Arxiv:1401.3724V1 [Math.NT] 15 Jan 2014

arXiv:1401.3724v1 [math.NT] 15 Jan 2014 h ue’ constant Euler’s the oaihi eis hoyo ucin facmlxvariab complex a of functions of theory series, logarithmic logarithmi arguments, Rational function, Gamma function, words: Key discussed. and derived also are fi function the Digamma Stiel involving generalized formulæ summation first related this the several complete Besides, evaluate we to article allows short which this theorem In way. this in expressed ucin.Tems nw n rqetyecutrdof encountered frequently and known most The functions. ovretfrRe for convergent ..Introduction I.1. notations and Introduction I. generali first the that conjectured was it Recently, the theory, plines. number analysis, Euler’s modern generalized in importance as standing known also constants, Stieltjes The Abstract rpitsbitdt odmteaia ora (initial journal mathematical good a to submitted Preprint h arn eisi egbuho of neighbourhood a in series Laurent the ets n oerltvl ipe ehp vnelementar of evaluation even the perhaps on simple, relatively some and ment(s) n htterol oei ipepl at pole simple a is pole only their that and otnain ti elkonthat known well is It continuation. ζ γ fntos hyaecasclyitoue sfloigs following as introduced classically are They –functions. 1 ( ∗ mi address: +33–4 Email fax: +33–970–46–28–33; Tel.: author. Corresponding The k hoe o h lsdfr vlaino h rtgeneral first the of evaluation closed–form the for theorem A / n ) ζ where , fntosaeoeo oeipratseilfntosi mo in functions special important more of one are –functions tete osat,gnrlzdElrscntns Spec constants, Euler’s generalized constants, Stieltjes ζ ( s [email protected] = ) k and s n ∑ ∞ = γ γ > ζ 1 h rtSiljsconstant Stieltjes first the , 1 ( n tete osata ainlarguments rational at constant Stieltjes ( s n ,admyb xeddt te oan of domains other to extended be may and 1, 1 1/2 = ) r oiieitgr uhthat such integers positive are s , ) , s γ − 1 1 ( 1/3 1 + ζ ) ( , n aolvV Blagouchine V. Iaroslav s ∑ ∞ = γ ) nvriyo oln France. Toulon, of University 0 1 and ζ s ( ( ( 2/3 = umsindt:Dcme 1 03 pi ,2019 7, April 2013) 11, December date: submission − s s , = 1 v ntefloigway following the in 1 ζ Irsa .Blagouchine) V. (Iaroslav ) ) = ) ( n ihrsde1 hycnb,teeoe xaddin expanded therefore, be, can They 1. residue with 1 , s n ( γ , γ s ! v 94–14–24–48. 1 1 − n ) ( ∑ h oaih fthe of logarithm the , ∞ = 1/4 r eoopi nteetr complex entire the on meromorphic are 1 0 ) ( n k ) n γ , < + γ n e tete osata ainlargument rational at constant Stieltjes zed 1 e rhgnlexpansions. orthogonal le, eries nerl,Mlse’ integrals, Malmsten’s integrals, c 1 , osat,aeo udmna n long– and fundamental of are constants, n v ( s eeaie tete osatadthe and constant Stieltjes generalized rst r fseilfntosadohrdisci- other and functions special of ory a eawy xrse ymasof means by expressed always be may , 3/4 ) jscntn tayrtoa argument. rational any at constant tjes ζ ,fnto.Ti ojcuewsbased was conjecture This function. y, ∗ rvossuyb eiiga elegant an deriving by study previous s fntosaeReanadHurwitz and Riemann are –functions , ) a osat,Nme hoy Zeta theory, Number constants, ial , γ 1 ( 1/6 s v enaayi n hoyof theory and analysis dern ytepicpeo analytic of principle the by Γ 6 = s ) fnto trtoa argu- rational at –function , 6 = 0, γ 1 .(1) 1. − ( 5/6 1, − ) ,... . . 2, hc ol be could which , ized s –plane and 1 ∞ ( 1)n(s 1)n ζ(s, v) = + ∑ − − γn(v) , s = 1. (2) s 1 n! 6 − n=0 respectively. Coefficients γn appearing in the regular part of expansion (1) are called Stieltjes constants, while those appearing in the regular part of (2), γn(v) , are called generalized Stieltjes constants. It is obvious that γn(1) = γn because ζ(s,1) = ζ(s). The study of these coefficients is an interesting subject and may be traced back to the works of Thomas Stieltjes and Charles Hermite [1, vol. I, letter 71 and following]. In 1885, first Stieltjes, and then Hermite, proved that

m lnnk lnn+1m γn = lim ∑ , n = 0,1,2,... (3) m ∞ k n + 1 → (k=1 − ) Later, this formula was also obtained or simply stated in works of Johan Jensen [36], [38], Jørgen Gram

[28], Godfrey Hardy [31], Srinivasa Ramanujan [2] and many others. From (3), it is visible that γ0 is the Euler’s constant γ. However, the study of other Stieltjes constants revealed to be much more difficult and, at the same time, interesting. In 1895, Franel [25], by using contour integration techniques, first, rediscovered the first Jensen’s formula for ζ(s) [the first expression in (47)], and then, showed that1

∞ 1 1 dx lnn(1 ix) lnn(1 + ix) γn = δ + − , n = 0,1,2,... (4) 2 n,0 i ˆ e2πx 1 1 ix − 1 + ix 0 − − The same integral formula was rediscovered in 1985 by Ainsworth and Howell who also provided a very detailed proof of it [6]. By the same line of reasoning, one can also deduce several similar formulæ2 +∞ n+1 1 π ln 2 ix γn = ± dx n = 0,1,2,... (5) − 2(n + 1) ˆ ch2πx ∞ − and ∞ ln 2 dx ln(1 ix) ln(1 + ix) γ = γ ln 2 + i −  1 − − 2 ˆ eπx + 1 1 ix − 1 + ix  0 −   ∞  2 2 2  ln 2 dx ln (1 ix) ln (1 + ix) γ = 2γ + γ ln 2 ln 2 + i −  2 − 1 − 3 ˆ eπx + 1 1 ix − 1 + ix  " # ( )  0 −  ∞ 3 3 3 2 ln 2 dx ln (1 ix) ln (1 + ix) γ3 = 3γ2 + 3γ1 ln 2 + γ ln 2 ln 2 + i πx −  − " − 4 # ˆ e + 1 ( 1 ix − 1 + ix )  0 −    ......    1See the related priority dispute between Jensen, Kluyver and Franel [25], [38], [37]. Besides, we corrected the Franel’s formula from [25] which was not valid for n = 0 [this correction comes from (8) and (9) here after]. 2The proof is analogous to that given for the formula (8) here after, except that the Hermite representation should be replaced by the third and second Jensen’s formulæ for ζ(s) (47) respectively.

2 ﬁrst of which is particularly simple. Other important results concerning the Stieltjes constants were obatined by studying series implying integer part of functions. In 1910, Giovanni Vacca discovered a new series for the Euler’s constant

∞ ( 1)k γ = ∑ − log2 k (6) k=2 k ⌊ ⌋ for which he gave a simple and elegant geometrical proof [57]. Glaisher, being puzzled by such an unusual demonstration, proposed an arithmetical proof of the same result [27]. Two years later, Hardy, by another analytical method, not only reproved the Vacca’s result (6), but also extended it to the ﬁrst Stieltjes constant

ln 2 ∞ ( 1)k γ1 = ∑ − log2 k 2 log2 k log2 2k 2 k=2 k ⌊ ⌋· − ⌊ ⌋ see [31]. Similar expressions for higher Stieltjes constants were later given by Kluyver [39] (Hardy, however, already mentioned this possibility in [31]). Apart from γ0, no closed–form expressions are known for γn; however, there are works devoted to their estimations [41], [9], [42], [35], and to the behaviour of their sign [11], [49]. In particular, Briggs in 1955 [11] demonstrated that there are inﬁnitely many changes of sign for them (later, Mitrovi´c[49] extended some results of Briggs). Gram [28], Liang & Todd [43], Ainsworth & Howell [6] and Kreminski [40] discussed some aspects related to the numerical computation of Stieltjes constants.3 As regards generalized Stieltjes constants, they are much less studied than the usual Stieltjes constants. In 1927, Wilton [60], showed that they can be given by an asymptotic representation of the same kind as (3)

m lnn(k + v) lnn+1(m + v) γn(v) = lim ∑ , n = 0,1,2,... v = 0, 1, 2,... m ∞ k n + 1 → (k=1 − ) 6 − − (7) which was also derived by Berndt in In 1972, Berndt, being inspired by the similar Lammel’s proof for the Riemann ζ–function [41], also proved this formula in [9]. Similarly to the Franel’s method of the derivation of (4), one may derive the following integral representation for the nth generalized Stieltjes constant ∞ n n 1 ln v n dx ln (v ix) ln (v + ix) γn(v) = ln v i − , n = 0,1,2,... 2v − n + 1 − ˆ e2πx 1 v ix − v + ix 0 − − (8) Re v > 0.4 A variant of this formula was also obtained by Mark Coffey [15], [19] (by the way, he also gave an equivalent of the formula (5) for the generalized Stieltjes constant). From both latter formulæ, it follows that γ (v) = Ψ(v). Take, for instance, (8) and put n = 0. Then, the latter equation takes 0 −

3In the latter work, author also considered aspects related to the generalized Stieltjes constants. 4Proof. Consider the well-known Hermite representation for ζ(s, v), see e.g. [8, vol. I, p. 26, Eq. 1.10(7)]. Recall ﬁrst that 2(v2 + x2) s/2 sin[s arctg(x/v)] = i[(v ix) s (v + ix) s], and then, expand 1 v s +(s 1) 1v1 s into the Laurent series − − − − − − 2 − − − − about s = 1. Performing the term–by–term comparison of the derived expansion with the Laurent series (2) yields (8). QED

3 the form ∞ 1 xdx γ (v) = ln v + 2 = Ψ(v) , n = 0,1,2,... (9) 0 2v − ˆ (e2πx 1)(v2 + x2) − 0 −

