### November 27, 2006

<ph f="cmbx">Periodicity and Jumps in Cohomology of R-Torsion-Free Groups</ph>

### Nansen Petrosyan

Abstract
• A discrete group $G$  has periodic cohomology over $R$  if there is an element $\alpha \in$  Ext ${}_{RG}^{*}\left(R,R\right)$  of positive degree and an integer $n\ge 0$  such that the cup product map by $\alpha$  induces an isomorphism of Ext ${}_{RG}^{i}\left(R,-\right)$  and Ext ${}_{RG}^{i+|\alpha |}\left(R,-\right)$  for $i\ge n$  . Adem and Smith [1] showed if $R=\mathbb{Z}$  , then this condition is equivalent to the existence of a finite dimensional free- $G$  -CW-complex homotopy equivalent to a sphere. It was conjectured by Olympia Talelli in [14], that if $G$  is also torsion-free then it must have finite cohomological dimension. In this paper we investigate the implied condition of jump cohomology over $R$  : $G$  has jump cohomology of height $k$  over $R$  if any subgroup $H$  with finite cohomological dimension over $R$  , has $c{d}_{R}\left(H\right)\le k$  . We prove that all solvable and $H\mathcal{ℱ}$  groups with jump cohomology over $R$  must have finite cohomological dimension over $R$  if they are $R$  -torsion-free.
1. Introduction
Definition 1.1. A discrete group $G$  has periodic cohomology over $R$  if there is a cohomology class $\alpha \in$  Ext ${}_{RG}^{*}\left(R,R\right)$  with $|\alpha |>0$  and an integer $n\ge 0$  such that the cup product map
$\alpha \cup -:$  Ext ${}_{RG}^{i}\left(R,M\right)\to$  Ext ${}_{RG}^{i+|\alpha |}\left(R,M\right)$
is an isomorphism for every $RG$  -module $M$  and every integer $i\ge n$  .
If $R=\mathbb{Z}$  , then we simply say that $G$  has periodic cohomology. In this case A. Adem and J. Smith [1] proved that this condition is equivalent to the existence of a finite dimensional free- $G$  -CW-complex homotopy equivalent to a sphere. If a group $G$  has periodic cohomology then it has periodic cohomology over any commutative ring $R$  with a unit. This definition of periodicity using the cup product differs from the general definiton of periodic cohomology of a group. Namely, a group $G$  has periodic cohomology after $k$  -steps if there exists an integer $q\ge 0$  such that the functors ${H}^{i}\left(G,-\right)$  and ${H}^{i+q}\left(G,-\right)$  are naturally equivalent for all $i>k$  . We will call this classical periodic cohomology. It is immediate that the cup product notion of periodic cohomology implies the classical periodic cohomology. It is a conjecture by O. Talelli that the two notions are the same. She has proved that the conjecture holds in the case of the $H\mathcal{ℱ}$  -groups introduced by P. Kropholler in [8]. This is the smallest class of groups which contains all finite groups and which contains all groups $G$  , whenever $G$  acts cellularly on a finite dimensional contractible CW-complex with all isotropy subgroups already in $H\mathcal{ℱ}$  .
This is a large class of groups. Among others, it contains all countable linear groups, all countable solvable groups and all groups with finite virtual cohomological dimension.
The following was conjectured by O. Talelli in [14], Conjecture III: Conjecture (O. Talelli). Every torsion-free discrete group $G$  with classical periodic cohomology after some steps has finite cohomological dimension.
The Conjecture was shown to be true for the class of $H\mathcal{ℱ}$  -groups by Mislin and Talelli in [9]. In our paper we investigate a weaker condition, implied by cohomological periodicity, for a discrete group $G$  .
Definition 1.2 A discrete group $G$  has jump cohomology over $R$  if there exists an integer $k\ge 0$  , such that if $H$  is any subgroup of $G$  with $c{d}_{R}\left(H\right)<\infty$  , then $c{d}_{R}\left(H\right)\le k$  . The bound $k$  will be called the jump height over R. If $R=\mathbb{Z}$  then we say $G$  has jump cohomology and a jump height k.
By naturality of the cup product it follows that if a group $G$  has periodic cohomology over $R$  , then $G$  has jump cohomology over $R$  . In section 2 we prove the following proposition which relates the geometric aspects of the notion to the jump cohomology.
Proposition 1.3. Let $G$  be a discrete group acting freely and properly on an $e$  dimensional CW-complex $E$  homotopy equivalent to ${S}^{n}$  and let $k=e-n$  . If $H$  is any subgroup of $G$  with $cd\left(H\right)<\infty$  ( $hd\left(H\right)<\infty$  ), then $cd\left(H\right)\le k$  (respectively $hd\left(H\right)\le k$  ).
In section 3 we investigate the classes of $H\mathcal{ℱ}$  -groups and solvable groups without $R$  -torsion. We prove the following two theorems.
Theorem 1.4. Let $G$  be a discrete group without $R$  -torsion with jump cohomology of height $k$  over $R$  . If $G$  is an $H\mathcal{ℱ}$  -group, then $c{d}_{R}\left(G\right)\le k$  .
In particular, any $H\mathcal{ℱ}$  -group $G$  has jump cohomology of height $k$  over $\mathbb{Q}$  if and only if $c{d}_{\mathbb{Q}}\left(G\right)\le k$  .
Theorem 1.5. Let $G$  be an $R$  -torsion-free solvable group with jump cohomology of height $k$  over $R$  , where $R$  is an integral domain. Then $G$  has finite cohomological dimension over $R$  and $G$  has finite Hirsch rank. These theorems give affirmative answers for the case of $H\mathcal{ℱ}$  -groups and solvable groups to the conjecture we believe is a natural generalization of the conjecture by Talelli. Conjecture. For every discrete group $G$  without $R$  -torsion the following are equivalent.
