Powers of Euler’s Product and Related Identities

Shaun Cooper, Michael Hirschhorn and Richard Lewis

January 25, 2001

1 Introduction
Ramanujan [17] made the conjectures that

p(5n+4) ≡ 0 (mod 5), p(7n+5) ≡ 0 (mod 7) and p(11n+6) ≡ 0 (mod 11),

where p(n) denotes the number of partitions of n (see [7], Chapter 19). He gave
(loc. cit.) simple proofs of the mod 5 and mod 7 congruences, and these proofs
are reproduced in [7]. They involve showing that if the ar (n) are defined by

X ∞
Y
(q)r∞ = ar (n)q n , where (q)∞ = (1 − q j ),
n≥0 j=1

then
a4 (5n + 4) ≡ 0 (mod 5),
and
a6 (7n + 5) ≡ 0 (mod 7).
Winquist [18] much later gave a similar proof of the mod 11 congruence by
showing that
a10 (11n + 6) ≡ 0 (mod 11).
Indeed, much more is true. As we shall see,
n
(1) a4 (5n + 4) = −5 a4 ,
5
 
n−1
(2) a6 (7n + 5) = 49 a6
7
and  
4 n−4
(3) a10 (11n + 6) = 11 a10 .
11

1
In these and all subsequent formulas, n is assumed to be an integer and ar (m/p)
is taken to be zero whenever m/p is not an integer.

These results can be generalised further. For example, it can be shown that
if p ≡ 5 (mod 6) is prime,

p2 − 1
   
n
(4) a4 pn + = −p a4 ,
6 p
if p ≡ 3 (mod 4) is prime,

p2 − 1
   
2 n
(5) a6 pn + = p a6 ,
4 p
while if p ≡ 11 (mod 12) is prime,

5p2 − 5
   
n
(6) a10 pn + = p4 a10 .
12 p

(1) is simply the case p = 5 of (4); (2) follows from (5) by taking p=7 and re-
placing n with n − 1, while (3) follows from (6) by taking p = 11 and replacing
n with n − 4.

Newman [13] generalised equations (4) – (6) and proved the following result.

Theorem (Newman) Suppose that r is one of the numbers 2, 4, 6, 8, 10, 14, 26.
Let p be a prime > 3 such that r(p + 1) ≡ 0 (mod 24). Let ∆ = (p2 − 1)/24,
and define ar (α) = 0 if α is not a non-negative integer. Then
 
(r/2)−1 n
ar (np + r∆) = (−p) ar .
p
Furthermore, there are no other values of r for which the theorem is true.

Newman’s proof of this theorem uses the theory of elliptic modular functions.

The purposes of this paper are as follows. First, we give an explicit, elementary
proof of Newman’s theorem for all cases except r = 26. Our proofs are explicit in
the sense that they utilise explicit series expansions for certain infinite products
which arise from Macdonald identities of ranks 1 and 2. They are elementary in
that they avoid the theory of elliptic modular functions. Our proof also shows
that more is true than Newman claimed for the cases r = 2, 10 and 14, and we
also obtain results for r = 1 and 3. We have not been able to use our methods
to prove the case r = 26.

Second, we obtain new results for infinite products of the types (q)r∞ (q 2 )s∞ ,

2
(q)r∞ (q 3 )s∞ and (q)r∞ (q 4 )s∞ for various integers r and s. These results arise out
of considering all of the Macdonald identities of ranks 1 and 2. Results for anal-
ogous infinite products arising from Macdonald identities of ranks higher than
2 appear not to exist.

Newman [14, 15] also considered products of the form (q)r∞ (q j )s∞ , although his
results are different. There has also been work on classifying eta-products whose
Fourier coefficients are multiplicative. See, for example, [2], [4, pp. 77–85], [5],
[6], [11] and [12]; some of these eta products arise in Conway and Norton’s [1]
“monstrous moonshine”.

We conclude with a list of conjectures that were suggested by computer search.

2 Statement of Results
Define the ceiling function by

dxe = min k
k≥x

where the minimum is taken over all integers k.

Theorem 1. Suppose r is an X integer and p is prime. Let ∆ = (p2 − 1)/24, let

j = dr/2e − 1 and let (q)r∞ = a(n)q n . Then the coefficients a(n) satisfy
n≥0
 
n j
a(pn + r∆) = p a
p

for the following values of r, p and :

r p 
1 if p≡1 or 11 (mod 12),
1 p>3
 −1 if p≡5 or 7 (mod 12)
1 if p≡7 or 11 (mod 12),
2 p ≡ 5, 7 or 11 (mod 12)
 −1 if p≡5 (mod 12)
1 if p≡1 (mod 4),
3 p>2
−1 if p≡3 (mod 4)
4 p ≡ 5 (mod 6) −1
6 p ≡ 3 (mod 4) 1
8 p ≡ 2 (mod 3) −1
10 p ≡ 7 or 11 (mod 12) 1
1 if p ≡ 11 (mod 12),
14 p ≡ 5 (mod 6)
−1 if p ≡ 5 (mod 12)

