12
$\begingroup$

I came across with the infinite series $$\sum_{n=1,3,5,\ldots}^{\infty} \frac{1}{n^4}= \frac{\pi^4}{96}$$ when calculating a problem about an infinite deep square well in quantum mechanics.

Mathematica gives the result in the title, which is enough for a physics problem. But I just want to find how to evaluate the series. I think this sum should be connected to $\zeta(4)=\pi^4/90$, but can't figure out their relation.

$\endgroup$
  • 2
    $\begingroup$ So it's basically $\zeta(4)$ with odd denominators? Hint: try getting the even denominators first, and then subtract them from $\zeta(4)$. $\endgroup$ – Akiva Weinberger Nov 9 '14 at 14:10
  • 1
    $\begingroup$ Also, please put the question in the post (as well as in the title, if you want)—on mobiles, math in the title doesn't render. $\endgroup$ – Akiva Weinberger Nov 9 '14 at 14:11
  • $\begingroup$ I wonder whether consideration on the mobiles' users is a good reason to not write mathematics in the title...but I've this hunch it is not. $\endgroup$ – Timbuc Nov 9 '14 at 14:14
26
$\begingroup$

$$\frac{\pi^4}{90}=\sum_{n=1}^\infty\frac1{n^4}=\sum_{n=1}^\infty\frac1{(2n)^4}+\sum_{n=1}^\infty\frac1{(2n-1)^4}=\frac1{16}\sum_{n=1}^\infty\frac1{n^4}+\sum_{n=1}^\infty\frac1{(2n-1)^4}\implies$$

$$\implies\sum_{n=1}^\infty\frac1{(2n-1)^4}=\left(1-\frac1{16}\right)\frac{\pi^4}{90}=\frac{\pi^4}{96} $$

$\endgroup$
5
$\begingroup$

You can also use the well known summation formula $$\sum_{n\in\mathbb{Z}}f\left(n\right)=-\sum\left\{ \textrm{residue of }\pi\cot\left(\pi z\right)f\left(z\right)\textrm{ at }f\left(z\right)\textrm{ poles}\right\} $$ which is a consequence of the residue theorem. So it is sufficient to note that $$ \sum_{n\geq1}\frac{1}{\left(2n-1\right)^{4}}=\frac{1}{2}\sum_{n\in\mathbb{Z}}\frac{1}{\left(2n-1\right)^{4}}=-\frac{1}{2}\left\{ \textrm{residue of }\frac{\pi\cot\left(\pi z\right)}{\left(2z-1\right)^{4}}\textrm{ at }\frac{1}{2}\right\} =\color{red}{\frac{\pi^{4}}{96}}.$$

$\endgroup$
2
$\begingroup$

Here is another proof I came up with. Start with the quadruple integral:

$$J=\int_{0}^{1}\int_{0}^{1}\int_{0}^{1}\int_{0}^{1} \frac{1}{1-x^2_1x^2_2x^2_3x^2_4}dx_1dx_2dx_3dx_4.$$

Replace the integrand with a geometric series as such: $$\frac{1}{1-x^2_1x^2_2x^2_3x^2_4}=\sum_{n=0}^{\infty}(x_1x_2x_3x_4)^{2n}.$$ Replace the integrand with the geometric series and integrate term by term to get:

$$J=\sum_{n=0}^{\infty} \frac{1}{(2n+1)^4}.$$

We now try to evaluate: $$J=\int_{0}^{1}\int_{0}^{1}\int_{0}^{1}\int_{0}^{1} \frac{1}{1-x^2_1x^2_2x^2_3x^2_4}dx_1dx_2dx_3dx_4.$$ Make the change of variables $$x_i=\frac{\sin(u_i)}{\cos(u_{i+1})},x_4=\frac{\sin(u_4)}{\cos(u_{1})}, 1\leq i \leq 3.$$ The Jacobian determinant turns out to be: $$\frac{\partial(x_1,...x_4)}{\partial(u_1,...u_4)}=1-x^2_1x^2_2x^2_3x^2_4$$ which cancels with the integrand. The region of integration is the (open) polytope $P$ defined by the constraints $$0<u_{i}+u_{i+1}<\frac{\pi}{2},0<u_{1}+u_{4}<\frac{\pi}{2}, u_{i}>0.$$ So this means $$J=\sum_{n=0}^{\infty} \frac{1}{(2n+1)^4}=\text{Vol}(P)$$ where $\text{Vol}$ means volume. I will consider the scaled polytope $V$ defined by constraints $$0<v_{i}+v_{i+1}<1,0<v_{1}+v_{4}<1, u_{i}>0.$$ You can show through change of variables $$u_i=\frac{\pi v_i}{2}, 1\leq i \leq 4$$ that $$\text{Vol}(P)=\left(\frac{\pi}{2} \right)^4 \text{Vol}(V).$$ So we will be done once we find $\text{Vol}(V).$ We will find $\text{Vol}(V)$ through probability.

