JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS 134, 1299140(1988)

Sums of Certain Series of

the Riemann Zeta Function*
H. M. SRIVASTAVA

Department of Mathematics, University of Victoria.

Victoria, British Columbia V8 W 2Y2, Canada
Submitted by J. L. Brenner

Received August 27, 1986

The object of the present paper is to investigate systematically several interesting

families of summation formulas involving infinite series of the Riemann zeta
function. Many of the various results, which are unified (and generalized) here in a
remarkably simple manner, have received considerable attention in recent years. We
also present a brief account of a number of analogous results associated with the
(Hurwitz’s) generalized zeta function. ‘X3 1988 Academic Press, Inc.

11 INTRODUCTION AND DEFINITIONS

A classical (over two centuries old) theorem of Christian Goldbach

(1690-1764) which was contained in a letter dated 1729 from Goldbach to
Daniel Bernoulli (1700-1782), has recently been revived as the problem
(see C221)
<op-w’=L (1.1)
<
where Y denotes the set of all nontrivial integer kth powers.
For the Riemann zeta function c(s) (see, e.g., Titchmarsh [26]),
Goldbach’s theorem (1.1) assumes the elegant form (cf. [22, p. 4031)

or, equivalently,

g f(m)) = 1, (1.3)

* This work was supported, in part, by the Natural Sciences and Engineering Research
Council of Canada under Grant A-7353.
129
0022-247X/88 $3.00
Copyright 0 1988 by Academic Press. Inc
All rights of reproduction in any form reserved
130 H.M.SRIVASTAVA

where f(x) =x- [x] denotes the fractional part of the real number x. In
fact, it is not difficult to show also that

,E2
W)‘f(i(k))=;, (1.4)

kt, f(Wk)) = iy and k~lf(i(2k+ l))=+ (1.5)

Formula (1.3), and hence also (1.1) and (1.2), and its such interesting
variations as (1.4) and (1.5) are, of course, equivalent to various (known or
easily derivable) sums of double series considered by, among others, Boole
[4, p. 105, Exercise lo], Stieltjes [24, p. 3001, Johnson [14, p. 4791,
Bromwich [5, p. 526, Example 61, Jordan [ 15, p. 3401, Chrystal [7, p. 422,
Exercise 181, Hansen [13, p. 3551, and Shallit and Zikan [22, p. 4021. In
the present paper we aim at investigating systematically several related
problems involving sums of series of c(s) and of the (Hurwitz’s) generalized
zeta function <(s, a) characterized, among other ways, by (cf. [8, p. 24,
Eq. 1.10(l)])
us7 1) = i(s), i(s, 4, = (2” - 1) C(s), (1.6)

(1.7)

and

(N= 1, 2, 3, ...). (1.8)

which, for N= 1, assumesa particularly simple (and useful) form.

2. UNIFICATIONS (AND GENERALIZATIONS)OF

THE SUMMATION FORMULAS (1.3) AND (1.4)

Making use of the familiar binomial expansion:

tk=(l-t)-A, ItI < 1, (2-l)

it is easily seen from the definitions of c(s) and ((s, a) that (cf. Ramanujan
[19, p. 78, Eq. (15)] and Apostol [2, p. 240, Eq. (7)])

{W+k)- l} tk=U12, 2-t), (tl <2. (2.2)

 SERIES OF THE RIEMANN ZETA FUNCTION 131

For fixed A # 1, the series in (2.2) converges absolutely for It( -C2. Thus, by
the principle of analytic continuation, the summation formula (2.2) is valid
for all (admissible) values of A.
Formula (2.2) provides a unification (and generalization) of ( 1.3) and
(1.4), and indeed also of a fairly large number of other summation formulas
scattered in the literature. For example, in view of the relationships in (1.6)
and (1.8) (2.2) with t = 1 gives us a known result (cf. [13, p. 356,
Eq. (54.4.1)] ) which generalizes ( 1.3), and a special case of (2.2) when
t = - 1 yields another known result (cf. [13, p. 356, Eq. (54.4.2)]) which
generalizes ( 1.4).
Several additional consequencesof the general summation formula (2.2)
are worthy of note. First of all, replace the summation index k in (2.2)
by k + 1, and set A = s - 1 and t = 1; we thus arrive immediately at the
alternative form of the aforementioned generalization of (1.3),

