JOURNAL OF NUMBER THEORY 25, 201-210 (1987)
Another Congruence for the Apkry Numbers
F. BEUKERS ’
Mathematisch Instituut, Universiteit Leiden,
Postbus 9512, 2300RA Leiden, The Netherlands
Communicated by M. Waldschmidt
Received May 29, 1985
In 1979 R. Apery introduced the numbers a,,=z:;; (;)-( ’ “ik)l in his irrationality
proof for l(3). We prove some congruences for these numbers, which extend
congruences previously published in J. Number Theory (Vol. 12. 14. 16, and 21).
Ii 1987 Academic Press. Inc.
Let, for any 12> 0,
These numbers have become interesting through Apery’s irrationality proof
of [( 3) and <( 2) (see [P] ). As a result, several people started looking more
closely at them and the first new properties to be discovered were of con-
gruence type (see, e.g., [CCC, Gel). Here we cite: for any prime p 2 5 and
m,rcN we have
aw’ , = amp,-I l (mod ~~~1,
bmp’-1= bmp ’ ~ I (mod p3’).
This was proved in [Bell by a brute force method. Another result for the
b,, is, that for p > 5 prime we have
4a’ - 2p (mod p) if p = a2 + b’, a odd
b(p- I)/2 =
{ 0 (mod P) if p=3 (mod4)
In fact, one can show a more general result; but here we restrict ourselves
to the simplest case. The proof (see [SB, Theorem 13.11) uses the interplay
between the numbers b, and the c-function of a certain algebraic K3
’ Present address: Mathematics Department, University of Utrecht, P.O. Box 80.010, 3508
TA Utrecht, The Netherlands.
201
0022-314X/87 $3.00
Copyright ID 1987 by Academic Press. Inc.
All rights of reproductmn in any form reserved.
202 [Link]
surface. Surprisingly, numerical experiment seems to suggest that the
congruences for b,, ~, ),z even hold mod p*.
In this paper we shall prove the following congruence. Let p 2 3 be prime
and write
f. y,q”=q fi (l-q2”)4(1-q4n)!
n=I n= I
Then
a(,, II/2=~p (modp).
For the most general statement see Theorem 4 of this paper. Although we
do not know all the details yet, this congruence must arise from the inter-
play between the numbers a, and the c-function of a certain algebraic
threefold. In this paper however, we give an ad hoc proof by relating the
generating function of the a,, to a certain modular form. We conclude this
introduction with the
Conjecture.
a,,-l,i2-~p (modp’).
1. SOME FACTS ON APBRY NUMBERS
The numbers
a,= f
k=O Q2(“i
“>
satisfy the recurrence
(n-t 1)3a,r+l =(34n3+51n2+27n+5)a,-n3a,p, n31
This implies that the generating function d(f) = C; a,,[” satisfies the
differential equation
2
Ld( t) = 0, L=(t4-34t3+t2) $ 3+(6t3-153f2+3t) ;
0 0
+ (7t2- 112t+ I)-$+([-5). (A)
The function d(t) can be interpreted as a period of a family of K3 surfaces
as described in [BP]. In the same paper it is also shown that any three
APkRY NUMBERS 203
solutions of (A) satisfy a homogeneous quadratic relation with constant
coefficients. In particular, if one chooses the solutions
y, = d(t) log t- 3 - 3t-9t2 + 103t3 + 5822t4 + ...>
y,= -+‘(t)log2t+y,logt-36t-558t2-11154t3+ . . ..
then
2y, y, + 9-v; = y:.
An old theorem of Fuchs [Fa, p. 5251 implies that (A) is the symmetric
square of a second order linear differential equation. In our particular
example we have
PROPOSITION 1. The space of solutions of (A) is spanned by the squares
of the solutions of
(t3-34t2+t)u”+(2t~-51t+1)u’+$(2-10)u=0. (B)
This fact, which can be verified by direct computation, was first noticed by
ApCry [Dw, pp. 255061.
