J. reine angew. Math.

549 (2002), 1—21 Journal für die reine und

angewandte Mathematik
( Walter de Gruyter
Berlin
New York 2002

Counting subgroups in nilpotent groups and

points on elliptic curves
By Marcus du Sautoy at Oxford

1. Introduction

We shall explain how to use zeta functions of groups to generate a hierarchy within
the class of finitely generated nilpotent groups indexed by certain subrings of the Gro-
thendieck ring of varieties.

The zeta function of a group was introduced by Grunewald, Segal and Smith in [7]
to provide a new invariant for a finitely generated nilpotent group G. It is defined as a
Dirichlet series encoding the number ae
n ðGÞ of all subgroups of index n in G:

P P
y
ze
G ðsÞ ¼ jG : Hjs ¼ ae s
n ðGÞn :
HeG n¼1

They also defined the normal zeta function of G:

P P
y
zG/ ðsÞ ¼ jG : Hjs ¼ an/ ðGÞns
H/G n¼1

where the coe‰cients an/ ðGÞ of the Dirichlet series record the number of normal subgroups
of index n in G. The expression as a sum over subgroups suggests that this is a natural non-
commutative generalization of the zeta function of a number field.

These zeta functions decompose as Euler products of local factors: for  A fe; /g
Q
zG ðsÞ ¼ 
zG; p ðsÞ
p prime

where

 P
y
zG; p ðsÞ ¼ apn ðGÞpns :
n¼0
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
2 Sautoy, Subgroups in nilpotent groups

The local factors were proved in [7] to be rational functions in ps . One of the major
open problems raised in [7] was the dependence on p of these local factors. The authors
speculated that the analogy with the zeta function of a number field might imply a finitely
uniform description for these local factors.

This was clarified in recent work with Grunewald [4]. We show that it is the Weil zeta
function counting points on varieties mod p on varieties which o¤ers the better analogy
rather than the zeta function of a number field. It is this recent work together with work of
myself and Loeser [5] on the concept of an associated motivic zeta function which reveals a
path from nilpotent groups to subrings of the Grothendieck ring of algebraic varieties.

In [4] we have provided an explicit formula for these local factors which depends on
counting points mod p on an explicit system of subvarieties Ei (i A T, T finite) of a variety Y
defined over Z: for each subset I of T there exists a rational function WI ðX ; Y Þ A QðX ; Y Þ
such that for almost all primes p
 P
ð1:1Þ zG; p ðsÞ ¼ cI ð pÞWI ð p; ps Þ
I HT

where

ð1:2Þ cI ð pÞ ¼ cardfa A Y ðFp Þ: a A Ei ðFp Þ if and only if i A I g:

The varieties Ei are the irreducible components corresponding to a resolution of sin-

gularities of a polynomial FG ðX Þ defined from a presentation for G (or rather its associated
Lie algebra L).

To attach some subring of the Grothendieck ring to the nilpotent group G, the idea is
to look at the ring generated by the varieties that we need to count points on mod p to get
an explicit expression like (1.1). However, this expression involves many choices: a choice
of a presentation for G, a choice of a resolution of singularities, and a choice of simplicial
decomposition of an associated cone in which we count lattice points. Not only that, but
non-isomorphic varieties can have the same number of points mod p. So on its own the
explicit expression derived in [4] is not enough to attach in some well-defined manner some
subring of the Grothendieck ring to G. It is the concept of an associated motivic zeta func-
tion attached to these zeta functions which allows one to canonically associate a subring of
the Grothendieck ring to G. This motivic zeta function takes its values in the Grothendieck
ring and by ‘‘taking the trace of Frobenius’’ of the motivic zeta function one recovers the

original zeta functions zG; p ðsÞ. The motivic zeta function is however independent of a pre-
sentation and resolution. Hence the subring generated by the coe‰cients of the zeta func-
tion is canonically associated to the group G.

Despite this theoretical work it was unclear what sort of varieties could possibly ap-

pear in expressions for zG; p ðsÞ. Indeed the speculations in the original paper of Grunewald,
Segal and Smith [7] implied that it was plausible that one only got varieties whose number
of points mod p were finitely uniform (i.e. given by a polynomial in p depending on some
finite partition of primes), e.g. rational varieties or Artin motives. So all nilpotent groups
would sit at the bottom of the hierarchy we are proposing.

In this paper we present examples which show that these zeta functions contain in
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 3

general a much richer algebraic geometry than simply rational varieties. In particular we
prove

Theorem 1.1. For each elliptic curve E ¼ y 2 þ x 3  Dx, define a nilpotent group GðEÞ
by the following presentation:
 
x1 ; x2 ; x3 ; x4 ; x5 ; x6 ; y1 ; y2 ; y3 : ½x1 ; x4 ¼ y3D ; ½x1 ; x5 ¼ y1 ; ½x1 ; x6 ¼ y2 ;
GðEÞ ¼ :
½x2 ; x4 ¼ y1 ; ½x2 ; x5 ¼ y3 ; ½x3 ; x4 ¼ y2 ; ½x3 ; x6 ¼ y1

Then there exist two rational functions P1 ðX ; Y Þ and P2 ðX ; Y Þ A QðX ; Y Þ such that for al-
most all primes p:
/ s s
zGðEÞ; p ðsÞ ¼ P1 ð p; p Þ þ jEðFp ÞjP2 ð p; p Þ:

A corollary to results proved in a previous paper [3] in the case of D ¼ 1 implies that
the rational functions P1 ðX ; Y Þ and P2 ðX ; Y Þ are non-zero, i.e. that one can’t avoid counting
points on the elliptic curve E. The proof depends on counting the number of normal sub-
groups of index p 5 and can easily be adapted to prove the same result for general D.

In section 5 we shall explain the concept of the associated motivic zeta function de-
veloped in [5] which will imply the following:

Corollary 1.2. The curve E is canonically attached to the nilpotent group GðEÞ.

The methods developed in this paper o¤er the hope to show that nilpotent groups can
involve arbitrary varieties. However there is one class of groups of which it is still con-
jectured that the associated varieties are all rational, namely free nilpotent groups (see [7]).
This conjecture has been demonstrated to have more significance than first realised. In [1]
and [2] it is explained why this conjecture relates to Higman’s PORC conjecture that the
number f ð p; nÞ of p-groups of order p n is given, for each fixed n, by polynomials in p de-
pending only on the residue class of p modulo some fixed integer Nn .

Acknowledgements. I would like to thank the Royal Society for support in the form
of a University Research Fellowship. I would also like to thank the Max Planck Institut in
Bonn where the group GðEÞ flashed into my head one evening and for invaluable conversa-
tions with Fritz Grunewald during my time at the MPI. Thanks are also due to my Ph.D.
student Christopher Voll who corrected a number of errors in the original manuscript.

