• No se han encontrado resultados

Computing large direct products of free groups in integral group rings

N/A
N/A
Protected

Academic year: 2023

Share "Computing large direct products of free groups in integral group rings"

Copied!
16
0
0

Texto completo

(1)

Computing large direct products of free groups in integral group rings



Angeldel Roand ManuelRuiz



Abstract

We construct explicitly a subgroup of nite minimalindex and min- imal rank inZGwhich is a direct products of free groups for each nite groupGfor which this is possible.

1 Introduction and preliminaries

Let G be a nite group. If G is abelian then the structure of the group of unitsZGof the integral group ringZGis well known by the work of Higman, Bass and Milnor. IfG is non abelian then generators of subgroups of nite index ofZG have been found for a large class of groups, (see e.g. [12] and [4]). However the structure of ZG(or of a subgroup of nite index) is not known except for some few groups. Some of the known cases appeared in [1], [2], [3], [7], [8] and [10].

In a series of three papers ([5], [9] and [6]) Leal, Jespers and the rst author characterized the groups G such that ZG contains a subgroup of nite index which is a direct product of free groups. In those papers the existence of such a large subgroup of ZG is proved theoretically. Let G be a group satisfying the mentioned property. The aim of these paper is to construct explicitly a subgroup of ZG which is a direct product of free groups and of minimal index in ZG.

We start with some notation:

The cyclic, dihedral and quaternion groups of order m are denoted by Cm,Dm and Qm respectively.

Let G be a group andR a ring. Then R denotes the group of units of Rand RGthe group ring of G with coecients in R. We refer to [12] for



The rstauthorhasb eenpartially supp orted bytheD.G.I.ofSpain andFundacion

SenecaofMurcia

(2)

notational matters concerning group rings. For any nite subsetH ofGlet Hb = jH1j P

h2Hh 2 QG. More generally, if : H ! Zis a map, then we set Hc = jH1j P

h2H (h)h. If g2G, then set bg=hcgi. The augmentation map is denoted by! :ZG! Z. If H is a normal subgroup of G then4(G;H) = Ker!H is the augmentation ideal modulo H. We set 4(G) = 4(G;G).

The rank ofG, denoted byr(G), is the minimum of the cardinalities of the generating subsets ofG. Set G = Hom(G;Z) and let inv :G! Gdenote the map given byg7!g?1.

If K is another group and ':G !Aut(K) is a group homomorphism, thenKo'Gdenotes the corresponding semidirect product.

The non abelian nite groups G such that ZGcontains a subgroup of nite index which is a direct product of free groups, are of the form G = HZ where Z is an elementary abelian 2-group and H is of one of the following types (see [6]):

(a) hx;yjx4 =y4= [x2;y] = [x;y2] = [x;[x;y]] = [y;[x;y]] = 1i,

(b) hx;y1;::: ;ynjx4 =yi2 = [yi;yj] = [x2;yi] = [[x;yi];yj] = [[x;yi];x] = 1i,

(c) hx;y1;::: ;ynjx4 =yi4=yi2[x;yi] = [yi;yj] = [x2;yi] = [yi2;x] = 1i, (d) hx;y1;::: ;ynjx2 =yi2= [yi;yj] = [[x;yi];yj] = [x;yi]2 = 1i,

(e) hx;y1;::: ;ynjx2 =yi4=yi2[x;yi] = [yi;yj] = [[x;yi];x] = 1i, (f) hx;y1;::: ;ynjx4 =yi4=x2y12=yi2[x;yi] = [yi;yj] = [yi2;x] = 1i, (g) hx;y1;::: ;ynjx4 =x2y4i =yi2[x;yi] = [yi;yj] = 1i,

(h) Uohxi where U is an elementary abelian 3-group, x has order 2 or 4 and(x) = inv.

(i) U oK where U is an elementary abelian 3-group, K =hx;yi= Q8 and(x) =(y) = inv.