Let Cr u,t” be a holomorphic solution at t = 0 of (B), then the u, satisfy
the recurrence relation
(n+ u2u,+1 =(34n2+17n+$)u,-(n-+)%4,-,.
This recurrence may give rise to another irrationality proof. In fact, we
claim
Claim. Let
3’2
then
Hz,> is irrational.
We shall give a proof of this claim in a forthcoming paper [Be 21.
We now substitute
t=w -9x) and u=(l-x)“ZU
1 --x
641’25 Z-6
204 F. BEUKERS
in (B) to obtain,
x(x-1)(9x-1)u”+(27x2-2Ox+1)0’+(9x-3)u=O. (Cl
Equation (C) however, is exactly the Picard-Fuchs equation associated to
the family of elliptic curves
E,: Y’+(l -3x) YX-x’(x-1) Y=X3. (1)
Two solutions of (C) are given by two independent periods of the differen-
tial form on E,. The holomorphic solution around x = 0 of (C) is given by
n 2 2k
u,= f:
k=O O( )k k’
(2)
For all these facts concerning Eq. (C) see [SB, Sects. 10, 111. The family E,
has essentially been taken from a list of Beauville [Bea] as the modular
family corresponding to the modular subgroup r,(6). That there exists
such a relation between Eq. (A) and the group r,(6) is an example of a
general theory of families of K3 surfaces with Picard number 19. A paper
on this subject is under preparation by C. Peters [Pe].
2. THE GROUP r,(6) AND ITS FUNCTIONS
The modular subgroup
has the following fundamental domain as pictured below. A set of
inequivalent cusps is given by ice, 0, f, f with widths given by 1, 6, 2, 3,
-s 1 -1 3 0 4 4
FIGURE 1
APkRY NUMBERS 205
respectively. All cusps are regular. Let r be the parameter of the complex
upper half plane and denote q = [Link] the transformationformulae in
[Ra, Chap. 93 for the Dedekind q-function
w=Y’~“nip -4”)
we observe that
8 v](T) 4
aT)= ($y ) (3q(32)
is a modular function with respect to r,(6). The values at the cusps are
given by
Since the only zero of $(z) is at r = ice and has order 1, the function I,+(Z)
generates the function field of modular functions on r,(6).
The @-vector space of modular forms of weight k with respect to f,(6)
will be denoted by M,(r,(6)). The space of weight k cusp forms will be
denoted by S,(r,(6)). Application of [Shi, Proposition 2.161 shows that
an element of M,(f i(6)) has k inequivalent zeros, counted with mul-
tiplicities.
The family E, of elliptic curves has singular Iibres at x = 0, 4, 1, cc and
the number of components in each Iibre is 6, 1, 3, 2, respectively. Now note
that the function x(t)= I,G(- l/6 t ) matches these multiplicities with the
widths of the cusps. Let f(x) be as in (2). According to [SB,
Theorem 14.91 with t, r, N, (; “,) replaced by x, $, 6, ( J’6 k), respectively, we
find that f($(z))~Mi(r,(6)). From [SB, Theorem 14.71 follows that
f((+(r)) has a zero of order 1 at r=$. Thus F(t)= (1 - $(z))f’($(r))
belongs to M2( r,(6)). The space M2(rl (6)) is spanned by
G*(T) - 6G2(6~), G,(T) - =,(2r), G,(T) - =,@T),
where
G,(T)=~-24 f nq"
n=, l-4"'
It is not hard to check that
F(T) = &j (=,(2r) - 3G2(3r) - 5(G,(r) - 66,(6r)))
=1+5q+13q*+23q~+29q4+30q~+ ‘..
206 F. BEUKERS
Notice that F( - l/67) = -6z2F(7). Moreover, from its definition follows
that F(7) has first order zeros at 7 = $, i.