• (1) $G$  has jump cohomology of height $k$  over $R$  .
• (2) $G$  has periodic cohomology over $R$  starting in dimension $k+1$  .
• (3) $c{d}_{R}\left(G\right)\le k$
This conjecture in particular would imply Talelli's Conjecture. It also says that every groups with periodic cohomology over $\mathbb{Q}$  must have $c{d}_{\mathbb{Q}}\left(G\right)<\infty$  . Periodic or jump cohomology is a rather strong condition for the cohomology of $G$  . In 3.7 we give the example of the Thompson group $F=〈{x}_{0},{x}_{1},{x}_{2},...|{x}_{i}^{-1}{x}_{n}{x}_{i}={x}_{n+1}$  for all $i  and $n\in \mathbb{N}〉$  , which is of type FP ${}_{\infty }$  . It has periodic cohomology with integral coefficients and vanishing cohomology with $\mathbb{Z}F$  -module coefficients, but it does not have jump cohomology over any ring $R$  , since it has an infinite rank abelian subgroup.
Next we apply Theorem 1.4 to obtain the following results, which may be of separate interest to the reader.
Corollary 1.6. If $G$  has a non-countable, torsion-free, solvable subgroup, then G does not have periodic cohomology over $R$  , where $R$  is an integral domain. In particular, $G$  cannot act freely and properly on a finite dimensional complex homotopy equivalent to a sphere.
Corollary 1.7. If $G$  is a torsion-free solvable group which acts freely and properly on an $\left(n+1\right)$  -dimensional complex $E$  homotopy equivalent to ${S}^{n}$  , then $G$  must be infinite cyclic or trivial.
Lastly, in section 4, we find an obstruction for a countable group acting on certain complexes to be free. We apply our Theorem 1.8 to a class of groups with periodic cohomology starting in dimension 2, obtaining Corollary 1.9. The statement of the Corollary also follows from [13], 3.4.
Theorem 1.8. Let $G$  be a finitely generated torsion-free group that acts freely and properly on an $n$  -dimensional CW-complex $X$  such that the $G$  -invariant submodule $\left({H}_{n-1}\left(X,R\right){\right)}^{G}$  is not trivial. Then $G$  is a free group if and only if Ext ${}_{RG}^{2}\left(R,{H}_{n}\left(X,R\right)\right)=0$  . Corollary 1.9. If $G$  is a torsion-free countable group that acts freely and properly on an $\left(n+1\right)$  -dimensional complex homotopy equivalent to ${S}^{n}$  , then $G$  must be free.
2. Periodicity and Boundedness
Unless otherwise specified, $R$  will denote any commutative ring with a unit.
A group $G$  is said to be $R$  -torsion-free if for every element $g\in G$  such that ${g}^{n}=1$  , where $n$  is a positive integer, $n\cdot 1$  is an invertible element in $R$  .
Definition 2.1. Cohomological dimension of a discrete group $G$  over $R$  , denoted $c{d}_{R}\left(G\right)$  , is defined as
inf $\left\{n:$  Ext ${}_{RG}^{i}\left(R,-\right)=0$  for $i>n\right\}$
Virtual cohomological dimension of $G$  over $R$  , denoted $vc{d}_{R}\left(G\right)$  , is the cohomological dimension of an $R$  -torsion-free subgroup of $G$  of finite index, if such a group exists.
Definition 2.2. A discrete group $G$  has periodic cohomology over $R$  if there is a cohomology class $\alpha \in$  Ext ${}_{RG}^{*}\left(R,R\right)$  with $|\alpha |>0$  and an integer $n\ge 0$  such that the cup product map
$\alpha \cup -:$  Ext ${}_{RG}^{i}\left(R,M\right)\to$  Ext ${}_{RG}^{i+|\alpha |}\left(R,M\right)$
is an isomorphism for every $RG$  -module $M$  and every integer $i\ge n$  . The smallest degree of such $\alpha$  will be the $\mathit{p}\mathit{e}\mathit{r}\mathit{i}\mathit{o}\mathit{d}$  of the periodic cohomology of $G$  , denoted $c{p}_{R}\left(G\right)$  . If $R=\mathbb{Z}$  , we say that $G$  has periodic cohomology with period $cp\left(G\right)$  .
The above notion of periodicity was used by Adem and Smith in [1]. They prove the generalized Wall's conjecture [9] assuming that the periodicity is induced by the cup product with a cohomology class in ${H}^{*}\left(G,\mathbb{Z}\right)$  .
Theorem 2.3 (Adem-Smith). A discrete group $G$  has periodic cohomology if and only if $G$  acts freely and properly on a finite dimensional complex homotopy equivalent to a sphere.
Let us investigate a weaker cohomological condition, which arises naturally, when a group $G$  has a periodic cohomology over a ring $R$  .
Definition 2.4. A discrete group $G$  has jump cohomology over $R$  if there exists an integer $k\ge 0$  , such that if $H$  is any subgroup of $G$  with $c{d}_{R}\left(H\right)<\infty$  , then $c{d}_{R}\left(H\right)\le k$  . The bound $k$  will be called the jump height over R. If $R=\mathbb{Z}$  then we simply say $G$  has jump cohomology and a jump height k. This next proposition gives a more geometric meaning to the jump height when periodicity is assumed over the ring of integers.
Proposition 2.5. Let $G$  be a discrete group acting freely and properly on an $e$  dimensional CW-complex $E$  homotopy equivalent to ${S}^{n}$  and let $k=e-n$  . If $H$  is any subgroup of $G$  with $cd\left(H\right)<\infty$  ( $hd\left(H\right)<\infty$  ), then $cd\left(H\right)\le k$  (respectively $hd\left(H\right)\le k$  ).
• Proof. We will show the proof of the proposition concerning cohomology.
The proof of the homological version is similar.
