10

Cauchy and Ramanujan both gave the formula:

$$ \sum_{\stackrel{m \in \mathbb{Z}}{m \neq 0}} \frac{\coth m \pi}{m^{4p+3}} = (2\pi)^{4p+3}\sum_{k=0}^{2p+2} (-1)^{k+1} \frac{B_{2k}}{(2k)!}\frac{B_{4(p+1)-2k}}{(4(p+1)-2k)!} $$

If we lightly manipulate things, we can use a bit of umbral calculus to simplify the formula

\begin{eqnarray} -\frac{[4(p+1)]!}{(2\pi)^{4p+3}}\sum_{\stackrel{m \in \mathbb{Z}}{m \neq 0}} \frac{\coth m \pi}{m^{4p+3}} &=& \sum_{k=0}^{2p+2} (-1)^{k} \binom{4(p+1)}{2k}B_{2k}B_{4(p+1)-2k} \\ &=& \frac{1}{2}\left[ (B+iB)^{2(p+1)}+(B-iB)^{2(p+1)}\right] \end{eqnarray}

It's still a mess, but at least we have normalization. Prove LHS = RHS

cactus314
  • 24,438

2 Answers2

6

Note that this sum is in fact $$2\sum_{m\ge 1} \frac{\coth m\pi}{m^{4p+3}}.$$

The sum term $$T_p(x) = 2\sum_{m\ge 1} \frac{\coth mx}{m^{4p+3}}$$ is harmonic and may be evaluated by inverting its Mellin transform.

Observe that $$\coth x = \frac{e^x+e^{-x}}{e^x-e^{-x}} = 1 + 2\frac{e^{-x}}{e^x-e^{-x}} = 1 + 2\frac{e^{-2x}}{1-e^{-2x}}.$$

This yields $$T_p(x) = 2\zeta(4p+3) + 4\sum_{m\ge 1} \frac{1}{m^{4p+3}} \frac{e^{-2mx}}{1-e^{-2mx}}.$$

We will now work with $$S(x) = \sum_{m\ge 1} \frac{1}{m^{4p+3}} \frac{e^{-2mx}}{1-e^{-2mx}}$$ and establish a functional equation for $S(x)$ that has $x=\pi$ as a fixed point.

Recall the harmonic sum identity $$\mathfrak{M}\left(\sum_{k\ge 1} \lambda_k g(\mu_k x);s\right) = \left(\sum_{k\ge 1} \frac{\lambda_k}{\mu_k^s} \right) g^*(s)$$ where $g^*(s)$ is the Mellin transform of $g(x).$

In the present case we have $$\lambda_k = \frac{1}{k^{4p+3}}, \quad \mu_k = k \quad \text{and} \quad g(x) = \frac{\exp(-2x)}{1-\exp(-2x)}.$$

We need the Mellin transform $g^*(s)$ of $g(x)$ which we compute as follows: $$\int_0^\infty \frac{\exp(-2x)}{1-\exp(-2x)} x^{s-1} dx = \int_0^\infty \sum_{q\ge 1} \exp(-2qx)\; x^{s-1} dx \\ = \sum_{q\ge 1} \int_0^\infty \exp(-2qx)\; x^{s-1} dx = \Gamma(s) \sum_{q\ge 1} \frac{1}{2^s q^s} = \frac{1}{2^s} \Gamma(s) \zeta(s).$$

with fundamental strip $\langle 1,\infty \rangle.$

Hence the Mellin transform $Q(s)$ of $S(x)$ is given by $$ Q(s) = \frac{1}{2^s} \Gamma(s) \zeta(s) \zeta(s+4p+3) \quad\text{because}\quad \sum_{k\ge 1} \frac{\lambda_k}{\mu_k^s} = \zeta(s+4p+3)$$ where $\Re(s+4p+3) > 1$ or $\Re(s) > -4p-2$.

