18
$\begingroup$

I mean the Laurent series at $s=1$.

I want to do it by proving $\displaystyle \int_0^\infty \frac{2t}{(t^2+1)(e^{\pi t}+1)} dt = \ln 2 - \gamma$,

based on the integral formula given in Wikipedia. But I cannot solve this integral except by using Mathematica. Tried complex analytic ways but no luck. Any suggestions? Thanks for your attention!

$\endgroup$
16
$\begingroup$

We can show that $$ \zeta(s)=\frac1{1 - 2^{1-s}}\sum_{n=1}^\infty\frac{(-1)^{n-1}}{n^s}\tag{1} $$ converges for $\mathrm{Re}(s)>0$ by applying the Generalized Dirichlet Test. To apply the test, we need to show that the partial sum of $(-1)^{n-1}$ is bounded, $|n^{-s}|\to0$, and $$ \sum_{n=1}^\infty\left|n^{-s}-(n+1)^{-s}\right|\tag{2} $$ converges.

When $s\in\mathbb{R}$, $n^{-s}$ head straight from $1$ to $0$, so the sum in $(2)$ is $1$. However, if $s\not\in\mathbb{R}$, then $n^{-s}$ spirals into $0$, and it is not immediately obvious that that spiral has finite length.

Let's look at how $n^{-s}$ spirals into $0$:

$\hspace{6pt}$(a)$\hspace{6pt}$ $\arg(n^{-s}) = -\log(n)\mathrm{Im}(s)$

$\hspace{6pt}$(b)$\hspace{6pt}$ $|n^{-s}| = n^{-\mathrm{Re}(s)} = e^{-\log(n)\mathrm{Re}(s)}$

Thus, $n^{-s}$ lies on the spiral $r = e^{t\theta}$ where the constant $t = \mathrm{Re}(s)/\mathrm{Im}(s)$. The length of this curve from $r=1$ to $r=0$ is easily computed to be $|s|/\mathrm{Re}(s)$. Thus, the total variation of $n^{-s}$, as given in $(2)$, is bounded by $|s|/\mathrm{Re}(s)$, and therefore, the sum in $(1)$ converges.


Consider $(1)$ to first order in $s-1$. $$ \frac1{1 - 2^{1-s}}=\frac1{s-1}\frac1{\log(2)}+\frac12+O(s-1)\tag{3} $$ and $$ \sum_{n=1}^\infty\frac{(-1)^{n-1}}{n^s}=\log(2)+(s-1)\sum_{n=1}^\infty(-1)^n\frac{\log(n)}{n}+O\left((s-1)^2\right)\tag{4} $$ Therefore, $$ \zeta(s)=\frac1{s-1}+\frac{\log(2)}{2}+\frac1{\log(2)}\sum_{n=1}^\infty(-1)^n\frac{\log(n)}{n}+O(s-1)\tag{5} $$ Next, we can use the Euler-Maclaurin Sum Formula to compute $$ \sum_{k=1}^n\frac{\log(k)}{k}=C+\frac{\log(n)^2}{2}+O\left(\frac{\log(n)}{n}\right)\tag{6} $$ and $$ \sum_{k=1}^n\frac1{k}=\log(n)+\gamma+O\left(\frac1n\right)\tag{7} $$ Then, we can use $(6)$ and $(7)$ to get $$ \begin{align} \sum_{n=1}^\infty(-1)^n\frac{\log(n)}{n} &=-\lim_{n\to\infty}\left(\sum_{k=1}^{2n}\frac{\log(k)}{k}-2\sum_{k=1}^n\frac{\log(2k)}{2k}\right)\\ &=\small-\lim_{n\to\infty}\left(\left(C+\frac{\log(2n)^2}{2}\right)-\left(C+\frac{\log(n)^2}{2}\right)-\log(2)(\log(n)+\gamma)\right)\\ &=\gamma\log(2)-\frac{\log(2)^2}{2}\tag{8} \end{align} $$ Combining $(5)$ and $(8)$ yields $$ \zeta(s)=\frac1{s-1}+\gamma+O(s-1)\tag{9} $$

