1: % VERSION FINALE DU 05/12/02

2: 

3: \documentclass[12pt]{article}

4: \usepackage{amsfonts,mathrsfs}

5: \textheight 23cm

6: \textwidth 16cm

7: \hoffset=-2cm

8: \voffset=-2cm

9: 

10: \def\d{{\rm d}}

11: \def\mi{{\rm i}}

12: \def\eps{\varepsilon}

13: \def\g{\mathop{\gamma}\nolimits}

14: \def\G{\mathop{\Gamma}\nolimits}

15: \def\D{\mathop{\Delta}\nolimits}

16: \def\Re{\mathop{\rm Re\,}\nolimits}

17: \def\Im{\mathop{\rm Im\,}\nolimits}

18: \def\e{\mathop{\rm e}\nolimits}

19: \def\Z{\mathop{\mathcal Z}\nolimits}

20: \def\Res{\mathop{\rm Res}\nolimits}

21: \def\hf{{\textstyle{1 \over 2}}}

22: \def\qt{{\textstyle{1 \over 4}}}

23: \def\fq{{\textstyle{5 \over 4}}}

24: 

25: \def\defi{\stackrel{\rm def}{=}}

26: \def\si{\!\!\! &}

27: \def\sf{& \!\!\!}

28: 

29: 

30: \title{Zeta functions for the Riemann zeros}

31: \author{{\bf Andr\'e Voros} \\

32: \\

33: CEA, Service de Physique Th\'eorique de Saclay\\

34: CNRS URA 2306\\

35: F-91191 Gif-sur-Yvette CEDEX (France)\\

36: { E-mail : {\tt voros@spht.saclay.cea.fr}}\\

37: \\

38: and\\

39: \\

40: Institut de Math\'ematiques de Jussieu--Chevaleret\\

41: CNRS UMR 7586\\

42: Universit\'e Paris 7\\

43: 2 place Jussieu,

44: F-75251 Paris CEDEX 05 (France)}

45: 

46: \begin{document}

47: \maketitle

48: 

49: {\abstract

50: A family of Zeta functions built as Dirichlet series over the Riemann zeros 

51: are shown to have meromorphic extensions in the whole complex plane, 

52: for which numerous analytical features 

53: (the polar structure, plus countably many special values) 

54: are explicitly displayed.

55: }

56: \bigskip

57: 

58: \centerline{\bf Fonctions Z\^eta pour les z\'eros de Riemann}

59: 

60: {\abstract

61: Certaines fonctions Z\^eta d\'efinies sur les z\'eros de Riemann, 

62: par une famille de s\'eries de Dirichlet, se prolongent \`a tout

63: le plan complexe en des fonctions m\'eromorphes, dont de nombreuses

64: caract\'eristiques peuvent \^etre explicit\'ees (la structure polaire,

65: mais aussi une infinit\'e de valeurs sp\'eciales).

66: }

67: \bigskip

68: 

69: {\bf Keywords :} Riemann zeta function, Riemann zeros, Dirichlet series, 

70: Hadamard factorization, meromorphic functions, Mellin transform.

71: \medskip

72: 

73: {\bf MSC2000 :} 11Mxx, 30B40, 30B50.

74: 

75: \bigskip

76: 

77: \section{Introduction}

78: 

79: This work proposes to investigate certain meromorphic functions 

80: defined by Dirichlet series over the nontrivial zeros $\{\rho\}$ 

81: of the Riemann zeta function $\zeta(s)$,

82: and to thoroughly compile their explicit analytical features.

83: If the Riemann zeros are listed in pairs as usual,

84: \begin{equation}

85: \label{ZER}

86: \{\rho = \hf \pm \mi \tau_k \}_{k=1,2,\ldots}, \quad

87: \{ \Re \, \tau_k \} \mbox{ positive and non-decreasing,}

88: \end{equation}

89: then the Dirichlet series to be mainly studied read as

90: \begin{equation}

91: \label{ZDef}

92: \Z(\sigma,v) \defi

93: \sum_{k=1}^\infty ({\tau_k}^2+v)^{-\sigma}, \qquad \Re \sigma > \hf,

94: \quad v> - {\tau_1}^2, 

95: \end{equation}

96: extended to meromorphic functions of $\sigma \in \mathbb C$, 

97: and parametrized by $v$

98: --- with emphasis on two cases, $v=\qt$ and especially $v=0$.

99: Their analysis will simultaneously yield some results for the variant series

100: \begin{equation}

101: \label{Hz}

102: {\mathfrak Z}(\sigma,a) \defi \sum_{k=1}^\infty ({\tau_k}+a)^{-2\sigma}, 

103: \qquad \Re \sigma > \hf \quad (\mbox{and, e.g., }|\arg a | \le \pi/2).

104: \end{equation}

105: 

106: Those Zeta functions are ``secondary": 

107: arising from the nontrivial zeros of a classic zeta function (here, $\zeta(s)$);

108: and ``generalized": they admit an auxiliary shift parameter

109: just like the Hurwitz zeta function 

110: ($\zeta(s,a) \defi \sum_0^\infty (n+a)^{-s}$).

111: Such ``$\zeta$-Zeta" functions have occasionally appeared in the literature, 

112: but mostly through particular cases or under very specific aspects.

113: On the other hand, their {\sl abundance of general explicit properties\/}

114: seems to have been largely ignored,

115: although it can be revealed by quite {\sl elementary\/} means (as we will do).

116: And with regard to the Riemann zeros, reputed to be highly elusive quantities, 

117: those properties constitute additional explicit information: this is enough 

118: to motivate a more comprehensive treatment (and bibliography) of the subject.

119: The present work just aims to do that, 

120: in a self-contained and very concrete way, as a kind of

121: ``All you ever wanted to know about $\zeta$-Zeta functions\ldots" handbook

122: (without prejudice to the usefulness of any single result by itself).

123: 

124: \medskip

125: 

126: We begin by developing the background and our motivations in greater detail.

127: 

128: First, if a {\sl Selberg zeta function\/} is used in place of $\zeta(s)$ 

129: from the start

130: (assuming the simplest setting of a compact hyperbolic surface $X$ here),

131: then the $\{ {\tau_k}^2+ \qt \}$ become the eigenvalues 

132: of the (positive) Laplacian on $X$, 

133: and the Zeta functions (\ref{ZDef}) turn into 

134: {\sl bona fide\/} spectral (Minakshisundaram--Pleijel) zeta functions,

135: for which numerous explicit results have indeed been displayed

136: (with the help of Selberg trace formulae:

137: cf. \cite{R} for $v \ge \qt$, \cite{S} for $v=\qt$, \cite{CV,Vz} for $v=0$).

138: 

139: Some transposition of those results to the Riemann case can then be expected, 

140: in view of the formal analogies between 

141: the trace formulae for Selberg zeta functions on the one hand,

142: and the Weil ``explicit formula" for $\zeta(s)$ on the other hand \cite{H}.

143: Indeed, a few symmetric functions over the Riemann zeros 

144: that resemble spectral functions have been well described, mainly 

145: $V(t) \defi \sum_\rho \e^{\rho t}$ (\cite{CR,G2}, \cite{JL2} chap.II).

146: Zeta functions like (\ref{ZDef}) have also been considered,

147: but almost solely to establish their meromorphic continuation 

148: to the whole $\sigma$-plane --- apart from the earliest occurrence we found:

149: a mention by Guinand \cite{G1} (see also \cite{CHA}) of the series 

150: $\sum_k \tau_k^{-s}$ ($\equiv \Z(s/2,v=0)$) on one side of a functional relation

151: (eq.(\ref{GFE}) below) arising as an instance 

152: of a generalized Poisson summation formula.

153: Later, Delsarte introduced that function again (as $\phi(s)$ in \cite{Dl})

154: to describe its poles qualitatively,

155: displaying (only) its principal polar part at $s=1$, as $(2\pi)^{-1}/(s-1)^2$;

156: Kurokawa \cite{K} made the same study at $v=\qt$, 

157: not only for $\zeta(s)$ but also for Dedekind zeta functions

158: and Selberg zeta functions for PSL$_2(\mathbb Z)$ [or congruence subgroups]

159: (then, Zeta functions like (\ref{ZDef}) occur 

160: within the {\sl parabolic\/} components);

161: and Matiyasevich \cite{MA} discussed the special values 

162: $\theta_n \equiv 2 \Z(n,\qt) \ (n\in {\mathbb N}^\ast)$.

163: Extensions in the style of the Lerch zeta function have also been studied

164: (\cite{F}, \cite{JL2} chap.VI).

165: 

166: Independently, Deninger \cite{Dn} and Schr\"oter--Soul\'e \cite{SS}

167: considered a {\sl different\/} Hurwitz-type family

168: (we keep their factor $(2\pi)^s$ just to avoid multiple notations),

169: \begin{equation}

170: \label{ZZet}

171: \xi (s,x) \defi (2\pi)^s \sum_\rho (x-\rho)^{-s} \qquad (\Re s >1),

172: \end{equation}

173: mainly to evaluate $\partial_s \xi (s,x)_{s=0}$ (as eq.(\ref{DZet1}) below);

174: earlier, Matsuoka \cite{Ms2}, then Lehmer \cite{L} had focused upon the sums 

175: \begin{equation}

176: \label{RN}

177: {\mathscr Z}_n \defi \sum_\rho \rho^{-n}, \quad n \in {\mathbb N}^\ast

178: \quad [\equiv (2\pi)^{-n} \xi(n,1), \mbox{ for } n \ne 1].

179: \end{equation}

180: Here we easily recover $\xi (s,x)$ from 

181: the {\sl other\/} Zeta function (\ref{Hz}) (but not the reverse), as

182: \begin{equation}

183: \label{Z2X}

184: \xi (s,\hf+y) \equiv (2\pi)^s 

185: \Bigl[ \e^{\mi\pi s/2} {\mathfrak Z} \bigl( \hf s, \mi y \bigr) +

186:      \e^{-\mi\pi s/2} {\mathfrak Z} \bigl( \hf s, -\mi y \bigr) \Bigr] .

187: \end{equation}

188: 

189: \bigskip

190: 

191: {\sl The present work proposes a broader, and unified, description 

192: for all those $\zeta$-Zeta functions, with a wealth of explicit results 

193: comparable to usual spectral zeta functions\/} \cite{Vz}.

194: 

195: Tools for the purpose could also be borrowed from spectral theory 

196: (trace formulae, etc.), but the objects under scrutiny are more singular here

197: (the Zeta functions for the Riemann zeros manifest double poles, 

198: vs simple poles in the Selberg case);

199: this then imposes so many adaptations upon the classic procedures

200: that a self-contained treatment of the Riemann case alone is actually simpler.

201: Even then, we cannot get maximally explicit outputs for all cases at once

202: (e.g., Weil's ``explicit formula" {\sl diverges\/} for $v \le \qt$),

203: and our analysis has to develop gradually.

204: 

205: So, we begin (Sec.2) by setting up a minimal abstract framework,

206: sufficient to handle $\Z(\sigma,v)$

207: (with a permanent distinction between properties in the half-planes

208: $\{ \Re \sigma < 1 \}$ and $\{ \Re \sigma > \hf \}$ respectively).

209: We next obtain a first batch of explicit results for the case $v=\qt$ (Sec.3),

210: then for general values of $v$ (Sec.4).

211: Specializing to the case $v=0$ in Sec.5, 

212: we reach an almost explicit meromorphic continuation formula for $\Z(\sigma,0)$

213: into the half-plane $\{ \Re \sigma < \hf \}$, 

214: which immediately implies many more properties of that function,

215: and is generalizable to $L$-series and other number-theoretical zeta functions.

216: In Sec.6 we exploit the latter results to sharpen the descriptions of 

217: both Hurwitz-type functions $\Z(\sigma,v)$ and ${\mathfrak Z}(\sigma,a)$.

218: Sec.7 provides a summary of the results; 

219: essentially, a Table collects the main formulae for $v=0$ and $\qt$,

220: also referring to the main text like an index.

221: (Which text can in some sense be viewed, and simply used, 

222: as a set of ``notes" for this Table~!)

223: Finally, Appendices A and B treat some subsidiary issues: 

224: a meromorphic continuation method for the Mellin transforms of Sec.2,

225: and certain numerical aspects.

226: 

227: \bigskip

228: 

229: For convenience, we recall the needed elementary results and notations 

230: \cite{AS,B,CHO}:

231: \begin{eqnarray}

232: \label{NOT}

233: B_n \si:\sf \mbox{Bernoulli numbers } (B_0=1,\ B_1=-\hf,

234: \ B_2=\textstyle{1 \over 6},\ldots; 

235: \ B_{2m+1} = 0 \mbox{ for }m=1,2,\ldots) \nonumber \\

236: (B_n(x) \si:\sf \mbox{Bernoulli polynomials}) \nonumber \\

237: \g \si=\sf \mbox{Euler's constant; \ Stieltjes constants : } 

238: \g_n \defi \!\! \lim_{M \to \infty} \biggl\{ 

239: \sum_{m=1}^M \! {(\log m)^n \over m}-{(\log M)^{n+1} \over n+1} \biggl\}

240: \nonumber \\

241: (\g_0 \si=\sf \g \approx 0.5772156649, \quad 

242: \g_1 \approx -0.0728158455, \quad \g_2 \approx -0.0096903632, \ldots) \\

243: E_n  \si:\sf \mbox{Euler numbers } (E_0=1,\ E_2=-1,\ E_4=5,\ldots;

244: \ E_{2m+1} = 0 \mbox{ for }m=0,1,\ldots). \nonumber 

245: \end{eqnarray}

246: [$E_n$ is also a standard  symbol for quantized energy levels,

247: a concept often invoked purely rhetorically about the Riemann zeros;

248: let us then insist that our (present) work is totally decoupled 

249: from such interpretations, 

250: whereas it sees the Euler numbers as truly essential~!]

251: 

252: Concerning the Riemann zeta function 

253: $\zeta(s) \defi \sum\limits_{k=1}^\infty k^{-s}$ \cite{T,Da,E},

254: we will need its special values,

255: \begin{equation}

256: \label{SVR}

257: \zeta(-n)= (-1)^n {B_{n+1} \over n+1} \quad (n=0,1,\ldots); \qquad 

258: \zeta(2m) = {(2 \pi)^{2m} \over 2(2m)!} |B_{2m}| \quad (m=1,2,\ldots);

259: \end{equation}

260: \begin{equation}

261: \label{Z0}

262: \zeta(0)=-\hf \qquad \mbox{and} \qquad \zeta'(0)=-\hf\log 2\pi,

263: \end{equation}

264: and its functional equation in the form 

265: \begin{equation}

266: \label{XI}

267: \Xi(s) \defi {\zeta(s) \over F(s)} \equiv {\zeta(1-s) \over F(1-s)}, \qquad

268: \qquad F(s) \defi {\pi^{s/2} \over s(s-1) \G(s/2)},

269: \end{equation}

270: where $\Xi(s)$ is an {\sl entire\/} function,

271: which is {\sl even\/} under the symmetry $s \leftrightarrow (1-s)$

272: (this expresses the functional equation), and only keeps the nontrivial zeros

273: of $\zeta(s)$.

274: 

275: In order to analyze $\zeta(s)$, 

276: we will be {\sl forced\/} to invoke a particular Dirichlet $L$-series as well

277: (associated with the Dirichlet character $\chi_4$ \cite{AS,B,Da})

278: \begin{equation}

279: \label{LDef}

280: \beta(s) \defi \sum_{k=1}^\infty (-1)^{k+1} (2k-1)^{-s} \quad [=L(\chi_4,s)];

281: \end{equation}

282: $\beta(s)$ extends to an entire function, having the special values 

283: \begin{equation}

284: \label{SVL}

285: \beta(-m)={E_m \over 2}, \qquad 

286: \beta(2m+1)= { (\pi/2)^{2m+1} \over 2 (2m)!} |E_{2m}| \qquad (m=0,1, \ldots);

287: \end{equation}

288: \begin{equation}

289: \label{L0}

290: \beta(0)= \hf \qquad \mbox{and} \qquad

291: \beta'(0)=-\textstyle{3 \over 2} \log 2 -\log \pi + 2 \log \G({1 \over 4}),

292: \end{equation}

293: and a functional equation expressible as

294: \begin{equation}

295: \label{xI}

296: \Xi_{\chi_4} (s) \defi 

297: \Bigl( {4 \over \pi} \Bigr)^{(s+1)/2} \G \Bigl( {s+1 \over 2} \Bigr) \beta(s)

298: \equiv \Xi_{\chi_4} (1-s)

299: \end{equation}

300: where the function $\Xi_{\chi_4}$ is entire 

301: and only keeps the nontrivial zeros of $\beta(s)$.

