I have asked this question at math stack exchange, however it did not get any traction. Still curious about the answer though.

Numerical evidence suggests that:

$$\lim_{N \to +\infty} \sum_{n=1}^N\frac{1}{n}\sum_{k=0}^n(-1)^k\ {n \choose k}\ \frac{1}{(k+1)^{s}}=s$$

or equivalently

$$\lim_{N \to +\infty} H\left( N \right)+\sum _{n=1}^{N} \left( { \frac {1}{n}\sum _{k=1}^{n}{\left( -1 \right) ^{k}{n\choose k}\frac { 1 }{ \left( k+1 \right) ^{s}}}} \right) = s$$

with $H(N)$ = the $N$-th Harmonic Number.

Convergence is quite slow, but clearly goes faster for negative $s$. Also, the computations for non-integer values of $s$ require high accuracy settings (I have used Maple, pari/gp and ARB).

However, according to Mathematica the series diverges by the "Harmonic series test", although when taking $s$ as an integer, it agrees on convergence.

Does this series converge for $s \in \mathbb{C}$ ?

Some numerical results below:

0.497702121, N = 100
0.499804053, N = 1000
0.499905919, N = 2000

-3.14160222, N = 100
-3.14159284, N = 1000
-3.14159272, N = 2000

2.45310498 - 1.94063637i, N = 100
2.33501943 - 2.09308517i, N = 1000
2.31996958 - 2.09923503i, N = 2000

We have to show that $$l(s):=\sum_{n=1}^\infty\frac1n\,S_n(s)=s,$$ where $$S_n(s):=\sum_{k=0}^n(-1)^k\binom nk\frac1{(k+1)^s} \\ =\sum_{k=0}^n(-1)^k\binom nk\int_0^\infty du\,u^{s-1}e^{-(k+1)u}/\Gamma(s) \\ =\int_0^\infty du\,u^{s-1}e^{-u}\sum_{k=0}^n(-e^{-u})^k\binom nk/\Gamma(s) \\ =\int_0^\infty du\,u^{s-1}e^{-u}(1-e^{-u})^n/\Gamma(s).$$ So, $$l(s)=\int_0^\infty du\,u^{s-1}e^{-u}\sum_{n=1}^\infty\frac1n\,(1-e^{-u})^n/\Gamma(s) \\ =\int_0^\infty du\,u^s e^{-u}/\Gamma(s) =\Gamma(s+1)/\Gamma(s) =s,$$ as desired.

  • $\begingroup$ very nice! So, the Mathematica divergence test wasn't correct after all. $\endgroup$ – Agno May 5 '20 at 17:08

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.