(2.3)

Formula (2.3) can easily be rewritten in a well-known form (cf. Landau

[16, p. 274, Eq. (3)] and Titchmarsh [26, p. 33, Eq. (2.14.1)]), which is
usually attributed to Edmund (Georg Hermann) Landau (1877-1938).
For f = - 1, (2.2) [with k replaced by k + 1, and L = s - 1] readily yields

i(s)=l+$&+ f (-1)“~‘&{i(s+W}.
k=l
(2.4)

which provides an interesting (presumably new) companion of Landau’s

formula referred to above; here, and in what follows, (s), = T(s + k)/T(s).
Setting t = 4 in (2.2) [with k replaced by k + 1, and I = s - 11, and
making use of (1.8) with a = 4 and N = 1, we obtain another series
representation for i(s),

W+k)- l},
which is believed to be new.
In their special cases when s = 2, Landau’s formula and its companion
(2.4) reduce simply to the summation formulas ( 1.3) and (1.4), respectively,
while (2.5) similarly yields an elegantly simple sum.
In view of (2.1), the summation formula (2.2) [with A = s and t = 4-J
readily yields the familiar result
132 H. M. SRIVASTAVA

which is attributed to Ramaswami (cf. [20, p. 166; 26, p. 33, Eq. (2.14.2)]).
Furthermore, in its special case when A = s and t = - i, (2.3) similarly gives
us the companion of (2.6),

(1-2’-“)5(s)=1- f (-&$dp, (2.7)

&=I

which was also obtained by Ramaswami [20, p. 1661.

Landau’s formula as well as (2.6) were rederived, using Eulerian integrals
for r-functions, by Menon [ 181.
In case we add (2.2) to itself (with t replaced by - t), we obtain a known
summation formula (cf. [ 13, p. 357, Eq. (54.6.3)]), while a similar subtrac-
tion yields the corresponding odd case of (2.2).
Various interesting further special even and odd cases of (2.2) are also
given in the literature. In particular, the special even casesof (2.2) when
t=+, t=f, and t+

were considered by Ramaswami [20, p. 167, Eq. (I), (3), and (4)] who
also gave a special odd case of (2.2) when t = f [20, p. 167, Eq. (2)], and by
Apostol [2] who proved various generalizations of Ramaswami’s results.
By assigning suitable numerical values to the variable s in some of the
aforementioned special even and odd casesof (2.2), Ramaswami [20] also
evaluated a number of special sums including, for example,

(2.8)

where y denotes the Euler-Mascheroni constant.

Formula (2.8) is contained in a memoir of 1781 by Leonhard Euler
(1707-1783) (cf. Glaisher [9, p. 28, Eq. (S)]); it was rederived by Wilton
[30, p. 921 who also gave a number of other sums. Furthermore, in view of
the special case x = 4 of the elementary identity

1.4
<1, (2.9)

the summation formula (2.8) is an immediate consequence of a well-known

result (also contained in Euler’s memoir of 1781 already referred to) which
was rederived in 1826 by Legendre ([ 17, p. 4341; see also Stieltjes [24,
p. 3021, and Glaisher 19, p. 28, Eq. (9)] who recalls both (2.8) and this
well-known result erroneously). Legendre [17, p. 434) also showed that
O” [(2k+ l)- 1
c = 1 -y+og2. (2.10)
&=I 2k+ 1
 SERIESOF THE RIEMANN ZETA FUNCTION 133

Johnson [14] presented alternative (direct) proofs of (2.10) and the

aforementioned well-known result, and obtained a number of additional
results including, for example, the sum (Johnson [ 14, p. 480, Eq. (S)]; see
also Verma and Kaur [28, p, 181, Eq. (D)]):

