Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have this integral related to a Laplace transform and I was wondering if anyone knows of a clever way to derive it. I know we usually look these up in a table, but this form is not in a table I have. The transform is:

$$\mathcal{L}\left\{ \frac{\sin 1/t}{\sqrt{t}} \right\}(s)=\int\limits_{0}^{\infty}\frac{\sin(1/t)}{\sqrt{t}}e^{-st}\,\mathrm dt, \;\ s>0.$$

I ran it through Maple and it gave me $$\sqrt{\frac{\pi}{s}}\sin(\sqrt{2s})e^{-\sqrt{2s}}.$$

A rather simplistic looking solution, but how to derive?

I did notice that the $$\frac{1}{t^{\frac{1}{2}}}=\sqrt{\frac{\pi}{s}},$$

since the Laplace of $$t^{k}=\frac{\Gamma(k+1)}{s^{k+1}}, \;\ k>-1.$$ This would mean

$$t^{\frac{-1}{2}}=\frac{\Gamma(\tfrac12)}{\sqrt{s}}=\sqrt{\frac{\pi}{s}}.$$

Perhaps it is too difficult to do by hand. I just thought it was interesting how to derive this if possible.

Thanks very much for your time, interest, and expertise.

share|improve this question
    
I think you are in good way. You can try to put sine power series expansion to equation. Then see what equation turns.en.wikipedia.org/wiki/Power_series –  Mathlover Mar 3 '12 at 10:32
    
Expanding in power series doesn't help since negative powers of $t$ appears in the series expansion. –  user17090 Mar 3 '12 at 11:20
    
Terminology nitpick: When you want to derive X, you start with some set of givens and conclude with X as your destination; what you want to do is evaluate the integral, not derive it. –  anon Mar 3 '12 at 11:43
    
Sorry about the terminology. I will use 'evaluate' from now on. By derive, I meant do it by hand. –  Cody Mar 3 '12 at 16:29
add comment

2 Answers

up vote 2 down vote accepted

Let's set $t:=x^2$ then : $$I=2\int_0^{\infty} \sin(\frac 1{x^2}) e^{-sx^2}dx=i\int_0^{\infty} \left(e^{-\frac i{x^2}}-e^{\frac i{x^2}}\right) e^{-sx^2}dx$$

At this point you may use following formula : $$J(a,s)=\int_0^{\infty} e^{-\frac a{x^2}-s x^2} dx=\sqrt{\frac{\pi}{4s}}e^{-2\sqrt{as}}$$

(right and left terms both verify $J(a,s)=-\frac{\partial J(s,a)}{\partial s}$, $\frac{\partial^2 J}{\partial a \partial s}=J$ and $J(0,s)=\sqrt{\frac{\pi}{4s}}$, this formula is usually provided with $a\gt 0$, $s\gt 0$ but may be extended by analytic continuation)

For a more serious justification of this formula of P. Laplace you may search it in Gradshteyn and Ryzhik's famous "Table of integrals, series and products" getting (3.325) and use Moll and other's heroic effort to justify them all! In this case see the chapter 5 of this paper of Albano, Amdeberhan, Beyerstedt and Moll.

so that we get :

$$I=i\sqrt{\frac{\pi}{4s}}\left(e^{-2\sqrt{is}}-e^{-2\sqrt{-is}}\right)$$

Use $\sqrt{i}=\frac{1+i}2$ and $\sqrt{-i}=\frac{1-i}2$ to get :

$$I=i\sqrt{\frac{\pi}{4s}}\left(e^{-(1+i)\sqrt{2s}}-e^{-(1-i)\sqrt{2s}}\right)$$

$$I=\sqrt{\frac{\pi}{s}}\sin(\sqrt{2s})e^{-\sqrt{2s}}$$

Simple looking things are sometimes not so simple and some of the previous steps are not so easy to justify! (if you remove the $\sqrt{t}$ at the denominator or put a $t$ or something else there you won't get a simple answer!)

share|improve this answer
    
Well yes, if we have such a formula then the rest is easy. I've spent the last 2 hours trying to evaluate such an integral, I just got it! –  Ragib Zaman Mar 3 '12 at 12:04
    
@RagibZaman: it is a classical integral from qft (path integrals). I am searching a reference... –  Raymond Manzoni Mar 3 '12 at 12:07
    
Thanks much, Raymond, and thanks for the links. I have the error function paper, so I will have to look it over. –  Cody Mar 3 '12 at 16:36
add comment

The Key Lemma: For all $a\in \mathbb{C}$ with $\Re(a)>0$ we have $$ \int^{\infty}_0 \frac{1}{\sqrt{t} } e^{- a (t+ 1/t)^2 } dt = e^{-2a} \sqrt{ \frac{\pi}{a} } .$$

Proof: Let the integral be denoted by $I$ and $t=u^2$ so that $$ I = 2 \int^{\infty}_0 e^{-a (u^2+1/u^2) } du= 2 e^{-2a} \int^{\infty}_0 e^{-a (u-1/u)} du. $$

By letting $ u=1/t$ we see that $$ \int^{\infty}_0 \frac{1}{u^2} e^{-a (u-1/u)^2} du = \int^{\infty}_0 e^{-a (t-1/t)^2 } dt $$

so then $$ I = e^{-2a} \int^{\infty}_0 \left(1 + \frac{1}{u^2} \right) e^{-a (u-1/u)^2 } du. $$

Now letting $ x= u-1/u $ gives $$ I = e^{-2a} \int^{\infty}_{-\infty} e^{-ax^2} = e^{-2a} \sqrt{ \frac{\pi}{a} } .$$


Idea for the rest of the proof: (I will come back soon and complete it.) Prove that $$\int^{\infty}_0 \frac{1}{\sqrt{t} } e^{-st + i/t } dt = \sqrt{\frac{\pi}{s}} e^{-(1-i)\sqrt{2s} } dt $$

from which the result follows by taking imaginary parts of both sides. To do this, let $ t= \frac{1-i}{\sqrt{2s} } x$ (This "substitution" can be made valid by integrating along a suitable sector in the complex plane). Then the resulting integral becomes of the form of our lemma (up to a constant multiple) and then we are basically done.

share|improve this answer
    
Thanks much Ragib. I would not have thought of that. I figured it was rather complicated even though the solution was ratehr simplistic-looking. –  Cody Mar 3 '12 at 16:34
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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