Journal of Number Theory 99 (2003) 361–372

http://www.elsevier.com/locate/jnt

Seven-modular lattices and a septic base

Jacobi identity
Heng Huat Chan,a,* Kok Seng Chua,b and Patrick Soléc
a
Department of Mathematics, National University of Singapore, Singapore 117543, Singapore
b
Institute of High Performance Computing, 1 Science Park Road, # 01-01, The Capricorn,
Singapore 117528, Singapore
c
CNRS-I3S, ESSI, Route des Colles, 06 903 Sophia Antipolis, France
Received 19 December 2001; revised 10 July 2002
Communicated by D. Goss

Abstract

A quadratic Jacobi identity to the septic base is introduced and proved by means of modular
lattices and codes over rings. As an application the theta series of all the 6-dimensional 7-
pffiffiffiffiffiffiffi
modular lattices with an Hermitian structure over Qð 7Þ are derived.
r 2002 Elsevier Science (USA). All rights reserved.

1. Introduction

For jqjo1; let

X
N
2
W2 ðqÞ ¼ qðnþ1=2Þ ;
n¼N

X
N
2
W3 ðqÞ ¼ qn
n¼N

*Corresponding author.
E-mail addresses: chanhh@math.nus.edu.sg (H.H. Chan), chuaks@ihpc.nus.edu.sg (K.S. Chua),
ps@essi.fr (P. Solé).

0022-314X/03/$ - see front matter r 2002 Elsevier Science (USA). All rights reserved.
PII: S 0 0 2 2 - 3 1 4 X ( 0 2 ) 0 0 0 6 9 - 0
362 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372

and
X
N
n 2
W4 ðqÞ ¼ ð1Þ qn :
n¼N

One of Jacobi’s famous identities states that

W43 ðqÞ ¼ W42 ðqÞ þ W44 ðqÞ: ð1:1Þ

This identity has been generalized in various ways, in particular in the context of
Vertex Operator Algebras [10].
Around 1991, Borwein and Borwein [4] discovered a cubic analogue of (1.1),
namely,

a3 ðqÞ ¼ b3 ðqÞ þ c3 ðqÞ;

where aðqÞ; bðqÞ; cðqÞ are the bi-dimensional theta series

X 2 þmnþn2
aðqÞ ¼ qm ;
m;nAZ

X 2 þmnþn2
bðqÞ ¼ omn qm ; o ¼ e2pi=3
m;nAZ

and
X 2 þðmþ1=3Þðnþ1=3Þþðnþ1=3Þ2
cðqÞ ¼ qðmþ1=3Þ :
m;nAZ

The aim of this note is to introduce another bi-dimensional generalization of

Jacobi’s identity (1.1). Our main result is

Theorem 1.1. Let

X 2 þmnþ2n2 Þ
AðqÞ ¼ q2ðm ;
m;nAZ

X mn m2 þmnþ2n2
BðqÞ ¼ ð1Þ q
m;nAZ

and
X 2 þðmþ1=2Þnþ2n2 Þ
CðqÞ ¼ q2ððmþ1=2Þ :
m;nAZ
 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372 363

Then

A2 ðqÞ ¼ B2 ðqÞ þ C 2 ðqÞ:

The proof techniques as in [19] are a combination of codes and lattices techniques.
In particular, we will follow Bachoc’s approach [1] for the construction of 7-modular
lattices from codes over the ring F2  F2 :
Like the Borweins’ cubic identity, our base 7 identity was unknown to Ramanujan
[2]. However, when calculating the theta series of a famous 7-modular lattice (related
to the polarization of the Klein curve) we encounter a formula akin to the ones in
[6,17]. It appears that our identity belongs to the septic analogue of Ramanujan’s
theories of elliptic functions to the alternative bases [3].

