arXiv:1709.05616v2 [math.GR] 12 Dec 2017 aeapiain fhsmto otechmlge ftetr the of cohomologies groups. the Galois to some method of abe and his finite of simp of [ cohomologies Takahashi applications are of gave instance, calculation which For the to resolutions, one. approach of new standard sorts the the than special In to convenient devoted classes. needs their papers often and plen of groups one by lot concrete a studied of is cohomologies actively there of is particular, it In Thus geo maticians. algebraic mathematics. theory, number of topology, branches in used group o widely of origins f is theory as ory the the arise of cohomologies to where related one representations, also was projective was b It It 1940’s in others. algebra. founded mological and was Hopf and Eckmann, spaces MacLane, acyclic of cohomologies on bu h eodchmlge a eueu ntesuyo cr of groups. study Chernikov the in of useful and be groups can cohomologies second the about bla rus(eto )adt s tfrcluaino co of calculation for it use to and irreducible 1) (Section groups abelian ls ota fTkhsi huhi em oeepii.We pro and of 2) explicit. cohomologies (Section more one for seems standard duality the it concerning with though resolution Takahashi, our of pare that to close s G a i 2010 o eidcelement periodic a For h hoyo ooooiso ruswsisie ytework the by inspired was groups of cohomologies of theory The h i forppri odsrb ahrsml resolution simple rather a describe to is paper our of aim The e od n phrases. and words Key = = P h OOOOISO IIEAEINGROUPS ABELIAN FINITE OF COHOMOLOGIES a ahmtc ujc Classification. Subject Mathematics Z eouin euei ocluaechmlge firreduci of cohomologies calculate to it use We resolution. Abstract. n hi duals. their and o i i =0 ( where , a | ) a − G i o 1 i ltie n hi ul Scin n ) u prahis approach Our 5). and 4 (Sections duals their and -lattices a 1 = i UI .DODADADIN .PLAKOSH I. ANDRIANA AND DROZD A. YURIY Let . G safiieaeingop n opr twt h standard the with it compare and group, abelian finite a is i ecntutasmlfidrslto o h trivial the for resolution simplified a construct We forders of G oooois nt bla rus resolution, groups, abelian finite cohomologies, = Q a fagroup a of i s =1 o 1. Introduction i G = Resolution i o eadrc rdc ffiiecci groups cyclic finite of product direct a be ( a 1 i 00,1G0 20K01. 18G10, 20J06, ), G G R ednt by denote we ltie Scin3.Teresults The 3). (Section -lattices = Z G , P = R co es hsthe- This sets. actor o ble [ ( x a 1 h re of order the ) einvestigations se er n other and metry G x , G ingop and groups lian h calculation the xesosand extensions -lattices -module ystallographic ]pooe a proposed 7] oooisof homologies 2 e n more and ler esm facts some ve yo mathe- of ty Eilenberg– y va module ivial x , . . . , fHurewicz of s G -lattice. locom- also h ho- the f o finite for s and ] a , 2 YU.DROZD,A.PLAKOSH
Pn be the set of homogeneous polynomials from P of degree n (including 0). We define a differential d : Pn → Pn−1 by the rule
s k1 k2 ks Ki k1 ki−1 ks dn(x1 x2 ...xs )= (−1) Cix1 ...xi ...xs , Xi=1 i−1 where Ki = j=1 kj and P ai − 1 if ki is odd,
Ci = sai if ki > 0 is even, 0 if ki = 0.
24 (2017), 144157 When speaking of the G-module Z, we always suppose that the elements of G act trivially.