We first find $$\text{Pr}\left(v\in V \cap v_1,...v_4<\frac{1}{2}\right),$$ which is the probability we pick an arbitrary point $v$ in $V$ where all coordinates are less than $\frac{1}{2}.$

$$\text{Pr}\left(v\in V \cap v_1,...v_4<\frac{1}{2}\right)=\left(\frac{1}{2}\right)^4=\frac{1}{16}.$$ This is the case because the open hypercube $I=(0,\frac{1}{2})^4 \subset V,$ which you can verify.

Now we find $$\text{Pr}\left(v\in V \cap \text{exactly one } v_i \geq \frac{1}{2}\right),$$ which is the probability we pick an arbitrary point $v$ in $V$ where exactly one coordinate is greater than or equal to $\frac{1}{2}.$ It turns out:

$$\text{Pr}\left(v\in V \cap \text{exactly one } v_i \geq \frac{1}{2}\right)=4\int_{\frac{1}{2}}^{1}\int_{0}^{1-v_1}\int_{0}^{1-v_1}\int_{0}^{\frac{1}{2}} 1 dv_3dv_4dv_2dv_1=\frac{1}{12}.$$ What I did here was compute $$\text{Pr}\left(v\in V \cap \text{only } v_1 \geq \frac{1}{2}\right)$$ and multiplied this answer by $4$ to account for all $4$ cases of this condition.

Now we find $$\text{Pr}\left(v\in V \cap \text{exactly two } v_i \geq \frac{1}{2}\right),$$ which is the probability we pick an arbitrary point $v$ in $V$ where exactly two coordinates are greater than or equal to $\frac{1}{2}.$ It turns out: $$\text{Pr}\left(v\in V \cap \text{exactly two } v_i \geq \frac{1}{2}\right)=4\int_{\frac{1}{2}}^{1}\int_{\frac{1}{2}}^{v_1}\int_{0}^{1-v_1}\int_{0}^{1-v_1} 1 dv_2dv_4dv_3dv_1=\frac{1}{48}.$$ What I did here was compute $$\text{Pr}\left(v\in V \cap \text{only } v_1,v_3 \geq \frac{1}{2}\cap v_3<v_1\right)$$ and multiplied this answer by $4$ to account for all $4$ cases of this condition. Keep in mind here that only two nonconsecutive coordinates can be greater than or equal $\frac{1}{2}$ at the same time. Thus, if we have more than two greater than or equal to $\frac{1}{2}$ at the same time, we will violate the constraints of $V$. Thus, we now just add up the computed probabilities to get: $$\text{Vol}(V)=\frac{1}{16}+\frac{1}{12}+\frac{1}{48}=\frac{1}{6},$$ which means $$\sum_{n=0}^{\infty} \frac{1}{(2n+1)^4}=\text{Vol}(U)=\left(\frac{\pi}{2}\right)^4\text{Vol}(V)=\left(\frac{\pi^4}{16}\right)\frac{1}{6}=\frac{\pi^4}{96}.$$

$\endgroup$
1
$\begingroup$

$\newcommand{\angles}[1]{\left\langle\,{#1}\,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{\mathrm{i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\Li}[1]{\,\mathrm{Li}_{#1}} \newcommand{\mrm}[1]{\,\mathrm{#1}} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ As shown by $\texttt{@Timbuc}$, you just need to know the sum $\ds{\sum_{n = 1}^{\infty}{1 \over n^{4}}}$. It has an interesting evaluation by starting from the $\ds{\cot\pars{z}}$ expansion $\ds{\pars{~\mbox{with}\ z \not= 0, \pm\pi,\pm 2\pi,\ldots~}}$:

\begin{align} \cot\pars{z} & = {1 \over z} + 2z\sum_{n = 1}^{\infty}{1 \over z^{2} - n^{2}\pi^{2}} = {1 \over z} - {2z \over \pi^{2}}\sum_{n = 1}^{\infty}{1 \over n^{2}} {1 \over 1 - z^{2}/ \pars{n^{2}\pi^{2}}} \\[5mm] & = {1 \over z} - {2z \over \pi^{2}}\sum_{n = 1}^{\infty}{1 \over n^{2}} \pars{1 + {z^{2} \over n^{2}\pi^{2}} + \cdots} = {1 \over z} - {2z \over \pi^{2}}\sum_{n = 1}^{\infty}{1 \over n^{2}} -{2z^{3} \over \pi^{4}}\color{#f00}{\sum_{n = 1}^{\infty}{1 \over n^{4}}} - \cdots \end{align}


\begin{align} z\cot\pars{z} & = 1 - {2z^{2} \over \pi^{2}}\sum_{n = 1}^{\infty}{1 \over n^{2}} -\bracks{{48 \over \pi^{4}}\color{#f00}{\sum_{n = 1}^{\infty}{1 \over n^{4}}}}{z^{4} \over 4!} - \cdots \end{align}
$$ \color{#f00}{\sum_{n = 1}^{\infty}{1 \over n^{4}}} = \left.-\,{\pi^{4} \over 48}\,\totald[4]{\bracks{z\cot\pars{z}}}{z} \right\vert_{\ z\ \to\ 0} = \pars{-\,{\pi^{4} \over 48}}\pars{-\,{8 \over 15}} = \color{#f00}{\pi^{4} \over 90} $$

$\endgroup$

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.