Correct application of Tannery's Theorem to Show Riemann Zeta $\zeta(s) \rightarrow 1$ as $\sigma \rightarrow \infty$?

143 Views Asked by At

The Riemann zeta function is described by the series $\zeta(s)=\sum 1/n^s$ for $\sigma>1$ where $s=\sigma+it$.

To show that the $\zeta(s)\rightarrow1$ as $\sigma\rightarrow\infty$ we might be tempted to use

$$\zeta(s)=\frac{1}{1^{s}}+\frac{1}{2^{s}}+\frac{1}{3^{s}}+\frac{1}{4^{s}}+\ldots$$

and say that as $\sigma\rightarrow\infty$, the magnitude of each term $|n^{-s}|=n^{-\sigma}$ tends to zero as $\sigma\rightarrow\infty$ for all n except $n=1$.

However - this assumes $\lim_{\sigma\rightarrow\infty}\sum 1/n^s = \sum \lim_{\sigma\rightarrow\infty} 1n^s$ which we haven't shown is true.

To do this we can use Tannery's Theorem (wiki, other).

Question: This question is about checking I have applied Tannery's theorem correctly.


Step 1

We start by stating Tannery's theorem to make sure we understand it. We start by defining:

$$S_j = \sum_{k}^{\infty} f_k(j) \;  \text{ converges for all } j$$ $$ \lim_{j \rightarrow \infty} f_k(j) = f_k \; \text{ exists} $$

If the following bounds apply,

$$| f_k(j) | \leq M_k$$ $$ \sum_{k=0}^{\infty} M_k < \infty $$

Then we can say,

$$ \lim_{j \rightarrow \infty} S_j= \sum_{k=0}^{\infty} f_k $$


Step 2

We cast our challenge, and take care with the changed variable names.

We are interested in

$$ \lim_{\sigma\rightarrow\infty} \sum_n 1/n^s = \lim_{\sigma\rightarrow\infty} \sum_n f_n(\sigma) $$

where $f_n(\sigma) = 1/n^s = 1/n^{\sigma+it}$.

We then establish the preliminary conditions:

$$\sum_{n}^{\infty} f_n(\sigma) \;  \text{ converges for all } \sigma>1$$

We are only interested in $\sigma>1$.

$$ \lim_{\sigma \rightarrow \infty} f_n(\sigma) = f_n \; \text{ exists} $$

Here:

$$ f_n = \begin{cases} 0 \text{ for } n>1\\ \\1 \text{ for } n=0 \end{cases} $$


Step 3

We then check the bounds:

$$| f_n(\sigma) | \leq M_n = 1/n^\sigma \text {converges for } \sigma > 1$$

$$ \sum M_n < \infty \text { for } \sigma >1$$


Step 4

Which finally allows us to say

$$ \lim_{\sigma \rightarrow \infty} \sum_n 1/n^s = \sum_{n} f_n = \sum 1+0+0+0 \ldots = 1 $$

1

There are 1 best solutions below

3
On BEST ANSWER

This is close. There's really just one issue you need to fix: $M_n$ cannot depend on $\sigma$.

A simple way to fix this is to assume $\sigma \geq 2$ and then define $M_n = 1/n^2$.