Theorem 1.1. P = (Pn, dn) is a free resolution of the G-module Z.
i Proof. If s = 1, it is well-known. If Ri = ZGi and P denotes such res- R s R olution for the group Gi, then = i=1 i and P is the tensor product of complexes s Pi. As all groups of cycles and boundaries in the com- i=1 N plexes Pi are free abelian, the claim follows from the K¨unneth relations [3, Theorem VI.3.1].N
2. Correspondence with standard resolution To apply Theorem 1.1, for instance, to extensions of groups, we have to compare it with the standard resolution, which is usually used for this purpose [2, 3]. So, in what follows, S denotes the normalized standard resolution for Z as R-module, { [g1, g2,...,gn] | gi ∈ G \ {1} } is the usual s basis of Sn such that the standard differential d is defined as
n s i dn[g1, g2,...,gn]= g1[g2,...,gn]+ (−1) [g1,...,gigi+1,...,gn]+ i=1 X n + (−1) [g1, g2,...,gn−1], setting [g1, g2,...,gn] = 0 if some gi = 1. Note that P0 = S0 = R. {i} 2 i−1 {o(a)} We denote a =1+ a + a + . . . a . Then sa = a ,
(2.1) a{i+k} = a{i} + aia{k}, in particular,
{m+o(a)} {m} m a = a + a sa. COHOMOLOGIESOFFINITEABELIANGROUPS 3
Theorem 2.1. There is a quasi-isomorphism σ : S → P such that
σ0 = id, s i−1 k1 k2 ks kj {ki} σ1[a1 a2 . . . as ]= aj ai xi, i=1 j=1 X Y s i i−1 j−1 (2.2) k1 k2 ks l1 l2 ls kq lr ki lj σ2[a1 a2 . . . as , a1 a2 . . . as ]= aq ar σ2[ai , aj ], q=1 r=1 Xi=1 Xj=1 Y Y 2 [(k + l)/oi]xi if i = j, k l where σ2[ai , aj]= 0 if i < j, {l} {k} aj ai xjxi if i > j 24 (2017), 144157
Since S and P are free resolutions of Z, σ induces isomoprhisms of coho- mologies Hn(HomR(S, M)) → Hn(HomR(P, M)). In particular, combining σ2 with cocycles from HomR(P2, M), we obtain the “usual” presentation of cocycles from H2(G, M).
Proof. Actually, we have to show that the diagram
s s d2 / d1 / S2 S1 S0
σ2 σ1 d2 / d1 / P2 P1 P0. is commutative. Then the set of homomorphisms {σ0,σ1,σ2} extends to a quasi-isomorphism σ : S → P. Note that gh − 1 = (g − 1) + g(h − 1) and ak − 1= a{k}(a − 1). Therefore,
s i−1 s k1 k2 ks k1 k2 ks kj ki d1[a1 a2 . . . as ]= a1 a2 . . . as − 1= aj (ai − 1) = i=1 j=1 X Y s i−1 s i−1 kj {ki} kj {ki} = aj ai (ai − 1) = aj ai d1xi, i=1 j=1 i=1 j=1 X Y X Y s hence d1σ1 = d1. Set (r)i = res(r, oi), the residue of r modulo oi. Then, for 0 ≤ k < oi, 0 ≤ l < oi,
s k l k l k+l k d2[ai , ai]= ai [ai] − [ai ] + [ai ], 4 YU.DROZD,A.PLAKOSH thus
s k l k {l} {(k+l)i} {l} σ1d2[ai , ai] = (ai ai − ai + ai )xi = k {l} {k+l} {l} = (ai ai − ai + ai + [(k + l)/oi]sai )xi =
= [(k + l)/oi]sai xi = 2 = d2([(k + l)/oi]xi ), so, if we set
k l 2 σ2[ai , ai] = [(k + l)/oi]xi , we have
k l s k l d2σ2[ai , ai]= σ1d2[ai , ai].
In the same way,
s k l k l k l k d2[ai , aj]= ai [aj] − [ai aj] + [ai ], thus, if i < j,
s k l k {l} {k} k {l} {k} σ1d2[ai , aj]= ai aj xj − ai xi − ai aj xj + ai xi = 0, while if i > j
s k l k {l} {l} l {k} {k} σ1d2[ai , aj]= ai aj xj − aj xj − ajai xi + ai xi = k {l} l {k} = (ai − 1)aj xj − (aj − 1)ai xi = {l} {k} = −d2(aj ai xjxi).