All throughout this paperG=HZ whereZ is an elementary abelian 2-group and H is of one of the types (a)-(i). The letter n is reserved to denote either the number ofy's for the groups of types (b)-(g) or the rank ofU for the groups of type (h) and (i). The rank ofZ is always denoted by k.

(3)

Our aim is to nd a concrete subgroupF =F0Qni=1Fiof nite index of

ZG, such thatF0 is free abelian andFi is free nonabelian and optimal in the sense that the index ofF inZGis minimal among all the possible subgroups ofZGwhich are a direct product of free groups . There is a natural way to obtain such a group if we do not impose the optimal condition. For every primitive central idempotent eofQGone of the following conditions holds:

(A) QGeis isomorphic to Q, an imaginary quadratic extension ofQ or a totally de nite quaternion algebra overQ.

(B) QGeis a totally de nite quaternion algebra over a real quadratic ex- tension ofQ.

(C) QGeis isomorphic to M2(Q).

LetA,B andC be the sets of primitive central idempotents ofQGof types (A), (B) and (C) respectively and I = A [B [C. For every e 2 I, let Oe be an order in QGe. Then Oe is nite if e2A, virtually in nite cyclic if e 2 B and virtually free nonabelian if e 2 C. Since O = Qe2IOe and

ZGare orders in QG, then O and ZGare commensurable, that is O\ZG has nite index in both. Therefore ZG\(1 +QGe) contains a subgroup of nite index Fe which is trivial if e 2A, in nite cyclic if e2 B and free nonabelian ife2C. ThenQe2IFe is a subgroup of nite index ofZGwith the desired structure. However this does not give information on how big the constructed group is. Surprisingly what we are going to prove in this paper is that this naive approach provides the optimal subgroup in almost all the cases. Our rst theorem is:

Theorem 1.1

Let G= HZ, where Z is an elementary 2-group and H is one of the types (a)-(i). Let B and C be the sets of primitive central idempotents of types (B) and (C) respectively. Then

1. F0 =ZG\(1 +QGfB), wherefB = P

f2Bf, is free abelian of rank jBj 2. If G6=D6;D8, then for all e2C

Fe=ZG\(1 +QGe)

is free non abelian and all the Fe's have the same rank.

3. F =F0 Q

e2CFe has nite index inZG

(4)

So the groupF =F0 Q

e2CFeof Theorem 1.1 has the desired properties. We prove Theorem 1.1 in Section 2. Besides we compute the ranks ofF0 and all theFe's for all the groups. Moreover every Fe is isomorphic to a subgroup of the modular group SL2(Z) that we compute in Proposition 2.1. Using that it is theoretically easy to compute generators for all the Fe's. Also generators ofF0 are easy to compute by using Proposition 2.3. So there is a method to obtain generators forF.

In Section 3 we prove that the group F computed in Section 2 is the best possible. Explicity we prove the following Theorem:

Theorem 1.2

Let G = HZ and F = F0  Q

e2CFe as in Theorem 1.1.

IfE =E0Qj2JEj is a subgroup of nite index of ZG, where E0 is free abelian and Ej is free non abelian for every j. Then

1. r(E0) =jBjand jCj=jJj.

Besides, if G6=D8;D6;Q8 and Q16 then 2. [ZG:F][ZG:E].

3. r(Ej)r(Fe) for every e2C and j2J.

Note that the exceptional groups D8, D6, Q12 and Q16 belong to the list of groups we are considering with the following parameters: k = 0 and n= 1 for all of them; then D8 is of type (d) or (e),D6 andQ12 are of type (h) withxof order 2 and 4 respectively and nally Q16is of type (g). These four groups have been studied separately in several papers and satisfactory results can be found in [3] forD8; [7] forD6; [10] forQ12 and [8] forQ16.

2 Large subgroups: Proof of Theorem 1.1

The statement (3) of Theorem 1.1 follows by the arguments in Section 1.

The subsequent Proposition 2.1 implies (2) and Proposition 2.3 implies (1).