3
Theorem 2. Suppose r and s are integers andX p is prime. Let ∆ = (p2 − 1)/24,
r 2 s
let j = d(r + s)/2e − 1 and let (q)∞ (q )∞ = a(n)q n . Then the coefficients
n≥0
a(n) satisfy  
n
a(pn + (r + 2s)∆) = pj a
p
for the following values of r, s, p and :

(r, s) p 
1 if p ≡ 5 or 7 (mod 8)
(1, 1) p ≡ 3, 5 or 7 (mod 8)
−1 if p ≡ 3 (mod 8)
(2, 2) p ≡ 3 (mod 4) −1
(3, 3) p ≡ 5 or 7 (mod 8) 1
1 if p ≡ 7 (mod 8)
(5, 5) p ≡ 3 (mod 4)
−1 if p ≡ 3 (mod 8)
(4, 2), (2, 4) p ≡ 7 or 11 (mod 12) 1
1 if p ≡ 13 or 19 (mod 24)
(3, 1), (1, 3) p ≡ 13, 17, 19 or 23 (mod 24)
 −1 if p ≡ 17 or 23 (mod 24)
1 if p ≡ 13 or 23 (mod 24)
(5, 1), (1, 5) p ≡ 13, 17, 19 or 23 (mod 24)
−1 if p ≡ 17 or 19 (mod 24)
(2, −1), (−1, 2) p > 2 1
1 if p ≡ 13 or 23 (mod 24)
(3, −1), (−1, 3) p ≡ 13, 17, 19 or 23 (mod 24)
 −1 if p ≡ 17 or 19 (mod 24)
1 if p≡5 (mod 8)
(5, −1), (−1, 5) p ≡ 5 or 7 (mod 8)
−1 if p≡7 (mod 8)
(4, −2), (−2, 4) p ≡ 3 (mod 4) 1
1 if p ≡ 1 (mod 6)
(5, −2), (−2, 5) p > 3
−1 if p ≡ 5 (mod 6)
(6, −2), (−2, 6) p ≡ 7 or 11 (mod 12) −1

1 if p ≡ 11 (mod 12)
(8, −2), (−2, 8) p ≡ 5 (mod 6)
 −1 if p≡5 (mod 12)
1 if p ≡ 13 or 19 (mod 24)
(7, −3), (−3, 7) p ≡ 13, 17, 19 or 23 (mod 24)
 −1 if p ≡ 17 or 23 (mod 24)
1 if p≡7 (mod 8)
(9, −3), (−3, 9) p ≡ 3 (mod 4)
−1 if p≡3 (mod 8)
(10, −4), (−4, 10) p ≡ 7 or 11 (mod 12) 1
(14, −4), (−4, 14) p ≡ 3 (mod 4) 1
1 if p ≡ 7 (mod 8)
(11, −5), (−5, 11) p ≡ 7 or 11 (mod 12)
−1 if p ≡ 3 (mod 8)

Theorem 3. Suppose r and s are integers andX p is prime. Let ∆ = (p2 − 1)/24,
r 3 s
let j = d(r + s)/2e − 1 and let (q)∞ (q )∞ = a(n)q n . Then the coefficients
n≥0

4
a(n) satisfy  
n
j
a(pn + (r + 3s)∆) = p a
p
for the following values of r, s, p and :

(r, s) p 
(1, 1) p ≡ 5 (mod 6) 1
(3, 3) p ≡ 5 (mod 6) 1
(7, 7) p ≡ 5 (mod 6) 1
1 if p ≡ 7 (mod 12)
(3, 1), (1, 3) p ≡ 7 or 11 (mod 12)
−1 if p ≡ 11 (mod 12)

Theorem 4. Suppose r and s are integers andX p is prime. Let ∆ = (p2 − 1)/24,
r 4 s
let j = d(r + s)/2e − 1 and let (q)∞ (q )∞ = a(n)q n . Then the coefficients
n≥0
a(n) satisfy  
n
a(pn + (r + 4s)∆) = pj a
p
for the following values of r, s, p and :

(r, s) p 
(1, 1) p≡7 or 11 (mod 12) 1
(2, 2) p≡7 or 11 (mod 12) −1
(3, 3) p≡3 (mod 4) 1
(3, 1), (1, 3) p≡5 (mod 6) −1
(5, −1), (−1, 5) p≡5 (mod 6) −1

3 Proofs
The following specialisations of Macdonald identities, listed in the appendix of
[10], will be used. Each formula below is stated together with its associated
root system. Thus, for example, the first formula is Macdonald’s specialisation
(c) for the root system BC1 .