Suppose $H$  is a subgroup of $G$  with finite cohomological dimension $m$  over $R$  . We can assume that the induced action of $H$  on ${H}^{*}\left(E,\mathbb{Z}\right)$  is trivial, otherwise we can pass to an index two subgroup of $H$  and use the Serre Theorem, [10]. There exists a free $\mathbb{Z}H$  -module $F$  such that ${H}^{m}\left(H,F\right)\ne 0$  .
We get the following oriented spherical fibration:
$E⟶E{×}_{H}EH\stackrel{\pi }{⟶}BH$
where $E{×}_{H}EH$  is the Borel Construction and it is homotopy equivalent to $E/H$  . The Gysin exact sequence of cohomology associated to this fibration shows:
${H}^{m+n}\left(H,F\right)\stackrel{{\pi }^{*}}{⟶}{H}^{m+n}\left(E/H,F\right)⟶{H}^{m}\left(H,F\right)⟶{H}^{m+n+1}\left(H,F\right)$
Since ${H}^{m+n}\left(H,F\right)={H}^{m+n+1}\left(H,F\right)=0$  the sequence shows that ${H}^{m+n}\left(E/H,F\right)\sim ={H}^{m}\left(H,F\right)$  . So ${H}^{m+n}\left(E/H,F\right)\ne 0$  . This implies $m+n\le e$  and $m\le k$  .
Remark. If there exist a finite dimensional free- $G$  -CW-complex $E$  homotopy equivalent to a sphere ${S}^{n}$  , then $G$  has a periodic cohomology over any ring $R$  . For we can always pass to an index two subgroup of $G$  , if necessary, so that we have a trivial action on ${H}^{*}\left(E,R\right)\sim =Hom\left({H}_{*}\left(E,\mathbb{Z}\right),R\right)$  . Then, as in the proof of the proposition, the Gysin exact sequence of cohomology over $R$  shows that the cup product with the Euler class of the spherical fibration induces an isomorphism in cohomology of $G$  after dimension $k$  , where $k$  is the jump height over $R$  .
This, together with Theorem 2.3, yields that, if a group $G$  has periodic cohomology, then it has periodic cohomology over any ring $R$  .
Now, suppose $G$  has periodic cohomology over $R$  starting in dimension $k+1$  . If $H$  is any subgroup of $G$  with finite cohomological dimension $h$  over $R$  , then by Shapiro's Lemma we have,
Ext ${}_{RH}^{k+1}\left(R,M\right)\sim =$  Ext ${}_{RG}^{k+1}\left(R,$  Coind ${}_{H}^{G}M\right)\stackrel{{\alpha }^{h}\cup -}{⟶}$  Ext ${}_{RG}^{k+h|\alpha |+1}\left(R,$  Coind ${}_{H}^{G}M\right)\sim =$  Ext ${}_{RH}^{k+h|\alpha |+1}\left(R,M\right)=0$
for any $RH$  -module $M$  . This establishes the following lemma.
Lemma 2.6. If a group $G$  has periodic cohomology over $R$  starting in dimension $k+1$  , then $G$  has a jump cohomology of height $k$  over $R$  .
Therefore, among the three cohomological conditions stated in the conjecture, jump cohomology is the weakest. We use the Serre Theorem to generalize these conditions to the virtual notions.
Definition 2.7. We will say that a group $G$  has virtual periodic cohomology over $R$  if it contains a finite index subgroup with periodic cohomology over $R$  . The period of the virtual periodic cohomology of $G$  is $vc{p}_{R}\left(G\right)=$  inf $\left\{$  period of $H|$  $|G:H|<\infty ,$  $H$  has periodic cohomology over R $\right\}$  .
Note that if $G$  has periodic cohomology over $R$  then $vc{p}_{R}\left(G\right)\le c{p}_{R}\left(G\right)$  . For example, if $G$  is a finite group then $vc{p}_{R}\left(G\right)=0$  for any ring $R$  . With these definitions, Proposition 2.4 can be generalized.
Corollary 2.8. Let $G$  be a discrete group with virtual periodic cohomology over $R$  . There exists an integer $k\ge 0$  , such that if $H$  is any subgroup of $G$  with $vc{d}_{R}\left(H\right)<\infty$  , then $vc{d}_{R}\left(H\right)\le k$  .
• Proof. Let ${G}^{\prime }$  be a finite index subgroup of $G$  with periodic cohomology over $R$  . Let $H$  be a subgroup of $G$  with $vc{d}_{R}\left(H\right)<\infty$  . Let ${H}^{\prime }$  be a finite index subgroup of $H$  with $c{d}_{R}\left({H}^{\prime }\right)<\infty$  . $|H:{H}^{\prime }\cap {G}^{\prime }|<\infty$  and $c{d}_{R}\left({H}^{\prime }\right)\le k$  , where $k$  is the jump height of ${G}^{\prime }$  over $R$  . Let $L={H}^{\prime }\cap {G}^{\prime }$  . Then, by the Serre Theorem $c{d}_{R}\left(L\right)=c{d}_{R}\left({H}^{\prime }\right)\le k$  . Hence, $vc{d}_{R}\left(H\right)\le k$  .
3. Special classes of Groups
The following class of groups was introduced by P. Kropholler in [8], p.
57-60. Our next theorem shows that they satisfy the conjecture stated in section 1.
Definition 3.1. Let $\mathcal{X}$  denote a class of groups. Define $H\mathcal{X}$  to be the smallest class of groups containing $\mathcal{X}$  with the property: if a group $G$  acts cellularly on a finite dimensional contractible CW-complex with all isotropy subgroups in $H\mathcal{X}$  , then $G$  is in $H\mathcal{X}$  .
