1)
i
d
i
.
The standard direct computation shows that
n

n+1
= 0 proving that C
n
(X) is
indeed a complex. Dene C
1
(X) = Z as the innite cyclic group generated by
( ) and dene
0
(x) = ( ) for any x X. Notice that this extended complex is
precisely the augmented complex
C
n
(X) C
2
(X)
2
C
1
(X)
1
C
0
(X)

Z 0.
with =
0
the augmentation homomorphism.
The action of G on X induces an action of G on C
n
(X) with n 0 given by
g (x
0
, . . . , x
n
) = (g x
0
, . . . , g x
n
)
which endows C
n
(X) with the structure of a Gmodule. We also let G act on
C
1
(X) = Z trivially.
For each x X and n 1 dene the map s
x
n
: C
n
(X) C
n+1
(X) given by
(3.1) s
x
n
(x
0
, . . . , x
n
) = (x, x
0
, . . . , x
n
).
We shall use later the following lemma (see [16, Proposition 1.1] and [1, 6])
Lemma 3.1. Let G
x
be the isotropy subgroup of x. Then s
x
n
is a G
x
homomorphism.
Proof. Let g G
x
. Then we have
s
x
n
_
g(x
0
, . . . , g
n
)
_
= s
x
n
(g x
0
, . . . , g x
n
) = (x, g x
0
, . . . , g x
n
)
= (g x, g x
0
, . . . , g x
n
) = g(x, x
0
, . . . , g
n
) = gs
x
n
(x
0
, . . . , g
n
).
Proposition 3.2. The augmented complex (C
(X),
) is acyclic. Hence it is a
Gresolution of the trivial Gmodule Z.
Proof. Let x X and consider the homomorphisms s
x
n
: C
n
(X) C
n+1
(X) given
by (3.1) for n 1. It is straightforward to see that d
0
s
x
n
= Id and d
i+1
s
x
n
=
s
x
n1
d
i
for i > 1. Therefore s
x
denes a contracting homotopy and the augmented

complex is acyclic.
Denote by (B
(X),
id
Z
) the complex given by
(3.2) B
(X) = C
(X)
Z[G]
Z.
As usual, here we see C
(X) as a right Gmodule by dening

(x
0
, . . . , x
n
) g := g
1
(x
0
, . . . , x
n
).
The groups
H
n
(G, X; Z) = H
n
(B
(X)),
are the homology groups of the permutation representation (G, X). The correspond-
ing cohomology groups were dened by Snapper in [16].
Remark 3.3. One can dene the homology groups of the permutation representa-
tion (G, X) with coecients in a Gmodule A by dening B
(X; A) = C
(X)
Z[G]
A and
H
n
(G, X; A) = H
n
(B
(X; A)),
but in the present article we only will use integer coecients.
The following lemma is a small generalization of DupontZickert [8, Lemma 1.3].
Lemma 3.4. Let

C
(X) be a Gsubcomplex of the augmented complex C
(X).
Suppose that for each cycle in

C
n
(X) there exists a point x() X such that
s
x()
n
()

C
n+1
(X), where s
x
n
is given by (3.1). Then

C
(X) is acyclic.
Same proof as [8, Lemma 1.3].
3.1. Group homology. If X is a free Gset, then C
n
(X) is a free Gmodule
and C
(X) is a free Gresolution of the trivial Gmodule Z [16, Remark 1.3]. In

particular, if X is the group G acting on itself by left multiplication, then C
(G) is
the standard free Gresolution of the trivial Gmodule Z (see for instance Brown
[4, I.5]). Hence the complex B
(G) computes the homology of G.

3.2. Hochschild relative group homology. If H is a subgroup of G and X =
G/H then we have that the homology of the complex B
(G/H) gives the Hochschild

relative homology groups
(3.3) H
i
([G : H]; Z) = H
i
(B
(G/H)), n = 0, 1, 2, . . . ,
dened by Hochschild in [10, 4]. We use the notation analogous to the one used by
Adamson in [1] for the corresponding cohomology groups. More generally, let X
(H)
be a Gset, such that the action of G on X is transitive and the set of isotropy
subgroups of points in X
(H)
is the conjugacy class of H in G. Then X
(H)
and
G/H are isomorphic as Gsets and we have that the homology of the transitive
permutation representation (G, X
(H)
) is the Hochschild relative group homology,
that is,
H
n
([G : H]; Z) = H
n
(G, X
(H)
; Z) = H
n
(B
(X
(H)
)), n = 0, 1, 2, . . . .
Notice that when H = 1 we recover the homology groups H
n
(G; Z) of G.
Analogous to group homology, the relative homology groups H
n
([G : H]; Z) can
be computed using relative projective resolutions. What follows summarizes the
8 JOS
AREZ AND JOS
results we need about relative homological algebra given in [10, 1, 2 and 4], for
the particular case of the ring Z[G] and its subring Z[H] for a discrete group G and
a subgroup H. For completeness, we also give proofs of some results which are only
indicated in [10].
An exact sequence of Z[G]homomorphisms between Z[G]modules,
N
i+1
ti+1
N
i
ti
N
i1
is called (Z[G], Z[H])exact if, for each i, the kernel of t
i
is a direct Z[H]module
summand of N
i
.
Proposition 3.5. A sequence N
i+1
ti+1
N
i
ti
N
i1
of Z[G]homomorphisms is
(Z[G], Z[H])exact, if and only if, for each i:
(1) t
i
t
i+1
= 0, and
(2) there exists a contracting Z[H]homotopy, i.e., a sequence of Z[H]homomorphisms
h
i
: M
i
M
i+1
such that t
i+1
h
i
+ h
i1
t
i
is the identity map of M
i
onto itself.
Proof. Let N
i+1
ti+1
N
i
ti
N
i1
be a (Z[G], Z[H])exact sequence of Z[G]
homomorphisms. As usual, we denote Z
i
= ker t
i
and B
i
= imt
i+1
. Since t
i
is (Z[G], Z[H])exact, it is an exact sequence of Z[G]homomorphisms. Hence
t
i
t
i+1
= 0 and Z
i
/B
i
= 0 for every i. Thus B
i
= Z
i
and the short sequences
(3.4) 0 Z
i+1
N
i+1
ti+1
Z
i
0
are exact. Since t
i
is (Z[G], Z[H])exact, we have that N
i+1
= Z
i+1
C
i+1
as
Z[H]modules for some Z[H]module C
i+1
. Hence the sequences (3.4) split and
there exists a contracting Z[H]homotopy (see [4, Proposition (0.3)]).
Conversely, suppose the sequence N
i+1
ti+1
N
i
ti
N
i1
of Z[G]homomorphisms
satises (1) and (2). By (1) we have that imt
i+1
Z
i
and by (2), if x Z
i
we
have that
(3.5) t
i+1
(h(x)) = x.
Thus x imt
i+1
and the sequence t
i
is exact, that is, Z
i
= B
i
. Hence the short
sequences
0 Z
i+1
N
i+1
ti+1
Z
i
0
are exact. Moreover, by (3.5) we also have that h[
Zi
is a section of t
i+1
and
therefore the sequence splits. Since h is an Z[H]homomorphism, we have that
Z
i+1
is a direct Z[H]module summand of N
i+1
. Thus t
i
is (Z[G], Z[H])exact.
A Gmodule A is said to be (Z[G], Z[H])projective if, for every (Z[G], Z[H])
exact sequence 0 U
p
V
q
W 0, and every Ghomomorphism : A W,
there is a Ghomomorphism
: A V such that q
= . This is shown in the

following commutative diagram
A
0

U
p
V
q
W

0.
Lemma 3.6 ([10, Lemma 2]). For every Z[H]module A, the Z[G]module Z[G]
Z[H]
A is (Z[G], Z[H])projective.
If N is any Gmodule, the natural map
: Z[G]
Z[H]
N N
g n gn
(3.6)
gives rise to an exact sequence (see [4, III (3.4)])
(3.7) 0 K
N
Z[G]
Z[H]
N N 0.
The map : N Z[G]
Z[H]
N given by n 1 n is an Hhomomorphism since
(hn) = 1 hn = h n = h(1 n) = h(n), for every h H.
Notice that if h is in Z[G] but not in Z[H] we cannot perform the second step, so
in general is not an Ghomomorphism. Since (n) = (1n) = n, is a section
of and it is an Hisomorphism of N onto a Z[H]module complement of K
N
in Z[G]
Z[H]
N, showing that the exact sequence is actually (Z[G], Z[H])exact.
If N is (Z[G], Z[H])projective, considering the identity map on N, there exists
: N Z[G]
Z[H]
N which makes the following diagram of Ghomomorphisms
commute
N
Id
0

K
N

Z[G]
Z[H]
N

N

0.
It follows that the sequence is (Z[G], Z[G])exact, so that N is Gisomorphic with
a direct Gmodule summand of Z[G]
Z[H]
N.
Proposition 3.7 ([10, page 249]). A direct Gmodule summand of a (Z[G], Z[H])
projective module is also (Z[G], Z[H])projective.
Proof. Let A be a direct Gmodule summand of the (Z[G], Z[H])projective module
N. Let p: N A be the projection and i : A N be the inclusion. Let
0 U V
q
W 0
be an (Z[G], Z[H])exact sequence and let : A W be an Ghomomorphism.
Consider the composition p: N W. Since N is (Z[G], Z[H])projective there
exists a Ghomomorphism : N V such that the following diagram commutes
N
p
0

U

V
q
W

0.
Thus
= i : A V is such that
q
= q i = p i = .
Therefore A is (Z[G], Z[H])projective.
Corollary 3.8 ([10, page 249]). A Gmodule N is (Z[G], Z[H])projective if and
only if it is Gisomorphic with a direct Gmodule summand of Z[G]
Z[H]
N, or if
and only if K
N
is a direct Gmodule summand of Z[G]
Z[H]
N.
10 JOS
AREZ AND JOS
A (Z[G], Z[H])projective resolution of a Z[G]module N is a (Z[G], Z[H])exact
sequence C
1
C
0
N 0 in which each C
i
is (Z[G], Z[H])projective.
Proposition 3.9 ([10, page 250]). Every Z[G]module N has a (Z[G], Z[H])
projective resolution.
Proof. By Lemma 3.6 Z[G]
Z[H]
N is (Z[G], Z[H])projective. Take C
0
= Z[G]
Z[H]
N, with the natural map Z[G]
Z[H]
N N which by the exact sequence (3.7) is
an epimorphism. Then proceed in the same way from the kernel K
N
of this map
in order to obtain C
1
= Z[G]
Z[H]
K
N
, with the map Z[G]
Z[H]
K
N
K
N

Z[G]
Z[H]
N, etc.
The (Z[G], Z[H])projective resolution obtained in the proof of Proposition 3.9
is called in [10, 2] the standard (Z[G], Z[H])projective resolution of N.
Proposition 3.10 ([10, page 260]). The augmented complex C
(G/H) is a (Z[G], Z[H])

projective resolution of Z.
Proof. The Z[G]module C
n
(G/H) as abelian group is generated by n + 1tuples
(g
0
H, . . . , g
n
H) of cosets of H. By Lemma 3.1 the homomorphism
s
H
n
: C
n
(G/H) C
n+1
(G/H)
s
H
n
(g
0
H, . . . , g
n
H) = (H, g
0
H, . . . , g
n
H)
is a Z[H]homomorphism. By the proof of Proposition 3.2 C
(G/H) is exact
and s
H
is a contracting Z[H]homotopy. Hence, by Proposition 3.5 C
(G/H) is
(Z[G], Z[H])exact. There is a Z[G]isomorphism
Z[G]
Z[H]
C
n1
(G/H) C
n
(G/H)
g (g
1
H, . . . , g
n
H) (gH, gg
1
H, . . . , gg
n
H).
Its inverse is given by
C
n
(G/H) Z[G]
Z[H]
C
n1
(G/H)
(g
0
H, . . . , g
n
H) g
0
(g
1
0
g
1
H, . . . , g
1
0
g
n
H).
Another representative of g
0
H is of the form g
0
h with h H and
(g
0
hH, . . . , g
n
H) g
0
h (h
1
g
1
0
g
1
H, . . . , h
1
g
1
0
g
n
H)
g
0
h(h
1
g
1
0
g
1
H, . . . , h
1
g
1
0
g
n
H)
g
0
(g
1
0
g
1
H, . . . , g
1
0
g
n
H),
so it is well dened. Hence by Lemma 3.6 each C
n
(G/H) is (Z[G], Z[H])projective
and therefore C
(G/H) is a (Z[G], Z[H])projective resolution of Z.

Proposition 3.11. Let G be a group and H be a subgroup of G. Consider Z as
a trivial Gmodule and let C
1
C
0
Z 0 be a (Z[G], Z[H])projective
resolution of Z. Then
H
n
([G : H]; Z) = H
n
(C

Z[G]
Z), n = 0, 1, 2, . . . .
Proof. Let A and B be two Gmodules and let C
1
C
0
A 0 be a
(Z[G], Z[H])projective resolution of A. In [10, 2] one of the denitions of the
relative Tor functor is
Tor
(Z[G],Z[H])
n
(A, B) = H
n
(C

Z[G]
B),
and it is proved that this denition is independent of the (Z[G], Z[H])projective
resolution of A. In [10, 4] the relative homology groups for (G, H) are dened by
H
n
([G : H]; Z) = Tor
(Z[G],Z[H])
n
(Z, Z),
and by Proposition 3.10 the complex C
(G/H) is a (Z[G], Z[H])projective resolu-

tion of Z.
Remark 3.12. Let G be a discrete group and let F be a family of subgroups.
By Proposition 2.4 we have that the simplicial chain complex of B
F
(G) is given
by B
(X
F
). In particular, given a group G and a subgroup H of G, consider the
family F(H) generated by H, then by Remark 2.5 we have that X
F(H)
= X
(H)
and
the simplicial chain complex of B
F(H)
(G) is precisely B
(X
(H)
)

= B
(G/H).
Then the following is immediate.
Proposition 3.13. Let G be a discrete group. Let H be a subgroup of G. Consider
the family of subgroups F(H) generated by H. Then
H
n
(B
F(H)
(G); Z)

= H
n
([G : H]; Z), n = 0, 1, 2, . . . ,
where B
F(H)
(G) is the orbit space of the classifying space E
F(H)
(G).
When H = 1 we recover the following well-known fact
Corollary 3.14. Let G be a group. Then
H
n
(BG; Z)

= H
n
(G; Z), n = 0, 1, 2, . . . .
So we can say that the space B
F(H)
(G) is a classifying space for the permutation
representation (G, X
(H)
), compare with Blowers [3] where a classifying space for an
arbitrary permutation representation is constructed for Snappers cohomology [16].
Proposition 3.15. Let K be a normal subgroup of G contained in H. Then we
have an isomorphism of relative homology groups
H
n
([G : H]; Z)

= H
n
([G/K : H/K]; Z), n = 0, 1, 2, . . . .
Proof. We have a bijection of sets
G/H

= (G/K)
_
(H/K)
gH gK(H/K),
which is equivariant with respect to the actions of G on G/H and of G/K on
(G/K)
_
(H/K) via the natural projection G G/K. This bijection induces
an isomorphism between the chain complexes B
(G/H) and B
_
(G/K)
_
(H/K)
_
which commutes with the boundary operators. Hence we get the desired isomor-
phism of homology groups.
Remark 3.16. The statement of Proposition 3.15 with coecients in an arbitrary
Z[G]module A is
H
i
([G : H]; A)

= H
i
([G/K : H/K]; A
K
),
where A
K
is the group of Kcoinvariants of A, compare with Adamson [1, Theo-
rem 3.2].
Corollary 3.17. If H is a normal subgroup of G, then we have an isomorphism
H
n
(G/H; Z)

