On the (non-)Contractibility of the Order Complex of the Coset Poset of an Alternating Group
MASSIMILIANOPATASSINI
ABSTRACT- Let Altkbe the alternating group of degreek. In this paper we prove that the order complex of the coset poset of Altkis non-contractiblefor a big family of k2N, including thenumbers of theformkpmwherem2 f3;. . .;35gand p>k=2. In order to prove this result, we show thatPG( 1) does not vanish, wherePG(s) is the Dirichlet polynomial associated to the groupG. Moreover, we extend the result to some monolithic primitive groups whose socle is a direct product of alternating groups.
MATHEMATICSSUBJECTCLASSIFICATION(2010). 20D30, 20P05, 11M41.
KEYWORDS. Probabilistic zeta functions; simplicial complexes; order complexes;
contractibility; coset posets; alternating groups; Brown conjecture.
1. Introduction
For a finitegroupG, le tC(G) be the set of (right) cosets of all proper subgroups of G, partially ordered by inclusion. Let DD(C(G)) bethe order complex ofC(G), so thek-dimensional faces ofDarechains of lengthk fromC(G). Thestudy ofDwas initiated in a paper of K. S. Brown (see [2]), who attributed therein to S. Bouc the observation that there exists a connection between the reduced Euler characteristic ~x(C(G)) and theDi- richlet polynomialPG(s) ofG, defined by
PG(s)X1
n1
an(G)
ns ; where an(G) X
HG;jG:Hjn
mG(H):
Here mG is theMoÈbius function of thesubgroup latticeof G, which is
(*) Indirizzo dell'A.: Via Rovede, 13, 31020 Vidor (TV), Italy.
E-mail: frapmass@gmail.com
defined inductively bymG(G)1, mG(H) P
K>HmG(K). The counterpart of the Dirichlet polynomial is called the probabilistic zeta function of G (see [1] and [6]).
In particular, Brown ([2], § 3) showed that PG( 1) ~x(C(G)):
It is a well-known fact that ifD(C(G)) is contractible, then its reduced Euler characteristic ~x(C(G)) is zero. Hence, ifPG( 1)60, then the simplicial complex associated to the groupGis non-contractible.
Moreover, in [2], Brown conjectured the following.
CONJECTURE1. If G is a finite group, then PG( 1)60. Hence the order complex of the coset poset of G is non-contractible.
The conjecture was proved for some families of groups, as the following theorem shows.
THEOREM2. Let G be a finite group.
(1) If G is soluble, then PG( 1)60(see[2]).
(2) If G is a classical group which does not contain non-trivial graph automorphisms, then PG( 1)60([9, Theorem 2]).
(3) If G is a Suzuki simple group or a Ree simple group, then PG( 1)60 ([7]).
Moreover, in our PhD thesis (see [8, Theorem 7.1]), we proved Con- jecture 1 for a class of monolithic primitive groups with socle isomorphic to a direct product of copies of a simple classical group with some conditions on the rank and the number of factors of the direct product.
In this paper we show the following.
THEOREM3. Let G be Symk orAltk. Let p be a prime number such thatk
25p5k 2.
(1) If k p35, then PG( 1)60.
(2) If p>2((k p)!)3, then PG( 1)60.
A proof of this result proceeds as follows (the general strategy is the same which we employed in [9]). Let r be a prime number and let f(s) P
n1
an
ns be a Dirichlet polynomial. We denote byf(r)(s) the Dirichlet
polynomial X
(n;r)1
an ns: Letpbe a prime number such thatk
25p5k 2. We have that PG(s)P(p)G(s)X
pjk
ak(G) ks :
The first summandP(p)G (s) collects the contribution given by the subgroups Hof X such thatHcontains a Sylow p-subgroup, which are intransitive subgroups of G (see [10, Lemma 9]). In [10, Proposition 12] an explicit formula for the Dirichlet polynomialP(p)G (s) is given. So, a careful analysis of the valueP(p)G ( 1), shows thatjP(p)G ( 1)jppin many cases (see Section 3), wherejajpis the greatest power ofpthat divides the integer numbera (see Section 2 for a more precise definition).
Sincendividesan(G) forn2N f0g(see Lemma 5), we have that X
pjn
an(G)n pp2; hence
jPG( 1)jp jP(p)G ( 1)X
pjn
an(G)njpp wheneverjP(p)G( 1)jpp. So we concludePG( 1)60.