2. Nilpotent groups and elliptic curves

Let LðEÞ be the class two nilpotent Lie algebra over Z of dimension 9 as a free
Z-module given by the following presentation:
 
x1 ; x2 ; x3 ; x4 ; x5 ; x6 ; y1 ; y2 ; y3 : ðx1 ; x4 Þ ¼ Dy3 ; ðx1 ; x5 Þ ¼ y1 ; ðx1 ; x6 Þ ¼ y2 ;
LðEÞ ¼
ðx2 ; x4 Þ ¼ y1 ; ðx2 ; x5 Þ ¼ y3 ; ðx3 ; x4 Þ ¼ y2 ; ðx3 ; x6 Þ ¼ y1

where all other commutators are defined to be 0. Then LðEÞ n Q is the Q-Lie algebra as-
sociated to the torsion-free finitely generated nilpotent group GðEÞ under the Mal’cev cor-
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
4 Sautoy, Subgroups in nilpotent groups

respondence. We can define the ideal zeta function associated to LðEÞ similarly to the
normal zeta function associated to GðEÞ:

P
y  
/
zLðEÞ; p ðsÞ ¼ ap/n LðEÞ pns
n¼0

where ap/n ðLÞ is the number of ideals of L of index p n . Section 4 of [7] confirms the following:

Proposition 2.1. For almost all primes p,

/ /
zGðEÞ; p ðsÞ ¼ zLðEÞ; p ðsÞ:

It therefore su‰ces to prove:

Theorem 2.2. There exist two rational functions P1 ðX ; Y Þ and P2 ðX ; Y Þ A QðX ; Y Þ

such that for p coprime to 2D:

/ s s
zLðEÞ; p ðsÞ ¼ P1 ð p; p Þ þ jEðFp ÞjP2 ð p; p Þ:

We begin by recalling something of the proof of (1.1) in [4] which will motivate the
direction for the proof of Theorem 2.2. The proof of this explicit formula breaks up into a
number of stages:
/
(1) We show how to express zLðEÞ; p ðsÞ as something we call a cone integral.
Cone integrals are defined for a set of cone integral data consisting of polynomials
D ¼ f f0 ðxÞ; g0 ðxÞ; . . . ; fl ðxÞ; gl ðxÞg by
Ð
ZD ðs; pÞ ¼ j f0 ðxÞj s jg0 ðxÞj jdxj
Vp ðDÞ

where
    
Vp ðDÞ ¼ x A Zpm : v fi ðxÞ e v gi ðxÞ for i ¼ 1; . . . ; l

and jdxj is the additive Haar measure on Zpm . The cone integrals in the case  of counting

ideals in a Lie algebra L of dimension d are defined as follows: let Cj ¼ cik ð jÞ be the
P
d
d d matrix defined by ðei ; ej Þ ¼ cik ð jÞek . Let M ¼ ðmij Þ be an upper triangular matrix
k¼1
whose rows we shall call mi and denote by M y the adjoint matrix. Then define polynomials
/
gijk ðmrs Þ in the entries of this triangular matrix to be the kth entry of the d-tuple mi Cj M y .
/
The cone conditions defining zL; p ðsÞ are given then by

  / 
Vp ðDÞ ¼ x A Zpm : vðm11


mdd Þ e v gijk ðmrs Þ for i; j; k ¼ 1; . . . ; d

/
and zL; p ðsÞ can be expressed in terms of the associated cone integral by

1 d
Ð
/
zL; p ðsÞ ¼ ð1  p Þ jm11


mdd j sd jm11
d1



md1d1 j jdxj:
Vp ðDÞ

Brought to you by | Sackler Library

Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 5

(2) If the polynomials involved in the cone integrals are all monomial then the cal-
culation reduces to a discrete problem about summing lattice points representing the valu-
ations of the variables mij satisfying various linear inequalities. We explain in the next
/
step why this is uniform in p. But in general the polynomials gijk ðmrs Þ are far from mono-
mial. To overcome this Q /problem we do a resolution of singularities on the polynomial
FL ðMÞ ¼ m11


mdd gijk ðmrs Þ which results in a partition of the resolved space into re-
i; j; k
gions on which the polynomial FL ðMÞ now becomes monomial. The partitioning of this
space naturally leads to the problem of counting points on the associated irreducible com-
ponents Ei ði A TÞ of the resolution of singularities of FL ðMÞ. This is the only place where
the evaluation of this integral depends in some essential way on p. Apart from throwing
away finitely many primes at various points (e.g. where the resolution has bad reduction)
the integral is uniform outside this partitioning.
P
(3) This reduces our integral to a finite sum cI ð pÞZI ðsÞ, where the coe‰cients
I HT
cI ð pÞ defined in (1.2) capture the essential dependence on p, and the ZI ðsÞ are cone in-
tegrals with respect to monomial polynomials of the following shape:

Ð
ZI ðsÞ ¼ jy1 j a0i1 sþb0i1


jyr j a0ir sþb0ir jdyj
Vp ðI Þ

where I ¼ fi1 ; . . . ; ir g and

Vp ðI Þ ¼ fy A ð pZp Þ m : vðjy1 j aji1


jyr j ajir Þ e vðjy1 j bji1


jyr j bjir Þ for j ¼ 1; . . . ; lg:

These in turn can be expressed as a sum over lattice points n ¼ ðn1 ; . . . ; nm Þ corresponding
to the valuations ni ¼ vðyi Þ:

P
ZI ðsÞ ¼ ð1  p1 Þ r p mr ð pða0i1 sþb0i1 þ1Þn1 


ða0ir sþb0ir þ1Þnr Þ
nAL

where
m
L ¼ fn A N>0 : aji1 n1 þ


þ ajir nr e bji1 n1 þ


þ bjir nr for j ¼ 1; . . . ; l g:

The evaluation of ZI ðsÞ is therefore independent of p thanks to the following propo-

sition contained in [4] which we shall use later on in this paper:
S
Proposition 2.3. Suppose there exists a finite partition Yi of R d defined by linear
iAS
inequalities with coe‰cients over Q and for each i A S linear functions ai ðxÞ and b i ðxÞ defined
over Q . Then there exist rational functions Hi ðX ; Y Þ A QðX ; Y Þ such that for each i A S
P
paðnÞsbðnÞ ¼ Hi ð p; ps Þ:
d
n A N XYi

These geometric progressions can be calculated by decomposing each cone Yi defined

by the linear inequalities into open simplicial cones with fundamental regions of volume 1.
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
6 Sautoy, Subgroups in nilpotent groups

Our strategy in proving Theorem 2.2 is that once we have established the shape of the
cone integrals depending on the presentation of LðEÞ, we shall perform a direct analysis to
partition our integral according to points on the elliptic curve such that the individual in-
tegrals reduce to sums of lattice points of the shape detailed in Proposition 2.3.

Our direct analysis calculates the integral by parts exploiting the fact that in a class
two nilpotent Lie algebra, the algebra splits into two abelian sections. We first integrate the
‘top’, the abelianisation, with respect to some fixed choice of basis for the centre. We then
integrate the ‘bottom’, the centre, with respect to the functions introduced in the evalua-
tion of the top. The analysis avoids the direct application of a resolution of singularities
although there is in fact a hidden blow-up at the heart of some of the case analysis in con-
sidering the integral of the bottom.

Note that the analysis could be carried through to its ultimate conclusion resulting in
an explicit evaluation of the zeta function. However the analysis in (3) involving a sim-
plicial decomposition of cones C, although uniform in p, results in general in a very com-
plicated case analysis. Since this paper has its focus in the dependence on p of these local
factors, we have chosen to subsume these complications under the umbrella of a uniform
calculation.