1 + 1 ln v 1 Ψ(v) − 4v 2 − 2 | {z } where the last integral was ﬁrst calculated by Legendre5. The demonstration of the same result from formula (7) may be found, for example, in [50]. For rational v, the 0th Stieltjes constant may be, therefore, expressed by means of the Euler’s constant γ and a ﬁnite combination of elementary functions (thanks to the Gauss’ Digamma theorem [26], [8, vol. I, p. 19, 1.7.3]). However, things are much § more complicated for higher generalized Stieltjes constants; currently, no closed–form expressions are known for them and little is known as to their arithmetical properties. Basic properties, such as the multiplication theorem

n 1 p 1 − l p ln n p − r r r ∑ γp v + = ( 1) n Ψ(nv) ln n + n ∑ ( 1) Cpγp r(nv) ln n , n − p + 1 − − − · l=0 r=0 p! n = 2,3,4,..., where Cr denotes the binomial coefﬁcient Cr = , and the recurrent rela- p p r!(p r)! tionship −

lnpv γ (v + 1) = γ (v) , p = 1,2,3,... v = 0, 1, 2,... (10) p p − v 6 − − may be both straightforwardly derived from those for the Hurwitz ζ–function, see e.g. [10, exercise no 64, Eq. (63)–(65)].6 In attempt to obtain other properties, several summation relations involving single and double infinite series were quite recently obtained in [13], [14]. Also, many important aspects regarding the Stieltjes constants were considered by Donal Connon in works [19], [18], [17]. Let now focus our attention on the first generalized Stieltjes constant. The most strong and per- tinent results in the field of its closed–form evaluation is the formula for the difference between the first generalized Stieltjes constant at rational argument and its reflected version

n 1 m m − 2πml l mπ γ γ 1 = 2π ∑ sin ln Γ π(γ + ln 2πn) ctg (11) 1 n − 1 − n n · n − n l=1 There are several ways to prove this important relationship.7 For instance, one can study the ﬁrst derivative of the functional equation for ζ(s, v), equation (22), with respect to s at rational v, see e.g. [7, p. 261, 12.9], [48, eq. (6)]. Another way consists in the study of Malmsten’s integrals at rational § arguments; we developed such a method in our precedent study [10, exercises no 63 and 67]. Recently, Coffey derived several interesting representations for the generalized Stieltjes constants and for their differences [16]. From one of these representations, one may conjecture that in some cases (author gave only two examples of such cases [16, p. 1821, Eqs. (3.33)–(3.34)]),not only the Γ–function at rational argument (which is more or less predictable from the preceding formula), but also the

5And not by Binet as stated in [8, vol. I, p. 18, Eq. 1.7.2(27)], see [55, vol. II, p. 190] and [10, exercise no 40, eq. (55)]. 6As regards the multiplication theorem, see e.g. [18, Eq. (6.6)] or [10, exercise no 64]. We can also ﬁnd its particular case for v = 1/n in [16, p. 1830, eq. (3.28)]. 7Accordingly to [4], it was ﬁrst proved by Almkvist and Meurman in a private communication.

4 second–order derivative of the Hurwitz ζ–function could be related, in some way, to the first generalized Stieltjes constant. However, these preliminary findings do not permit to precisely identify their roles in the general problem of the closed–form evaluation of the first Stieltjes constant at any rational argument (the problem which we come to solve here).

Very recently, it has been conjectured that similarly to the Digamma theorem for γ0(v), the ﬁrst generalized Stieltjes constant γ1(v) at rational v may be expressed by means of the Euler’s constant γ, the ﬁrst Stieltjes constant γ1 , the logarithm of the Γ–function and some “relatively simple” function, that is to say

m 1 r − l γ = f (l, m, γ) + ∑ α (r, m) ln Γ + γ , r = 1,2,..., m 1, (12) 1 m l · m 1 − l=1 where αl(r, m) are real coefficients and f (l, m, γ) is some “relatively simple” function, see [10, exer- o 1 1 1 1 2 3 5 cise n 64]. For seven rational values of v in the range (0,1), namely for 6 , 4 , 3 , 2 , 3 , 4 and 6 , it has been shown in [10, exercise no 64] that this “relatively simple” function is elementary.8 In this short manuscript, we extend these precedent researches by providing a theorem which allows to evaluate the first generalized Stieltjes constant at any rational argument in a closed–form and by precisely identifying this “relatively simple” function. The latter consists of elementary functions (containing the Euler’s constant γ as well) and of the reflected sum of two second–order derivatives of the Hur- witz ζ–function at zero ζ (0, p) + ζ (0,1 p), number p being rational in the range (0,1). Curiously ′′ ′′ − enough, the derived theorem represents also the finite Fourier series for the first generalized Stieltjes constant, so that classic Fourier analysis tools may be used at their full strength. With the help of the latter, we derive several summation formulæ including summation with trigonometric functions and summation with square. Obviously, the same method can be applied to other discrete functions allowing similar representations. In particular, its application to the little–known Malmsten’s representation for the Ψ–function yields several beautiful summation formulæ for the Digamma function which are derived in appendices.

I.2. Notations Throughout the manuscript, following abbreviated notations are used: γ = 0.5772156649 . . .for the Euler’s constant, γn for the nth Stieltjes constant, γn(p) for the nth generalized Stieltjes constant at point p, x for the integer part of x, tg z for the tangent of z, ctg z for the cotangent ⌊ ⌋ of z, ch z for the hyperbolic cosine of z, sh z for the hyperbolic sine of z, th z for the hyperbolic tangent of z, cth z for the hyperbolic cotangent of z.9 In order to avoid any confusion between compositional inverse and multiplicative inverse, inverse trigonometric and hyperbolic functions are denoted as 1 1 1 arccos, arcsin, arctg, . . . and not as cos− , sin− , tg− , . . . Writings Γ(z), Ψ(z), ζ(s) and ζ(s, v) denote respectively the Γ–function, the Ψ–function (or Digamma function), the Riemann ζ–function and the Hurwitz ζ–function. When referring to the derivatives of the the Hurwitz ζ-function, we always refers to the derivative with respect to its first argument s (unless otherwise specified). Re z and Im z denote, respectively, real and imaginary parts of z. Natural numbers are defined in a traditional way

as a set of positive integers, which is denoted by Æ. Kronecker symbol of arguments l and k is denoted

8Further to remarks we received after the publication of [10], we note that similar closed–form expressions for γ(1/2), γ(1/4), γ(3/4) and γ(1/3) were also obtained in [17, pp. 17–18]. 9Most of these notations come from Latin, e.g “ch” stands for cosinus hyperbolicus, “sh” stands for sinus hyperbolicus, etc.

5 by δ . Letter i is never used as index and is √ 1. By Malmsten’s integral we mean any integral of l,k − the form ∞ R(sh px, ch px) ln x · dx ˆ R(sh x, ch x) 0 where R denotes a rational function and the parameter p is such that the convergence is guaranteed. Moreover, if p is rational, then by an appropriate change of variable, the above Malmsten’s integral may be reduced to the following ln ln–integrals

∞ P(u) ln ln u du + ... ∞ ˆ Q(u) R(sh px, ch px) ln x · dx = 1 ˆ R(sh x, ch x)  1  R(y) 1 0  ln ln dy + ... ˆ S(y) y 0   where P(u), Q(u), R(y) and S(y) are polynomials in u and y respectively, see [10] for more details. Other notations are standard.

II. Evaluation of the ﬁrst generalized Stieltjes constant at rational argument

II.1. Generalized Stieltjes constants and their relationships to Malmsten’s integrals The formula (11) provides a closed–form expression for the difference of two first Stieltjes constants at rational arguments. It should be therefore interesting to investigate if there could be some expressions containing other combinations of Stieltjes constants. In our previous work [10], we already demonstrated that some Malmsten’s integrals are connected with the first generalized Stieltjes constants. This connection was quite fruitful and permitted not only to prove by another method the known relationship (11), but also to evaluate the first generalized Stieltjes constant γ1(p) at 1 1 1 1 2 3 5 p = 2 , 3 , 4 , 6 , 3 , 4 , 6 by means of elementary functions, the Euler’s constant γ, the first Stieltjes o constant γ1 and the Γ–function, see for more details [10, exercise n 64]. Taking into account that afore- mentioned manuscript was quite long, many results and theorems were given as exercises with hints and without rigorous proofs. Below, we provide several useful proofs and unpublished results (given as lemmas and corollaries) showing that Malmsten’s integrals of the first and second orders may be expressed by means of the first generalized Stieltjes constants. This connection between Malmsten’s integrals and Stieltjes constants is crucial and plays the central role in the proof of the main theorem of this manuscript.

Lemma 1. For any Re p < 1 and Re a > 1, | | − ∞ xa 1(ch px 1) Γ(a) 1 p 1 p − − dx = ζ a, + ζ a, + 2 2a 1 ζ(a) (13) ˆ sh x 2a 2 − 2 2 2 − − 0 1 Proof. From elementary analysis it is well-known that sh− x, for Re x > 0, may be represented by the following geometric series

∞ 1 (2n+1)x = 2 ∑ e− , Re x > 0. sh x n=0

6 This series, being uniformly convergent, can be integrated term–by–term. Hence

∞ ∞ xa 1(ch px 1) ∞ − − dx = xa 1 e (2n+1 p)x + e (2n+1+p)x 2e (2n+1)x dx ˆ ∑ ˆ − − − − − sh x n=0 − 0 0 n o ∞ 1 1 2 = Γ(a) ∑ + (2n + 1 p)a (2n + 1 + p)a − (2n + 1)a n=0 −

Γ(a) 1 p 1 p 1 = ζ a, + ζ a, + 2ζ a, , 2a 2 − 2 2 2 − 2 where the integral on the left converges if Re p < 1 and Re a > 1. In order to obtain (13), it sufﬁces | | − to notice that ζ a, 1 = (2a 1) ζ(a) . QED 2 − Corollary 1. For any p lying in the strip Re p < 1, we always have | | ∞ (ch px 1) ln x 1 p π πp − dx=(γ + ln 2) Ψ + + ln 2 tg ˆ sh x · 2 2 − 2 2 0 1 1 p 1 1 p + γ2 + γ γ + γ . (14) 1 − 2 1 2 2 − 2 1 2 − 2 This result is straightforwadly obtained from Lemma 1 by differentiating (13) with respect to a, and then by making a 1. In order to evaluate the limit in the right–hand side, we make use of Laurent series (1) and (2). → Another Malmsten’s integral of the ﬁrst order whcih also contains Stieltjes constants appear in the next Lemma.

Lemma 2. For any Re p < 1 and Re a > 1, | | − ∞ xa 1 sh px Γ(a) 1 p 1 p − dx = ζ a, + ζ a, ˆ ch x 2a 2 2 − 2 − 2 0 1 a 1 p 1 a 1 p 2 − ζ a, + + 2 − ζ a, − 4 4 4 − 4 Proof. Analogous to that of Lemma 1.