With a little more work we can obtain a more general result. Let us note that if N is a normal subgroup of a finite group G, then PG(s)PG=N(s)PG;N(s) (see [2]), where
PG;N(s)X
n1
an(G;N)
ns ; with an(G;N) X
HG;jG:Hjn;
NHG;
mG(H):
Thus, if 1N05N15. . .5NlG is a chief series of G, applying the above formula repeatedly, we obtain
PG(s)Yl 1
i0
PG=Ni;Ni1=Ni(s):
If the chief factorNi1=Niis abelian, by a result of [4], we have that PG=Ni;Ni1=Ni(s)1 ci
jNijs
where ci is the number of complements of Ni1=Ni in G=Ni. Hence PG=Ni;Ni1=Ni( 1)60.
If the chief factorNi1=Niis non-abelian, then there exists a monolithic primitive groupLisuch that
P(r)Li;soc(Li)(s)P(r)G=Ni;Ni1=Ni(s)
for each prime divisor r of the order of the socle soc(Li) of Li (see, for example, [3, Proposition 16]).
The above discussion suggests that in order to prove Conjecture 1, we can focus our attention on the monolithic primitive groups. In particular, here we prove the following.
THEOREM 4. Let G be a primitive monolithic group with socle N isomorphic to a direct product of n copies of the alternating groupAltk with k8. Let p be a prime number such that k
25p5k 2. If p>2n((k p)!)2n1, then PG( 1)60.
2. Some useful results and definitions
Letabe an integer and letrbe a prime number. We denote byjajrthe r-part ofa, i.e.jajrriwherei2Nsuch thatridividesabutri1does not dividea. Moreover, we setj0jr0.
Ifbc=dis a rational number for somec2Z;d2N f0g, then we let jbjrjcjr
jdjr.
We record here an important result on the MoÈbius function of the subgroup lattice ofG.
LEMMA5 ([5], Theorem; 4.5). Let A be a finite group and B a subgroup of A. The indexjNA(B):BjdividesmA(B)jA:BA0j.
In particular, if X is an almost simple group with socleGand His a subgroup ofG, then Lemma 5 yieldsjG:HjdividesmG(H)jG:NG(H)j. So, in particular,ndividesan(X;G).
We can say some more words on the Dirichlet polynomial of a mono- lithic primitive groupLwith non-abelian socleN. Assume thatSis a simple component of L, define XNL(S)=CL(S) and n jL:NL(S)j. We have that NSn. SinceSsoc(X), assume that SX. The following result shows a connection between the Dirichlet polynomialsPL;N(s) andPX;S(s).
THEOREM6 (See [11, Theorem 5]).
P(r)L;N(s)P(r)X;S(ns n1) for each prime divisor r of the order of S.
LetXbe Symkor Altk withk8. A key role in our paper is played by the Dirichlet polynomial P(p)X;S(s), which can be expressed by an explicit formula.
PROPOSITION7 (See [10, Proposition 12]). Let k8. Let X be either Altk orSymk, and let p be a prime number such thatk
25p5k 2. Then P(p)X;S(s) X
v2Ck:klp
m(v) n(v)i(v)s 1:
HereCkis the set of partitions of k, i.e. non-decreasing sequences of pos- itive integers whose sum is k, we have v(k1;. . .;kl)2Ck for some lk;k1;. . .;kl2N f0gsuch that k1. . .klk, and we define
m(v)( 1)l 1(l 1)!; i(v) k!
Ql
i1ki!
; n(v)Yk
i1
vi!;
wherevi jfj:kj igj.
3. The Dirichlet polynomialP(p)(s)
Let XSymk or Altk for k8. For this section, we let P(s)PX;soc(X)(s). The aim of this section is to find the valuejP(p)(1 n)jp forn2 a natural number.
Letpbe a prime number such thatk
25p5k 2. By Proposition 7, we have that
P(p)(1 n) X
v2Ck:klp
m(v)
n(v)i(v) n X
v(k1;...;kl)2Ck:klp
( 1)l 1(l 1)!(k!)n Qk
i1vi!(Ql
i1ki!)n
Xk p
j0
g(j;n) k!
