Zeta-function derivation of Euler-MacLaurin sum rules (Q1820193)

The author considers generalizations of the Euler-MacLaurin summation formula \[ \sum^{M}_{m=1}f(m)-\int^{M}_{0}f(x) dx=\sum^{\infty}_{k=0}(1/k!) [f^{(k)}(0)-f^{(k)}(M)]\zeta (-k). \] Here \(\zeta\) (s) is the Riemann zeta-function. Usually the values \(\zeta\) (-k) are expressed by the Bernoulli numbers. Unfortunately it is not stated in this paper for which class of functions this formula is correct, because the author justifies an essential commutation of two summations only by referring to some of his previous results. On the other hand it seems to be inappropriate to work only with functions f which have convergent Taylor series. The announced generalizations consist in replacing f(m) by \((-1)^{m+1} f(m)\) and \(\zeta\) (s) \(by(1- 2^{1-s})\zeta (s)\) on the other side, and some similar changes. The proofs are sometimes only sketched or purely formal, so when the divergent series \(\sum^{\infty}_{m=1}m^ k\), \(k\in {\mathbb{N}}\), is replaced by \(\zeta\) (-k).