Proof of Theorem 2.2. Let

0 1
0 0 D
@
Cð1Þ ¼ 1 0 0 A;
0 1 0
0 1
1 0 0
Cð2Þ ¼ @ 0 0 1 A;
0 0 0
0 1
0 1 0
@
Cð3Þ ¼ 0 0 0 A:
1 0 0

Then for i; j ¼ 1; 2; 3 we have

ðxi ; xjþ3 Þ ¼ Ci1 ð jÞy1 þ Ci2 ð jÞy2 þ Ci3 ð jÞy3

¼ Cj1 ðiÞy1 þ Cj2 ðiÞy2 þ Cj3 ðiÞy3 :

From the analysis described above in (1) from [4] we can express the zeta function
/
zLðEÞ; p ðsÞ for all primes p by the following cone integral:

1 9
Ð
/
ð2:1Þ zLðEÞ; p ðsÞ ¼ ð1  p Þ jm11 j s1


jm66 j s6 jn1 j s7 jn2 j s8 jn3 j s9 jdmj jdnj
Vp

where dm and dn are additive Haar measures on Tr6 ðZp Þ and Tr3 ðZp Þ respectively and Vp
consists of all pairs of matrices
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 7
0 0 11
n1 a b
B @ C
ðM; NÞ ¼ @ðmij Þ; 0 n2 c AA A Tr6 ðZp Þ Tr3 ðZp Þ
0 0 n3
j j j
satisfying: for i ¼ 1; . . . ; 6; e ¼ 0 or 3 and j ¼ 1; 2; 3 there exists ðlieþ1 ; lieþ2 ; lieþ3 Þ A Zp3 such
that
j j j
ð2:2Þ ðmieþ1 ; mieþ2 ; mieþ3 ÞCð jÞN y ¼ ðlieþ1 n1 n2 n3 ; lieþ2 n1 n2 n3 ; lieþ3 n1 n2 n3 Þ

where N y is the adjoint matrix

0 1
n2 n3 an3 ac  n2 b
Ny ¼ @ 0 n3 n1 cn1 A:
0 0 n1 n2

The integral (2.1) is the same as the sum:

P
pM1 s


pM6 ðs5Þ pN1 ðs6Þ pN2 ðs7Þ pN1 ðs8Þ mðM1 ; . . . ; M6 ; N1 ; N2 ; N3 Þ
M1 ;...; M6 ; N1 ; N2 ; N3 A N

 
where mðM1 ; . . . ; M6 ; N1 ; N2 ; N3 Þ is the measure of the set of ðmij Þi<j ; a; b; c satisfying
(2.2) with mii replaced by p Mi and nj replaced by p Nj .

3. Integrating by parts: the top

We look to establish an expression for mðM1 ; . . . ; M6 ; N1 ; N2 ; N3 Þ by calculating the

measure modulo the matrix for the centre N. We integrate by parts by performing the in-
tegration on the matrix for the abelianisation. It will su‰ce by symmetry to calculate values
for the following functions: for each M1 and
0 1
p N1 a b
N¼@ p N2 c A A Tr3 ðZp Þ
p N3

(1) Z1 ðM1 ; NÞ ¼ mðW1 Þ where W1 consists of ðm2 ; m3 Þ A Zp2 such that for each
j ¼ 1; 2; 3 there exists ðl1j ; l2j ; l3j Þ A Zp3

ð3:1Þ ð p M1 ; m2 ; m3 ÞCð jÞN y ¼ ðl1j p N1 þN2 þN3 ; l2j p N1 þN2 þN3 ; l3j p N1 þN2 þN3 Þ;

(2) Z2 ðM2 ; NÞ ¼ mðW2 Þ where W2 consists of m3 A Zp such that for each j ¼ 1; 2; 3

there exists ðl1j ; l2j ; l3j Þ A Zp3

ð3:2Þ ð0; p M2 ; m3 ÞCð jÞN y ¼ ðl1j p N1 þN2 þN3 ; l2j p N1 þN2 þN3 ; l3j p N1 þN2 þN3 Þ;

(3) Z3 ðM3 ; NÞ ¼ 1 if there exists ðl1j ; l2j ; l3j Þ A Zp3 such that

ð0; 0; p M3 ÞCð jÞN y ¼ ðl1j p N1 þN2 þN3 ; l2j p N1 þN2 þN3 ; l3j p N1 þN2 þN3 Þ

and 0 otherwise.
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
8 Sautoy, Subgroups in nilpotent groups

Once we have done this, the measure can be expressed as

mðM1 ; . . . ; M6 ; N1 ; N2 ; N3 Þ
 3
Ð P M ðMþ1Þ
¼ Z1 ðM1 ; NÞZ2 ðM2 ; NÞZ3 ðM3 ; NÞ Z1 ðM; NÞð p p Þ
ða; b; cÞ M

Z1 ðM4 ; NÞZ2 ðM5 ; NÞZ3 ðM6 ; NÞjdaj jdbj jdcj:

3.1. Calculating Z 1 (M 1 , N ). We shall begin by calculating a value for the function

Z1 ðM1 ; NÞ.

Condition (3.1) is equivalent to the following conditions:

ð3:3Þ p M1 ; m2 ; m3 1 0 mod p N1 ;
0 1
0 a p N1
B C
ð3:4Þ ð p M1 ; m2 ; m3 Þ@ a 0 0 A 1 0 mod p N1 þN2 ;
p N1 0 a
0 1
Dp N1 þN2 ac  p N2 b cp N1
B C
ð3:5Þ ð p M1 ; m2 ; m3 Þ@ ac  p N2 b p N1 þN2 0 A 1 0 mod p N1 þN2 þN3 :
cp N1 0 ac  p N2 b

Since ðp M1 ; m2 ; m3 Þ A p N1 Zp3 by (3.3) we get that


0 if M1 < N1 ;
Z1 ðM1 ; NÞ ¼
p2N1 ZðM1  N1 ; NÞ if N1 e M1

where Z 1 ðM1 ; NÞ is the measure of ðm2 ; m3 Þ A Zp2 satisfying

0 1
a p N1
0
M1 B C
ð3:6Þ ð p ; m2 ; m3 Þ@ a
0 0 A 1 0 mod p N2 ;
p N1
0 a
0 1
Dp N1 þN2 ac  p N2 b cp N1
B C
ð3:7Þ ð p M1 ; m2 ; m3 Þ@ ac  p N2 b p N1 þN2 0 A 1 0 mod p N2 þN3 :
cp N1 0 ac  p N2 b

We can write this as the problem of calculating the measure of ðm2 ; m3 Þ A Zp2 satisfying:

ð p M1 ; m2 ; m3 ÞðS1 ; S2 Þ 1 0 mod p N2 þN3

where
0 1
0 ap N3 p N1 þN3 Dp N1 þN2 ac  p N2 b cp N1
B C
ð3:8Þ ðS1 ; S2 Þ ¼ @ ap N3 0 0 ac  p N2 b p N1 þN2 0 A:
p N1 þN3 0 ap N3 cp N1 0 ac  p N2 b
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 9

Once we have the existence of one solution ð p M1 ; M2 ; M3 Þ then all the other solutions
are of the form ðp M1 ; M2 ; M3 Þ þ ð0; m2 ; m3 Þ where ðm2 ; m3 Þ is a solution of