(2.11)

Formulas (2.10) and (2.11), together, imply the classical result (con-
tained in the aforementioned 1781 memoir by Euler)

(2.12)

which has appeared in several subsequent works by, for example, Glaisher
[9, p. 28, Eq. (4)], Johnson [14, p. 478, Eq. (4)], Bromwich [S, p. 526,
Example 61, Wilton [30, p. 933, Barnes and Kaufman [3], where it is
posed as a problem, and Verma and Kaur [28, p. 181, Eq. (A)], where it is
rederived in a standard manner.
We conclude this section by recalling the formula (cf. Glaisher [9, p. 27,
Eq. (l)] and Johnson [14, p. 478, Eq. (3)])

cm --&~(W}=IJ,
k=2
k-l
(2.13)

which was given in Euler’s memoir of 1769 and also the results contained
in Wilton’s work [30],

= log(2n) - 1; (2.14)
k:, k(?:)l )

(2.15)

Obviously, Euler’s formula (2.13) follows immediately upon subtracting

(2.12) from (1.2). Formula (2.15) on the other hand, complements such
sums as (2.8).

3. ADDITIONAL CONSEQUENCES OF THE SUMMATION FORMULA (2.2)

Replacing the summation index k in (2.2) by k + 2, setting ,I = s - 1, and

dividing each side by t2, if we differentiate the resulting equation with
respect to t, using the formula (1.7), we finally obtain
134 H. M. SRIVASTAVA

c0-w)k+I{[(s+k+
c l)- l} tk ’
k=,(k+2Y
= f -2([(s, 2- f) + i(s) - 1)
2t-3
-z{i(s-1,2-f)-i(s-l)+lj, O<Ifl<2, (3.1)

or, equivalently,

x {((s- 1, 1 -t)-gs- l)}, 0-C (t( < 1.

(3.2)
For t = - 1, (3.1) readily yields one of the two main results in a recent
paper by Singh and Verma [23] which incidentally follows also from (2.4)
above. Furthermore, (3.2) would formally reduce, when t + - 1, to the sum

k(S)/c + I
x(kUs+k+ 11, Re(s)< 1, (3.3)

which happens to be the other main result in [23]. These obvious

consequencesof (3.1) and (3.2) were proven in [23] in a markedly different
manner.
By assigning suitable special values to the variable t in (3.1) and (3.2),
we can deduce a large number of sums of series involving the zeta function.
For example, for t = 1, (3.1) immediately yields a series representation
which is deducible also from Landau’s formula referred to already.

4. SUMMATION FORMULAS INVOLVING SERIES OF [(k)/k

In the theory of r-functions, there exist two fairly well-known series of

[(k)/k (see, e.g., Erdelyi et al. [8, p. 45, Eq. 1.17(2)] and Jordan [lS, p. 62,
Eq. (2)]; see also Abramowitz and Stegun [1, p. 256, Eq. (6.1.33)-J).
For t -+ 1, the first one of these series reduces immediately to a classical
result (see Jordan [15, p. 621 and Erdelyi et al. [8, p. 45, Eq. 1.17(3)], and
the second one with t = 1 yields another sum (cf. Verma [27]; see also
Verma and Kaur [28, p. 182, Eq. (1 )]) which is, of course, equivalent to
the aforementioned classical result.
 SERIESOF THE RIEMANN ZETA FUNCTION 135

The special case of the second one of the aforementioned f-function

results when t = - 1 gives us the well-known sum (2.12) which, in
conjunction with the special case t = 1, would immediately yield the
summation formulas (2.10) and (2.11). Furthermore, the obvious special
casesof the first one when t = + f, together, yield the results like (2.8) and
(2.15).
Finally, we set t = + 3 in the aforementioned r-function result, and we
obtain the sums

and
1
=ilogrr+;(l-y), (4.2

which do not seem to have been recorded earlier.