302: 

303: 

304: \section{General Delta and Zeta functions of order $<1$}

305: 

306: \subsection{Admissible sequences and Delta functions}

307: 

308: Throughout this work, a numerical sequence $\{x_k\}$

309: (systematically labeled by {\sl positive\/} integers $k=1,2,\ldots$) 

310: will be called {\sl admissible of order\/} $\mu_0<1$ if:

311: 

312: (i) $0 < x_1 \le x_2 \le \cdots,\quad  x_k \uparrow +\infty$ ;

313: 

314: \noindent (or: complex $x_k \to \infty$ with $|\arg x_k|$ sufficiently bounded

315: \cite{QHS,I}, to provide an unconditionally valid framework ``in case of need");

316: 

317: (ii) $\sum\limits_k |x_k|^{-1} < \infty$, 

318: making the following Weierstrass product {\sl converge\/},

319: \begin{equation}

320: \label{DDef}

321: \D(z \mid \{x_k\}) \defi \prod_k (1+ z /x_k) 

322: \qquad (\forall z \in \mathbb C) ;

323: \end{equation}

324: it then defines an {\sl entire\/} ``Delta" function $\D(z)$

325: (we omit the argument $\{x_k\}$ except when an ambiguity may result);

326: 

327: (iii) for $z \to\infty$, $\log \D(z)=o(|z|^{\mu_0+\delta}) \ \forall \delta>0 $,

328: and it admits a complete {\sl uniform asymptotic expansion\/}

329: in some sector $\{ |\arg z| < \theta \}$, of a form 

330: governed by some strictly decreasing sequence of real exponents $\{\mu_n\}$, as

331: \begin{eqnarray}

332: \label{DAs}

333: \log \D(z) \sim \sum_{n=0}^\infty 

334: (\tilde a_{\mu_n} \log z + a_{\mu_n})  z^{\mu_n} \qquad (z \to \infty) \\

335: \mbox{with } \mu_0 > \mu_1 > \cdots,\quad \mu_n  \downarrow -\infty, \quad 

336: \mbox{and }  0< \mu_0 < 1 \nonumber

337: \end{eqnarray}

338: (``generalized Stirling expansion", 

339: by extension from the case $x_k=k$ \cite{JL1});

340: such a uniform expansion is repeatedly differentiable in $z$.

341: 

342: Then, the Dirichlet series

343: \begin{equation}

344: \label{Zdef}

345: Z(\sigma \mid \{x_k\}) \defi \sum_{k=1}^\infty x_k^{-\sigma} \qquad 

346: (\mbox{convergent for } \Re \sigma > \mu_0)

347: \end{equation}

348: defines the {\sl Zeta function of\/} $\{x_k\}$, holomorphic in that half-plane.

349: The point $\sigma=1$ has to lie in the latter by assumption (ii), 

350: which imposes $\mu_0 < 1$; then $(\log \D)$ can moreover be expanded

351: term by term in eq.(\ref{DDef}) to yield the Taylor series

352: \begin{equation}

353: \label{DTay}

354: \log \D(z) \equiv 

355: \sum_{m=1}^\infty (-1)^{m-1} {Z(m) \over m} z^m

356: \qquad (\mbox{convergent for } |z|<x_1).

357: \end{equation}

358: 

359: Motivations: the idea here is to assume certain properties 

360: for an entire function $\D(z)$ of order $\mu_0<1$,

361: so as to generate a function $Z(\sigma)$ meromorphic in all of $\mathbb C$

362: with poles at the $\mu_n$, their maximum order $r$ being 2 here

363: (as dictated by the specific form of eq.(\ref{DAs})).

364: Such $\D(z)$ are very special instances of zeta-regularized infinite products,

365: for which much more general frameworks exist (e.g., \cite{JL1,JL2,I}).

366: However, singularities and essential complications definitely

367: increase each time $r$ or the integer part $[\mu_0]$ can become larger

368: (chiefly, the formalism leaves $([\mu_0]+1)$ meaningful integration constants

369: undetermined).

370: Efficiency then commands to minimize both latter parameters 

371: (subject to $r \ge 1$ and $\mu_0>0$);

372: specially, $\mu_0<1$ is simpler to handle than any $\mu_0 \ge 1$.

373: In this respect, spectral zeta functions frequently have $\mu_0 \ge 1$

374: (e.g., for Laplacians on compact Riemannian manifolds, 

375: $\mu_0 = \hf \times$[dimension]) but only simple poles,

376: hence the simplifying assumption $r=1$ is suitable for them \cite{V,QHS};

377: by contrast, the present functions $\Z(\sigma,v)$ 

378: will accept the lower value $\mu_0=\hf$ but will need $r=2$.

379: Another difference is that for eigenvalue spectra,

380: a ``partition function" $\sum_k \e^{-tx_k}$ is a natural starting point; 

381: in the Riemann case, that type of function ($V(t)=\sum_\rho \e^{\rho t}$) 

382: exhibits a more remote and contrived structure \cite{CR}, 

383: while Delta-type functions are easily accessible

384: (by simple alterations of eq.(\ref{XI})).

385: Thus, Riemann zeros and standard eigenvalue spectra 

386: have several {\sl mutually singular\/} features, 

387: making their unified handling rather cumbersome.

388: 

389: \subsection{Meromorphic continuation of Zeta functions}

390: 

391: The bounds $\log \D(z)=O(z^1)$ for $z \to 0$, and $O(z^{\mu_0})$ for $z\to+\infty$,

392: with $\mu_0<1$, imply that the following Mellin transform of $\log \D(z)$,

393: \begin{equation}

394: \label{IMel}

395: I(\sigma) \defi \int_0^{+\infty} \log \D(z) z^{-\sigma-1} \d z 

396: \qquad \qquad (\mu_0 < \Re \sigma <1)

397: \end{equation}

398: converges to a holomorphic function of $\sigma$ in the stated vertical strip.

399: 

400: Then, by standard arguments (see App.A, and \cite{J,V,C}),

401: $I(\sigma)$ extends to a meromorphic function on either side of that strip:

402: 

403: \noindent $a)$ by virtue of the expansion (\ref{DAs}) for $z \to \infty$, 

404: $I(\sigma)$ extends to all $\Re \sigma<1$, with 

405: \begin{equation}

406: \label{IP2}

407: \mbox{(at most) double poles at } \sigma=\mu_n, \quad 

408: \mbox{polar parts} = 

409: {\tilde a_{\mu_n} \over (\sigma-\mu_n)^2} + {a_{\mu_n} \over (\sigma-\mu_n)};

410: \end{equation}

411: 

412: \noindent $b)$ just because $\log \D(z)$ has a Taylor series at $z=0$,

413: $I(\sigma)$ extends to all $\Re \sigma>\mu_0$, with

414: \begin{equation}

415: \label{IP1}

416: \mbox{(at most) simple poles at } \sigma =m \in \mathbb N^\ast, \quad

417: \mbox{of residues } -(\log \D)^{(m)}(0)/m!\ .

418: \end{equation}

419: 

420: Now, for $\mu_0 < \Re \sigma <1$, the integral (\ref{IMel}) 

421: can also be done term by term after inserting the expansion (\ref{DDef}), giving

422: \begin{equation}

423: \label{ZMel}

424: {\sigma \sin \pi \sigma\over \pi} I(\sigma) \equiv 

425: \sum_{k=1}^\infty x_k^{-\sigma} \equiv Z(\sigma).

426: \end{equation}

427: The meromorphic extension of $I(\sigma)$ then entails that of $Z(\sigma)$ 

428: to the whole $\sigma$-plane as well, 

429: i.e., to the half-planes $\{ \Re \sigma <1 \}$ by $a)$,

430: resp. $\{ \Re \sigma > \mu_0 \}$ by $b)$ independently, so that:

431: 

432: \noindent $a')$ any {\sl non-integer\/} pole $(\mu_n\notin \mathbb Z)$ 

433: of $I(\sigma)$ generates at most a double pole for $Z(\sigma)$, with

434: \begin{equation}

435: \label{ZP2}

436: Z(\mu_n+ \eps) = 

437: \Bigl[ {\mu_n \sin \pi \mu_n \over \pi} \tilde a_{\mu_n} \Bigr] {1 \over \eps^2}

438: + \Bigl[ {\mu_n \sin \pi \mu_n \over \pi} a_{\mu_n} + 

439: \Bigl( {\sin \pi \mu_n \over \pi} + \mu_n \cos \pi \mu_n \Bigr) \tilde a_{\mu_n} \Bigr]

440: {1 \over \eps} \ [+ \mbox{regular part}];

441: \end{equation}

442: 

443: \noindent $a'')$ any {\sl negative integer\/} pole $(\mu_n=-m)$ of $I(\sigma)$ 

444: generates at most a simple pole for $Z(\sigma)$,

445: through partial cancellation with the zeros of $(\sin \pi \sigma)$, with

446: \begin{eqnarray}

447: \label{ZP1}

448: (m \in \mathbb N^\ast): \qquad Z(-m+\eps) \si=\sf 

449: (-1)^m \Bigl[ -{m \tilde a_{-m} \over \eps} + (\tilde a_{-m} - m a_{-m}) \Bigr]+O(\eps)

450: \nonumber\\

451: \label{ZRFP}

452: i.e., \mbox{ residue:} \quad \Res_{\sigma=-m} Z(\sigma) 

453: \si=\sf (-1)^{m+1} m \tilde a_{-m}\ ;\\

454: \label{ZT}

455: \mbox{ finite part:} \quad {\rm FP}_{\sigma=-m} Z(\sigma) 

456: \si=\sf (-1)^m(\tilde a_{-m} - m a_{-m})

457: \end{eqnarray}

458: (we call eqs.(\ref{ZT}) ``trace identities" by extension from spectral theory,

459: specially when $\tilde a_{-m}=0$, in which case explicit finite values 

460: for the $Z(-m)$ result);

461: 

462: \noindent $a''')$ any pole at $\sigma=0$ of $I(\sigma)$ gets fully cancelled 

463: by the double zero of $(\sigma \sin \pi \sigma)$, giving

464: \begin{equation}

465: \label{ZP0}

466: Z(\eps) = \tilde a_0 + a_0 \eps + O(\eps^2) \qquad \Longrightarrow \qquad

467: Z(0)=\tilde a_0, \quad Z'(0)=a_0.

468: \end{equation}

469: 

470: \noindent $b')$ each {\sl positive integer\/} pole of $I(\sigma)$

471: ($\sigma=m \in \mathbb N^\ast$, from eq.(\ref{IP1}))

472: gets cancelled by a zero of $(\sin \pi \sigma)$, giving

473: \begin{equation}

474: \label{ZTay}

475: Z(m)=(-1)^{m-1}(\log \D)^{(m)}(0)/(m-1)!

476: \end{equation}

477: but this last output just duplicates with the previous Taylor formula

478: (\ref{DTay}).

479: 

480: Ultimately, all the poles of $Z(\sigma)$ 

481: lie in a {\sl single\/} decreasing sequence $\{ \mu_n \}_{n \in \mathbb N}$, 

482: and have maximum order $r=2$ under the specific assumption (\ref{DAs}).

483: 

484: \section{Delta function based at $s=0$, 

485: and the Zeta function $Z(\sigma) \defi \Z(\sigma,v=\qt)$}

486: 

487: We now apply the previous framework to the Riemann zeros 

488: upon just slight changes with regard to the usual factorization

489: of $\zeta(s)$ around $s=0$.

490: 

491: \subsection{Basic facts and notations}

492: 

493: The most convenient starting point is the entire function (\ref{XI}),

494: which keeps precisely the non-trivial zeros $\{\rho\}$ of $\zeta(s)$, 

495: and has the familiar Hadamard product representation \cite{E}

496: \begin{equation}

497: \label{CHAD}

498: \Xi(s) = \e^{Bs} \prod_\rho (1-s /\rho) \e^{s /\rho}, 

499: \qquad B \defi [\log \Xi]'(0) = \log 2\sqrt\pi\ -1 -\hf\g .

500: \end{equation}

501: 

502: Zeros are henceforth grouped in pairs as 

503: $\{\rho = \hf \pm \mi \tau_k\}_{k=1,2,\ldots}$ according to eq.(\ref{ZER}).

504: Their corresponding counting function,

505: ${\mathcal N}(T) \defi {\rm card} \{\tau_k \mid \Re \tau_k < T \}$,

506: follows a well-known estimate $\overline {\mathcal N}(T)$, as \cite{T,E}

507: \begin{equation}

508: \label{GE}

509: {\mathcal N}(T) \sim \overline {\mathcal N}(T) \defi

510: {T \over 2 \pi}\Bigl[ \log {T \over 2 \pi} -1 \Bigr] \qquad

511: (T \to +\infty).

512: \end{equation}

513: 

514: Thus, the sequence of zeros itself could be admissible of order 1 at best;

515: fortunately, a transformed sequence $\{x_k\}$ and its Zeta function

516: will immediately arise:

517: \begin{equation}

518: \label{Z1}

519: x_k \defi \qt +{\tau_k}^2 , \qquad {\rm and} \qquad

520: Z(\sigma) \defi Z(\sigma \mid \{x_k\}) = \sum_k {x_k}^{-\sigma}

521: \end{equation}

522: (the latter series will converge for $ \Re \sigma > \hf $, 

523: again by the estimate (\ref{GE})).

524: 

525: Indeed, once the zeros have been reordered in pairs, it first follows that

526: $ {\mathscr Z}_1 \defi \sum\limits_\rho \rho^{-1} = 

527: \sum\limits_k {x_k}^{-1} = Z(1) $

528: (convergent sums), and then,

529: $ \Xi(s) \equiv \e^{[B+ Z(1)]s} \prod\limits_k [1+ s(s-1)/x_k]$;  

530: the parity property $\Xi(s)=\Xi(1-s)$ thereupon imposes $Z(1)=-B$, hence

531: \begin{equation}

532: \label{Z11}

533: Z(1) = {\mathscr Z}_1 = -[\log \Xi]'(0) = -\log 2\sqrt\pi\ +1 +\hf \g ,

534: \end{equation}

535: a classic result (\cite{Da}, ch.12; \cite{E}, Sec.3.8). 

536: All in all, the product formulae and functional equation boil down to 

537: \begin{equation}

538: \label{PROD}

539: {\zeta(s) \over F(s)} \equiv \D(\lambda \mid \{x_k\}) 

540: \equiv {\zeta(1-s) \over F(1-s)} 

541: \qquad \qquad \Bigl[ F(s) \defi {\pi^{s/2} \over s(s-1) \G(s/2)} \Bigr] ,

542: \end{equation}

543: \begin{equation}

544: \label{CONV}

545: {\rm where} \qquad 

546: \D(\lambda \mid \{x_k\}) \equiv \Xi(s)\ \ {\rm with\ \ } \lambda \defi s(s-1):

547: \end{equation}

548: i.e., $\Xi(s)$ has been rewritten as an infinite product 

549: $\D(\lambda \mid \{x_k\})$

550: which is {\sl manifestly even\/} (under $s \longleftrightarrow (1-s)$),

551: and will qualify as a Delta function of order $\mu_0=\hf$

552: {\sl in the variable\/} $z=\lambda$, 

553: a much simpler situation than $\mu_0=1$ (naively suggested by eq.(\ref{CHAD})).

554: We now derive the ensuing properties of $Z(\sigma)$ 

555: (as compiled in Sec.7, Table 1).

556: 

557: 

558: \subsection{Properties of $Z(\sigma)$ for $\Re \sigma <1$}

559: 

560: As basic initial result, the sequence $\{ x_k \defi \qt +{\tau_k}^2 \}$ 

561: {\sl is admissible of order\/} $\hf$: 

562: it obviously fulfills assumptions (i)--(ii);

563: the function $\Xi(s)$ is entire of order 1 in $s$, hence $\hf$ in $\lambda$;

564: finally, a large-$\lambda$ expansion of the form (\ref{DAs}) for 

565: $\log \D (\lambda \mid \{x_k\})$ is easily obtained as follows.

566: In eq.(\ref{PROD}), for $s \to \infty \ (|\arg s| < \pi/2)$, 

567: $-\log F(s)$ can be replaced by its complete Stirling expansion  

568: and $\log \zeta(s) = O(s^{-\infty})$ can be deleted, giving

569: \begin{equation}

570: \label{Z1as}

571: \log \D (\lambda) \sim \log \lambda - {\log \pi \over 2} s 

572: + {s \over 2} (\log s -\log 2 -1) - {1 \over 2} \log {s \over 2} 

573: +{1 \over 2} \log 2 \pi 

574: + \sum_{m \ge 1} {B_{2m} 2^{2m-1}\over 2m(2m-1)} s^{-2m+1};

575: \end{equation}

576: whereupon $s$ is to be substituted by the relevant solution branch 

577: of $ s(s-1)=\lambda $, namely

578: \begin{equation}

579: \label{SUB}

580: s = \hf + \sqrt{\lambda + \qt}  \sim

581: \lambda ^{1/2} + \hf + 

582: \sum_{n=1}^\infty {2^{-2n}\G(3/2) \over n! \G(-n+3/2)}\lambda ^{1/2-n} .

583: \end{equation}

584: The resulting $\lambda$-expansion then has all the required properties, 

585: with the exponents $\{\mu_n=\hf (1-n)\}$ --- giving the order $\mu_0 = \hf$ --- 

586: and leading coefficients

587: \begin{eqnarray}

588: \label{TAa1}

589: \tilde a_{1/2} \si=\sf \qt \qquad \qquad \qquad \qquad \qquad

590: a_{1/2} = - \hf (1 + \log 2\pi) \\

591: \label{TAa2}

592: \tilde a_0 \si=\sf \textstyle {7 \over 8}  \qquad \qquad \qquad  \qquad \qquad

593: a_0 = \qt \log 8\pi \\

594: \label{TAa3}

595: \tilde a_{-1/2} \si=\sf \textstyle {1 \over 32}  \qquad \qquad \qquad \qquad \quad

596: a_{-1/2} = -{1 \over 16} \log 2\pi- {1 \over 48} ,\quad  etc.