5
 X 2
f1 (q) := (q)∞ = (−1)(α−1)/6 q (α −1)/24
BC1 (c)
α≡1 (mod 6)
X 2
f2 (q) := (q)3∞ = αq (α −1)/8
A1
α≡1 (mod 4)
α α2
X
f3 (q) := (q)2∞ /(q 2 )∞ = (−1) q B1 (c)
α X 2
f4 (q) := (q 2 )2∞ /(q)∞ = q (α −1)/8
B1 (b)
α≡1 (mod 4)
X 2
f5 (q) := (q)5∞ /(q 2 )2∞ = αq (α −1)/24
BC1 (a)
α≡1 (mod 6)
X 2
f6 (q) := (q 2 )5∞ /(q)2∞ = (−1)α−1 αq (α −1)/3
BC1 (d)
α≡1 (mod 3)
1 X 2
−αβ+β 2 −1)/3
f7 (q) := (q)8∞ = (α + β)(2α − β)(2β − α)q (α A2
2
α≡1 (mod 3)
β≡1 (mod 3)
1 X 2
+β 2 −5)/12
f8 (q) := (q)10
∞ = αβ(α2 − β 2 )q (α B2 (a)
6
α≡2 (mod 6)
β≡1 (mod 6)
1 X 2
+β 2 −10)/16
f9 (q) := (q)5∞ (q 2 )5∞ = αβ(α2 − β 2 )q (α B2∨ (a)
24
α≡3 (mod 8)
β≡1 (mod 8)
1 X 2
+β 2 −10)/40
f10 (q) := (q)14 2 4
∞ /(q )∞ = αβ(α2 − β 2 )q (α BC2 (a)
24
α≡3 (mod 10)
β≡1 (mod 10)
1 X 2
+β 2 −5)/5
f11 (q) := (q 2 )14 4
∞ /(q)∞ = (−1)α+β+1 αβ(α2 − β 2 )q (α BC2 (d)
6
α≡2 (mod 5)
β≡1 (mod 5)
f12 (q) := (q)14
∞
1 X 2
+αβ+β 2 −21)/36
= αβ(α + β)(α − β)(2α + β)(α + 2β)q (α G2
3240
α≡4 (mod 12)
β≡1 (mod 12)
f13 (q) := (q)7∞ (q 3 )7∞
1 X 2
+αβ+β 2 −7)/6
= αβ(α + β)(α − β)(2α + β)(α + 2β)q (α G∨
2
120
α≡2 (mod 6)
β≡1 (mod 6)

Formulas for f1 , · · · , f4 are classical consequences of the Jacobi triple product

identity. Formulas for f5 and f6 are consequences of the quintuple product
identity, and are stated explicitly by Ramanujan [16] and Macdonald [10]. As
noted by Dyson [3], the identity for f7 goes back to Klein and Fricke [9, p. 373].
The series for f8 is originally due to Winquist [18]; also see [8]. The one for f12
is due to Atkin (unpublished), and Dyson [3, p. 651] gives Atkin’s formula for
some of the coefficients a26 (n) in the expansion of (q)26 ∞.

6
All of Theorems 1 – 4 may be proved either by appealing directly to one of the
specialisations of Macdonald’s identities above, or by combining two of the sin-
gle series identities f1 , · · · , f6 . The following four proofs illustrate our procedure
in detail. The proof of the case r = 2, p ≡ 5 (mod 12) of Theorem 1 is more
complicated than the others, so is given at the end of this section.

Proof of Theorem 1 in the case r = 1.

We have X 2
(q)∞ = f1 (q) = (−1)(α−1)/6 q (α −1)/24 .
α≡1 (mod 6)

Thus X X
a(n) = (−1)(α−1)/6 = (−1)(α−1)/6 .
α≡1 (mod 6) α≡1 (mod 6)
(α2 −1)/24=n α2 =24n+1

Therefore
p2 − 1
  X
a(pn + ∆) = a pn + = (−1)(α−1)/6 .
24
α≡1 (mod 6)
α2 =24pn+p2

The condition α2 = 24pn + p2 implies α2 ≡ 0 (mod p) and so α ≡ 0 (mod p).

Let 
1 if p ≡ 1 (mod 6)
λ=
−1 if p ≡ −1 (mod 6)
Let α = λpα0 . Then α0 is an integer. Also, α0 ≡ 1 (mod 6) since λp ≡ 1
(mod 6) and α ≡ 1 (mod 6).
Next, modulo 2,
α−1 λpα0 − 1
=
6 6
α0 − 1 (λp − 1)α0
= +
6 6
α0 − 1 0p−λ
= + λα
6 6
α0 − 1 p − λ p−λ
= + + (λα0 − 1)
6 6 6
α0 − 1 p − λ
≡ + .
6 6
2
Furthermore, the condition α2 = 24pn + p2 becomes α0 = 24n/p + 1. Therefore

p2 − 1
  X
a pn + = (−1)(α−1)/6
24
α≡1 (mod 6)
α2 =24pn+p2

7
 X 0
= (−1)(p−λ)/6 (−1)(α −1)/6
α0 ≡1 (mod 6)
α0 2 =24n/p+1
 
n
= (−1)(p−λ)/6 a
p
 
n
= a ,
p
where 
1 if p ≡ 1 or 11 (mod 12)
=
−1 if p ≡ 5 or 7 (mod 12).
This completes the proof of Theorem 1 in the case r = 1.