Consider the modular function
1- 911/(7)
t(7) = $(7)
l-$(7).
It is not hard to verify that
Note, that t( - l/67) = t(7), and t(0) = t(ico) = 0, t(4) = t(f) = a3, both zeros
and poles being of order 1. Now, let &(t) be the generating function of the
Aptry numbers u,. Since
l-9*(7)
,C4(f(7))=yQ1’
( $(7) 1-$(7) ) =F(7)
we observe that d(t(7)) E M,(r,(6)), it has simple zeros at 7 = f, f, and the
first few terms of its Fourier expansion read
.d(t(z))= 1+5q+ 13q’+23q3+29q4+30q5+ ‘.‘.
PROPOSITION 2. Let
Then we have
d’(A2) djb= (~/(27)~ ~(47)~ - 9~(67)~ q( 127)4) dq.
ProoJ: Consider the function
G(7)=
(2>2.
(&>2d2(t(r))f
It is straightforward to check that t - ’ dt/dr E M,(r,(6)). We know that
t(s) has simple zeros at 7 = 0, ice, simple poles at 7 = 1, f and d(t(z))
has simple zeros at 7 = +, f. Hence G(7) E S,(r,(6)). Moreover,
G( - l/67) = 647RG(7). Two more forms in S,(r,(6)) having the same sym-
metry with respect to 7 + -l/67 are given by
~(7)~ ~(27)’ + 81q(37)’ r/(67)’ = q - 8q2+ 93q3+ 64q4+ .
APkRYNUMBERS 207
and
~(2)~ am 17(3r)~ ~(62)~ = q* - 4q3- 2q4 + . . ..
The Fourier expansion of G(r) starts with
G(z)=q-26q* + 165q3 + 100q4 + . . ..
Note that this last expansion coincides up to at least fourth order with the
expansion of
~(2)~ n(2r)’ + 81q(3r)* I*- 18q(~)~ am ~j(3r)~ ~(62)~
= (~(2)~ q(2Q4 - 9~(3r)~ am)*.
Hence the difference between these two forms has a zero of order at least 4
at r = ice. By symmetry there is also a zero of order at least 4 at z = 0.
Together with the zeros at T = t, $ we now have a form with at least 10
zeros. Since a nontrivial form in S&f ,(6)) has exactly 8 zeros, we conclude
that
G(T) = (~(2)~ am - 9~(3r)~ am)*
and our proposition now follows after taking square roots and replacing z
by 2r.
3. THE CONGRUENCES
PROPOSITION 3. Let p be a prime and
w(t)=f’h,t”-’ dt
a differential form with 6, E Z,, the p-adic integers, Vn. Let t(u) = C;” A,u”,
A, E Z,, Vn and suppose
Suppose there exist c1,, BPEZ, with p 1BPsuch that
bmp’- apbrnp,+l+fi,b,,-z-0 (modp’), Qm,rEN, (3)
Then
cmpr- IX~C,,,~~-I+ /lpc,,,pr-~ = 0 (mod p’), Vm,reN. (4)
Moreover, if A, is a p-adic unit then the congruences(4) imply (3).
208 F. BEUKERS
Remark. We use the convention b,$ = 0 if mp” # Z.
Proof Note that the congruences (3) are equivalent to
Note also
Since
t(u)“” = t(uP)” + npf,(u), .f?l(u) E aIun
we find
A o(t(u)“) = r z d(t(uP)“) + a, df,(u)
P
=$ 41(UP)) + Q(u), $(u) E Z,[u]
Similarly
L 4f(f4p2) =; O(f(UP2))+ dl+h(u),
l)(u) EZ,[u].
P
Hence (5) implies
where p(u) E Z,[uj. The latter equality is equivalent to
c mpr- ~l~c,,,~,- I + ~pc,pr-~ = 0 (mod p’), Vm,rEN.
If A, is a p-adic unit, we can write u = CF B, t”, B, E Z, and work
backwards.