= H
n
([G : H]; Z), n = 0, 1, 2, . . . .
12 JOS
AREZ AND JOS
3.2.1. Natural Gmaps and induced homomorphism. Let H and K be subgroups
of G. There exists a Gmap G/H G/K if and only if there exists a G such
that a
1
Ha K and is given by
R
a
: G/H G/K,
gH gaK.
Any Gmap G/H G/K is of the form R
a
for some a G such that a
1
Ha K
and R
a
= R
b
only if ab
1
K, see tom Dieck [20, Proposition I(1.14)].
Let H and K be subgroups of G such that H is conjugate to a subgroup of K,
then there is a Gmap
h
K
H
: X
(H)
X
(K)
.
This induces a Ghomomorphism
(h
K
H
)
: C
(X
(H)
) C
(X
(K)
),
which in turn induces a homomorphism of homology groups
(3.8) (h
K
H
)
: H
n
([G : H]; Z) H
n
([G : K]; Z).
Remark 3.18. Let H and K be subgroups of G. Consider the families F(H) and
F(K) generated by H and K respectively and suppose that F(H) F(K). Then
there exists a Gmap unique up to Ghomotopy E
F(H)
(G) E
F(K)
(G). Notice
that F(H) F(K) implies that H is conjugate to a subgroup of K and therefore
there exists a Gmap h
K
H
: X
(H)
= X
F(H)
X
(K)
= X
F(K)
. Using the Simplicial
Construction of Proposition 2.4 we can see the Gmap E
F(H)
(G) E
F(K)
(G) as
the simplicial Gmap given in nsimplices by
(x
0
, . . . , x
n
)
_
h
K
H
(x
0
), . . . , h
K
H
(x
n
)
_
, x
i
X
F(H)
, i = 0, . . . , n.
This map induces a canonical map B
F(H)
(G) B
F(K)
(G) between the correspond-
ing Gorbit spaces. This map in turn induces a canonical homomorphism in ho-
mology
H
n
(B
F(H)
(G); Z) H
n
(B
F(K)
(G); Z),
which by Proposition 3.13 corresponds to the homomorphism (h
K
H
)
in (3.8).
Let C
h
K
H
=
(X
(H)
) be the subcomplex of C
(X
(H)
) generated by tuples mapping
to dierent elements by the homomorphism h
K
H
. We call this subcomplex the h
K
H
subcomplex of C
(X
(H)
). As before, set
B
h
K
H
=
(X
(H)
) = C
h
K
H
=
(X
(H)
)
Z[G]
Z.
Lemma 3.19. Let H and K be subgroups of G such that H is conjugate to a
subgroup of K and suppose K has innite index in G (and therefore also H). Then
the h
K
H
subcomplex C
h
K
H
=
(X
(H)
) is acyclic.
Proof. Since K has innite index in G, given an ncycle =
n
i
(x
i
0
, . . . , x
i
n
)
in C
h
K
H
=
n
(X
(H)
) there exists y() X
(K)
such that y() is dierent from all the
h
K
H
(x
i
j
), j = 0, . . . , n. Let x() h
K
H
1
(y) X
(H)
, then we have that s
x()
n
()
C
h
K
H
=
n+1
(X
(H)
) and by Lemma 3.4 we get the result.
Proposition 3.20. Let H and K be subgroups of G such that H is conjugate to a
subgroup of K and suppose K has innite index in G (and therefore also H). Then
H
n
(B
h
K
H
=
(X
(H)
))

= H
n
([G : H]; Z), n = 0, 1, 2, . . . .
Proof. Since K has innite index in G by Lemma 3.19 we have that C
h
K
H
=
(X
(H)
)
is a resolution of Z.
Since C
(X
(H)
) is (Z[G], Z[H])exact, ker
n
is a direct Hmodule summand of
C
n
(X
(H)
), so we can write
C
n
(X
(H)
) = ker
n
N
for some Hmodule N (actually N = coker
n+1
). Clearly we can also write
C
(X
(H)
) as a direct sum of Gmodules
(3.9) C
n
(X
(H)
) = C
h
K
H
=
n
(X
(H)
) C
h
K
H
=
n
(X
(H)
)
where C
h
K
H
=
n
(X
(H)
) is generated by tuples (x
0
, . . . , x
n
) such that h
K
H
(x
i
) = h
K
H
(x
j
)
for some i ,= j. Restricting the action to H this can be seen as a direct sum of
Hmodules. Hence we can see C
h
K
H
=
(X
(H)
) as a direct sum of Hmodules
C
h
K
H
=
(X
(H)
) =
_
C
h
K
H
=
(X
(H)
) ker
_
C
h
K
H
=
(X
(H)
) N
_
,
and since ker
[
C
h
K
H
=
(X
(H)
)
=
_
C
h
K
H
=
(X
(H)
) ker
_
we have that C
h
K
H
=
(X
(H)
) is
also (Z[G], Z[H])exact.
Finally, by Proposition 3.10 C
n
(X
(H)
) is a (Z[G], Z[H])projective module for
n 0, by (3.9) and Proposition 3.7 we have that C
h
K
H
=
n
(X
(H)
) is also a (Z[G], Z[H])
projective module for n 0. Hence C
h
K
H
=
(X
(H)
) is a (Z[G], Z[H])projective
resolution of Z and therefore B
h
K
H
=
(X
(H)
) computes H
([G : H]; Z).

Remark 3.21. In the case H = K, we can take h
H
H
to be the identity. Hence
C
h
H
H
=
(X
(H)
) is generated by tuples of distinct elements, so we just denote it by
C
=
(X
(H)
). By Proposition 3.20 we have that
H
n
(B
=
(X
(H)
))

= H
n
([G : H]; Z), n = 0, 1, 2, . . . .
3.3. Takasu relative group homology. Recall that the homology of a discrete
group G is equal to the homology of its classifying space BG (Corollary 3.14). Let
H be a subgroup of G, the classifying space BH can be regarded as a subspace of
the classifying space BG. We dene the Takasu relative homology groups denoted
by H
n
(G, H), as the homology of the pair H
n
(BG, BH; Z), or equivalently, let
Cof(i) denote the cobre (mapping cone) of the map BH BG induced by the
inclusion i : H G and take the reduced singular homology groups

H
n
(Cof(i); Z)
for n = 0, 1, 2, . . . . Algebraically this is done as follows (see the Topological Remark
in Weibel [21, p. 19]). Let B
(H) and B
(G) be as in (3.2) and let i
: B
(H)
B
(G) be the chain map induced by the inclusion. The mapping cone of i
is the
chain complex
D
n
(G, H) := B
n1
(H) B
n
(G)
14 JOS
AREZ AND JOS
with boundary map given by the matrix
_
n1 0
i n
_
. Hence the Takasu relative
homology groups are given by
H
n
(G, H; Z) = H
n
(D
(G, H)).
There is a description of these relative homology groups in terms of Gprojective
(in particular Gfree) resolutions: Let H be a subgroup of G. Given a Gmodule
N we consider the epimorphism given in 3.6
: Z[G]
Z[H]
N N
g n gn
and dene the Gmodule
I
(G,H)
(N) = ker .
We have that I
(G,H)
(N) is a covariant exact functor from the category of left G
modules to itself with respect to the variable N, see Takasu [18, Proposition 1.1
(i)].
Consider Z as a trivial Gmodule. For any free Gresolution F
of I
(G,H)
(Z),
we have a canonical isomorphism
(3.10) H
n
(G, H)

= H
n1
(F

Z[G]
Z), n = 1, 2, . . . .
That is, H
n
(G, H) = Tor
Z[G]
n1
(I
(G,H)
(Z), Z), see Takasu [18, Def. 2 (i) & Proposi-
tion 3.2] or Zickert [22, Theorem 2.1].
Remark 3.22. Notice that Z[G]
Z[H]
Z

= Z[G/H] = C
0
(G/H) and with this
isomorphism the epimorphism corresponds to the augmentation : C
0
(G/H)
Z. Therefore
I
(G,H)
(Z) = ker .
The relative homology groups t in the long exact sequence [18, Proposition 2.1]
(3.11) H
i
(H; Z) H
i
(G; Z) H
i
(G, H; Z) H
i1
(H; Z) . . .
which corresponds to the long exact sequence for the pair (BG, BH).
Let H and K be a subgroup of G such that H K G. There exists an induced
homomorphism [18, Proposition 1.2]
I
(G,H)
(Z) I
(G,K)
(Z),
which in turn, by the functoriality of Tor
Z[G]
n1
induces a homomorphism
H
n
(G, H; Z) H
n
(G, K; Z).
3.4. Comparison of Hochschild and Takasu relative group homologies.
Hochschild and Takasu relative group homologies do not coincide in general. The
following examples where suggested to us by Francisco Gonzalez Acu na. Consider
G = Z Z and H = Z. Since Z is normal in Z Z by Corollary 3.17 we have that
H
n
([Z Z : Z]; Z) = H
n
(Z Z/Z; Z) = H
n
(Z; Z) = H
n
(S
1
; Z).
While on the other hand, since B(Z Z) = T
2
, where T
2
is the 2torus and
B(Z) = S
1
,
H
n
(Z Z, Z; Z) = H
n
(B(Z Z), B(Z); Z) =

H
n
(T
2
/S
1
; Z) =

H
n
(S
2
S
1
; Z),
where

H
n
denotes reduced homology.
From the denitions of H
n
(G, H; Z) and H
n
([G : H]; Z) in terms of resolutions
we get a canonical homomorphism between them.
Let F
be any free Gresolution of I

(G,H)
(Z). By Remark 3.22 we have that the
complex
C
3
(G/H)
3
C
2
(G/H)
2
C
1
(G/H)
1
I
(G,H)
(Z),
is a Gresolution of I
(G,H)
(Z). By the Comparison Theorem for Resolutions (see
for instance Weibel [21, Theorem 2.2.6] or Brown [4, Lemma I.7.4]) there is a chain
map F
n
C
n+1
(G/H), n 0, unique up to chain homotopy which by (3.10) and
(3.3) induces a homomorphism
(3.12) H
n
(G, H; Z) H
n
([G : H]; Z), n = 2, 3, . . . .
The following proposition gives a case when both relative homologies agree.
Proposition 3.23. Let H and K be subgroups of G such that H is conjugate to
a subgroup of K and suppose K has innite index in G. Also assume that for any
g / K we have H gHg
1
= I where I G is the identity element. Consider
the h
K
H
subcomplex C
h
K
H
=
(X
(H)
) dened in Subsection 3.2.1. Then C
h
K
H
=
(X
(H)
) is
a free resolution of I
(G,H)
(Z) and therefore we have an isomorphism
H
n
(G, H; Z)

= H
n
(B
h
K
H
=
(X
(H)
)), n = 2, 3, . . . .
Proof. Firstly, we claim that C
h
K
H
=
n
(X
(H)
) is a free Gmodule for n 1. By
Remark 2.1 it is enough to compute the isotropy subgroup of an nsimplex. Without
loss of generality we can consider an nsimplex = (x
0
, . . . , x
n
) C
h
K
H
=
n
(X
(H)
)
such that G
x0
= H since the Gorbit of any nsimplex has an element of this form,
and isotropy subgroups of elements in the same Gorbit are conjugate. We have
that G
xi
= g
i
Hg
1
i
for some g
i
G and g
0
= I. The isotropy subgroup of is
given by
G
=
n
i=0
g
i
Hg
1
i
.
By the denition of C
h
K
H
=
(X
(H)
) we have that g
i
/ K for i = 1, . . . , n and by
hypothesis the intersection of H with any conjugate gHg
1
with g / K is the
identity. Therefore G
= I and C
h
K
H
=
n
(X
(H)
) is a free Gmodule for n 1.
Now, we have that C
h
K
H
=
0
(X
(H)
) = C
0
(X
(H)
) and by Lemma 3.19 the augmented
h
K
h
subcomplex
C
h
K
H
=
2
(X
(H)
)
2
C
h
K
H
=
1
(X
(H)
)
1
C
0
(X
(H)
)

Z 0
is exact, so im
1
= ker . Finally, by Remark 3.22 we have that ker = I
(G,H)
(Z)
and therefore
C
h
K
H
=
2
(X
(H)
)
2
C
h
K
H
=
1
(X
(H)
)
1
I
(G,H)
(Z) 0
is a free Gresolution of I
(G,H)
(Z).
Corollary 3.24. Let H and K be subgroups of G such that H is conjugate to a
subgroup of K. Also assume that for any g / K we have H gHg
1
= I where
I G is the identity element. Then
H
n
(G, H; Z)

= H
n
([G : H]; Z), n = 2, 3, . . . .
16 JOS
AREZ AND JOS
4. The group SL
2
(C) and some of its subgroups
We denote by C
the multiplicative group of the eld of complex numbers. Now

we study the particular case when G = SL
2
(C) and H is one of the following
subgroups:
I =
__
1 0
0 1
__
,
T =
__
a 0
0 a
1
_
a C
_
,
U =
__
1 b
0 1
_
b C
_
,
P =
__
1 b
0 1
_
b C
_
,
B =
__
a b
0 a
1
_
a C
, b C
_
.
By abuse of notation, we denote by I the identity matrix and also the subgroup of
G which consists only of the identity matrix. Denote by

G = G/ I = PSL
2
(C).
Given a subgroup H of G denote by

H the image of H in

G. Notice that

U =

P.
We denote by g the element of

G with representative g G. We shall use this
notation through the rest of the article except in Section 10. As usual we consider
all groups with the discrete topology.
We list some known facts about the these groups. Their proofs are in Lang [11].
SL
2
(C) is generated by elementary matrices [11, Lem. XIII.8.1].
B is a maximal proper subgroup [11, Proposition XIII.8.2].
Bruhat decomposition: [11, XIII 8, p. 538]. Let w =
_
0 1
1 0
_
. There is
a decomposition of SL
2
(C) into disjoint subsets
SL
2
(C) = B BwB.
The subgroups U and P are normal in B and we have the exact sequences
I U
i
B T

= C
I (4.1)
I P
i
B T

= C
I.
PSL
2
(C) is a simple group [11, Theorem XIII.8.4]. Hence, I is the only
normal subgroup of SL
2
(C).
Lemma 4.1. Let g / B. Then
U gUg
1
= I.
P gPg
1
= I.
Proof. Since g / B by Bruhat decomposition we have that g can be written as
g = g
1
wg
2
, g
1
, g
2
B.
Then we have that
gUg
1
= g
1
wg
2
Ug
1
2
w
1
g
1
1
= g
1
wUw
1
g
1
1
.
Let us analyze the elements in g
1
wUw
1
g
1
1
. Consider h = (
1 e
0 1
) U and g
1
=
_
a b
0 a
1
_
B, so a C
. We have that
g
1
whw
1
g
1
1
=
_
1 +a
1
be b
2
e
a
2
e 1 a
1
be
_
.
The only way to have g
1
whw
1
g
1
1
U is to have e = 0 and in that case
g
1
whw
1
g
1
1
= I. Analogously if h =
_
1 e
0 1
_
P then
g
1
whw
1
g
1
1
=
_
1 +a
1
be b
2
e
a
2
e 1 a
1
be
_
,
and the only way to have g
1
whw
1
g
1
1
P is to have e = 0 and in that case
g
1
whw
1
g
1
1
= I.
Now we give models for the Gsets X

(H)
with H = U, P, B.
Remark 4.2. Recall that for H = P, B we have bijections of sets G/H

=

G/

H
which are equivariant with respect to the actions of G on G/H and of

G on

G/

H
via the natural projection G

G. Thus, we have that X
(H)
= X
(

H)
as sets, the
subgroup will indicate whether we are considering the action of G or

G on it.
4.1. The Gset X
(U)
. Consider the action of G on C
2
(0, 0) given by left matrix
multiplication.
Proposition 4.3. The group G acts transitively on C
2
(0, 0).
Proof. Let (x, y) and (z, w) be elements of C
2
(0, 0). Then a matrix g = (
a b
c d
)
SL
2
(C) which send (x, y) to (z, w) is given by:
If x ,= 0: we have two cases:
w ,= 0:
a =
zw +xy
xw
, b =
x
w
, c =
w
x
, d = 0.
That is
g =
_
zw+xy
xw

x
w
w
x
0
_
w = 0: which implies that z ,= 0
a =
z by
x
, c =
y
z
, d =
x
z
, b = any complex number.
That is (simplifying taking b = 0)
g =
_
z
x
0
y
z
x
z
_
If x = 0: which implies y ,= 0, then we have
b =
z
y
, d =
w
y
, a =
_
0 if z ,= 0,
y
w
if w ,= 0.
c =
_
y
z
if z ,= 0,
0 if w ,= 0.
18 JOS
AREZ AND JOS
That is
g =
_
0
z
y
y
z
w
y
_
if z ,= 0,
g =
_
y
w
z
y
0
w
y
_
if w ,= 0.
This proves the transitivity of the action.
Hence we have that
Proposition 4.4. The isotropy subgroup of (1, 0) is U. Therefore, there is a G
isomorphism between SL
2
(C)/U and X
(U)
given by
SL
2
(C)/U X
(U)
gU g
_
1
0
_
.
Proof. Let g = (
a b
c d
) SL
2
(C). Then
_
a b
c d
__
1
0
_
=
_
a
c
_
=
_
1
0
_
implies that a = 1 and c = 0, but since the determinant of g is 1 we have that d = 1
and therefore g = (
1 b
0 1
) U.
Therefore we can set X
(U)
= C
2
(0, 0).
4.2. The Gset X
(P)
. Now consider the action of Z
2
on C
2
(0, 0) such that
1 sends the pair (x, y) to its antipodal point (x, y). Consider the orbit space
C
2
(0, 0)/Z
2
and denote by [x, y] = (x, y), (x, y) the orbit of the pair (x, y).
Let g G and [x, y] C
2
(0, 0)/Z
2
. We dene an action of G on C
2
(0, 0)/Z
2
given by
g
_
x
y
_
=
_
g
_
x
y
__
.
This gives a well-dened action since
_
a b
c d
__
x
y
_
=
_
ax +by
cx +dy
_
_
ax by
cx dy
_
=
_
a b
c d
__
x
y
_
.
Since I acts as the identity, this action descends to an action of