(k j)!
n
;
where
g(j;n) X
v(k1;...;kl 1)2Cj
( 1)l 1(l 1)!
Q
i1vi! lQ1
i1ki! n
forj0 (the setC0consists of the empty partitionv( ), sog(0;n)1).
Consider the polynomial fn(x)g(0;n)Xk p
j1
g(j;n) Yj 1
i0
(xk p i)
!n
inQ[x]. Clearlyfn(p)P(p)(1 n). We want to show thatxdividesfn(x) in Q[x], butx2does not dividefn(x) inQ[x]. This implies thatjP(p)(1 n)jpp if the coefficienta1ofxinfn(x) is such thatja1jp1 (see the proof of The- orem 11).
First of all, we prove the following formula, which is very useful in order to find some properties ofg(j;n).
PROPOSITION8. If m1, then Xm
j0
g(j;n) ((m j)!)n0
PROOF. Let us rewrite the sum in another way. Let
v(k1;. . .;kl)2Cm. For each h2 f1;. . .;lg there exists vh
(k1;. . .;kh 1;kh1;. . .;kl)2Cm kh. Clearlyvappears in the termg(m;n) andvhappears in the termg(m kh;n). In particular, the contribution of vh ing(m kh;n) is
( 1)l 1(l 1)!
n(vh) Q
i6hki!n:
Moreover, since n(v) ji:kikhjn(vh) we have that the previous ex- pression becomes
( 1)l 1(l 1)!ji:kikhj n(v) Q
i6hki!n :
Furthermore, it is clear that if (k1;. . .;kl0)t2Cm0 for somem05m, then there exists a unique v2Cm such that tvh for some
h2 f1;. . .;l01g: indeed, there exists h2 f1;. . .;l01g such that kh 1m m0kh(with the convention thatk00 andkl01k) and so v(k1;. . .;kh 1;m m0;kh;. . .;kl0).
Let Hv fh2 f1;. . .;lg:h1 or kh 15khg (so fk1;. . .;klg fkh:h2Hvg and if h1;h22Hv, then kh1kh2 if and only if h1h2).
We get Xm
j0
g(j;n)
((m j)!)n X
(k1;...;kl)v2Cm
( 1)ll!
n(v) Yl
i1
ki!
!nX
h2Hv
( 1)l 1(l 1)!ji:kikhj n(v) Y
i6h
ki!
!n 1 (ki!)n 0
BB BB B@
1 CC CC CA
X
(k1;...;kl)v2Cm
( 1)l(l 1)!(l X
h2Hv
ji:kikhj)
n(v) Yl
i1
ki!
!n 0
since P
h2Hv
ji:kikhj lby definition. p
Thanks to the recursive formula we obtained forg(j;n), we can prove the following inequalities.
PROPOSITION9. We have that
05( 1)m 1g(m 1;n)=25( 1)mg(m;n)( 1)m 1g(m 1;n)1 for m1and n2.
PROOF. Let us prove the proposition by induction onm. Ifm1, the claim holds by definition ofg(m;n). Assume thatm>1. By induction, it is enough to prove that
( 1)mg(m;n)=25( 1)m1g(m1;n)( 1)mg(m;n) By Proposition 8, we have that
g(m1;n) Xm
j0
g(j;n) ((m1 j)!)n; hence
( 1)m1g(m1;n)( 1)mg(m;n)Xm 1
j0
( 1)m j ( 1)jg(j;n) ((m1 j)!)n: (y)
Letg0(j;n)( 1)m j ( 1)jg(j;n)
((m1 j)!)n. Now, by induction we have ( 1)jg(j;n)5 ( 1)j 1g(j 1;n)=250
for 1jm. Hence, form jodd, we get
g0(j;n)g0(j 1;n)5( 1)j 1g(j 1;n)2 (m j2)n 2((m j2)!)n 0 forn1. This implies that
X
m 1
j0
( 1)m j ( 1)jg(j;n)
((m1 j)!)nag0(0;n)
X
1jm 1;m jodd
g0(j;n)g0(j 1;n)50 where a0 if m is even, a1 otherwise (when m is odd, note that g0(0;n)50). So this proves that ( 1)m1g(m1;n)( 1)mg(m;n).