Corollary 2. For any Re p < 1, | | ∞ sh px ln x 1 πp 1 p 1 p · dx = π(γ + ln 2) tg (γ + 2ln2) Ψ + Ψ ˆ ch x 2 2 − 4 4 − 4 − 4 0 1 p 1 p 1 p 1 p + γ γ + γ + γ + , 1 2 − 2 − 1 2 2 − 1 4 − 4 1 4 4 This result can be shown in the same way as that in Corollary 2.

By the same line of reasoning, one may also prove that following logarithmic integrals may be

7 expressed in terms of ﬁrst generalized Stieltjes constants.

∞ sh px ln x 1 πp 1 p 1 p · dx = π(γ + ln 2) tg + γ γ + ˆ sh x − 2 2 1 2 − 2 − 1 2 2 0 ∞ ch px ln x 1 1 p 1 p 1 p 1 p 1 · dx = γ + + γ γ + γ ln22 ˆ ch x 2 1 2 2 1 2 − 2 − 1 4 4 − 1 4 − 4 − 2 0 1 p π πp π π πp + ln 2 Ψ + + (γ + ln 2) tg (γ + 2ln2) ctg · 2 2 2 2 − 2 4 − 4 ∞ sh2 px ln x 1 · dx = ln π ln sin πp + p γ1(p) γ1(1 p) (γ + ln 2)(1 πp ctg πp) ˆ sh2x 2 − − − − − 0 n o ∞ ch px ln x p p p p p · dx = γ1 γ1 1 γ1 + γ1 1 ˆ ch2x 2 2 − − 2 − 4 − 4 0 πp πp πp πp (γ + 2ln2) csc + ln 2 ctg + ln tg − 2 2 · 2 4 n o where parameter p should be such that Re p < 1 in the ﬁrst three integrals and Re p < 2 in the | | | | fourth one.

II.2. Malmsten’s series and Hurwitz’s reﬂection formula We now show that the integral form Lemma 1 may be also evaluated via a trigonometric series.

Lemma 3. For any 1 < p < 1 and Re a < 1, − | | ∞ xa 1(ch px 1) πa ∞ cos pπn 1 − − dx = πasec ( 1)n − (15) ˆ ∑ 1 a sh x 2 n=1 − n − 0 Proof. The Mittag–Lefﬂer theorem is a fundamental theorem in the theory of functions of a complex variable and allows to expand meromorphic functions into a series accordingly to its poles.10 Ap- plication of this theorem to the meromorphic function (ch pz 1)/ sh z, p ( 1, +1), having only n − ∈ − ﬁrst–order poles at z = πni, n Z, with residue ( 1) (cos πpn 1), leads to the following expansion ∈ − − ∞

ch pz 1 n cos pπn 1 Z − = 2z ∑ ( 1) 2 2−2 , z C , z = πni , n . sh z n=1 − z + π n ∈ 6 ∈

which is uniformly convergent on the entire complex z–plane except discs z πin < ε, n Z, of | − | ∈ arbitrary small radius ε. Therefore

∞ ∞ xa 1(ch px 1) ∞ xa πa ∞ cos pπn 1 − − dx = 2 ( 1)n(cos pπn 1) dx = πasec ( 1)n − ˆ ∑ ˆ 2 2 2 ∑ 1 a sh x n=1 − − x + π n 2 n=1 − n − 0 0

1 a a 1 1 2 π n − sec 2 πa | {z } (16)

10For more details, see [47], [58, pp. 147–148, no 994–1002], [24, Chap. V, § 27, no 27.10-2], [54, Chap. VII, p. 175], [44].

8 which holds only for 1 < p < 1 and Re a < 1 (the elementary integral in the middle, whose − | | evaluation is due to Euler, is convergent only in the strip Re a < 1, see e.g. [58, p. 126, no 880], [21, | | p. 197, no 856.2], [3, p. 256, no 6.1.17], [29, p. 67, no 587], [44, p. 51]). However, the above equality can be analytically continued for other values of a: the integral is the analytic continuation of the sum for Re a > 1 , while the sum analytically continues the integral for Re a 6 1 . We obviously − have to expect trouble with the right–hand part at a = 1, 3, 5, . . . because of the secant. Since ± ± ± when a = 1, 3, 5, . . . the sum in the right–hand side converges, these points are poles of the ﬁrst − − − order for the analytic continuation of integral (16). In contrast, for a = 1, 3, 5, . . ., the integral on the left remains bounded, and thus, these points are removable singularities for the right–hand side of (16). In other words, formally ∑( 1)n(cos pπn 1)na 1, n > 1 , must vanish identically for any odd − − − positive a (exactly as η(1 a), the result which has been derived by Euler, see e.g. [22, p. 85]). These − matters are treated in detail in the next Corollary. QED

Corollary 3. For 0 < p < 1 and Re a < 1,

∞ cos2πpn πa Γ a ∑ 1 a = (a)(2π)− cos ζ(a, p) + ζ(a,1 p) n − 2 − n=1 n o  (17a,b)  ∞  sin 2πpn a πa ∑ = Γ(a)(2π)− sin ζ(a, p) ζ(a,1 p) n1 a 2 − − n=1 −  n o  Proof. In view of the fact the alternating ζ–function η(s) may be reduced to the Riemann ζ–function and by making use of the well–known reﬂection formula for the Riemann ζ–function11 ζ(1 s) = − Γ s 1 2ζ(s) (s)(2π)− cos 2 πs, we may continue (16) as follows

∞ ∞ a πa n cos pπn 1 a πa n cos pπn a π sec ∑ ( 1) 1 a − = π sec ∑ ( 1) 1 a (2 1)ζ(1 a) 2 n=1 − n − 2 (n=1 − n − − − − )

πa ∞ cos pπn a n a Γ = π sec ∑ ( 1) 1 a 2(1 2− ) (a)ζ(a) 2 n=1 − n − − −

Comparing the latter expression to the result of Lemma 1 gives

∞ n cos pπn a πa 1 p 1 p ∑ ( 1) = Γ(a)(2π)− cos ζ a, + + ζ a, − n1 a 2 2 2 2 − 2 n=1 − Writing in this expression 2p 1 instead of p yields immediately (17a). Now, by partially differenti- − ating (17a) with respect to p and by remarking that aΓ(a) = Γ(a + 1), and then, by writing a instead of a + 1, we arrive at (17b). Note also that both sums (17a,b) may be analytically continued to other domains of a by means of expressions in corresponding right parts. QED

11There is an interesting history related to this famous relationship. It was first proposed by Leonhard Euler in 1749 in [22], who derived it by the method of mathematical induction. Then, it was independently rediscovered and rigorously proved by Bernhard Riemann in 1859 [51], [20, p. 861]. Very similar reflection formulæ were obtained by Carl Malmsten in 1842 [45] and in 1846 [46], as well as by Oscar Schlömilch in 1849 [52], [53], [34], [32, p. 23], [10]. By the way, Malmsten, unlike Riemann, remarked that the proved formula is analogous to that already conjectured by Euler. Thus it seems strange that Riemann ignored not only the existence of the Euler’s work [22], but also that of Malmsten’s works [45] and [46], the latter of which was published in one of the most known mathematical journals of that time.

9 Nowadays, the results (17a,b) seem to be not particularly well–known (for instance, advanced calculators such as Wolfram Alpha Pro expresses both series in terms of polylogarithms). Notwith- standing, equation (17b) can be found in an old Malmsten’s work published as early as 1849 [46, p. 17, eq. (48)], and (17a) is a straightforward consequence of (17b).12

Corollary 4. If we notice that π π Γ(a) = = sin πa Γ(1 a) 2 sin 1 πa cos 1 πa Γ(1 a) · − 2 · 2 · − then, the sum of (17a) with (17b) leads to the well–known formula

Γ( ) ∞ ∞ 2 1 a πa cos2πpn πa sin 2πpn < 6 < ζ(a, p) = −1 a sin ∑ 1 a + cos ∑ 1 a , 0 p 1, Re a 1 (2π) − " 2 n=1 n − 2 n=1 n − # (18) which is usually attributed to Adolf Hurwitz who derived it in 1881, see [34, p. 93]13, [59, p. 269], [44, p. 107], [56, p. 37], [9, p. 156], [8, vol. I, p. 26, Eq. 1.10(6)].14

Remark It is quite rarely emphasized that the latter representation coincide with the trigonometric Fourier series for ζ(a, p). Remarking this permits to immediately derive several integral formulæ, whose demonstration by other means is more difﬁcult

1 ζ(a, p) dp = 0 ˆ 0  1   a 1 πa  ζ(a, p) cos2πpndp = Γ(1 a)(2πn) − sin ˆ − 2 0 1 πa  Γ a 1  ζ(a, p) sin 2πpndp = (1 a)(2πn) − cos ˆ − 2 0   for Re a < 1 and n = 1, 2, 3, . . .. Furthermore, in virtue of the Parseval’s theorem, we have

1 2 2 2a 2 1 ζ (a, p) dp = 2Γ (1 a)(2π) − ζ(2 2a) , Re a < 1, a = (19) ˆ − − 6 2 0

Differentiating this formula with respect to a and then setting a = 0, yields:

1 1 1 γ + ln 2π ζ (2) 2 p ln Γ(p) ln 2π dp = ′ ˆ 2 − · − 2 6 − π2 0 ζ(0,p) ζ′(0,p) | {z } | {z }

12Moreover, Malmsten derived his reﬂection formulæ (see footnote 11) precisely from this equality. 13Hurwitz derived all his results for the function f (s, a) which is related to the modern Hurwitz ζ–function as f (s, a) s ≡ fm(s, a)= m− ζ(s, a/m), see [34, p. 89]. 14There is a slight error in this formula in the latter reference: it remains valid not only for Re a < 0, but also for Re a < 1.

10 Whence, accounting for the well–known result15

1 1 ln Γ(x) dx = ln 2π ˆ 2 0 we obtain 1 ζ (2) γ 2ln2π x ln Γ(x) dx = ′ − ˆ 2π2 − 12 0 Integration by parts of the latter expression leads to the antiderivatives of ln Γ(x) which are currently not well–studied yet.