G. By the
denition of the action of G on C
2
(0, 0)/Z
2
and Proposition 4.3 we have
Corollary 4.5. The group G acts transitively on C
2
(0, 0)/Z
2
.
Proposition 4.6. The isotropy subgroup of [1, 0] is P. Therefore, there is a G
isomorphism between SL
2
(C)/P and C
2
(0, 0)/Z
2
given by
SL
2
(C)/P C
2
(0, 0)/Z
2
gP g
_
1
0
_
.
Proof. Let g = (
a b
c d
) SL
2
(C) such that
_
a b
c d
__
1
0
_
=
_
a
c
_
=
_
1
0
_
.
This implies that a = 1 and c = 0, but since det g = 1 we have that d = 1 and
therefore g = (
1 b
0 1
) P.
Therefore we can set X
(P)
= C
2
(0, 0)/Z
2
. By Remark 4.2 we also have that
X
(
P)
= C
2
(0, 0)/Z
2
.
There is another model for X
(P)
which we learned from Ramadas Ramakrish-
nan. Consider the set Sym of 2 2 non-zero symmetric complex matrices with
determinant zero. The set Sym is given by matrices of the form
(4.2) Sym =
__
x
2
xy
xy y
2
_
(x, y) X
(U)
_
.
Let g SL
2
(C) and S Sym. We dene an action of G on Sym by
g S = gSg
T
,
where g
T
is the transpose of g. The action is well-dened because transpose con-
jugation preserves symmetry and the determinant function is a homomorphism.
Since I acts as the identity, this action descends to an action of

G.
Proposition 4.7. The group G acts transitively on Sym.
Proof. Let (
x
2
xy
xy y
2
) and (
z
2
zw
zw w
2
) be elements of Sym. Then the matrix g
= (
a b
c d
)
SL
2
(C) which send (
x
2
xy
xy y
2
) to (
z
2
zw
zw w
2
) is given by:
If x ,= 0: we have two cases:
w ,= 0:
a =
zw +xy
xw
, b =
x
w
, c =
w
x
, d = 0.
That is
g
=
_
zw+xy
xw
x
w
w
x
0
_
If w = 0: which implies that z ,= 0
a =
z
x
, c =
y
z
, d =
x
z
, b = 0
That is
g
=
_
z
x
0
y
z

x
z
_
If x = 0: which implies that y ,= 0, then we have
b =
z
y
, d =
w
y
, a =
_
0 if z ,= 0,
y
w
if w ,= 0.
c =
_
y
z
if z ,= 0,
0 if w ,= 0.
20 JOS
AREZ AND JOS
That is
g
=
_
0
z
y
y
z

w
y
_
if z ,= 0,
g
=
_
y
w

z
y
0
w
y
_
if w ,= 0.
Remark 4.8. Notice that the matrices g
found in the proof of Proposition 4.7 are

the negatives of the corresponding matrices g found in the proof of Proposition 4.3,
that is g
= g. In the proof of Proposition 4.7 one can also take the corresponding
matrices g.
Proposition 4.9. The isotropy subgroup of (
1 0
0 0
) Sym is P. Therefore, there is
a Gisomorphism between SL
2
(C)/P and Sym given by
SL
2
(C)/P Sym
gP g
_
1 0
0 0
_
g
T
.
Proof. Let g = (
a b
c d
) SL
2
(C) such that
_
a b
c d
__
1 0
0 0
__
a c
b d
_
=
_
a
2
ac
ac c
2
_
=
_
1 0
0 0
_
.
This implies that a = 1, c = 0 and d = 1 since det g = 1. Thus g = (
1 b
0 1
) P
as claimed.
We denote

X
(P)
= Sym to distinguish it from X
(P)
. By Remark 4.2 we have
that Sym is also a model for

G/
P. We use the notation

X
(
P)
= Sym to distinguish
it from X
(
P)
and emphasize the action of

G.
Corollary 4.10. The sets X
(P)
and

X
(P)
are isomorphic as Gsets.
Proof. This is immediate from Propositions 4.6 and 4.9. We can give an explicit
isomorphism by using the transitivity of the actions of G on each of these Gsets.
Let [x, y] X
(P)
, then
_
x
y
_
=
_
x z
y w
__
1
0
_
,
where z and w are complex numbers such that xw yz = 1. On the other hand,
we have
_
x
2
xy
xy y
2
_
=
_
x z
y w
__
1 0
0 0
__
x y
z w
_
We dene the isomorphism by
: X
(P)

X
(P)
_
x
y
_
_
x
2
xy
xy y
2
_
.
(4.3)
It is equivariant because
_
a b
c d
__
x
y
_
=
_
ax +by
cx +dy
_
is sent by the isomorphism to
_
(ax +by)
2
(ax +by)(cx +dy)
(ax +by)(cx +dy) (cx +dy)
2
_
=
_
a b
c d
__
x
2
xy
xy y
2
__
a c
b d
_
.
4.3. The Gset X

(B)
. Let

C = C be the Riemann sphere. Let LF(
C) be
the group of fractional linear transformations on

C. Let
: G = SL
2
(C) LF(
C)
g =
_
a b
c d
_
(g)(z) =
az +b
cz +d
,
be the canonical homomorphism. Let g G act on

C by the corresponding
fractional linear transformation (g). It is well-known that this action is tran-
sitive. Abusing of the notation, given g G and z

C we denote the action by
g z = (g)(z).
We can also identify CP
1
with

C via [z
1
: z
2
]
z1
z2
, where [z
1
: z
2
] CP
1
is
written in homogeneous coordinates. In this case an element g = (
a b
c d
) G acts in
an element [z
1
: z
2
] in CP
1
by matrix multiplication
g [z
1
: z
2
] =
_
a b
c d
__
z
1
z
2
_
= [az
1
+bz
2
: cz
1
+dz
2
].
Proposition 4.11. The isotropy subgroup of

C is B. Therefore, there is a
Gisomorphism between SL
2
(C)/B and

C given by
SL
2
(C)/B

C
gB g .
Proof. Let g = (
a b
0 a
1 ) B. Then
g [1 : 0] =
_
a b
0 a
1
__
1
0
_
=
_
a
0
_
= [a : 0] = [1 : 0].
It is easy to see that all the elements in G that x [1 : 0] are of this form, i.e., that
they are elements in B.
Therefore we set X
(B)
=

C. Again, by Remark 4.2 we also have that X
(

B)
=

C.
4.4. The explicit Gmaps. The inclusions
(4.4) I U P B
induce Gmaps
G G/U G/P G/B
g gU gP gB.
Using the models X
(H)
for the Gsets G/H with H = U, P, B given in the previous
subsections we give the explicit Gmaps between them.
22 JOS
AREZ AND JOS
Let g = (
a b
c d
) G. We have the Gmaps
(4.5)
G
h
U
I
X
(U)
h
P
U
X
(P)
h
B
P
X
(B)
g =
_
a b
c d
_

g
_
1
0
_
=
_
a
c
_

g
_
1
0
_
=
_
a
c
_

g [1 : 0] = g =
a
c
.
Notice that h
P
U
: X
(U)
X
(P)
is just the quotient map given by the action of Z
2
.
On the other hand, we have that
(4.6) h
B
U
= h
B
P
h
P
U
,
where h
B
U
is the Hopf map
h
B
U
: X
(U)
X
(B)
h
B
U
(a, c) =
a
c
. (4.7)
Using

X
(P)
instead of X
(P)
we have
(4.8) G
h
U
I
X
(U)
h
P
U

X
(P)
h
B
P
X
(B)
_
a b
c d
_

(a, c)

_
a
2
ac
ac c
2
_

a
2
ac
=
ac
c
2
=
a
c
.
We have that
h
B
P

h
P
U
= h
B
U
.
It is also useful to write the Gmap

h
P
U
:

X
(P)
X
(B)
in terms of the entries of
the matrix in

X
(P)
without writing it in the form given in (4.2). Let (
r t
t s
)

X
(P)
,
that is, (
r t
t s
) ,= (
0 0
0 0
) and rs = t
2
. Then we have
h
P
U
:

X
(P)
X
(B)
_
r t
t s
_
r
t
=
t
s
.
Remark 4.12. For the case of

G we have practically the same

Ghomomorphisms
as in (4.5) and (4.8) except that X
(
U)
= X
(
P)
.
Remark 4.13. Consider X
(B)
and its inverse image under the Hopf map
(4.7)
(h
B
U
)
1
() = (x, 0) [ x C
X
(U)
,
which corresponds to the rst coordinate complex line minus the origin. Since by
Proposition 4.11 the isotropy subgroup of X
(B)
under the action of G is B,
we have that (h
B
U
)
1
() is a B-invariant subset of X
(U)
. Since the short exact
sequence (4.1) splits, any element of B can be written in a unique way as the
product of an element in U and an element in T
(4.9)
_
a b
0 a
1
_
=
_
1 ab
0 1
__
a 0
0 a
1
_
.
It is easy to see that U xes pointwise the points of (h
B
U
)
1
(), while T acts freely
and transitively on (h
B
U
)
1
() where the matrix
_
a b
0 a
1
_
acts multiplying (x, 0) by
a C
.
Another way to interpret this is to write X
(B)
in homogeneous coordinates
[ : 0], then the elements in U x and the homogeneous coordinates [ : 0],
while an element in
_
a b
0 a
1
_
T x but multiply the homogeneous coordinates
by a C
obtaining the homogeneous coordinates [a : 0].

More generally, for any point z X
(B)
, its isotropy subgroup G
z
is a conjugate
of B, which can be written as the direct product of the corresponding conjugates
of U and T, which we denote by U
z
and T
z
. Writing z X
(B)
in homogeneous
coordinates [z : ] the elements of U
z
x z and the homogeneous coordinates, while
the elements of T
z
x z but multiply the homogeneous coordinates by a constant.
Remark 4.14. Analogously, consider X
(B)
and its inverse image under the
Gmap h
B
P
(h
B
P
)
1
() = [x, 0] [ x C
X
(P)
.
By (4.9) any element of P can be written in a unique way as the product of an
element in U and an element in T
(4.10)
_
1 b
0 1
_
=
_
1 b
0 1
__
1 0
0 1
_
.
Given a representative (x, 0) of [x, 0] (h
B
P
)
1
(), the matrix
_
1 0
0 1
_
T changes
the sign of the representative to (x, 0) but xes its class [x, 0] and the matrix
_
1 b
0 1
_
U xes any representative of [x, 0], thus it xes the class itself. Therefore.
the elements in P x pointwise the points in (h
B
P
)
1
() while T acts transitively
on (h
B
P
)
1
() with isotropy Z
2
, where the matrix
_
a b
0 a
1
_
acts multiplying [x, 0]
by a C
obtaining [ax, 0].

If we use instead

h
B
P
the inverse image of X
(B)
is given by
(
h)
B
P
1
() =
__
x 0
0 0
_
x C

X
(P)
.
The elements in P x pointwise the points in (
h
B
P
)
1
() while T acts transitively
on (
h
B
P
)
1
() with isotropy Z
2
, where the matrix
_
a b
0 a
1
_
acts multiplying
_
x 0
0 0
_
by a
2
with a C
obtaining
_
a
2
x 0
0 0
_
.
More generally, for any point z X
(B)
, its isotropy subgroup G
z
is a conju-
gate of B, and let P
z
denote the corresponding conjugate of P. The elements of
P
z
x pointwise the points in (
h
B
P
)
1
(z) while T
z
acts transitively on (
h
B
P
)
1
(z)
multiplying by a constant.
4.5. Canonical homomorphisms. As in Section 2, we denote by F(H) the family
of subgroups of G generated by H. The inclusions (4.4) induce the inclusions of
families of subgroups of G
F(I) F(U) F(P) F(B)
and in turn, these inclusion give canonical Gmaps between classifying spaces
(4.11) EG E
F(U)
(G) E
F(P)
(G) E
F(B)
(G)
which are unique up to Ghomotopy. Taking the quotient by the action of G we
get canonical maps
BG B
F(U)
(G) B
F(P)
(G) B
F(B)
(G).
24 JOS
AREZ AND JOS
The homomorphisms induced in homology give the sequence
H
i
(BG) H
i
(B
F(U)
(G)) H
i
(B
F(P)
(G)) H
i
(B
F(B)
(G)),
which by Proposition 3.13 is the same as the sequence of homomorphisms
H
i
(G)
(h
U
I
)
H
i
([G : U])
(h
P
U
)
H
i
([G : P])
(h
B
P
)
H
i
([G : B]).
Recall that we denote by

G = G/ I = PSL
2
(C) and given a subgroup H of G
we denote by

H the image of the H in

G. Notice that

U =

P. Analogously, the
inclusions

I

P

B induce a sequence of homomorphisms
H
i
(

G) H
i
([

G :

P]) H
i
([

G :

B]).
The relation between the coset sets of G and

G can be shown in the following
diagram
G

G/U

G/P =

G/
P

G/B =

G/

B
G = G/ I

G/
T = G/T.
In turn, by Proposition 3.15 this induces the following commutative diagram of

relative homology groups
(4.12) H
i
(G)

H
i
([G : U])

H
i
([G : P])

H
i
([G : B])
H
i
(

G)

H
i
([G : T]).
4.6. Relative homology of SL

2
(C). The following proposition states that for the
cases of G and U and

G and

P the Hochschild and Takasu relative homology groups
coincide, compare with [8, Remark. 3.6] and [22, 7]).
Consider the h
B
U
subcomplex C
h
B
U
=
(X
(U)
) and the h
P
subcomplex C
h
P
=
(X
(
P)
)
dened in Subsection 3.2.1.
Proposition 4.15. We have isomorphisms
H
n
(G, U; Z)

= H
n
(B
h
B
U
=
(X
(U)
)), n = 2, 3, . . . .
H
n
(

G,

P; Z)

= H
n
(B
h
B=
(X
(
P)
)), n = 2, 3, . . . .
Proof. By Lemma 4.1 U gUg
1
= I for any g / B and

P g

Pg
1
=

I for any
g /

B. Hence, the result follows by Proposition 3.23.
Putting together Propositions 3.20 and 4.15 we get the following corollary.
Corollary 4.16.
H
n
(G, U; Z)

= H
n
([G : U]; Z), n = 2, 3, . . . .
H
n
(

G,

P; Z)

= H
n
([

G :

P]; Z), n = 2, 3, . . . .
5. Invariants for finite volume hyperbolic 3manifolds
In this section we generalize the construction given in Cisneros-MolinaJones [5]
to dene invariants
H
(M) H
3
([SL
2
(C) : H]; Z) of a complete oriented hyperbolic
3manifold of nite volume M, where H is one of the subgroups P or B of SL
2
(C)
dened in Section 4.
5.1. Hyperbolic 3manifolds and the fundamental class of

M. Consider
the upper half space model for the hyperbolic 3space H
3
and identify it with the
set of quaternions z +tj [ z C, t > 0 . Let H
3
= H
3

C be the standard
compactication of H
3
. The group of orientation preserving isometries of H
3
is
isomorphic to PSL
2
(C) and the action of
_
a b
c d
_
PSL
2
(C) in H
3
is given by the
linear fractional transformation
(w) = (aw +b)(cw +d)
1
, w = z +tj, ad bc = 1,
which is the Poincare extension to H
3
of the complex linear fractional transforma-
tion on