Now, by (y), it remains to prove that ( 1)m1g(m1;n) ( 1)mg(m;n)=2
( 1)mg(m;n)=2Xm 1
j0
( 1)m j ( 1)jg(j;n) ((m1 j)!)n >0:
Again, by induction we have that ( 1)jg(j;n)=25( 1)j1g(j1;n) for 0jm 1. Hence, form jeven, we have that
g0(j;n)g0(j 1;n)>( 1)jg(j;n)(m2 j)n 2 ((m2 j)!)n 0 forn1. Moreover, we get
( 1)mg(m;n)=2g0(m 1;n)>( 1)mg(m;n)2n 2 1 2n 1 0 forn2. This implies that
( 1)mg(m;n)=2Xm 1
j0
( 1)m j ( 1)jg(j;n) ((m1 j)!)n
bg0(0;n)( 1)mg(m;n)=2g0(m 1;n)
X
1jm 2;m jeven
g0(j;n)g0(j 1;n)>0
where b1 if m is even, b0 otherwise (when m is even, note that g0(0;n)>0). So this proves that ( 1)m1g(m1;n)>( 1)mg(m;n)=2. p
Now, we are ready to prove the main result of this section.
PROPOSITION10. Let n2. We have that x divides fn(x)inQ[x], but x2 does not divide fn(x)inQ[x].
Letmk p. To prove thatxdividesfn(x) is enough to show that fn(0)Xm
j0
g(j;n) m!
(m j)!
n
0;
which follows from Proposition 8. In order to prove thatx2does not divide fn(x) we may show that the coefficient ofxinfn(x) does not vanish. It is easy to realize that the coefficient ofxinfn(x) is:
Xm
j0
ng(j;n) m!
(m j)!
nXj 1
i0
1 m i (y) By Proposition 8, we can subtract
Xm
j0
ng(j;n) m!
(m j)!
nmX1
i0
1 m i0
from (y), hence proving that (y) does not vanish is equivalent to prove that X
m 1
j0
g(j;n) m!
(m j)!
nmX1
ij
1 m i60:
Let
ha;b(m;n)Xb
ja
g(j;n) m!
(m j)!
nXm 1
ij
1 m i: Let 0jm 2. Note that
m!
(m j 1)!
n Xm 1
ij1
1
m i2 m!
(m j)!
n Xm 1
ij
1 m i>0 since
(m j)n mX1
ij1
1
m i2Xm 1
ij
1 m i>0:
Hence, by Proposition 9 we get ( 1)j1g(j1;n) m!
(m j 1)!
n Xm 1
ij1
1 m i>
( 1)jg(j;n) m!
(m j)!
nXm 1
ij
1 m i; i.e. ( 1)j1hj;j1(m;n)>0. This implies thath0;m 1(m;n)50 ifmis even, and h3;m 1(m;n)>0 if m is odd. Since g(0;n) g(1;n)1 and g(2;n)1 2 n, it is easy to see thath0;2(m;n)>0 (forn2 andm3).
Thus we conclude that
h0;m 1(m;n)h0;2(m;n)h3;m 1(m;n)>0
ifmis odd. The proof is complete. p
We can finally prove the main theorem.
THEOREM 11. Let k8 and let p be a prime number such that k=25p5k 2. Assume that p>n((k p)!)n1. Then
jP(p)( n1)jpp:
PROOF. Let mk p. Let fn(x)a0a1x. . .atxt for some aiai=bi2Qwith (ai;bi)1 (ifai0, letbi1) andt2N.
By the proof of Proposition 10 we have that a1 nh0;m 1(m;n).
Moreover, we have seen that ( 1)m 1h0;m 1(m;n)>0 and ( 1)m 1h0;m 2(m;n)50, hence jhm 1;m 1(m;n)j>jh0;m 1(m;n)j. So we have
ja1j njhm 1;m 1(m;n)j njg(m 1;n)j(m!)nn(m!)n beingjg(m 1;n)j 1 (by Proposition 9).