For every a;b2Zset

?(b) = SL2(Z)\



1 +b



Z Z

Z Z

 

;

a(b) = SL2(Z)\



1 +b Z aZ

Z Z

 

and

a(b) = SL2(Z)\



1 +b

 aZ Z aZ aZ

 

:

(5)

If ? is a subgroup of SL2(Z), then b? denotes the image of ? in PSL2(Z).

Proposition 2.1

Letebe a primitive central idempotent ofQGof type (C).

1. Assume that H is of type (a)-(g). Then Ge ' D8 and there are a;b2 Gesuch that fe;a;b;abg is an integral basis of ZGe. Furthermore if e = 0e+ 1a+ 2b+ 3ab, with 0; 1; 2; 3 2 Zand 2 ZG, then

!( ) 0+ 1+ 2+ 3mod 2.

(i) If H is of type (a) then e=hx\2;y2i(1?cG0)Zc for some non trivial

2hx2;y2i and some 2Z. Moreover Fe '2(2k+3), so that Fe

is free of rank1 + 23k+6,

(ii) Assume thatHis of type (b)-(f). For everyi= 1;::: ;nletti = [x;yi].

Let:G0!Gbe given by (ti11ti22tinn) =y1i1y2i2yinn. Then e=eS;'; =cx2(Sb?cG0)\(S)'Zc

where S is a maximal subgroup of G0, 2 Z and ':(S)! Zis a map such that '2S. Moreover, if H is of type (f), then t12S. IfH is of type (b) or (c), then Fe '2(22n+k) which is free of rank 1 + 26n+3(k?1).

If H is of type (d)-(f) then Fe '2(22n+k?1). which is free of rank 1+ 26n+3(k?2), unlessH is of either type (d) or (e),n= 1 and k= 0, or equivalentlyG'D8.

(iii) Assume thatHis of type (g). For everyi= 1;::: ;nletti be the image of [x;yi] in G=hx2i. Let :G=hx2i !G be given by (ti11ti22tinn) = yi11y2i2ynin. Then

e=eS;'; = (Sb?Gc0)\(S)'Zc

where  : G ! G=hx2iis the projection, S is a maximal subgroup of G0 containinghx2i, 2Z and ':(S)!Zis a map such that '2(S=hx2i). MoreoverFe'2(22n+k), so that Fe is free of rank 1 + 26n+3(k?1)

2. Assume that H is of type (h) or (i). ThenGe'D6 and there exist a;b2Ge such that

ZG(1?a)e=Ze(1?a)Za(1?a)Zb(1?a)Zab(1?a):

(6)

(i) IfHis of type (h) withxof order 2 thene=eS = (Sb?Ub)Zc whereS is a maximal subgroup of U and 2Z. Moreover Fe '3(2k3n?1) which is free unless k = 0 and n = 1, that is to say, G = S3. So if G 6= S3 then the rank of Fe is 33n?4 if k = 0, and 23k?133n?4 otherwise.

(ii) IfHis of type (h) with xof order 4 thene=eS =cx2(Sb?Ub)Zc where S is a maximal subgroup of U and 2Z. Then Fe '3(2k+13n?1) which is free of rank 1 + 23k+233(n?1).

(iii) IfH is of type (i), thene=eS; ;= (Sb?Gc0)Zch\xyiwhereSis a max- imal subgroup ofU, 2Z and 2hxyi. ThenFe'3(2k+23n?1), which is free of rank 1 + 23k+533(n?1).

Proof.

The proof is done separately for each case. The rst step in each case is to identify the elements ofC. We leave this part to the reader. Then for eache2C one has to make a good selection of aand b. This selection induces a ring isomorphism:QGe!M2(Q) and using this isomorphism one identi es (Fe) which happen to be the free subgroup of the modular group in the proposition. Cases (a)-(g) are very similar. We only do case (c) and let the reader check that similar arguments work for the remaining cases. Cases (h) and (i) are also similar so we only do case (i).