C given by
_
a b
c d
_
. Recall that isometries of hyperbolic 3space H
3
can be
of three types: elliptic if xes a point in H
3
; parabolic if xes no point of H
3
and
xes a unique point of

C and hyperbolic if xes no point of H
3
and xes two points
of

C.
A subgroup of SL
2
(C) or PSL
2
(C) is called parabolic if all its elements corre-
spond to parabolic isometries of H
3
xing a common point in

C. Since the action of
SL
2
(C) (or PSL
2
(C)) in

C is transitive and the conjugates of parabolic isometries
are parabolic [15, (4.7.1)] we can assume that the xed point is the point at innity
which we denote by its homogeneous coordinates = [1 : 0] and therefore par-
abolic subgroups are conjugate to a group of matrices of the form
_
1 b
0 1
_
, with
b C, or its image in PSL
2
(C). In other words, a parabolic subgroup of SL
2
(C)
or PSL
2
(C) is conjugate to a subgroup of P or

P respectively.
A complete oriented hyperbolic 3manifold M is the quotient of the hyperbolic
3space H
3
by a discrete, torsion-free subgroup of orientation preserving isome-
tries. Since is torsion-free, it acts freely on H
3
[15, Theorem 8.2.1] and therefore
it consist only of parabolic and hyperbolic isometries. Notice that since H
3
is con-
tractible it is the universal cover of M and therefore
1
(M) = and M is a K(, 1),
ie M = B, the classifying space of . To such an hyperbolic 3manifold we can
associate a geometric representation :
1
(M) = PSL
2
(C) given by the in-
clusion, which is canonical up to equivalence. This representation can be lifted
to a representation : SL
2
(C) [6, Proposition 3.1.1]. There is a one-to-one
correspondence between such lifts and spin structures on M. We identify with
a subgroup of SL
2
(C) using the representation : SL
2
(C) which corresponds
to the spin structure.
Let M be an orientable complete hyperbolic 3manifold of nite volume. Such
manifolds contain a compact 3manifold-with-boundary M
0
such that M M
0
is
the disjoint union of a nite number of cusps. Each cusp of M is dieomorphic to
T
2
(0, ), where T
2
denotes the 2torus, see for instance [15, page 647, Corollary 4
and Theorem 10.2.1]. The number of cusps can be zero, and this case corresponds
when the manifold M is a closed manifold.
Let M be an oriented complete hyperbolic 3manifold of nite volume with d
cusps, with d > 0. Each boundary component T
2
i
of M
0
denes a subgroup
i
of
1
(M) which is well dened up to conjugation. The subgroups
i
are called the
26 JOS
AREZ AND JOS
peripheral subgroups of . The image of
i
under the representation : SL
2
(C)
given by the inclusion is a free abelian group of rank 2 of SL
2
(C). The subgroups
i
are parabolic subgroups of SL
2
(C). Hence we have that
i
F(P). Therefore
the image of
i
under the representation : PSL
2
(C) is contained in F(

P).
Let M = H
3
be a non-compact orientable complete hyperbolic 3manifold of
nite volume. Let : H
3
H
3
= M be the universal cover of M. Consider the
set ( of xed points of parabolic elements of in

C and divide by the action of .
The elements of the resulting set

( are called the cusp points of M.
Remark 5.1. No hyperbolic element in has as xed point any point in (, oth-
erwise the group would not be discrete [15, Theorem 5.5.4].
Let

Y = H
3
( and consider

M =
Y . If M is closed ( = and

M = M, if
M is non-compact we have that

M is the end-compactication of M which is the
result of adding the cusp points of M. We get an extension of the covering map
to a map :

Y

M.
Consider as well the one-point-compactication M
+
of M which consists in iden-
tifying all the cusps points of

M to a single point. Since M is homotopically equiv-
alent to the compact 3manifold-with-boundary M
0
we have that M
+
=

M/
( =
M
0
/M
0
. By the exact sequence of the pair (
M,

() we have that H
3
(
M; Z)

=
H
3
(
M,

(; Z) and therefore we have that
H
3
(
M; Z)

= H
3
(
M,

(; Z)

= H
3
(
M/
(; Z)

= H
3
(M
+
; Z)

= H
3
(M
0
, M
0
; Z)

= Z.
We denote by [
M] the generator and call it the fundamental class of

M.
5.2. Invariants of hyperbolic 3manifolds of nite volume. Let M be a com-
pact oriented hyperbolic 3manifold. To the canonical representation :
1
(M)
PSL
2
(C) corresponds a map B: M BPSL
2
(C) where BPSL
2
(C) is the clas-
sifying space of PSL
2
(C). There is a well-known invariant [M]
PSL
of M in the
group H
3
(PSL
2
(C); Z) given by the image of the fundamental class of M under
the homomorphism induced in homology by B.
As we said before, we generalize the construction given in Cisneros-MolinaJones
[5] to extend this invariant when M is a complete oriented hyperbolic 3manifold
of nite volume (ie M is compact or with cusps) to invariants
H
(M), but in this
case
H
(M) takes values in H
3
([SL
2
(C) : H]; Z), where H is one of the subgroups
P or B of G = SL
2
(C) dened in Section 4.
Let be a discrete torsion-free subgroup of SL
2
(C). The action of on the
hyperbolic 3space H
3
is free and since H
3
is contractible, by Theorem 2.3 it is a
model for E.
The action of on

Y is no longer free. The points in ( have as isotropy subgroups
the peripheral subgroups
1
, . . . ,
d
of or their conjugates and any subgroup in
F(
1
, . . . ,
d
) xes only one point in (. Therefore, by Theorem 2.3 we have that

Y
is a model for E
F(1,...,
d
)
().
We have the following facts:
Since I F(
1
, . . .
d
) there is a map H
3

Y unique up to
homotopy. We can use the inclusion.
By Proposition 2.7 res
G
EG is a model for E. Therefore, there is a

homotopy equivalence H
3
res
G
EG which is unique up to homotopy.

Since F(
1
, . . .
d
) = F(P)/ F(B)/ we have maps
Y res
G
E
F(P)
(G) res
G
E
F(B)
(G)
which are unique up to homotopy.
Remark 5.2. Since F(
1
, . . .
d
) = F(P)/ by Proposition 2.7 we have that the
space res
G
E
F(P)
(G) is a model for E
F(1,...,
d
)
(). Therefore, the map

Y
res
G
E
F(P)
(G) is in fact a homotopy equivalence.
Combining the previous maps with the Gmaps given in (4.11) we have the
following commutative diagram (up to equivariant homotopy)
EG

E
F(P)
(G)

E
F(B)
(G)
H
3
Taking the quotients by SL

2
(C) and we get the following commutative diagram
(5.1) EG

E
F(P)
(G)

E
F(B)
(G)
H
3
BG

B
F(P)
(G)

B
F(B)
(G)
M
f
where f = B: B BG is the map between classifying spaces which on funda-

mental groups induces the representation : SL
2
(C) of M, and

P
and

B
are given by the compositions
P
:

M E
F(P)
(G)/ B
F(P)
(G),
B
:

M E
F(B)
(G)/ B
F(B)
(G),
and they are well-dened up to homotopy.
The maps

P
and

B
induce homomorphisms
(
P
)
: H
3
(
M; Z) H
3
(B
F(P)
(G); Z),
(
B
)
: H
3
(
M; Z) H
3
(B
F(B)
(G); Z).
We denote by
P
(M) and
B
(M) the canonical classes in H
3
(B
F(P)
(G); Z) and
H
3
(B
F(B)
(G); Z) respectively, given by the images of the fundamental class [
M] of
P
(M) = (
P
)
([
M]),
B
(M) = (
B
)
([
M]).
By the commutativity of the lower triangle in (5.1) we have that
P
(M) is sent to
B
(M) by the canonical homomorphism from H
3
(B
F(P)
(G); Z) to H
3
(B
F(B)
(G); Z).
28 JOS
AREZ AND JOS
Thus, by Proposition 3.13 we have the following
Theorem 5.3. Given a complete oriented hyperbolic 3manifold of nite volume
M we have well-dened invariants
P
(M) H
3
([G : P]; Z),
B
(M) H
3
([G : B]; Z).
Moreover, we have that
B
(M) = (h
B
P
)
P
(M)
_
,
where (h
B
P
)
: H
n
([G : P]; Z) H
n
([G : B]; Z) is the homomorphism described in
(3.8).
Proposition 5.4. The invariants
P
(M) and
B
(M) of M only depend on the
canonical representation : PSL
2
(C) and not on the lifting : SL
2
(C).
In other words, they are independent of the choice a spin structure of M.
Proof. Notice that in diagram (5.1) we can replace G by

G. Since by Proposi-
tion 3.15 H
3
([

G :

P]; Z)

= H
3
([G : P]; Z) and H
3
([

G :

B]; Z)

= H
3
([G : B]; Z), by
(4.12) we get the same invariants
P
(M) and
B
(M).
Remark 5.5. The invariants
P
(M) and
B
(M) extend the invariant [M]
PSL
for
M closed in the following sense: when M is compact

M = M, by the commutativity
of the lower diagram in (5.1) (using

G instead of G) and by Remark 3.18 we have
that
(
P
)
= (h
I
)
,
(
B
)
= (h
I
)
,
where (h
I
)
and (h
I
)
are the homomorphisms described in (3.8). Thus
P
(M) = (h
I
)
([M]
PSL
),
B
(M) = (h
I
)
([M]
PSL
).
6. Relation with the extended Bloch group
In the present section we recall the denitions of the Bloch and extended Bloch
groups and the Bloch invariant. We see that the Bloch group is isomorphic to
H
3
([G : B]; Z) and under this isomorphism the Bloch invariant is the invariant
B
(M).
6.1. The Bloch group. The pre-Bloch group T(C) is the abelian group generated
by the formal symbols [z], z C 0, 1 subject to the relation
(6.1) [x] [y] +
_
y
x
_
_
1 x
1
1 y
1
_
+
_
1 x
1 y
_
= 0, x ,= y.
This relation is called the ve term relation. By DupontSah [7, Lemma 5.11] we
also have the following relations in T(C)
(6.2) [x] =
_
1
1 x
_
=
_
1
1
x
_
=
_
1
x
_
=
_
x
x 1
_
= [1 x].
Using this relations it is possible to extend the denition of [x] T(C) allowing
x

C and removing the restriction x ,= y in (6.1). This is equivalent [7, after
Lemma 5.11] to dene T(C) as the abelian group generated by the symbols [z],
z

C subject to the relations
[0] = [1] = [] = 0,
[x] [y] +
_
y
x
_
_
1 x
1
1 y
1
_
+
_
1 x
1 y
_
= 0.
The pre-Bloch group can be interpreted as a Hochschild relative homology group.
The action of G on

C by fractional linear transformations (see 4.3) is not only
transitive but triply transitive, that is, given four distinct points z
0
, z
1
, z
2
, z
3
in
C, there exists an element g PSL

2
(C) such that
g z
0
= 0, g z
1
= , g z
2
= 1, g z
3
= z
where z = [z
0
: z
1
: z
2
: z
3
] is the cross-ratio of z
0
, z
1
, z
2
, z
3
given by
(6.3) [z
0
: z
1
: z
2
: z
3
] =
(z
0
z
3
)(z
1
z
2
)
(z
0
z
2
)(z
1
z
3
)
.
In other words, the orbit of a 4tuple (z
0
, z
1
, z
2
, z
3
) of distinct points in

C under
the diagonal action of G is determined by its cross-ratio.
If we extend the denition of the cross-ratio to [z
0
: z
1
: z
2
: z
3
] = 0 whenever
z
i
= z
j
for some i ,= j, we get a well dened homomorphism
: B
3
(X
(B)
) = B
3
(
C) T(C)
(z
0
, z
1
, z
2
, z
3
)
G
[z
0
: z
1
: z
2
: z
3
]
(6.4)
where (z
0
, z
1
, z
2
, z
3
)
G
denotes the Gorbit of the 3simplex (z
0
, z
1
, z
2
, z
3
) C
3
(X
(B)
).
It is easy to see that the ve term relation (6.1) is equivalent to the relation
4
i=0
(1)
i
[z
0
: : z
i
: : z
4
] = 0.
By the triply transitivity of the action of G on

C we have that B
2
(X
(B)
)

= Z and
B
3
(X
(B)
) consists only of cycles. Thus induces an isomorphism, compare with
[17, Lemma 2.2]
(6.5) H
3
([G : B]; Z) = H
3
(B
(X
(B)
)

= T(C).
Remark 6.1. If we consider the rst denition of the pre-Bloch group where for
the generators [z] of T(C) we only allow z to be in C 0, 1, each generator [z]
corresponds to the Gorbit of a 4tuple (z
0
, z
1
, z
2
, z
3
) of distinct points in

C. In
this case we have that
(6.6) H
3
(B
=
(X
(B)
))

= T(C),
and the isomorphism between (6.5) and (6.6) is given by Remark 3.21. Also using
this denition of the pre-Bloch group it is possible to prove that it is isomorphic to
the corresponding Takasu relative homology group:
Proposition 6.2.
H
3
(G, B; Z) = H
3
([G : B]; Z)

= T(C).
Proof. By [18, Theorem 2.2] we have that H
3
(G, B; Z)

= H
2
(G, I
(G,B)
(Z)) and by
[7, (A27), (A28)] we also have H
2
(G, I
(G,B)
(Z))

= T(C).
30 JOS
AREZ AND JOS
The Bloch group B(C) is the kernel of the map
: T(C)
2
Z
(C
)
[z] z (1 z).
(6.7)
6.2. The Bloch invariant. An ideal simplex is a geodesic 3simplex in H
3
whose
vertices z
0
, z
1
, z
2
, z
3
are all in H
3
=

C. We consider the vertex ordering as part
of the data dening an ideal simplex. By the triply transitivity of the action of G
on H
3
the orientation-preserving congruence class of an ideal simplex with vertices
z
0
, z
1
, z
2
, z
3
is given by the cross-ratio z = [z
0
: z
1
: z
2
: z
3
]. An ideal simplex is
at if and only if the cross-ratio is real, and if it is not at, the orientation given
by the vertex ordering agrees with the orientation inherited from H
3
if and only if
the cross-ratio has positive imaginary part.
From (6.3) we have that an even (ie orientation preserving) permutation of the
z
i
replaces z by one of three so-called cross-ratio parameters,
z, z
=
1
1 z
, z
= 1
1
z
,
while an odd (ie orientation reversing) permutation replaces z by
1
z
,
z
z 1
, 1 z.
Thus, by the relations (6.2) in T(C) we can consider the pre-Bloch group as being
generated by (congruence classes) of oriented ideal simplices.
Let M be a non-compact orientable complete hyperbolic 3manifold of nite
volume. An ideal triangulation for M is a triangulation where all the tetrahedra
are ideal simplices.
Let M be an hyperbolic 3manifold and let
1
, . . . ,
n
be the ideal simplices
of an ideal triangulation of M. Let z
i
C be the parameter of
i
for each i.
These parameters dene an element (M) =
n
i=1
[z
i
] in the pre-Bloch group. The
element (M) T(C) is called the Bloch invariant of M.
Remark 6.3. Neumann and Yang dened the Bloch invariant using degree one
ideal triangulations, in that way it is dened for all hyperbolic 3manifolds of nite
volume, even the compact ones, see NeumannYang [14, 2] for details.
In [14, Theorem 1.1] it is proved that the Bloch invariant lies in the Bloch
group B(C). An alternative proof of this fact is given in Cisneros-MolinaJones [5,
Cor. 8.7].
Remark 6.4. By (6.5) we have that H
3
([G : B]; Z)

= T(C) and in [5, Theo-
rem 6.1] it is proved that
B
(M) is precisely the Bloch invariant (M) of M, see
Subsection 8.1.
6.3. The extended Bloch group. Given a complex number z we use the con-
vention that its argument Arg z always denotes its main argument < Arg z
and Log z always denotes a xed branch of logarithm, for instance, the principal
branch having Arg z as imaginary part.
Let be an ideal simplex with cross-ratio z. A attening of is a triple of
complex numbers of the form
(w
0
, w
1
, w
2
) = (Log z +pi, Log(1 z) +qi, Log(1 z) Log z pi qi)
with p, q Z. The numbers w
0
, w
1
and w
2
are called log parameters of . Up to
multiples of i, the log parameters are logarithms of the cross-ratio parameters.
Remark 6.5. The log parameters uniquely determine z. Hence we can write a
attening as [z; p, q]. Note that this notation depends on the choice of logarithm
branch.
Following [13] we assign cross-ratio parameters and log parameters to the edges
of a attened ideal simplex as indicated in Figure 1.
z
z
z
z
z
0
z
2
z
3
z
1
z
w
2
w
0
w
2
w
0
z
0
z
2
z
3
z
1
w
1
w
1
Figure 1. Cross-ratio and log parameters of a attened ideal simplex
Let z
0
, z
1
, z
2
, z
3
and z
4
be ve distinct points in