So, if we set
k l 0 if i < j, σ2[ai , aj]= {l} {k} (aj ai xjxi if i > j we have
k l s k l 24(2017), 144157d2σ2[ai , aj]= σ1d2[ai , aj] for i 6= j. s Let now σ2 is defined by the rule (2.2). We check that d2σ2 = σ1d2 for s = 3. The general case is analogous, though a bit cumbersome. We write COHOMOLOGIESOFFINITEABELIANGROUPS 5 a, b, c instead of a1, a2, a3 and x,y,z instead of x1,x2,x3. Then s i j r k l s σ1d2[a b c , a b c ]= i j r k l s i+k j+l r+s i j r = σ1(a b c [a b c ] − [a b c ] + [a b c ]) = = aibjcr(a{k}x + akb{l}y + akblc{s}z)− − a{i+k}x − ai+kb{j+l}y − ai+kbj+lc{r+s}z+ i+k i+k j+l + [(i + k)/oa]sax + a [(j + l)/ob]sby + a b [(r + s)/oc]scx+ + a{i}x + aib{j}y + aibjc{r}z = i j r {k} {i+k} {i} = (a b c a − a + a + [(i + k)/oa]sa)x+ i k j r {l} k {j+l} {j} k + a (a b c b − a b + b + a [(j + l)/ob]sb)y+ i j k l r {s} k l {r+s} {r} k l + a b (a b c c − a b c + c + a b [(r + s)/oc]sc)z, while
i j r k l s d2σ2[a b c , a b c ]= i {k} {j} i j {k} {s} i+k j {l} {r} 24(2017), 144157 = d2 − a a b xy − a b a c xz − a b b c yz+ 2 i+k 2 i+k j+l 2 + [(i + k)/oa]x + a [(j + l)/ob]y + a b [(r + s)/oc]z = = −ai(ak − 1)b{j}y + ai(bj − 1)a{k}x − ai(ak − 1)bj c{r}z+ + aibj(cr − 1)a{k}x − ai+kbj(bl − 1)c{r}z + ai+kbj(cr − 1)bk}y, i+k i+k j+l + [(i + k)/oa]sax + a [(j + l)/ob]sby + a b [(r + s)/oc]scx = i {k} i j r {k} (−a a + a b c a + [(i + k)/oa]sa)x+ i k {j} {j} k j r {l} k j {l} k + a (−a b + b + a b c b − a b b + a [(j + l)/ob]sb)y+ i j {r} k l {r} k l + a b (c − a b c + a b [(r + s)/oc]sc)z. Relations (2.1) immediately imply that both results are equal.
3. Cohomologies of G-lattices. In this section G denotes a finite group, R = ZG. Recall that a G-lattice (or an integral representation of G) is a G-module M such that its abelian group is free of finite rank. They also say that M is a lattice in the QG- module M˜ = Q ⊗Z M. Two G-lattices M,N are said to be of the same genus if Mp ≃ Np24 (2017), 144157 for each prime p, where Mp = Zp ⊗Z M (Zp = { r/z | r ∈ Z, s ∈ Z \ pZ }). Then they write M ∨ N. We also set ∗ −1 M = HomZ(M, Z), where G acts by the rule gf(u)= f(g u). We denote by Hˆ n(G, M) the Tate cohomologies of G with coefficients in M [2, 3]. Let
dn dn−1 d2 d1 F : ···→ Fn −→ Fn−1 −−−→ . . . −→ F1 −→ F0 → 0 6 YU.DROZD,A.PLAKOSH be a free resolution of Z, where all modules Fn are finitely generated, ∗ ∗ d∗ ∗ ∗ ∗ d1 ∗ d2 n−1 ∗ dn ∗ F : 0 → F0 −→ F1 −→ . . . −−−→ Fn−1 −→ Fn → . . . ∗ be the dual complex, d0 : F0 → F0 be the composition of the maps F0 → ∗ ∗ ∗ ∗ coker d1 ≃ Z ≃ ker d0 → F0. Set F−n = Fn−1, d−n = dn. The sequence
+ dn dn−1 d2 d1 d0 F : . . . → Fn −→ Fn−1 −−−→ . . . −→ F1 −→ F0 −→
d0 d−1 d−2 d−n ∗ d−n −→ F−1 −−→ F−2 −−→ . . . −−→ F−n −−→ F−n−1 → . . . is called a complete resolution for the group G. Then Hˆ n(G, M) are just the + cohomologies of the complex HomR(F , M). If F0 = R and the surjection F0 → Z maps g to 1, then F−1 ≃ R and d0 is just the trace, i.e. the multiplication by trG = x∈G x. It is the case for the resolutions F and S. Proposition 3.1. Let GP be a finite group, M,N be G-lattices such that M ∨ N. Then Hˆ n(G, M) ≃ Hˆ n(G, N) for all n. Proof. It is known that all groups Hˆ n(G, M) (n> 0) are periodic of period ˆ n ˆ n #(G), hence H (G, M) ≃ p|#(G) H (G, M)p. Moreover, as Zp is flat over Z ˆ n ˆ n , H (G, M)p ≃ H (G, MpL). It implies he claim.