2. Notations and definitions

2.1. Codes

Let R4 denote the ring with 4 elements F2 þ vF2 where v2 ¼ v: This ring contains
two maximal ideals v and ðv þ 1Þ: Observe that both of R4 =ðvÞ and R4 =ðv þ 1Þ are F2 :
The Chinese Remainder Theorem tells us that

R4 ¼ ðvÞ"ðv þ 1Þ;

so that ring R4 DF2  F2 : An explicit isomorphism is

va þ ðv þ 1Þb/ða; bÞ; a; bAF2 :

A code over R4 is an R4 -submodule of Rn4 :

pffiffiffiffiffiffiffi
Let K :¼ Qð 7Þ be the quadratic number field with ring of integers O ¼ Z½a
;
with a2 þ a þ 2 ¼ 0: Then we can regard R4 as O=ð2Þ when v is the image of a under
reduction modulo 2. Denote by a bar the conjugation which fixes F2 and swaps v and
1 þ v: The natural scalar product induced by the hermitian scalar product of Cn is
then given by
X
xi yi :
i

The Bachoc weight as defined in [1, Definition 3.1] is of course wB ð0Þ ¼ 0: But more
surprisingly wB ðvÞ ¼ wB ð1 þ vÞ ¼ 2 and wB ð1Þ ¼ 1: Define the Bachoc composition of
x say, ni ðxÞ; i ¼ 0; 1; 2; as the number of entries in x of Bachoc weight i: In terms of
Bachoc composition, we have

wB ¼ n1 þ 2n2 :
364 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372

The symmetrized weight enumerator (swe) is then the polynomial of three

variables X ; Y ; Z; defined by
X
sweC ðX ; Y ; ZÞ ¼ X n0 ðxÞ Y n1 ðxÞ Z n2 ðxÞ :
xAC

2.2. Lattices

An n-dimensional lattice L is a discrete subgroup of Rn : Its theta series is

X
yL ðqÞ ¼ qxx ;
xAL

Pn
where x  x ¼ i¼1 x2i ; and q ¼ expðpizÞ; zAH; where

H :¼ fzAC j Im z > 0g:

The lattice L is called integral if it is contained in its dual Ln defined as

Ln :¼ fyARn ; 8xAL; x  yAZg:

A topic of current interest in research is the study of modular lattices. The salient
property of these lattices introduced by Quebbemann [16] is that their theta series is a
modular form for a suitable subgroup of the modular group. Specifically, an integral
pffiffiffi
lattice L is said to be c-modular [1,13] for some prime c if L is isometric to cLn :

Theorem 2.1 (Quebbemann [16, Theorem 7]). Let

Y
N
3 3
D6 ðqÞ :¼ q2 ð1  q2n Þ ð1  q14n Þ :
n¼1

The theta series of an even 7-modular lattice is an isobaric polynomial in the two
pffiffiffi
variables Að qÞ and D6 :

The special cusp form D6 is called a CM-form in [15, (3.b)] where an expansion as a
twisted theta series attached to the quadratic form ½1; 0; 7
is given.

3. Preliminaries

Define the construction AK ðCÞ as the preimage in On of CDRn4 under reduction

modulo 2. Specifically,

AK ðCÞ :¼ fyAOn j y ðmod 2ÞACg:

 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372 365

pffiffiffi
Theorem 3.1. If CDRn4 is a self-dual code then the lattice AK ðCÞ= 2 is even 7-
modular.

Proof. The assertion follows by Bachoc [1, Proposition 3.6] and can alternatively be
derived directly by checking that O is 7-modular for the bilinear form

%
ðx; yÞ/TrK ðxyÞ:

To compute the theta series of a lattice AK ðCÞ as a function of sweC we need to

define some auxiliary theta series. Following [1], we introduce
X
y0 ðqÞ ¼ qxx% ;
xA2O

X
y1 ðqÞ ¼ qxx%
xA1þ2O

and
X
y2 ðqÞ ¼ qxx% :
xAaþ2O

We quote [1, Proposition 4.2] in the case at hand. &

Theorem 3.2 (Bachoc). If CDRn4 is a code of length n; then the theta series of the
lattice AK ðCÞ satisfies

yAK ðCÞ ¼ sweC ðy0 ðqÞ; y1 ðqÞ; y2 ðqÞÞ:

4. Proof of Theorem 1.1

We first express AðqÞ; BðqÞ and CðqÞ in term of Jacobi’s functions Wi ðqÞ; i ¼
1; 2; 3:

Lemma 4.1. Let AðqÞ; BðqÞ and CðqÞ be given as in Theorem 1.1. Then

AðqÞ ¼ W3 ðq2 ÞW3 ðq14 Þ þ W2 ðq2 ÞW2 ðq14 Þ; ð4:1Þ

BðqÞ ¼ W4 ðqÞW4 ðq7 Þ ð4:2Þ

366 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372

and

CðqÞ ¼ W2 ðq2 ÞW3 ðq14 Þ þ W3 ðq2 ÞW2 ðq14 Þ: ð4:3Þ

Proof. Identity (4.1) can be found, for example, in [5, p. 1738]. For the proofs of the
subsequent identities, we will need the following simple identity, namely, for any odd
integer n;
X
N
m n2
ð1Þ qðmþ2Þ ¼ 0: ð4:4Þ
m¼N

This identity follows immediately from [20, p. 464].

Next, note that
X
N
mn m2 þmnþ2n2
BðqÞ ¼ ð1Þ q
m;n¼N

X
N
mþn ðmþnÞ2 þDn2
¼ ð1Þ q 2 4 ;
m;n¼N

with D ¼ 7; and that

X
N
n 2 =4 X
N
m n2
BðqÞ ¼ ð1Þ qDn ð1Þ qðmþ2Þ
n¼N m¼N

X n 2 =4 X
N
m n2
¼ ð1Þ qDn ð1Þ qðmþ2Þ
nA2Z m¼N

X n 2 =4 X
N
m n2
þ ð1Þ qDn ð1Þ qðmþ2Þ
nA2Zþ1 m¼N

X n 2 =4 XN
m n2
¼ ð1Þ qDn ð1Þ qðmþ2Þ ¼ W4 ðqD ÞW4 ðqÞ
nA2Z m¼N

by (4.4).
Finally, rewrite CðqÞ as

X
N
n2 D 2
CðqÞ ¼ q2ððmþ2Þ þ n Þ
4
m;n¼N
X n2 D 2 X n2 D 2
¼ q2ððmþ2Þ þ n Þ
4 þ q2ððmþ2Þþ n Þ
4
mAZ;nA2Z mAZ;nA2Zþ1

¼ W2 ðq ÞW3 ðq Þ þ W3 ðq ÞW2 ðq14 Þ:

2 14 2
&
 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372 367

We proceed to express y0 ðqÞ; y1 ðqÞ; and y2 ðqÞ as functions of AðqÞ; BðqÞ; and CðqÞ:

Proposition 4.2. For all zAH we have

y0 ðqÞ ¼ Aðq2 Þ

and

y1 ðqÞ ¼ Cðq2 Þ:

Proof. If we set x ¼ m  na; then the norm form becomes

xx% ¼ m2 þ mn þ 2n2 ¼: NF ðm; nÞ: ð4:5Þ

If we set x ¼ 1 þ 2ðm  naÞ; then the norm form becomes

2
xx% ¼ 4½ðm þ 1=2Þ þ ðm þ 1=2Þn þ 2n2
: &

In view of these expressions it is natural to look for an expression for y2 involving

q2 : To that end, we shall require the following duplication formulas:

Lemma 4.3. For all zAH we have

AðqÞ ¼ 2Aðq2 Þ  Bðq2 Þ

and

BðqÞ ¼ Aðq2 Þ  Cðq2 Þ:

Proof. From (4.1) and (4.3), we find, after some simplification, that

Aðq2 Þ  Cðq2 Þ ¼ ðW3 ðq4 Þ  W2 ðq4 ÞÞðW3 ðq28 Þ  W2 ðq28 ÞÞ

¼ W4 ðqÞW4 ðq7 Þ ¼ BðqÞ:

This completes the proof of the second assertion. To derive the first assertion, we
claim that

X
N
2 þ2mnþ4n2 Þ
AðqÞ þ Bðq2 Þ ¼ 2 q2ð2m : ð4:6Þ
m;n¼N
368 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372

Indeed
X
N
mn 2 þmnþ2n2 Þ
AðqÞ þ Bðq2 Þ ¼ ð1 þ ð1Þ Þq2ðm
m;n¼N
" #
X X 2 þmnþ2n2 Þ
¼2 þ q2ðm
m;nA2Z m;nA2Zþ1
" #
X X X 2 þmnþ2n2 Þ
¼2 þ þ q2ðm :
mA2Z;nAZ m;nA2Zþ1 mA2Z;nA2Zþ1

From (4.4), we deduce that

!
X X 2 þmnþ2n2 Þ
 q2ðm
m;nA2Zþ1 mA2Z;nA2Zþ1
X 2 þmnþ2n2 Þ
¼ q2ðm
mAZ;nA2Zþ1
X n2 D 2
¼ q2ððmþ2Þ þ n Þ
4 : &
mAZ;nA2Zþ1

Proposition 4.4. For all zAH; we have

y2 ðqÞ ¼ Aðq2 Þ  Bðq2 Þ:

Proof. Writing x ¼ a þ 2ðm  naÞ we see that

2
xx% ¼ 4½m2 þ 2ðn  1=2Þ þ mðn  1=2Þ

or in other words,
X
y2 ðqÞ ¼ qNF ðm;nÞ ;
m;nAZ
m even;n odd

where NF ðm; nÞ is given by (4.5). Introduce for convenience

X
y3 ðqÞ ¼ qNF ðm;nÞ :
m;nAZ
m odd;n odd

Since

Z  Z ¼ fm; nAZ j m even; n eveng,fm; nAZ j m odd; n oddg

,fm; nAZ j m odd; n eveng,fm; nAZ j m even; n oddg;

 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372 369

we may split the sum over ðm; nÞAZ  Z in A and B into four sums and obtain the
following system of two equations in y2 ðqÞ and y3 ðqÞ:

Aðq1=2 Þ ¼ y2 ðqÞ þ Aðq2 Þ þ Cðq2 Þ þ y3 ðqÞ

and

BðqÞ ¼ y2 ðqÞ þ Aðq2 Þ  Cðq2 Þ þ y3 ðqÞ:

Solving for y2 we find that

pffiffiffi
2y2 ðqÞ ¼ Að qÞ  BðqÞ  2Cðq2 Þ:

The result follows by Lemma 4.3. &

We shall require the following lemma.

Lemma 4.5. For all zAH we have

yðpffi2OÞ2 ðqÞ ¼ ð2Aðq2 Þ  Bðq2 ÞÞ :

2

Proof. By definition

yðpffi2OÞ2 ðqÞ ¼ A2 ðqÞ:

Applying Lemma 4.3 for AðqÞ; we complete the proof of the lemma.
We now complete the proof of Theorem 1.1.

Proof of Theorem 1.1. By Scharlau [18] there exists—up to isometry—a unique 7-

modular lattice of dimension 4 over Z: An immediate candidate is O2 : By Theorem
pffiffiffi
3.1 another candidate is AK ðC2 Þ= 2 where C2 is the length 2 self-dual code with
generator matrix ½1; 1
:
Now the swe of that code is computed in [1, p. 102] and evaluated as

sweC2 ðX ; Y ; ZÞ ¼ X 2 þ Y 2 þ 2Z 2 :

