We know the Dirichlet $\eta$-function is defined as the analytic continuation of $$\eta(s) = \sum_{n=1}^\infty \frac{(-1)^{n-1}}{n^s} \quad \Re(s)>0$$ I find an identity for the values of this function at negative integers: $$\eta(-n)+ \sum_{k=0}^n {n\choose k}(-1)^k \eta(k-n) =0.$$ where $1<n\in\mathbb{Z}$.

I found this using some "computation" on divergent series. But this wrong computation cannot serve as a proof.

I think this identity should be true, though I haven't found a simple proof.

Do you know anything about this identity? If it's true, how to prove it? Is there any generalization (for arbitrary $s$ rather than $-n$) of this identity? Could you provide me some reference?

  • 2
    $\begingroup$ See the two answers for this question. $\endgroup$ Jan 16 '14 at 19:41
  • $\begingroup$ I haven't checked it, but you can quite easily derive such formulas for negative values via the functional equation of eta. $\endgroup$ Jan 16 '14 at 21:08
  • $\begingroup$ Since my answer was no proof but just a visualization I've just deleted it. $\endgroup$ Jan 26 '14 at 14:25

1) Bernoulli numbers, $(n=0,1,2,...; B^n:=B_n)$, (DLMF = http://dlmf.nist.gov/24)

\begin{align*} & B = [1, -1/2, 1/6, 0, -1/30, 0, ...] \\ &\iff e^{Bx} = x/(e^x-1) \\ &\iff e^{(B+1)x} - e^{Bx} = x \\ &\iff (B+1)^{n+1} - B^{n+1} = \delta_{n,0} \qquad{(\rm DLMF\; 24.4.1,}\,x=0) \\ &\iff \cosh(Bx) = (x/2)\coth(x/2) \quad\&\quad \sinh(Bx) = -x/2 \end{align*}

2) Bernoulli polynomials at $x=1/2$, $(n=0,1,2,...; B^n:=B_n)$ \begin{align*} & e^{(B+1/2)x} = e^{Bx} (e^{x/2}+1) - e^{Bx} = 2 (x/2)/(e^{x/2}-1) - x/(e^x-1) \\ &\iff (B+1/2)^n = (2^{1-n}-1) B_n \qquad{(\rm DLMF\; 24.4.27}) \\ &\iff (2B+1)^n = (2-2^n) B_n \end{align*}

3) Dirichlet's $\eta(s)$ function and Riemann's $\zeta(s)$ function, $(n=0,1,2,...)$

\begin{align*} \eta(-n) &= (1-2^{n+1}) \zeta(-n) \\ \zeta(-n) &= (-1)^n B_{n+1}/(n+1)\qquad({\rm DLMF\; 25.6.3, avec}\;n=0^*) \end{align*} *Jonathan Sondow, Analytic continuation of Riemann's zeta function and values at negative integers via Euler's transformation of series http://www.ams.org/journals/proc/1994-120-02/S0002-9939-1994-1172954-7

4) Récurrence pour $\eta(-n), (n=0,1,2,...; B^n:=B_n)$ \begin{align*} &\;(-1)^n (n+1) \left[ \;\eta(-n) + \sum_{k=0}^n {n\choose k} (-1)^{n-k} \eta(-k) \;\right] \\ &= (1-2^{n+1}) B_{n+1} + \sum_{k=0}^n {n\choose k} (n+1)/(k+1) (1-2^{k+1}) B_{k+1} \\ &= (1-2^{n+1}) B_{n+1} + \sum_{k=0}^n {n+1\choose k+1} (1-2^{k+1}) B_{k+1} \\ &= (1-2^{n+1}) B_{n+1} + \sum_{k=0}^{n+1} {n+1\choose k} (1-2^k) B_k \\ &= \left[(2-2^{n+1}) B_{n+1} - (2B+1)^{n+1}\right] + \left[(B+1)^{n+1} - B^{n+1}\right] \\ &= 0 + \delta_{n,0} \\ &= (-1)^n (n+1) \left[\; \delta_{n,0} \;\right] \qquad{\rm CQFD} \end{align*} Jacques Gélinas

  • $\begingroup$ (For those unfamiliar with what's going on: Look up the umbral calculus.) $\endgroup$ Oct 20 '14 at 2:53
  • $\begingroup$ I don't understand this line (Google Translate: Je ne comprends pas cette ligne): $e^{(B+1/2)x} = e^{Bx} (e^{x/2}+1\; -1) = 2 (x/2)/(e^{x/2}-1) - x/(e^x-1)$ $\endgroup$ Oct 20 '14 at 3:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.