Proof of Theorem 1 in the case r = 4

We have
2
+3β 2 −4)/24
X
(q)4∞ = (q)∞ (q)3∞ = f1 (q)f2 (q) = (−1)(α−1)/6 βq (α .
α≡1 (mod 6), β≡1 (mod 4)

Thus
X X
a(n) = (−1)(α−1)/6 β = (−1)(α−1)/6 β.
α≡1 (mod 6), β≡1 (mod 4) α≡1 (mod 6), β≡1 (mod 4)
(α2 +3β 2 −4)/24=n α2 +3β 2 =24n+4

Therefore
p2 − 1
  X
a(pn + 4∆) = a pn + = (−1)(α−1)/6 β.
6
α≡1 (mod 6), β≡1 (mod 4)
α2 +3β 2 =24pn+4p2

The
 condition
 α2 + 3β 2 = 24pn + 4p2 implies α2 + 3β 2 ≡ 0 (mod p). Since
−3
= −1 [7, Theorem 96], this implies α, β ≡ 0 (mod p). If p ≡ 5
p
(mod 12), let α = −pα0 , β = pβ 0 , while if p ≡ 11 (mod 12), let α = −pα0 ,
β = −pβ 0 . Then α0 , β 0 are integers and α0 ≡ 1 (mod 6), β 0 ≡ 1 (mod 4).
If p ≡ 5 (mod 12) then, modulo 2,
α−1 −pα0 − 1 α0 − 1 p+1 α0 − 1 α0 − 1
= = − α0 ≡ + α0 ≡ +1
6 6 6 6 6 6
and 0
(−1)(α−1)/6 β = −(−1)(α −1)/6 pβ 0 ,
while if p ≡ 11 (mod 12),
α−1 −pα0 − 1 α0 − 1 p+1 α0 − 1
= = − α0 ≡
6 6 6 6 6

8
and once again, 0
(−1)(α−1)/6 β = −(−1)(α −1)/6 pβ 0 .
2 2
Furthermore, the condition α2 +3β 2 = 24pn+4p2 becomes α0 +3β 0 = 24n/p+
4. Therefore
p2 − 1
  X
a pn + = (−1)(α−1)/6 β
6
α≡1 (mod 6), β≡1 (mod 4)
α2 +3β 2 =24pn+4p2
X 0
= −p (−1)(α −1)/6 β 0
α0 ≡1 (mod 6), β 0 ≡1 (mod 4)
α0 2 +3β 02 =24n/p+4
 
n
= −p a .
p
This completes the proof of the case r = 4 of Theorem 1.

Proof of Theorem 1 in the case r = 8

We have
1 X 2 2
(q)8∞ = f7 (q) = (2α − β)(2β − α)(α + β)q (α −αβ+β −1)/3 .
2
α,β≡1 (mod 3)

Therefore
1 X
a(n) = (2α − β)(2β − α)(α + β)
2
α,β≡1 (mod 3)
α2 −αβ+β 2 =3n+1

and so
p2 − 1
 
1 X
a(pn + 8∆) = a pn + = (2α − β)(2β − α)(α + β).
3 2
α,β≡1 (mod 3)
α2 −αβ+β 2 =3pn+p2

Consider first the case p = 2. The right hand side of α2 − αβ + β 2 = 3pn + p2 is

even, and this implies α and β are both even. Let α = −2α0 , β = −2β 0 . Then
2 2
α0 and β 0 are integers, α0 , β 0 ≡ 1 (mod 3) and α0 − α0 β 0 + β 0 = 3n/2 + 1.
Therefore
1 X
a(2n + 1) = −8(2α0 − β 0 )(2β 0 − α0 )(α0 + β 0 )
2 0 0
α ,β ≡1 (mod 3)
α0 2 −α0 β 0 +β 02 =3n/2+1
n
= −8a .
2
Now consider the case p ≡ 2 (mod 3), p > 2, p prime. The equation α2 −
αβ + β 2 = 3pn + p2 is equivalent to (α + β)2 + 3(α − β)2 = 12pn + 4p2 and so

9
  
2 2 −3
(α + β) + 3(α − β) ≡ 0 (mod p). Since = −1, [7, Theorem 96] this
p
implies α + β ≡ 0 (mod p) and α − β ≡ 0 (mod p). Consequently α, β ≡ 0
(mod p). Let α = −pα0 , β = −pβ 0 . Then α0 , β 0 are integers, α0 , β 0 ≡ 1
2 2
(mod 3) and α0 − α0 β 0 + β 0 = 3n/p + 1. Therefore
p2 − 1
 
1 X
a pn + = −p3 (2α0 − β 0 )(2β 0 − α0 )(α0 + β 0 )
3 2 0 0
α ,β ≡1 (mod 3)
α0 2 −α0 β 0 +β 02 =3n/p+1
 
n
3
= −p a .
p
Combining the two cases, we see that if p ≡ 2 (mod 3) is prime, then
p2 − 1
   
3 n
a pn + = −p a .
3 p
This completes the proof of Theorem 1 in the case r = 8.

Proof of Theorem 2 in the case (r, s) = (1, 5)

We have
(q 2 )5∞ X 2
+8β 2 −11)/24
(q)∞ (q 2 )5∞ = (q)3∞ 2
= f2 (q)f6 (q) = αβ(−1)β−1 q (3α .
(q)∞
α≡1 (mod 4)
β≡1 (mod 3)

Thus X
a(n) = αβ(−1)β−1 .
α≡1 (mod 4), β≡1 (mod 3)
3α2 +8β 2 =24n+11

Therefore
11p2 − 11
  X
a(pn + 11∆) = a pn + = αβ(−1)β−1 .
24
α≡1 (mod 4), β≡1 (mod 3)
3α2 +8β 2 =24pn+11p2

The condition 3α2 + 8β 2 = 24pn + 11p2 implies 3α2 + 8β 2 ≡ 0 (mod p),

therefore 6α2 + (4β)2 ≡ 0 (mod p). By [7, Theorems 85, 93 and 96], we have
     
−6 2 −3 1 if p ≡ 1, 5, 7 or 11 (mod 24)
= =
p p p −1 if p ≡ 13, 17, 19 or 23 (mod 24).