597: \end{eqnarray}

598: This expansion can be computed to any power $\mu_n$ in principle;

599: still, a reduced general formula for $a_{\mu_n}$ looks inaccessible this way;

600: by contrast, all the $ \tilde a_{\mu_n}$ with $\mu_n \ne 0$ arise from 

601: the single substitution of eq.(\ref{SUB}) into the prefactor 

602: ${s \over 2}$ of $ \log s \ (\sim \hf\log \lambda) $ in eq.(\ref{Z1as}),

603: giving 

604: \begin{equation}

605: \label{TA}

606: \tilde a_{1/2-m}= {2^{-2m-2}\G(3/2) \over m! \G(-m+3/2)}, 

607: \qquad \tilde a_{-1-m}=0, \qquad m=0,1,2,\ldots .

608: \end{equation}

609: 

610: Now, by Sec.2.2$a'$--$a'''$), 

611: a large-$\lambda$ expansion for $\log \D (\lambda)$ 

612: translates into explicit properties of $Z(\sigma)$ for $\Re \sigma<1$.

613: Here, $Z(\sigma)$ gets a double pole at each {\sl half\/}-integer $\hf-m

614: \ (m \in \mathbb N)$, with principal polar term 

615: ${(-1)^m \over \pi} (\hf -m) \tilde a_{1/2-m}/

616: (\sigma +m-\hf)^2$, 

617: and is regular elsewhere; the leading pole $\sigma= \hf$ has 

618: \begin{equation}

619: \label{ZPP}

620: \mbox{full polar part} \ =

621: {1 \over 8 \pi}{1 \over (\sigma-\hf)^2} - 

622: {\log 2 \pi \over 4 \pi}{1 \over \sigma-\hf}.

623: \end{equation}

624: At $\sigma=0$, eqs.(\ref{TAa2}) and (\ref{ZP0}) 

625: deliver {\sl two\/} explicit values,

626: \begin{equation}

627: \label{Z10}

628: \textstyle

629: Z(0) = \tilde a_0 = {7 \over 8}, \qquad \qquad 

630: Z'(0) = a_0 = \qt \log 8\pi \ (\approx 0.806042857) ;

631: \end{equation}

632: the latter makes the {\sl Stirling constant\/}

633: (the value $\exp-[Z'(0)]$, or regularized determinant) 

634: also explicitly known for this sequence $\{x_k\}$.

635: 

636: Further quantities, 

637: tied to the  yet unspecified general coefficients $a_{\mu_n}$,

638: will also acquire explicit closed forms:

639: the polar terms of order $1/(\sigma +m-\hf)$

640: and the finite values $Z(-m)$ for all $m \in \mathbb N$, see Table 1;

641: but we will need a more indirect approach (Sec.6.1).

642: 

643: \subsection{Properties of $Z(\sigma)$ for $\Re \sigma> \hf $}

644: 

645: Now by Sec.2.2$b'$), $Z(\sigma)$ is holomorphic

646: in the half-plane $\Re \sigma> \hf $,

647: and the values $Z(n)$ for $n=1,2,\ldots$ 

648: lie in the Taylor series of $ \log \D (\lambda) $ at $\lambda=0$,

649: which can be specified here through eq.(\ref{PROD}).

650: 

651: \medskip

652: 

653: We first expand $ \log \D (\lambda) $ in powers of $s$, where $\lambda=s(s-1)$.

654: Eq.(\ref{CHAD}) directly implies

655: \begin{equation}

656: \label{ZZ}

657:  \log \D (\lambda)  \equiv \log \Xi(s)=

658: -\sum_{n=1}^\infty {{\mathscr Z}_n \over n} s^n \qquad 

659: ({\mathscr Z}_n \equiv \sum_\rho \rho^{-n});

660: \end{equation}

661: then, from eq.(\ref{PROD}), $\log[\zeta(s)/F(s)]$ and 

662: $\log[\zeta(1-s)/F(1-s)]$ (expressing the functional equation)

663: respectively yield the two Taylor series

664: \begin{eqnarray}

665: \label{EX1}

666: -\log \sqrt \pi \ s + \log \G(1+s/2) + \log (1-s) + \log[-2 \zeta(s)]

667: \qquad \qquad \qquad \qquad \qquad \qquad \nonumber\\

668:  = -\log \sqrt \pi \, s + \Bigl[ 

669: -{ \g \over 2} s + \sum_{n=2}^\infty {(-1)^n \zeta(n) \over 2^n n}s^n \Bigr]

670: - \sum_{n=1}^\infty {1 \over n} s^n 

671: + \sum_{n=1}^\infty {(\log |\zeta|)^{(n)}(0) \over n!} s^n , \\ 

672: \label{EX2}

673: \log (1-s) + \Bigl[ {\log \pi \over 2} (s-1) +

674: \log \G \bigl( {1-s \over 2} \bigr) \Bigr] + \log [-s \zeta(1-s)] 

675: \qquad \qquad \qquad \qquad \nonumber\\

676: = \Bigl({\g \over 2} + \log 2 \sqrt \pi -1 \Bigr) s +

677: \sum_{n=2}^\infty {(1-2^{-n}) \zeta(n) - 1 \over n} s^n 

678: - \sum_{n=1}^\infty { \g^{\rm c}_{n-1} \over (n-1)! } s^n .

679: \end{eqnarray}

680: The $\g^{\rm c}_n$ in the last line are {\sl cumulants\/} 

681: for the Stieltjes constants $ \g_n $ of eq.(\ref{NOT}) \cite{AS} 

682: (cf. also the $\eta_n$ in ref. \cite{BL}, Sec. 4), i.e., 

683: \begin{eqnarray}

684: \label{LAU}

685: & \displaystyle \log [-s \zeta(1-s)] \equiv

686: -\sum_{n=1}^\infty {\g^{\rm c}_{n-1} \over (n-1)! } s^n 

687: \qquad vs \qquad

688: -s \zeta(1-s) \equiv 

689: 1 - \sum_{n=1}^\infty {\g_{n-1} \over (n-1)! } s^n  \\

690: &(\g^{\rm c}_0=\g_0=\g, \quad \g^{\rm c}_1 = \g_1 + \hf\g^2 \approx 0.260362078,

691: \quad \g^{\rm c}_2 = \g_2 + 2\g \g_1 + {2 \over 3} \g^3 \approx 0.034459088,

692: \ldots).&

693: \nonumber

694: \end{eqnarray}

695: The identification of the three series (\ref{ZZ})--(\ref{EX2})

696: at each order $s^n$ now yields a countable sequence of 3-term identities:

697: the first one just restores the result 

698: ${\mathscr Z}_1 = -B$ as in eq.(\ref{Z11});

699: then, the subsequent ones likewise express the higher ${\mathscr Z}_n$ 

700: in {\sl two\/} ways,

701: \begin{equation}

702: \label{ZRH}

703: {\mathscr Z}_n =

704: 1 - (-1)^n 2^{-n} \zeta(n) - {(\log |\zeta|)^{(n)}(0) \over (n-1)!}

705: = 1 - (1-2^{-n}) \zeta(n) + { n \over (n-1)!} \g^{\rm c}_{n-1} ,

706: \quad n=2,3,\ldots

707: \end{equation}

708: That short argument subsumes several earlier results.

709: The rightmost and center expressions in eq.(\ref{ZRH}) amount to formulae for

710: ${\mathscr Z}_n$ by Matsuoka \cite{Ms2} 

711: and Lehmer (\cite{L}, eq.(12)) respectively;

712: the implied relations between 

713: the derivatives $\zeta^{(n)}(0)$ and the Stieltjes constants $\g_n$

714: were also discussed in \cite{A,CHO},

715: together with Euler--Maclaurin formulae for the $\zeta^{(n)}(0)$

716: which parallel the specification of the $\g_n$ in eq.(\ref{NOT}).

717: 

718: \medskip

719: 

720: As for the values $Z(m)$ themselves, 

721: they are given by eq.(\ref{ZTay}) now using $\lambda=s(s-1)$ 

722: as expansion variable, i.e.,

723: \begin{equation}

724: \label{ZN}

725: Z(m) = {(-1)^{m-1} \over (m-1)!} \Bigl[ 

726: \Bigl( {1 \over 2s-1}{\d \over \d s} \Bigr)^m \log \Xi (s) 

727: \Bigr] _{s=0 {\rm \ or\ } 1} \qquad (m=1,2,\ldots) .

728: \end{equation}

729: But alternatively, $Z(m) \equiv {\mathscr Z}_m + \bigl[ 

730: \mbox{a {\sl finite\/} linear combination of the } 

731: \{ {\mathscr Z}_n \}_{n=1,\ldots,m-1} \bigr]$, and vice-versa: as shown below,

732: \begin{equation}

733: \label{SZ}

734: {{\mathscr Z}_n \over n} = \!\! \sum_{0 \le \ell \le n/2} \!\! 

735: (-1)^\ell {n-\ell \choose \ell} { Z(n-\ell) \over n - \ell} \quad

736: \Longleftrightarrow \quad 

737: Z(m) = \sum_{\ell=0}^{m-1} {m+\ell-1 \choose m-1} {\mathscr Z}_{m-\ell}

738: \end{equation}

739: ($Z(1)={\mathscr Z}_1,\ Z(2)={\mathscr Z}_2+2 {\mathscr Z}_1, \ldots$).

740: The ${\mathscr Z}_n$ being already known from eqs.(\ref{Z11}) and (\ref{ZRH}), 

741: $Z(m)$ then reduces to an explicit affine combination (over the rationals) 

742: of $B=[\log \Xi]'(0)$, $\zeta(n)$, and 

743: either $(\log |\zeta|)^{(n)}(0)$ or $\g^{\rm c}_{n-1}$ for $1< n \le m$

744: (see Table 1), as stated earlier by Matiyasevich \cite{MA}. 

745: (More recently, $Z(1)$ and $Z(2)$ also got revived 

746: in studies of the distribution of primes \cite{RS,Lb}).

747: 

748: Since ref.\cite{MA} uses eq.(\ref{SZ}) for $Z(m)$ without mentioning any proof,

749: we sketch one.

750: First, if $x \defi \rho(1-\rho)$, the expansion of the identity

751: $\log \bigl[ (1-s/\rho)(1-s/(1-\rho)) \bigr] \equiv \log \bigl[ 1-s(1-s)/x \bigr] $

752: in powers of $s$ yields 

753: $ [ \rho^{-n}+(1-\rho)^{-n}] /n  \equiv 

754: \sum_{0 \le \ell \le n/2} (-1)^\ell {n-\ell \choose \ell} x^{-n+\ell}/(n-\ell)$

755: for $n=1,2,\ldots$. By recursion, this must invert in the form 

756: $x^{-m} \equiv \sum\limits_{n=1}^m V_{m,n} [ \rho^{-n}+(1-\rho)^{-n}] $;

757: then, $\sum\limits_{n=1}^m V_{m,n} \rho^{-n}$ has to be the singular part 

758: in the Laurent series of $x^{-m}$ around $\rho=0$, 

759: resulting in $V_{m,n}={2m-n-1 \choose m-1}$. 

760: Now, summing both sets of identities over the Riemann zeros $\{ \rho \}$ 

761: yields the stated decompositions (\ref{SZ}).

762: (We stress that their finite character is specific to the $Z(m)$ as opposed

763: to all other values $\Z(m,v)$, $v \ne \qt$.)

764: \medskip

765: 

766: {\sl Note added in proof.\/} ---

767: For completeness, we quote two other sets of identities 

768: for the sums ${\mathscr Z}_n$ \cite{VZ}:

769: $$ 2 {\mathscr Z}_k = - \sum_{\ell=k+1}^\infty 

770: {\ell-1 \choose k-1} {\mathscr Z}_\ell \qquad \qquad 

771: \mbox{for each odd } k \ge 1 $$

772: (a countable sequence of ``sum rules", easy but unreported,

773: which allow to eliminate any {\sl finite\/} subset of odd values); 

774: and the connection to {\sl Li's coefficients\/} (cf. [39], thm 2),

775: $ \lambda_n {\stackrel{\rm def}{=}} \sum_\rho \, [1-(1-1/\rho)^n] $

776: [which allow to recast 

777: the Riemann Hypothesis as $ \lambda_n >0 \ (\forall n)$]:

778: $$ \lambda_n = \sum_{j=1}^n (-1)^{j+1} {n \choose j} {\mathscr Z}_j

779: \qquad \Longleftrightarrow \qquad 

780: {\mathscr Z}_n = \sum_{j=1}^n (-1)^{j+1} {n \choose j} \lambda_j \qquad

781: (n=1,2,\ldots) . $$

782: 

783: 

784: \section{Generalized Zeta functions 

785: $\Z(\sigma,v) \equiv Z(\sigma \mid \{ {\tau_k}^2+v \} )$}

786: 

787: We begin to discuss the Hurwitz-like generalizations of the preceding case

788: obtained by shifting the {\sl squared\/} parameters ${\tau_k}^2$.

789: The obviously allowed translations

790: ($x_k \mapsto x_k+v'',\quad  v''>-x_1$ in the real case) preserve the notion of 

791: admissible sequences (with their values of $\mu_0,\ r$);

792: that validates the earlier definition

793: (\ref{ZDef}), as $\Z(\sigma,v) = Z(\sigma \mid \{ {\tau_k}^2+v \} )$.

794: 

795: The corresponding transformation of Delta functions 

796: (as Hadamard products of order $<1$)

797: only involves an explicit constant denominator to preserve 

798: their specific normalization $\D(0)=1$, as

799: \begin{equation}

800: \label{SHD}

801: \D(\lambda \mid \{ {\tau_k}^2+v \} ) \equiv 

802: \D(\lambda + v-v' \mid \{ {\tau_k}^2+v' \} ) \big/

803: \D(v-v' \mid \{ {\tau_k}^2+v' \} ).

804: \end{equation}

805: Alternatively, we might have opted to normalize Delta functions as

806: {\sl zeta-regularized\/} infinite products, i.e., 

807: \begin{equation}

808: \label{DRG}

809: \D_{\rm zr}(\lambda) \defi

810: \exp [-\partial_\sigma Z(\sigma \mid \{ {\tau_k}^2+\lambda \} )]_{\sigma=0} \, ,

811: \end{equation}

812: which are fully translation-covariant, but at the same time less explicit.

813: The overall benefit of this normalization is then dubious 

814: within the restricted scope of this work; but here,

815: it explains a dichotomy between algebraic and transcendental properties 

816: of Zeta functions, which roughly follows our overall division between 

817: $\{ \Re \sigma <1 \}$ and $\{ \Re \sigma > \mu_0 \}$ properties, but not quite.

818: 

819: \bigskip

820: 

821: Covariance implies that if we translate $\lambda \mapsto (\lambda + v)$,

822: the expansion of $\log \D_{\rm zr}(\lambda)$ 

823: around the {\sl invariant\/} point $\lambda=+\infty$ 

824: can be recomputed to any order by straight substitution,

825: yielding {\sl explicit polynomials\/} in $v$ as coefficients.

826: When $\mu_0<1$ as here, then 

827: $ \D_{\rm zr}(\lambda) \equiv \e^{-a_0} \D(\lambda)$ \cite{V},

828: hence the previous statement holds for the expansion (\ref{DAs}) 

829: {\sl minus its term of order\/} $\lambda^0$;

830: i.e., all the shifted coefficients $\tilde a_{\mu_n}(v), \ a_{\mu_n}(v)$ 

831: will be polynomial excepting $a_0(v)$.

832: For $\log \D(\lambda \mid \{ {\tau_k}^2+v \} )$ specifically, 

833: eqs.(\ref{TAa1}--\ref{TA}) (at $v=\qt$) imply that

834: \begin{equation}

835: \label{AA}

836: \tilde a_{1/2},\ a_{1/2},\ \tilde a_0  \mbox{ stay constant \quad (as well as }

837: \tilde a_{-1} \equiv \tilde a_{-2} \equiv \cdots \equiv 0);

838: \end{equation}

839: \begin{equation}

840: \label{TAG}

841: \tilde a_{1/2-m}(v) \equiv {\G(3/2) \over \G(-m+3/2)}{v^m \over 4\,m!}.

842: \end{equation}

843: For functions like $\Z(\sigma,v)$, the consequences are that their polar parts 

844: and ``trace identities" will depend polynomially on $v$;

845: furthermore, a single (fixed-$v$) large-$\lambda$ expansion, 

846: such as eq.(\ref{Z1as}) for 

847: $\log \D(\lambda \mid \{ {\tau_k}^2+{1 \over 4} \} )$, 

848: suffices to express those $v$-dependences in full. 

849: Precisely here, by eq.(\ref{AA}): $\Z(\sigma ,v)$ keeps its rightmost 

850: ($\sigma=\hf$) full polar part {\sl constant\/} (and given by eq.(\ref{ZPP})),

851: as well as its value at 0 ($\Z(0,v) \equiv {7 \over 8}$);

852: all its other poles (of order 2, except at $v=0$) keep fixed locations.

853: (As a by-product, any difference function 

854: $[\Z(\sigma ,v)-\Z(\sigma ,v_0)]$ is holomorphic for $\Re \sigma > -\hf$).