In [8] many properties of these classes of groups, such as subgroup and extension closure, closure under countable direct unions and free product, were shown. The main interest to us is the hierarchical description of $H\mathcal{X}$  -groups defined by operations ${H}_{\alpha }$  for each ordinal $\alpha$  inductively:
${H}_{0}\mathcal{X}=\mathcal{X}$  For each $\beta >0$  , ${H}_{\beta }\mathcal{X}$  is the class of groups $G$  which act cellularly on a finite dimensional contractible CW-complex $X$  such that for each cell $\sigma$  of $X$  the isotropy group ${G}_{\sigma }$  is in ${H}_{\alpha }\mathcal{X}$  for some $\alpha <\beta$  .
It is immediate that a group $G$  is a $H\mathcal{X}$  -group if and only if there is an $\alpha$  such that $G$  is in ${H}_{\alpha }\mathcal{X}$  . Let $\mathcal{ℱ}$  denote the class of all finite groups.
We obtain the following result pertaining $H\mathcal{ℱ}$  -groups.
Theorem 3.2. Let $G$  be a discrete group without $R$  -torsion with jump cohomology of height $k$  over $R$  . If $G$  is an $H\mathcal{ℱ}$  -group, then $c{d}_{R}\left(G\right)\le k$  .
In particular, any $H\mathcal{ℱ}$  -group $G$  has jump cohomology of height $k$  over $\mathbb{Q}$  if and only if $c{d}_{\mathbb{Q}}\left(G\right)\le k$  .
First we need a lemma.
Lemma 3.3. If a discrete group $G$  acts freely and cellularly on an $n$  -dimensional contractible complex $X$  , such that for any cell stabilizers of $X$  , $c{d}_{R}\left({G}_{\sigma }\right)\le k$  , then $c{d}_{R}\left(G\right)\le k+n$  .
• Proof. By considering the double complex $Ho{m}_{RG}\left({P}_{*},{C}_{R}^{*}\left(X,M\right)\right)$  , where ${P}_{*}$  is a projective resolution of $R$  over $RG$  and ${C}_{R}^{*}\left(X,M\right)$  is the cellular cochain complex of $X$  with $RG$  -module coefficients $M$  (see for example [4] p. 164) we can derive the first-quadrant spectral sequence:
${E}_{1}^{p,q}=$  Ext ${}_{RG}^{q}\left(R,{C}_{R}^{p}\left(X,M\right)\right)⟹$  Ext ${}_{RG}^{p+q}\left(R,{C}_{R}^{*}\left(X,M\right)\right)$
where Ext ${}_{RG}^{*}\left(R,{C}_{R}^{*}\left(X,M\right)\right)$  is the cohomology of the total complex associated to the double complex.
For each $p$  -cell $\sigma$  of $X$  there exists a ${G}_{\sigma }$  -module ${R}_{\sigma }$  . This module is isomorphic to $R$  additively, and ${G}_{\sigma }$  acts on it through the orientation character. Let ${M}^{\sigma }=Ho{m}_{R}\left({R}_{\sigma },M\right)$  . Let ${X}_{p}$  denote the collection of all the $p$  -cells and let ${\Sigma }_{p}$  be a set of representatives of all the $G$  -orbits in ${X}_{p}$  . We have the following decomposition:
${C}_{R}^{p}\left(X,M\right)=Ho{m}_{R}\left({C}_{p}^{R}\left(X\right),M\right)={\oplus }_{\sigma \in {X}_{p}}Ho{m}_{R}\left({R}_{\sigma },M\right)=$  ${\oplus }_{\sigma \in {\Sigma }_{p}}Ho{m}_{R}\left($  Ind ${}_{{G}_{\sigma }}^{G}{R}_{\sigma },M\right)={\oplus }_{\sigma \in {\Sigma }_{p}}$  Coind ${}_{{G}_{\sigma }}^{G}{M}^{\sigma }$
Now, by Shapiro's lemma,
Ext ${}_{RG}^{q}\left(R,{C}_{R}^{p}\left(X,M\right)\right)\sim ={\oplus }_{\sigma \in {\Sigma }_{p}}$  Ext ${}_{R\left[{G}_{\sigma }\right]}^{q}\left(R,{M}^{\sigma }\right)$  .
Since $X$  is contractible the two cochain complexes ${C}_{R}^{*}\left(X,M\right)$  and ${C}_{R}^{*}\left(pt.,M\right)$  are homotopy equivalent. Hence,
Ext ${}_{RG}^{*}\left(R,{C}_{R}^{*}\left(X,M\right)\sim =$  Ext ${}_{RG}^{*}\left(R,{C}_{R}^{*}\left(pt.,M\right)\right)=$  Ext ${}_{RG}^{*}\left(R,M\right)$  .
The spectral sequence then becomes:
${E}_{1}^{p,q}={\oplus }_{\sigma \in {\Sigma }_{p}}$  Ext ${}_{R\left[{G}_{\sigma }\right]}^{q}\left(R,{M}^{\sigma }\right)⟹$  Ext ${}_{RG}^{p+q}\left(R,M\right)$
It follows that ${E}_{1}^{p,q}=0$  if $p>n$  or $q>k$  . We can infer inductively for $i=0,1,2,...$  that the differentials ${d}_{i}^{p,q}=0$  and the terms of the spectral sequence ${E}_{i}^{p,q}=0$  if $p>n$  or $q>k$  . Therefore, ${E}_{\infty }^{p,q}\sim ={E}_{1}^{p,q}=0$  if $p>n$  or $q>k$  and Ext ${}_{RG}^{p+q}\left(R,M\right)=0$  if $p>n$  and $q>k$  . Thus $c{d}_{R}\left(G\right)\le n+k$  .
• Proof of Theorem. We will show that $c{d}_{R}\left(G\right)\le k$  by transfinite induction, using the hierarchical discription of $H\mathcal{ℱ}$  -groups.
If $G$  is in ${H}_{0}\mathcal{ℱ}=\mathcal{ℱ}$  , then $G$  is a finite group. So, $c{d}_{R}\left(G\right)=0$  .