Therefore p ≡ 13, 17, 19 or 23 (mod 24) implies α ≡ 0 (mod p) and 4β ≡ 0

(mod p), and consequently β ≡ 0 (mod p). Let
pα0 pβ 0
 
if p ≡ 1 (mod 4) if p ≡ 1 (mod 6)
α= 0 ,β = .
−pα if p ≡ −1 (mod 4) −pβ 0 if p ≡ −1 (mod 6)

10
 2 2
Then α0 ≡ 1 (mod 4), β 0 ≡ 1 (mod 6) and 3α0 + 8β 0 = 24n/p + 11. Fur-
0
thermore β ≡ β 0 (mod 2), so αβ(−1)β−1 = p2 α0 β 0 (−1)β −1 , where

1 if p ≡ 13 or 23 (mod 24)
= .
−1 if p ≡ 17 or 19 (mod 24)
Therefore
11p2 − 11
   
X 0 n
a pn + = p2 α0 β 0 (−1)β −1
= p2 a .
24 p
α0 ≡1 (mod 4), β 0 ≡1 (mod 3)
3α0 2 +8β 02 =24n/p+11

This completes the proof of the case (r, s) = (1, 5) of Theorem 2.

Proof of Theorem 1 in the case r = 2

We have
X 2 2
(q)2∞ = f1 (q)2 = (−1)(α+β−2)/6 q (α +β −2)/24 ,
α, β≡1 (mod 6)

thus X
(7) a(n) = (−1)(α+β−2)/6
α, β≡1 (mod 6)
α2 +β 2 =24n+2
and therefore
p2 − 1
  X
a(pn + 2∆) = a pn + = (−1)(α+β−2)/6 .
12
α, β≡1 (mod 6)
α2 +β 2 =24pn+2p2

The condition α2 + β 2 = 24pn + 2p2 implies α2 + β 2 ≡ 0 (mod p).

Case 1 

−1
If p ≡ 7 or 11 (mod 12), then = −1 [7, Theorem 82]. Consequently,
p
α, β ≡ 0 (mod p). If p ≡ 7 (mod 12), let α = pα0 , β = pβ 0 , while if
p ≡ 11 (mod 12), let α = −pα0 , β = −pβ 0 . Then α0 and β 0 are integers,
0 0
α0 , β 0 ≡ 1 (mod 6) and (−1)(α+β−2)/6 = (−1)(α +β −2)/6 . Furthermore, the
2 2
condition α2 + β 2 = 24pn + 2p2 becomes α0 + β 0 = 24n/p + 2. Therefore
p2 − 1
  X
a pn + = (−1)(α+β−2)/6
12
α, β≡1 (mod 6)
α2 +β 2 =24pn+2p2
X 0 0
= (−1)(α +β −2)/6

0 0
α , β ≡1 (mod 6)
α0 2 +β 02 =24n/p+2
 
n
= a .
p

11
Case 2  
−1
If p ≡ 5 (mod 12), then = +1, and the condition α2 +β 2 ≡ 0 (mod p)
p
does not imply α, β ≡ 0 (mod p). Therefore a different method is required.
Case 2 is a corollary of the following Lemma.

LemmaX Suppose p ≡ 5 (mod 12) is prime, let ∆k = (p2k − 1)/24, and let
2
(q)∞ = a(n)q n . Then the coefficients a(n) satisfy
n≥0
(i) a(p2k n + 2∆k ) = (−1)k a(n) if 12n + 1 6≡ 0 (mod p),
(ii) a(p2k+1 n + 2∆k+1 ) = 0 if n 6≡ 0 (mod p).

p2 − 1
Theorem (Case 2) With p ≡ 5 (mod 12), ∆ = ∆1 = and (q)2∞ =
X 24
a(n)q n ,
n≥0
 
n
a(pn + 2∆) = −a .
p
Proof of Theorem  
n
If p 6 | n then by part (ii) of the Lemma, a(pn + 2∆) = 0 = −a .
p
If p|n, we can write n = pn0 . We are then trying to show

a(p2 n0 + 2∆) = −a(n0 ).

We can write 12n0 + 1 = pr m with m 6≡ 0 (mod p) for some r ≥ 0.

Suppose r = 2k. Then m ≡ 1 (mod 12) and
  2k  
2 0 2 p m−1
a(p n + 2∆) = a p + 2∆
12
  2k
p2 − 1
 
2 p m−1
= a p +
12 12
2k+2
   
2k+2 m−1 p −1
= a p +
12 12
   
2k+2 m − 1
= a p + 2∆k+1
12
 
m−1
= (−1)k+1 a
12

while
p2k m − 1
 
a(n0 ) = a
12

12
 p2k − 1
   
2k m−1
= a p +
12 12
   
2k m−1
= a p + 2∆k
12
 
m−1
= (−1)k a .
12
On the other hand, if r = 2k + 1 then m ≡ p (mod 12) and
  2k+1
p2 − 1
 
p m−1
a(p2 n0 + 2∆) = a p2 +
12 12
2k+4
   
m−p p −1
= a p2k+3 +
12 12
   
m−p
= a p2k+3 + 2∆k+2
12
= 0
and
p2k+1 m − 1
 
0
a(n ) = a
12
p2k+2 − 1
   
2k+1 m−p
= a p +
12 12
   
2k+1 m − p
= a p + 2∆k+1
12
= 0.
This completes the proof of the Theorem.