The theta series of AK ðC2 Þ can then be computed on applying Theorem 3.2. By the
preceding discussion it should equal yðpffi2OÞ2 ðqÞ computed in Lemma 4.5 as a function
of AðqÞ; BðqÞ; and CðqÞ: This yields
2 2
ð2Aðq2 Þ  Bðq2 ÞÞ ¼ A2 ðq2 Þ þ C 2 ðq2 Þ þ 2ðAðq2 Þ  Bðq2 ÞÞ ;

which reduces to the desired identity. &

370 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372

5. Seven-modular lattices in dimension 6

5.1. Extremal case

The lattice called

* A26 in Craig’s notation [8, Chapter 8, p. 223],
* P6 in Barnes notation [13, p. 131],
* J in Cohen’s notation [7]
is registered as a perfect 7-modular lattice in dimension 3 over O of determinant 73 ;
kissing number 42 and norm 4 in Nebe–Sloane Catalogue of lattices [11]. According
to [18] an extremal (i.e. norm 4) 7-modular lattice in dimension 3 over O is unique.
The construction of [14, Section 3, p. 237] attributed to Serre shows that A26 ¼
AK ðC3 Þ: Here C3 ¼ vR3 þ ð1 þ vÞR> 3 where R3 stands for the binary linear code of
generator ½1; 1; 1
: It occurred [14] in relation to the Jacobian of the Klein curve.
Another geometric construction, using Mordell–Weil lattices can be found in [9]
where the theta series is computed (using Quebbemann’s theorem) as
3
yA2 ðqÞ ¼ AðqÞ  6D6 ðqÞ:
6

Combining this information with [1] where it is shown that sweC3 ðX ; Y ; ZÞ ¼

X 3 þ Z 3 þ 3XZ 2 þ 3ZY 2 ; we obtain

Theorem 5.1. The theta series of A26 is

yA2 ðqÞ ¼ 5A3 ðq2 Þ  9A2 ðq2 ÞBðq2 Þ þ 6Aðq2 ÞB2 ðq2 Þ
6

þ 3C 2 ðq2 ÞAðq2 Þ  3C 2 ðq2 ÞBðq2 Þ  B3 ðq2 Þ:

A third evaluation of yA2 can be obtained by using Ramanujan modular

6 QN
equations to the base 7. Let f ðqÞ :¼ j¼1 ð1  qj Þ: On applying Lemma 2.2 of [5] we
get

f 7 ðq2 Þ f 7 ðq14 Þ
yA2 ðqÞ ¼ þ 7D6 ðqÞ þ 49q2 :
6 f ðq Þ
14 f ðq2 Þ

On the other hand, according to Ranghachari [17, p. 370] the theta series of the
lattice An6 admits the similar expression

f 7 ðq2 Þ f 7 ðq14 Þ
yAn ðqÞ ¼ þ 7D6 ðqÞ þ 7q2 :
6 f ðq Þ
14 f ðq2 Þ

This comes from the fact that both theta series are invariant under G0 ð7Þ with the
same quadratic character [6]. The corresponding space of weight 3 modular forms is
 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372 371

three dimensional and spanned by the three functions

f 7 ðq2 Þ f 7 ðq14 Þ
; D6 ðqÞ and q2 :
f ðq14 Þ f ðq2 Þ

5.2. Norm 2

There are two other self-dual R4 -codes in length 3; namely C3;2 :¼ v/½0; 1; 1
S þ
ðv þ 1Þ/½0; 1; 1
S> with weight enumerator

X 3 þ X 2 Z þ XY 2 þ 2XZ 2 þ Y 2 Z þ 2Z 3

and C3;3 :¼ v/½0; 0; 1

S þ ðv þ 1Þ/½0; 0; 1
S> with weight enumerator

X 3 þ 3X 2 Z þ 3XZ 2 þ Z 3 :

Since AK ðC3;2 Þ and AK ðC3;3 Þ seem to have the same theta series

ð1 þ 6q2 þ 24q4 þ 56q6 þ 114q8 þ 168q10 þ 280q12

þ 294q14 þ 444q16 þ Oðq18 ÞÞ

we are led to conjecture the cubic relation

 2A3 ðq2 Þ þ A2 ðq2 ÞBðq2 Þ þ 2Aðq2 ÞC 2 ðq2 Þ þ 2Aðq2 ÞB2 ðq2 Þ

 C 2 ðq2 ÞBðq2 Þ  B3 ðq2 Þ ¼ 0

which is equivalent to

ð2Aðq2 Þ  Bðq2 ÞÞðA2 ðq2 Þ  B2 ðq2 Þ  C 2 ðq2 ÞÞ ¼ 0;

which is certainly true.