Assume that H is of type (c) and let e=eS;'; 2C as in (1.ii). Then HS =hx2;G0;(S);Ziis a normal subgroup of index 4 ofGand ift2G0nS, then 1;x;y = (t);xy is a transversal of G modulo HS. Set a = xe and b=ye. Every element 2ZGcan be written as

= 0+ 1x+ 2y+ 3xy

where i 2 ZGand Supp( i)  HS. Moreover if h 2 HS, then he = 1 and therefore e = 0e+ 1a+ 2b + 3ab for some 0; 1; 2; 3 2 Z and !( ) = P3i=0!( i)  0 + 1 + 2 + 3mod 2. It is easy to see that Ge 'D8 and e;a;b;abis an integral basis of ZGe. Then there is an isomorphism:QGe!M2(Q) given by

( e) =

 0? 1+ 2? 3 2( 2? 1) 1+ 3 0+ 1? 2+ 3



(see [3] or [12]). Since the support of e is HS, ie 2 ZG for every i and considering the coecient of 1 in this element one deduce that 22n+k j i. Let i= i=22n+k. Then = 1 + 22n+k( 0+ 1x+ 2y+ 3xy)e, so that

( ) = 1 + 22n+k

 0? 1+ 2? 3 2( 2? 1) 1+ 3 0+ 1? 2+ 3



:

(7)

Therefore ( ) = 1 + 22n+k

 a 2b c d



with a  dmod 2. Note that this implies that the determinant of ( ) is 1. Thus (Fe)  2(22n+k) and by solving a system of linear equation one can easily verify that the equal- ity holds. Moreover 2(22n+k) is a subgroup of index 2 of ?(22n+k) and

?(22n+k)'?(2\2n+k) is a subgroup of index 26n+3k?4of?(2) (see [11]). Sinced the last group is free of rank 2 then 2(22n+k) is free of rank 1+26n+3(k?1). Now assume that H is of type (i). Then e = eS; ; as in (2.iii). Set a = ue and b = xe where u 2 U nS. Then HS = hxy;S;Zi is a normal subgroup ofGand 1;u;u2;x;xu;xu2is a right transversal of GmoduloHS. For the rst part we argue as in the previous case by noticing thathe=e for everyh2HSand (1+u+u2)e= 0. On the other handZGe=ZG(1?a)e so that (1?a)e;a(1?a)e;b(1?a)e;ba(1?a) is an integral basis of ZGe. Moreover, as in [12] or [7], there is an isomorphism :QGe! M2(Q) that associates 0(1?a)e+ 1(1?a)e+ 2b(1?a)e+ 3ba(1?a)ewith

 3( 0? 1? 2) ? 0+ 2 1+ 2 2? 3 3( 0?2 1? 2? 3) 3( 1+ 2)



Furthermore, if 2 ZG\(1 +QGe) = Fe then ?1 2 (G;hui)e=

ZG(1?a)eand

= 1 + ( ?1)e= 1 + ( 0(1?a) + 1a(1?a)e+ 2b(1?a) + 3ba(1?a)e with i 2 Z. Unfortunately in this case, unlike the previous one, the sup- ports of the base elements intersect. However this diculty can be overcome as follows. The coecients of 1, u, x and xu in ?1 are 0 = 2k+2 30n?1, 1 = 2k+2 31n?1, 2 = 2k+2 32n?1 and 3 = 2k+2 33n?1, respectively. Thus

( ) = 1 + 2k+23n?1

 3( 0? 1? 2) ? 0+ 2 1+ 2 2? 3 3( 0?2 1? 2? 3) 3( 1+ 2)



so that by similar arguments as in the previous paragraph one shows that

(Fe) = 3(2k+23n?1). The ranks can be computed as in the previous case.

The classical quaternion algebra over an arbitrary ring Ris denoted by

H(R).

In order to prove statement (3) of Theorem 1.1 we need information about the central idempotents of QG of type (B). This is the role of next Lemma. The proof is straightforward.