$\endgroup$
  • $\begingroup$ from here, how does one proceed in showing that $\dfrac{\zeta'(s)}{\zeta(s)} = \dfrac{1}{1-s} + \gamma + O(s-1)$? Thanks! $\endgroup$ – Eric Jul 17 '13 at 18:56
  • 2
    $\begingroup$ @Eric: $(9)$ gives $$ \begin{align} \log(\zeta(s)) &=\log\left(\frac1{s-1}+\gamma+O(s-1)\right)\\ &=-\log(s-1)+\log\left(1+\gamma(s-1)+O(s-1)^2\right)\\ &=-\log(s-1)+\gamma(s-1)+O(s-1)^2 \end{align} $$ With proper assumptions on the smoothness of $\zeta$, we can differentiate this to give $$ \frac{\zeta'(s)}{\zeta(s)}=\frac1{1-s}+\gamma+O(s-1) $$ $\endgroup$ – robjohn Jul 17 '13 at 20:34
  • $\begingroup$ Of course! Please excuse my lack of creativity. =| Thanks for this!!! $\endgroup$ – Eric Jul 17 '13 at 23:19
  • 1
    $\begingroup$ @Eric: $(4)$ is the Taylor expansion at $s=1$. The constant term is the value at $s=1$: $$\sum_{n=1}^\infty\frac{(-1)^{n-1}}{n}=\log(2)$$ The $(s-1)$ term is the first derivative at $s=1$: $$\sum_{n=1}^\infty\frac{(-1)^n\log(n)}{n}$$ which is computed in $(8)$. $\endgroup$ – robjohn Jul 18 '13 at 13:43
  • 1
    $\begingroup$ @Henry: When $s\in\mathbb{R}$, $n^{-s}$ follows a straight line (along the real axis) from $1$ to $0$, as opposed to the case where $s\not\in\mathbb{R}$, in which $n^{-s}$ spirals from $1$ to $0$ (which is discussed in the next sentence). Remember we are assuming $\mathrm{Re}(s)\gt0$. $\endgroup$ – robjohn Aug 25 '14 at 20:01
6
$\begingroup$

There is a nice proof in Titchmarsh's "The Theory of the Riemann Zeta Function": putting $\,[x]=\,$ the greatest integer not greater than $\,x\in\mathbb R\,$, we get: $$\lim_{x\to 1^+}\left[\zeta(s)-\frac{1}{s-1}\right]\stackrel{\mathbf{(7)}}=\int_1^\infty\frac{[x]-x+\frac{1}{2}}{x^2}\,dx+\frac{1}{2}=\\=\int_1^\infty\frac{[x]-x}{x^2}+\frac{1}{2}\int_1^\infty\frac{dx}{x^2}+\frac{1}{2}=\int_1^\infty\frac{[x]-x}{x^2}\,dx+1=$$$$=\lim_{n\to\infty}\left[\sum_{m=1}^{n-1}\left(\int_m^{m+1}\frac{[x]dx}{x^2}-\int_m^{m+1}\frac{dx}{x}\right)+1\right]=$$$$\lim_{n\to\infty}\left[\sum_{m=1}^{n-1}m\left(\int_m^{m+1}\frac{dx}{x^2}\right)-\log n+1\right]=$$$$=\lim_{n\to\infty}\left[\left(1-\frac{1}{2}+1-\frac{2}{3}+...+1-\frac{m-1}{m}\right)+1-\log n\right]=$$$$=\lim_{n\to\infty}\left(\sum_{m=1}^{n-1}\frac{1}{m+1}+1-\log n\right)=\lim_{n\to\infty}\left(\sum_{m=1}^n\frac{1}{m}-\log n\right)=:\gamma$$

Finally, since we know $\,\displaystyle{\lim_{s\to 1^+}(s-1)\zeta(s)=1}\,$, we have that $\,s=1\,$ is a simple pole of $\,\zeta(s)\,$ with residue $\,1\,$, so the above gives the free coefficient of the Laurent expansion of $\,\zeta(s)\,$ around $\,1$

Proof of (7): We use the next form of Abel's summation by parts formula (all the time, $\,n\in\mathbb N\,$):

Lemma: Let $\,\phi(x)\,$ be any function with continuous derivative in $\,[a,b]\,$, then $$\sum_{a< n\leq b}\phi(n)=\int_a^b\phi(x) dx+\int_a^b\left(x-[x]-\frac{1}{2}\right)\phi'(x)dx+\left(a-[a]-\frac{1}{2}\right)\phi(a)-\left(b-[b]-\frac{1}{2}\right)\phi(b)\,\,\,\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\color{blue}{(1)}$$