Proposition 5.2. The lattices AK ðC3;2 Þ and AK ðC3;3 Þ are isometric. Their theta series is

8A3 ðq2 Þ  12A2 ðq2 ÞBðq2 Þ þ 6Aðq2 ÞB2 ðq2 Þ  B3 ðq2 Þ:

Proof. The first assertion follows by inspection of Schulze–Pillot’s database of

Hermitian lattices [12]. The second assertion follows on applying Theorem 2.2 to
sweC3;3 :
To conclude there are exactly two classes of O lattices in (real) dimension 6 and
they can both be constructed by using codes over F2  F2 : &
372 H.H. Chan et al. / Journal of Number Theory 99 (2003) 361–372

References

[1] C. Bachoc, Application of coding theory to the construction of modular lattices, J. Combin. Theory
Ser. A 78 (1997) 92–119.
[2] B.C. Berndt, email to Heng Huat Chan, September 2001.
[3] B.C. Berndt, S. Bhargava, F.G. Garvan, Ramanujan’s theories of elliptic functions to alternative
bases, Trans. Amer. Math. Soc. 347 (11) (1995) 4163–4244.
[4] J.M. Borwein, P.B. Borwein, A cubic counterpart of Jacobi’s identity and the AGM, Trans. Amer.
Math. Soc. 323 (1991) 691–701.
PN 2 2
[5] H.H. Chan, Y.L. Ong, On Eisenstein series and m;n¼N qm þmnþ2n ; Proc. Amer. Math. Soc. 127 (6)
(1999) 1735–1744.
[6] K.S. Chua, The root lattice An ; and Ramanujan’s circular summation of theta functions, Proc. Amer.
Math. Soc. 130 (1) (2002) 1–8.
[7] A.M. Cohen, Finite complex reflection groups, Ann. Sci. École Norm. Sup. (4) 9 (3) (1976) 379–436.
[8] J.H. Conway, N.J.A. Sloane, Sphere packings, lattices and groups, third edition, in: Grundlehren der
Mathematischen Wissenschaften, Vol. 290, Springer, New York, 1999.
[9] N.D. Elkies, The Klein quartic in coding theory, in: S. Levy (Ed.), The Eightfold Way, Vol. 35,
Mathematical Sciences Research Institute Publications, Cambridge University Press, Cambridge,
1999.
[10] I. Frenkel, J. Lepowsky, A. Meurman, Vertex Operator Algebras and the Monster, in: Pure and
Applied Mathematics, Vol. 134, Academic Press, Boston, MA, 1988.
[11] http://www.research.att.com/~njas/lattices/P6.5.html
[12] http://www.math.uni-sb.de/~ag-schulze/Hermitian-lattices/D-7/
[13] J. Martinet, Les réseaux parfaits des espaces euclidiens, Mathématiques, Masson, Paris, 1996.
[14] B. Mazur, Arithmetic on curves, Bull. Amer. Math. Soc. 14 (1986) 207–259.
[15] K. Ono, On the circular summation of the eleventh powers of Ramanujan’s theta function, J. Number
Theory 76 (1999) 62–65.
[16] H-G. Quebbemann, Modular lattices in Euclidean spaces, J. Number Theory 54 (2) (1995) 190–202.
[17] S.S. Rangachari, On a result of Ramanujan on theta functions, J. Number Theory 48 (1994) 364–372.
[18] R. Scharlau, B. Hemkemeier, Classification of integral lattices with large class number, Math.
Comput. 67 (1998) 737–749.
[19] P. Solé, P. Loyer, Un lattices, Construction B; and AGM iterations, European J. Combin. 19 (1998)
227–236.
[20] E.T. Whittaker, G.N. Watson, A Course of Modern Analysis, Cambridge University Press,
Cambridge, 1996.