Analytic Continuation of Riemann's Zeta Function and Values at Negative Integers Via Euler's Transformation of Series Author(s): Jonathan

Sondow Reviewed work(s): Source: Proceedings of the American Mathematical Society, Vol. 120, No. 2 (Feb., 1994), pp. 421-424 Published by: American Mathematical Society Stable URL: http://www.jstor.org/stable/2159877 . Accessed: 15/11/2011 05:04
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Mathematical Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the American Mathematical Society.

http://www.jstor.org

PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 120, Number 2, February 1994

ANALYTICCONTINUATION OF RIEMANN'S ZETA FUNCTION AND VALUES AT NEGATIVE INTEGERS VIA EULER'S TRANSFORMATION OF SERIES
JONATHANSONDOW (Communicatedby William W. Adams) provides We provethat a seriesderivedusing Euler'stransformation the analytic continuationof C(s) for all complex s 76 1 . At negative integers the series becomes a finite sum whose value is given by an explicit formula for Bernoullinumbers.
ABSTRACT.

1. INTRODUCTION

Euler computed the values of the zeta function at the negative integers ussum ing both Abel summation (75 years before Abel) and the Euler-Maclaurin formula. (Comparisonof these values with those he found at the positive even integers led him to conjecture the functional equation 100 years before Riemann!) Euler also used a third method, his transformation of series or (E) summation (see ?2), to calculate C(-n), but only for n = 0, 1, 2, and 4. (See [1; 4, ?1.5; 5, volume 14, pp. 442-443, 594-595; volume 15, pp. 70-90; 7; 9, ??1.3, 1.6, 2.2, 2.3; 10; 14, Chapter III, ??XVII-XX].) We observe in ?4 that this last method in fact yields
C(-n)

(-l)nBn+1/(n + 1) for all n > O

but we require an explicit formula for Bernoulli numbersthat was discovered a century after Euler. In ?3 we justify the method by proving that a series used in ?4 gives the analytic continuation of 4(s) for all s : 1 . Similar results for approximationsto Euler's transformationare obtained in ?5, as well as an evaluation of C'(0)/I(0) = log 27r. In a paper in preparation, the author will apply the method to other zeta functions and to Dirichlet L-series.
2. EULER'S TRANSFORMATION SERIES OF

Any convergentseries of complex numbers, written with alternatingsigns as A = a, - a2 + a3 ,

Received by the editors December 8, 1991 and, in revised form, June 3, 1992; presentedto the AmericanMathematicalSociety on January10, 1992, at Baltimore,MD. 1991 MathematicsSubjectClassification.Primary 1IM06, 40G05, 11B68.
Society ( 1994 AmericanMathematical 0002-9939/94 $1.00 + $.25 per page

421

422

JONATHANSONDOW

can also be written in the form

] A = a1 '[(a, - a2 )- (a2 - a3) + a,+ Repeating the process on the series in brackets, we have

A = la, + 1(a, and in general (1) ~~00

(1)
1
(-ln-1

a2) +I

[(a, - 2a2+ a3)-

(a2- 2a3+

a4) +***]

k-lA = 0

ZE

\0%\f.1af

al +

E(L)
n=1

n-2Akan

= where AOan an and

kk Akan =Akl an _Akl an+i
=

Z,(1)m
m=O

an+m

approaches 0 as k

for k > 1. It is proved in [11, ?33B] that the sum of the last series in (1) -* oo, so that
00

(2)

Z(-1)n-an
1

Z 2JI,
0

00Aa

which is Euler's transformationof series. (See [5, volume 10, pp. 222-227; 9, ?4.6; 11, ??35B, 59, 63].) 3. ANALYTIC CONTINUATIONOF C(S) Instead of working directly with C(s), which for a = Re(s) > 1 is given by C(s) = 1- + 2-5 + 3-s + , let us consider the alternating series
..

, (s) = 1-5 - 2-5 + 3-s _ which converges for a > 0 (see ?5). Applying the Euler transformation, we (3) have, for a > 1,
(1- 21-s)(s)=
Z

