I'm wondering how to integrate the so-called integral using Residue theorem,as it has a pole of second order on the real axis(not simple) so we cannot use $\pi i Res(@ z=0)$.Would you please give me a hint?($a,b>0$)
-
How about this one? – MAh2014 Jan 19 '16 at 11:34
-
Another form is $$-\frac{1}{2}I=\int_0^\infty \frac{\sin\left(\frac{1}{2}(a+b)x\right)\sin\left(\frac{1}{2}(a-b)x\right)}{x^2}{dx}.$$ – pshmath0 Jan 19 '16 at 11:38
-
Can you not use a keyhole contour? – pshmath0 Jan 19 '16 at 11:42
-
I just want it to be solved!what should we do after this? – MAh2014 Jan 19 '16 at 11:48
-
Related: http://math.stackexchange.com/questions/61828 (but not via residue calculus) – mrf Jan 19 '16 at 13:02
-
You can check that now $x=0$ is a simple pole on real-axis. – Wang Kah Lun Jan 19 '16 at 13:46
5 Answers
Consider the integral
$$\oint_C dz \frac{e^{i a z}-e^{i b z}}{z^2} $$
where $C$ is a semicircle in the upper half plane of radius $R$ with a small semicircular detour at $z=0$ of radius $\epsilon$ into the upper half plane. Then the integral is equal to
$$PV \int_{-R}^R dx \frac{e^{i a x}-e^{i b x}}{x^2} +i R \int_0^{\pi} d\theta \, e^{i \theta} \frac{e^{i a R e^{i \theta}} - e^{i b R e^{i \theta}}}{R^2 e^{i 2 \theta}} + i \epsilon \int_{\pi}^0 d\phi \, e^{i \phi} \frac{e^{i a \epsilon e^{i \phi}} - e^{i b \epsilon e^{i \phi}}}{\epsilon^2 e^{i 2 \phi}}$$
where $PV$ denotes the Cauchy principal value of the integral.
As $R \to \infty$, the second integral vanishes. This is so because the magnitude of the integral is bounded by
$$\frac1{R} \int_0^{\pi} d\theta \, e^{-\min{(a,b)} R \sin{\theta} } \le \frac{2}{R} \int_0^{\pi/2} d\theta \, e^{-2 \min{(a,b)} R \theta/\pi} \le \frac{\pi}{\min{(a,b)} R^2}$$
We now consider the third integral as $\epsilon \to 0$. In this case, we Taylor expand the numerator and find that the integral has a limit:
$$i \epsilon \int_{\pi}^0 d\phi \, e^{i \phi} \frac{i a \epsilon e^{i \phi} - i \epsilon b e^{i \phi}}{\epsilon^2 e^{i 2 \phi}} = -\pi(b-a)$$
By Cauchy's theorem, the contour integral is zero. Thus, in these limits - and taking the real part of both sides - we find that
$$\int_{-\infty}^{\infty} dx \frac{\cos{ax}-\cos{b x}}{x^2} = \pi (b-a) $$
or
$$\int_{0}^{\infty} dx \frac{\cos{ax}-\cos{b x}}{x^2} = \frac{\pi}{2} (b-a) $$

