Let $\alpha>1$ and $z>0$. I am looking for a solution to the following Laplace transofrm: \begin{equation} \alpha\cdot\mathcal{L}\left\{{_{2}F_{1}(1-\alpha,1;2;-t)}\right\}(z)=\mathcal{L}\left\{\frac{(1+t)^\alpha-1}{t}\right\}(z) =\int_{0}^{\infty}\frac{(1+t)^\alpha-1}{t}\,e^{-zt}\,\mathrm{d}t \end{equation} I found a solution to the Laplace transform of ${_{2}F_{1}}(-t)$ here but it does not seems to work.

I also found this (Gradshteyn & Ryzhik, ed. 8, #7.522.1): \begin{equation} \int_{0}^{\infty}e^{-\lambda x}x^{\gamma-1}{_{2}F_1}(\alpha,\beta;\delta;-x)\,\mathrm{d}x=\frac{\Gamma(\delta)\lambda^{-\gamma}}{\Gamma(\alpha)\Gamma(\beta)}E(\alpha\,;\beta:\gamma\,;\delta:\lambda), \end{equation} for $\Re \lambda>0$ and $\Re \gamma>0$ where $E(p\,;\alpha_{r}:q\,;\varrho_{s}:x)$ is the MacRobert E-Function. Apparently, the E-function canbe expressed as MeijerG; however, I am having trouble understanding how to do so. In the Table of Integral Transforms (vol. 1, pg. 384) we have \begin{equation} E(p\,;\alpha_{r}:q\,;\beta_{s}:x)= G_{\,q+1,p}^{\,p,1}\left(x\,\middle\vert \begin{array}{c} 1,\beta_1,\dots,\beta_q\\ \alpha_1,\dots,\alpha_p \end{array} \right). \end{equation}


You can convert the hypergeometric function to the Meijer G-function and use the formula for the integral of a product of two linear G-functions: $$\alpha \int_0^\infty {_2F_1}(1 - \alpha, 1; 2; -t) e^{-z t} dt = \\ \frac \alpha {\Gamma(1 - \alpha)} \int_0^\infty G_{2, 2}^{1, 2} \left(t \middle| {0, \alpha \atop 0, -1} \right) G_{0, 1}^{1, 0} \left(z t \middle| {- \atop 0} \right) dt = \\ \frac \alpha {\Gamma(1 - \alpha)} G_{2, 3}^{3, 1} \left(z \middle| {0, 1 \atop 0, 0, -\alpha} \right).$$ In terms of more elementary function, it'll be $$\frac z {1 + \alpha} \,{_2F_2}(1, 1; 2, 2 + \alpha; z) + \Gamma(\alpha + 1) z^{-\alpha} (-z)^\alpha \Gamma(-\alpha, -z) + \\ \ln z + \pi \csc (\pi \alpha) z^{-\alpha} (-z)^\alpha - \psi(\alpha) - \pi \cot \pi \alpha - \frac 1 \alpha,$$ $\alpha>0, \operatorname{Re} z > 0$. For integral $\alpha_0$, this is still valid as a limit, with the singular terms yielding $$\lim_{\alpha \to \alpha_0}( \pi \csc (\pi \alpha) z^{-\alpha} (-z)^\alpha - \pi \cot \pi \alpha) = \ln(-z) - \ln z.$$

| cite | improve this answer | |
  • 1
    $\begingroup$ How did you convert the G function to more elementary functions? Was it done in mathematica? Or do you have an identity to reference? $\endgroup$ – Aaron Hendrickson Jun 17 '18 at 23:38
  • 2
    $\begingroup$ To derive this identity, write the G-function by definition as the integral of $\Gamma(1+y) \Gamma(-y)^2 \Gamma(-\alpha-y) \Gamma(1-y)^{-1} z^y$. The integral has to be taken over the right loop. When $\alpha$ isn't an integer, the sum of the residues over the simple poles at $y=-\alpha+k$ will give ${_1F_1}(-\alpha;-\alpha+1;z)$, which simplifies to the incomplete gamma function; the sum over the simple poles at $y=k$ will give ${_2F_2}(1,1;2,2+\alpha;z)$; and the double pole at $y=0$ will give the logarithmic and the polygamma terms. $\endgroup$ – Maxim Jun 18 '18 at 23:37
  • $\begingroup$ Thanks. I will take a look at it. $\endgroup$ – Aaron Hendrickson Jun 19 '18 at 0:11

Wolfram Language all the way:



$$\frac{-\mathcal{L}_t\left[\frac{1}{t}\right](z)+\frac{z \, _2F_2(1,1;2,a+2;z)}{a+1}+\frac{\pi z^{-a} \csc (\pi a) \left((-1)^{a+1} z^a \Gamma (1-a,-z)+e^z\right)}{\Gamma (1-a)}+\pi e^{i \pi a} \csc (\pi a)-\psi ^{(0)}(-a)-\gamma }{a}$$ Here $\gamma$ is the Euler constant, and $\psi^{(0)}(-\alpha)$ is the PolyGamma function. You still have an LT in here, unfortunately, but it seems reduced compared with the original.

| cite | improve this answer | |
  • $\begingroup$ Thank you. I believe I have found a more compact way to express the solution to the integral. If the solution can be converted to MeijerG that would be accepted. $\endgroup$ – Aaron Hendrickson Apr 10 '18 at 13:37

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.