(8)

Lemma 2.2

1. If G is not of type (g) or (i) thenB =;. 2. Assume that G is of type (g) and let

K=hy12y22;::: ;y12yn2;Zi=fy21i1y22i2:::y2nin :Xn

t=1it0 mod 2gZ:

For every 2K and every2in, let n=(y12yi2), K() =hy12y2;::: ;y1nyn;Zi and T=fyP1 nj=2jijyi22:::ynin : 0ij 1g:

ThenT is a transversal ofhx2;K()imodulo K. Moreover each element of B is of the form

f =f;= (1?cx2)K[();

for some 2K and 2K() so that KKer and ;(kt) =(k)(t) for everyk2K and t2T. Furthermore

Gf;=ha=xf;b=y1f ja2 =b4;b8= 1;aba?1=b?1i'Q16;

B = ff;b;b2;b3;a;ab;ab2;ab3g is an integral basis of ZGf and the map

:QGf !H(Q[p2]) given by

(Pg2B gg) = f +p22( b? b3) + ( a+p22( ab? ab3))i+ ( b2 +p22( b+ b3))j+ ( ab2+p22( ab+ ab3))k is a ring isomorphism.

3. Assume that Gis of type (i). Then every element of B is of the form fS; = (1?cx2)(Sb?Ub)Zc

where S is a maximal subgroup of U and 2 Z Moreover GfS; ' Q12 and if X = xf, Y = yf and W = wf, with w 2 U n S, then B =

ff;X;Y;XY;W;WX;WY;WXYg is an integral basis ofZGf and the map

:QGf !H(Q[p3]) given by

(Pg2B gg) = 12[2 1? W +p3 XY W + (2 XY ? XY W ?p3 W)i (2 X ? XW ?p3 Y W)j+ (2 Y ? Y W +p23 XW)k] is a ring isomorphism.

(9)

Next proposition provides a proof of statement (3) of Theorem 1.1 and also computes the rank ofF0.

Proposition 2.3

The group F0 is embedded in the centre of QG. IfH is of type (g) then F0 is free abelian of rank 22n+k?2; if H is of type (i) then F0 is free abelian of rank 2k?1(3n?1). In the remaining cases F0 = 1.

Proof.

By Lemma 2.2, we may assume thatGis of type (g) or (i). By the comments prior to Theorem 1.1, it is enough to show that for everyf 2B, F0f is embedded in the centre ofQGf and realizing that the rank claimed forF0coincides with the cardinality ofB that can be easily computed using Lemma 2.2.

Assume that H is of type (g). Then fB = 1?cx2. If 2 F0 then ?1 2 (G;hx2i) and hence ( ?1)fB  0 mod 2, in ZGfB. Therefore, if f 2B, then ( ?1)f 0 mod 2, in ZGf and, by Lemma 2.2, ( ) is a unit ofH(Z[p2]), whereis the isomorphism of Lemma 2.2. Using that all the units ofH(Z[p2]) are of the form u;ui;ujoruk, where u2Z[p2], one easily deduces that( )2Z[p2] and hence is central.

Assume now thatGis of type (i). ThenfB= (1?cx2)(1?Ub). Therefore, F0 (G;hX2i)\(G;U) and we argue as in the previous case.

3 Optimality: Proof of Theorem 1.2

By Theorem [12, Theorem 30.1] the following is a torsionfree normal com- plement of the trivial units ofZG:

V =ZG\(1 + (G)(G;G0))

It is well known thatV is free nonabelian if G=D8 orG=D6. Using this fact and Proposition 2.1, it is easy to prove:

Lemma 3.1

For everye2C, V e is torsionfree.

Set

Fe0 =fu2ZG:ue=efor everye2Cg and for everye2C, let

Fee =fu2ZG:uf =f for allf 2Cnfegg: Plainly Fe0 =Fee1\Fee2 for every two di erente1 and e2 in C.

(10)

The group of type (h) with n= 1,k= 0 and x of order 4 is denoted by C3oC4.

The crux of our argument relies on the following technical lemma.

Lemma 3.2

IfG6=C3oC4 andG6=Q16, then for every e2C, Fee\V  F0Fe and Fe0\V F0.

Proof.

Lete2C.