- 138,521
-
I have problem understanding these inequalities$\frac1{R} \int_0^{\pi} d\theta , e^{-\min{(a,b)} R \sin{\theta} } \le \frac{2}{R} \int_0^{\pi/2} d\theta , e^{-2 \min{(a,b)} R \theta/\pi} \le \frac{\pi}{\min{(a,b)} R^2}$ – MAh2014 Jan 19 '16 at 12:05
-
-
I think there must be a $cos (\theta)$in the exponent of $e^{-\min{(a,b)} R \sin{\theta} } \le \frac{2}{R} \int_0^{\pi/2} d\theta $?!!! – MAh2014 Jan 19 '16 at 12:18
-
@MAh2014: Nope. This is an estimate, and one of the exponents dominates the other one over the integration region. The cosine piece then gets eliminated because we are taking magnitudes of the integrand (which is justified by Cauchy-Schwartz). – Ron Gordon Jan 19 '16 at 12:21
-
-
1@MAh2014: I have no idea what you are asking. Look at the inequality I posted in my first comment. Aside from that, I am not a tutoring service and have no time to hold your hand through this. If this is a problem in a course in complex analysis, you should be able to follow what I have provided. If you are having trouble, then consult a teacher or a tutor. – Ron Gordon Jan 19 '16 at 12:37
-
Sorry I thought you have transform $\theta$ to $2 \theta $ to change the interval to $\frac{\pi}{2}$ but I've got it now :). – MAh2014 Jan 19 '16 at 13:22
Double integral $$ \begin{aligned}\int_0^{\infty} \frac{\cos (a x)-\cos (b x)}{x^2} d x = & \int_0^{\infty} \frac{1}{x} \int_a^b \sin (x y) d y d x \\ = & \int_a^b \int_0^{\infty} \frac{\sin (x y)}{x} d x d y \\ = & \int_a^b \frac{\pi}{2} d y\\=&\frac{\pi}{2}(b-a) \end{aligned} $$

- 20,421
It can be seen as a limit case of Frullani's theorem, but it also follows from: $$\int_{0}^{+\infty}\left(\frac{\sin x}{x}\right)^2\,dx \stackrel{i.b.p.}{=}\int_{0}^{+\infty}\frac{\sin(2x)}{x}\,dx=\frac{\pi}{2} $$ since: $$ \int_{0}^{+\infty}\frac{1-\cos(ax)}{x^2}\,dx = 2\int_{0}^{+\infty}\left(\frac{\sin\frac{ax}{2}}{x}\right)^2\,dx. $$

- 353,855
Using the trigonometric identity in my comment above... Let $y=1/x$. Then $-dx/x^2=dy$ and we have
$$-\frac{1}{2}I=\int_0^{\infty}\sin\left(\frac{1}{2}(a+b)/y\right)\sin\left(\frac{1}{2}(a-b)/y\right)dy.$$
Mathematica gives the primitve to be, which can probably be obtained using i.b.p., $$-\frac{a \text{Si}\left(\frac{a}{y}\right)}{2}-\frac{1}{2} y \cos \left(\frac{a}{y}\right)+\frac{b \text{Si}\left(\frac{b}{y}\right)}{2}+\frac{1}{2} y \cos \left(\frac{b}{y}\right),$$ where $\text{Si}$ is the sine integral, which evaluates over the limits to be $$I=\frac{\pi}{2}\left(\left|b\right|-\left|a\right|\right)$$ for all $a,b\in\mathbb{R}$. Not a proof, but an alternative outline to the residue calculus perhaps.

- 10,565
By integration by part, we reduce the power 2. $$ \begin{aligned} & \int_{0}^{\infty} \frac{\cos (a x)-\cos (b x)}{x^{2}} d x \\ =&-\int_{0}^{\infty}[\cos (a x)-\cos (b x)] d\left(\frac{1}{x}\right) \\ =&\left.-\left[\frac{\cos (a x)-\cos b x}{x}\right]_{0}^{\infty}+\int_{0}^{\infty} \frac{-a \sin (a x)+b \sin (b x)}{x}dx\right]\\=&-a \int_{0}^{\infty} \frac{\sin (a x)}{x} d x+b\int_{0}^{\infty} \frac{\sin (b x)}{x} d x \end{aligned} $$ Using the formula, $$ \int_{0}^{\infty} \frac{\sin (k x)}{x} d x=\frac{\pi}{2} \operatorname{sgn}(k), $$ We can conclude that $$ \int_{0}^{\infty} \frac{\cos (a x)-\cos (b x)}{x^{2}} d x =\frac{\pi}{2}(|b|-|a|) $$

- 20,421