With a view to simplifying the derivation of such summation formulas
as (2.8), (2.10), (2.1l), and (2.15), we can make use of some well-known
consequencesof the r-function results recorded, among others, by Hansen
[13, p. 356, Eq. (54.5.3) and (54.5.8)].

5. SUMMATION FORMULAS INVOLVING SERIES OF [(k)/(k+ 1)

By differentiating the r-function results, referred to in the preceding

section, and applying an elementary integral [ 11, p. 661, Entry 6.441(l)], it
is readily seen that

1 +;y-;log(2n), (5.1)

which was proved by Suryanarayana [25], and again by Singh and Verma
[23, p. 3, Sect.43. The method of derivation of (5.1) by these earlier
workers is fairly standard in the theory of the Riemann zeta function. By
the same method, Suryanarayana [25, p. 143, Eq. (14)] claimed to have
summed the obviously divergent alternating series

,;, t-1)” i(k),

whose corrected (convergent) version is precisely the same as the well-

known result (1.4).
136 H.M.SRIVASTAVA

In a manner similar to our derivation of (5.1), we can show also that

kIf2
(-1)” (5.2)
and
O” i(k)-1 3 1
c -=Z-Zy-;log(2n),
k=2 k+l
(5.3)

which, together, yield certain obvious sums in the even and odd cases.
Formula (5.2) follows trivially from the known result (5.1).
Formulas (5.2) and (5.3), and indeed also the well-known result (2.12),
happen to be the main results in a recent paper by Verma and Kaur [28,
p. 181, Eq. (A), (B), and (C)] who also state an erroneous version of the
sum in the odd case resulting from (5.2) and (5.3) [28, p. 181, Eq. (F)].
Moreover, the summation formulas (5.2) and (5.3) appeared more recently
as a problem (see [6]).
Finally, we obtain the sum

,c, <(2k)2-2k==log2, (5.4)

2k+1 2 2
which, in conjunction with the even case resulting from (5.2) and (5.3),
yields (see Robbins [21])

; log(27t) = 1 - f f(2m+l)~“f~(2m+l)~‘+ ..*I. (55)

m=l i

6. MISCELLANEOUSSUMMATION FORMULAS AND GENERALIZATIONS

The various summation formulas established in the preceding sections

can be suitably applied to deduce sums of series involving, for example,
c(k)/{k(k+ l)}. In particular, the summation formulas (2.12) and (5.3)
lead us easily to the sum

while (4.2) and (5.2) would yield the sum

 SERIES OF THE RIEMANN ZETA FUNCTION 137

By assigning appropriate special values to the arbitrary constants i and

p, we can obtain a number of interesting summation formulas as immediate
consequencesof (6.1) and (6.2). For instance, the special case of (6.1) when
,I= -p = 1 yields a known sum recorded by Chrystal [7, p. 372, Eq. (IS)].
By integrating a certain well-known series [S, p. 45, Eq. 1.17(2)] from
-z to z, we have (cf., e.g., [ 13, p. 356, Eq. (54.5.5)] for an alternate form)
1 z
k$ CG’k) z2k j- logf(l+r)dr, 0<[2[<1, (6.3 1
k(2k-t l)=; z
which obviously contains Wilton’s formula (2.14) as a special case. In a
similar manner. we obtain the following unification (and generalization) of
the summation formulas (6.1) and (6.2) and indeed also of several other
results,

