1
$\begingroup$

I would like to know the asymptotics of the following sequences of integrals: $$ I_n = \displaystyle { \int _0 ^{+ \infty} \dfrac{t^n}{(t + i)^{n + 1}} \cdot e^{-t} \ dt } $$

Using a deformation of path with Cauchy theorem, I've shown that $$ I_n = \displaystyle { \int _{i \mathbb{R}^+} \dfrac{t^n}{(t + i)^{n + 1}} \cdot e^{-t} \ dt = \int _0 ^{+ \infty} \dfrac{t^n}{(t + 1)^{n + 1}} \cdot e^{-it} \ dt } $$

I therefore have tried using Laplace method, but I have been unable to conclude anything...

Moreover, I didn't manage to use computer software to have some reliable numeric values, because of the oscilating function $t \mapsto e^{-it}$. So, I haven't got any conjecture...

$\endgroup$
3
$\begingroup$

The saddle point $t^\ast$ is obtained by solving $f'(t)=0$ for $f(t)=-t+n\ln t-(n+1)\ln(t+i)$, for large $n$ we find $f(t^\ast)=-2\sqrt{in}$, so we arrive at the approximation for the integral $I_n\approx\exp(-2\sqrt{in})$.

The plot compares exact (gold) and approximate (blue) absolute values for $n$ up to 200 (multiply horizontal axis labels by 10).

$\endgroup$
4
  • $\begingroup$ Nice cart, to confirm theory! I find two possibilities to the saddle point $t^\star$: $\dfrac{-1 - i \pm \sqrt{2i (2n + 1)}}{2}$. Which one should we consider? $\endgroup$ – MathTolliob Jun 15 at 18:35
  • $\begingroup$ Do you have a link on with theory and example on Laplace method? I thought it applies on integral $\displaystyle { \int _a ^b g(t) e^{-\lambda f(t)} \ dt } $, with a function $f$ non-depending on the variable $\lambda$? $\endgroup$ – MathTolliob Jun 15 at 18:35
  • $\begingroup$ It looks to me that the picture shows divergence of the blue and gold plots. $\endgroup$ – Brendan McKay Jun 16 at 4:09
  • 1
    $\begingroup$ Using the saddle node $\dfrac{-1 -i + \sqrt{2i(2n + 1)}}{2}$, I conjecture that $ I_n \underset{n \longrightarrow + \infty}{\sim} \dfrac{\sqrt{\pi}}{n^{\frac{1}{4}}} e^{i \big ( \frac{1}{2} - \frac{\pi}{8} \big )} e^{-2 \sqrt{i n}} $. This has been confirmed by numerical result for $n = 2^k$ for $k \in [\![0;14]\!]$ $\endgroup$ – MathTolliob Jun 16 at 12:14

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.