I met this integral in my group:
$$\int_0^\infty \frac{x}{(e^x-e^{-x})^m}dx\qquad 0<m<2$$
I tried: $$\int_0^\infty \frac{x}{(e^x-e^{-x})^m}dx=\int_0^\infty \frac{xe^{-mx}}{(1-e^{-2x})^m}dx$$
let $e^{-2x}=t$,$x=\frac{-\ln t}{2}$ so the integral is equal to:
$$\int_1^0\frac{(\frac{-\ln t}{2})t^{\frac m2}}{(1-t)^m}(-\frac{1}{2t})dt=-\frac14\int_0^1(1-t)^{-m}t^{\frac m2-1}\ln tdt $$
But how to calculate this new integral? Who can help me. Thanks !
Assume that $0<m<1$.
Let $$ F(s)=\int_0^{+\infty}\frac{1}{(1-e^{-2x})^m}e^{-s x}\,dx. $$ Then, your integral equals $-F'(m)$.
The substitution you already did shows that $$ F(s)=\frac{1}{2}\int_0^1 (1-t)^{(1-m)-1}t^{s/2-1}\,dt=\frac{1}{2}\beta(1-m,s/2). $$ What remains is to calculate $$ -F'(m)=-\frac{d}{ds}\frac{1}{2}\beta(1-m,s/2)\Bigr|_{s=m} $$ But (see here for example) $$ \frac{d}{dx}\beta(y,x)=\beta(y,x)\bigl[\psi_0(x)-\psi_0(x+y)\bigr], $$ where $\psi_0$ denotes the Polygamma function. Hence, $$ -F'(m)=-\frac{1}{4}\beta(1-m,m/2)\bigl[\psi_0(m/2)-\psi_0(1-m/2)\bigr] $$ By the reflection principle for $\psi_0$ $$ \psi_0(1-z)-\psi_0(z)=\pi\cot(\pi z), $$ we find that
Edit
As kindly pointed out to me by @JanG, the original integral $F(s)$ diverges for $1\leq m<2$ (I was originally not reading the question well, and just assumed $0<m<1$).
To extend the equality in the yellow box above to $1\leq m<2$, we might argue as follows: The integral in the left-hand side is convergent for $0<\text{Re}\,m<2$, and depends analytically on $m$. The right-hand side depends analytically on $m$ (note that the singularity at $m=1$ of the beta function is canceled by the cotangent), and can be analytically continued from $0<\text{Re}\,m<1$ to $0<\text{Re}\,m<2$. Since the expressions agree for $0<m<1$, they agree for all $0<\text{Re}\,m<2$, by the identity theorem of analytic functions.