Now, we want to show that jP(p)( n1)jpp. Note that since p>k pandbidivides (k p)!m!(by definition offn(x)), we have that jaijp jaijp. By Proposition 10 we have thata00, hencejP(p)( n1)jp jfn(p)jpp. For a contradiction, assume that jP(p)( n1)jp>p. Then jfn(p)jpp2, henceja1jpp, so
ja1j ja1jpp>n(m!)n1 ja1jm! ja1kb1j ja1j;
a contradiction. p
When the numberk pis small, we know explicitly the coefficient ofx infn(x). For example, we have thatjP(p)( n1)jppifpdoes not appear in Table 1 (for k p7 and n2 f2;4;6g). Further calculations can be done, but the coefficient ofxin fn(x) grows very quickly and it has very large prime factors (for example, ifk p35, the coefficient ofxinf2(x) has a prime divisor of 50 digits).
TABLE1. Exceptions forp.
k p n2 n4 n6
3 23 31, 53 109, 617
4 677 1047469 19, 31, 5051
5 71, 103 643, 148579 3889, 595523689
6 7, 13 23, 269, 89714671 2357, 4584299, 35430211 7 863897 181, 2005732476817 70353197, 1633443829219
However, there are many prime numbers betweenk=2 andk 2 ifkis big enough. For example, we have the following.
PROPOSITION12. Let8k106. If k613, then there exists a prime number p such thatjP(p)( 1)jpp.
4. The main result
We can now prove the main result.
THEOREM13. Let G be a monolithic primitive group with socle iso- morphic to Altnk for k8. Let p be a prime number such that k=25p5k 2.
(1) If p>2n((k p)!)2n1, then PG;soc(G)( 1)60.
(2) If n1and k p35, then PG;soc(G)( 1)60.
(3) If n1and8k107, then PG;soc(G)( 1)60.
PROOF. By Theorem 6, we have that
jP(p)G;soc(G)( 1)jp jP(p)X;soc(X)(1 2n)jp;
whereXNG(S)=CG(S) andSis a simple component ofG. By Theorem 11, ifp>2n((k p)!)2n1, thenjP(p)X;soc(X)(1 2n)jpp. By Lemma 5, we have
thatldividesal(G;soc(G)) forl1, hence
jPG;soc(G)( 1) P(p)G;soc(G)( 1)jpp2; thus we get the claim.
Ifn1 and k p35, the coefficient ofx in fn(x) is known, so by direct computation, there exists a prime number p0 (not necessarily dif- ferent from p) such that k=25p05k 2 such that jP(pX;soc(X)0) ( 1)jp0p0 (except whenk13). Arguing as above, we get the claim (using a direct computation for k13). Finally, if n1 and 8k106, arguing as
above, by Proposition 12 we have the claim. p
REFERENCES
[1] N. BOSTON, A probabilistic generalization of the Riemann zeta function, Analytic Number Theory,1(1996), pp. 155-162.
[2] K. S. BROWN,The coset poset and the probabilistic zeta function of a finite group, J. Algebra,225(2000), pp. 989-1012.
[3] E. DETOMI- A. LUCCHINI,Crowns and factorization of the probabilistic zeta function of a finite group, J. Algebra,265(2003), pp. 651-668.
[4] W. GASCHUÈTZ, Zu einem von B. H. und H. Neumann gestellten Problem, Math. Nachr.,14(1955), pp. 249-252.
[5] T. HAWKES- M. ISAACS- M. OÈZAYDIN, On the MoÈbius function of a finite group, Rocky Mountain Journal,19(1989), pp. 1003-1034.
[6] A. MANN, Positively finitely generated groups, Forum Math., 8 (1996), pp. 429-459.
[7] M. PATASSINI, The Probabilistic Zeta function of PSL2(q), of the Suzuki groups2B2(q)and of the Ree groups2G2(q), Pacific J. Math.,240(2009), pp.
185-200.
[8] M. PATASSINI,On the Dirichlet polynomial of the simple group of Lie type, UniversitaÁ di Padova, 2011, http://paduaresearch.cab.unipd.it/3272/1/Phd_The- sis.pdf,
[9] M. PATASSINI,On the (non-)contractibility of the order complex of the coset poset of a classical group, J. Algebra,343(2011), pp. 37-77.
[10] M. PATASSINI,Recognizing the non-Frattini abelian chief factors of a finite group from its Probabilistic Zeta function, Accepted by Comm. Algeb., 2011.
[11] P. JIMEÂNEZ SERAL, Coefficient of the Probabilistic Zeta function of a monolithic group, Glasgow J. Math.,50(2008), pp. 75-81.
Manoscritto pervenuto in redazione il 9 Settembre 2011.