We denote by DM the dual G-module DM = HomZ(M, T), where T = Q/Z. Proposition 3.2. Let M be a G-lattice. Then (3.1) Hˆ n−1(G, DM) ≃ DHˆ −n(G, M), (3.2) Hˆ n(G, DM) ≃ Hˆ n+1(G, M ∗), (3.3) Hˆ n(G, M ∗) ≃ DHˆ −n(G, M). If M = Z, (3.3) coincides with [3, Theorem XII.6.6]. Proof. (3.1) follows from [3, Corollary XII.6.5]. Consider the exact sequence 0 → Z → Q → T → 0. As M is free abelian, it gives the exact sequence of G-modules ∗ 0 → M → HomZ(M, Q) → DM → 0. n Hˆ (G, HomZ(M, Q)) = 0 for all n, since the multiplication by #(G) is an automorphism of HomZ(M, Q), whence we obtain (3.2). (3.3) follows from (3.1) and (3.2). We also need some information on cohomologies of direct products. Proposition 3.3. Let N be a normal subgroup of G, F = G/N and gcd(#(N), #(F )) = 1. For every G-module M and all n (3.4) Hˆ n(G, M) ≃ Hˆ n(N, M)F ⊕ Hˆ n(F, M N ). COHOMOLOGIESOFFINITEABELIANGROUPS 7
Proof. As #(G) annihilates all Hn(G, M) if n> 0 and the same is true for N and F , in the Hochschild–Serre spectral sequence Hp(F,Hq(N, M)) =⇒ Hn(G, M) all terms with p> 0 and q > 0 are zero. Hence, if n> 0, Hˆ n(G, M) ≃ H0(F, Hˆ n(N, M)) ⊕ Hˆ n(F,H0(N, M)) = =Hˆ n(N, M)F ⊕ Hˆ n(F, M N ).
Suppose now that the claim holds for Hˆ n. Choose an exact sequence 0 → L → P → M → 0, where P is a free ZG-module. Then Hˆ n−1(G, M) ≃ Hˆ n(G, L) ≃ Hˆ n(N,L)F ⊕ Hˆ n(F,LN ). As P is also free as ZN-module, Hˆ n(N,L) ≃ Hˆ n−1(N, M). On the other hand, there are exact sequences 0 → LN → P N → M ′ → 0 and 0 → M ′ → M N → M N /M ′ → 0, ′ N N ′ where M is the image of the map P → M . Obviously, M ⊇ trN M, thus #(N)(M N /M ′) = 0, whence Hˆ n(F, M N /M ′) = 0. Therefore, Hˆ n−1(F, M N ) ≃ Hˆ n−1(F, M ′) ≃ Hˆ n(F,LN ), since P N is a free ZF -module. So the isomorphism (3.4) holds for Hˆ n−1, hence for all values of n.
Corollary 3.4. Let G = G1 × G2 with gcd(#(G1), #(G2)) = 1, M = M1 ⊗Z M2, where Mi is a Gi-lattice (i = 1, 2). Then ˆ n ˆ n G2 G1 ˆ n H (G, M) ≃ H (G1, M1) ⊗Z M2 ⊕ M1 ⊗Z H (G2, M2).