C and let
i
denote the ideal
simplices (z
0
, . . . , z
i
, . . . , z
4
). Let (w
i
0
, w
i
1
, w
i
2
) be attenings of the simplices
i
.
Every edge [z
i
z
j
] belongs to exactly three of the
i
and therefore has three associ-
ated log parameters. The attenings are said to satisfy the attening condition if
for each edge the signed sum of the three associated log parameters is zero. The
sign is positive if and only if i is even.
From the denition we have that the attening condition is equivalent to the
following ten equations:
(6.8)
[z
0
z
1
] : w
2
0
w
3
0
+w
4
0
= 0 [z
0
z
2
] : w
1
0
w
3
2
+w
4
2
= 0
[z
1
z
2
] : w
0
0
w
3
1
+w
4
1
= 0 [z
1
z
3
] : w
0
2
+w
2
1
+w
4
2
= 0
[z
2
z
3
] : w
0
1
w
1
1
+w
4
0
= 0 [z
2
z
4
] : w
0
2
w
1
2
w
3
0
= 0
[z
3
z
4
] : w
0
0
w
1
0
+w
2
0
= 0 [z
3
z
0
] : w
1
2
+w
2
2
+w
4
1
= 0
[z
4
z
0
] : w
1
1
+w
2
1
w
3
1
= 0 [z
4
z
1
] : w
0
1
+w
2
2
w
3
2
= 0
The extended pre-Bloch group

T(C) is the free abelian group generated by at-
tened ideal simplices subject to the relations:
Lifted ve term relation:
(6.9)
4
i=0
(1)
i
(w
i
0
, w
i
1
, w
i
2
) = 0,
if the attenings satisfy the attening condition.
Transfer relation:
[z; p, q] + [z; p
, q
] = [z; p, q
] + [z; p
, q].
32 JOS
AREZ AND JOS
The extended Bloch group

B(C) is the kernel of the homomorphism
:

T(C)
2
Z
(C)
(w
0
, w
1
, w
2
) w
0
w
1
.
7. Mappings via configurations in X
(P)
In this section following ideas in Dupont and Zickert [8, 3] we dene a homo-
morphism
H
3
([G : P]; Z)

B(C).
We use the models for X
(U)
, X
(P)
, X
(B)
and the explicit G-maps between them
described in Section 4. We simplify notation by setting
h
U
= h
B
U
: X
(U)
X
(B)
h
P
= h
B
P
: X
(P)
X
(B)
.
Consider the h
H
subcomplexes with H = U, P, B: C
h
U
=
(X
(U)
), C
h
P
=
(X
(P)
) and
C
=
(X
(B)
) dened in Subsection 3.2.1.
Since h
P
U
, h
U
and h
P
are Gequivariant they induce maps
(h
P
U
)
: C
h
U
=
(X
(U)
) C
h
P
=
(X
(P)
),
(h
U
)
: C
h
U
=
(X
(U)
) C
=
(X
(B)
),
(h
P
)
: C
h
P
=
(X
(P)
) C
=
(X
(B)
),
We start dening a homomorphism C
h
U
=
3
(X
(U)
)

T(C) which descends to a
homomorphism
C
h
P
=
3
(X
(P)
)

T(C).
We assign to each 4tuple (v
0
, v
1
, v
2
, v
3
) C
h=
3
(X
(U)
) a combinatorial attening
of the ideal simplex
_
h
U
(v
0
), h
U
(v
1
), h
U
(v
2
), h
U
(v
3
)
_
in such a way that the com-
binatorial attenings assigned to tuples (v
0
, . . . , v
i
, . . . , v
4
) satisfy the attening
condition.
Given v
i
= (v
1
i
, v
2
i
) X
(U)
we denote by
det(v
i
, v
j
) = det
_
v
1
i
v
2
i
v
1
j
v
2
j
_
= v
1
i
v
2
j
v
2
i
v
1
j
.
Let (v
0
, v
1
, v
2
, v
3
) C
h
U
=
3
(X
(U)
). As was noticed in [8, Section 3.1], the cross-ratio
parameters z,
1
1z
and
z1
z
of the simplex
_
h
U
(v
0
), h
U
(v
1
), h
U
(v
2
), h
U
(v
3
)
_
can be
expressed in terms of determinants
z =
_
h
U
(v
0
) : h
U
(v
1
) : h
U
(v
2
) : h
U
(v
3
)
=
_
v
1
0
v
2
0
v
1
3
v
2
3
__
v
1
1
v
2
1
v
1
2
v
2
2
_
_
v
1
0
v
2
0
v
1
2
v
2
2
__
v
1
1
v
2
1
v
1
3
v
2
3
_ =
det(v
0
, v
3
) det(v
1
, v
2
)
det(v
0
, v
2
) det(v
1
, v
3
)
,
(7.1)
1
1 z
=
_
h
U
(v
1
) : h
U
(v
2
) : h
U
(v
0
) : h
U
(v
3
)
=
_
v
1
1
v
2
1
v
1
3
v
2
3
__
v
1
2
v
2
2
v
1
0
v
2
0
_
_
v
1
1
v
2
1
v
1
0
v
2
0
__
v
1
2
v
2
2
v
1
3
v
2
3
_ =
det(v
1
, v
3
) det(v
0
, v
2
)
det(v
0
, v
1
) det(v
2
, v
3
)
,
(7.2)
z 1
z
=
_
h
U
(v
0
) : h
U
(v
2
) : h
U
(v
3
) : h
U
(v
1
)
=
_
v
1
0
v
2
0
v
1
1
v
2
1
__
v
1
2
v
2
2
v
1
3
v
2
3
_
_
v
1
0
v
2
0
v
1
3
v
2
3
__
v
1
2
v
2
2
v
1
1
v
2
1
_ =
det(v
0
, v
1
) det(v
2
, v
3
)
det(v
0
, v
3
) det(v
2
, v
1
)
.
Hence we also have
(7.3)
1 z
z
=
_
h
U
(v
0
) : h
U
(v
2
) : h
U
(v
3
) : h
U
(v
1
)
=
_
v
1
0
v
2
0
v
1
1
v
2
1
__
v
1
2
v
2
2
v
1
3
v
2
3
_
_
v
1
0
v
2
0
v
1
3
v
2
3
__
v
1
1
v
2
1
v
1
2
v
2
2
_ =
det(v
0
, v
1
) det(v
2
, v
3
)
det(v
0
, v
3
) det(v
1
, v
2
)
.
We have that
h
U
(v
i
) ,= h
U
(v
j
)
v
1
i
v
2
i
v
1
j
v
2
j
,= 0
v
1
i
v
2
j
v
2
i
v
1
j
v
2
i
v
2
j
,= 0 det(v
i
, v
j
) ,= 0.
Then all the previous determinants are non-zero.
We dene
(7.4) v
i
, v
j
=
1
2
Log det(v
i
, v
j
)
2
.
Using formulas (7.1), (7.2), and (7.3), we assign a attening to (v
0
, v
1
, v
2
, v
3
)
C
h
U
=
3
(X
(U)
) by setting
(7.5)
w
0
(v
0
, v
1
, v
2
, v
3
) = v
0
, v
3
+ v
1
, v
2
v
0
, v
2
v
1
, v
3
,
w
1
(v
0
, v
1
, v
2
, v
3
) = v
0
, v
2
+ v
1
, v
3
v
0
, v
1
v
2
, , v
3
,
w
2
(v
0
, v
1
, v
2
, v
3
) = v
0
, v
1
+ v
2
, v
3
v
0
, v
3
v
1
, v
2
.
Recall that for w C
we have that
(7.6)
1
2
Log w
2
=
_
_
Log w +i if Arg w (,
2
],
Log w if Arg w (
2
,

2
],
Log w i if Arg w (
2
, ].
By the addition theorem of the logarithm [2, 3.4] and (7.6) we have that:
w
0
= Log z +ik, w
1
= Log
_
1
1 z
_
+il, w
2
= Log
_
1 z
z
_
+im,
for some integers k, l and m. Hence, w
0
, w
1
and w
2
are respectively logarithms
of the cross-ratio parameters z,
1
1z
and
z1
z
up to multiples of i and clearly we
have that w
0
+ w
1
+ w
2
= 0. Therefore ( w
0
, w
1
, w
2
) is a attening in

T(C) of the
ideal simplex
_
h
U
(v
0
), h
U
(v
1
), h
U
(v
2
), h
U
(v
3
)
_
.
34 JOS
AREZ AND JOS
This denes a homomorphism
: C
h
U
=
3
(X
(U)
)

T(C)
(v
0
, v
1
, v
2
, v
3
) ( w
0
, w
1
, w
2
).
Lemma 7.1. Let v
i
, v
j
X
(U)
. Then det(v
i
, v
j
) is invariant under the action of
G on X
(U)
.
Proof. Let
_
a b
c d
_
SL
2
(C) and set
v
i
= ( v
1
i
, v
2
i
) =
_
a b
c d
__
v
1
i
v
2
i
_
= (av
1
i
+bv
2
i
, cv
1
i
+dv
2
i
)
We have that
(7.7)
det( v
i
, v
j
) = det
_
v
1
i
v
2
i
v
1
j
v
2
j
_
= v
1
i
v
2
j
v
2
i
v
1
j
= (av
1
i
+bv
2
i
)(cv
1
j
+dv
2
j
) (av
1
j
+bv
2
j
)(cv
1
i
+dv
2
i
)
= (ad bc)(v
1
i
v
2
j
v
2
i
v
1
j
) since ad bc = 1
= det(v
i
, v
j
).
Hence the homomorphism descends to the quotient by the action of SL

2
(C)
: B
h
U
=
3
(X
(U)
)

T(C).
Now suppose that ( w
0
0
, w
0
1
, w
0
2
),. . . ,( w
4
0
, w
4
1
, w
4
2
) are attenings dened as above
for the simplices
_
h
U
(v
0
), . . . ,

h
U
(v
i
), . . . , h
U
(v
4
)
_
. We must check that these at-
tenings satisfy the attening condition, that is, we have to check that the ten
equations (6.8) are satised. We check the rst equation, the others are similar:
[z
0
z
1
]:
w
2
0
= v
0
, v
4
+ v
1
, v
3
v
0
, v
3
v
1
, v
4
,
w
3
0
= v
0
, v
4
v
1
, v
2
+ v
0
, v
2
+ v
1
, v
4
,
w
4
0
= v
0
, v
3
+ v
1
, v
2
v
0
, v
2
v
1
, v
3
,
Having veried all the ten equations, it now follows from [8, Theorem 2.8] or [13,
Lemma 3.4] that sends boundaries to zero and we obtain a homomorphism
: H
3
(B
h
U
=
(X
(U)
))

T(C).
7.1. The homomorphism descends to C
h
P
=
3
(X
(P)
). Given v
i
= (v
1
i
, v
2
i
)
X
(U)
we denote by v
i
= h
P
U
(v
i
) = [v
1
i
, v
2
i
] its class in X
(P)
.
Remark 7.2. Notice that if v
i
= (v
1
i
, v
2
i
) X
(U)
, then
det(v
i
, v
j
) = det
_
v
1
i
v
2
i
v
1
j
v
2
j
_
= v
1
i
v
2
j
v
2
i
v
1
j
= det(v
i
, v
j
), (7.8)
but in the other hand
det(v
i
, v
j
) = det
_
v
1
i
v
2
i
v
1
j
v
2
j
_
= v
1
i
v
2
j
+v
2
i
v
1
j
= det(v
i
, v
j
). (7.9)
Thus, the quantity det(v
i
, v
j
) is just well-dened up to sign. However, its square
det(v
i
, v
j
)
2
is well-dened.
By Lemma 7.1, (7.8) and (7.9) we have:
Lemma 7.3. Let v
i
, v
j
X
(P)
. Then det(v
i
, v
j
)
2
is invariant under the action of
G on X
(P)
.
So, we dene
(7.10) v
i
, v
j
=
1
2
Log det(v
i
, v
j
)
2
.
Let (v
0
, v
1
, v
2
, v
3
) C
h=
3
(X
(P)
). The homomorphism descends to a well-
dened homomorphism
: C
h
P
=
3
(X
(P)
)

T(C)
(v
0
, v
1
, v
2
, v
3
) ( w
0
, w
1
, w
2
).
by assign to (v
0
, v
1
, v
2
, v
3
) the attening of the ideal simplex
_
h
P
(v
0
), h
P
(v
1
), h
P
(v
2
), h
P
(v
3
)
_
=
_
h
U
(v
0
), h
U
(v
1
), h
U
(v
2
), h
U
(v
3
)
_
given by (7.5) and we obtain a homomorphism
: H
3
(B
h
P
=
(X
(P)
))

T(C).
Proposition 7.4. The image of : H
3
(B
h
P
=
(X
(P)
))

T(C) is in

B(C).
Proof. Dene a map : B
h
P
=
2
(X
(P)
)
2
Z
(C) by
(v
0
, v
1
, v
2
)
G
v
0
, v
1
v
0
, v
2
v
0
, v
1
v
1
, v
2
+ v
0
, v
2
v
1
, v
2
.
Recall that the extended Bloch group

B(C) is the kernel of the homomorphism
:

T(C)
2
Z
(C)
(w
0
, w
1
, w
1
) w
0
w
1
.
A straightforward computation shows that the following diagram
B
h
P
=
3
(X
(P)
)

T(C)

B
h
P
=
2
(X
(P)
)

2
Z
(C)
is commutative. This means that cycles are mapped to

B(C) as desired.
Therefore : H
3
(B
h
P
=
(X
(P)
))

B(C). Then by Proposition 3.20 we have a
homomorphism
(7.11) : H
3
([G : P]; Z)

B(C).
36 JOS
AREZ AND JOS
7.2. Using

X
(P)
. Sometimes is useful to express the homomorphism using

X
(P)
instead of X
(P)
since with

X
(P)
we do not have to worry about equivalence classes
and representatives. Again, we simplify notation by setting
h
P
=

h
B
P
:

X
(P)
X
(B)
.
Let A
i
, A
j

X
(P)
, then we have that
A
i
=
_
r
i
t
i
t
i
s
i
_
, r
i
s
i
= t
2
i
.
Dene
DS(A
i
, A
j
) = r
i
s
j
2t
i
t
j
+r
j
s
i
.
Recall from Corollary 4.10 that the Gisomorphism between X
(P)
and

X
(P)
is
given by
: X
(P)

X
(P)
,
_
u
v
_
_
u
2
uv
uv v
2
_
.
Lemma 7.5. Let v
i
and v
j
in X
(P)
. Then
DS
_
(v
i
), (v
j
)
_
= det(v
i
, v
j
)
2
.
Proof. Let v
i
= [u
i
, v
i
] and v
j
= [u
j
, v
j
] in X
(P)
. We have that
(v
i
) =
_
u
2
i
u
i
v
i
u
i
v
i
v
2
i
_
, (v
j
) =
_
u
2
j
u
j
v
j
u
j
v
j
v
2
j
_
Then
DS
_
(v
i
), (v
j
)
_
= u
2
i
v
2
j
2u
i
v
i
u
j
v
j
+u
2
j
v
2
i
= (u
i
v
j
u
j
v
i
)
2
= det(v
i
, v
j
)
2
.
Combining Corollary 4.10 and Lemma 7.5 we get the following corollary which
can also be proved with a straightforward but tedious computation.
Corollary 7.6. Let A
i
, A
j

X
(P)
and g G. Then DS(A
i
, A
j
) is Ginvariant,
that is,
DS(gA
i
g
T
, gA
j
g
T
) = DS(A
i
, A
j
).
We dene
A
i
, A
j
=
1
2
Log DS(A
i
, A
j
).
So by Lemma 7.5, given v
i
, v
j
X
(P)
we have that (v
i
), (v
j
) = v
i
, v
j
.
Let (A
0
, A
1
, A
2
, A
3
) C
h
P
=
3
(

X
(P)
). Then the homomorphism
: C
h
P
=
3
(

X
(P)
)