Assume now that for a fixed $\beta$  , and all $\alpha <\beta$  , $c{d}_{R}\left(H\right)\le k$  for any subgroup $H$  of $G$  in ${H}_{\alpha }\mathcal{ℱ}$  . Suppose $G$  is an ${H}_{\beta }\mathcal{ℱ}$  -group. By assumption there is a finite dimensional contractible $G$  -CW-complex $X$  such that for any cell $\sigma$  of $X$  the isotropy group ${G}_{\sigma }$  is an ${H}_{\alpha }\mathcal{ℱ}$  -group for some $\alpha <\beta$  . So, by induction, $cd\left({G}_{\sigma }\right)\le k$  . Let $n$  be the dimension of $X$  . Now, by Lemma 3.3, $c{d}_{R}\left(G\right)\le k+n$  . Hence, by hypothesis, $c{d}_{R}\left(G\right)\le k$  .
Our next goal is to establish the conjecture for first nilpotent then solvable groups. For this we need to assume $R$  is an integral domain.
Note that by the previous theorem this is true for countable $R$  -torsion-free solvable groups, for they are $H\mathcal{ℱ}$  -groups, for arbitrary commutative rings $R$  . So, our next theorem extends this fact to uncountable solvable groups when $R$  is a domain. In fact, in this setting, if we also assume $G$  is torsion-free, our results show that $G$  cannot be uncountable. We denote by $\mathbb{F}$  the fraction field of $R$  , so $\mathbb{F}$  is a field of an arbitrary characteristic.
Lemma 3.4. Let $1={G}_{0}\subset {G}_{1}\subset$  ... $\subset {G}_{n}=G$  be an upper central series of $G$  . If $G$  is an $R$  -torsion-free group, then so are the factor groups ${G}_{i}/{G}_{i-1}$  , $i=1,..,n$  .
• Proof. Recall that an upper central series for $G$  can be defined inductively by letting ${G}_{i}$  to be a normal subgroup of $G$  such that ${G}_{i}/{G}_{i-1}\sim =Z\left(G/{G}_{i-1}\right)$  , where $Z\left(G/{G}_{i-1}\right)$  is the center of the quotient group $G/{G}_{i-1}$  .
We will prove the lemma by induction. Suppose, for all $j\le i$  , ${G}_{j}/{G}_{j-1}$  are $R$  -torsion-free. We need to show that the same is true for the quotient group ${G}_{i+1}/{G}_{i}$  . Suppose otherwise; that there exists an nontrivial element $\overline{x}\in {G}_{i+1}/{G}_{i}$  such that ${\overline{x}}^{m}=1$  and $m\cdot {1}_{R}$  is not invertible in the ring $R$  . This implies that ${x}^{m}\in {G}_{i}$  and $x\in {G}_{i+1}\{G}_{i}$  .
Now for any element $g$  of $G$  , $\left[g,x{\right]}^{m}\equiv \left[g,{x}^{m}\right]\equiv 1$  mod ${G}_{i-1}$  . Since $\left[g,x\right]\in {G}_{i}$  , the induction shows $\left[g,x\right]\in {G}_{i-1}$  . So, in the quotient group $G/{G}_{i-1}$  , $\left[\overline{g},\overline{x}\right]=1$  for all $\overline{g}\in G/{G}_{i-1}$  . Therefore, by definition, $\overline{x}\in {G}_{i}/{G}_{i-1}$  and $x\in {G}_{i}$  , a contradiction.
Proposition 3.5. Let $G$  be a nilpotent group without $R$  -torsion and with jump cohomology of height $k$  over $R$  . Then the Hirsch rank $h\left(G\right)<\infty$  , $c{d}_{R}\left(G\right)\le k$  and Tor ${}_{h{d}_{R}\left(G\right)}^{RG}\left(R,\mathbb{F}\right)\sim =\mathbb{F}$  , where $h{d}_{R}\left(G\right)$  is the homological dimension of $G$  over $R$  .
• Proof. Let $1={G}_{0}\subset {G}_{1}\subset$  ... $\subset {G}_{n}=G$  be a central series of $G$  with $R$  -torsion-free factor groups. We use induction on $n$  .
For $n=0$  , $h{d}_{R}\left({G}_{0}\right)=c{d}_{R}\left({G}_{0}\right)=0$  and Tor ${}_{0}^{R\left[{G}_{0}\right]}\left(R,\mathbb{F}\right)={\mathbb{F}}^{{\omega }_{0}}$  , where $\mathbb{F}$  is the trivial $G$  -module and ${\mathbb{F}}^{{\omega }_{0}}$  is the trivial $G/{G}_{0}$  -module $\mathbb{F}$  . Let us assume $h{d}_{R}\left({G}_{i}\right)={h}_{i}$  , $h\left({G}_{i}\right)<\infty$  , $c{d}_{R}\left({G}_{i}\right)\le k$  and Tor ${}_{{h}_{i}}^{R{G}_{i}}\left(R,\mathbb{F}\right)={\mathbb{F}}^{{\omega }_{i}}$  , where ${\mathbb{F}}^{{\omega }_{i}}$  is the trivial $G/{G}_{i}$  -module $\mathbb{F}$  . Let $A\subset {G}_{i+1}/{G}_{i}$  and $A\sim ={\mathbb{Z}}^{m}$  . Let ${G}_{i+1}^{\prime }={\phi }^{-1}\left(A\right)$  , where $\phi :{G}_{i+1}\to {G}_{i+1}/{G}_{i}$  is the canonical quotient map. So there exists an extension, $1\to {G}_{i}\to {G}_{i+1}^{\prime }\to A\to 1$  .