Gi Gi Proof. As Mi are free abelian, ⊗ZMi is an exact functor and M = Mi ⊗Z n n Mj (j 6= i). Hence Hˆ (Gi, M) ≃ Hˆ (Gi, Mi) ⊗Z Mj, where j 6= i. So the claim is just a reformulation of Proposition 3.3 for this special case.
4. Cohomologies of irreducible G-lattices A G-lattice M is called irreducible if there are no submodules 0 6= N ⊂ M such that M/N is torsion free (i.e. again a G-lattice). Equivalently, M˜ = Q ⊗Z M is a simple QG-module. If G is a finite abelian group, then any simple QG-module is defined by a group homomorphism ρ : G → K×, where K is a cyclotomic field and the image of ρ generates the ring of integers of K. Therefore, any two G-lattices in K are of the same genus [4], so have the same cohomologies. In particular, if M is a G-lattice in K, so is M ∗, hence M ∗ ∨ M and (4.1) Hˆ n(G, M) ≃ Hˆ n(G, M ∗) ≃ DHˆ −n(G, M) ≃ DHˆ n−1(G, DM). 8 YU.DROZD,A.PLAKOSH
The subgroup of periodic elements of K is cyclic and generated by a primi- tive root of unity ζ. Hence, there is an element a ∈ G such that ρ(a)= ζ. Let s oi G = i=1 Ci, where Ci = h ai | ai = 1 i are cyclic groups. We can suppose that a1 = a. Set o = o1. Changing he generators ai, we can make ρ(ai) = 1 Q ′ ′ for i 6= 1. Let G = h a2, a3, . . . , as i, so G = C1 × G . Then M ≃ M1 ⊗Z Z, ′ where M1 is M considered as C1-module and Z is the trivial G -module. Note that M G = 0, as ζv = v implies v = 0. Hence Hˆ 0(G, M) = 0. Consider the o−1 k o−1 k trace T = g∈G g = ( k=0 a )( g∈G′ g). Obviously, k=0 ζ = 0, hence −1 m T M = 0. It implies that Hˆ (G, M)= H0(G, M)= M/(ζ −1)M. If o = p P P kP P fore some m, then also o(ζ) = p for some k, whence NK/Q(1 − ζ) = p [1] −1 and Hˆ (G, M) = H0(G, M) ≃ Z/pZ. If o(ζ) is not a degree of a prime ˆ −1 number, then NK/Q(1 − ζ) = 1 and H (G, M) = H0(G, M) = 0 (it also follows from Corollary 3.4).. Let a finite abelian group G be a direct product G1 × G2 and the orders of G1 and G2 be coprime. If Ki (i = 1, 2) is a cyclotomic field arising from a simple QGi-module, then K = K1 ⊗Q K2 is again a field, hence a simple QG-module, and all simple QG-modules arise in this way. If Mi (i = 1, 2) is a Gi-lattice in Ki, then M = M1⊗ZM2 is a G-lattice in K, unique up to genus. n Corollary 3.4 shows that Hˆ (G, M) = 0 if neither M1 nor M2 is trivial. If M1 n n is non-trivial and M2 is trivial, then Hˆ (G, M) ≃ Hˆ (G1, M1), and if both n n n M1 and M2 are trivial, then Hˆ (G, M) ≃ Hˆ (G1, Z) ⊕ Hˆ (G2, Z). Thus we only need to consider the case of p-groups. Note also that T = p Tp and T is the quasicyclic p-group, i.e. the direct limit lim Z/pmZ with p −→m L respect to the natural embeddings Z/pm Z → Z/pm+1Z . Hence, if M is finitely generated, DM ≃ p DMp, where DpM = HomZ(M, Tp). If M is a lattice, the additive group of DpM is a direct product of several copies of L n n Tp. Moreover, if G is a p-group, Hˆ (G, DqM)=0and DqHˆ (G, M)=0for q 6= p, so we can always replace D by Dp in all formulae from Proposition 3.2. s mk So, let G = k=1 Gk, where Gk is a cyclic group of order p . We calculate cohomologies of a non-trivial irreducible G-lattices. Actually, it is easier to calculateQ homologies.