A-p z
-- logf(2+t)dt, o< \z( <2, (6.4)
z s0
which, in view of an elementary integral, would yield (6.1) and (6.2) in its
special cases when z = - 1 and z = 1, respectively.
In the special case of (6.3) when z = 4, if we evaluate the resulting integral
by means of the aforementioned integral, we shall readily obtain a
summation formula given by Wilton 130, p. 911. Wilton’s result was posed
as a problem over four decades later (see [12]); it follows immediately
upon setting a = 1 in Burnside’s formula (cf., e.g., Wilton [30, p. 91,
Eq. (3)]; see also Erdelyi et al. [8, p. 48, Eq. 1.18(11)]),

kf+,1(2k3
‘) ,(;;; 1)=log(2ni+(2n-l)jlogjuf)-1}
- 2 log r(a), Re(a)> -4, (6.5)

involving the generalized zeta function [(s, a). As a matter of fact, Wilton
[30] rederived Burnside’s formula (6.5) as a consequence of the following
straightforward generalization of certain expansions like (2.3):
i+k-I
[(A+k,a)tk=[(L,a-t), (tl < Ial. (6.6)
k >
138 H. M. SRIVASTAVA

In terms of the generalized zeta function i(s, a), it is also known that
(cf. Whittaker and Watson [29, p. 276J; see also Gradshteyn and Ryzhik
[ll, p. 1074, Entry 9.5321)

= log r(a + t) -log r(a) - t*(u), JtJ< ]a(, (6.7)

where ll/(z) = P(z)/r(z).

Since $( 1) = -7, (6.7) would reduce immediately to certain well-known
results in [ 1, 8, 151 upon setting a = 1 and a = 2.
By employing the rather elementary techniques illustrated fairly fully in
this section, and in the preceding sections, we can easily derive appropriate
generalizations of the various summation formulas considered in this paper
as useful consequences of (6.6) and (6.7). We are thus led to several
generalizations of the type recorded by Hansen [ 13, p. 358, Eq. (54.11.2),
(54.11.3), and (54.11.4)]. It is easily seen from (6.7) and one of these
generalizations that

kz,w*$$$j (-(k, u)zk = 1 log f(u + z) -p log T(a)

x =logT(u+r)dr, O<)z(<JuJ, (6.8)

s0

which, upon setting z = - 1 and evaluating the resulting integral, yields the
summation formula:

-;(i-~)log(2x)-(I-p)(u-l)

x {log(u- l)- I}, larg(a - l)] <K. (6.9)

For A = -p = 1, this last result (6.9) reduces immediately to Binet’s

formula (cf. Whittaker and Watson [29, p. 261, Example 18-J; see also
ErdClyi et al. [8, p. 48, Eq. 1.18(10)]),
 SERIESOF THE RIEMANN ZETA FUNCTION 139

,:,c && ((I?, a) = 2 log r(a) - (2a - 1) log(a - 1)

-log(2rr) + 2(a - l), larg(u - l)[ < rc, (6.10)

which, for a = 2, yields the aforementioned known sum [7, p. 372,
Eq. (18)l.
It is not difficult to deduce, in a manner similar to what we have
indicated already, the following interesting generalization of several results
including, for example, Burnside’s formula (6.5),

-; (A-2/L) log(2x)

- (la-p(2u- l)] log

Re(u)> -i, (6.11)

which indeed yields (6.5) for i = p - 1 = 0.

Finally, by integrating a known result [13, p. 358, Eq. (54.11.3)] from
t = 0 to t = 1, we obtain the summation formula

=(a-l)log(u-l)-ulogu
k:, :::);2;::,
+ $(a) + 1, larg(u - 1)1< n. (6.12)
This last result (6.12) reduces, when a -+ 1, to an elegant summation
formula considered, for instance, by Glaisher [ 10, p. 9, Sect. 181 and
Ramanujan [ 19, p. 731.

REFERENCES

1. M. ABRAMOWITZ AND I. A. STEGUN, “Handbook of Mathematical Functions with

Formulas, Graphs, and Mathematical Tables,” National Bureau of Standards,
Washington, DC, 1964.
2. T. M. APOSTOL, Some series involving the Riemann zeta function, Proc. Amer. Math. Sot.
5 (1954), 239-243.
3. E. R. BARNEYAND W. E. KAUFMAN, The Euler-Mascheroni constant, Amer. Math.
Monthly 72 (1965), 1023.
4. G. &33LE, “A Treatise on the Calculus of Finite Differences,” 2nd ed., Macmillan Rr Co.,
London, 1872; reprinted by Dover, New York, 1960.
140 H. M. SRIVASTAVA