Proof of Lemma
We have X
a(n) = (−1)(α1 +α2 −2)/6 .
α1 ≡1, α2 ≡1 (mod 6)
(α21 +α22 −2)/24=n

Now let the Gaussian integer α be defined by α = α1 + iα2 and let N(α) =
α21 + α22 = αα, tr(α) = α1 + α2 and ε(α) = (−1)(tr(α)−2)/6 .
Then X
a(n) = ε(α).
α≡1+i (mod 6)
N(α)=24n+2

Also, X
a(p2k n + 2∆k ) = ε(α)
α≡1+i (mod 6)
2k
N(α)=p (24n+2)

13
since
p2k − 1
24(p2k n + 2∆k ) + 2 = 24(p2k n + ) + 2 = p2k (24n + 2).
12
So now we must consider the equation
2k
N(α) = p (24n + 2),
or,
αα = p2k (24n + 2).
Now it can be shown that p ≡ 5 (mod 12) can be uniquely factored
p = ππ
with π ≡ 1 + 2i (mod 6).
Thus we have to consider
αα = p2k (24n + 2) = π 2k π 2k (24n + 2)
with 12n+1 6≡ 0 (mod p). There are just three possibilities, since factorisation
is (essentially) unique in Z[i]. They are: both π and π|α, π 6 | α and π 6 | α.
In the first case, p|α, so write α = −pα0 . Then α0 ≡ 1 + i (mod 6), N(α0 ) =
p2k−2 (24n + 2) and ε(α) = ε(α0 ).
In the second, we have π 2k |α, so write
α = −iπ 2 α01 , α01 = −iπ 2 α02 , · · · , α0k−1 = −iπ 2 α0 .
Then α01 , α02 , · · · , α0k−1 , α0 ≡ 1 + i (mod 6), N(α0 ) = 24n + 2, π 6 | α0 and
ε(α) = −ε(α01 ) = · · · = (−1)k ε(α0 ).
Note that since 12n + 1 6≡ 0 (mod p), the restriction π 6 | α0 is superfluous.
Similarly in the third case, π 2k |α, so write
α = iπ 2 α01 , α01 = iπ 2 α02 , · · · , α0k−1 = iπ 2 α0 .
Then α01 , α02 , · · · , α0k−1 , α0 ≡ 1 + i (mod 6), N(α0 ) = 24n + 2, π 6 | α0 , again
superfluous, and
ε(α) = −ε(α01 ) = · · · = (−1)k ε(α0 ).

Thus we have
X X X
a(p2k n + 2∆k ) = ε(α) + ε(α) + ε(α)
p|α π6 |α π6 |α
X X
= ε(α0 ) + 2(−1)k ε(α0 )
0 0
α ≡1+i (mod 6) α ≡1+i (mod 6)
2k−2 0
0
N(α )=p (24n+2) N(α )=24n+2

= a(p2(k−1) n + 2∆k−1 ) + 2(−1)k a(n).

14
It now follows by induction on k that a(p2k n + 2∆k ) = (−1)k a(n).

A similar argument shows that if n 6≡ 0 (mod p),

a(p2k+1 n + 2∆k+1 ) = a(p2(k−1)+1 n + 2∆k ) + (−1)k a(pn + 2∆),

and hence by induction on k,

1 + (−1)k
 
a(p2k+1 n + 2∆k+1 ) = a(pn + 2∆).
2
Now, X
a(pn + 2∆) = ε(α)
α≡1+i (mod 6)
N(α)=p(24n+2p)

p2 − 1
since 24(pn + 2∆) + 2 = 24(pn + ) + 2 = 24pn + 2p2 = p(24n + 2p). So
12
we need to consider
N(α) = p(24n + 2p),
or,
αα = p(24n + 2p).
If n 6≡ 0 (mod p) then π|α or π|α, but not both. In the first case, write
α = πα0 , in the second α = −iπα0 . In either case, α0 ≡ 3 + i (mod 6) and
N(α0 ) = 24n + 2p.
So
X X
a(pn + 2∆) = ε(α) + ε(α)
π|α π|α
X X
= ε(πα0 ) + ε(−iπα0 )
0 0
α ≡3+i (mod 6) α ≡3+i (mod 6)
0 0
N(α )=24n+2p N(α )=24n+2p
= 0.

(To see this, consider α0 = a + ib in both sums.) This completes the proof of
the lemma.
√
Much the same argument as the one given here, with the ring Z[ −2] (also
a unique factorisation domain) in place of Z[i], may be used to prove the case
r = s = 1, p ≡ 3 (mod 8) of theorem 2.