T(C)
(A
0
, A
1
, A
2
, A
3
) ( w
0
, w
1
, w
2
),
is given by assigning to (A
0
, A
1
, A
2
, A
3
) the attening dened by
w
0
= A
0
, A
3
+ A
1
, A
2
A
0
, A
2
A
1
, A
3
,
w
1
= A
0
, A
2
+ A
1
, A
3
A
0
, A
1
A
2
, A
3
,
w
2
= A
0
, A
1
+ A
2
, A
3
A
0
, A
3
A
1
, A
2
,
which by Lemma 7.5 is the same as the one given in (7.5) and by Proposition 7.4
we obtain the homomorphism (7.11).
7.3. The fundamental class of M in H
3
(PSL
2
(C); Z). Recall that by Neu-
mann [13, Theorem 2.6] we have that

B(C) is isomorphic to H
3
(PSL
2
(C); Z). Given
a complete oriented hyperbolic 3manifold of nite volume M in Subsection 5.2 we
construct a canonical class
P
(M) H
3
([G : P]; Z). Taking the image of
P
(M)
under the homomorphism : H
3
([G : P]; Z)

B(C) we obtain a canonical class
[M]
PSL
=
_
P
(M)
_

B(C)

= H
3
(PSL
2
(C); Z).
We call this class the PSLfundamental class of M.
Hence we have the following commutative diagram
H
3
([G : U]; Z)

(h
P
U
)
T(C)
H
3
([G : P]; Z)

(h
P
)
H
3
(
M; Z)
(
P
)
B
)
H
3
([G : B]; Z)

T(C)
Remark 7.7. Recall that by Corollary 4.16 and Proposition 3.15 we have the
isomorphism H
3
([G : P]; Z)

= H
3
(

G,

P; Z). In [13, Proposition 14.3] Neumann
proves that the long exact sequence (3.11) gives rise to a split exact sequence
(7.12) 0 H
3
(

G; Z)
i
H
3
(

G,

P; Z)

H
2
(

P; Z) 0.
We will see in Remark 9.11 that the homomorphism denes a splitting of sequence
(7.12), that is, i
= id. Hence, in the case M is closed the lower square of diagram

(5.1) induces in homology the following commutative diagram
H
3
(

G; Z)
i

H
3
([G : P]; Z)

H
3
(M; Z)
f=B
H
3
(
M; Z)
which shows that for M closed both denitions of [M]

PSL
coincide.
8. Computing
P
(M) and
B
(M) using an ideal triangulation of M
Let M be a non-compact orientable complete hyperbolic 3manifold of nite
volume. Let : H
3
H
3
= M be the universal cover of M. Then M lifts to
an exact, convex, fundamental, ideal polyhedron P for [15, Theorem 11.2.1]. An
ideal triangulation of M gives a decomposition of P into a nite number of ideal
tetrahedra (z
i
0
, z
i
1
, z
i
2
, z
i
3
), i = 1, . . . , n. Since T = gP [ g is an exact tessella-
tion of H
3
[15, Theorem 6.7.1], this decomposition of P gives an ideal triangulation
of H
3
.
As in Subsection 5.1 let ( be the set of xed points of parabolic elements of
in H
3
=

C and consider

Y = H
3
(, which is the result of adding the vertices of
the ideal tetrahedra of the ideal triangulation of H
3
. Hence we can consider

Y as
38 JOS
AREZ AND JOS
a simplicial complex with 0simplices given by the elements of (

C. The action
of G on

Y induces an action of G on the tetrahedra of the ideal triangulation of
Y . Taking the quotient of

Y by we obtain

M and we get an extension of the
covering map to a map :

Y

M. The orbits (z
0
, z
1
, z
2
, z
3
)
of the tetrahedra
(z
0
, z
1
, z
2
, z
3
) of the ideal triangulation of

Y correspond to the tethahedra of the
ideal triangulation of

M. The orbit set

( of ( corresponds to the cusps points of
M, we suppose that M has d cups, so the cardinality of

( is d.
8.1. Computation of
B
(M). Using the simplicial construction of E
F
(G) given
by Proposition 2.4 we have that a model for E
F(B)
(G) is the geometric realization
of the simplicial set whose nsimplices are the ordered (n + 1)tuples (z
0
, . . . , z
n
)
of elements of X
F(B)
=

C and the ith face (respectively, degeneracy) of such a
simplex is obtained by omitting (respectively, repeating) z
i
. The action of g G
on an nsimplex (z
0
, . . . , z
n
) gives the simplex (gz
0
, . . . , gz
n
).
Considering

Y as the geometric realization of its ideal triangulation and since
its vertices are elements in (

C we have that the map
B
:

Y E
F(B)
(G) in
diagram (5.1) is given by the (geometric realization of the) equivariant simplicial
map
Y

B
E
F(B)
(G)
(z
0
, z
1
, z
2
, z
3
) (z
0
, z
1
, z
2
, z
3
).
This induces the map

B
:

M B
F(B)
(G) in diagram (5.1). Furthermore, this in-
duces on simplicial 3chains the homomorphism (
B
)
: C
3
(
M) B
3
(B
F(B)
(G)) =
B
3
(X
(B)
) (see Remark 3.12) which we can compose with homomorphism (6.4) to
get
C
3
(
M) = C
3
(
Y )
B
)
B
3
(X
(B)
)

T(C)
(z
0
, z
1
, z
2
, z
3
)
(z
0
, z
1
, z
2
, z
3
)
G
[z
0
: z
1
: z
2
: z
3
]. (8.1)
where (z
0
, z
1
, z
2
, z
3
)
(resp. (z
0
, z
1
, z
2
, z
3
)
G
) denotes the orbit (respectively G
orbit) of the 3simplex (z
0
, z
1
, z
2
, z
3
) in C
3
(
Y ) (respectively in B
3
(X
(B)
)) and [z
0
:
z
1
: z
2
: z
3
] is the cross-ratio parameter of the ideal tetrahedron (z
0
, z
1
, z
2
, z
3
).
Let (z
i
0
, z
i
1
, z
i
2
, z
i
3
)
, i = 1, . . . , n, be the ideal tetrahedra of the ideal triangulation

of

M and let z
i
= [z
i
0
: z
i
1
: z
i
2
: z
i
3
] C be the cross-ratio parameter of the ideal
tetrahedron (z
i
0
, z
i
1
, z
i
2
, z
i
3
) for each i. Then the image of the fundamental class [
M]
under the homomorphism in homology given by (8.1) is given by
H
3
(
M)
(
B
)
H
3
([G : B]; Z)

=
T(C)
[
M] =
n
i=1
(z
i
0
, z
i
1
, z
i
2
, z
i
3
)
i=1
(z
i
0
, z
i
1
, z
i
2
, z
i
3
)
G

n
i=1
[z
i
],
and we have that the invariant
B
(M) under the isomorphism corresponds to the
Bloch invariant (M), see Cisneros-MolinaJones [5, Theorem 6.1].
8.2. Computation of
P
(M). We want to give a simplicial description of the
map
P
:

Y E
F(P)
(G) in diagram (5.1). For this, we also use the Simplicial
Construction of Proposition 2.4 to give a model for E
F(P)
(G) as the geometric
realization of the simplicial set whose nsimplices are the ordered (n+1)tuples of
elements of X
F(P)
= X
(P)
(or X
F(P)
=

X
(P)
) (see Remark 3.12 and Subsection 4.2).
The ith face (respectively, degeneracy) of such a simplex is obtained by omitting
(respectively, repeating) the ith element. The action of g G on an nsimplex is
the diagonal action.
Since the vertices of

Y are elements in (

C, to give a simplicial description of
the map
P
:

Y E
F(P)
(G) is enough to give a map
: ( X
(P)
, or : (

X
(P)
and dene
(8.2)
Y

P
E
F(P)
(G)
(z
0
, z
1
, z
2
, z
3
) ((z
0
), . . . , (z
3
)).
For i = 1, . . . , d, every cusp point c
i

( of M corresponds to a orbit of (.
Choose c
i
( in the orbit corresponding to c
i

(. Now choose an element
v
i
(
h
B
P
)
1
(c
i
) X
(P)
(or A
i
(
h
B
P
)
1
(c
i
)

X
(P)
) and dene
(8.3)
: ( X
(P)
or : (

X
(P)
c
i
v
i
c
i
A
i
and extend equivariantly by
(g c
i
) = g v
i
, or (g c
i
) = gA
i
g
T
.
Remark 8.1. Suppose that for every cusp point c
i

( we have chosen c
i
(

C
in the orbit corresponding to c
i
. Using homogeneous coordinates we can write
c
i
= [z
i
: w
i
].
So, one way to choose v
i
(or A
i
) is given by
v
i
= [z
i
, w
i
], or A
i
=
_
z
2
i
z
i
w
i
z
i
w
i
w
2
i
_
.
The isotropy subgroups of the c
i
are conjugates of the peripheral subgroups
i
and they consist of parabolic elements, that is, elements in a conjugate of P,
and by Remark 4.14 they x pointwise the elements of (
h
B
P
)
1
(c
i
). On the other
hand, by Remark 5.1 no hyperbolic element in has as xed point any point in
(. Therefore is a well-dened equivariant map and the map
P
in (8.2) is
a well-dened equivariant map. Since any two such maps are homotopic,
P
is independent of the choice of the c
i
and the A
i
up to homotopy. This
induces the map

P
:

M B
F(P)
(G) in diagram (5.1) and choosing dierent c
i
and A
i
we obtain homotopic maps. Thus, this induces a canonical homomorphism
in homology (
P
)
: H
3
(
M) H
3
([G : P]; Z) independent of choices.
Let (z
i
0
, z
i
1
, z
i
2
, z
i
3
)
, i = 1, . . . , n, be the ideal tetrahedra of the ideal triangulation

of

M. The image of the fundamental class [
M] under (
P
)
is given by
H
3
(
M)
(
P
)
H
3
([G : P]; Z)
[
M] =
n
i=1
(z
i
0
, z
i
1
, z
i
2
, z
i
3
)

P
(M) =
n
i=1
_
(z
i
0
), (z
i
1
), (z
i
2
), (z
i
3
)
_
G
(8.4)
40 JOS
AREZ AND JOS
obtaining an explicit formula for the invariant
P
(M). Furthermore, applying the
homomorphism (7.11) we get an explicit formula for the PSLfundamental class of
M
(8.5) [M]
PSL
= (
P
(M))
=
n
i=1
_
w0
_
(z
i
0
),(z
i
1
),(z
i
2
),(z
i
3
)
_
, w1
_
(z
i
0
),(z
i
1
),(z
i
2
),(z
i
3
)
_
, w2
_
(z
i
0
),(z
i
1
),(z
i
2
),(z
i
3
)
_
_
.
9. The invariant
P
(M) is Zickerts fundamental class
In [22] Zickert denes a complex

C
(

G,

P) of truncated simplices and proves
that the complex

B
(

G,

P) =

C
(

G,

P)
Z[G]
Z computes the Takasu relative ho-
mology groups H
(

G,

P; Z) [22, Corollary 3.8]. This complex is used to dene a
homomorphism : H
3
(

G,

P; Z)

B(C) [22, Theorem 3.17]. Given an ideal trian-
gulation of an hyperbolic 3manifold and using a developing map of the geometric
representation to give to each ideal simplex a decoration by horospheres, it is de-
ned a fundamental class in the group H
3
(

G,

P; Z) [22, Corollary 5.6]. A priori,
this fundamental class depends on the choice of decoration, but it is proved that
its image under the homomorphism is independent of the choice of decoration
[22, Theorem 6.10]. In fact, in [22] it is considered the more general situation of a
tame 3manifold with a boundary parabolic PSL
2
(C)representation, this will be
considered in the following section.
In this section we compare the results in [22] with our construction of the in-
variant
P
(M). We give an explicit isomorphism between the complexes

C
(

G,

P)
and C
h
P
=
(X
(
P)
); this gives another proof of Corollary 4.16. Under this isomor-
phism, the homomorphisms and in (7.11) coincide. Moreover, we prove that
Zickerts fundamental class in H
3
(

G,

P; Z) coincides with the invariant
P
(M) in
H
3
([

G :

P]; Z) under the aforementioned isomorphism. This proves that Zickert
fundamental class is independent of the choice of decoration. Choosing two dier-
ent decorations, Zickerts construction indeed gives two dierent cycles in

B
3
(

G,

P)
but they are homologous. From this, it is obvious that the image of such cycles
under coincide.
Remark 9.1. Notice the dierence of notation, in [22] G = PSL
2
(C) and P
corresponds to the subgroup of PSL(C) given by the image of the group of upper
triangular matrices with 1 in the diagonal, that is, in our notation to the subgroup
U =

P.
9.1. The complex of truncated simplices. Let be an nsimplex with a ver-
tex ordering given by associating an integer i 0, . . . , n to each vertex. Let
denote the corresponding truncated simplex obtained by chopping o disjoint reg-
ular neighborhoods of the vertices. Each vertex of is naturally associated with
an ordered pair ij of distinct integers. Namely, the ijth vertex of is the vertex
near the ith vertex of and on the edge going to the jth vertex of .
Let be a truncated nsimplex. A

Gvertex labeling g
ij
of assigns to the
vertex ij of an element g
ij

G satisfying the following properties:
(i) For xed i, the labels g
ij
are distinct elements in

G mapping to the same left
Pcoset.
(ii) The elements g
ij
= ( g
ij
)
1
g
ji
are counterdiagonal, that is, of the form
_
0 c
1
c 0
_
.
Let

C
n
(

G,

P), n 1, be the free abelian group generated by

Gvertex labelings of
truncated nsimplices.
Remark 9.2. Since X
(
P)
is

Gisomorphic to the set of left

Pcosets, using the
homomorphism h
I
given in (4.5) (see Remark 4.12) property (i) means that for
xed i we have
h
I
( g
ij
) = [a
i
, c
i
]
for some xed element [a
i
, c
i
] X
(
P)
and for all j ,= i. By the denition of h
I
we
have that for xed i all the g
ij
have the form
g
ij
=
_
a
i
b
ij
c
i
d
ij
_
, with j ,= i and b
ij
,= b
ik
and d
ij
,= d
ik
for j ,= k.
Left multiplication endows

C
n
(

G,

P) with a free Gmodule structure and the
usual boundary map on untruncated simplices induces a boundary map on

C
n
(

G,

P),
making it into a chain complex. Dene
(

G,

P) =

C
(

G,

P)
Z[G]
Z.
Let g
ij
be a

Gvertex labeling of a truncated nsimplex . We dene a

Gedge
labeling of assigning to the oriented edge going from vertex ij to vertex kl the
labeling ( g
ij
)
1
g
kl
. It is easy to see that for any g

G, the

Gvertex labelings
of given by g
ij
and g g
ij
have the same

Gedge labelings. Hence, a

Gedge
labeling represents a generator of

B
(

G,

P). The labeling of an edge going from
vertex i to vertex j in the untruncated simplex is denoted by g
ij
, and the labeling
of the edges near the ith vertex are denoted by
i
jk
. These edges are called the long
edges and the short edges respectively. By properties (i) and (ii) in the denition
of

Gvertex labelings of a truncated simplex, the
i
jk
s are nontrivial elements
in

P and the g
ij
s are counterdiagonal. Moreover, from the denition of

Gedge
labelings we have that the product of edge labeling along any two-face (including
the triangles) is

I.
In [22, Corollary 3.8] it is proved that the complex

B
(

G,

P) computes the groups
H
(

G,

P). For this result, it is not necessary to have property (ii), nor to ask to
have distinct elements in property (i) in the denition of

Gvertex labelings [22,
Remark 3.2]. The reason for asking this extra properties on the

Gvertex labelings
is to be able to assign to each generator a attening of an ideal simplex. We shall
see that this corresponds to our use in Section 7 of the complex C
P
=
(X
(
P)
) rather
than the complex C
(X
(
P)
) to assign to each generator a attening of an ideal
simplex.
In what follows we need a more explicit version of [22, Lemma 3.5].
Lemma 9.3 ([22, Lemma 3.5]). Let g
i