By the Hochschild-Serre spectral sequence and the Maximal Principle,
Tor ${}_{m+{h}_{i}}^{R\left[{G}_{i+1}^{\prime }\right]}\left(R,\mathbb{F}\right)\sim =$  Tor ${}_{m}^{RA}\left(R,$  Tor ${}_{{h}_{i}}^{R\left[{G}_{i}\right]}\left(R,\mathbb{F}\right)\right)\sim =$  Tor ${}_{m}^{R\left[{G}_{i+1}^{\prime }/{G}_{i}\right]}\left(R,{\mathbb{F}}^{{\omega }_{i}}\right)$
Since ${G}_{i+1}^{\prime }/{G}_{i}\sim ={\mathbb{Z}}^{m}$  ,
Tor ${}_{m}^{R\left[{G}_{i+1}^{\prime }/{G}_{i}\right]}\left(R,{\mathbb{F}}^{{\omega }_{i}}\right)\sim ={\wedge }^{m}\left({G}_{i+1}^{\prime }/{G}_{i}\otimes {\mathbb{F}}^{{\omega }_{i}}\right)\sim ={\wedge }^{m}\left({\mathbb{Z}}^{m}\otimes {\mathbb{F}}^{{\omega }_{i}}\right)\sim =\mathbb{F}$
where ${\wedge }^{*}\left({G}_{i+1}^{\prime }/{G}_{i}\otimes {\mathbb{F}}^{{\omega }_{i}}\right)$  is the exterior algebra of the $\mathbb{F}$  -vector space ${G}_{i+1}^{\prime }/{G}_{i}\otimes {\mathbb{F}}^{{\omega }_{i}}$  . This shows Tor ${}_{m+{h}_{i}}^{R\left[{G}_{i+1}^{\prime }\right]}\left(R,\mathbb{F}\right)\ne 0$  . On the other hand by the Maximal Principle, $h{d}_{R}\left({G}_{i+1}^{\prime }\right)\le m+{h}_{i}$  . Thus $h{d}_{R}\left({G}_{i+1}^{\prime }\right)=m+{h}_{i}$  .
Since ${G}_{i+1}^{\prime }$  has finite cohomological dimension over $R$  and $h{d}_{R}\left({G}_{i+1}^{\prime }\right)\le c{d}_{R}\left({G}_{i+1}^{\prime }\right)$  , the hypothesis imply $m+{h}_{i}\le k$  , showing $m\le k$  . This proves that ${G}_{i+1}/{G}_{i}$  is a finite rank abelian group. Thus, ${G}_{i+1}$  has finite Hirsch rank, and since ${G}_{i+1}/{G}_{i}$  is an $R$  -torsion-free abelian group, ${G}_{i+1}/{G}_{i}$  has finite cohomological dimension over $R$  . So, again by the Maximal Principal, we have $c{d}_{R}\left({G}_{i+1}\right)<\infty$  , hence $c{d}_{R}\left({G}_{i+1}\right)\le k$  . Let $m=hd\left({G}_{i+1}/{G}_{i}\right)$  . Again by the Hochschild-Serre spectral sequence
Tor ${}_{m+{h}_{i}}^{R\left[{G}_{i+1}\right]}\left(R,\mathbb{F}\right)\sim =$  Tor ${}_{m}^{R\left[{G}_{i+1}/{G}_{i}\right]}\left(R,$  Tor ${}_{{h}_{i}}^{R\left[{G}_{i}\right]}\left(R,\mathbb{F}\right)\right)\sim =$  Tor ${}_{m}^{R\left[{G}_{i+1}/{G}_{i}\right]}\left(R,{\mathbb{F}}^{{\omega }_{i}}\right)\sim ={\wedge }^{m}\left({G}_{i+1}/{G}_{i}\otimes {\mathbb{F}}^{{\omega }_{i}}\right)$
The last isomorphism follows from the fact that ${G}_{i+1}/{G}_{i}$  is an $R$  -torsion-free abelian group and ${\mathbb{F}}^{{\omega }_{i}}$  has trivial ${G}_{i+1}/{G}_{i}$  action. Also, since ${G}_{i+1}/{G}_{i}$  is central in $G/{G}_{i}$  , the induced action of $G/{G}_{i+1}$  on ${\wedge }^{m}\left({G}_{i+1}/{G}_{i}\otimes {\mathbb{F}}^{{\omega }_{i}}\right)$  is trivial. So the module ${\wedge }^{m}\left({G}_{i+1}/{G}_{i}\otimes {\mathbb{F}}^{{\omega }_{i}}\right)$  is ${\mathbb{F}}^{{\omega }_{i+1}}$  . This completes the induction.
Theorem 3.6. Let $G$  be an R-torsion-free, solvable group with jump cohomology of height $k$  over $R$  , where $R$  is an integral domain. Then $c{d}_{R}\left(G\right)\le k$  and $G$  has finite Hirsch rank.
• Proof. By Proposition 3.5 all abelian subgroups of $G$  have finite rank.
So by [3], p. 105, $G$  has a unique maximal nilpotent normal subgroup $N$  (the Hirsch-Plotkin-radical) such that $G/N$  contains a finitely generated free abelian group $A$  of finite index, $1\to N\to G\stackrel{\phi }{\to }G/N\to 1$  .
Since, by Proposition 3.5 $h\left(N\right)<\infty$  , this immediately implies that $h\left(G\right)<\infty$  .
Since $A$  is a finitely generated free group it is isomorphic to ${\mathbb{Z}}^{m}$  for some positive integer $m$  . Let ${G}^{\prime }={\phi }^{-1}\left(A\right)$  . Since $A$  is a finite index subgroup of $G/N$  , ${G}^{\prime }$  is a finite index subgroup of $G$  . By the Serre Theorem, $c{d}_{R}\left({G}^{\prime }\right)=c{d}_{R}\left(G\right)$  . By Proposition 3.4, $c{d}_{R}\left(N\right)=n<\infty$  .
So the Maximal Principal applied to the extension $1\to N\to {G}^{\prime }\to {\mathbb{Z}}^{m}\to 1$  shows $c{d}_{R}\left({G}^{\prime }\right)\le n+m$  .