Intersecting the fundamental strip and the half-plane from the zeta function term we find that the Mellin inversion integral for an expansion about zero is $$\frac{1}{2\pi i} \int_{3/2-i\infty}^{3/2+i\infty} Q(s)/x^s ds$$ which we evaluate in the left half-plane $\Re(s)<3/2.$

The plain zeta function term cancels the poles of the gamma function term at even negative integers $2q\le -2$ and the compound zeta function term the poles at odd negative integers $2q+1\le -4p-5$. We are left with the contributions from $s=1$, $s=0$ and $s=-1$ which are

$$\begin{align} \mathrm{Res}(Q(s)/x^s; s=1) & = \frac{1}{2x}\zeta(4p+4) \\ \mathrm{Res}(Q(s)/x^s; s=0) & = -\frac{1}{2}\zeta(4p+3) \\ \mathrm{Res}(Q(s)/x^s; s=-1) & = \frac{1}{6} x\zeta(4p+2). \end{align}$$

The remaining contributions are $$\sum_{q=1}^{2p+1} \mathrm{Res}(Q(s)/x^s; s=-2q-1) \\ = \sum_{q=1}^{2p+1} 2^{2q+1} x^{2q+1} \frac{(-1)^{2q+1}}{(2q+1)!} \zeta(-2q-1) \zeta(4p+2-2q) \\ = \sum_{q=1}^{2p+1} 2^{2q+1} x^{2q+1} \frac{1}{(2q+1)!} \frac{B_{2q+2}}{2q+2} (-1)^{2p+1-q+1} \frac{B_{4p+2-2q}(2\pi)^{4p+2-2q}}{2(4p+2-2q)!} \\ = \sum_{q=1}^{2p+1} 2^{2q+1} x^{2q+1} \frac{B_{2q+2}}{(2q+2)!} (-1)^{-q} \frac{B_{4p+2-2q}(2\pi)^{4p+2-2q}}{2(4p+2-2q)!} \\ = \sum_{q=2}^{2p+2} 2^{2q-1} x^{2q-1} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}(2\pi)^{4p+4-2q}}{2(4p+4-2q)!} \\ = 2^{4p+2} \sum_{q=2}^{2p+2} x^{2q-1} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}\pi^{4p+4-2q}}{(4p+4-2q)!} \\ = 2^{4p+2} \sum_{q=2}^{2p+2} x^{2q-1} \pi^{4p+4-2q} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!}$$

and the one from the pole of the compound zeta function term at $s=-4p-2$ which we'll do in a moment.

Now some algebra shows that setting $q=0$ and $q=1$ in the sum produces precisely the values that we obtained earlier for the poles at $s=1$ and $s=-1$ so we may extend the sum to start at zero, keeping only the residue from the pole at $s=0$ to get $$2^{4p+2} \sum_{q=0}^{2p+2} x^{2q-1} \pi^{4p+4-2q} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!}.$$

We have thus established that $$S(x) = -\frac{1}{2}\zeta(4p+3) + \mathrm{Res}(Q(s)/x^s; s=-4p-2) \\ + 2^{4p+2} \sum_{q=0}^{2p+2} x^{2q-1} \pi^{4p+4-2q} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!} \\ + \frac{1}{2\pi i} \int_{-4p-4-i\infty}^{-4p-4+i\infty} Q(s)/x^s ds.$$

To treat the integral recall the duplication formula of the gamma function: $$\Gamma(s) = \frac{1}{\sqrt\pi} 2^{s-1} \Gamma\left(\frac{s}{2}\right) \Gamma\left(\frac{s+1}{2}\right).$$

which yields for the integral $$\int_{-4p-4-i\infty}^{-4p-4+i\infty} \frac{1}{2\sqrt\pi} \Gamma\left(\frac{s}{2}\right) \Gamma\left(\frac{s+1}{2}\right) \zeta(s) \zeta(s+4p+3)/x^s ds.$$

Furthermore observe the following variant of the functional equation of the Riemann zeta function: $$\Gamma\left(\frac{s}{2}\right)\zeta(s) = \pi^{s-1/2} \Gamma\left(\frac{1-s}{2}\right) \zeta(1-s)$$

which gives for the integral $$\int_{-4p-4-i\infty}^{-4p-4+i\infty} \frac{\pi^{s-1}}{2} \Gamma\left(\frac{s+1}{2}\right) \Gamma\left(\frac{1-s}{2}\right) \zeta(1-s) \zeta(s+4p+3)/x^s ds \\ = \int_{-4p-4-i\infty}^{-4p-4+i\infty} \frac{\pi^{s}}{2} \frac{1}{\sin(\pi/2(s+1))} \zeta(1-s) \zeta(s+4p+3)/x^s ds.$$