Claim 1. If f is a primitive central idempotent, such thatQGf is com- mutative or isomorphic to H(Q), then V f =f.

If QGf is commutative, then f(1?cG0) = 0 and the claim follows. If

QGf ' H(Q) then Gf ' Q8 = ha;bi and T = ff;a;b;abg is a rational basis of QGf. Moreover for every g 2 G0, gf = f. This implies that every element of (G)(G;G0)f is of the form 2 f for some 2 (G).

Moreover f = Pt2T tt, where t 2 ZW, W being the kernel of the canonical map G ! Gf ! Gf=ha2i. Then tf = tf where t 2 Zand t  !( t) mod 2. Therefore Pt2T t  !( ) = 0 mod 2. Thus, if u 2 V, thenuf = 1 + 2Pt2T tt, withPt t even. Since the unique units ofH(Z) are1,i,j and ij, we conclude that u=f. This proves Claim 1.

Claim 2. If G6=Q12 andf 2A, then (Fee\V)f =f.

By Claim 1, we may assume that QGf is not commutative andQGf 6=

H(Q). This implies that H is of type (h) with x of order 4 and QGf is isomorphic to the generalized quaternion algebraA=Q[i;j:i2 =?1;j2=

?3;ij=k=?ji] (see [9]). By Proposition 2.1 and [9]

e=eS; =cx2(Sb?Ub)Zc and f =fS1; 1 = (1?cx2)(cS1?Ub)Zc;

where S and S1 are two maximal subgroups of U and ; 1 2 Z. Fix y 2U nS1. Then B1 =ff;a=xf;b=yf;abg is an integral basis of ZGf and there is an isomorphism:QGf !Aso that(a) =iand(b) = 1+2j. Therefore(ZGf) is the subring ofAgenerated by iand 1+2j. Furthermore the only units of this ring are1,i,1?2j,1+2j,i1?2j and i1+2j [10].

Letu2V \Fee. We have to prove that uf =f.

Assume rst that (S; ) 6= (S1; 1). Let e0 = eS1; 1. Then B2 =

fe0;xe0;ye0;xye0g is an integral basis of ZGe0. Moreover for an 2 ZG the coecients of f and e0in the basisB1 and B2 are pairwise congruent modulo 2. Since (u?1)e0= 0, the coecients of (u?1)f in the basis B1 are even. Therefore uf = 1 + 2(Pt2B1 tt) is a unit in ZGf. By the previous

(11)

paragraphuf =f. Ifuf =?f, then (u?1)f =?2f. Sinceu2V, we can write (u?1)f = ( 0+ 1x)1?2x2cS1Zd1 where 0; 12ZGand its support is embedded in a xed transversal ofGmodulo hx2;S1;Zicontaining 1. Then the coecient of 1 in (u?1)f is 2k+1 3m?1, where is the coecient of 1 in 0. However the coecient of?2f is?2k13m, which yields to a contradiction.

Thusuf =f as desired.

Now assume that S = S1. By Claim 1 and the previous paragraph we have

u?1 = (u?1)(fS; +eS; ) = (u?1)(Sb?Ub)Zc =X3

i=0 ixi(Sb?Ub)Zc with i=P2j=0 ijyj 2hyi. The coecient ofxiyj in u?1 is

3 ij ? i0? i1? i2

2k3n = ij

2k3n?1 2Z:

Therefore 2k3n?1 j ij and hence uf 2 f + 2k3n?1ZGf is a unit of ZGf. If n > 1, then uf = f. Assume now that n = 1. Since we are assuming that G6=C3oC4, then k 1 and we argue as at the end of the previous paragraph to prove thatuf =f. This nishes the proof of Claim 2.

Now we prove thatFe0\V F0. By Claim 1, the result is obvious unless H is of type (h) and the order of x is 4. In this case if G6=Q12, then the cardinality ofC is greater than 2 and the result is a consequence of Claim 2. Now we prove Fee \V  F0 Fe for every e 2 C. If H is neither of type (g) or (i), thenB =;and hence the Lemma is a direct consequence of Claim 2.