Theorem 4.1. Let M be a non-trivial irreducible G-lattice. Then ν(n,s) Hn(G, M) ≃ (Z/pZ) , where
n −s (4.2) ν(n,s) = (−1)n . i Xi=0 COHOMOLOGIESOFFINITEABELIANGROUPS 9
Note that for fixed n the value of ν(n,s) is a polynomial of degree n with respect to s with the leading coefficient (n!)−1. For instance, ν(0,s) = 1, ν(1,s)= s − 1, s2 + s + 2 ν(2,s)= , 2 s3 + 5s − 6 ν(3,s)= . 6
′ ′ s−1 Proof. We consider G as a direct product G × Gs, where G = i=1 Gi, and suppose that Gs acts trivially on M. Then M can be considered as the ′ ′ ′ Q outer tensor product M ×Z Z, where M = M considered as G -module and Z is considered as trivial Gs-module. Then we can use the K¨unneth formula [2, Corollary V.5.8]:
n ′ ′ (4.3) Hn(G, M) ≃ Hi(G , M ) ⊗Z Hn−i(Gs, Z) ⊕ Mi=0 n−1 Z ′ ′ ⊕ Tor1 (Hi(G , M ),Hn−i−1(Gs, Z)) . i=0 M Recall that, for a cyclic group C = Z/pmZ,
H0(C, Z)= Z; Z/pmZ if n is odd, Hn(C, Z)= (0 if n is even; while for a non-trivial irreducible lattice M
Z/pZ if n is even, Hn(C, M)= (0 if n is odd, that is
1 if n is even, ν(n, 1) = (0 if n is odd.
Moreover,
H0(G, M)= Z/pZ, that is
ν(0,s) = 1. 10 YU.DROZD,A.PLAKOSH
Thus (4.1) is valid for n = 0 and for s = 1, the minimal values of n and s. Therefore, the K¨unneth formula implies that Hn(G, M) ≃ (Z/pZ)ν(n,s) for some ν(n,s). Moreover, it implies that n ν(n,s)= k=0 ν(n,s − 1) = ν(n,s − 1) + ν(n − 1,s) Hence we can prove (4.1)P by induction, supposing that it is true for ν(n,s−1) and ν(n − 1,s). Then we have ν(n,s)= ν(n,s − 1) + ν(n − 1,s)= n n−1 −s + 1 −s = (−1)n − (−1)n = i i i=0 i=0 Xn X −s + 1 −s = (−1)n − = i i − 1 i=0 Xn −s = (−1)n . i Xi=0 −1 0 Note that in this case Hˆ (G, M)= H0(G, M) and Hˆ (G, M) = 0. The formulae (4.1) and (4.2) give the following result. Corollary 4.2. If M is a non-trivial irreducible G-lattice, then Hˆ n(G, M) ≃ Hˆ n−1(G, DM) ≃ (Z/pZ)ν(|n|−1,s). Analogous calculations give the known result for the trivial G-module Z (cf. [6, 7]).
Theorem 4.3. If n 6= 0 and m1 ≥ m2 ≥···≥ ms, then n ˆ n s mk ν(|n|−1,k)+(−1) (4.4) H (G, Z) ≃ k=1(Z/p Z) . ˆ 0 m s Recall that H (G, Z) ≃ Z/p LZ, where m = k=1 mk.
Proof. First of all, the K¨unneth formula (4.3)P implies that Hn(G, Z) is a direct sum of µ(n,s) cyclic groups so that n µ(n,s)= µ(i, s − 1) + ε, Xi=1 where
1 if n is odd, ε = (0 if n is even, whence
µ(n,s)= µ(n,s − 1) + µ(n − 1,s) − (−1)n−1. COHOMOLOGIES OF FINITE ABELIAN GROUPS 11
Using inductiion by s, we obtain that
µ(n,s)= ν(n,s) + (−1)n, hence
µ(n,s)= µ(n,s − 1) + ν(n − 1,s).
i ms Note that all groups H (Gs, Z) are of period p . Therefore, by (4.3),
′ ms r Hn(G, Z) ≃ Hn(G , Z) ⊕ (Z/p Z) for some r. Together with the formula for µ(n,s), it gives that
n ′ ms ν(n−1,s)−(−1) Hn(G, Z) ≃ Hn(G , Z) ⊕ (Z/p Z) .