P =
_
a
i
b
i
c
i
d
i
_
P and g
j

P =
_
a
j
b
j
c
j
d
j
_
P be
Pcosets satisfying the condition g

i

B ,= g
j

B. There exists unique coset representa-
tives g
i
x
ij
and g
j
x
ji
satisfying the condition that ( g
i
x
ij
)
1
g
j
x
ji
be counterdiagonal
given by
(9.1) g
i
x
ij
=
_
a
i
aj
aicjajci
c
i
cj
aicjajci
_
, g
j
x
ji
=
_
a
j
ai
ajciaicj
c
j
ci
ajciaicj
_
.
42 JOS
AREZ AND JOS
Proof. We start by reproducing the proof of [22, Lemma 3.5] since it saves compu-
tations. Let g
1
i
g
j
=
_
a b
c d
_
, and let x
ij
=
_
1 pij
0 1
_
and x
ji
=
_
1 pji
0 1
_
. We have
x
1
ij
g
1
i
g
j
x
ji
=
_
a cp
ij
ap
ji
+b p
ij
(cp
ji
+d)
c cp
ji
+d
_
.
Since g
i

B ,= g
j

B, it follows that c is nonzero. This implies that there exists unique
complex numbers p
ij
and p
ji
such that the above matrix is conterdiagonal. They
are given by
(9.2) p
ij
=
a
c
, p
ji
=
d
c
.
Now, using the explicit expressions for g
i
and g
j
we have
(9.3) g
1
i
g
j
=
_
a
j
d
i
b
i
c
j
d
i
b
j
b
i
d
j
a
i
c
j
a
j
c
i
a
i
d
j
c
i
b
j
_
=
_
a b
c d
_
.
Substituting in (9.2) we get
p
ij
=
a
j
d
i
b
i
c
j
a
i
c
j
a
j
c
i
, p
ji
=
a
i
d
j
b
i
c
j
a
i
c
j
a
j
c
i
.
Hence we have
g
i
x
ij
=
_
a
i
b
i
c
i
d
i
_
_
1
ajdibicj
aicjajci
0 1
_
=
_
a
i
aj
aicjajci
c
i
cj
aicjajci
_
,
g
j
x
ji
=
_
a
j
b
j
c
j
d
j
_
_
1
aidjbjci
ajciaicj
0 1
_
=
_
a
j
ai
ajciaicj
c
j
ci
ajciaicj
_
.
Notice that g
i
x
ij
is a well dened element in

G, if we change the signs of a
i
and c
i
the whole matrix changes sign, while if we change the signs of a
j
and c
j
the matrix
does not change. Analogously for g
j
x
ji
.
Remark 9.4. Notice that the expressions for g
i
x
ij
and g
j
x
ji
given in (9.1) only
depend on the classes [a
i
, c
i
] and [a
j
, c
j
] in X
(
P)
, so indeed they only depend on the
left

Pcosets g
i

P and g
j

P. Also notice that by (9.3) the condition g
i

B ,= g
j

B is
equivalent to a
i
c
j
a
j
c
i
,= 0 which is equivalent to h
P
( g
i

P) =
ai
ci
,=
aj
cj
= h
P
( g
j

P).
Corollary 9.5. Let be a truncated nsimplex. A generator of

C
n
(

G,

P), ie a
Gvertex labeling g
ij
of has the form
g
ij
=
_
a
i
aj
aicjajci
c
i
cj
aicjajci
_
, i, j 1, . . . , n, j ,= i, a
i
c
j
a
j
c
i
,= 0,
and the class [a
i
, c
i
] X
(
P)
corresponds to the left

Pcoset associated to the ith
vertex of . Hence, a generator of

B
n
(

G,

P), ie a

Gedge labeling of has the
form

i
jk
=
_
1
a
k
cjajc
k
(aicjajci)(aic
k
a
k
ci)
0 1
_
, i, j, k 1, . . . , n, i ,= j, k, j ,= k, (9.4)
g
ij
=
_
0
1
aicjajci
a
i
c
j
a
j
c
i
0
_
, i, j 1, . . . , n, i ,= j. (9.5)
Proof. Follows immediately from Remark 9.2, Lemma 9.3 and Remark 9.4.
Corollary 9.6. There is a

Gisomorphism of of chain complexes
C
h
P
=
n
(X
(
P)
)

C
n
(

G,

P)
_
[a
0
, c
0
], . . . , [a
n
, c
n
]
_
g
ij
=
_
a
i
aj
aicjajci
c
i
cj
aicjajci
_
.
Hence, there is an isomorphism of chain complexes B
h
P
=
n
(X
(
P)
)

=

B
n
(

G,

P) where
the

Gorbit
_
[a
0
, c
0
], . . . , [a
n
, c
n
]
_
G
corresponds to the

Gedge labeling given by (9.4)
and (9.5).
Proof. This is a rened version of [22, Corollary 3.6] and follows from Corollary 9.5.
By direct computation it is easy to see that the isomorphism is

Gequivariant.
The only thing that remains to prove is that the isomorphism commutes with the
boundary maps of the complexes, which is an easy exercise.
As we mention above, the isomorphism given in Corollary 9.6 induces an iso-
morphism H
([

G :

P]; Z)

= H
(

G,

P; Z) obtaining another proof of Corollary 4.16.
Remark 9.7. From Corollary 9.6, to represent a generator of C
h
P
=
n
(X
(
P)
) we just
need 2(n + 1) complex numbers, while to represent a generator of

C
n
(

G,

P) we
need 2(n+1)
2
because there is a lot of redundant information in g
ij
, the entries b
ij
and d
ij
in g
ij
, see Remarks 9.2 and 9.4. So it is more ecient to use the complex
C
h
P
=
(X
(
P)
) than the complex

C
(

G,

P) to compute H
([

G :

P]; Z)

= H
(

G,

P; Z).
Another advantage is that by Proposition 3.15 we can work with the action of
G rather than with the action of

G, as we did in Section 7.
Remark 9.8. If we denote by v
i
= [v
1
i
, v
2
i
] an element in X
(
P)
as in Subsection 7.1,
we have that the isomorphism of Corollary 9.6 is written as
(v
0
, . . . , v
n
) g
ij
=
_
_
v
1
i
v
1
j
det(vi,vj)
v
2
i
v
2
j
det(vi,vj)
_
_
,
where (v
0
, . . . , v
n
) is an (n+1)tuple of elements of X
(
P)
such that det(v
i
, v
j
)
2
,= 0
for i ,= j, see Remark 9.4. We also have that in this notation the

Gedge labeling
given by (9.4) and (9.5) is written as

i
jk
=
_
1
det(v
k
,vj)
det(vi,vj) det(vi,v
k
)
0 1
_
, i, j, k 1, . . . , n, i ,= j, k, j ,= k,
g
ij
=
_
0
1
det(vi,vj)
det(v
i
, v
j
) 0
_
, i, j 1, . . . , n, i ,= j. (9.6)
Notice that although det(v
i
, v
j
) is only well-dened up to sign, see Remark 7.2,
we get well-dened elements in

G. The fact that the matrices (9.4) and (9.5) of
the

Gedge labeling are constant under the action of

G is because det(v
i
, v
j
) is
invariant (up to sign) under the action of

G, see Lemma 7.1.
44 JOS
AREZ AND JOS
9.2. Decorated ideal simplices and attenings. Also in [22] it is proved that
there is a one-to-one correspondence between generators of

B
3
(

G,

P) and congru-
ence classes of decorated ideal simplices.
Remember that the subgroup

P xes H
3
and acts by translations on any
horosphere at . A horosphere at is endowed with the counterclockwise ori-
entation as viewed from . Since

G acts transitively on horospheres, we get an
orientation on all horospheres.
A horosphere together with a choice of orientation-preserving isometry to C is
called an Euclidean horosphere [22, Denition 3.9]. Two horospheres based at the
same point are considered equal if the isometries dier by a translation. Denote by
H() the horosphere at at height 1 over the bounding complex plane C, with the
Euclidean structure induced by projection. We let

G act on Euclidean horospheres
in the obvious way, this action is transitive and the isotropy subgroup of H() is
P. Hence the set of Euclidean horospheres can be identied with the set

G/
P of
left

Pcosets, which is

Gisomorphic to X
(
P)
, where an explicit Gisomorphism is
given by
(9.7)
Euclidean horospheres X
(
P)
H() [1, 0],
and extending equivariantly using the action of

G.
A choice of Euclidean horosphere at each vertex of an ideal simplex is called a
decoration of the simplex. Having xed a decoration, we say that the ideal simplex
is decorated. Two decorated ideal simplices are called congruent if they dier by an
element of

G.
Using the identication of Euclidean horospheres with left

Pcosets, we can see
a decorated ideal simplex as an ideal simplex with a choice of a left

Pcoset for
each vertex of the ideal simplex.
Proposition 9.9. Generators in C
h
P
=
3
(X
(
P)
) are in one-to-one correspondence
with decorated simplices. Thus, generators of B
h
P
=
3
(X
(
P)
) are in one-to-one corre-
spondence with congruence classes of decorated simplices.
Proof. Consider the homomorphism (h
P
)
: C
h
P
=
3
(X
(
P)
) C
=
3
(X
(

B)
) and con-
sider a generator (v
0
, . . . , v
3
) of C
h
P
=
3
(X
(
P)
). Its image
_
(h
P
)
(v
0
), . . . , (h
P
)
(v
3
)
_
is a 4tuple of distinct points in X
(

B)
, so it determines a unique ideal simplex in H
3
.
Moreover, v
i
represents a left

Pcoset which corresponds to the vertex (h
P
)
(v
i
)
of such ideal simplex. Hence (v
0
, . . . , v
3
) represents a decorated simplex.
This together with the isomorphism given in Corollary 9.6 proves [22, Theo-
rem 3.13] that there is a one-to-one correspondence between generators of

B
3
(

G,

P)
and congruence classes of decorated ideal simplices, see [22, Remark 3.14].
For a matrix g =
_
a b
c d
_
, let c(g) denote the entry c. Let be a generator of
B(

G,

P). By (9.6) we have that c( g
ij
) = det(v
i
, v
j
), that is, it is only well-dened
up to sign. But we have that
(9.8) c( g
ij
)
2
= det(v
i
, v
j
)
2
is a well-dened non-zero complex number. Squaring formulas (7.1), (7.2) and (7.3)
and using (9.8) we get
c( g
03
)
2
c( g
12
)
2
c( g
02
)
2
c( g
13
)
2
= z
2
,
c( g
13
)
2
c( g
02
)
2
c( g
01
)
2
c( g
23
)
2
=
_
1
1 z
_
2
,
c( g
01
)
2
c( g
23
)
2
c( g
03
)
2
c( g
12
)
2
=
_
1 z
z
_
2
,
which are the formulas of [22, Lemma 3.15]. Now, our choice of logarithm branch
denes a square root of c( g
ij
)
2
, see [22, Remark 3.4], given by
Log c( g
ij
) =
1
2
Log c( g
ij
)
2
=
1
2
Log det(v
i
, v
j
)
2
,
which is the denition of v
i
, v
j
given in (7.10).
Proposition 9.10. The following diagram commutes
H
3
([

G :

P]; Z)

B(C)
H
3
(

G,

P)
where is the homomorphism given in [22, Theorem 3.17], is the homomorphism

given in (7.11) and the vertical arrow is given by the isomorphism of Corollary 9.6.
Proof. The denition of given by formula [22, (3.6)] coincides with the denition
of given by (7.5) via the isomorphism of Corollary 9.6.
Remark 9.11. In [22, Proposition 6.12] Zickert proves that denes a splitting
of the sequence (7.12). This together with Proposition 9.10 proves the claim made
in Remark 7.7 that denes a splitting of the sequence (7.12).
9.3. The fundamental class. Now we compare the construction of Zickers fun-
damental class in H
3
(

G,

P; Z) with our computation of the invariant
P
(M) given
in Subsection 8.2.
As in Section 8, consider an hyperbolic 3manifold of nite volume M and let
:
1
(M) PSL
2
(C) be the geometric representation. Let :

Y

M be the
extension of the universal cover of M to its end-compactication. A developing
map of is a equivariant map
D:

Y H
3
sending the points in ( to H
3
. Let c

( and for each lift c ( of c, let H(D(c)) be
an Euclidean horosphere based at D(c). The collection H(D(c))
c
1
(c)
of Eu-
clidean horospheres is called a decoration of c if the following equivariance condition
is satised:
H(D( c)) = ()H(D(c)), for
1
(M), c
1
(c).
A developing map of together with a choice of decoration of each c

( is called
a decoration of .
By [22, Corollary 5.16] a decoration of denes a fundamental class F(M)
in H
3
(

G,

P; Z). This can be seen as follows. The decoration of endows each
3simplex of M with the shape of a decorated simplex. By [22, Theorem 3.13]
each congruence class of these decorated simplices corresponds to a generator of
B
3
(

G,

P) which is a truncated simplex with a

Gedge labeling. The decoration and
46 JOS
AREZ AND JOS
the

Gedge labelings respect the face pairings so this gives a well-dened cycle
in

B
3
(

G,

P), see [22, p. 518] for details.
Theorem 9.12. The fundamental class F(M) corresponds to the invariant
P
(M)
under the isomorphisms given in Corollary 9.6.
Proof. In Subsection 8.2 the inclusion

Y H
3
is a developing map of the geo-
metric representation :
1
(M) PSL
2
(C). Using the bijection between the set
of horospheres and

X
(
P)
given in (9.7) the map given in (8.3) corresponds to
a decoration of . So not only endows each 3simplex of M with the shape of
a decorated simplex, but also gives the equivariant map
P
in (8.2). This in-
duces the map

P
:

M B
F(P)
(G). Composing the homomorphism on complexes
induced by

P
with the isomorphism of complexes of Corollary 9.6 we get
C
3
(
M)
(
P
)
B
h
P
=
3
(X
(
P)
)

=

B
3
(

G,

P).
The image of the fundamental cycle [
M] of

M gives the cycle (8.4) in B
h
P
=
3
(X
(
P)
)
which represents
P
(M) and this corresponds to a well-dened cycle in

B
3
(

G,

P)
which represents the fundamental class F(M) in H
3
(

G,

P; Z).
Corollary 9.13. The fundamental class F(M) H
3
(

G,

P; Z) does not depend on
the choice of decoration of .
Remark 9.14. Choosing a dierent decoration of , that is, a dierent equivariant
map
in (8.3), we get a dierent equivariant map
P
, a dierent homomor-
phism of complexes (
P
)
and a dierent cycle
in

B
3
(

G,

P); but by the universal
property of E
F(P)
(G) we have that
P
and
P
are -homotopic and therefore the
cycles and
in

B
3
(

G,

P) are homologous. Consider the natural homomorphism
: H
3
(

G,

P; Z) H
2
(

P; Z) of sequence (7.12), the images of and
under this
homomorphism give dierent but homologous cycles in H
2
(

P; Z), compare with
[22, Remark 5.19]. Thus we have that the image
_
F(M)
_
H
2
(

P; Z) is also an
invariant of M.
From Corollary 9.13 now it is obvious that the image of the fundamental class
F(M) H
3
(

G,

P; Z) under the homomorphism : H
3
(

G,

P; Z)

B(C) is inde-
pendent of the choice of decoration [22, Theorem 6.10] and by Theorem 9.12 and
Proposition 9.10 it coincides with the PSLfundamental class dened in Subsec-
tion 7.3.
Remark 9.15. Notice that the image of the PSLfundamental class [M]
PSL
un-
der the homomorphism (h
I
)
: H
3
(PSL
2
(C); Z) H
3
([PSL
2
(C) :

T]; Z) in di-
agram (4.12) is also an invariant of M which we can denote by
T
(M). This
invariant is sent to the classical Bloch invariant
B
(M) by the homomorphism
h
T
: H
3
([PSL
2
(C) :

T]; Z) H
3
([PSL
2
(C) :

B]; Z). It would be interesting to see
which information carries this invariant and if it is possible to obtain it directly
from the geometric representation of M.
10. (G, H)representations
Our construction also works in the more general context of (G, H)representations
of tame manifolds considered in [22]. Here we give the basic denitions and facts,
for more detail see [22, 4].
A tame manifold is a manifold M dieomorphic to the interior of a compact
manifold M. The boundary components E
i
of M are called the ends of M. The
number of ends can be zero to include closed manifolds as tame manifolds with no
ends.
Let M be a tame manifold. We have that
1
(M)

=
1
(M) and each end E
i
of M denes a subgroup
1
(E
i
) of
1
(M) which is well dened up to conjugation.
These subgroups are called peripheral subgroups of M.
Let

M be the compactication of M obtained by identifying each end of M to
a point. We call the points in

M corresponding to the ends as ideal points of M .
Let

M be the compactication of the universal cover

M of M obtained by adding
ideal points corresponding to the lifts of the ideal points of M . The covering map
extends to a map from