Now put $s=-4p-2-u$ to get $$\int_{2-i\infty}^{2+i\infty} \frac{\pi^{-4p-2-u}}{2} \frac{1}{\sin(\pi/2(-4p-2-u+1))} \\ \times \zeta(u+4p+3) \zeta(1-u)/x^{-4p-2-u} du \\ = \frac{x^{4p+2}}{\pi^{4p+2}} \int_{2-i\infty}^{2+i\infty} \frac{\pi^{-u}}{2} \frac{1}{\sin(\pi/2(-(u+1)))} \\ \times \zeta(u+4p+3) \zeta(1-u)/x^{-u} du \\ = - \frac{x^{4p+2}}{\pi^{4p+2}} \int_{2-i\infty}^{2+i\infty} \frac{\pi^{-u}}{2} \frac{1}{\sin(\pi/2(u+1))} \\ \times \zeta(u+4p+3) \zeta(1-u)/x^{-u} du.$$

We have established the functional equation $$S(x) = -\frac{1}{2}\zeta(4p+3) + \mathrm{Res}(Q(s)/x^s; s=-4p-2) \\ + 2^{4p+2} \sum_{q=0}^{2p+2} x^{2q-1} \pi^{4p+4-2q} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!} -\left(\frac{x}{\pi}\right)^{4p+2} S\left(\frac{\pi^2}{x}\right).$$

Setting $x=\pi$ we have $$S(\pi) = -\frac{1}{2}\zeta(4p+3) + \mathrm{Res}(Q(s)/\pi^s; s=-4p-2) \\ + 2^{4p+2} \pi^{4p+3} \sum_{q=0}^{2p+2} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!} -\left(\frac{\pi}{\pi}\right)^{4p+2} S\left(\pi\right)$$ and hence $$S(\pi) = -\frac{1}{4}\zeta(4p+3) + \frac{1}{2} \mathrm{Res}(Q(s)/\pi^s; s=-4p-2) \\ + 2^{4p+1} \pi^{4p+3} \sum_{q=0}^{2p+2} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!}.$$

To conclude we treat the residue that we have defered until now. Recall the alternate form of $Q(s)/\pi^s,$ $$\frac{\pi^{s}}{2} \frac{1}{\sin(\pi/2(s+1))} \zeta(1-s) \zeta(s+4p+3)/\pi^s \\ = \frac{1}{2} \frac{1}{\sin(\pi/2(s+1))} \zeta(1-s) \zeta(s+4p+3).$$ It follows that the residue at $s=-4p-2$ is $$\frac{1}{2} \frac{1}{\sin(\pi/2(-4p-1))} \zeta(4p+3) = \frac{1}{2} \frac{1}{\sin(-\pi/2)} \zeta(4p+3) = -\frac{1}{2} \zeta(4p+3).$$

This finally yields for $S(\pi)$ $$S(\pi) = -\frac{1}{2}\zeta(4p+3) + 2^{4p+1} \pi^{4p+3} \sum_{q=0}^{2p+2} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!}.$$

Computing $T_p(\pi)$ we thus obtain $$2\zeta(4p+3) -4\times\frac{1}{2}\zeta(4p+3) + 2^{4p+3} \pi^{4p+3} \sum_{q=0}^{2p+2} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!}$$ or $$(2\pi)^{4p+3} \sum_{q=0}^{2p+2} \frac{B_{2q}}{(2q)!} (-1)^{q+1} \frac{B_{4p+4-2q}}{(4p+4-2q)!},$$ QED.

The inspiration for this calculation is from the paper "Mellin Transform and its Applications" by Szpankowski.

Marko Riedel
  • 61,317
  • Thank you for these kind words. – Marko Riedel Jun 04 '15 at 20:50
  • Again as in your another post you made a connection between $S(x)$ and $S(\pi^2/x)$. But how did you relate each other since integral with $Q(s)$ and the one with sine in the denominator are note the same? I would be glad where was the point of establishing the functional relation. – Machinato Aug 04 '16 at 20:49
  • Nice derivation. Btw, Is this your own proof of did you read it somewhere? – BooleanCoder Apr 15 '21 at 04:11
  • @BooleanWick. The proof is mine but I make no claim to originality of the method. For sources consult "Mellin Transform Asymptotics" by Flajolet and Sedgewick INRIA RR 2956 as well as the Mellin transform material in Szpankowski's "Average Case Analysis of Algorithms on Sequences" as mentioned in the post. – Marko Riedel Apr 15 '21 at 16:18
  • I also recommend "On Plouffe's Ramanujan Identities" by L. Vepstas. – Marko Riedel Apr 15 '21 at 17:50
4