ðm2 ; m3 ÞðR1 ; R 2 Þ 1 0 mod p N2 þN3

and
!
ap N3 0 0 ac  p N2 b p N1 þN2 0
ð3:9Þ ðR1 ; R 2 Þ ¼ :
p N1 þN3 0 ap N3 cp N1 0 ac  p N2 b

Once we have a solution then the value of Z 1 ðM1 ; NÞ is the measure of this set which
is

p U1 þU2 2ðN2 þN3 Þ

where

ð3:10Þ U1 ¼ minfu1 ; N2 þ N3 g
U2 ¼ minfu2 ; N2 þ N3 g;

and

u1 ¼ minfvðdet X Þ: X is a 1 1 minor of ðR1 ; R 2 Þg;

u2 ¼ minfvðdet X Þ: X is a 2 2 minor of ðR1 ; R 2 Þg  u1 :

Put b~ ¼ ac  p N2 b then in this case

u1 ¼ minfvðaÞ þ N3 ; vðb~Þ; N1 þ N2 ; N1 þ N3 ; vðcÞ þ N1 g;

8   9
< 2 vðaÞ þ N3 ; vðacp 1 3  b~p 1 3 Þ ¼ vðbÞ þ N1 þ N2 þ N3 ; >
> N þN N þN
=
u2 ¼ min 2N1 þ N2 þ N3 ; vðaÞ þ vðb~Þ þ N3 ; vðaÞ þ N1 þ N2 þ N3 ;  u1 :
>
: >
;
~2 ~
vðcÞ þ 2N1 þ N2 ; vðb Þ; vðbÞ þ N1 þ N2

We have to calculate a condition on M1 that we have such a solution. Let H1 be the

minimal value of M1 such that there is a solution of

ðp M1 ; m2 ; m3 ÞðS1 ; S2 Þ 1 0 mod p N2 þN3 :

Then the measure of all the solutions of ðm1 ; m2 ; m3 ÞðS1 ; S2 Þ 1 0 mod p N2 þN3 is

pH1 þU1 þU2 2ðN2 þN3 Þ :

But we have another expression for this measure, namely it is

p W1 þW2 þW3 3ðN2 þN3 Þ

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
10 Sautoy, Subgroups in nilpotent groups

where

ð3:11Þ W1 ¼ minfw1 ; N2 þ N3 g;
W2 ¼ minfw2 ; N2 þ N3 g;
W3 ¼ minfw3 ; N2 þ N3 g;
and

w1 ¼ minfvðdet X Þ: X is a 1 1 minor of ðS1 ; S2 Þg;

w2 ¼ minfvðdet X Þ: X is a 2 2 minor of ðS1 ; S2 Þg  w1 ;
w3 ¼ minfvðdet X Þ: X is a 3 3 minor of ðS1 ; S2 Þg  w1  w2 :

This provides us with a smooth way to understand the value of H1 .

An analysis of the 18 1 1 minors of ðS1 ; S2 Þ, 45 2 2 minors and 20 3 3 minors

of ðS1 ; S2 Þ reveals that

w1 ¼ minfvðaÞ þ N3 ; vðb~Þ; N1 þ N2 ; N1 þ N3 ; vðcÞ þ N1 g;

8   9
> ~
< 2 vðaÞ þ N3 ; vðaÞ þ vðbÞ þ N3 ; vðaÞ þ vðcÞ þ N1 þ N3 ; > =
w2 ¼ min vðaÞ þ N 1 þ N 2 þ N 3 ; vðb~Þ þ N 1 þ N 3 ; 2vð ~
b Þ; 2N 1 þ 2N 2 >  w1 ;
;
>
: ~ ;
vðbÞ þ vðcÞ þ N1 ; 2N1 þ 2N3 ; 2vðcÞ þ 2N1 ; vðbÞ þ N1 þ N2 ; ~
8   9
>
> 3 vðaÞ þ N3 ; vðaÞ þ N1 þ 2N3 þ vðbÞ þ N2 ; >
>
>
> >
>
>
> vðaÞ þ 2N þ 2N þ N ; 2vðaÞ þ vð ~
b Þ þ 2N ; >
>
>
> 1 3 2 3 >
>
>
> >
>
>
> 2N 1 þ N 2 þ 2N 3 þ vðbÞ; 3N 1 þ 2N 3 þ N 2 ; >
>
>
> >
>
>
> 3N þ N þ 2N ; 2N þ N þ vðcÞ þ N þ vðbÞ; >
>
< 1 3 2 1 3 2 =
w3 ¼ min 3N3 þ N2 þ N3 þ vðcÞ; 2vðaÞ þ N1 þ N2 þ 2N3 ;  w1  w2 :
>
> >
>
>
> vðaÞ þ vðcÞ þ 2N1 þ N2 þ N3 ; vðaÞ þ 2vðbÞ þ N3 ; > ~ >
>
> >
>
>
> vðaÞ þ vð ~Þ þ N1 þ N2 þ N3 ; vðaÞ þ 2N1 þ 2N2 þ N3 ; >
b >
>
>
>
> >
>
>
> vðbÞ þ vð ~
b Þ þ N þ N þ N ; vðbÞ þ 2N þ 2N þ N ; >
>
>
> 1 2 3 1 2 3 >
>
>
:  2 2  >
;
v b~  ðcp 1 Þ p 1 2 þ Db~ð p 1 2 Þ
3 N N þN N þN

(We have used in our analysis that minfvðX þ Y Þ; vðX Þg ¼ minfvðX Þ; vðY Þg and
minfvðX 2 Þ; vðXY Þ; vðY 2 Þg ¼ minfvðX 2 Þ; vðY 2 Þg. Also recall that we have assumed that D
is a unit in Zp since p is coprime to D.)

In conclusion we have

Proposition 3.1.

0 if M1 < U1 þ U2  ðW1 þ W2 þ W3 Þ þ N1 þ N2 þ N3 ;
Z1 ðM1 ; NÞ ¼
p U1 þU2 2ðN1 þN2 þN3 Þ otherwise;

where U1 ; U2 ; W1 ; W2 ; W3 are defined as above in (3.10) and (3.11).

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 11

3.2. Calculating Z 2 (M 2 , N ). As in the previous analysis


0 if M2 < N1 ;
Z2 ðM2 ; NÞ ¼ N1
p Z2 ðM2  N1 ; NÞ if N1 e M2 ;

where Z2 ðM2 ; NÞ is the measure of m3 A Zp satisfying

0 1
a p N1
0
M2 B C
ð3:12Þ ð0; p ; m3 Þ@ a0 0 A1 0 mod p N2 ;
p N1
0 a
0 1
Dp N1 þN2 ac  p N2 b cp N1
B C
ð3:13Þ ð0; p M2 ; m3 Þ@ ac  p N2 b p N1 þN2 0 A1 0 mod p N2 þN3 :
cp N1 0 ac  p N2 b

We can write the evaluation of Z2 ðM2 ; NÞ as the problem of calculating the measure
of m3 A Zp satisfying:

ð0; p M2 ; m3 ÞðS1 ; S2 Þ 1 0 mod p N2 þN3

where ðS1 ; S2 Þ was defined in (3.8).

