A NOTE ON THE IRRATIONALITY OF C(2) AND C(3)

F. BEUKERS

1. Introduction

At the "Journees Arithmetiques" held at Marseille-Luminy in June 1978,

R. Apery confronted his audience with a miraculous proof for the irrationality of
((3) = l~ 3 + 2~ 3 + 3~ 3 + .... The proof was elementary but the complexity and the
unexpected nature of Apery's formulas divided the audience into believers and
disbelievers. Everything turned out to be correct however. Two months later a com-
plete exposition of the proof was presented at the International Congress of Mathe-
maticians in Helsinki in August 1978 by H. Cohen. This proof was based on the
lecture of Ape*ry, but contained ideas of Cohen and Don Zagier. For a more extensive
record of this little history I refer to A. J. van der Poorten [1]. Apery's proof will be
published in Acta Arithmetica.
In this note we give another proof for the irrationality of ((3) which is shorter
and, I think, more elegant. This proof is achieved by means of double and triple
integrals, the shape of which is motivated by Apery's formulas. Like Apery's proof it
also works for {(2), which is of course already known to be transcendental since it
equals n2/6. Most of the integrals that appear in the proof are improper. The manipu-
l-E

lations with these integrals can be justified if one replaces I by and by letting e
tend to zero. 0 6

2. Throughout this paper we denote the lowest common multiple of 1, 2, ..., n

by dn. The value of dn can be estimated by
(j _ FT p[log n/Jog p] < T-T -log n/log p _ ^ ( n )
Prime Prime
p^n p^n

and the latter number is smaller than 3" for sufficiently large n.

LEMMA 1. Let r and s be non-negative integers. Ifr>s then,

l I
r f xr ys
(a) ~^—dxdy
J J I—xy
0 0

is a rational number whose denominator is a divisor of dr2.

I I

(b) f ( ~l xy
oo
is a rational number whose denominator is a divisor ofdr3.

Received 18 November, 1978.

[BULL. LONDON MATH. SOC, 11 (1979), 268-272]
 A NOTE ON THE IRRATIONALITY OF ( ( 2 ) AND ( ( 3 ) 269

Ifr = s, then

f f *""/ 1 1
(c) — •xy
— dxdy = C(2)- — - . . . - — ,
JooJ l-xy r r*

n
1 1
log xv [ 1 1

0 0

Remark. In case r = 0, we let the sums \~2 + ...+r~2 and l ~ 3 + . . . + r ~ 3 vanish.

Proo/. Let a be any non-negative number. Consider the integral

— xy ^^- a)
JooJ \-xy
Develop (1— xy)~l into a geometrical series and perform the double integration.
Then we obtain
1
(2)

Assume that r > s. Then we can write this sum as

1 ( 1 1 ) 1 ( 1 1 )
k=o r - s \ k + s + < r + l k + r +a+\) r - s \ s + \ + o '" r+ a j '

If we put a = 0 then assertion (a) follows immediately. If we differentiate with respect

to a and put a = 0, then integral (1) changes into

and summation (3) becomes

n oo
log
-
I- xy

- 1 I 1 1
2
r-s {(s+l) ' r2
Assertion (b) now follows straight away.
Assume r = s, then by (1) and (2),
i i

— dxdy= Z 2 *
J J l—xyxy kk==o o (k + r + a+l)
oo
By putting a = 0 assertion (c) becomes obvious. Differentiate with respect to a and
put a = 0. Then we obtain
1 i

logxy
/• /- logxy . . , , « -2
r r

0 0

which proves assertion (d).

270 F. BEUKERS

THEOREM 1. ((2) is irrational.

Proof. For a positive integer n consider the integral

r r l-y) pn(x) dxdy>

J J l-xy
oo
f d )»
where Pn(x) is the Legendre-type polynomial given by n\Pn(x) = {—} x"(l — x)".
[dxj
Note that Pn(x)e l[x]. In this proof we shall denote the double integration by the
single sign J. It is clear from Lemma 1 that integral (4) equals (An+Bn£(2))dn~2 for
some An e Z and BneZ. After an w-fold partial integration with respect to x integral (4)
changes into

yn(l-y)nxn(l-x)n
— dxdy.
(l—xy)n

It is a matter of straightforward computation to show that

^ I 1 for all 0 < x < 1, 0 <

l-xy I 2 J

Hence integral (4) is bounded in absolute value by

( V 5 - 1 )5" r 1 ( V 5 - 1 )5n
} dxdy = C(2).
{ 2 j J l-xy { 2 j

Since integral (5) is non-zero we have

0 < |,

and hence
, ( V 5 - 1 )5» • I 7 5 - 1 \5" t5\»
2 2L
0 < \An + Bn C(2)| < d [• Y~) C(2) < 9" { — — | C(2) < J - J
for sufficiently large n. This implies the irrationality of £(2), for if C(2) was rational the
expression in modulus signs would be bounded below independently of n.
THEOREM 2. ((3) is irrational.

Proof. Consider the integral

noo
—logxy
1— xy
^ Pn(x)Pn(y)dxdy, (6)
 A NOTE ON THE IRRATIONALITY OF £(2) AND £(3) 271

where n\Pn(x) = {—} x"(i— x)n. It is clear from Lemma 1 that integral (6) equals
[dxj

(An+Bn £(3))dn~3 for some An e I, Bnel. By noticing that

1
— log xy r 1
Z>
\-xy ~ J l-(\-xy)z

integral (6) can be written as

r P (x)P (y)
— dxdydz,
J 1 — (1— xy)z
where J denotes the triple integration. After an «-fold partial integration with respect
to x our integral changes into

f \xyzy\i-xy vn\j)
— dxdydz.

Substitute

1-z
w =

We obtain

r P (v)
(
{l-xy{l_wy "
(1— xy)w
J 1— (1—
After an K-fold partial integration with respect to y we obtain
n
f (lxyyn(ly)nw"(lw)n
x\L-x)y\i-yyw\i-wr
—— dxdydw. (7)

It is straightforward to verify that the maximum of

x(i -x)y(l -y)w(l ~ w)(l - (1 -xy)w)-i

occurs for x = y and then that

x(l — x)y(l — y)w(l — w)

(72-1)4 for all 0 ^ x, y, w < 1.
1 — (1— xy)w

Hence integral (6) is bounded above by

(V2-1)4" f—— -dxdydw = (V2-1)4" f f Z^i^dxdy = 2(V2-1)4" C(3).

J 1 — ( 1 — xv)w JJ 1— xv
272 A NOTE ON THE IRRATIONALITY OF £(2) AND ( ( 3 )

Since integral (7) is not zero we have

0 < \An + BnC(3)\dn-2 < 2C(3)(V2-1) 4 ",
and hence

0 < K + iU(3)l < 2C(3K 3 (V2-1) 4 » < 2C(3)27"(V2-1)4"

for sufficiently large n, which implies the irrationality of £(3).

Reference
1. A. J. van der Poorten, "A proof that Euler missed . . . Apery's proof of the irrationality of £(3)'
To appear.

Department of Mathematics,
University of Leiden,
Postbus 9512,
2300 RA Leiden,
Netherlands.