4 Remaining Proofs
The remaining cases of Theorems 1 – 4 can be proved in a similar fashion to
the previous section, using f1 , · · · , f13 and various combinations of f1 , · · · , f6 as

15
follows.

For Theorem 1, use

(q)∞ = f1 (q)
(q)2∞ = f1 (q)2
(q)3∞ = f2 (q)
(q)4∞ = f1 (q)f2 (q)
(q)6∞ = f2 (q)2
(q)8∞ = f7 (q)
(q)10
∞ = f8 (q)
(q)14
∞ = f12 (q)

For Theorem 2, use

(q)∞ (q 2 )∞ = f1 (q)f1 (q 2 )
(q)2∞ (q 2 )2∞ = f2 (q 2 )f3 (q)
(q)3∞ (q 2 )3∞ = f2 (q)f2 (q 2 )
(q)5∞ (q 2 )5∞ = f9 (q)

(q)4∞ (q 2 )2∞ = f6 (q 1/2 )f6 (−q 1/2 ) (q)2∞ (q 2 )4∞ = f5 (q 2 )f6 (−q)
(q)3∞ (q 2 )∞ = f1 (q 2 )f2 (q) (q)∞ (q 2 )3∞ = f1 (q)f2 (q 2 )
(q)5∞ (q 2 )∞ = f2 (q 2 )f5 (q) (q)∞ (q 2 )5∞ = f2 (q)f6 (q)
(q)2∞ /(q 2 )∞ = f3 (q) (q 2 )2∞ /(q)∞ = f4 (q)
(q)3∞ /(q 2 )∞ = f1 (q)f3 (q) (q 2 )3∞ /(q)∞ = f1 (q 2 )f4 (q)
(q)5∞ /(q 2 )∞ = f1 (q 2 )f5 (q) (q 2 )5∞ /(q)∞ = f1 (q)f6 (q)
(q)4∞ /(q 2 )2∞ = f3 (q)2 (q 2 )4∞ /(q)2∞ = f4 (q)2
(q)5∞ /(q 2 )2∞ = f5 (q) (q 2 )5∞ /(q)2∞ = f6 (q)
(q)6∞ /(q 2 )2∞ = f1 (q)f5 (q) (q 2 )6∞ /(q)2∞ = f1 (q 2 )f6 (q)
(q)8∞ /(q 2 )2∞ = f2 (q)f5 (q) (q 2 )8∞ /(q)2∞ = f2 (q 2 )f6 (q)
(q)7∞ /(q 2 )3∞ = f3 (q)f5 (q) (q 2 )7∞ /(q)3∞ = f4 (q)f6 (q)
(q)9∞ /(q 2 )3∞ = f2 (q 1/2 )f2 (−q 1/2 ) (q 2 )9∞ /(q)3∞ = f2 (q 4 )f2 (−q)
(q)10 2 4
∞ /(q )∞ = f5 (q)2 (q 2 )10
∞ /(q)∞
4
= f6 (q)2
(q)∞ /(q 2 )4∞
14
= f10 (q) (q )∞ /(q)4∞
2 14
= f11 (q)
(q)11 2 5
∞ /(q )∞ = f5 (q 1/2 )f5 (−q 1/2 ) (q 2 )11
∞ /(q)∞
5
= f6 (q 2 )f5 (−q).
For Theorem 3, use

(q)∞ (q 3 )∞ = f1 (q)f1 (q 3 )
(q)3∞ (q 3 )3∞ = f2 (q)f2 (q 3 )
(q)7∞ (q 3 )7∞ = f13 (q)

(q)3∞ (q 3 )∞ = f1 (q 3 )f2 (q) (q)∞ (q 3 )3∞ = f1 (q)f2 (q 3 )

16
For Theorem 4, use

(q)∞ (q 4 )∞ = f1 (q)f1 (q 4 )
(q)2∞ (q 4 )2∞ = f1 (q 2 )f6 (−q)
(q)3∞ (q 4 )3∞ = f2 (q)f2 (q 4 )

(q)3∞ (q 4 )∞ = f1 (q 4 )f2 (q) (q)∞ (q 4 )3∞ = f1 (q)f2 (q 4 )

(q)5∞ /(q 4 )∞ = f3 (q 2 )f5 (q) (q 4 )5∞ /(q)∞ = f4 (q)f6 (q 2 )

5 Conjectures
The following conjectures were suggested by computer search.

Conjecture 1. Theorem 2 holds for the following additional values of r, s, p

and :
(r, s) p 
(9, 9) p ≡ 7 (mod 8) 1
1 if p ≡ 7 (mod 8)
(7, 3), (3, 7) p ≡ 7 or 11 (mod 12)
−1 if p ≡ 3 (mod 8)
(7, 1), (1, 7) p = 31 −1

1 if p≡7 (mod 24)
(3, −1), (−1, 3) p ≡ 7 or 11 (mod 24)
 −1 if p ≡ 11 (mod 24)
1 if p≡7 (mod 8)
(7, −1), (−1, 7) p ≡ 7 or 11 (mod 12)
−1 if p≡3 (mod 8)
(13, −5), (−5, 13) p = 31 −1

