2
$\begingroup$

How to prove the following inequality $$\forall t>0,\quad\int^\infty_0 \sin(rt)\frac{r^3}{\sinh^2(r)} dr\leq c \big(te^{-At}\big)$$ for some constants $A>0,c>0$

$\endgroup$

2 Answers 2

5
$\begingroup$

This is to complement the answer by Carlo Beenakker by showing that $$I(t)\le\pi^4 te^{-\pi t}\tag{1}\label{1}$$ for real $t\ge0$, where $I(t)$ is the integral in question.

Indeed, according to Carlo Beenakker, $$I(t)=\frac{\pi ^3}{8} \frac{\pi t \cosh \pi t+2\pi t-3 \sinh \pi t}{\sinh^4(\pi t/2)}.$$ This expression for $I(t)$ can be obtained from formula 3.986.4 of Gradshteyn and Ryzhik by using substitutions $\pi x=r$ and $\beta=\pi t/2$, to get $$J(t):=\int_0^\infty\frac{1-\cos rt}{\sinh^2 r}\,dr =\frac{\pi t}2\,\coth\frac{\pi t}2-1,$$ and then noting that $I(t)=-J'''(t)$.

So, using the substitution $t=\dfrac{\ln(1+x)}\pi$, one rewrites inequality \eqref{1} as $$d(x):=\left(8 x^3+19 x^2+18 x+6\right) \ln(x+1)-3 x (x+1)^2 (x+2)\le0\tag{2}\label{2}$$ for real $x\ge0$.

In turn, inequality \eqref{2} follows immediately because $d(0)=d'(0)=d''(0)=d'''(0)=0$ and $$d''''(x)=-\frac{2 x \left(36 x^3+120 x^2+139 x+58\right)}{(x+1)^4}\le0$$ for real $x\ge0$. $\quad\Box$

One may note that $I(t)\sim\pi^4 te^{-\pi t}$ as $t\to\infty$, so that the upper bound $\pi^4 te^{-\pi t}$ on $I(t)$ in \eqref{1} is exact.

$\endgroup$
4
  • $\begingroup$ Thank you a lot for your help@ Iosif Pinelis. $\endgroup$ Commented Jul 5, 2022 at 19:47
  • $\begingroup$ It'is clear for me thank you. $\endgroup$ Commented Jul 8, 2022 at 16:56
  • $\begingroup$ Thank you lot@Iosif Pinelis. $\endgroup$ Commented Jul 9, 2022 at 10:37
  • $\begingroup$ I voted on it five days ago. Dear @Iosif Pinelis. $\endgroup$ Commented Jul 18, 2022 at 17:41
6
$\begingroup$

$$\int^\infty_0 \frac{r^3\sin rt}{\sinh^2 r} \,dr=\tfrac{1}{8} \pi ^3 \frac{\pi t \cosh \pi t+2\pi t-3 \sinh \pi t}{\sinh^4(\pi t/2)}, $$ which decays as $e^{-\pi t}$ for large $t$, while it vanishes $\propto t$ for small $t$, so choosing $A$ a bit smaller than $\pi$ will satisfy the inequality for all $t>0$ for a sufficiently large $c$.

I checked that $A=3$ and $c=100$ works.

The exponential decay $\propto e^{-\pi t}$ could be obtained more directly from the pole of the integrand at $r=i\pi$, since the residue decays as $e^{irt}$.

$\endgroup$
5
  • 1
    $\begingroup$ You also need to remark it goes to $0$ as $t \to 0$. $\endgroup$ Commented Jul 5, 2022 at 11:18
  • 1
    $\begingroup$ right, added the small-$t$ limit, thanks $\endgroup$ Commented Jul 5, 2022 at 11:20
  • $\begingroup$ Thank you a lot @Carlo Beenakker. But is there a manner to prove the inequality whitout calculating. $\endgroup$ Commented Jul 5, 2022 at 11:32
  • 1
    $\begingroup$ yes, all that matters is the exponential decay $\propto e^{-\pi t}$, which follows from the pole at $r=i\pi$. $\endgroup$ Commented Jul 5, 2022 at 11:45
  • $\begingroup$ Please, could you detail more $\endgroup$ Commented Jul 5, 2022 at 11:54

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .