8

So, my teacher wants us to compute the value of the Fresnel integral:

$$\int_0^\infty\cos(x^2)dx=\sqrt{\frac{\pi}{8}}$$

The problem is that we cannot use complex analysis to prove that and we should do that using the Euler identity: $$\int_0^\infty\cos(x^2)dx=\frac{1}{2}\int_0^\infty e^{iw^2}dw+\frac{1}{2}\int_0^\infty e^{-iw^2}dw$$

But I have that integral of $e^{-iw^2}$ and I cannot solve that :(

I am an engineering student, so basically, I only have the "basic" calculus, just simple/double/triple integrals, some notions about series, ODE's and PDE's, but nothing as deep as in the regular Math degree, so probably there's no need to use hard stuff to figure this out.

Pedro
  • 122,002

6 Answers6

11

We have to prove: $$\int_{0}^{\infty} \cos(ax^2) dx =\sqrt{\frac{\pi}{8a}}$$ Now, we consider the LHS. $$ LHS = \int_{0}^{\infty} \cos(ax^2) dx $$

Now, we make the substitution: $$x \rightarrow x^{\frac{1}{4}}$$

Therefore, we get: $$ LHS =\frac{1}{4}\int_{0}^{\infty} x^{-\frac{3}{4}}\cos(a\sqrt{x}) dx $$ Now by Maclaurin series, $$\cos (a\sqrt{x}) = \sum\limits_{n=0}^{\infty} \frac{(-1)^n(a\sqrt x)^{2n}}{2n!}$$ This can also be written as: $$\cos (a\sqrt x) = \sum\limits_{n=0}^{\infty} \frac{(-x)^n(a)^{2n}n!}{2n!n!}$$ On plugging the value into LHS, we get: $$ LHS =\frac{1}{4} \int_{0}^{\infty} x^{-\frac{3}{4}} \sum\limits_{n=0}^{\infty} \frac{(-x)^n(a)^{2n}n!}{2n!n!} dx$$ Now, by Ramanujan's Master Theorem, we get $$ LHS = \frac{1}{4} \int_{0}^{\infty} x^{-\frac{3}{4}}\sum\limits_{n=0}^{\infty}\frac{(-x)^n(a)^{2n}n!}{2n!n!} dx = \frac{1}{4}\frac{\Gamma(\frac{1}{4})\Gamma(\frac{3}{4})}{\sqrt{a}\Gamma(\frac{1}{2})}$$ Therefore, by Euler's Reflection Formula, we get $$\int_{0}^{\infty}\cos(ax^2) dx =\sqrt{\frac{\pi}{8a}}$$

10

You can use double integrals and change of variables. Consider the surface $e^{-y^2}\cos(x^2)$ in the positive octant, and determine the volume bounded by it. Compute the integrals both in polar coordinates and in Cartesian coordinates, and equate the values. Then you can solve for the value of your integral.

To do it in Cartesian coordinates, you have a product of the Fresnel integral, $I$ and the Gaussian integral with value $\frac{\sqrt{\pi}}{2}$. Then

$$V = \frac{\sqrt{\pi}}{2}I$$

By expressing $V$ with polar cooridinates, and using the substitution $u = p^2\cos^2(\theta)$ then the substitution $\tan(\theta)=t$, you get the integral

$$V = \frac{1}{2}\int_{0}^{\infty}\frac{t^2}{t^4+1}dt$$ Using partial fractions you can get that this is $\frac{\pi\sqrt{2}}{8}$. Then by equating your two values of $V$ you can solve for I and get $$I = \sqrt{\frac{\pi}{8}}$$ as desired.

J.G.
  • 115,835
  • I'm very sorry, but could you be a little more specific? hahaha – Rafael Vignoli Dec 09 '16 at 16:32
  • Oh I got it now, thanks.. The only problem is that we didn't learn that Gaussian integral, I believe that sadly that is not the way, :( – Rafael Vignoli Dec 09 '16 at 16:50
  • You could show the Gaussian integral by essentially the same methods... But then your problem would take up several pages of computation. It may be a horrible way to go, but it is a way. – Retired account Dec 09 '16 at 16:54
7

Here is a (non-rigorous) proof with Fourier Transforms I discovered. It involves double integrals mostly.