$\color{red}{\text{Proof}}$: Taking $\,a=n\,,\,b=n+1\,$ and integrating by parts we get at once: $$\int_n^{n+1}\left(x-n-\frac{1}{2}\right)\phi'(x)dx=\left.\left(x-n-\frac{1}{2}\right)\phi(x)\right|_n^{n+1}-\int_n^{n+1}\phi(x)dx=$$ $$=\frac{1}{2}\left(\phi(n)+\phi(n+1)\right)-\int_n^{n+1}\phi(x)dx\Longrightarrow$$ $$\Longrightarrow \int_n^{n+1}\phi(x)dx+\int^{n+1}_n\left(x-[x]-\frac{1}{2}\right)\phi'(x)dx+\left(n-[n]-\frac{1}{2}\right)\phi(n)-\left(n+1-[n+1]-\frac{1}{2}\right)\phi(n+1)=\phi(n+1)=\sum_{n<m\leq n+1}\phi(m)\,\,,\,\,m\in\mathbb N$$

which shows both that the formula works for the above particular case and that it's enough to check for the case $\,n\leq a<b\leq n+1\,$ , so again integrating by parts: $$\int_a^b\left(x-n-\frac{1}{2}\right)\phi'(x)dx=\left(b-n-\frac{1}{2}\right)\phi(b)-\left(a-n-\frac{1}{2}\right)\phi(a)-\int_a^b\phi(x)dx$$ Comparing with the equality promised by the lemma, we see the RHS of $\,(1)\,$ above reduces to $$-\left(b-[b]-\frac{1}{2}\right)\phi(b)+\left(b-n-\frac{1}{2}\right)\phi(b)=\left([b]-n\right)\phi(b)$$ and this equals zero unless $\,b=n+1\,$, but then the last expression above equals $\,\phi(b)=\phi(n+1)\,$, which is the LHS in $\,(1)\,\;\;\;\;\;\;\;\square$

Now $\,(7)\,$ follows from the above with $$a_n=1\,\,,\,\forall n\in\mathbb N\,\,,\,\,\phi(n):=n^{-s}\quad\text{ and }\quad\,\,A(x):=\sum_{0\leq n\leq x}a_n=[x]$$ and we get

$$\zeta(s):=\sum_{n=1}^\infty\frac{1}{n^s}=\sum_{n=1}^\infty a_n\phi(n)=s\int_1^\infty\frac{[x]dx}{x^{1+s}}$$

Finally, we just write $$\frac{1}{s-1}=\int_1^\infty\frac{dx}{x^s}$$

$\endgroup$
  • $\begingroup$ Nice math and nice LaTeX. Hope you had already done this and had it ready, instead of entering it manually. $\endgroup$ – marty cohen Jul 30 '12 at 0:23
  • $\begingroup$ Thanks @martycohen. In fact I wrote the above from scratch because of two main reason: (1) I've been fooling a lot around Riemann's zeta function and wanted to clear out to myself some technical stuff in some parts, and (2) I wanted back then to continue practicing my LaTeX skills in this site. $\endgroup$ – DonAntonio Jul 30 '12 at 2:39
6
$\begingroup$

In my previous answer, it is shown that $$ \zeta(s)=\frac1{s-1}+\gamma+O(s-1)\tag{1} $$ We can compute more terms of the Laurent expansion using the formula $$ \begin{align} (-1)^n\frac{\mathrm{d}^n}{\mathrm{d}s^n}\left[(s-1)\sum_{k=1}^m\frac1{k^s}\right] &=(s-1)\sum_{k=1}^m\frac{\log(k)^n}{k^s}-n\sum_{k=1}^m\frac{\log(k)^{n-1}}{k^s}\tag{2} \end{align} $$ Euler-Maclaurin gives $$ \begin{align} \sum_{k=1}^m\frac{\log(k)^n}{k^s} &=\int_1^m\frac{\log(x)^n}{x^s}\,\mathrm{d}x+c_n(s)+O\left(\frac{\log(m)^n}{m^s}\right)\\ &=\int_0^{\log(m)}x^ne^{(1-s)x}\,\mathrm{d}x+c_n(s)+O\left(\frac{\log(m)^n}{m^s}\right)\\ &=\frac1{(s-1)^{n+1}}\int_0^{(s-1)\log(m)}x^ne^{-x}\,\mathrm{d}x+c_n(s)+O\left(\frac{\log(m)^n}{m^s}\right)\tag{3} \end{align} $$ Since $c_n(s)=\sum\limits_{k=1}^\infty\frac{\log(k)^n}{k^s}-\int_1^\infty\frac{\log(x)^n}{x^s}\,\mathrm{d}x$, we have $$ c_n'(s)=-c_{n+1}(s)\tag{4} $$ and since $\frac{\log(x)^n}{x^s}$ increases then decreases for $x\ge1$, $$ \left|c_n(s)\right|\le2\sup\limits_{x\ge1}\frac{\log(x)^n}{x^s}=2\left(\frac n{es}\right)^n\tag{5} $$