2C(s) - 2 2

2+-S

(4)
=

00

E
0

1-(j)2-s+(i()3-s-

+ (-1)j(j)(j+1)
2j+1

Theorem. The analytic continuation of C(s) for all complex s :$ 1 is given by the product
(5) (5)

~~~~~~~~~~~~~coAI

C(s) = (1 - 21-s)-l

1:
0

s 12i

in which the series convergesabsolutely and uniformly on compact sets to an

entire function. Proof. Fix k > 0. Evidently
n

I
...**

Akns= -(s)k

(n + xi + ***+ Xk )-s-k

dxj

dXk

OF ANALYTICCONTINUATION RIEMANN'SZETAFUNCTION

423

for k = 1, 2,..., Hence, (6)

k
=

where

(W)k

denotes the product s(s + 1)...(s + k - 1).

\jAkn-sl < I(S)kI/na+k whenever a + k 0, 1, 2,..., where (s)=

> 0,

1. Now let S be a compact set in the half

k, and let M, denote the maximum of I(S)kI/na+k on S. Then plane a> (6) implies that E Mn dominates the series 100

(7)

1) 1: ~~(- n-leakn-s n=1

on S. It follows, using the triangle inequality, that the Euler transform of Z Mn dominates the Euler transform of (7), which, since AjAk = Aj+k, is A jAkI-Vs
j=O
E 2+1 2 00 E

ll2+-ken

j =kZ2i+1

Multiplying this by 1/2k and adding

k-I AlsI-s/21+1

produces the series in

(4), which, since k is arbitrary,therefore converges absolutely and uniformly on compact sets to an entire function. Since the series in (3) has zeros at the (simple) poles of (1 - 21-s)-I except at s = 1 (for a direct proof see [12]), the theorem follows. 4. EVALUATION OF C(-m) Let m be a positive integer or 0. Note that (-rm)j = 0 and, hence, Ai lm0 for j > m. Thus when s = -im the series in (5) becomes a finite sum. Its value is given by a formula for Bernoulli numbers that Carlitz [2] attributesto Worpitzky[15] (see also [3]), namely, the second equality in the following. Corollary. For m = 0, 1, 2, ....
=

1
- 2m+
E

) 2i~' +1(m (_l)m Bm+I

Alternatively, one can view 4(-m) = (-l)mBm+j/(m + 1) as known, which gives a proof of Worpitzky'sformula (compare [6]).
5. APPROXIMATIONS TO EULER'S TRANSFORMATION

Note that (1), (3), and (6) imply (without using (2)) that for k > 1 the product (8)
C(s) = (1 - 21s A) I

+ 2 Z(l)n-1Akn5)

providesthe analytic continuationof C(s) on the puncturedhalfplane a > 1- k, s :$ 1 where the infinite series convergesabsolutely and uniformlyon compact sets to a holomorphic function. Moreover,except that the convergencewill not be absolute in the strip -k < a < 1 - k this remains truefor k > 0 and a > -k,

424

JONATHAN SONDOW

s $ 1. (Proof Grouping terms in pairs in the even partial sums of the second summation, we have
2N N

E(2N
n=1

_n-lAkn-s = E(Ak(2n
n=1 N

- 1)-5 - Ak(2n)-s)

=E Ak+l(2n
n=1

1)-s.

Then it follows from (6) that both even and odd partial sums converge as required.) Since we can use (8) with k > m + 1 to evaluate 4(-m), the approximations (1) to Euler's transformationyield everything it does except formula
(5).

As an example, take k = 1 in (8):

~(s)=(1-21-s5

1I2+

Z()n-l(n-s

(n +1)-s))

for a > -1, s 1. (This formula appears in Hardy's proof of the functional = equation [8; 13, ?2.2] and gave the idea for the present note.) Thus C(O) -1/2 and, using Wallis's product for ir/2, 2 24 4 -1 N"O) = 2 1 2 + =i log 2h z, which log + log r pr s which figuresin the Hadamard product representationof the zeta function.
REFERENCES 1. R. Ayoub, Euler and the zeta function, Amer. Math. Monthly 81 (1974), 1067-1086. 2. L. Carlitz,Remarkon a formulafor the Bernoullinumbers,Proc. Amer.Math. Soc. 4 (1953), 400-401. 3. E. Cesairo,Transformations algebriques le calcul des differences, par Nouv. Ann. Math. (3) 5 (1886), 489-492. 4. H. M. Edwards,Riemann's zeta function, Academic Press, New York, 1974. 5. L. Euler, OperaOmnia, Series Prima, Teubner,Leipzig and Zurich, 1911-1956. 6. H. L. Garabedian,A new formula for the Bernoulli numbers,Bull. Amer. Math. Soc. 46 (1940), 531-533. 7. D. Goss, A simple approachto the analytic continuationand values at negativeintegersfor Riemann's zeta function, Proc. Amer. Math. Soc. 81 (1981), 513-517. 8. G. H. Hardy,A newproofof thefunctional equationfor the zeta-function, Mat. TidsskriftB 1922, 71-73. 9. , Divergentseries, Oxford Univ. Press, London, 1949. 10. M. Kline, Euler and infiniteseries, Math. Mag. 56 (1983),307-315. 11. K. Knopp, Theoryand applicationof infiniteseries, Blackie & Sons, London, 1951. 12. J. Sondow, The zeros of the alternatingzeta function on a = 1 , preprint. 13. E. C. Titchmarsh, The theoryof the Rz'emann zeta-function,Oxford Univ. Press, London,
1951.

14. A. Weil, Numbertheory,an approachthroughhistory,Birkhauser,Basel, 1984. 15. J. Worpitzky, Studien uber die Bernoullischenund EulerschenZahlen, J. Reine Angew. Math. 94 (1883), 203-232.
DEPARTMENT OF MATHEMATICS, YESHIVA UNIVERSITY, NEW YORK, NEW YORK

10033

Current address: 209 West 97th St., New York, New York, 10025