Remark. The congruences (3) imply, that C b,t”/n is a logarithm of a
formal group law over Z,. Our Proposition comes down to the fact that for
a logarithm of an isomorphic formal group law, C c,t”/n, the c, satisfy the
same congruences as the b, (see [SB, Theorem A.9 of the Appendix]).
APkRY NUMBERS 209
THEOREM 4. Let
u(n)=,_,(~)‘(“:“)’
i na”
and let
f y,q”=q fi (1 -q2”)4(1 -q4”)?
n=l n=l
Then, for any odd prime p and any m, r E N, m odd, we have
~(~)-~,,(mpr~-1)+p3a(mpr~-1)~0 (modp’).
Proof: Since the function
q fi (1 -q2”)“(1 -q4n)4, 4 = e2nir
n=l
is the unique cusp form in S,(r,(8)) (see [Ko]), its corresponding
Dirichlet series has an Euler product
see [Ogg, Chap. IV]. Let
Y/(~z)~ ~(42)~ - 91(6~)~~( 12~)~ = C y”,,q”.
n=l
Then
~~,~=,~,yn-n~-3=~is,~-f f 2
n=in
1
=(1-32-T) fl
podd1 -YpP-“+P3-2s
This Euler product implies for any prime p 2 5,
ympr- ypy”mp’-’+p3j+*=o Vm,rEN
and
0 if 9 I m3’
ym3r- y3jjm3rm, + 27y”,,,-2 =
-9Y, if 3 II m3’, Vm, rE N.
210 F. BEUKERS
Hence, for all odd primes p,
From Proposition 2 we know that
and Proposition 3 now implies our congruences.
REFERENCES
[Bell F. [Link] congruences for the Aptry numbers, J. Number Theory 21 (1985),
141-I 55.
[Be21 F. BEUKERS, Irrationality proofs using modular forms, JournCes arithmttiques
BesanCon 1985. AstPrisque, to appear.
[Bea] A. BEAUWLLE, Les families stables de courbes elliptiques sur P’ admettant quatre
libres singuliires, C. R. Acad. Sci. Paris 294 (1982), 657.
[BPI F. BEUKERSANI) C. A. M. PETERS.A family of K3 surfaces and r(3). J. Reine Anget+,.
Math. 351 (1984), 42-54.
[CCC1 S. CHOWLA, J. [Link] M. COWLES,Congruence properties of ApCry numbers,
.I. Number Theory 12 (1980), 188%190.
lDw1 B. DWORK, On Apkry’s differential operator, “Groupe d’itude d’analyse ultramktri-
que.” Paris, 1980/1981.
IFal G. FANO, Uber lineare homogene Differentialgleichungen, Mafh. Annal. 53 (1900).
493-590.
[Gel I. GESSEL,Some congruences for the Aptry numbers, J. Number Theory 14 (1982).
362-368.
[Kol M. KOIKE, On McKay’s conjecture, Nagoya Math. J. 95 (1984), 85-89.
mg1 A. OGG, “Modular Forms and Dirichlet Series,” Benjamin, New York, 1969.
CPel C. A. M. PETERS. Monodromy and Picard-Fuchs equations for families of
K3-surfaces and elliptic curves, preprint.
[PI A. J. VAN DER POORTEN, A proof that Euler missed... Ap&y’s proof of the
irrationality of c(3). Math. Infelligencer 1 (1979). 195-203.
CRal H. RADEMATHER,“Topics in Analytic Number Theory,” Springer, New York/Berlin,
1973.
WI J. STIENSTRA AND F. BEUKERS, On the Picard-Fuchs equation and the formal
Brauer group of certain elliptic K3 surfaces, Mafh. Ann. 271 (1985), 269-304.
[Shi] G. SHIMIJRA, Introduction to the arithmetic theory of automorphic forms, Iwanami
Shoten, Princeton, 1971.