Example 3.7. Let $F$  denote the Thompson group defined by the presentation
$〈{x}_{0},{x}_{1},{x}_{2},...|{x}_{i}^{-1}{x}_{n}{x}_{i}={x}_{n+1}$  for all $i  and $n\in \mathbb{N}〉$
This group is finitely presented for it also has the presentation
$〈{x}_{0},{x}_{1}|{x}_{2}^{{x}_{0}}={x}_{3},{x}_{3}^{{x}_{1}}={x}_{4}〉$
Brown and Goeghegan [5] showed that it is also of type FP ${}_{\infty }$  , i.e.
it has a projective $\mathbb{Z}G$  -resolution of $\mathbb{Z}$  , where each module is finitely generated. They also show ${H}_{n}\left(F,\mathbb{Z}\right)=\mathbb{Z}\oplus \mathbb{Z}$  for $n\ge 1$  and ${H}^{n}\left(F,\mathbb{Z}F\right)=0$  for all $n$  . This may suggest that this group has a periodic cohomology over some ring. But, in fact, $F$  cannot have periodic or jump cohomology over any $R$  , since it has an infinite rank abelian subgroup ${\mathbb{Z}}^{\infty }$  . Namely, $〈{x}_{0}{x}_{1}^{-1},{x}_{2}{x}_{3}^{-1},{x}_{4}{x}_{5}^{-1},...〉\subseteq F$  . This reinstates our belief that periodicity and jumps are rather special conditions on the cohomology of a discrete group, since countable groups containing the Thompson group are the only class of groups, known to us, not to be $H\mathcal{ℱ}$  .
Corollary 3.8. If $G$  has a non-countable, torsion-free, solvable subgroup, then G does not have periodic cohomology over $R$  , where $R$  is an integral domain. In particular, $G$  cannot act freely and properly on a finite dimensional complex homotopy equivalent to a sphere.
• Proof. Let $H$  be a non-countable, torsion-free, solvable subgroup of $G$  .
By [3], Lemma 7.9, p. 100, $H$  has infinite Hirsch rank. So by Theorem 3.6 $H$  cannot have periodic cohomology over $R$  . Specifically, $G$  does not have periodic cohomology, therefore cannot act freely and properly on a finite dimensional complex homotopy equivalent to a sphere.
Corollary 3.9. If $G$  is a torsion-free solvable group that acts freely and properly on an $\left(n+1\right)$  -dimensional complex $E$  homotopy equivalent to ${S}^{n}$  , then $G$  must be infinite cyclic or trivial.
• Proof. By Proposition 2.4, it follows that any subgroup $H$  of $G$  with finite cohomological dimension must have $cd\left(H\right)\le$  dim $\left(E\right)-n=1$  .
But, by Theorem 3.6 $cd\left(G\right)<\infty$  . Therefore $cd\left(G\right)\le 1$  . Now Theorem of Stallings and Swan [12, Theorem A] says that $G$  is a free group. The infinite cyclic group is the only nontrivial solvable free group.
Corollary 3.10. Let $G$  be a virtually $R$  -torsion-free, virtually solvable group with virtually periodic cohomology over an integral domain $R$  .
Then $G$  has finite virtual cohomological dimension over $R$  .
4. Free and Proper Actions
This next results apply only to countable torsion-free groups. We find an obstruction for a group to be free in a specific setting. It would be interesting to know whether the following theorem holds for not necessarily countable groups and what implications there are when the obstruction does not vanish.
Theorem 4.1. Let $G$  be a finitely generated torsion-free group that acts freely and properly on an $n$  -dimensional CW-complex $X$  such that the $G$  -invariant submodule $\left({H}_{n-1}\left(X,R\right){\right)}^{G}$  is not trivial. Then $G$  is a free group if and only if Ext ${}_{RG}^{2}\left(R,{H}_{n}\left(X,R\right)\right)=0$  .
• Proof. If $G$  is a free group then $c{d}_{R}\left(G\right)\le 1$  , so Ext ${}_{RG}^{2}\left(R,{H}_{n}\left(X,R\right)\right)$  must be trivial.
Conversely, suppose Ext ${}_{RG}^{2}\left(R,{H}_{n}\left(X,R\right)\right)=0$  . Let $\left({C}_{*}\left(X\right),{d}_{*}\right)$  be the chain complex of $X$  , where ${C}_{i}\left(X\right)$  is the free $R$  -module generated by the $i$  -cells of $X$  . We proceed by induction on the number of generators of $G$  with the case of $G$  having exactly one generator understood.
There exists an exact sequence of $\mathbb{Z}G$  -modules,
$0\to {d}_{n}\left({C}_{n}\left(X\right)\right)\to$  ker $\left({d}_{n-1}\right)\to {H}_{n-1}\left(X,R\right)\to 0$
and an associated long exact sequence,
Ext ${}_{RG}^{0}\left(R,$  ker $\left({d}_{n-1}\right)\right)\to$  Ext ${}_{RG}^{0}\left(R,{H}_{n-1}\left(X,R\right)\right)\to$  Ext ${}_{RG}^{1}\left(R,{d}_{n}\left({C}_{n}\left(X\right)\right)\right)\to$  Ext ${}_{RG}^{1}\left(R,$  ker $\left({d}_{n-1}\right)\right)$
The first term of the long exact sequence is the $G$  -invariant submodule (ker $\left({d}_{n-1}\right){\right)}^{G}$  . This module is trivial, for it is a submodule of the $G$  -invariant submodule $\left({C}_{n-1}\left(X\right){\right)}^{G}$  . But $G$  is infinite and ${C}_{n-1}\left(X\right)$  is a free $RG$  -module, hence $\left({C}_{n-1}\left(X\right){\right)}^{G}=0$  . Thus the second term, the $G$  -invariant submodule $\left({H}_{n-1}\left(X,R\right){\right)}^{G}$  , of the long exact sequence injects into Ext ${}_{RG}^{1}\left(R,{d}_{n}\left({C}_{n}\left(X\right)\right)\right)$  . This shows that Ext ${}_{RG}^{1}\left(R,{d}_{n}\left({C}_{n}\left(X\right)\right)\right)\ne 0$  .
