Here's the limit.


Can we prove that this doesn't converge for any $z$ where $\Im(z)\ne0$?

  • $\begingroup$ The convergence In C even now is not clear or less obvious at all $\endgroup$ – zeraoulia rafik Jan 6 '17 at 22:07

This is just a note.

There is a Riemann sum involved here: $$\lim_{n\rightarrow\infty}R_n(z):=\lim_{n\rightarrow\infty}\,\,\frac1n\sum_{j=1}^n\sqrt\frac{n}{j}\sin\left(z\log{\frac{n}{j}}\right) =-\int_0^1\frac{\sin(z\log x)}{\sqrt{x}}\,dx=\frac{2z}{1+4z^2};$$ valid (at least) for $z\in\mathbb{R}$. In view of this, the OP's question amounts to: how fast does the Riemann sums converge to the integral with respect to $n$; that is, $$R_n(z)-\int_0^1\frac{-\sin(z\log x)}{\sqrt{x}}\,dx=O(??) \tag1$$ Some numerical evidence suggests that (even for real $z\neq0$) the order of convergence is weaker than $\frac1n$, which means the OP's limit does not converge.


The sum can be carried out in closed form in terms of harmonic numbers:

$$\frac{1}{n}\sum _{j=1}^n \sqrt{\frac{n}{j}} \sin \left(z \log \left(\frac{n}{j}\right)\right)=-\frac{i}{2 \sqrt{n}}\left(n^{i z} H_n^{\left(i z+\frac{1}{2}\right)}-n^{-i z} H_n^{\left(\frac{1}{2}-i z\right)}\right)$$

Now we can expand the harmonic numbers at $n=\infty$. The leading terms are:

$$\frac{4 z}{4 z^2+1}+\frac{1}{2} i \sqrt{\frac{1}{n}} \left(n^{-i z} \zeta \left(\frac{1}{2}-i z\right)-n^{i z} \zeta \left(\frac{1}{2}+i z\right)\right)$$


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.