# Contour integration problem

#### ijmbarr

1. Homework Statement
I am trying to evaluate the integral:
$$I = \int^{\infty}_{-\infty} \frac{z e^{irz}\ }{(z-k)(z+k)} dz$$

The way I attepted it was to use contour integration around a semicircle in the top half of the argan diagram, with two small indents above the poles. This means that the contour doesn't contain either pole, but has contributions from the indents around them.

I calculated the residue at each pole to be:

$$R(k) = e^{irk}/2$$

$$R(-k) = e^{-irk}/2$$

Using the method here: http://en.wikipedia.org/wiki/Residue_(complex_analysis)#Calculating_residues

From jordans lemma, the contribution from the semicircle tends to zero as its radius tends to infinity, and the small indents around each pole contribute $$- i\! \pi \! residue$$. The overall intergral should be:

$$I = i \pi (e^{irk}\ + e^{-irk})/2$$

However the answers I have say that the intergral is

$$I = i \pi e^{irk}$$

Is there an error somewhere that I'm missing? Does it depend where you choose the contour to be?

2. Homework Equations

3. The Attempt at a Solution

[Edited because of typos]

Last edited:
Related Calculus and Beyond Homework Help News on Phys.org

#### xepma

It's impossible for the residue to have a z-dependence, so try to explain in more detail how you calculate them. Then we can better judge where the error lies.

#### ijmbarr

Sorry, that was a typo, there should be no z there. The residues should read:
$$R(k) = e^{irk}/2$$

$$R(-k) = e^{-irk}/2$$

They were worked out using
$$res(f,c) = \frac{1}{(n-1)!} \lim_{z \to c} \frac{d^{n-1}}{dz^{n-1}}\left( f(z)(z-c)^{n} \right )$$

Both the poles I'm looking at are first order, n = 1. The overall integral is then:

$$R(-k) = i \pi (e^{irk} + e^{-irk})/2$$

Apologies for the mistake.

#### latentcorpse

you have two simple poles, one at +k and one at -k. Your contour is a smemicircle in the upper half plane which INCLUDES the pole at +k but doesn't include the one at -k.

The Cauchy Residue Theorem tells us that

$\int_{\Gamma} \frac{ze^{irz}}{(z-k)(z+k)}dz = 2 \pi i \sum_{j=1}^k res(f)(a_j)$

where $\Gamma$ is the contour and we only sum over the singularities contained INSIDE $\Gamma$ - by this means the pole at -k doesn't affect the value of your contour integral.

so we get $\int_{\Gamma} \frac{ze^{irz}}{(z-k)(z+k)}dz = 2 \pi i \frac{e^{irk}}{2}=i \pi e^{irk}$ using the value of the residue at z=+k.

then just argue that the part of your integral that's not on the real axis goes to 0 as R goes to infinity as you've done above and then its just the contribution on the real line that is left so you conclude

$\int_{-\infty}^{\infty} \frac{xe^{irx}}{(x-k)(x+k)}dx = i \pi e^{irk}$

#### ijmbarr

So why would you choose the contour to go around the pole at +k rather than both poles or none of them? does this mean the integral has several possible values?

Also, while the pole at -k isn't include in the contour, doesn't the half circle that goes around it, whose radius tends to zero have some contribution? As in "Mathematical methods for physicists", Riley Hobson Bence chapter 24?

#### ijmbarr

Any ideas? Is the integral Multivalued?

#### xepma

Have you tried calculating the contribution of the small arc and see if it tends to zero? Let me know if you need some help on it.

#### weejee

I think your result is correct.
It seems there is a deviation from k, such that $$k \rightarrow k+i\eta$$ where $$\eta$$ is a positve infinitesimal.
(i.e. the poles are a little bit off from the real axis)

Last edited:

#### latentcorpse

what page of chapter 24???

also if you see the integral you want the value of goes from $-\infty$ to $+\infty$ so part of your contour is going to need to lie on the real axis. If you draw the two poles and put part of your contour on the real axis, there is no way to now make a closed contour that contains both poles in its interior, is there?

secondly - the only reason we take a semi circle is because its relatively easy to deal with - you just use Jordan's lemma to see how the semi circular arc behaves. you could take a ridiculous contour that goes all over the complex plane and provided it includes that pole you'll get teh same answer but it's going to be basically impossible to evaluate.

hope this helps.

oh and the integrand isn't multivalued as far as i can tell...

#### ijmbarr

Thanks for the replies. The section I was referring to is page 864 RHB third edition.

I'm treating the small indents around the poles as semicircles A, so their contribution to the integral is:

$\int_{A} f(z) dz$

Where f(z) is the function being integrated. Expanding it as a laurent series:

$$\sum^{\infty}_{n = -m} a_{n} \int_{A} (z - z_{0})^{n} dz$$

Provided the pole is simple (as in this case) then the function can be written as the sum of a analytic function and $$a_{-1} (z - z_{0})^{-1}$$. When you carry out the integration, the analytic part vanishes and you're left with, written in polar form:

$\int_{A} f(z) dz = \int _{0} ^{\pi} a_{-1} \frac{i r e^{i\theta}}{re^{ir\theta}} d\theta$

Which just gives:

$\int_{A} f(z) dz = i \pi a_{-1}$

Where r is the radius of the semicircle and the extra $$ire^{ir\theta}$$ comes from the substitution of z = $$re^{ir\theta}$$.

Then as r tends to zero the integral along the real axis becomes continuous.

#### latentcorpse

i didn't read your initial post very thoroughly and started treating it with poles at +-ik.

i actually can't see any problems with what you've done.

i hadn't seen that result on p864. i would just have used jordan's lemma on both of the small semicircular deviations as well - guess you could give that a try...

#### ijmbarr

Doesn't jordan's lemma only apply as the radius tends to infinity?

#### weejee

As long as the poles(k,-k) lie exactly on the real axis and as long as we are dealing with the Cauchy principal value, ijmbarr's original result is correct.

The poles must be a little bit off from the real axis(k -> k+i*eta).
To be specific, the pole at k should be slightly higher than the real axis, and the one at -k should be slightly lower than it.
This often happens when we deal with the Green's function in scattering theory.

I think either you didn't read the book carefully enough or the book is wrong.

Last edited:

#### ijmbarr

Why do you choose the k+ to be higher and the k- pole to be lower? It seems like you could just as easily choose it the other way around and change the result?

#### weejee

Why do you choose the k+ to be higher and the k- pole to be lower? It seems like you could just as easily choose it the other way around and change the result?
I totally agree with you. One should specify in advance whether he wants to use k+i*eta or k-i*eta.

#### ijmbarr

I mean how does the physics of the situation determine which you choose?

#### latentcorpse

for those small indentations you should be able to apply Jordan's lemma with $r \rightarrow 0$

### Physics Forums Values

We Value Quality
• Topics based on mainstream science
• Proper English grammar and spelling
We Value Civility
• Positive and compassionate attitudes
• Patience while debating
We Value Productivity
• Disciplined to remain on-topic
• Recognition of own weaknesses
• Solo and co-op problem solving