Assume now that H is of type (g) and G 6= Q16, so that either n > 1 or k > 0. The sum of the elements of B is fB = 1?cx2. Without lost of generality one may assume that

e=Rb(Sb?Gc0)Zc

where R=hy2;::: ;yni,S =hy22;::: ;yn2i and 2Z. The support of e is L=hy12;y2;::: ;yniZandTe=f1;x;y1;xy1gis a transversal ofGmodulo L.

Let u 2 V \Fee and = u?1. If l 2 L, then le = e. Therefore, e= 0e with 0 2 ZG, Supp( 0)  Te and !( 0)  !( ) = 0 mod 2. By Claim 2,u= 1+ 0e+ fB. It is enough to prove that g? gx2 is even for

(12)

everyg2G. Indeed, in this case fB 2ZGand sou= (1 + 0e)(1 + fB), 1 + 0e2Fe and 1 + fB2F0.

Let g2 Gand t 2Te so thatg tmodL. Then the coecient of g in u?1 is



0t

22n+k + g? gx2

2 :

Therefore 0t= 22n+k?1 t for some t2Zand g = g? x2g  tmod 2.

So it is enough to show that t is even for everyt2Te.

To obtain our goal we are going to consider the image of uf under the isomorphism=f :QGf 'H(Q(p2)), given in Lemma 2.2, for everyf 2 B. Recall that an element ofB is of the formf =f;= (1?cx2)K\(); as in Lemma 2.2. We use all the notation of that lemma. Then f =Pb2B bb where

b= X

k2K;(k)( bk? bkx2):

Note thatK()\Lhas index 2 inK() and hence the cardinality ofK()\L is 22(n?1)+k. Since either n > 1 or k > 0, these cardinality is even. Since g1x2 ? g1x2  g2x2 ? g2x2 mod 2 if g1 g2 modL,

t= X

k2K()\L;(k)( th? thx2) + X

k2K()nL;(k)( th? thx2) is even. By Lemma 2.2,(ZGf)H(Z[p2]). Since every unit ofH(Z[p2]) is of the formu,ui,uj orukwithu a unit ofZ[p2] then by Lemma 2.2 we deduce that

a= b2 = ab2 = ab= ab3 = b+ b3 = 0:

Writing these equations in terms of the 's we obtain a system of linear

equations X

h2K();(h)( th? thx2) = 0;

for everyt=x;y12;xy21;xy1;xy31. For a xed 2Kandk2Kthe previous system of linear equations becomes

X

t12T(t1)X

k2K(k)( tt1k? tt1kx2) = 0:

But the matrix ((t1))t12T;2(K()=K) is a Hadamard matrix, that is a matrix of 1's and ?1's with orthogonal rows. In particular its determinant

Referencias

Documento similar

Abstract: We extend the classical Stallings theory (describing subgroups of free groups as automata) to direct products of free and abelian groups: after introducing enriched

No obstante, como esta enfermedad afecta a cada persona de manera diferente, no todas las opciones de cuidado y tratamiento pueden ser apropiadas para cada individuo.. La forma

The second case is a non-trivial generalization of the formalism consisting in substituting the structure group U(1) of phase invariance by a bigger group T (to account for

1 Instituto de F´ısica Te´orica y Computacional Carlos I, Facultad de Ciencias, Universidad de Granada, Campus de Fuentenueva, Granada 18002, Spain.. 2 IFIC, Centro Mixto Universidad

(Notice that A is of Kleinian type if and only if every simple quotient of A is of Kleinian type.) Then we use this characterization and results from [8] to characterize the

group identities) in the set of symmetric elements (resp. symmetric units), and when these identities are transferred to the whole group ring (resp. unit group). We consider

In this article we construct free groups and subgroups of finite index in the unit group of the integral group ring of a finite non-abelian group G for which every

Polcino Milies, Commutativity of skew symmetric elements in group rings, Proc.. Giambruno, C.Polcino Milies, Unitary units and skew elements in