The left term of $(2)$ is $(s-1)$ times $(3)$ $$ \begin{align} (s-1)\sum_{k=1}^m\frac{\log(k)^n}{k^s} &=\frac1{(s-1)^n}\int_0^{(s-1)\log(m)}x^ne^{-x}\,\mathrm{d}x+(s-1)c_n(s)\\ &+(s-1)O\left(\frac{\log(m)^n}{m^s}\right)\tag{6} \end{align} $$ The right term of $(2)$ is $n$ times $(3)$ at $n-1$ $$ \begin{align} n\sum_{k=1}^m\frac{\log(k)^{n-1}}{k^s} &=\frac{n}{(s-1)^n}\int_0^{(s-1)\log(m)}x^{n-1}e^{-x}\,\mathrm{d}x+nc_{n-1}(s)+O\left(\frac{\log(m)^{n-1}}{m^s}\right)\\ &=\log(m)^nm^{1-s}+\frac1{(s-1)^n}\int_0^{(s-1)\log(m)}x^ne^{-x}\,\mathrm{d}x+nc_{n-1}(s)\\ &+O\left(\frac{\log(m)^{n-1}}{m^s}\right)\tag{7} \end{align} $$ Putting $(2)$, $(6)$, and $(7)$ together, we get $$ \begin{align} (-1)^n\frac{\mathrm{d}^n}{\mathrm{d}s^n}(s-1)\zeta(s) &=\lim_{m\to\infty}\left[(s-1)\sum_{k=1}^m\frac{\log(k)^n}{k^s}-n\sum_{k=1}^m\frac{\log(k)^{n-1}}{k^s}\right]\\[6pt] &=(s-1)c_n(s)-nc_{n-1}(s)\tag{8} \end{align} $$ Evaluating $(8)$ at $s=1$ gives $$ \begin{align} \left.(-1)^n\frac{\mathrm{d}^n}{\mathrm{d}s^n}(s-1)\zeta(s)\,\right|_{\,s=1} &=-nc_{n-1}(1)\\ &=\lim_{m\to\infty}\left[\log(m)^n-n\sum_{k=1}^m\frac{\log(k)^{n-1}}k\right]\tag{9} \end{align} $$ Applying $(9)$, we get the Laurent series for $\zeta(s)$ at $s=1$ to be $$ \bbox[5px,border:2px solid #C0A000]{\zeta(s)=\frac1{s-1}+\sum_{n=0}^\infty\frac{(1-s)^n}{n!}\,\gamma_n}\tag{10} $$ where $$ \gamma_n=\lim_{m\to\infty}\left[\sum_{k=1}^m\frac{\log(k)^n}k-\frac{\log(m)^{n+1}}{n+1}\right]\tag{11} $$

$\endgroup$
  • 1
    $\begingroup$ This was (is) a great supplement to the earlier post! $\endgroup$ – Mark Viola Oct 5 '16 at 19:16
0
$\begingroup$

This is my third answer to this question, but I just came up with this much simpler derivation.