M to

M. We choose a point in M as a base point of

M
and one of its lifts as base point of

M. With the base points xed, the action of
1
(M) on

M by covering transformations extends to an action on

M which is not
longer free. The stabilizer of a lift e of an ideal point e corresponding to and end
E
i
is isomorphic to a peripheral subgroup
1
(E
i
). Changing the lift e corresponds
to changing the peripheral subgroup by conjugation.
Let G be a discrete group, let H be any subgroup and consider the family of
subgroups F(H) generated by H. Let M be a tame manifold, a representation
:
1
(M) G is called a (G, H)representation if the images of the peripheral
subgroups under are in F(H).
In the particular case when G = PSL
2
(C) and H =

P a (G, H)representation
:
1
(M) PSL
2
(C) is called boundary-parabolic.
The geometric representation of a hyperbolic 3manifold is boundary parabolic.
For further examples see Zickert [22, 4].
Let M be a tame nmanifold with d ends and let :
1
(M) G be a (G, H)
representation. Let be the image of
1
(M) in G under , also denote by
i
the image of the peripheral subgroup
1
(E
i
) under and consider the family F =
F(
1
, . . . ,
d
) of subgroups of G. On the other hand, dene
i
=
1
(
i
) and
consider the family F
= F
1
, . . . ,
d
) of subgroups of
1
(M).
Proposition 10.1. Consider the classifying space E
F
() as a
1
(M)space dening
the action by
x = () x,
1
(M), x E
F
().
Then, with this action E
F
() is a model for the classifying space E
F
(
1
(M)).
Proof. Consider the set
F
dened in Subsection 2.1. It is enough to see that
F
seen as a
1
(M)set using is
1
(M)isomorphic to
F
. By the denition of
and
i
we have that
/ ker

= and
i
/ ker

=
i
. Then
= (
/ ker )
_
(
i
/ ker )

= /
i
.
Therefore
F
=
d
i=1
=
d
i=1
/
i
=
F
.
So now we can use
F
in the simplicial construction of E
F
(
1
(M)) and we obtain
precisely E
F
().
48 JOS
AREZ AND JOS
Since the action of
1
(M) on

M has as isotropy subgroups the peripheral sub-

groups
1
(E
i
) and
1
(E
i
) F
there is a
1
(M)map unique up to
1
(M)homotopy
(10.1)
F
:

M E
F
(
1
(M))

= E
F
().
Now consider the classifying space E
F(H)
(G), by Proposition 2.7, restricting the
action of G to we have that res
G
E
F(H)
(G)

= E
F(H)/
(). On the other hand, we
have that F F(H)/, so we have a map unique up to homotopy
(10.2)
F
: E
F
() E
F(H)
(G).
Composing (10.1) with (10.2) we obtain a equivariant map unique up to
homotopy
(10.3)
M E
F(H)
(G).
Taking the quotients by the actions of
1
(M) and G we get a map unique up to
homotopy given by the composition
:

M E
F(H)
(G)/ B
F(H)
(G).
Denote by
H
() the image of the fundamental class [
M] of

M under the map
induced in homology by

(10.4)
H
n
(
M; Z)
(
)
H
n
(B
F(H)
(G); Z)
[
M]
H
().
Thus, by Proposition 3.13 we have:
Theorem 10.2. Given an oriented tame nmanifold with a (G, H)representation
:
1
(M) G we have a well-dened invariant
H
() H
n
([G : H]; Z).
As before, one can compute the class
H
() using a triangulation of M. A trian-
gulation of a tame manifold M is an identication of

M with a complex obtained
by gluing together simplices with simplicial attaching maps. A triangulation of M
always exists and it lifts uniquely to a triangulation of

M.
Let M be a tame nmanifold with d ends and let :
1
(M) G be a (G, H)
representation. In [22, 5.2], given a triangulation of M it is constructed a (G, H)
cocycle, see [22, 5.2] for the denition, which denes a fundamental class F() in
H
n
(G, H; Z). The construction of the (G, H)cocycle depends on a decoration of
by conjugation elements. Such decorations are given as follows: for each ideal point
e
i

M choose a lifting e
i

M and assign to this lifting an element g

i
( e
i
) G, or
rather an Hcoset g
i
( e
i
)H, then extend equivariantly by
g
i
( e
i
) = ()g
i
( e
i
),
1
(M).
Let J denote the set of ideal point in

M. Notice that a decoration by conjugation

elements is equivalent to give a equivariant map
: J G/H.
The map
denes explicitly the map (10.3) and using the triangulation of M

gives also explicitly the homomorphism (10.4) as in Subsection 8.2.
Remark 10.3. For general G and H we do not necessarily have that H
n
(G, H; Z)
coincides with H
n
([G : H]; Z), see Subsection 3.4. The construction of the (G, H)
cocycle a priori depends on the choice of decoration of by conjugation elements,
so in principle, choosing dierent decorations one can obtain dierent fundamental
classes in H
n
(G, H; Z), in that case, all this fundamental classes are mapped to
H
() H
n
([G : H]; Z) under the canonical homomorphism (3.12) since
H
()
does not depend on the choice of decoration because the map (10.3) given by
the decoration is unique up to homotopy. So in this general context it is more
appropriate to use Hochschild relative group homology than Takasu relative group
homology because we obtain invariants independent of choice.
10.1. Boundary-parabolic representations. In the case of boundary-parabolic
representations of tame 3manifolds we can use a developing map with a decoration
to compute
P
(). Let M be a tame 3manifold and let be a boundary-parabolic
representation. A developing map of is a equivariant map D
M H
3
sending
the ideal points of

M to H
3
=

C. Taking a suciently ne triangulation of M it is
always possible to construct a developing map of [22, Theorem 4.11]. Let ( be the
image under D
of the set of ideal points J of
M. A decoration of is a equivariant
map
: ( X
(
P)
which can be obtained assigning a Euclidean horosphere to
each element of ( as in Subsection 9.3 or as in Remark 8.1. Again, the decoration
denes explicitly the map (10.3) which gives explicitly the homomorphism (10.4)
with G = PSL
2
(C) and H =

P. Since by Corollary 4.16 H
3
(PSL
2
(C),

P; Z)

=
H
3
([PSL
2
(C) :

P]; Z) in this case F() =
P
() is independent of the choice of
decoration.
As in the case of the geometric representation we can dene the PSLfundamental
class of by
[]
PSL
= (
P
())

T(C)

= H
3
(PSL
2
(C); Z).
Remark 10.4. As in Remarks 9.14, the image
_
F()
_
H
2
(

P; Z) under the
homomorphism of sequence (7.12) is also an invariant of .
The image of
P
() under (h
P
)
: H
3
([PSL
2
(C) :

P; Z) H
3
([PSL
2
(C) :
B]; Z)

= T(C) gives an invariant
B
() in the Bloch group B(C). The invariant
B
() can be computed using a developing map as in Subsection 8.1.
Also as in Remarks 9.15 the image of the PSLfundamental class []
PSL
under
the homomorphism (h
I
)
: H
3
(PSL
2
(C); Z) H
3
([PSL
2
(C) :

T]; Z) in diagram
(4.12) is also an invariant
T
() of .
11. Complex volume
Recall that Rogers dilogarithm is given by
L(z) =
_
z
0
Log(1 t)
t
dt +
1
2
Log(z) Log(1 z).
In [13, Proposition 2.5] Neumann denes the homomorphism
L:

T(C) C/
2
Z,
[z; p, q] L(z) +
i
2
_
q Log z +p Log(1 z)
_

2
6
(11.1)
50 JOS
AREZ AND JOS
where [z; p, q] denotes elements in the extended pre-Bloch group using our choice of
logarithm branch, see Remark 6.5, but (11.1) is actually independent of this choice,
see [22, Remark 1.9]. In [13, Theorem 2.6 or Theorem 12.1] Neumann proves that
under the isomorphism H
3
(PSL
2
(C); Z
=

B(C) the homomorphism

L corresponds
to the CheegerChernSimons class.
Since by Proposition 9.10 the homomorphism : H
3
([PSL
2
(C) :

P]; Z)

B(C)
corresponds to the map : H
3
(PSL
2
(C),

P; Z)

B(C) we have that for the geo-
metric representation of a complete oriented hyperbolic 3manifold of nite volume
M we have
L
_
P
(M)
_
= i
_
Vol(M) +i CS(M)
_
C/i
2
Z,
where Vol(M) is the volume of M and CS(M) = 2
2
cs(M) R/
2
Z with cs(M)
the ChernSimons invariant of M. Usually Vol(M) +i CS(M) is called the complex
volume of M, see Neumann [13] and Zickert [22] for details.
For the case of a boundary-parabolic representation :
1
(M) PSL
2
(C) of a
tame 3manifold M, following Zickert [22, 6] we dene the complex volume of the
representation by
i
_
Vol() +i CS()
_
=

L
_
(M)
_
.
Applying

L to a formula for []
PSL
analogous to (8.5) in Subsection 8.2, but using
the decoration
: ( X
(
P)
of of Subsection 10.1, we obtain an explicit formula
for the complex volume of .
Acknowledgments. Both authors would like to thank Ramadas Ramakrishnan
for many fruitful discussions and suggestions about this work and for his hospitality
during the authors visits to the Abdus Salam International Centre for Theoretical
Physics, Trieste, Italy. Both authors also thank Francisco Gonzales Acu na for his
interest in this work and many illustrative talks.
The second author thanks Walter Neumann for many useful discussions and for
his kind hospitality during his visit to Columbia University as part of his sabbatical
year. He also thanks Christian Zickert for the discussions about his work and his
interest in this article. He also would like to thank Wolfgang L uck and Haydee
Aguilar for the interesting conversations related to this work.
The rst author was partially supported by CONACYT-Becas-Mixtas en el
extranjero para becarios CONACyT nacionales 2011-2012, Mexico. The second
author is a Regular Associate of the Abdus Salam ICTP. He was partially supported
by CONACYT and UNAM-DGAPA-PASPA, Mexico during his sabbatical visits.
References
[1] Iain T. Adamson. Cohomology theory for non-normal subgroups and non-normal elds. Proc.
Glasgow Math. Assoc., 2:6676, 1954.
[2] Lars V. Ahlfors. Complex Analysis. McGraw-Hill, 1979.
[3] James V. Blowers. The classifying space of a permutation representation. Trans. Amer. Math.
Soc., 227:345355, 1977.
[4] Kenneth S. Brown. Cohomology of groups, volume 87 of Graduate Texts in Mathematics.
Springer-Verlag, New York, 1994. Corrected reprint of the 1982 original.
[5] Jose Luis Cisneros-Molina and John D. S. Jones. The Bloch invariant as a characteristic class
in B(SL
2
(C), T). Homology Homotopy Appl., 5(1):325344, 2003.
[6] Marc Culler and Peter B. Shalen. Varieties of group representations and splittings of 3-
manifolds. Ann. of Math. (2), 117(1):109146, 1983.
[7] Johan L. Dupont and Chih-Han Sah. Scissors congruences. II. J. Pure Appl. Algebra,
25(2):159195, 1982.
[8] Johan L. Dupont and Christian K. Zickert. A dilogarithmic formula for the Cheeger-Chern-
Simons class. Geom. Topol., 10:13471372 (electronic), 2006.
[9] F. T. Farrell and L. E. Jones. Isomorphism conjectures in algebraic K-theory. J. Amer. Math.
Soc., 6(2):249297, 1993.
[10] G. Hochschild. Relative homological algebra. Trans. Amer. Math. Soc., 82:246269, 1956.
[11] Serge Lang. Algebra, volume 211 of Graduate Texts in Mathematics. Springer-Verlag, New
York, third edition, 2002.
[12] Wolfgang L uck. Survey on classifying spaces for families of subgroups. In Innite groups:
geometric, combinatorial and dynamical aspects, volume 248 of Progr. Math., pages 269322.
Birkhauser, Basel, 2005.
[13] Walter D. Neumann. Extended Bloch group and the Cheeger-Chern-Simons class. Geom.
Topol., 8:413474 (electronic), 2004.
[14] Walter D. Neumann and Jun Yang. Bloch invariants of hyperbolic 3-manifolds. Duke Math.
J., 96(1):2959, 1999.
[15] John G. Ratclie. Foundations of hyperbolic manifolds, volume 149 of Graduate Texts in
Mathematics. Springer, New York, second edition, 2006.
[16] Ernst Snapper. Cohomology of permutation representations. I. Spectral sequences. J. Math.
Mech., 13:133161, 1964.
[17] A. A. Suslin. K
3
of a Field and the Bloch Group. Proceedings of the Steklov Institute of
Mathematics, 4:217239, 1991.
[18] Satoru Takasu. Relative homology and relative cohomology theory of groups. J. Fac. Sci.
Univ. Tokyo. Sect. I, 8:75110, 1959.
[19] Tammo tom Dieck. Transformation Groups and Representation Theory. Lecture Notes in
Mathematics 766. Springer-Verlag, Berlin-Heidelberg, 1979.
[20] Tammo tom Dieck. Transformation Groups. Number 8 in Studies in Mathematics. Walter de
Gruyter, Berlin, 1987.
[21] Charles A. Weibel. An Introduction to Homological Algebra. Cambridge University Press,
1994.
[22] Christian K. Zickert. The volume and Chern-Simons invariant of a representation. Duke Math.
J., 150(3):489532, 2009.
Instituto de Matem aticas, Unidad Cuernavaca, Universidad Nacional Autonoma de
M exico, Avenida Universidad sin n umero, Colonia Lomas de Chamilpa, Cuernavaca, More-
los, M exico
E-mail address: fenix@matcuer.unam.mx
Instituto de Matem aticas, Unidad Cuernavaca, Universidad Nacional Autonoma de
M exico, Avenida Universidad sin n umero, Colonia Lomas de Chamilpa, Cuernavaca, More-
los, M exico
E-mail address: jlcm@matcuer.unam.mx

Arciniega

Cargado por

Información del documento

Descripción original:

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

Arciniega

Cargado por

Copyright:

Formatos disponibles

INVARIANTS OF HYPERBOLIC 3-MANIFOLDS IN RELATIVE

AREZ AND JOS

AREZ AND JOS

AREZ AND JOS

AREZ AND JOS

Proposition 3.2. The augmented complex (C

denes a contracting homotopy and the augmented

(X) as a right Gmodule by dening

(X) be a Gsubcomplex of the augmented complex C

(X) is a free Gresolution of the trivial Gmodule Z [16, Remark 1.3]. In

(G) computes the homology of G.

(G/H) gives the Hochschild

AREZ AND JOS

= . This is shown in the

AREZ AND JOS

(G/H) is a (Z[G], Z[H])

is a contracting Z[H]homotopy. Hence, by Proposition 3.5 C

(G/H) is a (Z[G], Z[H])projective resolution of Z.

(G/H) is a (Z[G], Z[H])projective resolu-

AREZ AND JOS

([G : H]; Z).

(G) be as in (3.2) and let i

AREZ AND JOS

be any free Gresolution of I

AREZ AND JOS

the multiplicative group of the eld of complex numbers. Now

Now we give models for the Gsets X

AREZ AND JOS

AREZ AND JOS

Remark 4.8. Notice that the matrices g

found in the proof of Proposition 4.7 are

P. We use the notation

4.3. The Gset X

AREZ AND JOS

obtaining the homogeneous coordinates [a : 0].

obtaining [ax, 0].

AREZ AND JOS

In turn, by Proposition 3.15 this induces the following commutative diagram of

4.6. Relative homology of SL

AREZ AND JOS

M] the generator and call it the fundamental class of

EG is a model for E. Therefore, there is a

EG which is unique up to homotopy.

Taking the quotients by SL

where f = B: B BG is the map between classifying spaces which on funda-

AREZ AND JOS

are the homomorphisms described in (3.8). Thus

C, there exists an element g PSL

AREZ AND JOS

AREZ AND JOS

AREZ AND JOS

Hence the homomorphism descends to the quotient by the action of SL

AREZ AND JOS

= id. Hence, in the case M is closed the lower square of diagram

which shows that for M closed both denitions of [M]

AREZ AND JOS

Y . Taking the quotient of

, i = 1, . . . , n, be the ideal tetrahedra of the ideal triangulation

, i = 1, . . . , n, be the ideal tetrahedra of the ideal triangulation

AREZ AND JOS

Pcosets satisfying the condition g

AREZ AND JOS

AREZ AND JOS