6
$\begingroup$

Let $\square_n=\{(x_1,\dots,x_n): 0\leq x_i\leq1,\, \forall i\}$ be an $n$-dimensional unit hypercube, and let $\Delta_n=\{(u_1,\dots,u_n):u_1+\cdots+u_n\leq\frac{\pi}2,\, u_i\geq0,\, \forall i\}$ be $n$-simplex.

Also, let $E_{2m}$ be secant numbers (even-index Euler numbers) given by $$\sum_{m\geq0}E_{2m}\frac{y^{2m}}{(2m)!}.$$

The following has been experimentally tested.

Question. What is the transformation that makes this equality possible? $$\int_{\square_{2m+1}}\frac{d\mathbf{x}}{1+x_1^2\cdots x_{2m+1}^2}= \frac{(2m+1)E_{2m}}2\int_{\Delta_{2m+1}}d\mathbf{u}$$

$\endgroup$
  • $\begingroup$ Just to be sure: should the $m$ in the definition of $\Delta_n$ be an $n$? $\endgroup$ – Vincent May 2 '17 at 9:51
  • $\begingroup$ @Vincent: Thank you, this has been edited. $\endgroup$ – T. Amdeberhan May 2 '17 at 11:52
10
$\begingroup$

This is only a partial answer. The Beukers-Kolk-Calabi change of variables $$x_1=\frac{\sin{u_1}}{\cos{u_2}},\;\;x_2=\frac{\sin{u_2}}{\cos{u_3}},\ldots, \;x_{n-1}=\frac{\sin{u_{n-1}}}{\cos{u_n}},\;\;x_n=\frac{\sin{u_n}}{\cos{u_1}}$$ has the Jacobian $$\frac{\partial(x_1,\ldots,x_n)}{\partial(u_1,\ldots,u_n)}= 1-(-1)^n\,x^2_1x^2_2\cdots x^2_n.$$ Therefore you integral is a volume of the polytope $\delta_{2m+1}=\left \{(u_1,\ldots,u_{2m+1}): u_i\ge 0,\; u_i+u_{i+1}\le \pi/2 \right \}$. Here $i=1,\ldots 2m+1$ and it is assumed that $u_i$ are indexed cyclically so that $u_{2m+2}=u_1$. It remains to relate the volumes of the polytope $\delta_{2m+1}$ and $(2m+1)$-simplex $\Delta_{2m+1}$. See https://arxiv.org/abs/math/0101168

P.S. The volume of $\delta_{2m+1}$ can be calculated by using (37) and (40) from https://arxiv.org/abs/1003.3602 and the result is $$Vol(\delta_{2m+1})=(-1)^m\frac{2^{2m-1}}{(2m)!}E_{2m}(1/2)\left(\frac{\pi}{2}\right)^{2m+1}=(-1)^m\frac{E_{2m}}{2(2m)!}\left(\frac{\pi}{2}\right)^{2m+1},$$ while (see, for example, https://eudml.org/doc/141172) $$Vol(\Delta_{2m+1})=\frac{1}{(2m+1)!}\left(\frac{\pi}{2}\right)^{2m+1}.$$ Therefore we indeed get the desired identity provided $(-1)^m$ is incorporated in the definition of Euler numbers through Euler polynomials (we used definitions from https://eudml.org/doc/49338 which doesn't incorporate $(-1)^m$).

$\endgroup$
  • 3
    $\begingroup$ It is well known that the volume of $\{u_i\geq 0, u_i+u_{i+1}\leqslant 1\}$ equals Euler number divided by factorial. Though I can not give a reference immediately. $\endgroup$ – Fedor Petrov May 2 '17 at 5:27
  • 1
    $\begingroup$ Yes, indeed if we mean the volume of the zigzag chain polytope (see, for example arxiv.org/abs/0912.4240). However, $\delta_n$ is not quite a zigzag chain polytope because of $n_n+u_1\le 1$ extra condition. $\endgroup$ – Zurab Silagadze May 2 '17 at 10:44
  • $\begingroup$ @ZurabSilagadze: This is cool. I wish we can give a direct transformation rule from $\square$ to $\Delta$ shapes, because your idea goes through $\square$ to $\delta$ and then followed by algebraic comparison of the volumes of $\Delta$ and $\delta$. $\endgroup$ – T. Amdeberhan May 2 '17 at 11:55
  • 1
    $\begingroup$ A reference for Fedor Petrov's comment is Richard P. Stanley, Two poset polytopes, Discrete Comput. Geom. 1 (1986), no. 1, 9–23, dedekind.mit.edu/~rstan/pubs/pubfiles/66.pdf. $\endgroup$ – Ira Gessel May 3 '17 at 20:07

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.