Define the Fourier Cosine Transform of $f(x)$ to be:

$$\mathcal{F}_{c}(f(x))=\frac{1}{\sqrt{2 \pi}}\int_{-\infty}^{\infty} f(y) \cos(xy) \ dy.$$ One can see that $\mathcal{F}_c$ is self-adjoint by checking the definition using the inner product: $$\langle f(x), g(x) \rangle= \int_{\mathbb{R}} f(x)g(x) \ dx. $$ The definition more specifically is $$\left \langle \mathcal{F}_c(f), g \right \rangle=\left \langle f, \mathcal{F}_c (g) \right \rangle.$$ To prove it is true, expand the left hand side into a double integral, and use Fubini's theorem to get the right hand side.

We compute the inner product of $$\left \langle \mathcal{F}_c \left(\cos(x) \right), \frac{1}{\sqrt{|x|}} \right \rangle$$ in two ways.

Expanding the inner product, we get the inner product equal to $$\left \langle \sqrt{2 \pi} \delta(1-x) , \frac{1}{\sqrt{|x|}} \right \rangle= \sqrt{2 \pi},$$ from the integral property of Dirac Delta function $\delta(x).$

On the other hand, exploiting the self-adjoint property, $$\sqrt{2 \pi}=\left \langle \cos(x), \mathcal{F}_c \left( \frac{1}{\sqrt{|x|}} \right ) \right \rangle =\frac{1}{\sqrt{2\pi}} \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} \frac{\cos(x) \cos(xy)}{\sqrt{|y|}} \ dy \ dx.$$

Apply the change of variables $$x=u, y=\frac{v}{u}$$ which has Jacobian $$\frac{\partial(x,y)}{\partial(u,v)}= \frac{1}{u}$$ to get

$$\frac{1}{\sqrt{2\pi}} \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} \frac{\cos(u) \cos(v)}{\sqrt{|uv|}} \ dv \ du.$$

Since the first half of the proof showed that this integral is $\sqrt{2 \pi},$ we see $$2\pi= \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} \frac{\cos(u) \cos(v)}{\sqrt{|uv|}} \ dv \ du= \left (\int_{-\infty}^{\infty} \frac{\cos(u)}{\sqrt{|u|}} \ du \right)^2.$$

So $$\sqrt{2 \pi}=\int_{-\infty}^{\infty} \frac{\cos(u)}{\sqrt{|u|}} \ du=2 \int_{0}^{\infty} \frac{\cos(u)}{\sqrt{u}} \ du. $$

As a result,

$$\frac{\sqrt{\pi}} {\sqrt{2}}= \int_{0}^{\infty} \frac{\cos(u)}{\sqrt{u}} \ du.$$ Lastly, let $u=z^2, du = 2z \ dz$ to get that

$$\frac{\sqrt{\pi}} {\sqrt{2}}= \int_{0}^{\infty} 2 \cos(z^2) \ dz,$$ so

$$\int_{0}^{\infty} \cos(z^2) \ dz =\frac{\sqrt{\pi}}{\sqrt{8}}.$$

  • 1
    OMG, you are a genius!!! This question is about Fourier Transforms basically.. I need to proof that the function $\frac{1}{\sqrt{x}}$ is self reciprocal, and when we do that we get that Fresnel integral and the teacher wants us to proof it! Thank you!!! – Rafael Vignoli Dec 09 '16 at 17:53
  • 1
    By self reciprocal I mean self adjoint I guess...English is not my native language, that is probably a false cognate – Rafael Vignoli Dec 09 '16 at 17:55
  • 1
    this: $$\sqrt{\frac{2}{\pi}}\int_0^\infty f(x)cos(tx)dx=f(t)=\sqrt{\frac{2}{\pi}}\int_0^\infty f(x)sin(tx)dx$$ $$f(x)=\frac{1}{\sqrt{x}}$$ – Rafael Vignoli Dec 09 '16 at 18:24
  • You are right, I wasn't familiar with the term self-reciprocal at all until I just looked it up. I thought about it as an eigenfunction. But self-reciprocal does not mean self-adjoint. – Vivek Kaushik Dec 09 '16 at 18:25
5