Corollary 5. In (18), the index n may be represented as n = mk + l, where for each k = 0,1,2,..., ∞, the index l runs over [1,2,..., m] and where m is some positive integer. Then, (18) may be written in the form:

2Γ(1 a) πa m ∞ cos2πp(mk + l) πa m ∞ sin 2πp(mk + l) ζ(a, p) = −1 a sin ∑ ∑ 1 a + cos ∑ ∑ 1 a (20) (2π) − " 2 l=1 k=0 (mk + l) − 2 l=1 k=0 (mk + l) − #

Now, let p be a rational part of m, i.e. p = r/m, where r and m are positive integers such that r 6 m. Then cos 2πp(mk + l) = cos(2πrl/m), and similarly for the sine. Hence, for positive rational p not greater than 1, equation (20) takes the form

r 2Γ(1 a) πa m 2πrl ∞ 1 πa m 2πrl ∞ 1 ζ a, = −1 a sin ∑ cos ∑ 1 a + cos ∑ sin ∑ 1 a m (2π) − " 2 m (mk + l) − 2 m (mk + l) − # l=1 k=0 l=1 k=0 ma 1ζ(1 a, l/m) − − 2Γ(1 a) πa m 2πrl | {z l } πa m 2πrl l = − sin ∑ cos ζ 1 a, + cos ∑ sin ζ 1 a, (2πm)1 a 2 m · − m 2 m · − m − " l=1 l=1 #

2Γ(1 a) m 2πrl πa l = − ∑ sin + ζ 1 a, , r = 1,2,..., m. (2πm)1 a m 2 · − m − l=1 (21) This equality holds in the entire complex a–plane for any positive integer m > 2. Furthermore, by putting in the latter formula 1 a instead of a, it may be rewritten as − r 2Γ(a) m 2πrl πa l ζ 1 a, = ∑ cos ζ a, , r = 1,2,..., m. (22) − m (2πm)a m − 2 · m l=1 In the case r = m, above formulæ reduce to the reﬂection formulæ for the Riemann ζ–function (simply use the multiplication theorem for the Hurwitz ζ–function, see e.g. [10]). Formulæ (21) and (22) are known as functional equations for the Hurwitz ζ–function and were both obtained by Hurwitz in the same article [34, p. 93] in 1881. By the way, the above demonstration also shows that they can be elementary derived from Malmsten’s results (17a,b) obtained as early as 1840ies.

15This result is straightforward from a similar Fourier series expansion for the logarithm of the Γ–function, see e.g. [8, vol. I, pp. 23–24, § 1.9.1]. This expansion, attributed erroneously to Ernst Kummer, was ﬁrst derived by Malmsten and colleagues from the Uppsala University in 1842. We discuss this interesting historical question in details in [10, Sect. 2.2, Fig. 2 and exercise no 20].

11 II.3. Closed–form evaluation of the first generalized Stieltjes constant at rational argument and some related results We now state the main result of this manuscript allowing to evaluate in a closed–form the first generalized Stieltjes constant at any rational argument. Theorem The first generalized Stieltjes constant of any rational argument in the range (0,1) may be expressed in a closed form via a finite combination of logarithms of the Γ–function, of second–order derivatives of the

Hurwitz ζ–function at zero, of the Euler’s constant γ, of the ﬁrst Stieltjes constant γ1 and of elementary functions: r 1 π πr γ = γ γ ln 2m ln22 ln 2 ln πm ln2m (γ + ln 2πm) ctg 1 m 1 − − − · − 2 − 2 m ( 1)r πr πr m + 1 πr m 1 − 1 ( 1)m+1 (3ln2 + 2 ln π) ln 2 π ln π csc sin sin − − 4 − − · − · m · m 2 · m 2 1 (m 1) 1 (m 1) 1 (m 1) ⌊2 − ⌋ 2πrl πl ⌊2 − ⌋ 2πrl πl ⌊2 − ⌋ 2πrl l +2(γ + ln 2πm) ∑ cos ln sin + π ∑ sin ln sin + 2π ∑ sin ln Γ · m · m m · m m · m l=1 l=1 l=1

1 (m 1) ⌊2 − ⌋ 2πrl l l + ∑ cos ζ′′ 0, + ζ′′ 0, 1 m · m − m l=1 (23) This elegant formula holds for any r = 1,2,3,..., m 1, where m is positive integer greater than 1. The − Stieltjes constants for other “periods” may be obtained from the recurrent relationship: ln v γ (v + 1) = γ (v) , v = 0, (24) 1 1 − v 6 see, e.g. [10, exercise no 64, Eq. (64)]. The above theorem is an equivalent of the Gauss’ Digamma theorem for the 0th Stieltjes constant γ (r/m) = Ψ(r/m). Two alternative forms of the same theorem are given in 0 − equations (35) and (38). Proof of the Theorem Consider the integral (13). Put 2p 1 instead of p and denote the resulting − integral via Ja(p): ∞ xa 1(ch [(2p 1)x] 1) Γ(a) J (p) − − − dx = ζ(a, p) + ζ(a,1 p) 2 2a 1 ζ(a) , 0 < Re p < 1 a ≡ ˆ sh x 2a − − − 0 (25) Let now p be rational p = r/m, where r and m are positive integers such that r < m. Then, the precedent equation becomes

r Γ(a) r r a J = ζ′′ a, + ζ′′ a,1 2 2 1 ζ(a) (26) a m 2a m − m − − The sum of ﬁrst two terms in curly brackets may be evaluated via the Hurwitz’ reﬂection formula (21): r r 2Γ(1 a) m 2πrl πa 2π(m r)l πa ζ′′ a, + ζ′′ a,1 = − ∑ sin + + sin − + m − m (2πm)1 a m 2 m 2 − l=1

l 4Γ(1 a) πa m 2πrl l ζ 1 a, = − sin ∑ cos ζ 1 a, × − m (2πm)1 a 2 · m · − m − l=1

12 Thus, by noticing that Γ(a)Γ(1 a) = 1 π csc 1 πa sec 1 πa , the integral J (r/m) takes the form: − 2 2 · 2 a r 2π πa m 2πrl l 2Γ(a) 2a 1 ζ(a) Ja = sec ∑ cos ζ 1 a, − (27) m 2a(2πm)1 a 2 · m · − m − 2a − l=1 f f3 1 f2 | {z } | {z } | {z } which is third expression for the integral Ja, other two expressions being given by (13) and (15). Let now study each term of the right part, denoted for brevity f1, f2 and f3 respectively, in a neighbourhood of a = 1. The ﬁrst and the third terms have poles of the ﬁrst order at this point, while the second term f2 is analytic at a = 1. Thus, in a neighbourhood of a = 1, terms f1 and f3 may be expanded in the Laurent series as follows

2 π2 f = 2 ln πm ln2 πm (a 1) + O(a 1)2 (28) 1 − a 1 − − 12 − · − − − ! and 1 π2 ln22 γ2 f = + ln 2 + γ (a 1) + O(a 1)2 (29) 3 a 1 12 − 2 − 2 − 1 · − − − ! while f2 may be represented by the following Taylor series

m 2πrl l m 2πrl l (a 1)2 m 2πrl l f = ∑ cos ζ 0, (a 1)∑ cos ζ′ 0, + − ∑ cos ζ′′ 0, 2 m · m − − m · m 2 m · m l=1 l=1 l=1 1 l/m ln Γ(l/m) 1 ln 2π 2 − − 2 | {z } m | {z } 2 m 3 1 2πrl l (a 1) 2πrl l 3 +O(a 1) = (a 1) ∑ cos ln Γ + − ∑ cos ζ′′ 0, + O(a 1) − − 2 − − m · m 2 m · m − l=1 l=1 (30) because m 2πrl ∑ cos = 0 r = 1,2,3,..., m 1 m − l=1   m 2πrl m ∑ l cos = , r = 1,2,3,..., m 1 l=1 · m 2 −   In the ﬁnal analysis, the substitution of (28), (29) and (30) into (27), yields the following representation for the integral Ja(r/m) in a neighbourhood of a = 1:

r πm π2 ln2πm γ2 ln22 J = ln + 2A (r)+(a 1) B (r) + 2A (r) ln πm + + + + γ a m 2 m − · − m m − 24 2 2 2 1 " # +O(a 1)2 − (31) where m 2πrl l Am(r) ∑ cos ln Γ ≡ m · m  l=1    m 2πrl l Bm(r) ∑ cos ζ′′ 0, ≡ m · m  l=1  

13 Now, if we look at the integral Ja(r/m) deﬁned in (26), we see that it is uniformly convergent and regular near a = 1 (see appendix C), and hence, may be expanded in the Taylor series about a = 1:

∂J (r/m) J (r/m) = J (r/m)+(a 1) a + O(a 1)2 (32) a 1 − ∂a − a=1

Equating right–hand sides of (31) and (32), and then, searching for terms with same powers of (a 1), − gives

∞ ch [(2p 1)x] 1 πm − − dx = ln + 2Am(r) ˆ sh x 2 0    ∞  (ch [(2p 1)x] 1) ln x π2 ln2πm γ2 ln22 − − dx = Bm(r) + 2Am(r) ln πm + + + + γ1 ˆ sh x − − 24 2 2 2 0   where p r/m. The sum A (r) may be reduced either to elementary functions (if using the reﬂection ≡ m formula for the logarithm of the Γ–function) or to the Ψ–function and the Euler’s constant γ (see appendix B). We, for the purpose of brevity, prefer to use the latter representation for Am(r). Thus, by using (52), the ﬁrst of the above integrals may be calculated as

∞ ch [(2p 1)x] 1 π πr r r − − dx = γ 2ln2 ctg Ψ , p ˆ sh x − − − 2 m − m ≡ m 0 while the second one reduces to ∞ (ch [(2p 1)x] 1) ln x m 2πrl l π πr r − − dx = cos ζ 0, γ + ln 2 + ctg + Ψ ln πm ˆ ∑ ′′ sh x − l=1 m · m − 2 m m 0

2 2 2 2 m 1 π ln πm γ ln 2 − 2πrl l π πr r + + + γ = ∑ cos ζ′′ 0, γ + ln 2 + ctg + Ψ ln πm − 24 − 2 2 2 1 − m · m − 2 m m l=1

ln2πm ln22 ln22π r + + , p − 2 2 2 ≡ m (33) where at the ﬁnal stage we separate the last term in the sum Bm(r) whose value is known

γ2 π2 ln22π ζ′′(0,1) = ζ′′(0) = γ + 1 2 − 24 − 2 But the integral (33) was also evaluated in (14) by means of ﬁrst generalized Stieltjes constants. Hence, the comparison of (14) to (33) yields

m 1 r r − 2πrl l r π πr γ + γ 1 = 2 ∑ cos ζ′′ 0, + 2(γ + ln 2πm) Ψ + ctg 1 m 1 − m m · m · m 2 m l=1 (34)

