I have two functions that I expect to be equal (where $P_{2l}$ are the even Legendre Polynomials):

$$F_{2l}(x)=x\, \tanh(\pi x/2)\left|\int_0^1 u^{i x-1}P_{2l}(u)\,du\right|^2$$

$$G_{2l}(x)=\frac{1}{2\pi}\, \int_{-\infty}^\infty dt\int^1_{-1}du\frac{P_{2l}(u)e^{-i x t}}{\cosh(t)-u}$$

I checked the Taylor expansion up to fifth order for $2l=0,2,4,6,8,10$.

Do you have any ideas or tips what a good approach could look like to prove the equality analytically? Beside trying standard transformations, I have been thinking about constructing a linear differential equation in x whose solution is given by one of them and showing that the other one also solves it. In particular, the second term is just a simple Fourier transform of the given integrals. The functions seem to be analytical (!?) and in fact for $F$, I did the integral for $\Im(x)<0$ and continued analytically (for the Taylor expansion).

Do you have any other tips or ideas to approach this problem?


$F_{2l}(x)$ can be calculated in closed form for $\text{Im}\ x<0$ using the integral from Gradsteyn and Ryzhik $$ \int_0^1u^\lambda P_{2l}(u)du=\frac{(-1)^l\Gamma\left(l-\frac{\lambda}{2}\right)\Gamma\left(\frac{1}{2}+\frac{\lambda}{2}\right)}{2\Gamma\left(-\frac{\lambda}{2}\right)\Gamma\left(l+\frac{3}{2}+\frac{\lambda}{2}\right)},\quad \text{Re}\ \lambda>-1, $$ $$ F_{2l}(x)=\frac{1}{4}x\, \tanh(\pi x/2)\left|\frac{\Gamma\left(l+\frac{1}{2}-\frac{ix}{2}\right)\Gamma\left(\frac{ix}{2}\right)}{\Gamma\left(\frac{1}{2}-\frac{ix}{2}\right)\Gamma\left(l+1-\frac{ix}{2}\right)}\right|^2=\\ \frac{1}{2}\left|\frac{\Gamma\left(l+\frac{1}{2}-\frac{ix}{2}\right)}{\Gamma\left(l+1-\frac{ix}{2}\right)}\right|^2,\quad \text{Im}\ x<0. $$ [Reflection formula for gamma function has been used to simplify the expression above]

Since $Q_n(z)=\frac{1}{2}\int_{-1}^{1}P_n(z)\frac{dy}{z-y}$ one can simplify $G_{2l}(x)$ as follows $$ G_{2l}(x)=\frac{1}{\pi}\, \int_{-\infty}^\infty Q_{2l}(\cosh t)e^{-ixt} dt. \tag{1} $$ This integral converges when $|\text{Im}\ x|<2l+1$, due to the asymptotics of Legendre function of the second kind $Q_{2l}(\cosh t)\sim e^{-(2l+1)t}$,$\ t\to +\infty$.

Now let's consider the following Fourier transform $$ I(t)=\int_{-\infty}^\infty\left|\frac{\Gamma\left(l+\frac{1}{2}-\frac{ix}{2}\right)}{\Gamma\left(l+1-\frac{ix}{2}\right)}\right|^2 e^{ixt}\ dx=\\ \int_{-\infty}^\infty\frac{\Gamma\left(l+\frac{1}{2}-\frac{ix}{2}\right)\Gamma\left(l+\frac{1}{2}+\frac{ix}{2}\right)}{\Gamma\left(l+1-\frac{ix}{2}\right)\Gamma\left(l+1+\frac{ix}{2}\right)} e^{ixt}\ dx, \quad t>0 $$ This can be calculated by contour integration as follows. Note that the only poles of the integrand in the upper half plane come from the factor $\Gamma\left(l+\frac{1}{2}+\frac{ix}{2}\right)$ in the numerator: $x_k=i(2l+k+1)$,$\ k=0,1,2,3,...$ The integrand decays as $1/|x|$ when $|x|\to+\infty$ and exponentially when $\text{Im}\ x\to +\infty$. Since $t>0$, one obtains closing the contour by a circle of large radius in the upper half plane $$ I(t)=2\pi i\ \sum_{k=0}^\infty \frac{(-1)^k\Gamma\left(2l+k+1\right)}{(i/2)k!\Gamma\left(2l+\frac{3}{2}+k\right)\Gamma\left(\frac{1}{2}-k\right)} e^{-(2l+k+1)t}. $$ [here we used that $\text{res}\ \Gamma(z)\Bigr|_{z=-m}=\frac{(-1)^m}{m!}$] This series can be expressed through hypergeometric function: $$ I(t)=\frac{4\sqrt{\pi}(2l)!e^{-(2l+1)t}}{\Gamma\left(2l+\frac{3}{2}\right)} \phantom{}_2F_1\left(2l+1,\frac{1}{2},2l+\frac{3}{2};e^{-2t}\right) $$ Applying Kummer's quadratic transformation one obtains $$ I(t)=\frac{4\sqrt{\pi}(2l)!}{\Gamma\left(2l+\frac{3}{2}\right)(2\cosh t)^{2l+1}} \phantom{}_2F_1\left(l+\frac{1}{2},l+1,2l+\frac{3}{2};\frac{1}{(\cosh t)^2}\right)=4Q_{2l}(\cosh t). $$ Now one can calculate $G_{2l}$ by taking inverse Fourier transform $$ G_{2l}(x)=\frac{1}{2}\cdot \frac{1}{2\pi}\, \int_{-\infty}^\infty 4Q_{2l}(\cosh t)e^{-ixt} dt=\frac{1}{2}\left|\frac{\Gamma\left(l+\frac{1}{2}-\frac{ix}{2}\right)}{\Gamma\left(l+1-\frac{ix}{2}\right)}\right|^2, \quad \text{for real}\ \ x.\tag{2} $$ But one can analytically continue (2) to the domain $|\text{Im}\ x|<2l+1$. This means that $F_{2l}(x)=G_{2l}(x)$, for $-2l-1<\text{Im}\ x<0$.


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.