5. T. J. PA. BROMWICH,“An Introduction to the Theory of Infinite Series,” 2nd ed.,

Macmillan & Co., London, 1926.
6. L. M. CHRISTOPHE, JR., M. V. FINN, AND J. A. CROW,Two series involving zeta function
values, Math. Mag. 59 (1986), 176-178.
7. G. CHRYSTAL, “Algebra: An Elementary Text-Book,” Part II, 7th ed., Chelsea, New York,
1964.
8. A. ERDBLYI, W. MAGNUS, F. OBERHETTINGER, AND F. G. TRICOMI,“Higher Transcenden-
tal Functions,” Vol. I, McGraw-Hill, New York, 1953.
9. J. W. L. GLAISHER,On the history of Euler’s constant, Messenger Math. 1 (1872), 25-30.
10. J. W. L. GLAISHER,Relations connecting quantities of the form 1 + 2-” + 3-” + 4-” &c.,
Messenger Math. 44 (19141915), l-10.
11. I. S. GRADSHTEYN AND I. M. RYZHIK, “Table of Integrals, Series, and Products,” corrected
and enlarged ed., Academic Press, New York, 1980.
12. R. GREENBERG, D. C. B. MARSH,AND A. E. DANESE,A zeta-function summation, Amer.
Math. Monthly 14 (1967), 8&81.
13. E. R. HANSEN,“A Table of Series and Products,” Prentice-Hall, Englewood Cliffs, NJ,
1975.
14. W. W. JOHNSON,Note on the numerical transcendents S, and s, =S, - I, Bull. Amer.
Math. Sot. 12 (1905-1906), 477-482.
15. C. JORDAN,“Calculus of Finite Differences,” 3rd ed., Chelsea, New York, 1965.
16. E. LANDAU, “Handbuch der Lehre von der Verteilung der Primzahlen,” 2nd ed., Chelsea,
New York, 1953.
17. A. M. LEGENDRE, “Trait6 des Fonctions Elliptiques et des Inttgrales Eultriennes,” Vol. 2,
Huzard-Courtier, Paris, 1826.
18. P. K. MENON,Some series involving the zeta function, Math. Student 29 (1961), 77-80.
19. S. RAMANUJAN, A series for Euler’s constant y, Messenger Math. 46 (19161917), 73-80.
20. V. RAMASWAMI,Notes on Riemann’s i-function, J. London Math. Sot. 9 (1934), 165-169.
21. H. ROBBINS,A remark on Stirling’s formula, Amer. Math. Monthly 62 (1955), 26-29.
22. J. D. SHALLITAND K. ZIKAN, A theorem of Goldbach, Amer. Marh. Monrhly 93 (1986),
402403.
23. R. J. SINCHAND D. P. VERMA,Some series involving Riemann zeta function, Yokohama
Marh. J. 31 (1983), 14.
24. T. J. STIELTJES,Table des valeurs des sommes S, =x.;” n-‘[, Acfa Math. 10 (1887),
299-302.
25. D. SURYANARAYANA, Sums of the Riemann zeta function, Math. Student 42 (1974)
141-143.
26. E. C. TITCHMARSH, “The Theory of the Riemann Zeta-Function,” Oxford Univ. Press
(Clarendon ), London, 1951.
27. D. P. VERMA,A note on Euler’s constant, Murh. Student 29 (1961), 14&141.
28. D. P. VERMAAND A. KAUR, Summation of some series involving Riemann zeta function,
Indian J. M&h. 25 (1983), 181-184.
29. E. T. WHITTAKERAND G. N. WATSON,“A Course of Modern Analysis,” 4th ed.,
Cambrbge Univ. Press, Cambridge, 1927.
30. J. R. WILTON,A proof of Bumside’s formula for log f(x + 1) and certain allied properties
of Riemann’s c-function, Messenger Math. 52 (1922-1923), 90-93.