There is also a way to obtain this formula using the partial fraction expansion of cotangent: $$ \frac{\pi}{2x} \coth(\pi x) - \frac{1}{2x^2} = \sum_{k=1}^{\infty} \frac{1}{k^2+x^2} $$ Let $m\geq 0$ be an integer, set $x=n$, divide by $n^{4m+2}$ and then sum from $n=1$ to infinity to get $$ \sum_{n=1}^{\infty} \frac{\pi}{2n^{4m+3}} \coth(\pi n) - \sum_{n=1}^{\infty} \frac{1}{2n^{4m+4}} = \sum_{n=1}^{\infty} \sum_{k=1}^{\infty} \frac{1}{n^{4m+2}(k^2+n^2)} $$ The sum on the right may be handled by switching the order of summation (justified by absolute convergence) and then rewriting things as follows: $$ \sum_{k=1}^{\infty} \sum_{n=1}^{\infty} \frac{1}{n^{4m}} \frac{1}{n^2(k^2+n^2)} = \sum_{k=1}^{\infty} \sum_{n=1}^{\infty} \frac{1}{n^{4m}k^2} \left(\frac{1}{n^2}-\frac{1}{k^2+n^2} \right) = \zeta(4m+2)\zeta(2) - \sum_{k=1}^{\infty} \sum_{n=1}^{\infty} \frac{1}{n^{4m}k^2(k^2+n^2)} $$ Repeating the procedure results in $$ \sum_{k=1}^{\infty} \sum_{n=1}^{\infty} \frac{1}{n^{4m+2}(k^2+n^2)} = \zeta(4m+2)\zeta(2) - \zeta(4m)\zeta(4) + \cdots + \zeta(2) \zeta(4m+2) - \sum_{k=1}^{\infty} \sum_{n=1}^{\infty} \frac{1}{k^{4m+2}(k^2+n^2)}$$ The final sum is the same as the first if we switch the dummy variables, and so we have that $$ 2\sum_{k=1}^{\infty} \sum_{n=1}^{\infty} \frac{1}{n^{4m+2}(k^2+n^2)} = 2 \sum_{k=0}^{m-1} (-1)^k \zeta(4m+2-2k) \zeta(2k+2) +(-1)^m \zeta^2(2m+2) $$ Replacing this in the equality above and cleaning up a bit gives the result $$ \sum_{n=1}^{\infty} \frac{\coth(\pi n)}{n^{4m+3}} = \frac{\zeta(4m+4)}{\pi} + \frac{2}{\pi} \sum_{k=0}^{m-1} (-1)^k \zeta(4m+2-2k)\zeta(2k+2) + \frac{(-1)^m}{\pi} \zeta^2(2m+2) $$ This may be expressed in the form you wrote by using the well known formula for zeta at the even integers.

Dave
  • 1,723
  • Formula $$ \sum_{n=1}^{\infty} \frac{\coth(\pi n)}{n^{4m+3}} = \frac{\zeta(4m+4)}{\pi} + \frac{2}{\pi} \sum_{k=0}^{m-1} \zeta(4m+2-2k)\zeta(2k+2) + \frac{(-1)^m}{\pi} \zeta^2(2m+2) $$ dosen't work for $m>1$ ? Gives incorrect values.? – Mariusz Iwaniuk May 04 '20 at 12:45
  • Which values are incorrect? – Dave May 05 '20 at 14:34
  • For example m=2. With Mathematica:{f[m_] := Zeta[4 m + 4]/Pi + 2/Pi*Sum[Zeta[4 m + 2 - 2 k]*Zeta[2 k + 2], {k, 0, m - 1}] + (-1)^ m/Pi*Zeta[2 m + 2]^2, f[2] // N} give me:2.38791 but correct answer is: 1.00424. – Mariusz Iwaniuk May 05 '20 at 14:44
  • I have corrected the formula. There was a $(-1)^k$ factor missing before. Thanks for catching the mistake – Dave May 05 '20 at 15:58
  • Thanks,now it's ok. – Mariusz Iwaniuk May 05 '20 at 16:05
  • 1
    @Dave Nice derivation btw. Is this your own approach or have you read this derivation somewhere else already? Thanks. – BooleanCoder Apr 07 '21 at 09:18
  • Thanks! I saw the method used on a special case here: https://math.stackexchange.com/questions/991066/method-of-proof-of-sum-limits-n-1-infty-tfrac-coth-n-pin7-tfrac19 and generalized it. The method also works for finding three other classes series if you start with the partial fractions for tanh, csch, and sech. – Dave Apr 25 '21 at 03:30