Once we have the existence of one solution ð0; p M2 ; M3 Þ then all the other solutions
are of the form ð0; p M2 ; M3 Þ þ ð0; 0; m3 Þ where m3 is a solution of

m3
ðp N1 þN3 0 ap N3 cp N1 0 ac  p N2 bÞ 1 0 mod p N2 þN3 :

Once we have a solution then the value of Z2 ðM2 ; NÞ is the measure of this set which
is p V1 ðN2 þN3 Þ where

ð3:14Þ V1 ¼ minfðN1 þ N3 Þ; vðaÞ þ N3 ; vðcÞ þ N1 ; vðb~Þ; ðN2 þ N3 Þg:

Again we have to calculate a condition on M2 that we have such a solution. Let H2 be

the minimal value of M2 such that there is a solution of

ð p M2 ; m3 ÞðR1 ; R 2 Þ 1 0 mod p N2 þN3

where ðR1 ; R 2 Þ was defined in (3.9).

Then the measure of all the solutions of ðm2 ; m3 ÞðR1 ; R 2 Þ 1 0 mod p N2 þN3 is

pH2 þV1 ðN2 þN3 Þ :

But we have another expression for this measure, namely it is

p U1 þU2 2ðN2 þN3 Þ

where U1 and U2 are defined above in (3.10).

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
12 Sautoy, Subgroups in nilpotent groups

Again this provides us with a smooth way to understand the value of H2 .

In conclusion we have:

Proposition 3.2.

0 if M2 < V1  ðU1 þ U2 Þ þ N1 þ N2 þ N3 ;
Z2 ðM2 ; NÞ ¼
p V1 ðN1 þN2 þN3 Þ otherwise;

where V1 is defined in (3.14) and U1 and U2 are defined in (3.10).

3.3. Calculating Z 3 (M 3 , N ). Similarly to the above we have


0 if M3 < N1 ;
Z3 ðM3 ; NÞ ¼
Z3 ðM3  N1 ; NÞ if N1 e M3 ;

where Z3 ðM3 ; NÞ ¼ 1 if
01
a p N1 0
M3 B C
ð3:15Þ 0 A 1 0 mod p N2 ;
ð0; 0; p Þ@ a
0
0 a p N1
0 1
Dp N1 þN2 ac  p N2 b cp N1
B C
ð3:16Þ ð0; 0; p M3 Þ@ ac  p N2 b p N1 þN2 0 A 1 0 mod p N2 þN3 :
cp N1 0 ac  p N2 b

We can see directly from the conditions (3.15) and (3.16) that Z3 ðM3 ; NÞ ¼ 1 if and only if

M3 f N2  N1 ; N2  vðaÞ;

M3 f N2 þ N3  N1  vðcÞ; N2 þ N3  vðac  p N2 bÞ;

and equals 0 otherwise.

If we return now to our integral we see that it is equal to

P
ð3:17Þ pM1 s . . . pM6 ðs5Þ pN1 ðs6Þ pN2 ðs7Þ pN3 ðs8Þ
M1 ;...; M6 ; N1 ; N2 ; N3 A N
 3
Ð P
Z1 ðM1 ; NÞZ2 ðM2 ; NÞZ3 ðM3 ; NÞ Z1 ðM; NÞð pM  pðMþ1Þ Þ
ða; b; cÞ M

Z1 ðM4 ; NÞZ2 ðM5 ; NÞZ3 ðM6 ; NÞjdaj jdbj jdcj:

4. Integrating by parts: the bottom

To calculate the integral over the matrix for the centre, we shall need to know the
value of
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 13
( )
ða; b; cÞ A Zp3 : vðaÞ ¼ A; vðbÞ ¼ B; vðb~Þ ¼ B~; vðcÞ ¼ C;
aA; B; B~; C; E; N1 ; N2 ¼m   :
v b~3  ðcp N1 Þ 2 p N1 þN2 þ Db~ð p N1 þN2 Þ 2 ¼ E

It will su‰ce to prove the following although our analysis provides precise informa-
tion about the linear functions and inequalities involved:
S
Proposition 4.1. There exists a finite partition Yi of R 7 defined by linear inequal-
iAS
ities with coe‰cients in Q and for each i A S, polynomials Pi ðX Þ and Qi ðX Þ and linear func-
tion ai ðxÞ and bi ðxÞ such that if D ¼ ðA; B; B~; C; E; N1 ; N2 Þ A N 7 X Yi then

aD ¼ aA; B; B~; C; E; N1 ; N2 ¼ Pi ð pÞp aI ðDÞ þ Qi ð pÞjEðFp Þjp bI ðDÞ :

Proof. Recall b~ ¼ ac  p N2 b. We can do this bit by parts by first fixing b~ and c and
calculating the measure of the corresponding a:
8
>
> 0 if B 3A þ C and N2  C > minðA; B  CÞ;
> A
< 1
  p ð1  p Þ if B 3A þ C and N2  C eminðA; B  CÞ;
m ðb=c þ p N2 C Zp Þ X p A Zp ¼
>
> p CN2 if B ¼ A þ C and A þ C < N2 ;
>
: A
p if B ¼ A þ C and A þ C < N2 :

Then
( )
ða; b; cÞ A Zp3 : vðaÞ ¼ A; vðbÞ ¼ B; vðb~Þ ¼ B~; vðcÞ ¼ C;
m  
v b~3  ðcp N1 Þ 2 p N1 þN2 þ Db~ð p N1 þN2 Þ 2 ¼ E
( )
  ðb~; cÞ A Zp2 : vðb~Þ ¼ B~; vðcÞ ¼ C;
¼ m ðb=c þ p N2 C Zp Þ X p A Zp p N2
m   :
v b~3  ðcp N1 Þ 2 p N1 þN2 þ Db~ð p N1 þN2 Þ 2 ¼ E

Hence it su‰ces to calculate the measure of

  
ð4:1Þ m ðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C; v b 3  ðcp N1 Þ 2 p N1 þN2 þ Dbð p N1 þN2 Þ 2 ¼ E

where now we write b for b~ and B for B~ to avoid too much notation.

The measure in (4.1) is the same as

  
p N1 m ðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C þ N1 ; v b 3  c 2 p N1 þN2 þ Dbð p N1 þN2 Þ 2 ¼ E :

Let N1 þ N2 ¼ N.

We run over three cases: (1) N e B; C; (2) B < N, B e C; (3) C < B; N.

(1) N e B; C. We make a transformation b 0 ¼ b=p N and c 0 ¼ c=p N :

mfðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C; vðb 3  c 2 p N þ Dbp 2N Þ ¼ Eg

¼ p2N mfðb 0 ; c 0 Þ A Zp2 : vðb 0 Þ ¼ B  N; vðc 0 Þ ¼ C  N; vðb 03  c 02 þ Db 0 Þ ¼ E  3Ng:

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
14 Sautoy, Subgroups in nilpotent groups

We need to calculate a value for

cB; C; E ¼ mfðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C; vðb 3  c 2 þ DbÞ ¼ Eg

(where we have replaced b 0 ; c 0 by b and c).

If we put

dB; C; E ¼ mfðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C; vðb 3  c 2 þ DbÞ f Eg

then cB; C; E ¼ dB; C; E  dB; C; Eþ1 .