+2(γ + ln 2) ln 2πm + ln2πm ln22 ln22π + 2γ2 + 2γ − − 1

14 for each r = 1,2,..., m 1. Adding this to (11) and simplifying the result ﬁnally gives − r 1 r γ = γ + γ2 + γ ln 2πm + ln 2π ln m + ln2m +(γ + ln 2πm) Ψ 1 m 1 · 2 · m (35) m 1 m 1 − 2πrl l − 2πrl l +π ∑ sin ln Γ + ∑ cos ζ′′ 0, m · m m · m l=1 l=1 This is the most simple form of the theorem which we are stating here and can be used as is. Notwith- standing, we may also notice that each of two sums from the right–hand side may be further simpli- ﬁed. Since each pair of terms which occupy symmetrical positions relatively to the center (except for l = m/2 when m is even) may be grouped together, the last sum may be reduced to m 1 − 2πrl l ∑ cos ζ′′ 0, = m · m l=1

1 (m 1) 2 − 2πrl l l ∑ cos ζ′′ 0, + ζ′′ 0, 1 , if m is odd  l=1 m · m − m  =  1  2 m 1  − 2πrl l l r 1 ∑ cos ζ′′ 0, + ζ′′ 0, 1 +( 1) ζ′′ 0, , if m is even m · m − m − 2  l=1   1  2 (m 1) r ⌊ − ⌋ 2πrl l l ( 1) m+1 =∑ cos ζ′′ 0, + ζ′′ 0, 1 − 1 ( 1) (3ln2 + 2 ln π) ln 2 m · m − m − 4 − − · l=1 (36) because ζ 0, 1 = 3 ln22 ln π ln 2, see e.g. [10, exercise no 24]. Similarly, ′′ 2 − 2 − 1 (m 1) m 1 2πrl l ⌊2 − ⌋ 2πrl l l ∑− sin ln Γ = ∑ sin ln Γ ln Γ 1 m · m m · m − − m l=1 l=1 2 ln Γ(l/m)+lnsin πl ln π m − 1 (m 1) 1 (m 1) ⌊2 − ⌋ 2πrl l ⌊2 − ⌋ 2π|rl πl {z } (37) = 2 ∑ sin ln Γ + ∑ sin ln sin m · m m · m l=1 l=1 πr πr m + 1 πr m 1 ln π csc sin sin − − · m · m 2 · m 2 because for natural n n x nx x ∑ sin(lx) = csc sin sin (n + 1) 2 · 2 · 2 l=1 h i see e.g. [29, no 58, p. 12]. Thus, by using (36) and (37), as well as the Gauss’ Digamma theorem, equation (35) reduces to (23). QED. In some cases, it may be more advantageous to have the complete ﬁnite Fourier series form. For this aim, it sufﬁces to take again (35) and to use the Malmsten’s representation for the Ψ–function, see

15 appendix B, formulæ (52)–(53). This yields the following expression r 1 γ = γ γ ln 2πm ln 2π ln m ln2m ln22π 1 m 1 − − · − 2 − m 1 − 2πrl l l(γ + ln 2πm) +π ∑ sin ln Γ + (38) m · m m l=1 m 1 − 2πrl l l + ∑ cos ζ′′ 0, 2(γ + ln 2πm) ln Γ m · m − m l=1 where r = 1,2,3,..., m 1, and m is positive integer greater than 1. − Corollary 6. For the ﬁrst generalized Stieltjes constant at rational argument take place following summation formulæ

m 1 − r 2πrk kπ m k k ∑ γ cos = γ + m(γ + ln 2πm) ln 2 sin + ζ′′ 0, + ζ′′ 0, 1 1 m · m − 1 · m 2 m − m  r=1   m 1  − r 2πrk π πm πk k ∑ γ1 sin = (γ + ln 2πm)(2k m) ln π ln sin + mπ ln Γ  r=1 m · m 2 − − 2 − m m   (39a,b)  for k = 1,2,3,..., m 1, where m is natural greater than 1. − Proof. Formula (38) represents the ﬁnite Fourier series of the kind (48). Comparing (38) to (48), we immediately identify 1 a (0) = γ γ ln 2πm ln 2π ln m ln2m ln22π , m 1 − − · − 2 −    l l a (l) = ζ 0, 2(γ + ln 2πm) ln Γ , l = 1,2,3,..., m 1  m ′′ (40)  m − m −  l l(γ + ln 2πm)  Γ bm(l) = π ln + , l = 1,2,3,..., m 1  m m −   Thus, in virtue of (49), for any k = 1,2,3,..., m 1, − m 1 m 1 − r 2πrk 1 2 2 − l ∑ γ cos = γ + γ ln 2πm + ln 2π ln m + ln m + ln 2π ∑ ζ′′ 0, 1 m · m − 1 · 2 − m r=1 l=1 2 d [(ns 1)ζ(s)] = 1 ln2m ln m ln 2π ds2 − s=0 − 2 − ·

m 1 | {z } − l k k +2(γ + ln 2πm) ∑ ln Γ m(γ + ln 2πm) ln Γ + ln Γ 1 m − m − m l=1 " # 1 1 πk (m 1) ln2π ln m ln π ln sin m 2 − − 2 − | {z } m k | {z } k kπ + ζ′′ 0, + ζ′′ 0, 1 = γ + m(γ + ln 2πm) ln 2 sin 2 m − m − 1 · m m k k + ζ′′ 0, + ζ′′ 0, 1 2 m − m (41)

16 where we respectively used the multiplication theorem for the Hurwitz ζ–function, see e.g. [10, exercise no 64], the Gauss’ multiplication theorem and the reﬂection formula for the logarithm of the Γ–function. Analogously, by (50), we deduce

m 1 − r 2πrk πm k k γ + ln 2πm ∑ γ sin = ln Γ ln Γ 1 + k (m k) 1 m · m 2 m − − m m − − r=1 ( ) (42) π πm πk k = (γ + ln 2πm)(2k m) ln π ln sin + mπ ln Γ 2 − − 2 − m m which holds for k = 1,2,3,..., m 1. QED. − Corollary 7. The Parseval’s theorem for the ﬁrst generalized Stieltjes constant at rational argument has the following form m 1 m 1 − 2 r 2 m − l l l ∑ γ = (m 1)γ mγ (2γ + ln m) ln m + Cm + ∑ ζ′′ 0, ζ′′ 0, + ζ′′ 0, 1 1 m − 1 − 1 2 m · m − m r=1 l=1

m 1 m 1 − l l l − l πl m(γ + ln 2πm) ∑ ln Γ ζ′′ 0, + ζ′′ 0, 1 + m(γ + ln 2πm) ∑ ζ′′ 0, ln sin − m · m − m m · m l=1 l=1

m 1 l m 1 l m m 1 l πl +mπ2 ∑− ln2Γ + 2π2(γ + ln 2πm) ∑− l ln Γ 4(γ + ln 2πm)2 π2 ∑− ln Γ ln sin m · m − 2 − m · m l=1 l=1 l=1 (43) which can be also written (simpliﬁed) as m 1 − 2 r 2 m m+1 3 2 ∑ γ = (m 1)γ mγ (2γ + ln m) ln m + Cm + 1 ( 1) ln 2 + ln π ln 2 1 m − 1 − 1 2 − − 2 · r=1 ( 3 π2 ln2π ln 2 π2 ln22 +(γ + ln 4πm) ln π + (3ln2 + 2 ln π)(γ + ln 2πm) ln π + (γ + ln 2πm) ln π × 2 4 − 2 2 ) m m π2 + ln π ln m (γ + ln 2πm) ln(4π2m) 2(γ + ln 2πm)2 (1 m) ln 2 + ln m ln π 2 · · − 2 − 2 · − 1 1 2 (m 1) 2 (m 1) 2 m 2 2 ⌊ − ⌋ 2 πl m⌊ − ⌋ l l + 4(γ + ln 2πm) π ∑ ln sin + ∑ ζ′′ 0, + ζ′′ 0, 1 2 − · m 2 · m − m l=1 l=1 1 1 2 (m 1) 2 (m 1) ⌊ − ⌋ l l πl 2⌊ − ⌋ 2 l 2 l + 2m(γ + ln 2πm) ∑ ζ′′ 0, + ζ′′ 0, 1 ln sin + mπ ∑ ln Γ + ln Γ 1 · m − m · m · m − m l=1 l=1

1 (m 1) 1 (m 1) ⌊2 − ⌋ l l ⌊2 − ⌋ l +2π2(γ + ln 2πm) ∑ l ln Γ ln Γ 1 + 2mπ2(γ + ln 2πm) ∑ ln Γ 1 · · m − − m · − m l=1 l=1 where, for the purpose of brevity, by Cm we designated an elementary function depending on m and containing

17 the Euler’s constant γ

C m(m 1) ln42 m(m 1)(2 ln m + 2γ + 3 ln π) ln32 m(m 2) ln2m ln22 + m ln3m ln 2 m ≡ − − − − − − · · 2m 2(m 1) ln π + γ(m 2) ln m ln22 m(m 1) 3 ln2 π + 4γ ln π + γ2 + 5 π2 + 1 π2 ln22 − − − · − − 12 6m h i m (m 5 ) ln π 3γ ln m2 ln 2 + 2m (1 m) ln2π (m 5 )γ ln π ln m ln 2 − − 2 − · − − − 2 · + 1 (6π2 + 24γ2)m + 4π2(1 m2) ln m 4(m 1) 3m ln3π + 6mγ ln2π + γπ2(m + 1) 12 − − − h 13 2 2 2 1 4 1 3 1 2 +(( 4 π + 3γ )m + π ) ln π ln 2 + 4 m ln m + m(γ + 2 ln π) ln m + 12 6m ln π i h 2 2 2 2 2 2 1 2 +18γm ln π + π m +(12γ + 3π )m + 2π ln m + 12 12mγ ln π i h + (12γ2 + 9π2)m + 4π2(1 m2) ln π + 2π2(2 + m2)γ ln m − i 1 (m 1) 2π2(4m + 1) ln2π + 4γπ2(m + 1) ln π π2γ2(m 2) − 12 − − − and where m is natural greater than 1.

Proof. Inserting Fourier series coefﬁcients (40) into (51) and proceeding analogously to (57)–(58), yields, after several pages of careful calculations and simpliﬁcations, the above result. The unique formula that should be used in addition to those employed in derivations (57)–(58) is

m 1 πl m m 1 πl m m 1 πl m[(1 m) ln 2 + ln m] ∑− l ln sin = ∑− ln sin = ln ∏− sin = − l=1 · m 2 l=1 m 2 l=1 m 2

Also, the fact that the reﬂected sum ζ (0, l/m) + ζ (0,1 l/m), as well as the function ln sin(πl/m), ′′ ′′ − are both invariant with respect to a change of summation’s index l m l greatly helps when sim- → − plifying formula (43). Note that in the second variant of the Parseval’s theorem (that which contains truncated sums), the term in big curly brackets vanishes when m is odd; this is because when m is even, the number of terms in each sum is odd, and the term corresponding to the factor l = m/2 does not have its pair.

