1
$\begingroup$

(H. Priestley complex Analysis Chapter 7 Exercise 9)

Suppose $f$ is holomorphic inside and on $\gamma(0,1)$. By integration around the usual keyhole like this one : Integration of $\ln $ around a keyhole contour , prove that:

$$\int_0^1 f(x)\,dx=\frac{1}{2i \pi} \int_{\gamma(0,1)} f(z) (\ln(z) - i \pi)\, dz$$

where $\ln(z)$ denotes the branch of the logarithm with imaginary part between $[0,2\pi]$


I am not sure how to tackle this question :(

the branch of the $\ln$ will be $\ln(z) = \ln (|z|) + i \theta$

I thought that maybe if I integrate $f(z) (\ln(z) - i \pi)$ along the entire contour I will get $0$ by Cauchy's theorem as it is holomorphic inside by design:

$$ \int_{\gamma(0,1)} f(z) (\ln(z) - i \pi)\, dz + \int_{\text{top edge}} f(z) (\ln(z) - i \pi)\, dz + \int_{\text{bottom edge}} f(z) (\ln(z) - i \pi)\, dz - \int_{\gamma(0,\epsilon)} f(z) (\ln(z) - i \pi)\, dz = 0,$$

i.e.

$$ \int_{\gamma(0,1)} f(z) (\ln(z) - i \pi) i\, dz + \int_{0}^{1} f(x) (\ln(x) - i \pi)\, dx + \int_{1}^{0} f(x) (\ln(x)+2i \pi - i \pi)\, dx - \int_{\gamma(0,\epsilon)} f(\epsilon e^{i\theta}) (\ln(\epsilon) +i\theta - i \pi)\, dz = 0.$$

That is, with a fortunate cancellation:

$$ \int_{\gamma(0,1)} f(z) (\ln(z) - i \pi)\, dz -2i \pi \int_{0}^{1} f(x)\, dx - \int_{\gamma(0,\epsilon)} f(\epsilon e^{i\theta}) (\ln(\epsilon) +i\theta - i \pi) \epsilon i e^{i\theta}\, d\theta = 0.$$

however I'm not sure about what happens with the very last integral, if the integrand were holomorphic apart from a simple pole then it would just be $2\pi \times\text{residue}$, i.e. $0$ since $f$ is holomorphic inside but with the $\ln()$ it makes it complicated...

$\endgroup$
0
$\begingroup$

You have tackled it correctly, all that is missing is that you see

$$\lim_{\epsilon \to 0} \int_{\gamma(0,\epsilon)} f(z)(\ln z - i\pi)\,dz = 0.$$

The standard estimate

$$\biggl\lvert \int_\Gamma g(z)\,dz\biggr\rvert \leqslant L(\Gamma)\cdot \sup \{ \lvert g(\zeta)\rvert : \zeta \in \operatorname{Tr}(\Gamma)\}$$

yields that, since $f$ is bounded, and

$$\lim_{\epsilon \to 0} \epsilon\cdot \lvert\ln\epsilon\rvert = 0.$$

The length of $\gamma(0,\epsilon)$ is bounded by $2\pi\epsilon$, and the integrand is bounded by $\lVert f\rVert_\infty(\pi + \lvert\ln\epsilon\rvert)$.

$\endgroup$
  • $\begingroup$ Of course ! Many thanks $\endgroup$ – user3203476 Jul 5 '15 at 13:57

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.