Hint: note that $$I=\int_{0}^{\infty}\cos\left(x^{2}\right)dx\stackrel{x^{2}=u}{=}\frac{1}{2}\int_{0}^{\infty}u^{-1/2}\cos\left(u\right)du $$ $$=\frac{1}{2\sqrt{\pi}}\int_{0}^{\infty}\cos\left(u\right)\int_{0}^{\infty}v^{-1/2}e^{-uv}dvdu $$ and now using the Fubini theorem we can exchange the integrals and get $$I=\frac{1}{2\sqrt{\pi}}\int_{0}^{\infty}v^{-1/2}\int_{0}^{\infty}\cos\left(u\right)e^{-uv}dudv=\textrm{Re}\left(\frac{1}{2\sqrt{\pi}}\int_{0}^{\infty}v^{-1/2}\int_{0}^{\infty}e^{u\left(i-v\right)}dudv\right) $$ $$\textrm{Re}\left(\frac{1}{2\sqrt{\pi}}\int_{0}^{\infty}\frac{v^{-1/2}}{v-i}dv\right)=\frac{1}{2\sqrt{\pi}}\int_{0}^{\infty}\frac{v^{1/2}}{v^{2}+1}dv\stackrel{z=\sqrt{v}}{=}\frac{1}{\sqrt{\pi}}\int_{0}^{\infty}\frac{z^{2}}{z^{4}+1}dz $$ and the last integral is simple to evaluate using partial fractions. Note that $$\int_{0}^{\infty}\frac{z^{2}}{z^{4}+1}dz=\frac{1}{2\sqrt{2}}\left(\int_{0}^{\infty}\frac{z}{z^{2}-\sqrt{2}z+1}-\frac{z}{z^{2}+\sqrt{2}z+1}dz\right) $$ $$=\frac{1}{2\sqrt{2}}\left(\lim_{a\rightarrow\infty}\frac{1}{2}\int_{0}^{a}\frac{2z-\sqrt{2}+\sqrt{2}}{z^{2}-\sqrt{2}z+1}-\frac{2z+\sqrt{2}-\sqrt{2}}{z^{2}+\sqrt{2}z+1}dz\right) $$ $$=\frac{1}{2\sqrt{2}}\left(\lim_{a\rightarrow\infty}\frac{1}{2}\int_{0}^{a}\frac{2z-\sqrt{2}}{z^{2}-\sqrt{2}z+1}-\frac{2z+\sqrt{2}-\sqrt{2}}{z^{2}+\sqrt{2}z+1}dz+\int_{0}^{a}\frac{\sqrt{2}}{z^{2}-\sqrt{2}z+1}dz\right)$$ and I think you can conclude by yourself from here.

Marco Cantarini
  • 33,062
  • 2
  • 47
  • 93
3

After using Euler's, use Gaussian Integral:

$$\int_{-\infty}^{\infty}e^{-a(x+b)^2}dx=\sqrt{ \frac{\pi}{a}}$$

or Integral of a Gaussian Function:

$$\int_{-\infty}^{\infty}ae^{-(x-b)^2/2c^2}dx=\sqrt{2}a\,|c|\sqrt{\pi}$$

BCLC
  • 13,459
0

$$ \begin{aligned} \int_{\infty}^{\infty}\left[\cos \left(t^2\right)-i \sin \left(t^2\right)\right] d t = & \int_{-\infty}^{\infty} e^{-t^2 i} d t \\ = & \int_{-\infty}^{\infty} e^{-\left[\frac{(1+i) t}{\sqrt{2}}\right]^2} d t \\ = & \frac{\sqrt{2}}{1+i} \sqrt{\pi} \\ = & \sqrt{\frac{\pi}{2}}(1-i) \end{aligned} $$ By comparing the real parts on both sides, we get $$ \int_{-\infty}^{\infty} \cos \left(t^2\right) d t=\sqrt{\frac{\pi}{2}} $$ Hence $$ \int_0^{\infty} \cos \left(t^2\right) d t=\frac{1}{2} \sqrt{\frac{\pi}{2}} $$

Lai
  • 20,421