By induction, we obtain that
n s mk ν(n−1,k)−(−1) Hn(G, Z) ≃ k=1(Z/p Z) . In view of (4.1), it is justL the formula (4.4).
5. Explicit formulae In this section we find explicit formulae for crossed homomoprhisms (el- ements of H1(G, M)) and cocycles (elements of H2(G, M)) for irreducible latticies and their duals (the latter are important, for instance, in study of Chernikov groups see [5]). We use the resolution defined in Section 1. m ms p i Let G = i=1 Gi, where Gi = h ai | ai = 1 i is a cyclic group of order mi oi = p . We set si = sa . For a cochain µ : Pn → M we denote by Q i ∂µ its coboundary, that is the composition µdn+1 : Pn+1 → M. Then, if ξ : P1 → M, i < j, 2 ∂ξ(x )= siξ(xi), (5.1) i ∂ξ(xixj) = (ai − 1)ξ(xj) − (aj − 1)ξ(xi). Thus ξ is a cocycle if and only if
siξ(xi) = 0 for all i, (5.2) (ai − 1)ξ(xj) = (aj − 1)ξ(xi) for all i 6= j.
If γ : P2 → M, i Thus γ is a cocycle if and only if 2 (ai − 1)γ(xi ) = 0 for all i, 2 siγ(xixj)= −(aj − 1)γ(x ), (5.3) i 2 sjγ(xjxi) = (ai − 1)γ(xj ), (aj − 1)γ(xixk) = (ai − 1)γ(xj xk) + (ak − 1)γ(xixj). Finally, if we identify an element u ∈ M with the homomorphism P0 → M which maps a to au, then ∂u(xi) = (ai − 1)u. mi First suppose that M = Z. Then the element si acts on M as p and the 1 formulae (5.2) show that H (G, Z) = 0. As ai − 1 acts as 0, the formulae (5.3) mean that γ is a cocycle if and only if γ(xixj) = 0. The formulae 2 mi (5.1) imply that, adding a coboundary, we can reduce γ(xi ) modulo p . 2 ms mi Therefore, H (G, Z) ≃ i=1 Z/p Z and generators of this group can be chosen as the cohomology classes of the cocycles γk : P2 → Z such that L 2 γk(xixj) = 0 for all i, j and γk(xi )= δik. For the dual module DpZ = Tp, the formulae (5.2) mean that ξ is a co- mi 1 ms cylce if and only if p ξ(xi) = 0. Hence H (G, Tp) ≃ i=1 Tmi , where mi mi Tmi = { u ∈ Tp | p u = 0 } (it is a cyclic group of order p ). As Tp is divisible, the formulae (5.1) imply that, adding a cobounL dary to a 2- 2 dimensional cocycle γ, one can always make γ(xi ) = 0. Then the for- mij mulae (5.3) mean that p γxixj = 0, where mij = min{mi,mj}. Hence 2 mij H (G, Tp) ≃ i 1 P s−1 s−1 ξ(xi) = λiu0, where λi ∈ Z/pZ. Hence H (G, DpM) ≃ 1 ≃ (Z/pZ) . In the same way, adding a coboundary to a 2-cocycle γ, we can make 2 γ(x1xi) = 0 for i > 1. Then the conditions (5.3) give (ζ − 1)γ(xi ) = 0 2 for all i, whence γ(xi ) = λiu0 (λi ∈ Z/pZ), and (ζ − 1)γ(xixj) = 0 for 2 1 Institute of Mathematics, National Academy of Sciences of Ukraine, Tereschenkivska 3, 01601 Kyiv, Ukraine E-mail address: [email protected], [email protected], [email protected] URL: www.imath.kiev.ua/∼drozd