855: We can specify such polynomial formulae in the half-plane $\Re \sigma < \hf$

856: still further, but only later by a different path (Sec.6).

857: 

858: \bigskip

859: 

860: By contrast, all formulae for $\Z(\sigma,v)$ 

861: in the half-plane $\{\Re \sigma > \hf \}$ refer to Taylor coefficients

862: of $\log \D_{\rm zr}(v+\lambda)$ around  $v$ {\sl finite\/},

863: which evolve {\sl transcendentally\/} with $v$: here they will only express 

864: in terms of $\log |\zeta(s)|$ (or $\log \Xi(s)$, from eq.(\ref{XI}))

865: and its derivatives at $s= \hf \pm  v^{1/2}$.

866: The first of those coefficients, $\log \D_{\rm zr}(v)$ itself ($= -a_0(v)$),

867: actually yields a special value lying at $\sigma=0$, by eq.(\ref{DRG}):

868: \begin{eqnarray}

869: \label{DETG}

870: \partial_\sigma  \Z(\sigma,v)_{\sigma=0} = a_0(v) 

871: \si=\sf a_0(\qt)-\log \D(v- \qt \mid \{ {\tau_k}^2 + \qt \} ) \nonumber\\

872: \si=\sf \qt \log 8\pi - \log \Xi(\hf \pm v^{1/2})

873: \end{eqnarray}

874: (a result which fully matches eq.(\ref{DZet1}) below 

875: for the Hurwitz-type function $\xi$ \cite{Dn,SS}).

876: Then, the Taylor coefficients of order $n \ge 1$

877: (identical for $\log \D(\lambda)$ and $\log \D_{\rm zr}(\lambda)$) yield

878: \begin{eqnarray}

879: \label{ZNV}

880: & \displaystyle \Z(n,v) = {(-1)^{n-1} \over (n-1)!} 

881: {\d^n \over \d v^n} \log \Xi \bigl( \hf \pm v^{1/2} \bigr) 

882: \qquad (n=1,2,\ldots), & \\

883: \label{Z1V}

884:  e.g., \qquad 

885: & \Z(1,v) = \pm \hf v^{-1/2} (\log \Xi)'(s=\hf \pm v^{1/2}) \quad (v \ne 0),

886: \qquad \Z(1,0) = \hf (\log \Xi)''(\hf) & \nonumber

887: \end{eqnarray}

888: (but we are in lack of more reduced closed forms for general $(n,v)$).

889: 

890: \medskip

891: 

892: In summary, the polar parts of $\Z(\sigma,v)$ and the special values 

893: $\{ \Z(-n,v) \}_{n \in \mathbb N}$ (``trace identities") 

894: have polynomial expressions in $v$; 

895: whereas $\{ \Z(n,v)\ (n \ne 0$) {\sl plus\/ $\partial_\sigma \Z(0,v)$ at $n=0 \}$}

896: are also special values, but only computable transcendentally. 

897: (This conclusion is moreover fully obeyed for typical spectral zeta functions.)

898: 

899: 

900: \section{Delta function based at $s= \hf$, and the Zeta function 

901: $\Z(\sigma) \defi \Z(\sigma,v=0)$}

902: 

903: An interesting option is now to shift the parameter $v$ 

904: from its initial value $\qt$ in $Z(\sigma)$, 

905: to the most symmetrical value $v=0$. 

906: By eq.(\ref{SHD}), the Hadamard product (\ref{CONV}) becomes based at $s=\hf$ as

907: \begin{equation}

908: \label{PRD}

909: \Xi(\hf + t) \equiv \Xi(\hf) \D (t^2 \mid \{{\tau_k}^2\}) \qquad

910: \mbox{with} \quad t \defi s - \hf , \quad t^2 = \lambda + \qt,

911: \end{equation}

912: \begin{equation}

913: \label{XIV}

914: \mbox{and} \quad

915: \Xi(\hf) = - \pi^{-1/4} \G(\fq) \zeta(\hf) \quad (\approx 0.994241556)

916: \end{equation}

917: (but very little is known about $\zeta(\hf)$ \cite{Ms1}

918: and we cannot make this constant factor any more explicit,

919: contrary to the special case $v=\qt$ where that factor was $\Xi(0)=1$).

920: 

921: The factorized representation (\ref{PROD}) then transforms to 

922: \begin{equation}

923: \label{FAC}

924: \D (t^2 \mid \{{\tau_k}^2\}) {{\bf D}(t) \over 1-2t} = \zeta(\hf + t), \qquad

925: {\bf D}(t) \defi 

926: \zeta(\hf) \G(\fq) \pi^{t/2} \big/ \G(\fq + \textstyle{t \over 2}).

927: \end{equation}

928: (This Delta function is closest to the determinant of Riemann zeros used

929: by Berry--Keating for other purposes \cite{BK}.)

930: 

931: We accordingly switch to the Zeta function of the sequence $\{{\tau_k}^2\}$

932: \cite{G1,Dl}, or in short,

933: \begin{equation}

934: \label{Z2}

935: \Z(\sigma) \defi \Z(\sigma,0)  

936: = \sum_{k=1}^\infty {\tau_k}^{-2\sigma} , \qquad \Re \sigma > \hf .

937: \end{equation}

938: Numerically, this new function looks almost indistinguishable 

939: from $Z(\sigma)$ (see App.B).

940: (Also, by Sec.4, $(Z-\Z)(\sigma)$ extends holomorphically to 

941: $\Re \sigma > - \hf$.)

942: By contrast, the {\sl meromorphic continuation\/} of $ \Z(\sigma) $ 

943: will prove to be distinctly simpler than that of $Z(\sigma)$,

944: thanks to specially explicit representation formulae 

945: in the half-plane $\{\Re \sigma < \hf \}$.

946: To obtain these, we now switch to a more powerful, special to $v=0$, approach

947: (whereas the earlier considerations would still describe $\Z(\sigma)$, 

948: but just to the same extent as $Z(\sigma)$).

949: 

950: \subsection{The shifted spectrum of trivial zeros}

951: 

952: The factor ${\bf D}(t)$ in eq.(\ref{FAC}) has the structure of a 

953: {\sl spectral determinant\/} built over the ``spectrum" of trivial zeros

954: {\sl in the variable\/} $(-t)$, namely $\{\hf+2k\}$

955: (${\bf D}(t)$ is not exactly the zeta-regularized determinant,

956: but again this will not matter here).

957: That spectral interpretation can be extended to the factor $(1-2t)^{-1}$, 

958: by treating the pole $t=\hf$ (of $\zeta(\hf + t)$)  

959: as a ``ghost eigenvalue" of multiplicity ($-1$). 

960: A major role of the spectrum of trivial zeros is to make 

961: $\log [{\bf D}(t)/(1-2t)]$ asymptotically cancel  

962: $\log \D(t^2 \mid \{{\tau_k}^2\})$ {\sl to all orders\/} when 

963: $t \to \infty$ in $|\arg t| < \pi/2$,

964: given that $\log \zeta(\hf + t)$ decreases exponentially there.

965: 

966: We therefore expect the {\sl spectral zeta function of the trivial zeros\/} 

967: (of $\zeta(\hf - t)$) to play an important role;

968: this ``shadow zeta function of $\zeta(s)$" (for short) involves

969: {\sl both $\zeta(s)$ itself and the partner function} $\beta(s)$,

970: in the combination

971: \begin{equation}

972: \label{SDef}

973: {\bf Z}(s) \defi \sum_{k=1}^\infty (\hf + 2k)^{-s} \equiv 

974: 2^s \Bigl[\hf \Bigl( (1-2^{-s}) \zeta(s) +\beta(s) \Bigr) -1 \Bigr]

975: \end{equation}

976: ($= 2^{-s} \zeta(s, \fq)$ in terms of the Hurwitz zeta function).

977: ${\bf Z}(s)$ has a single simple pole at $s=1$, of residue $\hf$, 

978: and admits the special values 

979: \begin{eqnarray}

980: \label{SVS1}

981: {\bf Z}(-n)

982: \si=\sf -{2^n \over n+1} B_{n+1}(\qt)-2^{-n} 

983: = \hf \Bigl[ (1-2^{-n}) {B_{n+1} \over n+1} + 2^{-n-1} E_n \Bigr] - 2^{-n}

984: , \quad n=0,1,\ldots \\

985: \label{SVS2}

986: {\bf Z}(n) \si=\sf {(-1)^n 2^{-n} \over (n-1)!} [\log \G]^{(n)}(\fq), 

987: \qquad \qquad \qquad \qquad \qquad \qquad  \qquad \qquad \qquad n=2,3,\ldots \\

988: \si=\sf \left\{  

989: { \hf \Bigl[ (2^{2m}-1) {(2\pi)^{2m} \over 2 (2m)!} |B_{2m}| 

990: + 2^{2m} \beta(2m) \Bigr] -2^{2m}, \quad n=2m  \atop

991:  \hf \Bigl[ (2^{2m+1}-1)  \zeta(2m+1) 

992: + {\pi^{2m+1} \over  2 (2m)!} |E_{2m}|  \Bigr] -2^{2m+1}, \quad n=2m+1} \right\}

993: m=1,2,\ldots \nonumber \\

994: \label{SVS3}

995: {\bf Z}(0) \si=\sf 

996: \textstyle -{3 \over 4} \qquad \mbox{and} \qquad {\bf Z}'(0) =  

997: -{7 \over 4} \log 2 - \hf \log \pi + \log \G \bigl( {1 \over 4} \bigr).

998: \end{eqnarray}

999: 

1000: Remark: literally, the framework of Sec.2 excludes the sequence of trivial zeros

1001: (of linear growth, and order $\mu_0=1$), 

1002: but the truly relevant function here will be ${\bf Z}(2 \sigma)$, 

1003: as Zeta function of the modified sequence $\{ (\hf + 2k)^2 \}$,

1004: which is admissible of order $\hf$ again. 

1005: 

1006: 

1007: \subsection{Meromorphic continuation formulae for $\Z(\sigma)$}

1008: 

1009: We start from a slight variant of the representation (\ref{ZMel}) 

1010: for $\Z(\sigma)$, obtained through an integration by parts upon 

1011: the Mellin formula (\ref{IMel}) 

1012: (where $z \equiv t^2$, by eq.(\ref{PRD})):

1013: \begin{equation}

1014: \label{JMel}

1015: \Z(\sigma) = {\sin \pi \sigma \over \pi} J(\sigma), \qquad

1016: J(\sigma) \defi \int_0^{+\infty} (t^2)^{-\sigma} 

1017: \d\,\log \D(t^2 \mid \{{\tau_k}^2\}) \quad 

1018: (\hf < \Re \sigma <1) .

1019: \end{equation}

1020: We next introduce a (regularized) {\sl resolvent trace\/} 

1021: for the spectrum of trivial zeros,

1022: \begin{equation}

1023: \label{RS}

1024: {\bf R}(t) \defi {\d \over \d t} \log {\bf D} (t) = \hf \Bigl[ \log \pi 

1025: - { \G' \over \G} \bigl(\fq + \textstyle{t \over 2} \bigr) \Bigr],

1026: \end{equation}

1027: which has a simple pole of residue $+1$ at each trivial zero of 

1028: $\zeta (\hf+t)$; 

1029: a corresponding function for the pole (``ghost") at $t={1\over 2}$ is 

1030: \begin{equation}

1031: \label{RG}

1032: R_{\rm g}(t) = -1 / (t - \hf) \qquad (\mbox{with residue } (-1)).

1033: \end{equation}

1034: Then, upon insertion of the factorization formula (\ref{FAC}), 

1035: eq.(\ref{JMel}) yields

1036: \begin{equation}

1037: \label{JV}

1038: J(\sigma) \equiv \int_0^{+\infty} t^{-2\sigma} \Bigl[ -{\bf R}(t) - R_{\rm g}(t)

1039: + {\zeta ' \over \zeta} \bigl(\hf+t \bigr) \Bigr] \d t .

1040: \end{equation}

1041: Now a crucial feature of the case $v=0$ is that {\sl this\/} integral 

1042: is also a Mellin transform with respect to the argument appearing 

1043: in the factorized form of $\zeta(s)$ 

1044: (namely the variable $t$, in eq.(\ref{FAC})).

1045: As a consequence, the contribution to $J$ from ${\bf R}(t)$ 

1046: (and also $R_{\rm g}(t)$) 

1047: can be neatly extracted and evaluated, in closed and interpretable form.

1048: Because the factor in brackets in eq.(\ref{JV}) is $O(t)$ at $t=0$

1049: (due to the functional equation), $J(\sigma)$ can be split as

1050: \begin{eqnarray}

1051: \label{JSR}

1052: J(\sigma)={\bf J}(\sigma)+J_{\rm r}(\sigma), \qquad {\bf J}(\sigma) \si\defi\sf 

1053: \int_0^{+\infty} t^{-2\sigma}[{\bf R}(0)-{\bf R}(t)] \d t \\

1054: J_{\rm r}(\sigma) \si\defi\sf 

1055: \int_0^{+\infty} t^{-2\sigma} \Bigl[ R_{\rm g}(0)-R_{\rm g}(t) +

1056: {\zeta ' \over \zeta} \bigl(\hf + t \bigr) 

1057: - {\zeta ' \over \zeta} \bigl(\hf \bigr) \Bigr] \d t , \nonumber

1058: \end{eqnarray}

1059: ({\sl this\/} splitting preserves the convergence strip 

1060: $\{ \hf < \Re \sigma <1\}$ for both resulting integrals).

1061: 

1062: $J_{\rm r}(\sigma)$ can be split still further, 

1063: once its integration path has been rotated by a small angle: 

1064: either $+\eps$ or $-\eps$, 

1065: in order to bypass the poles of $R_{\rm g}(t)$ and of

1066: ${\zeta ' \over \zeta} \bigl(\hf + t \bigr)$ at $t = \hf$,

1067: \begin{eqnarray}

1068: \label{JGZ}

1069: J_{\rm r}(\sigma) = J_{\rm g}^\pm (\sigma) + J_\zeta^\pm (\sigma), \qquad  

1070: J_{\rm g}^\pm (\sigma) \si\defi\sf

1071: \int_0^{+\e^{\pm \mi \eps} \!\infty} t^{-2\sigma} [R_{\rm g}(0)-R_{\rm g}(t)] 

1072: \d t \\

1073:  J_\zeta^\pm (\sigma) \si\defi\sf \int_0^{+\e^{\pm \mi \eps} \!\infty} t^{-2\sigma} 

1074: \Bigl[ {\zeta ' \over \zeta} \bigl(\hf + t \bigr) 

1075: - {\zeta ' \over \zeta} \bigl(\hf \bigr) \Bigr] \d t . \nonumber

1076: \end{eqnarray}

1077: 

1078: Now, ${\bf J}(\sigma),\ J_{\rm g}^\pm (\sigma)$ can be straightforwardly

1079: transformed into Hankel contour integrals and then 

1080: {\sl computed in closed form\/} (by the residue calculus), giving

1081: \begin{equation}

1082: \label{JSG}

1083: {\bf J}(\sigma) = -{\pi\, {\bf Z}(2\sigma) \over \sin 2 \pi \sigma}, \qquad

1084: J_{\rm g}^\pm (\sigma) = {\pi\, 2^{2\sigma} 

1085: \e ^{\mp 2 \pi \mi \sigma} \over \sin 2 \pi \sigma} ,

1086: \end{equation}

1087: both of which are explicit functions, meromorphic in the whole plane;

1088: chiefly, ${\bf J}$ brings in the shadow zeta function (\ref{SDef}).

1089: 

1090: Then, again upon back-and-forth integrations by parts, 

1091: $J_\zeta^\pm (\sigma)$ continue to

1092: \begin{equation}

1093: \label{JZ}

1094: J_\zeta^\pm (\sigma) \equiv \int_0^{+\e^{\pm \mi \eps} \!\infty} t^{-2\sigma} 

1095: {\zeta ' \over \zeta} \bigl(\hf + t \bigr) \d t \qquad 

1096: \mbox{analytic for } -\infty < \Re \sigma < \hf ,

1097: \end{equation}

1098: and (cf. eq.(\ref{IP1}))

1099: these integrals admit meromorphic extensions to the whole plane, with

1100: \begin{equation}

1101: \label{JP}

1102: \mbox{simple poles at } \sigma={n \over 2}, \quad \mbox{of residues } 

1103: -{1 \over 2} {(\log |\zeta|)^{(n)} (\hf) \over (n-1)!}, \quad

1104: n=1,2,\ldots

1105: \end{equation}

1106: (the difference 

1107: $J_\zeta^+(\sigma) - J_\zeta^-(\sigma)\equiv 2 \mi \pi 2^{2\sigma}$ 

1108: is entire).

1109: 

1110: \medskip

1111: 

1112: All in all, we finally get two complex conjugate Mellin representations 

1113: for $ \Z(\sigma) $:

1114: \begin{equation}

1115: \label{ZC}

1116: \Z(\sigma) = 

1117: {-{\bf Z}(2\sigma)+ 2^{2\sigma} \e^{\mp 2 \pi \mi \sigma} \over 2 \cos \pi \sigma}

1118: + {\sin \pi \sigma \over \pi} \int_0^{+\e^{\pm \mi \eps} \!\infty} t^{-2\sigma}

1119: {\zeta ' \over \zeta} \bigl(\hf + t \bigr) \d t ;

1120: \end{equation}

1121: and one real principal-value integral representation given by their half-sum,

1122: \begin{equation}

1123: \label{ZR1}

1124: \Z(\sigma) = 

1125: {-{\bf Z}(2\sigma)+ 2^{2\sigma} \cos 2 \pi \sigma \over 2 \cos \pi \sigma}

1126: + {\sin \pi \sigma \over \pi} \ {\bf -} \!\!\!\!\!\! \int_0^{+\infty} t^{-2\sigma}

1127: {\zeta ' \over \zeta} \bigl(\hf + t \bigr) \d t ,

1128: \end{equation}

1129: (each of the above converges in the full half-plane $\{\Re \sigma < \hf \}$).

1130:  

1131: \noindent Another real form can be obtained with a regular integrand, 

1132: directly from eq.(\ref{JSR}): 

1133: \begin{equation}

1134: \label{ZR2}

1135: \Z(\sigma) = 

1136: - {{\bf Z}(2\sigma) \over 2 \cos \pi \sigma}

1137: + {\sin \pi \sigma \over \pi} \int_0^{+\infty} t^{-2\sigma}

1138: \Bigl[ {\zeta ' \over \zeta} \bigl(\hf + t \bigr)+ {1 \over t - \hf} \Bigr]

1139: \d t ,

1140: \end{equation}

1141: however this last integral only converges in the strip 

1142: $\{0 < \Re \sigma < \hf \}$.

1143: 

1144: Remarks: 

1145: 

1146: - as analytical extension formulae, eqs.(\ref{ZC}--\ref{ZR2})

1147: are precise {\sl counterparts of the functional equation\/} for $\zeta(s)$;

1148: they also stand as more explicit forms of Guinand's functional relation 

1149: for $\Z(\sigma)$ \cite{G1}, as discussed below (eq.(\ref{GFE})).

1150: 

1151: - a similar formula exists for the function $\xi(s,x)$ of eq.(\ref{ZZet})

1152: (\cite{Dn}, middle of p.149), only requiring $\Re x >1$

1153: (which precisely avoids the problem raised above by the pole of $\zeta(s)$);

1154: in comparison, the present results correspond to the fixed value $x=\hf$, 

1155: since $\Z(\sigma) \equiv 

1156: (2\pi)^{-2\sigma}(2\cos \pi \sigma)^{-1} \xi(2\sigma,\hf)$ by eq.(\ref{Z2X});

1157: 

1158: \medskip

1159: 

1160: As we will elaborate next,

1161: analytical properties of $\Z(\sigma)$ in the half-plane $\{\Re \sigma < \hf \}$

1162: are made totally straightforward by the Mellin formulae (\ref{ZC}--\ref{ZR2})

1163: (while $\Z(\sigma)$ is holomorphic in the half-plane $\{\Re \sigma > \hf \}$, 

1164: where its defining series (\ref{Z2}) converges).

1165: Detailed results are also recollected in fully reduced form in Sec.7, Table 1.

1166: 

1167: \subsection{Properties of $\Z(\sigma)$ for $\Re \sigma < 1$}

1168: 

1169: A few leading properties of $\Z(\sigma)$ in the half-plane 

1170: $\{\Re \sigma < 1 \}$ emerge more easily as special cases from Sec.4

1171: (although they can be drawn from eq.(\ref{ZC}) as well):

1172: 

1173: - $\sigma =\hf$ is a {\sl double pole\/}, 

1174: with the {\sl same full polar part\/} as for $Z(\sigma)$, eq.(\ref{ZPP});

1175: 

1176: - by specializing eqs.(\ref{AA}),(\ref{DETG}),

1177: \begin{equation}

1178: \label{Z20}

1179: \textstyle

1180: \Z(0)=Z(0)={7 \over 8} ; \qquad 

1181: \Z'(0) = Z'(0) - \log \Xi \bigl( \hf \bigr) \ (\approx 0.811817944).

1182: \end{equation}

1183: 

1184: Otherwise, a Mellin representation like (\ref{ZC}) gives a better global view 

1185: of $\Z(\sigma)$ over the whole half-plane $\{\Re \sigma < \hf\}$.

1186: Indeed, its non-elementary part (the integral (\ref{JZ})) 

1187: becomes {\sl regular\/} there,

1188: hence can be {\sl ignored\/} both for the polar analysis and 

1189: (thanks to the $\sin \pi \sigma$ factor) 

1190: for the ``trace identities" at integer $\sigma$: 

1191: all such information lies then in the {\sl first term\/} alone,

1192: accessible by mere inspection. We thus obtain that:

1193: 

1194: - $\Z(\sigma)$ only has {\sl simple poles\/} 

1195: at the negative {\sl half\/}-integers $\sigma=\hf-m$, with residues

1196: \begin{equation}

1197: \label{Z2P}

1198: {\mathcal R}_m  = {(-1)^m \over 2 \pi} [{\bf Z}(1-2m) + 2^{1-2m}]

1199: \equiv {(-1)^m \over 8 \pi m} (1-2^{1-2m}) B_{2m} , \qquad  m=1,2,\ldots

1200: \end{equation}

1201: (hence, only the leading pole $\sigma=\hf$ stays double);

1202: 

1203: - at the negative integers, the {\sl ``trace identities"\/} read as 

1204: \begin{equation}

1205: \label{Z2T}

1206: \Z(-m) = {(-1)^m \over 2} [-{\bf Z}(-2m) + 2^{-2m}]

1207: \equiv (-1)^m 2^{-2m} (1 - {1 \over 8} E_{2m}),  \qquad m=0,1,\ldots

1208: \end{equation}

1209: (both formulae (\ref{Z2P}),(\ref{Z2T}) were fully reduced 

1210: using eq.(\ref{SVS1})).

1211: 

1212: - a ($\sigma \to -\infty$) {\sl asymptotic formula\/} follows 

1213: for $J_\zeta^\pm(\sigma)$,

1214: from the term-by-term substitution of the Euler product for $\zeta(s)$ 

1215: into the integrand of eq.(\ref{ZC}), giving

1216: \begin{equation}

1217: \label{ZAs}

1218: J_\zeta^\pm(\sigma) \sim 

1219: -\G(1-2\sigma) \sum_{n \ge 2} \Lambda(n) n^{-1/2} (\log n)^{2\sigma-1}, 

1220: \quad \sigma \to -\infty

1221: \end{equation}

1222: where as usual, $\Lambda(n) = \log p$ if $n=p^r$ for some prime $p$, else 0.

1223: (An asymptotic formula for $\Z(\sigma)$ itself then follows from

1224: eq.(\ref{ZC}) and the functional equations for $\zeta(s),\ \beta(s)$.)

1225: 

1226: Remark: in our notations, Guinand's functional relation for $\Z(\sigma)$

1227: \cite{G1} reads as

1228: \begin{equation}

1229: \label{GFE}

1230: \Z(\sigma) = 

1231: - {{\bf Z}(2\sigma) \over 2 \cos \pi \sigma}

1232: - {Z_{\rm p}(1-2 \sigma) \over 2 \G(2 \sigma) \cos \pi \sigma},

1233: \end{equation}

1234: where 

1235: $Z_{\rm p}(1-2 \sigma) \defi \!\!

1236:  \lim\limits_{T \to +\infty} \biggl\{ \displaystyle

1237: \sum_{2 \le n < \e^T} \!\!\!\! \Lambda(n) n^{-1/2} (\log n)^{2\sigma-1}

1238: - \int_0^T  \e^{x/2} x^{2\sigma-1} \d x \biggr\} $ 

1239: --- subject to the Riemann Hypothesis \cite{CHA} --- clearly specifies 

1240: a (real-valued) resummation of the {\sl divergent\/} series in eq.(\ref{ZAs})

1241: (the asymptotic series for $- J_\zeta^\pm(\sigma) /\G(1-2\sigma) $).

1242: Eq.(\ref{GFE}) was only asserted for $0< \Re \sigma < \qt$,

1243: with no clue as to the analytic structure 

1244: of either $\Z(\sigma)$ or $Z_{\rm p}(1-2 \sigma)$ elsewhere.

1245: The present formulae (\ref{ZC}--\ref{ZR2}) are thus 

1246: resummed versions of eq.(\ref{GFE}), with a definitely more explicit content.

1247: 

1248: \subsection{Properties of $\Z(\sigma)$ for $\Re \sigma > \hf$}

1249: 

1250: As stated before, $\Z(\sigma)$ is regular in the half-plane 

1251: $\{\Re \sigma >\hf\}$, where analytical results are identities directly

1252: obtainable by expanding the logarithm of the functional relation (\ref{FAC}) 

1253: in Taylor series around $t=0$. 

1254: Here we will extract those results from the Mellin representation (\ref{ZC}),

1255: invoking the meromorphic properties of its integral term in the whole plane 

1256: as given by eq.(\ref{JP}).

1257: 

1258: - for half-integer $\sigma=\hf +m$: 

1259: the residues of the two summands in (\ref{ZC}) have to cancel 

1260: given that $\Z(\sigma)$ is analytic in the half-plane; 

1261: this imposes 

1262: \begin{eqnarray}

1263: \label{ZID}

1264: (\log |\zeta|)^{(2m+1)} \bigl( \hf \bigr) \si=\sf

1265: (2m)!\ [ {\bf Z}(2m+1) + 2^{2m+1}],\quad  m=1,2,\ldots \\

1266: (\si=\sf -2^{-2m-1} (\log \G)^{(2m+1)} (\qt) \ ), \nonumber

1267: \end{eqnarray}

1268: which simply amounts to $(\log \Xi)^{(2m+1)}\bigl( \hf \bigr) = 0$

1269: (itself a consequence of the functional equation 

1270: $\Xi(\hf+t) \equiv \Xi(\hf-t)$);

1271: that result further reduces, using eq.(\ref{SVS2}), to the identity

1272: \begin{equation}

1273: \label{ZId}

1274: (\log |\zeta|)^{(2m+1)} \bigl( \hf \bigr) =

1275: \hf (2m)! \, (2^{2m+1}-1)  \zeta(2m+1) + \qt \pi^{2m+1} |E_{2m}| .

1276: \end{equation}

1277: The case $m=0$ is singular, but $(\log \Xi)'(\hf)=0$ directly yields

1278: \begin{equation}

1279: \label{ZI1}

1280: {\zeta' \over \zeta}(\hf) = \hf \Bigl[ \log \pi - (\log \G)' (\qt) \Bigr] =

1281: \hf \log 8\pi + {\pi \over 4} +{\g \over 2} \quad (\approx 2.68609171).

1282: \end{equation}

1283: 

1284: - for integer $\sigma=m$, the pole of the integral is cancelled 

1285: by the zero of $\sin \pi \sigma$, 

1286: and the following explicit relation results,

1287: \begin{equation}

1288: \label{Z2V}

1289: 2 (-1)^{m+1} \Z(m) 

1290: - {1 \over (2m-1)!} (\log |\zeta|)^{(2m)} \bigl( \hf \bigr) =

1291:  {\bf Z}(2m) - 2^{2m}, \quad  m=1,2,\ldots.

1292: \end{equation}

1293: which can also be further reduced with the help of eq.(\ref{SVS2}), see Table 1.

1294: 

1295: Unfortunately, we hardly know anything else about the values 

1296: $(\log |\zeta|)^{(n)} \bigl( \hf \bigr),\ n =0,1,\ldots$ 

1297: (cf. \cite{Ms1} for $\zeta(\hf)$).

1298: To supplement the relation (\ref{ZId}) with $\zeta(n)$ for $n$ odd,

1299: we can only refer to other formulae for $\zeta(2m+1)$ (compiled in \cite{BBC}),

1300: and to Euler--Maclaurin formulae for $\zeta^{(n)}(s)$

1301: (valid at $s=\hf$) with related numerical data \cite{A,CHO,BBC}.

1302: So, even at $v=0$, the transcendental values $\partial_\sigma \Z(0,v)$

1303: (eq.(\ref{Z20})) and $\Z(m,v)$ currently remain more elusive than at

1304: the (exceptional) point $v=\qt$ (Sec.3.3).

1305: Furthermore, we found no reference at all to those values 

1306: (i.e., $\Z'(0)$, $\Z(m)$) in the literature.

1307: 

1308: \subsection{Speculations and generalizations}

1309: 

1310: The results of Secs.5.2--4 for $\Z(\sigma)$ are similar to those yielded by

1311: the ``sectorial" trace formula for the analogous spectral zeta function

1312: $\Z_X(\sigma)$ over a compact hyperbolic surface $X$ \cite{CV,Vz}. 

1313: The present formulae for the Riemann case nevertheless show 

1314: several distinctive features.

1315: 

1316: \noindent - as announced end of Sec.2.1, the sequence $\{ {\tau_k}^2 \}$ 

1317: and the analogous spectrum of the Laplacian on $X$ 

1318: have mutually singular features:

1319: the former has the parameter values $\mu_0=\hf, \ r=2$ 

1320: ($\Z(\sigma)$ has its leading pole {\sl double\/}, at $\sigma=\hf$),

1321: whereas the latter more precisely has $\mu_0=1, \ r=1$ 

1322: ($\Z_X(\sigma)$ has all its poles {\sl simple\/}, starting at $\sigma=1$),

1323: hence this spectral analogy for the Riemann zeros holds only partially;

1324: 

1325: \noindent - in the continuation formulae (\ref{ZC}--\ref{ZR2}), 

1326: $\zeta (s)$ itself {\sl reenters\/} 

1327: as an {\sl additive\/} component of the shadow zeta function ${\bf Z}(s)$. 

1328: This is an altogether different incarnation of $\zeta (s)$ from its initial,

1329: {\sl multiplicative\/} involvement, 

1330: which remains in the integral term and indirectly through the zeros,

1331: in the left-hand side $\Z(\sigma)$. 

1332: It is curious to find two such copies of $\zeta (s)$ to coexist in one formula, 

1333: especially with the additive $\zeta (2 \sigma)$ represented 

1334: in its critical strip; 

1335: 

1336: %\noindent - formally, from eqs.(\ref{JZ},\ref{ZC}), the Riemann Hypothesis

1337: %would require $\sin \pi \sigma \, J_\zeta^\pm (\sigma)$ to define 

1338: %{\sl tempered\/} distributions on the vertical line $\{ \Re \sigma = \qt\}$.

1339: 

1340: \noindent - however, those formulae relative to $\zeta(s)$ 

1341: are not fully closed as they also invoke the other Dirichlet series

1342: $\beta (2 \sigma)$ 

1343: (as second additive component in the shadow zeta function ${\bf Z} (2 \sigma)$).

1344: The question then arises whether $\beta(s)$ and other zeta functions

1345: can be handled on the same footing as $\zeta(s)$ (as in \cite{K}),

1346: so we now outline a possible extension of eqs.(\ref{ZC}--\ref{ZR2}). 

1347: 

1348: \medskip

1349: 

1350: We assume that $\tilde \zeta (s)$ is a Dirichlet zeta or $L$-series, having:

1351: 

1352: \noindent - a single pole, at $s=1$ and of order $q$ (typically, $q=0$ or 1); 

1353: 

1354: \noindent - the asymptotic property $\log \tilde \zeta(s) = o(s^{-N})$ 

1355: for all $N\ (s \to +\infty)$;

1356: 

1357: \noindent - a functional equation of the form

1358: \begin{equation}

1359: \label{FE}

1360: \tilde {\D} (t^2) {\tilde {\bf D} (t) \over (1-2t)^q} = 

1361: \tilde \zeta(\hf + t), \quad \mbox{where}

1362: \end{equation}

1363:  

1364: \noindent - $\tilde {\D} (t^2)$ is an entire function of order $<1$ 

1365: {\sl in the variable\/} $t^2$, and

1366: 

1367: \noindent - $\tilde {\bf D} (t)$ is an entire function 

1368: with all of its zeros lying in the half-plane $\{ \Re t < 0 \}$.

1369: 

1370: Then the Zeta functions $\Z_{\tilde \zeta}(\sigma)$ 

1371: (for the zeros of $\tilde {\D}$) and

1372: ${\bf Z}_{\tilde \zeta}(\sigma)$ (for the zeros of $\tilde {\bf D}$) 

1373: are related by this formula corresponding to eq.(\ref{ZR2})

1374: (we omit the others),

1375: \begin{equation}

1376: \label{ZR}

1377: \Z_{\tilde \zeta}(\sigma) = 

1378: -{{\bf Z}_{\tilde \zeta} (2\sigma) \over 2 \cos \pi \sigma}

1379: + {\sin \pi \sigma \over \pi} \int_0^{+\infty} t^{-2\sigma}

1380: \Bigl[ {\tilde \zeta ' \over \tilde \zeta} \bigl(\hf + t \bigr)

1381: + {q \over t - \hf} \Bigr] \d t .

1382: \end{equation}

1383: 

1384: Apart from $ \zeta(s) $ itself, with eq.(\ref{ZR2}), 

1385: the next independent example is $\beta(s)$. 

1386: Its functional equation (\ref{xI}) has the form (\ref{FE}) with $q=0$

1387: (no pole) and $ \tilde {\D} (t^2) = \Xi_{\chi_4} (\hf + t) $,  

1388: $\tilde {\bf D} (t) = \bigl( {\pi \over 4} \bigr) ^{{3 \over 4} + {t \over 2}} 

1389: \big / \G \bigl( {3 \over 4}+ {t \over 2} \bigr) $;

1390: its spectrum of trivial zeros (for $\beta(\hf -t)$) is 

1391: $\{-\hf + 2k \}\quad ( = {3 \over 2},\ {7 \over 2}, \cdots)$,

1392: giving as shadow zeta function 

1393: \begin{equation}

1394: \label{SLD}

1395: {\bf Z}_\beta(\sigma)= \sum_{k=1}^\infty (-\hf +2 k)^{-s} \equiv 

1396: 2^s \Bigl[\hf  \bigl((1-2^{-s}) \zeta(s) -\beta(s)\bigr)  \Bigr] .

1397: \end{equation}

1398: Under $q=0$, all Mellin representations (\ref{ZC}--\ref{ZR2}) 

1399: coalesce into the single regular form

1400: \begin{equation}

1401: \label{LR}

1402: \Z_\beta(\sigma) = 

1403: -{{\bf Z}_\beta (2\sigma) \over 2 \cos \pi \sigma}

1404: + {\sin \pi \sigma \over \pi} \int_0^{+\infty} t^{-2\sigma}

1405: {\beta ' \over \beta} \bigl(\hf + t \bigr) \d t  \qquad (\Re \sigma < \hf),

1406: \end{equation}

1407: and all consequences previously drawn for $\Z(\sigma)$ 

1408: have analogs for $\Z_\beta(\sigma)$.

1409: 

1410: Various such integral representations will naturally {\sl add\/},

1411: whenever the initial zeta functions combine nicely under multiplication.

1412: For instance, eqs.(\ref{ZR2}) and (\ref{LR}) add up to:

1413: \begin{equation}

1414: \label{ZLR}

1415: (\Z_\zeta + \Z_\beta)(\sigma) = 

1416: -{ 2^{2\sigma} [(1-2^{-2\sigma}) \zeta (2\sigma) -1] \over 2 \cos \pi \sigma}

1417: + {\sin \pi \sigma \over \pi} \int_0^{+\infty} t^{-2\sigma}

1418: \Bigl[ \Bigl( {\zeta ' \over  \zeta} + {\beta ' \over  \beta} \Bigr)

1419: \bigl(\hf + t \bigr) + {1 \over t - \hf} \Bigr] \d t .

1420: \end{equation}

1421: Here, the shadow zeta function purely invokes $\zeta(s)$;

1422: on the other hand, under the integral sign we now find 

1423: $\bigl[ {(\beta \zeta)'  \over \beta \zeta} \bigr] \bigl( \hf +t \bigr) $

1424: so that the new multiplicative zeta function is $\beta(s) \zeta(s)$, 

1425: also recognized as ${1 \over 4}$ times $Z_4 (s)$, 

1426: the {\sl zeta function of the ring of Gaussian integers\/} $\mathbb Z [\mi]$

1427: \cite{C}. Hence eq.(\ref{ZLR}) becomes

1428: \begin{equation}

1429: \label{Z4R}

1430: \Z_{Z_4}(\sigma) = 

1431: -{ 2^{2\sigma} [(1-2^{-2\sigma}) \zeta (2\sigma) -1] \over 2 \cos \pi \sigma}

1432: + {\sin \pi \sigma \over \pi} \int_0^{+\infty} t^{-2\sigma}

1433: \Bigl[ {{Z_4}' \over Z_4} \bigl(\hf + t \bigr)

1434: + {1 \over t - \hf} \Bigr] \d t ;

1435: \end{equation}

1436: thus, to isolate $\zeta(s)$, here in the additive position, 

1437: we again had to allow a different zeta function elsewhere, 

1438: this time $Z_4(s)$ in the multiplicative position.

1439: 

1440: Likewise, by subtracting eq.(\ref{LR}) from (\ref{ZR2}) instead,

1441: we could get the shadow zeta function to be $\beta(s)$; then the counterpart

1442: of eqs.(\ref{ZId}),(\ref{ZI1}) is a {\sl fully explicit\/} identity, 

1443: \begin{equation}

1444: \label{ZB}

1445: (\log |\zeta|)^{(2m+1)} \bigl( \hf \bigr) -

1446:  (\log \beta)^{(2m+1)} \bigl( \hf \bigr) 

1447: = \hf \pi^{2m+1} |E_{2m}| + \delta_{m,0} \log 2, \quad m=0,1,\ldots ,

1448: \end{equation}

1449: whereas each of the two left-hand-side terms separately needs $\zeta(2m+1)$

1450: (or $\g$ for $m=0$).

1451: 

1452: 

1453: \section{More about the Hurwitz-type functions}

1454: 

1455: The purpose of this Section is twofold. 

1456: First, we analyze the Zeta functions 

1457: $\Z(\sigma,v) = \sum_k ({\tau_k}^2+v)^{-\sigma}$

1458: more explicitly over the half-plane $\{ \Re \sigma <1 \}$ than in Sec.4,

1459: by exploiting the latest special properties of the function $\Z(\sigma,0)$

1460: (with new results even for the case $v=\qt$).

1461: Then, by the same approach, we (briefly) discuss the other 

1462: Hurwitz-type Zeta functions ${\mathfrak Z}(\sigma,a)$ and $\xi(s,x)$,

1463: defined through eqs.(\ref{Hz}) and (\ref{ZZet}) respectively.

1464: 

1465: \subsection{Further properties of $\Z(\sigma,v)$ for $\Re \sigma <1$}

1466: 

1467: To describe the Hurwitz-type function $\Z(\sigma,v)$ better,

1468: we now systematically expand it in terms of $\Z(\sigma)$, as

1469: \begin{equation}

1470: \label{SHZ0}

1471: \Z(\sigma,v) = \sum_{k=0}^\infty 

1472: ({\tau_k}^2)^{-\sigma} \Bigl( 1 + {v \over {\tau_k}^2} \Bigr)^{-\sigma}

1473: = \sum_{\ell=0}^\infty {\G(1-\sigma) \over \ell !\G(1-\sigma-\ell)}

1474: \Z(\sigma+\ell) \, v^\ell \qquad (|v| < {\tau_1}^2).

1475: \end{equation}

1476: Such an expansion can be formulated around any reference point $v_0$, 

1477: but it will be specially useful for $v_0=0$ as above.

1478: For instance, coupled with eq.(\ref{ZC}) (say), it can express the meromorphic

1479: continuation of the general $\Z(\sigma,v)$ to $\{ \Re \sigma < \hf \}$, 

1480: while we lack an analog of eq.(\ref{ZC}) itself for any $v \ne 0$.

1481: 

1482: For the polar structure of $\Z(\sigma,v)$ at $\sigma=-m+\hf,\ m \in\mathbb N$, 

1483: the series (\ref{SHZ0}) reduces to

1484: \begin{equation}

1485: \label{HP}

1486: \Z(-m+\hf+\eps,v) = 

1487: \sum_{\ell=0}^m {\G(\hf+m-\eps) \over \ell ! \G(\hf+m-\ell-\eps)}

1488: \Z(-m+\ell+\hf+\eps) \, v^\ell

1489: \quad [\mbox{+regular part for } \eps \to 0];

1490: \end{equation}

1491: then, importing the polar structure of $\Z(\sigma)$ from eq.(\ref{Z2P}), we get

1492: \begin{equation}

1493: \label{Hp}

1494: \Z(-m+\hf+\eps,v) = {1 \over 8 \pi} {\G(m+\hf) \over m! \G(\hf)} v^m \,\eps^{-2}

1495:  + {\mathcal R}_m(v) \,\eps^{-1} + O(1) \qquad(\eps \to 0),

1496: \end{equation}

1497: just by brute-force polar expansion of the right-hand side of eq.(\ref{HP}). 

1498: Here, the polar part of order 2 at every $(-m+\hf)$ is clearly induced 

1499: by the only such part of $\Z(\sigma)$ (at $\sigma=\hf$),

1500: through the term with $\ell =m$ in eq.(\ref{HP}); whereas the residue 

1501: ${\mathcal R}_m(v)$ is built from all residues of $\Z(\sigma)$ 

1502: at poles with $\sigma \ge -m+\hf$, as

1503: \begin{equation}

1504: \label{HR}

1505: {\mathcal R}_m(v) = -{\G(m+\hf) \over m! \G(\hf)} 

1506: \Bigl[ {1 \over 4 \pi} \sum_{j=1}^m {1 \over 2j-1}

1507: + {\log 2\pi \over 4 \pi} \Bigr] v^m 

1508: + \sum_{j=1}^m{\G(\hf+m) \over (m-j)!\G(\hf+j)}

1509: {\mathcal R}_j \, v^{m-j} ,

1510: \end{equation}

1511: (the residues ${\mathcal R}_j$ of $\Z(\sigma)$ at $-j+\hf$

1512: are known from eq.(\ref{Z2P})).

1513: 

1514: \noindent Remark: for $m=0$, the full polar part (\ref{ZPP}) at $\sigma=\hf$,

1515: independent of $v$, is recovered.

1516: 

1517: When $\sigma \in -{\mathbb N}$, the series (\ref{SHZ0}) also terminates, as 

1518: \begin{equation}

1519: \label{TIG}

1520: \Z(-m,v) \equiv 

1521: \sum_{\ell=0}^m {m \choose \ell} \Z(-m+\ell)\,v^\ell \qquad (m \in \mathbb N) ,

1522: \end{equation}

1523: so that explicit ``trace identities" for general $v$ derive from 

1524: those for $v=0$ (eq.(\ref{Z2T})). (For $v=\qt$, this result simplifies further,

1525: see Table 1.)

1526: 

1527: Remark: in view of eqs.(\ref{ZP2}) and (\ref{ZT}), 

1528: the latter two results now imply a general-$n$ formula 

1529: for the coefficients $a_{(1-n)/2}(v)$ 

1530: in the large-$\lambda$ expansion (\ref{DAs}) 

1531: of $\log \D(\lambda \mid \{ {\tau_k}^2+v \})$.

1532: (Hitherto we had such a formula just at $v=0$, not even at $v=\qt$, 

1533: and knew only the {\sl other\/} coefficients $\tilde a_{(1-n)/2}(v)$ 

1534: for any $v$ and $n$, by eqs.(\ref{AA}--\ref{TAG}).)

1535: 

1536: Our initial emphasis on the special case $v=\qt$ might now seem misplaced:

1537: why didn't we operate at once from $v=0$~?

1538: In the first place, we saw the case $v=\qt$ arise more readily 

1539: from the standard product representation of $\zeta(s)$. 

1540: But mainly, the case $v=\qt$ also enjoys certain special properties, 

1541: this time with the values $Z'(0)$ and $Z(n)$ (Sec.3.3),

1542: and since these evolve from {\sl transcendental\/} functions of $v$ (Sec.4),

1543: their expansions (\ref{SHZ0}) around $v=0$ are now {\sl infinite\/}.

1544: So, each case $v=0$ and $v=\qt$ has its own exceptional features,

1545: the former in the half-plane $\{ \Re \sigma<1 \}$,

1546: and the latter for $\sigma \in \mathbb N$.

1547: 

1548: \subsection{The Hurwitz-type functions ${\mathfrak Z}(\sigma,a)$ and $\xi (s,x)$}

1549: 

1550: The function ${\mathfrak Z}(\sigma,a)$ as defined by eq.(\ref{Hz}) 

1551: is {\sl a priori\/} more singular than $\Z(\sigma,a)$ 

1552: (the sequence $\{ \tau_k \}$ itself has $r=2$ and $\mu_0=1$,

1553: which would require a formalism more elaborate than in Sec.2).

1554: Fortunately, ${\mathfrak Z}(\sigma,a)$ can also be analyzed directly

1555: through its expansion around 

1556: ${\mathfrak Z}(\sigma,0) \equiv \Z(\sigma)$, by analogy with eq.(\ref{SHZ0})

1557: (see also \cite{HKW}):

1558: \begin{equation}

1559: \label{SHZ1}

1560: {\mathfrak Z}(\sigma,a) = \sum_{k=0}^\infty 

1561: {\tau_k}^{-2\sigma} \Bigl( 1 + {a \over \tau_k} \Bigr)^{-2\sigma}

1562: = \sum_{\ell=0}^\infty {\G(1-2\sigma) \over \ell ! \G(1-2\sigma-\ell)}

1563: \Z(\sigma + \hf \ell) \, a^\ell  \quad (|a|<\tau_1).

1564: \end{equation}

1565: This formula generates a pole for ${\mathfrak Z}(\sigma,a)$

1566: now at every {\sl half-integer\/} $\hf(1-n),\ n \in\mathbb N$, according to:

1567: \begin{equation}

1568: \label{HP1}

1569: {\mathfrak Z}(\hf(1-n)+\eps,a) = 

1570: \sum_{\ell=0}^n {\G(n-2\eps) \over \ell ! \G(n-\ell-2\eps)}

1571: \Z(\hf(1-n+\ell)+\eps) \, a^\ell \quad + O(\eps) .

1572: \end{equation}

1573: Differences with eq.(\ref{HP}) arise due to the factor

1574: $\G(n-2\eps)/\G(n-\ell-2\eps)$ vanishing whenever $\ell \ge n >0$.

1575: Only the polar part at $\sigma=\hf$ remains the same as for $\Z(\sigma,v)$

1576: (of order $r=2$ and independent of $a$, given by eq.(\ref{ZPP})); 

1577: all other poles $\hf(1-n)$ of ${\mathfrak Z}(\sigma,a)$ are now {\sl simple\/}, 

1578: of residues

1579: \begin{equation}

1580: \label{RES1}

1581: r_n(a) = -{1 \over 4 \pi n} \, a^n +

1582: \sum_{0< 2m \le n} { n-1 \choose 2m-1 } {\mathcal R}_m \, a^{n-2m},

1583: \quad n=1,2,\ldots 

1584: \end{equation}

1585: (again, ${\mathcal R}_m$ is the residue given by eq.(\ref{Z2P})).

1586: In addition, at $\sigma=0$ the $\eps$-expansion of eq.(\ref{HP1}) 

1587: captures the {\sl finite part\/} too:

1588: \begin{equation}

1589: \label{RES0}

1590: r_1(a)=\Res_{\sigma=0} {\mathfrak Z}(\sigma,a) = -{a \over 4 \pi}; 

1591: \quad \mbox{finite part: }

1592: {\rm FP}_{\sigma=0} \,{\mathfrak Z}(\sigma,a) = 

1593: {7 \over 8} + { \log 2\pi \over 2\pi } \, a .

1594: \end{equation}

1595: 

1596: \medskip

1597: 

1598: As for the function $\xi(s,x)$,

1599: if we express it by eq.(\ref{Z2X}) in terms of ${\mathfrak Z}(\sigma,a)$,

1600: then we find this combination to be {\sl less singular\/} overall:

1601: by mere substitution of eq.(\ref{HP1}), 

1602: $\xi (s,x)$ shows a {\sl simple\/} pole at $s=1$, of residue $-\pi$ \cite{SS},

1603: and all other possible poles at $s=1-n,\ n=1,2,\ldots $ {\sl cancel out\/}, 

1604: resulting in the holomorphy of $\xi (s,x)$ for all $s \ne 1$ 

1605: with the computable finite values (``trace identities")

1606: \begin{equation}

1607: \label{TI1}

1608: \xi (1-n,\hf+y) = {2 \over (2\pi)^{n-1}} \Biggl[ -\pi { r_n( \mi y) \over \mi^n}

1609: + \sum_{0 \le 2m < n} (-1)^m { n-1 \choose 2m } \Z(-m) y^{n-2m-1} \Biggr] ,

1610: \quad n=1,2,\ldots 

1611: \end{equation}

1612: (An alternative evaluation follows from Deninger's continuation formula 

1613: for $\xi(s,x)$ (\cite{Dn}, middle of p.149), as

1614: $\xi (1-n,\hf+y) = (2\pi)^{1-n} 

1615: \bigl[ (y+\hf)^{n-1} + (y-\hf)^{n-1} + 2^{n-1} B_n(\qt+ \hf y) \bigr] $,

1616: whose agreement with eq.(\ref{TI1}) can be verified.)

1617: 

1618: As for special values: first, $\partial_s \xi (s,x)_{s=0}$ 

1619: is expressible as well, in terms of $\zeta(x)$ \cite{Dn,SS}:

1620: \begin{eqnarray}

1621: \label{DZet1}

1622: \partial_s \xi (s,x)_{s=0} \si=\sf \log \,2^{1/2} (2\pi)^2 - \log \Xi(x) \\

1623: \label{DZet2}

1624: \Longleftrightarrow \qquad 

1625: -\partial_s \bigl[ \sum_\rho (x-\rho)^{-s} \bigr]_{s=0}

1626: \si=\sf \log \Xi(x) + \hf (\log 2 \pi) \, x - \hf \log 4 \pi

1627: \end{eqnarray}

1628: (the equivalence of the two forms follows from eqs.(\ref{ZZet}) 

1629: and (\ref{TI1}) for $n=1$, i.e., $\xi (0,x) = \hf (x+3)$).

1630: Now, the exponentiated left-hand side of eq.(\ref{DZet2}) precisely defines  

1631: the zeta-regularized product $\tilde {\D}_{\rm zr}(x)$

1632: built upon the sequence $\{\rho\}$ of Riemann zeros,

1633: while the right-hand side mainly involves $\Xi(x)$ of eq.(\ref{CHAD}).

1634: So, eq.(\ref{DZet2}) is converting a zeta-regularized product

1635: ($\tilde {\D}_{\rm zr}(x)$) to Hadamard product form.

1636: As an aside, we now verify that such a conversion formula is entirely 

1637: {\sl fixed by universal rules\/} for (complex) {\sl admissible sequences\/},

1638: specialized here to $r=1$ (as in \cite{V,QHS}) and $\mu_0=1$ ---

1639: since the Zeta functions $\xi(s,x)$ have just a simple pole at $s=1$.

1640: Those rules yield these two prescriptions: 

1641: $\log \tilde {\D}_{\rm zr}(x) \equiv \log \Xi(x)-(\alpha x+\beta)$,

1642: {\sl and\/} the large-$x$ expansion of $\log \tilde {\D}_{\rm zr}(x)$

1643: shall only retain {\sl canonical\/} (or {\sl standard\/}) terms, namely: 

1644: $c_\mu x^\mu$ for $1>\mu \notin \mathbb N,\ c_1 x(\log x -1),\ c_0 \log x$.

1645: Those conditions together fix $(\alpha,\beta)$ uniquely, and here,

1646: eqs.(\ref{CONV}--\ref{Z1as}) for $\log \Xi(x)$ as input 

1647: precisely lead to eq.(\ref{DZet2}) as output.

1648: 

1649: Likewise, the special values $\xi (n,x),\ n=1,2,\ldots $ are expressible

1650: in terms of [the higher Laurent coefficients of] $\zeta(x)$, e.g.,

1651: by applying residue calculus to Deninger's continuation formula 

1652: (\cite{Dn}, p.149). 

1653: 

1654: Thus, a fair degree of structural parallelism finally shows up

1655: between the two Hurwitz-like families $\Z(s,v)$ and $\xi (s,x)$ \cite{VZ}.

1656: 

1657: \section{Recapitulation of main results}

1658: 

1659: In way of conclusion, Table 1 collates the analytical results found

1660: for the two $\zeta$-Zeta functions $ Z(\sigma) \ (=\Z(\sigma,v=\qt))$ 

1661: and $ \Z(\sigma) \ (=\Z(\sigma,0))$. Furthermore, corresponding results 

1662: for the general $\Z(\sigma,v)$ were derived in Sec.4 (for $\Re \sigma \ge 0$) 

1663: and 6.1 (for $\Re \sigma < 1$), 

1664: and partly extended to the functions ${\mathfrak Z}(\sigma,a)$ and $\xi (s,x)$

1665: in Sec.6.2.

1666: 

1667: \begin{table}

1668: \begin{tabular}  {ccc}

1669: \hline \\[-12pt]

1670: $\sigma$ & 

1671: $Z(\sigma) = \sum\limits_{k=1}^\infty ({\tau_k}^2+\qt)^{-\sigma} \quad [v=\qt]$

1672: & $\Z(\sigma) = \sum\limits_{k=1}^\infty {\tau_k}^{-2\sigma} \quad [v=0]$

1673: \\[10pt]

1674: \hline \\[-12pt]

1675: $-m$ & $ (-1)^{m+1} 2^{-2m-3} 

1676: \sum\limits_{\ell=0}^m {m \choose \ell} (-1)^{\ell}  {E_{2(m-\ell)}}

1677: ^{\ {\scriptsize(\ref{TIG},\ref{Z2T})}} $ & 

1678: $ (-1)^m 2^{-2m} (1 - {1 \over 8} E_{2m})\ ^{{\scriptsize(\ref{Z2T})}} $ \\[10pt]

1679: $-m+\hf +\eps$ & 

1680: $ \!\! [ {2^{-2m} \over 8 \pi} {\G(m+1/2) \over m! \G(1/2)}] \,\eps^{-2}

1681:  + {\mathcal R}_m(\qt) \, \eps^{-1} +O (1) \, ^{{\scriptsize(\ref{Hp},\ref{HR})}} $ & 

1682: $ [{(-1)^m \over 8 \pi m} (1-2^{1-2m}) B_{2m}] \,\eps^{-1} + O(1) 

1683: \ ^{{\scriptsize(\ref{Z2P})}} $ \\[3pt]

1684: $\vdots$ & $\vdots$ & $\vdots$ \\[3pt]

1685: $-1$ & $-1/16$ & $-9/32$ \\[6pt]

1686: $-\hf +\eps$ & 

1687: ${1 \over 64 \pi} \,\eps^{-2} - [{3 \log 2 \pi + 4 \over 96 \pi}] \,\eps^{-1} 

1688: +O(1)$ &

1689: $-{1 \over 96 \pi} \,\eps^{-1} +O(1)$ \\[6pt]

1690: $0$ & $7/8 \ ^{{\scriptsize(\ref{Z10})}} $ & $7/8 \ ^{{\scriptsize(\ref{Z20})}} $ \\[3pt]

1691: \hline \\[-12pt]

1692: ${\mbox{\sl derivative} \atop \mbox{\sl at 0}}$ & 

1693: $Z'(0) = \qt \log 8\pi \ ^{{\scriptsize(\ref{Z10})}} $ &  

1694: $\Z'(0) = \log \bigl[ 2^{11/4} \pi^{1/2} \G (\qt)^{-1} |\zeta (\hf)|^{-1} \bigr]

1695: \ ^{{\scriptsize(\ref{Z20})}} $ \\[6pt]

1696: $+\hf +\eps$ & 

1697: ${1 \over 8 \pi} \,\eps^{-2} - {\log 2 \pi \over 4 \pi} \,\eps^{-1} + O(1)

1698: \ ^{{\scriptsize(\ref{ZPP})}} $ & 

1699: ${1 \over 8 \pi} \,\eps^{-2} - {\log 2 \pi \over 4 \pi} \,\eps^{-1} + O(1)

1700: \ ^{{\scriptsize(\ref{ZPP},\ref{AA})}} $ \\[6pt]

1701: $+1$ & $-\hf \log 4\pi + 1 + \hf \g \equiv {\mathscr Z}_1 \ ^{{\scriptsize(\ref{Z11})}} $ & 

1702: $ \hf (\log |\zeta|)''(\hf) + {1 \over 8}\pi^2 +\beta(2) -4$ \\[4pt]

1703: $+2$ & 

1704: $\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!

1705: \biggl\{ \! {\textstyle -\log 4\pi+3+\g -(\log |\zeta|)''(0)-{1 \over 24} \pi^2

1706: \atop \textstyle \!\!\!\!\!\!\!\!\!\!\!

1707: -\log 4\pi+3+\g + \, 2\g_1 + \g^2 -{1 \over 8}\pi^2 } \biggr\} $ &

1708: %^{{\scriptsize(\ref{ZRH},\ref{Z123})}} $ &

1709: $ -{1 \over 12} (\log |\zeta|)^{(4)} (\hf)  - {1 \over 24}\pi^4 

1710: - 4 \beta(4) + 16 $ \\

1711: $\vdots$ & $\vdots$ & $\vdots$ \\

1712: $m$ & $ \sum\limits_{\ell=0}^{m-1} {m+\ell-1 \choose m-1} {{\mathscr Z}_{m-\ell}} 

1713: ^{\ {\scriptsize(\ref{SZ})}}$ &

1714: $\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!

1715: \matrix{ (-1)^m \Bigl\{ -{1 \over 2(2m-1)!} (\log |\zeta|)^{(2m)} (\hf) \cr

1716: \qquad\qquad\qquad \quad

1717: {} - \qt \bigl[ (2^{2m}-1) \zeta (2m) + 2^{2m} \beta (2m) \bigr] \cr

1718: \ \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad 

1719: {} + 2^{2m} \Bigr\} ^{{\scriptsize(\ref{Z2V})}} } $ \\[-13pt]

1720:  & \multicolumn{2}{l}{ $\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!

1721: \Bigl[ {\mathscr Z}_n \equiv \Biggl\{ {\!\textstyle 

1722: 1 - (-1)^n 2^{-n} \zeta(n) - {(\log |\zeta|)^{(n)}(0) \over (n-1)!}

1723: \atop \textstyle \!

1724: 1 - (1-2^{-n}) \zeta(n) + { n \over (n-1)!} \g^{\rm c}_{n-1} } \Biggr\}

1725: \; (n=2,3,\ldots) \Bigr] ^{\, {\scriptsize(\ref{ZRH})}} $ } \\[12pt]

1726: \hline

1727: \end{tabular}

1728: \caption{Analytical results for $\zeta$-Zeta functions of the Riemann zeros.

1729: Notations: see eqs.(\ref{ZER}),(\ref{NOT}), (\ref{LDef}) [for $\beta(s)$],

1730: (\ref{LAU}) [for $\g^{\rm c}_n$]; ${\mathscr Z}_m \equiv \sum_\rho \rho^{-m}$;

1731: $m$ stands for any {\sl positive\/} integer, and $\eps \to 0$. 

1732: As an indexing tool,

1733: the superscripts refer to the relevant equation numbers in the main text.

1734: [In the very last formula (for $\Z(m)$), one can still use 

1735: eq.(\ref{SVR}) for $\zeta(2m)$]. 

1736: }

1737: \end{table}

1738: 

1739: The other novel results we have developed here concern $\Z(\sigma)$

1740: in the half-plane $\{ \Re \sigma < \hf \}$:

1741: the analytical continuation formulae (\ref{ZC}--\ref{ZR2}),

1742: and the $\sigma \to -\infty$ asymptotic formula (\ref{ZAs}) as corollary.

1743: We also came across two elementary (but unfamiliar to us) formulae 

1744: concerning $\zeta(s)$ itself: eqs.(\ref{ZId}),(\ref{ZB}).

1745: 

1746: Appendix B gives information on some numerical aspects 

1747: based on our use of the 100,000 first Riemann zeros 

1748: (made freely available on the Web by A.M. Odlyzko \cite{O}, 

1749: to whom we express our gratitude). 

1750: We also wish to thank C. Deninger, J.P. Keating, P. Leb\oe uf, V. Maillot,

1751: C. Soul\'e, and the Referee, 

1752: for helpful references and comments.

1753: 

1754: 

1755: \section*{Appendix A: Meromorphic Mellin transforms}

1756: 

1757: We briefly recall the meromorphic continuation argument for a Mellin transform

1758: like eq.(\ref{IMel}), 

1759: $I(\sigma) \defi \int_0^{+\infty} L(z) z^{-\sigma-1} \d z$,

1760: assuming the function $L(z)$ to be regular on $\mathbb R^+$ (for simplicity), with 

1761: \begin{equation}

1762: \label{LAs}

1763: L(z)=O(z^{\nu_0})  \quad (z \to 0^+), \mbox{ and} \quad

1764: L(z) \sim \sum_{n=0}^\infty 

1765: (\tilde a_{\mu_n} \log z + a_{\mu_n})  z^{\mu_n} \quad (z \to +\infty)

1766: \end{equation}

1767: as in eq.(\ref{DAs}) (asymptotic estimates are repeatedly differentiable); 

1768: and crucially, $\mu_0 < \nu_0$.

1769: 

1770: \medskip

1771: 

1772: Sequential directed integrations by parts can be used 

1773: (see \cite{J,V,C} for details).

1774: 

1775: Step 1: $I(\sigma)$ converges for $\mu_0 < \Re z < \nu_0$, and in that strip,

1776: \begin{equation}

1777: \label{LI1}

1778: I(\sigma) =  \int_0^{+\infty} [L(z) z^{-\mu_0}]' 

1779: { z^{\mu_0 -\sigma} \over \sigma-\mu_0 } \, \d z .

1780: \end{equation}

1781: If $\tilde a_{\mu_0}=0$, this suffices:

1782: the new integral actually converges for  $\mu_1 < \Re z < \nu_0$

1783: (thanks to $z^{\mu_0 +1}[L(z) z^{-\mu_0}]' =O( z^{\mu_1} \log z)$ 

1784: for $z \to \infty$),

1785: hence $I(\sigma)$ is manifestly meromorphic in that wider strip, 

1786: with a simple pole at $\sigma=\mu_0$ of residue 

1787: \begin{equation}

1788: \int_0^{+\infty} [L(z) z^{-\mu_0}]' \, \d z = a_{\mu_0} ;

1789: \end{equation}

1790: furthermore, in the complementary strip $\{ \mu_1 < \Re z < \mu_0 \}$,

1791: backward integration by parts now yields

1792: \begin{equation}

1793: I(\sigma) = \int_0^{+\infty} [L(z) - a_{\mu_0}z^{\mu_0})] z^{-\sigma-1} \d z .

1794: \end{equation}

1795: Then the whole argument can be restarted from here, 

1796: to extend $I(\sigma)$ further (across $\{ \Re \sigma = \mu_1 \}$), and so on:

1797: the case $r=1$ thus gets settled.

1798: 

1799: Step 2: if $\tilde a_{\mu_0} \ne 0$, 

1800: one more integration by parts upon eq.(\ref{LI1}) yields

1801: \begin{equation}

1802: I(\sigma) = \int_0^{+\infty} [z [L(z) z^{-\mu_0}]' ]'

1803: {z^{\mu_0 -\sigma} \over (\sigma-\mu_0)^2}\,\d z \qquad (\mu_1 < \Re z < \nu_0).

1804: \end{equation}

1805: All previous arguments then carry over, but the pole is now {\sl double\/}, with

1806: \begin{equation}

1807: \mbox{principal polar cefficient:} \quad 

1808: \int_0^{+\infty} [z [L(z) z^{-\mu_0}]' ]' \, \d z = \tilde a_{\mu_0} ,

1809: \end{equation}

1810: and residue $-\int_0^{+\infty} [z [L(z) z^{-\mu_0}]' ]' \log z \, \d z$

1811: (from the residue calculus);

1812: integrations by parts (backwards, and split) reduce the latter to

1813: \begin{equation}

1814: \mbox{residue:} \quad \int_0^1  [L(z) z^{-\mu_0}]' \, \d z + 

1815: \int_1^{+\infty}  \{ z[L(z) z^{-\mu_0}]' - \tilde a_{\mu_0} \}{1 \over z}\, \d z

1816: = a_{\mu_0} .

1817: \end{equation}

1818: The last two formulae thus generate eq.(\ref{IP2}) for the leading double pole,

1819: and so on for $r=2$.

1820: 

1821: (More generally, if the factor of $z^{\mu_n}$ in the expansion (\ref{LAs})

1822: is a polynomial of degree $p_n$ in $\log z$, 

1823: then $\mu_n$ becomes a pole of order $(p_n+1)$ for $I(\sigma)$ \cite{J}.)

1824: 

1825: \medskip

1826: 

1827: As for meromorphic continuation in the other direction 

1828: (across $\{ \Re \sigma = \nu_0 \}$),

1829: it works likewise if $L(z)$ admits a $z \to 0$ expansion:

1830: the previous arguments apply upon exchanging the bounds 0 and $+\infty$

1831: under $\sigma \mapsto -\sigma$.

1832: E.g., in the regular case, $L(z)$ expands in an entire series at $z=0$, 

1833: and step 1 suffices (as in the main text, where $\nu_0=1$).

1834: 

1835: 

1836: \section*{Appendix B: Numerical aspects}

1837: 

1838: We complete our analytical study by describing some {\sl very heuristic\/}

1839: numerical work with $Z(\sigma )$ and $\Z(\sigma )$,

1840: mostly in the range $\{ \sigma \ge 0 \}$, 

1841: and focusing on the simpler case of $\Z(\sigma )$.

1842: (The same ideas apply for any generalized Zeta function $\Z(\sigma,v)$

1843: and for complex $\sigma$, but the formulae get more involved.)

1844: Here, the Riemann Hypothesis is {\sl de facto\/} implied throughout

1845: (there being no numerical counter-example).

1846: 

1847: Numerically, $Z(\sigma)$ looks almost indistinguishable 

1848: from $\Z(\sigma)$ for $\sigma \ge 0$,

1849: because ${\tau_k}^2 \gg {1 \over 4} \ (\forall k)$

1850: (an empirical fact; already, ${\tau_1}^2 \approx 199.790455$).

1851: If we expand $Z(\sigma)$ in terms of $\Z(\sigma)$ according to

1852: eq.(\ref{SHZ0}), and make the roughest approximations, we get that

1853: $Z(\sigma) \approx \Z(\sigma) [1- \sigma / (4 {\tau_1}^2) ]$: i.e., 

1854: the very first correction term is only of relative size $\approx \sigma/800$.

1855: As other related numerical observations:

1856: \begin{equation}

1857: 0 < \Z(\sigma) -Z(\sigma) < 0.0003 \mbox{ for all real }\sigma > 0 ;

1858: \end{equation}

1859: \begin{equation}

1860: A \defi 4[Z'(0)-\Z'(0)] \approx -0.0231003495 \quad vs \quad 

1861: B \equiv -Z(1)  \approx -0.0230957090 

1862: \end{equation}

1863: (not only is $A$ small, but moreover, 

1864: $A \equiv 4 \log \Xi(\hf)$ by eq.(\ref{Z20}) 

1865: and $B \equiv [\log \Xi]'(0)$ by eq.(\ref{Z11}),

1866: hence $|B-A|<5 \times 10^{-6}$ reflects how {\sl little\/}

1867: the function $\log \Xi(s)$ deviates from the parabolic shape $As(1-s)$ 

1868: over the interval $[0,1]$). 