We shall suppose that p 3 2.

Lemma 4.2. (1) If E e minfB; 2Cg then dB; C; E ¼ pC pB ð1  p1 Þ 2 .

(2) If B 3 0 and minfB; 2Cg < E then dB; C; E ¼ 0 unless B ¼ 2C when

dB; C; E ¼ pE pC ð1  p1 Þ.

(3) If B ¼ 0 then if 1 e E e 2C we have d 0; C; E ¼ 2pE pC ð1  p1 Þ.

 
(4) If B ¼ 0 then d 0; 0; E ¼ p E1 d 0; 0; 1 ¼ p E1 jEðFp Þj  1 .

Proof. (1) is clear. When B 3 2C, (2) follows since we can’t get any solutions. (3)
forces the value of b ¼ G1 mod p E . The remaining cases depend on the following quanti-
tative version of Hensel’s lemma applied to the non-singular curve E ¼ Y 2 þ X 3  DX :

Lemma 4.3. Let E f 1, p 3 2. Let ðb; cÞ A Z=p E Z with b 3  c 2 þ Db ¼ 0 mod p E .

Then there exist exactly p elements ðb1 ; c1 Þ A Z=p Eþ1 Z with b 1 b1 ; c 1 c1 mod p E and
b13  c12 þ Db1 ¼ 0 mod p Eþ1 .

Proof. Put b1 ¼ b þ bp E and c1 ¼ c þ gp E . We are required to count how many

pairs ðb; gÞ A f0; . . . ; p  1g 2 there are satisfying:

ð4:2Þ b13  c12 þ Db1 ¼ 0 mod p Eþ1 :

We know that b 3  c 2 þ Db ¼ tp E for some t. Hence expanding the equation (4.2)

and pulling out the common power of p E we have to solve:

t þ bðD  3b 2 Þ  2gc ¼ 0 mod p:

If c ¼ 0 mod p then bðD  b 2 Þ ¼ 0 mod p which implies that D  3b 2 3 0 mod p (as-

suming p 3 2Þ. Hence we get exactly p lifts to solutions ðb1 ; c1 Þ. This completes Lemma
4.3.

This lemma implies that whenever E > B; C; 0 then dB; C; Eþ1 ¼ p1 dB; C; E :

If B ¼ 2C < E (B ¼ 2C ¼ E is already covered) then d2C; C; E ¼ pEþð2Cþ1Þ d2C; C; 2Cþ1 .

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 15

If C > 0 then d2C; C; 2Cþ1 ¼ pC pð2Cþ1Þ ð1  p1 Þ since once c is determined this forces the
value of b mod p 2Cþ1 .
 
If B ¼ 0 then d 0; 0; E ¼ p E1 d 0; 0; 1 ¼ p E1 jEðFp Þj  1 . This completes the proof of
Lemma 4.2.

Hence we can determine dB; C; E once we know jEðFp Þj.

(Note that although E ¼ Y 2 þ X 3  DX is non-singular, it does not have normal

crossings with the varieties X ¼ 0 and Y ¼ 0. We can see the normal crossing issue here in
a di¤erence between dBþ1; C; Eþ1 and dB; Cþ1; Eþ1 and their relationship to dB; C; E . Namely
dB; Cþ1; Eþ1 ¼ p2 dB; C; E whilst for example d2Cþ2; Cþ1; Eþ1 ¼ p2 d2C; C; E .)

(2) B < N, B e C. Using a transformation b 0 ¼ p N =b and c 0 ¼ c=b it will su‰ce by

the same analysis as in case 1 to calculate a value for

dB; C; E ¼ mfðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C; vð1 3  c 2 b þ Db 2 Þ f Eg

in the case that B f 1 and C f 0.

Since B f 1 the following lemma is clear:

Lemma 4.4. dB; C; E ¼ 0 unless E ¼ 0 in which case dB; C; E ¼ pBC ð1  p1 Þ 2 .

(3) C < B; N. Using a transformation b 0 ¼ b=c and c 0 ¼ p N =c it will su‰ce to cal-

culate for B; C > 0:

dB; C; E ¼ mfðb; cÞ A Zp2 : vðbÞ ¼ B; vðcÞ ¼ C; vðb 3  c þ Dbc 2 Þ f Eg:

Lemma 4.5. Suppose B; C > 0. Then

(1) if E e minf3B; Cg then dB; C; E ¼ pCB ð1  p1 Þ 2 ;

(2) if 3B 3 C and minf3B; Cg < E then dB; C; E ¼ 0;

(3) if 3B ¼ C then dB; 3B; E ¼ pEB ð1  p1 Þ.

Proof. (1) and (2) are clear. (3) follows from the quantitative version of Hensel’s
Lemma for the non-singular curve X 3  Y þ DXY 2 :

Lemma 4.6. Let E f 1, p 3 2. Let ðb; cÞ A Z=p E Z with b 3  c þ Dbc 2 ¼ 0 mod p E .

Then there exist exactly p elements ðb1 ; c1 Þ A Z=p Eþ1 Z with b 1 b1 ; c 1 c1 mod p E and
b13  c1 þ Dbc12 ¼ 0 mod p Eþ1 .

Proof. Put b1 ¼ b þ bp E and c1 ¼ c þ gp E . We are required to count how many

pairs ðb; gÞ A f0; . . . ; p  1g 2 there are satisfying:

ð4:3Þ b13  c1 þ Dbc12 ¼ 0 mod p Eþ1 :

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
16 Sautoy, Subgroups in nilpotent groups

We know that b 3  c þ Dbc 2 ¼ tp E for some t. Hence expanding the equation (4.3)
and pulling out the common power of p E we have to solve:

t þ bðDc 2  3b 2 Þ  2gð1 þ 2DbcÞ ¼ 0 mod p:

If Dc 2  3b 2 ¼ 0 then 2Dbc 2 =3  c ¼ cð2Dbc=3  1Þ ¼ 0. So 2Dbc  1 3 0 assuming

p 3 2. Hence we get exactly p lifts to solutions ðb1 ; c1 Þ. This completes the proof of Lemma
4.6.

This lemma implies that whenever E > B; C; 1 then dB; C; Eþ1 ¼ p1 dB; C; E .

If E e minf3B; Cg then dB; C; E ¼ pC pB ð1  p1 Þ 2 .

If 3B 3 C and minf3B; Cg < E then dB; C; E ¼ 0.

So the only interesting case is dB; 3B; E . Our lemma implies that

dB; 3B; E ¼ pEþð3Bþ1Þ dB; 3B; 3Bþ1 ¼ pEþð3Bþ1Þ pB pð3Bþ1Þ ð1  p1 Þ ¼ pEB ð1  p1 Þ:

Remark 1. Note that the case distinction in (1), (2) and (3) reflects a blow-up of the
variety b 3  c 2 n þ Dbn 2 at the singular point ð0; 0; 0Þ. Each case distinction represents the
image of the variety in the three separate charts that define the blow-up.

The above analysis su‰ces to prove Proposition 4.1.