Remark From the above formulæ, we see that the sum of ζ (0, p) with its reflected version ζ (0,1 ′′ ′′ − p), at positive rational p less than 1, plays the fundamental role in the evaluation of the first generalized Stieltjes constant at rational argument. In other words, the transcendence of the latter is mainly defined by the expression ζ (0, p) + ζ (0,1 p) at rational p.16 We do not know which is the transcen- ′′ ′′ − dence of such a sum, but it is not unreasonable to expect that it is lower than that of solely ζ′′(0, p). Furthermore, in our previous work [10], we demonstrated that this sum has several comparatively simple integral and series representations; below, we briefly present some of them. In exercises no 20– 21, we dealt with integral Φ(ϕ), which we, unfortunately, could not reduce to elementary functions

16By neglecting, in such a context, the transcendence of the logarithm of the Γ–function.

18 (despite of its simple and naive appearance). Written in terms of this integral, the above sum reads17

ζ′′(0, p) + ζ′′(0,1 p) =π ctg2πp 2 ln Γ(p) + ln sin πp +(2p 1) ln 2π ln π − · − − n ∞ o e x ln x − dx ˆ ch x cos2πp − (44) 0 2ln2π ln 2 sin πp +  − ·  ∞   ln ln x 2 2 dx  ˆ x(x 2x cos2πp + 1)  1 −   Other representations for this sum involve one of the most basic logarithmico–trigonometric series (but never evaluated in terms of elementary functions, except several particular cases)

∞ cos2πpn ln n ζ′′(0, p) + ζ′′(0,1 p) = 2(γ + ln 2π) ln 2 sin πp + 2 ∑ · − − n=1 n as well as antiderivatives of the ﬁrst generalized Stieltjes constant Γ1(p)

ζ′′(0, p) + ζ′′(0,1 p) = (3ln2 + 2 ln π) ln 2 4Γ (1/2) + 2Γ (p) + 2Γ (1 p) − − − 1 1 1 − see [10, exercise no 22]. By the way, the latter formula, inserted into (23), gives an equation which is in some way analogous the Malmsten’s representation for the Digamma function (52) [in the sense that for rational arguments it provides a connection between the function and its derivative]. Other representations for ζ (0, p) + ζ (0,1 p) may also invlove integrals ′′ ′′ − ∞ ∞ ∞ ln x2 + p2 arctg x/p ln2(x + ip) ln2(x ip) ln2(x ip) · dx , − − dx and p.v. ± dx ˆ 2πx ˆ 2πx ˆ 2πx e 1 e 1 e 1 0 − 0 − 0 − which may be easily deduced from the Hermite representation for the Hurwitz ζ–function, see e.g. [8, vol. I, p. 26, Eq. 1.10(7)]. Besides, attempts to obtain Jensen’s like formualæ, see [38], [44, p. 103], could probably bring some new ideas. In fact, the last integral in (44), after a change of variable x 1/x, → may be also written as

∞ 1 ln ln x x ln ln 1 dx = · x dx (45) ˆ x(x2 2x cos2πp + 1) ˆ x2 2x cos2πp + 1 1 − 0 − If we look attentively at the latter integral, one may notice that it is the derivative with respect to s, when s 1, of → 1 s 1 1 x ln − · x dx ˆ x2 2x cos2πp + 1 0 − The latter integral appears in old Malmsten’s works [45, pp. 20–25] and [46, p. 12]. In particular, from

17Put in [10, Eq. 49] ϕ = π(2p 1). −

19 equations (31) [45, p. 21] and (29) [46, p. 12] it follows that

1 ∞ s 1 1 x ln − Γ(s) sh π(2p 1)x cos(s arctg x) x − 2 · dx = dx ˆ x 2x cos2πp + 1 − sin 2πp ˆ sh πx · (x2 + 1)s/2 0 − 0 ∞ Γ(s) sh π(2p 1)x (1 + ix)s + (1 ix)s = − − dx (46) − 2 sin 2πp ˆ sh πx · (x2 + 1)s 0 +∞ Γ(s) sh π(2p 1)x dx = − , Re s > 0. − 2 sin 2πp ˆ sh πx · (1 ix)s ∞ ± − Integrals in the right–hand side are very similar to Jensen’s formulæ for ζ(s) derived between 1893 and 1895 in [37] and [38]. Taking into account that these references are hard to ﬁnd and that the same formulæ were later reprinted with misprints18, we ﬁnd it useful to reproduce them here as well

π/2 ∞ 1 1 (cos θ)s 2 sin sθ 1 1 sin(s arctg x) dx ζ(s) = + + 2 − dθ = + + 2 s 1 2 ˆ e2π tg θ 1 s 1 2 ˆ (e2πx 1)(x2 + 1)s/2 − 0 − − 0 − ∞ ∞ s 1 s s s 1 2 − s 1 dx (1 + ix) (1 ix) 2 − s sin(s arctg x) dx ζ(s) = + i 2 − πx − s− = 2 (47) s 1 ˆ e + 1 · (x2 + 1) s 1 − ˆ (eπx + 1)(x2 + 1)s/2 − 0 − 0

+∞ ∞ π 1 dx π 2s 2 cos (s 1) arctg x ζ(s) = = − − dx ˆ 2 1 s 1 ˆ (s 1)/2 2 1 2(s 1) ch πx · + ix − s 1 (x2 + 1) − ch πx − ∞ 2 − 0 2 − s C, s = 1, where final simplifications were later done by Lindelöf, [44, p. 103], who also gave ∈ 6 details of their derivation.19 Application of contour integration methods to integrals (46) seems quite attractive as well (especially if p is rational), but the branch point at i is really annoying. ± Finally, note that all above representations remain valid everywhere in the strip 0 < Re p < 1, so it is not impossible that for rational p they could be further simplified or reduced to more convenient forms.

A. Some results from the theory of ﬁnite Fourier series

Finite Fourier series are well-known and widely used in discrete mathematics, numerical analysis, engineering sciences (especially in signal and image processing) and in a lot of related disciplines. Unlike usual Fourier series, which are essentially variants or particular cases of a same formula, ﬁnite Fourier series may take quite different forms and expressions. For instance, in engineering sciences, one usually deals with the following 2m–points Fourier series

m 1 am(0) − πrl πrl r am(m) fm(r) = + ∑ am(l) cos + bm(l) sin + ( 1) , r = 0,1,2,...,2m 1, m Æ , 2 m m − 2 − ∈ l=1

18 2πt t πt t In the well-known monograph [8, vol. I], in formula (13) on p. 33, “(e + 1)− ” should be replaced by “(e + 1)− ”. 19Jensen, in [38], did not provide proofs for these formulæ; he only stated that he had “found them in his notes”, and added that they can be easily derived by the Cauchy’s residue theorem. By the way, the ﬁrst of these three formulæ was also obtained by Franel [25], [38], [37].

20 With the help of orthogonality relations, one may determine the coefﬁcients in this expansion:

2m 1 1 − πrk am(k) = ∑ fm(r) cos , k = 0,1,2,..., m m m  r=1   2m 1  1 − πrk bm(k) = ∑ fm(r) sin , k = 1,2,3,..., m 1 m m −  r=1   as well as derive the Parseval’s theorem

1 2m 1 a2 (0) m 1 a2 (m) ∑− f 2 (r) = m + ∑− a2 (l)+ b2 (l) + m , m m 2 m m 2 r=1 l=1 see for more details [30, Chapter 6]. In contrast, in our researches, we encounter the following (m 1)–points finite Fourier series − m 1 − 2πrl 2πrl fm(r) = am(0)+ ∑ am(l) cos + bm(l) sin , r = 1,2,3,..., m 1, m Æ , (48) m m − ∈ l=1 for which inversion formulæ and Parseval’s theorem are quite different. Let, first, derive the inversion for- mulæ for the coefficients of this series. Multiplying both sides by cos(2πrk/m), where k = 1,2,3,..., m 1, and − summing over r [1, m 1], gives ∈ − m 1 m 1 m 1 m 1 − 2πrk − − 2πrl − 2πrl 2πrk ∑ fm(r) cos = ∑ am(0)+ ∑ am(l) cos + ∑ bm(l) sin cos r=1 m r=1 " l=1 m l=1 m # m

m 1 m 1 m 1 m 1 m 1 − 2πrk − − 2πrl 2πrk − − 2πrl 2πrk = am(0) ∑ cos + ∑ am(l) ∑ cos cos + ∑ bm(l) ∑ sin cos (49) r=1 m l=1 r=1 m · m l=1 r=1 m · m

1 1 m(δl,k+δl,m k) 1 0 − 2 − − m 1 | {z− } m | {z } | {z } = am(0) ∑ am(l)+ am(k)+ am(m k) − − 2 − l=1 n o Similarly, multiplying both sides of (48) by sin(2πrk/m), where k = 1,2,3,..., m 1, and summing over r − ∈ [1, m 1], yields − m 1 m 1 m 1 m 1 − 2πrk − − 2πrl − 2πrl 2πrk ∑ fm(r) sin = ∑ am(0)+ ∑ am(l) cos + ∑ bm(l) sin sin r=1 m r=1 " l=1 m l=1 m # m

m 1 m 1 m 1 m 1 m 1 − 2πrk − − 2πrl 2πrk − − 2πrl 2πrk = am(0) ∑ sin + ∑ am(l) ∑ cos sin + ∑ bm(l) ∑ sin sin (50) r=1 m l=1 r=1 m · m l=1 r=1 m · m

0 0 1 2 m(δl,k δl,m k) m − − = b |(k) {zb (m} k) | {z } | {z } 2 m − m − n o

21 Finally, Parseval’s equality for the ﬁnite series (48) reads: m 1 m 1 m 1 m 1 2 m 1 m 1 m 1 − 2 − − 2πrl − 2πrl − 2 − − 2πrl ∑ fm(r)= ∑ am(0)+ ∑ am(l) cos + ∑ bm(l) sin = ∑ am(0)+ 2am(0) ∑ am(l) ∑ cos r=1 r=1 " l=1 m l=1 m # r=1 l=1 r=1 m

1 m 1 m 1 m 1 m 1 m 1 − − − 2πrl − − − 2πrl 2πrn +2am(0) ∑ bm(l) ∑ sin +2 ∑ ∑ am(l)bm(n) ∑ cos sin | {z } l=1 r=1 m l=1 n=1 r=1 m · m

0 0 m 1 m 1 m 1 m 1 m 1 m 1 − − − 2πrl 2πrn − − − 2πrl 2πrn + ∑ ∑ am(l)am(|n) ∑{z cos } cos + ∑ ∑| bm(l)bm({zn) ∑ sin } sin l=1 n=1 r=1 m · m l=1 n=1 r=1 m · m

1 1 2 m(δl,n+δl,m n) 1 2 m(δl,n δl,m n) − − 2 − − | m 1 {z m 1 } m 1 | {z } 2 − − m − 2 2 = (m 1)am(0) 2am(0) ∑ am(l) ∑ am(l) + ∑ am(l)+ am(l)am(m l)+ bm(l) bm(l)bm(m l) − − − " # 2 − − − l=1 l=1 l=1 h i (51) Note that in above formulæ the performance of the sums’ permutation is permitted since all series are ﬁnite.