On the other hand there exists an exact sequence of $RG$  -modules,
$0\to {H}_{n}\left(X,R\right)\to {C}_{n}\left(X\right)\to {d}_{n}\left({C}_{n}\left(X\right)\right)\to 0$
The long exact sequence associated to this gives,
Ext ${}_{RG}^{1}\left(R,{H}_{n}\left(X,R\right)\right)\to$  Ext ${}_{RG}^{1}\left(R,{C}_{n}\left(X\right)\right)\to$  Ext ${}_{RG}^{1}\left(R,{d}_{n}\left({C}_{n}\left(X\right)\right)\right)\to$  Ext ${}_{RG}^{2}\left(R,{H}_{n}\left(X,R\right)\right)$
According to our assumption Ext ${}_{RG}^{2}\left(R,{H}_{n}\left(X,R\right)\right)=0$  , so the term Ext ${}_{RG}^{1}\left(R,{C}_{n}\left(X\right)\right)$  surjects onto the module Ext ${}_{RG}^{1}\left(R,{d}_{n}\left({C}_{n}\left(X\right)\right)\right)$  , which is nontrivial. Therefore Ext ${}_{RG}^{1}\left(R,{C}_{n}\left(X\right)\right)\ne 0$  . Now since $G$  is finitely generated and ${C}_{n}\left(X\right)$  is a free $RG$  -module, we must have Ext ${}_{RG}^{1}\left(R,RG\right)\ne 0$  . So by [6] $G$  must split as a nontrivial free product of groups.
Suppose then $G=H*K$  where $H,K\ne G$  . Since,
Ext ${}_{RG}^{2}\left(R,M\right)=$  Ext ${}_{RH}^{2}\left(R,M\right)\oplus$  Ext ${}_{RK}^{2}\left(R,M\right)$
for any $RG$  -module $M$  , it follows then
Ext ${}_{RH}^{2}\left(R,{H}_{n}\left(X,R\right)\right)=$  Ext ${}_{RK}^{2}\left(R,{H}_{n}\left(X,R\right)\right)=0$  .
Also by Grusko's Theorem [6] there exist a fewer number of generators for each of the groups $H$  and $K$  than the number of generators of $G$  .
So we can apply induction on these subgroups to show that they are free.
Corollary 4.2. If $G$  is a torsion-free countable group that acts freely and properly on an $\left(n+1\right)$  -dimensional complex $X$  homotopy equivalent to ${S}^{n}$  , then $G$  must be free.
• Proof. First, let us assume $G$  is finitely generated. ${H}_{n}\left(X\right)=\mathbb{Z}$  . If the induced action of $G$  on ${H}_{n}\left(X\right)$  is nontrivial then $G$  must contain an index two subgroup $H$  which acts trivially on ${H}_{n}\left(X\right)$  . Thus, there always exists such $H\subseteq G$  acting trivially ,with $\left[G:H\right]\le 2$  . Also, since ${H}_{n+1}\left(X\right)=0$  , ${H}^{2}\left(H,{H}_{n+1}\left(X\right)\right)=0$  . By Theorem 4.1 it follows that $H$  is free. Since $H$  is a finite index subgroup, $G$  must be free.
In general, since $G$  is a countable group, it is a countable union of an ascending chain of finitely generated subgroups. By the Theorem of Berstein, [7] p. 132, we have $cd\left(G\right)\le 2$  . Since $G$  acts on an $\left(n+1\right)$  -dimensional complex homotopy equivalent to ${S}^{n}$  , by Proposition 2.5, $cd\left(G\right)\le 1$  . Therefore it is free.
References
[1] A. Adem and J. Smith, Periodic complexes and group actions.
Annals of Math. (2) (154) (2001).
[2] R.C. Alperin and P.B. Shalen, Linear groups of finite cohomological dimension, Invent. Math. 66 (1982) 89-98.
[3] R. Bieri, Homological Dimension of Discrete Groups, Queen Mary College, London, UK, 1981.
[4] K. Brown, Cohomology of Groups, Graduate Text in Math. 87, Springer-Verlag, New York (1982).
[5] K. Brown and R. Geoghegan, An infinite-dimensional torsion-free FP ${}_{\infty }$  group, Invent. Math. 77 (1984) 367-381.
[6] D. Cohen, Groups of Cohomological Dimension One, Lecture Notes in Mathematics, 245. Berlin Heidelberg New-York, Springer-Verlag (1972). [7] K. Gruenberg, Cohomological Topics in Group Theory, Lecture Notes in Mathematics, 143. Berlin Heidelberg New-York, Springer-Verlag (1970).
[8] P. Kropholler, On groups of type FP ${}_{\infty }$  , J. Pure. Appl. Algebra 90(1993), 55-67.
[9] G. Mislin and O. Talelli, On groups which act freely and properly on finite dimensional homotopy spheres GAMA Proc. LMS 275(2000).
[10] J.-P. Serre, Cohomologie des groupes discretes. Annals of Math. Studies 70 , Princeton (1971).
[11] Stallings, On torsion free groups with infinitely many ends. Annals of Math 88(1968), 312-334.
[12] R. G. Swan, Groups of Cohomological Dimension One, J. Algebra 12 (1969), 585-601.
[13] O. Talelli , On Groups with Periodic Cohomology after 1-step, J. Pure. Appl Algebra 30(1983), 85-93.
[14] O. Talelli , Periodicity in Group Cohomology and Complete Resolutions, to appear in Bull. of LMS. ?? (????).