We now return to the proof of Theorem 2.2. We combine the expressions for
Z1 ðM1 ; NÞ, Z2 ðM2 ; NÞ, Z3 ðM3 ; NÞ, the analysis of aA; B; B~; C; E; N1 ; N2 with equation (3.17).
S
Together they imply that there exists a finite partition Yi of R 14 defined by linear in-
iAS
equalities with coe‰cients in Q and for each i A S, polynomials Pi ðX Þ and Qi ðX Þ and
linear function ai ðxÞ, b i ðxÞ, gi ðxÞ and di ðxÞ such that for p coprime to 2D, putting
L ¼ ðA; B; B~; C; E; M1 ; . . . ; M6 ; N1 ; N2 ; N3 Þ,
P P
/
zLðEÞ; p ðsÞ ¼ Pi ð pÞp ai ðLÞþgi ðLÞs þ QI ð pÞjEðFp Þj p bi ðLÞþdi ðLÞs :
i A S L A N 14 XYi

Proposition 2.3 implies then that there exist two rational functions P1 ðX ; Y Þ and
P2 ðX ; Y Þ A QðX ; Y Þ such that for p coprime to 2D:

/ s s
zLðEÞ; p ðsÞ ¼ P1 ð p; p Þ þ jEðFp ÞjP2 ð p; p Þ:

Hence Theorem 2.2 is proved.

5. Motivic zeta functions

In this section we show how to use the zeta function of a group to canonically as-
sociate to the group a subring of the Grothendieck ring. In particular we show why the
elliptic curve E ¼ y 2 þ x 3  Dx is canonically associated to GðEÞ.
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 17

Let us recall the definition of the Grothendieck ring M of algebraic varieties over a
field k. This is the ring generated by symbols ½S , for each S an algebraic variety over k,
with the relations

(1) ½S ¼ ½S 0 if S is isomorphic to S 0 ;

(2) ½S ¼ ½SnS 0 þ ½S 0 if S 0 is closed in S; and

(3) ½S S 0 ¼ ½S ½S 0 .

We denote by L ¼ ½Ak1 the Lefschetz motive.

It is tempting given an expression (valid for almost all p)

P
zG; p ðsÞ ¼ cI ð pÞWI ð p; ps Þ
I HT

where

cI ð pÞ ¼ cardfa A Y ðFp Þ: a A Ei ðFp Þ if and only if i A I g

to associate the subring of the Grothendieck ring generated by the varieties. But how ca-
nonical or unique is this? In general it is not unique. It is possible to have non-isomorphic
varieties with the same number of points mod p. For example, suppose that calculating
/
zGðEÞ; p ðsÞ by an alternative method we produce a formula of the form

/ 0 s 0 0 s
zGðEÞ; p ðsÞ ¼ P1 ð p; p Þ þ jE ðFp ÞjP2 ð p; p Þ

where E 0 is another elliptic curve. Then it does not mean that E and E 0 are isomorphic—
distinct varieties over Q may have the same number of points in Fp . In fact this is the case
if the elliptic curves E and E 0 are isogenous. Much deeper is the fact, due to Faltings [6],
that if, for almost all primes p, jEðFp Þj ¼ jE 0 ðFp Þj then E and E 0 are isogenous. Isogenous
elliptic curves define the same Chow motive. Therefore even without appealing to the for-
malism of motivic zeta functions, we may deduce:

Theorem 5.1. The Chow motive of E is canonically associated with GðEÞ.

However by using recent work with Loeser [5] on the concept of a motivic zeta func-
tion we can in fact show that E is canonically associated to GðEÞ. In [5] we define a motivic
zeta function associated to a Z-Lie algebra. This is a power series with coe‰cients in the
Grothendieck ring of algebraic varieties.

Let L be a Lie algebra over Z. Let Xe ðkÞ (respectively X / ðkÞÞ denote the class
of k½½t -subalgebras (respectively ideals) of L n k½½t where k is a finite extension of Q.
Let An ðX  ÞðkÞ denote the set of subalgebras or ideals in X  ðkÞ of codimension n in
L n k½½t . It is shown in [5] why An ðX  Þ is a constructible set of the Grassmannian
GrðL n k½½t =t n L n k½½t Þ and hence ½An ðX  Þ defines an element of the Grothendieck
ring. The motivic zeta function encodes the subalgebras or ideals of L n k½½t and is de-
fined as follows:
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
18 Sautoy, Subgroups in nilpotent groups

 P
y
PLnQ½½t ðTÞ ¼ ½An ðX  Þ T n A M½½T :
n¼0

Denote by Mloc the ring M½L1 obtained by localization and define by M½T loc the
subring of Mloc ½½T generated by Mloc ½T and the series ð1  L a T b Þ1 with a A Z and b A N.

It was shown in [5] that the power series PLnQ½½t ðTÞ is a rational function belonging to
M½T loc . Let L be the subring of M generated by the Lefschetz motive L.

We make the following:

Definition 5.2. The space of varieties Me ðLÞ (respectively M / ðLÞ) of L counting

subalgebras (respectively ideals) is defined to be the smallest L-submodule of M containing
the coe‰cients ½An ðX  Þ .


The rationality of PLnQ½½t ðTÞ implies that this module is in fact finitely generated.

In [5] it is shown that by ‘‘taking the trace of Frobenius’’ of the motivic zeta function
 
of PLnQ½½t ðTÞ that one can recover the local zeta functions zLnZ p
ðsÞ for almost all primes.
Let us explain this in more detail. For any variety X over Q, one can choose a model X of
X over Z, and consider the number of points np ðX Þ of the reduction of X modulo p, for p a
prime number. Of course, for some prime numbers p, np ðX Þ may depend of the model X,
but, for a given X , the numbers np ðX Þ are well defined for almost all p. If we denote by P
the set of all primes, the sequenceQ Þ is well defined as an element of the ring ZP0 , where,
np ðXL
for any ring R, we set R P0 :¼ R= R. Moreover, counting points being additive for
pAP pAP
disjoint unions and multiplicative for products, the sequence np ðX Þ in ZP0 only depends
on the class of X in M and may be extended to a ring morphism n: M ! ZP0 . Setting
np ðL1 Þ ¼ 1=p, one may extend uniquely n to a ring morphism n: Mloc ! QP0 .

What is the relationship then between M  ðLÞ and the varieties that we have to count

points on mod p to calculate zLnZ p
ðsÞ?

Let Ei (i A T, T finite) be the subvarieties arising from the resolution of singularities

of the polynomial FL ðX Þ associated to a presentation of L that arose in proving the explicit

expression for zLnZ p
ðsÞ in [4]. (Note that the polynomial FL ðX Þ depends on a choice of
presentation for L and we also have a choice in the resolution we take in general.) As
proved in [5] the proof of the explicit formula in [4] translates into an explicit expression for

PLnQ½½t ðTÞ of the same form
 P
PLnQ½½t ðTÞ ¼ ½EI WI ðL; TÞ
I LT

T S
where WI ðX ; Y Þ A QðX ; Y Þ are rational functions, and EI ¼ Ei and EI ¼ EI n Ej .
iAI j A TnI

When I ¼ j, we have Ej ¼ Y . Note however that the definition of PLnQ½½t ðTÞ is indepen-
dent of any choices made in the calculation.