B. Malmsten’s ﬁnite Fourier series representation for the Ψ–function and some related summa- tions

In 1842 Carl Malmsten found an interesting representation for the Ψ–function of a rational argument:

r π πr m 1 2πrl l Ψ = γ ln 2πm ctg 2 ∑− cos ln Γ , r = 1,2,..., m 1. (52) m − − − 2 m − m · m − l=1 where m is natural, see [45, p. 57, Eq. (70)], [10, Eq. (23)]. Malmsten did not notice that this formula may be further simpliﬁed and reduced to the Gauss’ Digamma theorem. However, on the other hand, if we recall that

m 1 2πrl m rπ ∑− l sin = ctg , r = 1,2,..., m 1. l=1 · m − 2 m − we may rewrite (52) as the complete ﬁnite–length Fourier series for the Ψ–function at rational argument:

r π m 1 2πrl m 1 2πrl l Ψ = γ ln 2πm + ∑− sin l 2 ∑− cos ln Γ , r = 1,2,..., m 1. (53) m − − m m · − m · m − l=1 l=1 This expression may be in some cases more suitable and more convenient to use than the usual Gauss’ Digamma theorem. For instance, from (53), thanks to semi–orthogonality properties of circular functions, we straightforwardly deduce following summation formulæ

m 1 r 2πrk kπ ∑− Ψ cos = m ln 2sin + γ , k = 1,2,..., m 1 m · m m −  r=1  m 1  − r 2πrk π  ∑ Ψ sin = (2k m) , k = 1,2,..., m 1  m · m 2 − −  r=1 1 (m 1) m 1 r π2(m2 3m + 2) ⌊2 − ⌋ πl  − Ψ2 2 2 2  ∑ = (m 1)γ + m(2γ + ln 4m) ln m m(m 1) ln 2 + − + 2m ∑ ln sin  m − − − 12 m  r=1 l=1  (54) 

22 Indeed, inserting expressions for coefﬁcients a (0) = γ ln 2πm, a (l) = 2 ln Γ(l/m) and b (l) = πl/m m − − m − m into (49), we have for the ﬁrst sum:

m 1 r 2πrk k k m 1 l kπ ∑− Ψ cos = γ + ln 2πm m ln Γ + ln Γ 1 + 2 ∑− ln Γ = γ + m ln 2sin m · m − m − m m m r=1 " # l=1

ln π lnsin πk 1 (m 1) ln 2π 1 ln m − m 2 − − 2 | {z } | {z } (55) where the final simplification is performed with the help of the reflection formula and the Gauss’ multiplication theorem for the logarithm of the Γ–function. Analogously, using (50) yields for the second sum:

m 1 r 2πrk m πk π(m k) π ∑− Ψ sin = − = (2k m) (56) m · m 2 m − m 2 − r=1 Finally, by (51) we derive the Parseval’s theorem for the Ψ–function of a discrete argument takes the following form 2 m 1 r m 1 l m 1 l ∑− Ψ2 = (m 1)(γ + ln 2πm)2 4(γ + ln 2πm) ∑− ln Γ 4 ∑− ln Γ m − − m − m r=1 l=1 " l=1 # 1 (m 1) ln 2π 1 ln m 2 − − 2 m 1 l l | l {z π2} m 1 π2 m 1 (57) +2m ∑− ln Γ ln Γ + ln Γ 1 + ∑− l2 ∑− l m · m − m m · − 2 · l=1 " # l=1 l=1 1 (m 1) π2(m2 3m + 2) ⌊2 − ⌋ πl =(m 1)γ2 + m(2γ + ln 4m) ln m m(m 1) ln22 + − + 2m ∑ ln2sin − − − 12 l=1 m where the ﬁrst sum from the second line, thanks to the symmetry of lnsin πl/m about l = m/2 and to the fact that ln sin πl/m = 0 for l = m/2, could be simpliﬁed as follows m 1 l l l m 1 l πl ln π ∑− ln Γ ln Γ + ln Γ 1 = ∑− ln Γ ln π lnsin = (m 1) ln 2π ln m m · " m − m # m · " − m # 2 − − l=1 l=1 h i 1 (m 1) m 1 l πl ln π ⌊2 − ⌋ πl πl ln π ∑− ln Γ lnsin = (m 1) ln 2π ln m ∑ ln π lnsin lnsin = (m 1) ln 2π ln m − m · m 2 − − − " − m # m 2 − − l=1 h i l=1 h i 1 (m 1) 1 (m 1) 1 (m 1) ⌊2 − ⌋ πl ⌊2 − ⌋ πl ln π ⌊2 − ⌋ πl ln π ∑ lnsin + ∑ ln2sin = (m 1) ln 4π 2 ln m + ∑ ln2sin − l=1 m l=1 m 2 − − l=1 m h i (58) because 1 (m 1) 1 (m 1) ⌊2 − ⌋ πl ⌊2 − ⌋ πl 1 m 1 ∑ lnsin = ln ∏ sin = − ln 2 + ln m l=1 m l=1 m 2 2 and where m 1 m 1 − m(m 1)(2m 1) − m(m 1) ∑ l2 = − − and ∑ l = − l=1 6 l=1 2 respectively, which completes the evaluation of the third formula in (54). In like manner, we may also derive similar summation formulæ for the Hurwitz ζ–function. Rewriting Hurwitz’ functional equation (21) in the form analogous to (48)

m 1 m 1 r a 1 2Γ(1 a) πa − 2πrl l πa − 2πrl l ζ a, = m − ζ(a)+ −1 a sin ∑ cos ζ 1 a, + cos ∑ sin ζ 1 a, m (2πm) − " 2 m · − m 2 m · − m # l=1 l=1

23 yields

m 1 − r 2πrk mΓ(1 a) πa k k ∑ ζ a, cos = −1 a sin ζ 1 a, + ζ 1 a, 1 ζ(a) m · m (2πm) − 2 · − m − − m −  r=1   m 1 Γ  − r 2πrk m (1 a) πa k k  ∑ ζ a, sin = −1 a cos ζ 1 a, ζ 1 a, 1  m · m (2πm) − 2 · − m − − − m  r=1 m 1 Γ2 m 1  − 2 r 2a 1 2 2m (1 a) − l l l  ∑ ζ a, = m − 1 ζ (a)+ 2−2a ∑ ζ 1 a, cos πa ζ 1 a, 1 ζ 1 a,  m − (2πm) − − m − · − − m · − m  r=1 l=1  (59)  which hold for any r = 1,2,3,..., m 1 and k = 1,2,3,..., m 1, where m is positive integer. − −

C. On the complex differentiability of the integral Ja(r/m) in a neighbourhood of a = 1

In order to show that Ja(r/m) is regular in a neighbourhood of a = 1, it sufﬁces to ascertain that in the region

a 1 + C of uniform convergence, the complex–valued function x , x Ê , a , is differentiable with respect to a. − ∈ ∈ The necessary and sufﬁcient condition that a complex–valued function be differentiable is that its both real and imaginary parts be differentiable, and that the Cauchy–Riemann equations20 hold. Denoting for brevity α Re a ≡ and β Im a, we separate the real part from the imaginary one of xa 1 as follows ≡ − a 1 α 1 iβ α 1 iβ ln x α 1 α 1 x − = x − x = x − e = x − cos(β ln x) + i x − sin(β ln x)

u(α,β) v(α,β) Veriﬁcation of the ﬁrst equation yields | {z } | {z }

∂u α 1 = x − ln x cos(β ln x) ∂α · ∂u ∂v  = ∂v α 1  ⇒ ∂α ∂β = x − cos(β ln x) ln x  ∂β  while the second one gives 

∂u α 1 = x − sin(β ln x) ln x ∂β − ∂u ∂v  = ∂v α 1  ⇒ ∂β − ∂α = x − ln x sin(β ln x)  ∂α · a 1 +  Thus, x , x Ê , is an entire function on the complex a–plane, which, together with the uniform convergence, − ∈ plenty guarantees the existence of the Taylor series for Ja(r/m) near a = 1. For a deeper study of regular (or holomorphic) and analytic functions, please refer to these classical complex analysis monographs: [23], [47], [54], [12], [33], [5].

References

[1] Correspondance d’Hermite et de Stieltjes. Vol. 1 and 2, Gauthier-Villars, Paris, 1905.

[2] Collected papers of Srinivasa Ramanujan, Cambridge, 1927.

[3] M. Abramowitz and I. A. Stegun, Handbook of mathematical functions with formula, graphs and mathematical tables [Applied mathematics series no 55], US Department of Commerce, National Bureau of Standards, 1961.

20For historical reasons, some author prefer to call these equations D’Alambert–Euler equations [47].

24 [4] V. Adamchik, A class of logarithmic integrals, Proceedings of the 1997 International Symposium on Symbolic and Algebraic Computation, pp. 1–8 (1997).

[5] L. Ahlfors, Complex Analysis (third edition), McGraw–Hill Science, USA, 1979.

[6] O. R. Ainsworth and L. W. Howell, An integral representation of the generalized Euler–Mascheroni constants, NASA Technical paper 2456, pp. 1–11 (1985).

[7] T. M. Apostol, Introduction to analytic number theory, Springer–Verlag New Yorc Inc., 1976.

[8] H. Bateman and A. Erd´elyi, Higher Transcendental Functions [in 3 volumes], Mc Graw–Hill Book Company, 1955.

[9] B. C. Berndt, On the Hurwitz Zeta–function, Rocky Mountain Journal of Mathematics, vol. 2, no. 1, pp. 151– 157 (1972).

[10] I. V. Blagouchine, Rediscovery of Malmsten’s integrals, their evaluation by contour integration methods and some related results, The Ramanujan Journal, vol. xx, no. xx, pp. 1–90 (2013, in press, DOI 10.1007/s11139- 013-9528-5).

[11] W. E. Briggs, Some constants associated with the Riemann Zeta–function, The Michigan Mathematical Jour- nal, vol. 3, issue 2, pp. 117–121 (1955/56).

[12] C. Carath´eodory, Theory of Functions of a Complex Variable [in 2 vols.], Chelsea Publishing Company, New York, USA, 1954.

[13] M. W. Coffey, New summation relations for the Stieltjes constants, Proceedings of the Royal Society A, vol. 462, pp. 2563–2573 (2006).

[14] M. W. Coffey, Series representations for the Stieltjes constants, arXiv, http://arxiv.org/abs/0905.1111v2 (2009).

[15] M. W. Coffey, The Stieltjes constants, their relation to the ηj coefﬁcients, and representation of the hurwitz zeta function, arXiv, http://arxiv.org/abs/0706.0343v2 (2009).

[16] M. W. Coffey, On representations and differences of Stieltjes coefﬁcients, and other relations, Rocky Mountain Journal of Mathematics, vol. 41, no. 6, pp. 1815–1846 (2011).

[17] D. F. Connon, The difference between two Stieltjes constants, arXiv, http://arxiv.org/abs/0906.0277 (2009).

[18] D. F. Connon, New proofs of the duplication and multiplication formulae for the gamma and the Barnes double gamma functions, arXiv, http://arxiv.org/abs/0903.4539 (2009).

[19] D. F. Connon, Some applications of the Stieltjes constants, arXiv, http://arxiv.org/abs/0901.2083 (2009).

[20] P. J. Davis, Leonhard Euler’s integral: A historical proﬁle of the Gamma function, American Mathematical Monthly, vol. 66, pp. 849–869 (1959).

[21] H. B. Dwigth, Tables of Integrals and Other Mathematical Data (3rd edition), The Macmillan Company, 1957.

[22] L. Euler, Remarques sur un beau rapport entre les séries des puissances tant directes que réciproques, His- toire de l’Académie Royale des Sciences et Belles–Lettres, année MDCCLXI, Tome 17, pp. 83–106, A Berlin, chez Haude et Spener, Libraires de la Cour et de l’Académie Royale, 1768 [read in 1749].

[23] M. A. Evgrafov, Analytic functions, W. B. Saunders Company, Philadelphia, USA, 1966.

25 [24] M. A. Evgrafov, Y. V. Sidorov, M. V. Fedoriuk, M. I. Shabunin and K. A. Bezhanov, A Collection of Problems in the Theory of Analytic Functions [in russian], Nauka, Moscow, USSR, 1969.

[25] J. Franel, Note no 245, L’Interm´ediaire des math´ematiciens, tome II, pp. 153–154 (1895).

αβ α(α + 1)β(β + 1) [26] C. F. Gauss, Disquisitiones generales circa seriem infinitam 1 + x + xx + 1 γ 1 2 γ(γ + 1) α(α + 1)(α + 2)β(β + 1)(β + 2) · · · x3 + etc, Commentationes Societatis Regiae Scientiarum Gottingensis recen- 1 2 3 γ(γ + 1)(γ + 2) tiores,· Classis· · Mathematicæ, vol. II, pp. 3–46 [republished later in “Carl Friedrich Gauss Werke”, vol. 3, pp. 265–327, Königliche Gesellschaft der Wissenschaften, Göttingen, 1866] (1813).

[27] G. W. L. Glaisher, On Dr. Vacca’s series for γ, The Quarterly journal of pure and applied mathematics, vol. 41, pp. 365–368 (1910).

[28] J. P. Gram, Note sur le calcul de la fonction ζ(s) de Riemann, Oversigt. K. Danske Vidensk. (Selskab Forhan- dlingar), pp. 305–308 (1895).

[29] N. M. Gunther and R. O. Kuzmin, A Collection of Problems on Higher Mathematics. Vol. 3 (4th edition) [in russian], Gosudarstvennoe izdatel’stvo tehniko–teoreticheskoj literatury, Leningrad, USSR, 1951.

[30] R. W. Hamming, Numerical methods for scientists and engineers, McGraw–Hill Book Company, 1962.

[31] G. H. Hardy, Note on Dr. Vacca’s seriesfor γ, The Quarterly journal of pure and applied mathematics, vol. 43, pp. 215–216 (reprinted also in vol. 4, pp. 475–476, of the Hardy’s “Collected papers”) (1912).

[32] G. H. Hardy, Divergent series, Oxford at the Clarendan press, 1949.

[33] P. Henrici, Applied and Computational Complex Analysis [in 3 vols.], Wiley, USA, 1974, 1977, 1986.

D 1 [34] A. Hurwitz, Einige Eigenschaften der Dirichlet’schen Functionen F(s)= ∑ ( ) s , die bei der Bestimmung n · n der Classenanzahlen bin¨arer quadratischer Formen auftreten, Zeitschrift f ¨ur Mathematik und Physik, vol. XXVII, pp. 86–101 (1882).

[35] M. I. Israilov, On the Laurent decomposition of Riemann’s zeta function [in Russian], Trudy Mat. Inst. Akad. Nauk. SSSR, vol. 158, pp. 98–103 (1981).

[36] J. L. W. V. Jensen, Sur la fonction ζ(s) de Riemann, Comptes-rendus hebdomadaires des s´eances de l’Acad´emie des sciences, tome 104, pp. 1156–1159 (1887).

[37] J. L. W. V. Jensen, Opgaver til løsning no 34, Nyt Tidsskrift for Matematik, Afdeling B, vol. 4, p. 54 (1893).

[38] J. L. W. V. Jensen, Note no 245. Deuxième réponse. Remarques relatives aux réponses du MM. Franel et Kluyver, L’Intermédiaire des mathématiciens, tome II, pp. 346–347 (1895).

[39] J. C. Kluyver, On certain series of Mr. Hardy, The Quarterly journal of pure and applied mathematics, vol. 50, pp. 185–192 (1927).

[40] R. Kreminski, Newton–cotes integration for approximating Stieltjes (generalized Euler) constants, Math. Comp., vol. 72, pp. 1379–1397 (2003).

[41] E. Lammel, Ein Beweis, dass die Rimannsche Zeta-funktion ζ(s) in s 1 6 1 keine Nullstelle besitzt, Univ. | − | Nac. Tacuman Rev. Ser. A, vol. 16, pp. 209–217 (1966).

[42] A. F. Lavrik, On the main term of the divisor’s problem and the power series of the Riemann’s zeta function in a neighbourhood of its pole [in Russian], Trudy Mat. Inst. Akad. Nauk. SSSR, vol. 142, pp. 165–173 (1976).

26 [43] J. J. Y. Liang and J. Todd, The Stieltjes constants, Journal of Research of the National Bureau of Standards— Mathematical Sciences, vol. 76B, nos. 3–4, pp. 161–178 (1972).

[44] E. Lindelöf, Le calcul des résidus et ses applications àla théorie des fonctions, Gauthier–Villars, Imprimeur Libraire du Bureau des Longitudes, de l’Ecole´ Polytechnique, Quai des Grands–Augustins, 55, Paris, 1905.

[45] C. J. Malmsten, T. A. Almgren, G. Camitz, D. Danelius, D. H. Moder, E. Selander, J. M. A. Grenander, S. Themp- tander, L. M. Trozelli, Alskade¨ Föräldrar, G. E. Ossbahr, D. H. Föräldrar, C. O. Ossbahr, C. A. Lindhagen, D. huldaste Moder, Alskade¨ Syskon, O. V. Lemke, C. Fries, L. Laurenius, E. Leijer, G. Gyllenberg, M. V. Morfader and A. Lin- deroth, Specimen Analyticum, theoremata quædam nova de integralibus definitis, summatione serierum earumque in alias series transformatione exhibens (Eng. trans.: “Some new theorems about the definite integral, summation of the series and their transformation into other series”) [Dissertation, in 8 parts], Upsaliæ, excudebant Regiæ academiæ typographi. Uppsala, Sweden, April–June 1842.

[46] C. J. Malmstén, De integralibus quibusdam definitis seriebusque infinitis (Eng. trans.: “On some definite integrals and series”), Journal f ür die reine und angewandte Mathematik, vol. 38, pp. 1–39 (1849 [work dated at May 1, 1846]).

[47] A. I. Markushevich, Theory of Functions of a Complex Variable (second edition, in 3 vol.), AMS Chelsea Publishing, American Mathematical Society, 2005.

[48] J. Miller and V. S. Adamchik, Derivatives of the Hurwitz Zeta function for rational arguments, Journal of Computational and Applied Mathematics, vol. 100, pp. 201–206 (1998).

[49] D. Mitrovi´c, The signs of some constants associated with the Riemann zeta function, Michigan Math. J., vol. 9, pp. 395–397 (1962).

[50] Z. Nan-You and K. S. Williams, Some results on the generalized Stieltjes constant, Analysis, vol. 14, pp. 147– 162 (1994).

[51] B. Riemann, Ueber die Anzahl der Primzahlen unter einer gegebenen Grosse, Monatsberichte der K¨oniglich Preußischen Akademie der Wissenschaften zu Berlin, pp. 136–144 (1859).

[52] O. Schlömilch, Uebungsaufgaben f ür Schüler. Lehrsatz, Grunert Archiv der Mathematik und Physik, vol. XII, issue IV, part XXXV, p. 415 (1849).

[53] O. Schl¨omilch, Ueber eine Eigenschaft gewisser Reihen, Zeitschrift f¨ur Mathematik und Physik, vol. III, pp. 130–132 (1858).

[54] M. R. Spiegel, Theory and problems of complex variables with an introduction to conformal mapping and its application, McGrow–Hill, 1968.

[55] A. M. Le Gendre, Exercices de calcul intégral sur divers ordres de transcendantes et sur les quadratures. Tomes I–III, Mme Ve Courcier, Imprimeur–Libraire pour les Mathématiques, rue du Jardinet, no 12, quartier Saint–André–des–Arc, Paris, 1811–1817.

[56] E. C. Titchmarsh, The theory of the Riemann Zeta–function (2nd edition), Clarendon Press, Oxford, 1986.

[57] G. Vacca, A new series for the Eulerian constant, The Quarterly journal of pure and applied mathematics, vol. 41, pp. 363–364 (1910).

[58] L. I. Volkovyskii, G. L. Lunts and I. G. Aramanovich, A Collection of Problems on Complex Analysis, Pergamon Press, United Kingdom, 1965.

27 [59] E. Whittaker and G. N. Watson, A course of modern analysis. An introduction to the general theory of inﬁnite processes and of analytic functions, with an account of the principal transcendental functions (third edition), Cambridge at the University Press, Great Britain, 1920.

[60] J. R. Wilton, A note on the coefﬁcients in the expansion ζ(s, x) in powers of s 1, The Quarterly journal of − pure and applied mathematics, vol. 50, pp. 329–332 (1927).