A Simple Derivation of the Laurent Series for Zeta $$ \begin{align} &\frac1{s-1}+\sum_{k=1}^mk^{-s}-\frac{m^{1-s}-1}{1-s}\tag1\\ &=\frac1{s-1}+\sum_{k=1}^m\frac1ke^{(1-s)\log(k)}-\frac{e^{(1-s)\log(m)}-1}{1-s}\tag2\\ &=\frac1{s-1}+\sum_{n=0}^\infty\left[\sum_{k=1}^m\frac1k\frac{(1-s)^n\log(k)^n}{n!}-\frac{(1-s)^n\log(m)^{n+1}}{(n+1)!}\right]\tag3\\ &=\frac1{s-1}+\sum_{n=0}^\infty\frac{(1-s)^n}{n!}\left[\sum_{k=1}^m\frac{\log(k)^n}k-\frac{\log(m)^{n+1}}{n+1}\right]\tag4 \end{align} $$ Explanation:
$(2)$: convert powers to exponentials
$(3)$: expand exponentials about $s=1$
$(4)$: pull out a common factor

Taking the limit as $m\to\infty$, for $s\gt1$, $$ \bbox[5px,border:2px solid #C0A000]{\zeta(s)=\frac1{s-1}+\sum_{n=0}^\infty\frac{(1-s)^n}{n!}\,\gamma_n}\tag5 $$ where $$ \bbox[5px,border:2px solid #C0A000]{\gamma_n=\lim_{m\to\infty}\left[\sum_{k=1}^m\frac{\log(k)^n}k-\frac{\log(m)^{n+1}}{n+1}\right]}\tag6 $$


A Rough Bound on the Coefficients

We can get the following bound on $\gamma_n$ using a truncated version of the Euler-Maclaurin Sum Formula with $f_n(x)=\frac{\log(x)^n}{x}$ $$ \begin{align} |\gamma_n| &=\lim_{m\to\infty}\left|\,\sum_{k=1}^mf_n(k)-\int_1^mf_n(x)\,\mathrm{d}x\,\right|\\ &=\lim_{m\to\infty}\left|\,\frac12(f_n(m)+f_n(1))+\int_1^mf_n'(x)\left(\{x\}-\tfrac12\right)\mathrm{d}x\,\right|\\ &\le\frac12[n=0]+\frac12\operatorname*{Var}_{(1,\infty)}(f_n)\\[3pt] &=\frac{n^n}{e^n}\tag7 \end{align} $$ $(7)$ guarantees a radius of convergence in $(5)$ of no less than $1$. Since $(s-1)\zeta(s)$ is entire, the radius of convergence of $(5)$ is actually $\infty$.

$\endgroup$
0
$\begingroup$

$\newcommand{\bbx}[1]{\,\bbox[15px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ \begin{align} &\bbox[10px,#ffd]{\int_{0}^{\infty}{2t \over \pars{t^{2} + 1}\pars{\expo{\pi t} + 1}}\,\dd t} \\[5mm] = &\ 2\int_{0}^{\infty}{t \over t^{2} + 1}\pars{{1 \over \expo{\pi t} + 1} - {1 \over \expo{\pi t} - 1}}\dd t + 2\int_{0}^{\infty}{t \over \pars{t^{2} + 1}\pars{\expo{\pi t} - 1}}\,\dd t \\[5mm] = &\ -4\ \underbrace{\int_{0}^{\infty}{t \over \pars{t^{2} + 1}\pars{\expo{2\pi t} - 1}}\dd t} _{\ds{-1/2 - \Psi\pars{1} \over 2}}\ +\ \underbrace{2\int_{0}^{\infty}{t \over \bracks{t^{2} + \pars{1/2}^{2}}\pars{\expo{2\pi t} - 1}}\,\dd t} _{\ds{\ln\pars{1/2} - {1 \over 2\pars{1/2}} - \Psi\pars{1 \over 2}}} \label{1}\tag{1} \end{align}

$\ds{\Psi}$ is the Digamma Function. $\ds{\Psi\pars{1} = -\gamma}$ where $\ds{\gamma}$ is the Euler-Mascheroni Constant. $\ds{\Psi\pars{1 \over 2} = -\gamma - 2\ln\pars{2}}$. In evaluating \eqref{1} I used the $\ds{\mathbf{\color{black}{6.3.21}}}$ A & S identity.

\eqref{1} becomes $$ \bbox[10px,#ffd]{\int_{0}^{\infty}{2t \over \pars{t^{2} + 1}\pars{\expo{\pi t} + 1}}\,\dd t} = \bbx{\ln\pars{2} - \gamma} \approx 0.1159 $$

$\endgroup$

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.