Consider the integral

$$\mathcal{I}=\int_0^t\left(\frac{Ae^{-\lambda t}-Ae^{-\lambda s}}{2}\right)^{2m+1}e^{-\epsilon(t-s)}ds,\tag{1}$$

for constants $A,\lambda,\epsilon,t\in\mathbb{R}$ and $m\in\mathbb{Z}^+$.

The intention of evaluating $\mathcal{I}$, is to find

\begin{equation} \begin{split} \mathcal{S}&=\int_0^tJ_1\left(Ae^{-\lambda t}-Ae^{-\lambda s}\right)e^{-\epsilon(t-s)}ds, \\&=\int_0^t\sum_{m=0}^{\infty}\frac{ (-1)^m}{m!(m+1)!}\left(\frac{Ae^{-\lambda t}-Ae^{-\lambda s}}{2}\right)^{2m+1}e^{-\epsilon(t-s)}ds,\\ &=\sum_{m=0}^\infty \frac{ (-1)^m}{m!(m+1)!}\mathcal{I}_m \end{split}\end{equation}

where $J_1$ denotes the first order Bessel function of first kind. Does a closed form exist for $\mathcal{S}$ (or Taylor series)? Could an asymptotic bound simplify things?

According to Mathematica, we have \begin{equation} \mathcal{I}=\frac{(A-Ae^{-\lambda t})^2 (Ae^{-\lambda t}-A)^{2m} 2^{-2m-1}e^{-\epsilon t}\cdot{}_2F_1\left(1,\frac{\epsilon+\lambda}{\lambda};\frac{\epsilon}{\lambda}-2m,e^{-\lambda t}\right)}{A\left(\epsilon-\lambda(2m+1)\right)}. \end{equation}

From this, is it possible to prove that $2\mathcal{S}\sim A\lambda^2\kappa e^{-\lambda t}-A\lambda e^{-\lambda t}$ as $t\rightarrow\infty$, for some $\kappa$?

  • $\begingroup$ There is no constraint on the sign of $\lambda$? it would be surprising that the asymptotic you suggest (I don't know if it is the right one) works for both positive and negative $\lambda$. $\endgroup$
    – username
    Nov 11, 2021 at 21:59
  • $\begingroup$ The asymptotic is a conjecture based on looking at numerical simulations of the integro-differential equations discussed in mathoverflow.net/questions/407878/…. I have imposed no restrictions for now, but it may very well be incorrect for some certain of $\lambda$. $\endgroup$
    – UNOwen
    Nov 12, 2021 at 7:43

1 Answer 1


If $\lambda>0$ and $\lambda>\epsilon$ one has $$\lim_{t\rightarrow\infty}e^{\epsilon t}\mathcal{S}=\int_0^\infty J_1\left(-Ae^{-\lambda s}\right)e^{\epsilon s}ds$$ $$\qquad=\frac{A/2}{\epsilon-\lambda} \, _1F_2\left(\frac{1}{2}-\frac{\epsilon}{2 \lambda};2,\frac{3}{2}-\frac{\epsilon}{2 \lambda};-A^2/4\right).$$ The asymptotics is different for other ranges of $\lambda,\epsilon$, for example, for $\lambda,\epsilon<0$ one has ${\cal S}\rightarrow -e^{-\epsilon t}J_1(Ae^{-\lambda t})/\epsilon$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.