1869: 

1870: \medskip

1871: 

1872: We now focus on the numerical evaluation of $\Z(\sigma)$ itself.

1873: The defining series (\ref{Z2}) converges more and more poorly 

1874: as $ \sigma \to \hf^+$ (with divergence setting in at $\sigma =\hf$). 

1875: We then replace a far tail of that series ($k>K$) by an integral 

1876: according to the integrated-density estimate (\ref{GE}),

1877: and formally obtain a kind of Euler--Maclaurin formula,

1878: \begin{eqnarray}

1879: \label{EMc}

1880: \Z (\sigma ) \si=\sf \lim_{K \to +\infty} S_K (\sigma ), \qquad S_K (\sigma ) \defi

1881: \sum_1^{K-1} {\tau_k}^{-2 \sigma }+ \hf {\tau_K}^{-2 \sigma } 

1882: + {\overline R}_K (\sigma ), \\

1883: {\overline R}_K (\sigma ) \si\defi\sf \int_{\tau_K}^{+\infty} T^{-2 \sigma }

1884: \d {\overline \mathcal N}(T) 

1885: = {1 \over 2 \pi} {{\tau_K}^{1-2 \sigma } \over 2 \sigma -1} 

1886: \Bigl[ \log {\tau_K \over 2 \pi}+{1 \over 2 \sigma -1}  \Bigr] \nonumber

1887: \end{eqnarray}

1888: (similar formulae can be written for $\Z'(\sigma)$, $\Z(\sigma,v)$,

1889: finite parts at $\sigma=\hf$, etc.).

1890: 

1891: \begin{table}

1892: \begin{tabular}  {ccc}

1893: \hline \\[-12pt]

1894: $\sigma$ & ~~ $Z(\sigma) = \sum\limits_{k=1}^\infty ({\tau_k}^2+\qt)^{-\sigma}

1895: \quad [v=\qt]~~ $ &

1896: ~~ $\Z(\sigma) = \sum\limits_{k=1}^\infty {\tau_k}^{-2\sigma}

1897: \quad [v=0] ~~ $ \\[10pt]

1898: \hline \\[-10pt]

1899: $-1$ & $-0.0625^\ast$ & $-0.28125^\ast$ \\

1900: $-3/4$ & $1.69388$ & $0.54319$ \\

1901: $- 1/4$ & $0.800805$ & $0.785321$  \\

1902: $0$ & $0.875^\ast$ & $0.875^\ast$ \\

1903: \hline \\[-10pt]

1904: {\sl derivative at 0} & {\sl 0.8060429} & {\sl 0.8118179} \\[3pt]

1905: $+1/4$ & $1.548829$ & $1.549060$ \\

1906: {\sl finite part at 1/2} & {\sl 0.251546} & {\sl 0.251637} \\

1907: $+3/4$ & $0.247730$ & $0.247760$ \\

1908: $+1$ & $0.0230957$ &  $0.0231050$ \\

1909: $+3/2$ &  $0.0007287$ &  $0.0007295$ \\

1910: $+2$ & $0.0000371$ & $0.0000372$ \\

1911: \hline

1912: \end{tabular}

1913: \caption{Numerical values for $\zeta$-Zeta functions of the Riemann zeros

1914: ($^\ast$: exact values).

1915: Implied precision is expected to hold, but not guaranteed.

1916: }

1917: \end{table}

1918: 

1919: The approximate remainder term $\overline R_K(\sigma )$ balances 

1920: the dominant {\sl trend\/} of the partial sums

1921: $\sum\limits_{k<K}{\tau_k}^{-2\sigma }$.

1922: It thereby accelerates the convergence of the partial sums in eq.(\ref{EMc}) 

1923: for $\sigma > \hf$, 

1924: while for $\sigma \le \hf$ it counters their divergent trend,

1925: so that $S_K (\sigma )$ converges (as $K \to +\infty$) when $\sigma >0$ 

1926: (\cite{G1}, p.116, last line).

1927: The next obstruction to convergence arises at $\sigma=0$ but is of another type:

1928: $S_K(\sigma ) $ displays {\sl erratic fluctuations\/} in $K$

1929: ({\sl roughly\/} of the order ${\tau_K}^{-2 \sigma } (\log \log \tau_K)^{1/2}$,

1930: according to \cite{O}, eq.(2.5.7)),

1931: and those numerically blow up indeed (as $K \to +\infty$) when $\sigma \le 0$.

1932: Further convergence now requires to perform a damping of those fluctuations 

1933: (as argued previously for ``chaotic" spectra \cite{BSV}).

1934: Here, a {\sl Cesaro averaging\/} (defined by 

1935: $\langle S \rangle_K \defi K^{-1} \sum\limits_1 ^K S_{K'} $) 

1936: {\sl appears\/} to work well initially (results can be verified at $\sigma=0$), 

1937: but not very far down:

1938: already at $\sigma = -0.25$, the fluctuations of $\langle S \rangle_K(\sigma)$

1939: itself retain a standard deviation $> 10^{-3}$ up to $K \approx 10^5$.

1940: So, instead of pursuing ever more severe (and unproven, after all) 

1941: numerical regularizations as $\sigma$ decreases below $\hf$,

1942: we advocate the switch to the continuation formulae (\ref{ZC}--\ref{ZR1}) 

1943: for numerical work as well. 

1944: Thus, we first tested eq.(\ref{ZR1}) against eq.(\ref{EMc}) for $\Z(+\qt)$, 

1945: then used it to evaluate $\Z(-\qt)$, 

1946: plus eq.(\ref{SHZ0}) with $v=\qt$ (3 terms sufficed) to obtain $Z(-\qt)$.

1947: 

1948: Table 2 gives a summary of the numerical results we obtained.

1949: (We found no earlier analogs, except for the {\sl other\/} special sums

1950: ${\mathscr Z}_n$ in \cite{Ms2}, \cite{L}.)

1951: 

1952: 

1953: \begin{thebibliography}{99}

1954: 

1955: \bibitem{AS} M. Abramowitz and I.A. Stegun, 

1956: {\it Handbook of Mathematical Functions\/}, chap. 23, Dover, New York (1965).

1957: 

1958: \bibitem{A} T.M. Apostol, {\it Formulas for Higher Derivatives of the

1959: Riemann Zeta Function\/}, Math. Comput. {\bf 44} (1985) 223--232.

1960: 

1961: \bibitem{BSV} N.L. Balazs, C. Schmit and A. Voros, 

1962: {\it Spectral Fluctuations and Zeta Functions\/},

1963: J. Stat. Phys. {\bf 46} (1987) 1067--1090.

1964: 

1965: \bibitem{BK} M.V. Berry and J.P. Keating,

1966: {\it A new asymptotic representation for $\zeta(\hf+\mi t)$ and quantum 

1967: spectral determinants\/}, Proc. R. Soc. Lond. {\bf A437} (1992) 151--173.

1968: 

1969: \bibitem{BBC} J.M. Borwein, D.M. Bradley, R.E. Crandall,

1970: {\it Computational strategies for the Riemann Zeta Function\/}, 

1971: J. Comput. Appl. Math. {\bf 121} (2000) 247--296, and refs. therein.

1972: 

1973: \bibitem{C} P. Cartier, {\it An Introduction to Zeta Functions\/}, 

1974: in: {\it From Number Theory to Physics\/},

1975: M. Waldschmidt, P. Moussa,  J.-M. Luck and C. Itzykson eds., 

1976: Springer-Verlag (1992) 1--63.

1977: 

1978: \bibitem{CV} P. Cartier and A. Voros, 

1979: {\it Une nouvelle interpr\'etation de la formule des traces de Selberg\/},

1980: C. R. Acad. Sci. Paris {\bf 307}, S\'erie I (1988) 143--148,

1981: and in: {\it The Grothendieck Festschrift\/} (vol. 2), eds. P. Cartier et al., 

1982: Progress in Mathematics, Birkh\"auser (1990) 1--67.

1983: 

1984: \bibitem{CHA} I.C. Chakravarty, {\it The Secondary Zeta-functions\/},

1985: J. Math. Anal. Appl. {\bf 30} (1970) 280--294.

1986: 

1987: \bibitem{CHO} B.K. Choudhury, {\it The Riemann zeta-function and its

1988: derivatives\/}, Proc. R. Soc. Lond. {\bf A450} (1995) 477--499,

1989: and refs. therein.

1990: 

1991: \bibitem{CR} H. Cram\'er, {\it Studien \"uber die Nullstellen der Riemannschen

1992: Zetafunktion\/}, Math. Z. {\bf 4} (1919) 104--130.

1993: 

1994: \bibitem{Da} H. Davenport, {\it Multiplicative Number Theory\/} 

1995: (3rd ed., revised by H.L. Montgomery), Graduate Texts in Mathematics {\bf 74},

1996: Springer-Verlag (2000).

1997: 

1998: \bibitem{Dl} J. Delsarte, {\it Formules de Poisson avec reste\/},

1999: J. Anal. Math. (Jerusalem) {\bf 17} (1966) 419--431 (Sec.7).

2000: 

2001: \bibitem{Dn} C. Deninger, 

2002: {\it Local L-factors of motives and regularized determinants\/},

2003: Invent. Math. {\bf 107} (1992) 135--150 (thm 3.3 and Sec.4).

2004: 

2005: \bibitem{E} H.M. Edwards, {\it Riemann's Zeta Function\/}, 

2006: Academic Press (1974).

2007: 

2008: \bibitem{B} A. Erd\'elyi (ed.), {\it Higher Transcendental Functions\/}

2009: (Bateman Manuscript Project), vols. I chap. 1 and III chap. 17, 

2010: McGraw-Hill, New York (1953).

2011: 

2012: \bibitem{F} A. Fujii, 

2013: {\it The zeros of the Riemann zeta function and Gibbs's phenomenon\/},

2014: Comment. Math. Univ. St. Paul. (Japan) {\bf 32} (1983) 229--248.

2015: 

2016: \bibitem{G1} A.P. Guinand, 

2017: {\it A summation formula in the theory of prime numbers\/},

2018: Proc. London Math. Soc. Series 2, {\bf 50} (1949) 107--119 (Sec. 4(A)).

2019: 

2020: \bibitem{G2} A.P. Guinand, 

2021: {\it Fourier reciprocities and the Riemann zeta-function\/},

2022: Proc. London Math. Soc. Series 2, {\bf 51} (1950) 401--414.

2023:  

2024: \bibitem{H} D.A. Hejhal, 

2025: {\it The Selberg trace formula and the Riemann zeta function\/},

2026: Duke Math. J. {\bf  43} (1976) 441--481.

2027: 

2028: \bibitem{I} G. Illies, {\it Regularized products and determinants\/},

2029: Commun. Math. Phys. {\bf 220} (2001) 69--94.

2030: 

2031: \bibitem{J} P. Jeanquartier, 

2032: {\it Transformation de Mellin et d\'eveloppements asymptotiques\/},

2033: Enseign. Math. II. Ser., {\bf 25} (1979) 285--308.

2034: 

2035: \bibitem{JL1} J. Jorgenson and S. Lang, 

2036: {\it Basic analysis of regularized series and products\/},

2037: Lecture Notes in Mathematics {\bf 1564}, Springer-Verlag (1993).

2038: 

2039: \bibitem{JL2} J. Jorgenson and S. Lang, 

2040: {\it Explicit formulas for regularized products and series\/},

2041: Lecture Notes in Mathematics {\bf 1593}, Springer-Verlag (1994), 

2042: and refs. therein.

2043: 

2044: \bibitem{K} N. Kurokawa, {\it Parabolic components of zeta functions\/},

2045: Proc. Japan Acad. {\bf 64}, Ser. A (1988) 21--24, and 

2046: {\it Special values of Selberg zeta functions\/}, in:

2047: {\it Algebraic K-theory and algebraic number theory\/} 

2048: (Proceedings, Honolulu 1987), M.R. Stein and R. Keith Dennis eds.,

2049: Contemp. Math. {\bf 83}, Amer. Math. Soc. (1989) 133--149.

2050: 

2051: \bibitem{Lb} P. Leb\oe uf, {\it Prime correlations and their fluctuations\/},

2052: Ann. Henri Poincar\'e {\bf 4} (2003) 635--660 ??

2053: (Special Proceedings Issue, TH-2002 Conference, Paris, July 2002).

2054: 

2055: \bibitem{L} D.H. Lehmer, 

2056: {\it The Sum of Like Powers of the Zeros of the Riemann Zeta Function\/}, 

2057: Math. Comput. {\bf 50} (1988) 265--273, and refs. therein.

2058: 

2059: \bibitem{MA} Yu.V. Matiyasevich, 

2060: {\it A relationship between certain sums over trivial and nontrivial 

2061: zeros of the Riemann zeta-function\/}, Mat. Zametki {\bf 45} (1989) 65--70 

2062: [Translation: Math. Notes (Acad. Sci. USSR) {\bf 45} (1989) 131--135.]

2063: 

2064: \bibitem{Ms1} Y. Matsuoka, 

2065: {\it On the values of the Riemann zeta function at half integers\/},

2066: Tokyo J. Math. {\bf 2} (1979) 371--377.

2067: 

2068: \bibitem{Ms2} Y. Matsuoka, 

2069: {\it A note on the relation between generalized Euler constants 

2070: and the zeros of the Riemann zeta function\/}, 

2071: J. Fac. Educ. Shinshu Univ. {\bf 53} (1985) 81--82,

2072: and {\it A sequence associated with the zeros of the Riemann zeta function\/}, 

2073: Tsukuba J. Math. {\bf 10} (1986) 249--254.

2074: 

2075: \bibitem{O} A.M. Odlyzko, {\it The $10^{20}$-th zero of the Riemann zeta 

2076: function and 175 million of its neighbors\/}, AT \& T report (1992) 

2077: unpublished, available on Web site {\tt http://www.research.att.com/\~{}amo/}.

2078: 

2079: \bibitem{QHS} J.R. Quine, S.H. Heydari and R.Y. Song, 

2080: {\it Zeta regularized products\/},

2081: Trans. Amer. Math. Soc. {\bf 338} (1993) 213--231.

2082: 

2083: \bibitem{R} B. Randol, {\it On the analytic continuation of the

2084: Minakshisundaram--Pleijel zeta function for compact Riemann surfaces\/},

2085: Trans. Amer. Math. Soc. {\bf 201} (1975) 241--246.

2086: 

2087: \bibitem{RS} M. Rubinstein and P. Sarnak, {\it Chebyshev's bias\/},

2088: Exp. Math. {\bf 3} (1994) 173--197.

2089: 

2090: \bibitem{SS} M. Schr\"oter and C. Soul\'e,

2091: {\it On a result of Deninger concerning Riemann's zeta function\/}, in: 

2092: {\it Motives\/}, Proc. Symp. Pure Math. {\bf 55} Part 1 (1994) 745--747.

2093: 

2094: \bibitem{S} F. Steiner, 

2095: {\it On Selberg's zeta function for compact Riemann surfaces\/},

2096: Phys. Lett. {\bf B 188} (1987) 447--454.

2097: 

2098: \bibitem{T} E.C. Titchmarsh, {\it The Theory of the Riemann Zeta-function\/}

2099: (2nd ed., revised by D.R. Heath-Brown), Oxford Univ. Press (1986).

2100: 

2101: \bibitem{V} A. Voros, 

2102: {\it Spectral functions, special functions and the Selberg zeta function\/},

2103: Commun. Math. Phys. {\bf 110} (1987) 439--465.

2104: 

2105: \bibitem{Vz} A. Voros, 

2106: {\it Spectral zeta functions\/}, in: 

2107: {\it Zeta Functions in Geometry\/} (Proceedings, Tokyo 1990),

2108: N. Kurokawa and T. Sunada eds., Advanced Studies in Pure Mathematics {\bf 21},

2109: Math. Soc. Japan (Kinokuniya, Tokyo, 1992) 327--358. 

2110: 

2111: \bibitem{BL} {\sl (added in proof)\/} E. Bombieri and J.C. Lagarias, 

2112: {\sl Complements to Li's Criterion for the Riemann Hypothesis\/},

2113: J. Number Theory {\bf 77} (1999) 274--287

2114: [the Stieltjes constants $\gamma_n$ are normalized differently therein].

2115: 

2116: \bibitem{VZ} {\sl (added in proof)\/} A. Voros,

2117: {\sl More zeta functions for the Riemann zeros\/}, Saclay preprint T03/078

2118: (June 2003), in: {\sl Frontiers in Number Theory, Physics and Geometry\/}

2119: (Proceedings, Les Houches, March 2003), B. Julia, P. Moussa and P. Vanhove eds.,

2120: Lecture Notes in Physics, Springer-Verlag (to be published).

2121: 

2122: \bibitem{HKW} {\sl (added in proof)\/} M. Hirano, N. Kurokawa and M. Wakayama,

2123: {\sl Half zeta functions\/}, J. Ramanujan Math. Soc. {\bf 18} (2003) 195--209. 

2124: [We understand that their polar parts for our function $\mathfrak Z$ should not

2125: have terms with $\g$ either (N. Kurokawa, private communication).]

2126: 

2127: \end{thebibliography}

2128: 

2129: 

2130: \end{document}

2131: