6

Let $a_1,a_2,\ldots\in\mathbb{C}$ and consider the Dirichlet series $\sum_{n=1}^\infty \dfrac{a_n}{n^z}$. Suppose the series converges for some $z_0$. Then why does it converge uniformly on every closed disk in the half-space $\Re z>\Re z_0$, where $\Re$ denotes the real part of a complex number?

I think it's reasonable that when the real part of $z$ increases, the denominator of $\dfrac{a_n}{n^z}$ increases (well, in some sense), so the convergence gets better. How can we actually prove it though?

JJ Beck
  • 2,696
  • 17
  • 36

2 Answers2

13

For simplifying notation, let $$S_k=\sum_{m\geq k}\frac{a_n}{n^{z_0}}$$ Now, we do the classical Abel's trick and for all $N,M\in \mathbb{N}$ with $N<M$, we obtain \begin{align*} \sum_{n=N}^M\frac{a_n}{n^z}&=\sum_{n=N}^M\frac{a_n}{n^{z_0}}\frac{1}{n^{z-z_0}}\\ &=\sum_{n=N}^M\frac{S_{n+1}-S_{n}}{n^{z-z_0}}\\ &=\sum_{n=N}^M\frac{S_{n+1}}{n^{z-z_0}}-\sum_{n=N}^M\frac{S_{n}}{n^{z-z_0}}\\ &=\sum_{n=N+1}^{M}S_n\left(\frac{1}{(n-1)^{z-z_0}}-\frac{1}{n^{z-z_0}}\right)+\frac{S_{M+1}}{M^{z-z_0}}-\frac{S_N}{N^{z-z_0}} \end{align*} after rutinary operations with each sumatory. Here, we have to observe that \begin{align*} \left|\frac{1}{(n-1)^{z-z_0}}-\frac{1}{n^{z-z_0}}\right|=\left|\int_{n-1}^{n}\frac{z-z_0}{x^{z-z_0+1}}\,dx\right|&\\\leq |z-z_0| \int_{n-1}^{n}\frac{dx}{x^{\text{Re}(z-z_0)+1}}&=\frac{|z-z_0|}{\text{Re}(z-z_0)}\left(\frac{1}{(n-1)^{\text{Re}(z-z_0)}}-\frac{1}{n^{\text{Re}(z-z_0)}}\right) \end{align*} under the assumption $\text{Re}(z-z_0)>0$. Therefore for $\text{Re}(z-z_0)>0$ and $M>N$ we obtain $$\left|\sum_{n=N}^M\frac{a_n}{n^z}\right|\leq \frac{\max_{N\leq n\leq M+1}|S_n|}{N^{\text{Re}(z-z_0)}}\left(2+\frac{|z-z_0|}{\text{Re}(z-z_0)}\right)$$ after seeing that \begin{align*} \left|\sum_{n=N}^M\frac{a_n}{n^z}\right|&=\sum_{n=N+1}^{M}|S_n|\left|\frac{1}{(n-1)^{z-z_0}}-\frac{1}{n^{z-z_0}}\right|+\frac{|S_{M+1}|}{M^{\text{Re}(z-z_0)}}+\frac{|S_N|}{N^{\text{Re}(z-z_0)}}\\ &\leq \left(\max_{N<n\leq M}|S_n|\right)\sum_{n=N+1}^{M}\left|\frac{1}{(n-1)^{z-z_0}}-\frac{1}{n^{z-z_0}}\right|+\frac{|S_{M+1}|}{M^{\text{Re}(z-z_0)}}+\frac{|S_N|}{N^{\text{Re}(z-z_0)}}\text{;} \end{align*} \begin{align*} \sum_{n=N+1}^{M}\left|\frac{1}{(n-1)^{z-z_0}}-\frac{1}{n^{z-z_0}}\right| &\leq \frac{|z-z_0|}{\text{Re}(z-z_0)}\sum_{n=N+1}^{M}\left(\frac{1}{(n-1)^{\text{Re}(z-z_0)}}-\frac{1}{n^{\text{Re}(z-z_0)}}\right)\\ &\leq \frac{|z-z_0|}{\text{Re}(z-z_0)}\left(\frac{1}{N^{\text{Re}(z-z_0)}}-\frac{1}{M^{\text{Re}(z-z_0)}}\right)\\ &\leq \frac{|z-z_0|}{\text{Re}(z-z_0)}\frac{1}{N^{\text{Re}(z-z_0)}}\text{,} \end{align*} since $\sum_{n=N+1}^{M}\left(\frac{1}{(n-1)^{\text{Re}(z-z_0)}}-\frac{1}{n^{\text{Re}(z-z_0)}}\right)$ is a telescoping series; and that \begin{align*} \left(\max_{N<n\leq M}|S_n|\right)\frac{|z-z_0|}{\text{Re}(z-z_0)}\frac{1}{N^{\text{Re}(z-z_0)}}&+\frac{|S_{M+1}|}{M^{\text{Re}(z-z_0)}}+\frac{|S_N|}{N^{\text{Re}(z-z_0)}}\\&\leq \frac{\max_{N\leq n\leq M+1}|S_n|}{N^{\text{Re}(z-z_0)}}\left(2+\frac{|z-z_0|}{\text{Re}(z-z_0)}\right)\text{,} \end{align*} due to $N<M$.

Now, we are ready for the final step, but before three remarks:

  1. Since $\text{Re}(z-z_0)>0$, $$\frac{1}{N^{\text{Re}(z-z_0)}}<1\text{.}$$
  2. Since every closed ball $B$ is compact and if contained in the considered region, the function $$z\mapsto \frac{|z-z_0|}{\text{Re}(z-z_0)}$$ is well defined and continuous over it, there is a constant $K_B$ such that for all $z\in B$, $$\frac{|z-z_0|}{\text{Re}(z-z_0)}\leq K_B\text{.}$$ Note that here the important part is that the considered function is bounded, for which we can take far bigger domains -e.g., $A(z_0,\theta)= \{z\in \mathbb{C}\,|\,\text{Re}(z)>\text{Re}(z_0)\text{ and }\text{arg}(z-z_0)\leq \theta\}$ for $\theta<\frac{\pi}{2}$-.
  3. Since $\sum \frac{a_n}{n^{z_0}}$ converges, $S_K$ converges for zero and thus for all $\varepsilon>0$, there is $C_\varepsilon>0$ such that $$|S_n|<\varepsilon$$ whenever $n\geq C_\varepsilon$.

Joining this three remarks and the previous formula, we obtain that given a close ball $B$ contained in the half plane $\text{Re}(z)>\text{Re}(z_0)$ and $\varepsilon>0$, for all $z\in B$ and $N,M\geq C_{\varepsilon/(2+K_B)}$, $$\left|\sum_{n=N}^M\frac{a_n}{n^z}\right|<\varepsilon\text{.}$$ Thus $\sum_{n=1}^\infty\frac{a_n}{n^z}$ is uniformly Cauchy in $B$ and therefore converges uniformly in B as desired.

  • In Remark 2, if we let Im(z-z0) nonzero and Re(z-z0) tend to 0, wouldn't the image of the map go to infinity? – Divide1918 Nov 24 '21 at 11:55
  • Also, what's S_K? – Divide1918 Nov 24 '21 at 11:55
  • @Divide1918 Regarding the first, note that |z-z_0|/Re(z-z_0)=1/\cos(arg(z-z_0)), so letting both Im(z-z_0) and Re(z-z_0) is not enough so that the quotient goes to infinity, as long as we can guarantee that |arg(z-z_0)| is bounded away from \pi/2. Regarding the second, S_K is S_k. Error in typing. – Josué Tonelli-Cueto Nov 24 '21 at 15:14
0

Apply what you know about complex series:

$$\text{Re}\,z>\text{Re}\,z_0\implies \sum_{n=1}^\infty\left|\frac{a_n}{n^z}\right|=\sum_{n=1}^\infty\frac{|a_n|}{n^{\text{Re}\,z}}\le\sum_{n=1}^\infty\frac{|a_n|}{n^{\text{Re}\,z_0}}=\sum_{n=1}^\infty\left|\frac{a_n}{n^{z_0}}\right|$$

and since the rightmost series converges then also ours does

DonAntonio
  • 211,718
  • 17
  • 136
  • 287
  • Is it true that if a complex series $\sum_{n=1}^\infty a_n$ converges, then $\sum_{n=1}^\infty |a_n|$ converges? Sorry if this is something obvious. – JJ Beck Nov 30 '13 at 16:43
  • No, of course not @JJBeck... that's false even for real series! – DonAntonio Nov 30 '13 at 16:44
  • So how can you say "since the rightmost series converges"? – JJ Beck Nov 30 '13 at 16:45
  • Oh, now I see you did not mention "absolute convergence" in your post. Sorry, I thought you were talking of absolute convergence (for analytic functions and stuff). – DonAntonio Nov 30 '13 at 16:47
  • That's all right.. please fix it if you can :) – JJ Beck Nov 30 '13 at 16:48
  • Ok, I tried to find as simple a proof as possible, but it is going to be a little work anyway. If you know/remember Dirichlet's Test for convergence of series, Abel's summation and stuff, then you'll have an edge with the following: http://www.proofwiki.org/wiki/Dirichlet_Series_Convergence_Lemma – DonAntonio Nov 30 '13 at 17:11