Hence M  ðLÞ is contained in the L-submodule generated by the varieties Ei ði A TÞ

which arise from a resolution of singularities of the polynomial FL ðX Þ defined from a pre-
sentation for L.
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 19
/
For the Lie algebra LðEÞ one can follow the calculation above for zLðEÞnZ p
ðsÞ to show
that it formally translates into the same result for the motivic zeta function of LðEÞ:
/
Theorem 5.3. PLðEÞnQ½½t ðTÞ ¼ P1 ðL; TÞ þ ½E P2 ðL; TÞ.
 
Corollary 5.4. M / LðEÞ is the L-submodule generated by ½E . The elliptic curve ½E
is canonically associated with LðEÞ and hence GðEÞ.

In [3] we analysed the subalgebras of index p 3 in LðEÞ to show that ze

LðEÞnZp ðsÞ is also
not finitely uniform. Although the conditions are little more complex, I would conjecture
the following:

Conjecture 5.5. There exist two rational functions P1 ðX ; Y Þ and P2 ðX ; Y Þ A QðX ; Y Þ

such that for almost all primes p:
s s
ze
LðEÞ; p ðsÞ ¼ P1 ð p; p Þ þ jEðFp ÞjP2 ð p; p Þ

and
e
PLðEÞnQ½½t ðTÞ ¼ P1 ðL; TÞ þ ½E P2 ðL; TÞ:
 
Hence Me LðEÞ is the additive subgroup generated by ½E and the Lefschetz motive L.

The method of considering minors of matrices developed in section 3 certainly has

some prospect for generalization. It seems likely that it will provide the key for example to
determining for a general class two nilpotent group or Lie algebra a candidate for the
smallest subring of the Grothendieck ring required to express the associated motivic zeta
function counting ideals or normal subgroups. In other words, to count these ideals or
normal subgroups can be reduced to counting points on the varieties from a resolution of
singularities of the polynomial defined by a product of the minors of these matrices.

Also the method of constructing a presentation of a Lie algebra from the determinants
of certain matrices seems also ripe for generalization. For example, suppose that we have
a variety V defined as the solution set of a homogeneous  polynomial
 f ðX1 ; . . . ; Xn Þ A Z½X of
degree r. Suppose we can define an r r matrix gij ðX Þ where gij ðX Þ are linear poly-
nomials with coe‰cients in Z with the property that
 
f ðX1 ; . . . ; Xn Þ ¼ det gij ðX Þ :

Define a Lie algebra

LðV Þ ¼ hA1 ; . . . ; Ar ; B1 ; . . . ; Br ; X1 ; . . . ; Xn : ½Ai ; Bj ¼ gij ðX Þi

 
where all other commutators are zero. Then the analysis above suggests that M / LðV Þ
should contain the variety ½V . However
 it is also likely to contain all the varieties defined
by the various minors of gij ðX Þ . The other point to bare in mind is of course that there
is the possibility that cancelling of varieties might occur which could see the variety V dis-
appear.

This at least raises the interesting problem:

Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
20 Sautoy, Subgroups in nilpotent groups

Problem.  Let f ðX
 1 ; . . . ; Xn Þ A Z½X be a homogeneous polynomial. Does there exist
an r r matrix gij ðX Þ where gij ðX Þ are linear polynomials with coe‰cients in Z with the
property that
 
f ðX1 ; . . . ; Xn Þ ¼ det gij ðX Þ ?
 
Does there exist an m m matrix gij ðX; ZÞ where gij ðX; ZÞ are linear polynomials with
coe‰cients in Z with the property that
 
Z mr f ðX1 ; . . . ; Xn Þ ¼ det gij ðX; ZÞ ?

The second question allows us some more flexibility and we can define an associated
Lie algebra:

Lð f Þ ¼ hA1 ; . . . ; Am ; B1 ; . . . ; Bm ; X1 ; . . . ; Xn ; Z: ðAi ; Bj Þ ¼ gij ðX; ZÞi:

Note that the elliptic curves we have considered of the form E ¼ y 2 þ x 3  Dx all
have complex multiplication. If you would like an example of a nilpotent group which in-
volves counting points mod p on an elliptic curve without complex multiplication, for ex-
ample E ¼ y 2 þ y þ x 3  x, I would conjecture the following:

Conjecture 5.6. Let G be the nilpotent group with presentation:

 
x1 ; x2 ; x3 ; x4 ; x5 ; x6 ; y1 ; y2 ; y3 : ½x1 ; x4 ¼ y3 ; ½x1 ; x5 ¼ y1 ; ½x1 ; x6 ¼ y2 ;
GðEÞ ¼ :
½x2 ; x4 ¼ y1 ; ½x2 ; x5 ¼ y3 ; ½x3 ; x4 ¼ y2 y3 ; ½x3 ; x6 ¼ y1

Then there exist two non-zero rational functions P1 ðX ; Y Þ and P2 ðX ; Y Þ A QðX ; Y Þ such that
for almost all primes p:
/ s s
zGðEÞ; p ðsÞ ¼ P1 ð p; p Þ þ jEðFp ÞjP2 ð p; p Þ

where E is the elliptic curve E ¼ y 2 þ y þ x 3  x without complex multiplication.

Note that the associated Lie algebra has a presentation

 
x1 ; x2 ; x3 ; x4 ; x5 ; x6 ; y1 ; y2 ; y3 : ðx1 ; x4 Þ ¼ y3 ; ðx1 ; x5 Þ ¼ y1 ; ðx1 ; x6 Þ ¼ y2 ;
LðEÞ ¼
ðx2 ; x4 Þ ¼ y1 ; ðx2 ; x5 Þ ¼ y3 ; ðx3 ; x4 Þ ¼ y2 þ y3 ; ðx3 ; x6 Þ ¼ y1

and
 
det ðxi ; x3þj Þ ¼ ðy22 y3 þ y2 y32 þ y13  y1 y32 Þ:

References

[1] M. P. F. du Sautoy, Zeta functions and counting finite p-groups, Electronic Research Announcements of the
American Math. Soc. 5 (1999), 112–122.
[2] M. P. F. du Sautoy, Counting p-groups and nilpotent groups, Inst. Hautes Ét. Sci. Publ. Math. 92 (2000), 63–
112.
Brought to you by | Sackler Library
Authenticated | [Link]
Download Date | 9/3/12 5:03 PM
 Sautoy, Subgroups in nilpotent groups 21

[3] M. P. F. du Sautoy, A nilpotent group and its elliptic curve: non-uniformity of local zeta functions of groups,
Israel J. Math. 126 (2001), 269–288.
[4] M. P. F. du Sautoy and F. J. Grunewald, Analytic properties of zeta functions and subgroup growth, Ann.
Math. 152 (2000), 793–833.
[5] M. P. F. du Sautoy and F. Loeser, Motivic zeta functions of infinite dimensional Lie algebras, École Poly-
technique preprint series 2000-12.
[6] G. Faltings, Endlichkeitssätze für abelsche Varietäten über Zahlkörpern, Invent. Math. 73 (1983), 349–366.
[7] F. J. Grunewald, D. Segal and G. C. Smith, Subgroups of finite index in nilpotent groups, Invent. Math. 93
(1988), 185–223.

Mathematical Institute, 24-29 St Giles, Oxford OX1 3LB

e-mail: dusautoy@[Link]

Eingegangen 1. August 2000

Brought to you by | Sackler Library

Authenticated | [Link]
Download Date | 9/3/12 5:03 PM