1 if p ≡ 23 (mod 24)
(15, −5), (−5, 15) p ≡ 19 or 23 (mod 24)
−1 if p ≡ 19 (mod 24)
(16, −6), (−6, 16) p ≡ 11 (mod 12) 1
(17, −7), (−7, 17) p ≡ 7 (mod 8) 1
(18, −8), (−8, 18) p ≡ 11 (mod 12) 1
1 if p ≡ 23 (mod 24)
(19, −9), (−9, 19) p ≡ 19 or 23 (mod 24)
−1 if p ≡ 19 (mod 24)

17
Conjecture 2. Theorem 3 holds for the following additional values of r, s, p
and :
(r, s) p 
(2, 2) p ≡ 5 (mod 6) −1

1 if p ≡ 5 (mod 12)
(5, 3), (3, 5) p ≡ 5 (mod 6)
−1 if p ≡ 11 (mod 12)
(3, −1), (−1, 3) p ≡ 5 (mod 6) 1
1 if p ≡ 5 (mod 12)
(5, −1), (−1, 5) p ≡ 5 (mod 6)
−1 if p ≡ 11 (mod 12)
(9, −1), (−1, 9) p ≡ 11 (mod 12) −1
(10, −2), (−2, 10) p ≡ 5 (mod 6) −1
(11, −3), (−3, 11) p ≡ 11 (mod 12) −1

Conjecture 3. Theorem 4 holds for the following additional values of r, s, p

and :
(r, s) p 
(5, 5) p ≡ 19 or 23 (mod 24) 1
(7, −1), (−1, 7) p ≡ 7 (mod 8) 1

Remark. Since this paper was written, Robin Chapman has proved the cases
(r, s) = (3, −1) and (r, s) = (−1, 3) of Conjecture 1.

This investigation was partially supported by a grant to the second author

from the ARC.

References
[1] J. H. Conway and S. P. Norton, ‘Monstrous moonshine’, Bull. London
Math. Soc., 11 (1979), 308–339.
[2] D. Dummit, H. Kisilevsky and J. McKay, ‘Multiplicative products of η-
functions’, Finite groups — coming of age (Montreal, Que., 1982), 89–98,
Contemp. Math., 45, Amer. Math. Soc., Providence, R.I., 1985.
[3] F. J. Dyson, ‘Missed opportunities’, Bull. Amer. Math. Soc., 78 (1972),
635–652.
[4] N. J. Fine, ‘Basic hypergeometric series and applications’, Mathematical
Surveys and Monographs, 27. Amer. Math. Soc., Providence, R.I., 1988.
[5] B. Gordon and K. Hughes, ‘Multiplicative properties of η-products. II’, A
tribute to Emil Grosswald: number theory and related analysis, 415–430,
Contemp. Math., 143, Amer. Math. Soc., Providence, R.I., 1993.

18
 [6] B. Gordon and D. Sinor, ‘Multiplicative properties of η-products’, Number
theory, Madras 1987, 173–200, Lecture Notes in Math., 1395, Springer,
Berlin-New York, 1989.
[7] G. H. Hardy and E. M. Wright, ‘An Introduction to the Theory of Num-
bers’, Oxford, 5th edition, 1989.
[8] M. D. Hirschhorn, ‘A generalisation of Winquist’s identity and a conjecture
of Ramanujan’, J. Indian Math. Soc. , 51 (1987), 49–55.
[9] F. Klein and R. Fricke, ‘Vorlesungen über die Theorie der elliptische Mod-
ulfunktionen’, Vol. 2, Teubner, Leipzig, 1892.
[10] I. G. Macdonald, ‘Affine root systems and Dedekind’s η-function’, Invent.
Math., 15 (1972), 91–143.
[11] Y. Martin, ‘Multiplicative η-quotients’, Trans. Amer. Math. Soc., 348
(1996), 4825–4856.
[12] G. Mason, ‘Elliptic systems and the eta-function’, Notas Soc. Mat. Chile,
8 (1989), 37–53.
[13] M. Newman, ‘An identity for the coefficients of certain modular forms’, J.
London Math. Soc., 30 (1955), 488–493.
[14] M. Newman ‘Modular forms whose coefficients possess multiplicative prop-
erties’, Ann. of Math., (2) 70 (1959), 478–489.
[15] M. Newman ‘Modular forms whose coefficients possess multiplicative prop-
erties. II.’, Ann. of Math., (2) 75 (1962), 242–250.
[16] S. Ramanujan, ‘On certain arithmetical functions’, Trans. Camb. Phil. Soc.,
22 (1916), 159–184.
[17] S. Ramanujan, ‘Some properties of p(n), the number of partitions of n’,
Proc. Camb. Philos. Soc., 19 (1919), 207–210.
[18] L. Winquist, ‘An elementary proof of p(11n + 6) ≡ 0 (mod 11)’, J. Com-
bin. Theory, Ser. A, 6 (1969), 56–59.
IIMS, Massey University, Albany Campus, Private Bag 102 904, North Shore
Mail Centre, Auckland, New Zealand
e-mail: s.cooper@massey.ac.nz

School of Mathematics, UNSW, Sydney, Australia 2052

e-mail: mikeh@maths.unsw.edu.au

SMS, University of Sussex, Brighton, BN1 9QH, UK

email: r.p.lewis@sussex.ac.uk

19