4
$\begingroup$

I have been trying to calculate the following integral using the Residue Theorem but end up with an answer that seems to contain an imaginary part while the integral should be purely real: $$I_1=\int_0^\infty \frac{\cos(a x)}{x^4+b^4}\,dx$$ In order to calculate this integral, I consider the contour integral $$I_2=\int_C \frac{e^{iaz}}{z^4+b^4}\,dz$$ Where $C$ is the semi-circle of radius $R$ in the upper half plane parameterized counter-clockwise. As $R\to\infty$, the semi-circular integral vanishes, and a symmetry argument shows that $I_1=I_2/2$.

Now, we can use the Residue Theorem to compute $I_2$, which has poles at $z=be^{i\pi/4}$ and $z=be^{3i\pi/4}$. After some computations, I obtain \begin{align}I_1&=\frac{\pi i}{4b^3} \left( e^{-3\pi ab /4}e^{-i\pi/4}+e^{-\pi ab /4}e^{-3i\pi/4} \right)\\ &=\frac{\pi i}{4b^3} \left( e^{-3\pi ab /4}e^{-i\pi/4}-ie^{-\pi ab /4}e^{-i\pi/4} \right) \\ &=\frac{\pi i}{4b^3} e^{-i\pi/4}\left( e^{-3\pi ab /4}-ie^{-\pi ab /4} \right) \\ &=\frac{\pi}{4b^3} e^{i\pi/4}\left( e^{-3\pi ab /4}-ie^{-\pi ab /4} \right) \\ \end{align} Now, it seems that unless $a=0$ ( and $b=0$, but at that point the integral is undefined), there will always be a non-zero imaginary part, which shouldn't occur because, as I said earlier, the integrand is purely real and is being integral on a real interval. I noted that there was an interesting symmetry in the exponentials, but haven't found a way to use that.

I thought that it was possible that a mistake related to phase occurred, but I wouldn't be sure where.

Any ideas?

$\endgroup$
2
  • $\begingroup$ The problem seems to be in the computations of the residues (I find $-\frac{(1+i)e^{-\frac{(1-i) a b}{\sqrt{2}}}}{4\sqrt{2}b^3}$ for the first residue). btw to simplify the analysis note that by a substitution then if the integral is $f(a,b)$ then we have $f(a,b) = \frac{1}{|b|^3}f(|ab|,1)$ so you can take $b=1$ and then put the $b$-dependency back in the end using this scaling. $\endgroup$
    – Winther
    Dec 22 '18 at 21:49
  • 1
    $\begingroup$ For reference: if you compute the residues correctly and simplify you should end up with $\frac{\pi e^{-\frac{| a b| }{\sqrt{2}}} \sin\left(\frac{|ab|}{\sqrt{2}}+\frac{\pi }{4}\right)}{2|b|^3}$ $\endgroup$
    – Winther
    Dec 22 '18 at 22:06
2
$\begingroup$

Alright, I found an entirely real answer which agrees with Winther's comment. I originally made a mistake when calculating my residues. Here we go:

\begin{align} I_2&= 2\pi i \sum \text{Res}_{\text{Im(}z)\ge 0 } \,f(z) \\ &= \frac{2\pi i}{4b^3}\left(\frac{e^{iabe^{3\pi i/4}}}{e^{i\pi /4}} + \frac{e^{iabe^{\pi i/4}}}{e^{3i\pi /4}} \right) \\ &= \frac{2\pi i}{4b^3}\left(e^{iab\left(-\frac{\sqrt{2}}{2}+i\frac{\sqrt{2}}{2}\right)-\frac{i\pi}{4}} + e^{iab\left(\frac{\sqrt{2}}{2}+i\frac{\sqrt{2}}{2}\right)-\frac{3i\pi}{4}} \right) \\ &=\frac{2\pi i}{4b^3}e^{-\frac{ab}{\sqrt{2}}} \left( e^{-i\left[\frac{ab}{\sqrt{2}} +\frac{\pi}{4}\right]}+ e^{i\left[\frac{ab}{\sqrt{2}} -\frac{3\pi}{4}\right]} \right) \\ &=\frac{2\pi i}{4b^3}e^{-\frac{ab}{\sqrt{2}}} \left( e^{-i\left[\frac{ab}{\sqrt{2}} +\frac{\pi}{4}\right]}- e^{i\left[\frac{ab}{\sqrt{2}} +\frac{\pi}{4}\right]} \right) \\ &=\frac{2\pi i}{4b^3}e^{-\frac{ab}{\sqrt{2}}} \left( -2 i\sin \left(\frac{ab}{\sqrt{2}} +\frac{\pi}{4}\right) \right) \\ &= \frac{\pi}{b^3} e^{-\frac{ab}{\sqrt{2}}}\sin \left(\frac{ab}{\sqrt{2}} +\frac{\pi}{4}\right) \end{align} using $I_1=I_2/2$ and demanding that $I(a)=I(-a)$ and $I(b)=I(-b)$, we finally obtain

$$I_1=\frac{\pi}{2b^3} e^{-\frac{|ab|}{\sqrt{2}}}\sin \left(\frac{|ab|}{\sqrt{2}} +\frac{\pi}{4}\right)$$

$\endgroup$
2
  • $\begingroup$ I think your forgot the b in the exponent on row 2 and 3 (as a typo I mean because you have it again on row 4) $\endgroup$
    – Ryukyu
    Dec 22 '18 at 23:26
  • $\begingroup$ Yes, I just fixed it. Thank you for the comment. $\endgroup$
    – Zachary
    Dec 23 '18 at 3:02
1
$\begingroup$

Not really an answer, just making things as compact as possible

First of all note that $$e^{i\pi/4}=\frac{1+i}{\sqrt2}$$ Hence $$I_1=\frac{\pi(1+i)}{4b^3\sqrt2}\bigg(\frac1{e^{3ab\pi/4}}-\frac{i}{e^{ab\pi/4}}\bigg)$$ Multiplying the RHS by $\frac{\exp(ab\pi/4)}{\exp(ab\pi/4)}$, $$I_1=\frac{\pi(1+i)}{4b^3e^{ab\pi/4}\sqrt2}\bigg(e^{\frac{ab\pi}4(1-3)}-ie^{\frac{ab\pi}4(1-1)}\bigg)$$ $$I_1=\frac{\pi(1+i)}{b^3e^{ab\pi/4}2^{5/2}}(e^{-ab\pi/2}-i)$$ Now let $$p=\frac\pi{b^3e^{ab\pi/4}2^{5/2}}\\ q=e^{-ab\pi/2}$$ So $$I_1=p(1+i)(q-i)$$ Thus $$I_1=p(1+q)+ip(1-q)$$ So now you have a distinct real and imaginary part. I guess you have to make more assumptions about the restrictions on $a$ and $b$ if you want $\text{Im}I_1=0$. Either that or you made a mistake.

$\endgroup$

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.