5
$\begingroup$

I was searching around the web for some information about integrals and I came across the formula:

$$\int_{-\infty}^\infty \frac{\ln(x^2)e^{\frac{-x^2}{2\sigma}}}{(2\pi)^\frac{1}{2}\sigma}dx= \ln(\sigma^2)-\gamma-\ln(2)$$

$\gamma =$ the Euler-Mascheroni Constant

I'm very unsure where the Euler-Mascheroni constant came from. I tried rearranging the integral to simpler terms but I end up getting:

$$\int_{-\infty}^\infty \ln|x|e^{-x^2}dx$$

which isn't overtly integrable. Where does this formula come from?

$\endgroup$
8
  • $\begingroup$ Your rearrangement is problematical, since $\log x$ is not defined for $x\le0$. $\endgroup$ – Gerry Myerson Oct 31 '17 at 4:02
  • $\begingroup$ If you used the fact that $\ln(x^2) = 2 \ln x$ while simplifying, then note that for $x$ negative, it must change to $2 \ln |x|$. $\endgroup$ – Teresa Lisbon Oct 31 '17 at 4:04
  • 1
    $\begingroup$ Are you very sure the exponent should not be $-x^2/(2\sigma^2)\text{ ?} \qquad$ $\endgroup$ – Michael Hardy Oct 31 '17 at 4:22
  • 2
    $\begingroup$ If we assume that this was supposed to say $$\int_{-\infty}^\infty \frac{\ln(x^2)e^{-x^2/(2\sigma^2)}}{(2\pi)^{1/2} \sigma}dx= \ln(\sigma^2)-\gamma-\ln(2),$$ then in probabilistic language, it says that if $X\sim N(0,\sigma^2),$ i.e. $X$ is normally distributed with expected value $0$ and standard deviation $\sigma,$ then $$\operatorname{E}(2\log|X|) = 2\log\sigma - \gamma-\log 2.$$ $\endgroup$ – Michael Hardy Oct 31 '17 at 4:26
  • 1
    $\begingroup$ $$ \begin{align} \int_{-\infty}^\infty \frac{\ln(x^2)e^{-x^2/(2\sigma^2)}}{(2\pi)^{1/2} \sigma}dx & = 2\log\sigma - \int_{-\infty}^\infty 2\log|(x/\sigma)| \frac{e^{(-1/2)(x/\sigma)^2}}{\sqrt{2\pi}} \, \frac{dx} \sigma \\ \\ & = 2\log \sigma - \int_{-\infty}^\infty 2\log|u| \frac{e^{-(1/2)u^2}}{\sqrt{2\pi}} \,du. \end{align} $$ Thus it is easily seen that the value of the integral is $2\log\sigma$ plus something not depending on $\sigma. \qquad$ $\endgroup$ – Michael Hardy Oct 31 '17 at 4:36
3
$\begingroup$

Throw out unecessary constants and since the integrand is even, the problem reduces to evaluate: $$I=\int_0^\infty \ln(x^2) e^{-ax^2} dx$$

Enforcing the subsitution $x=(u/a)^{1/2}$ gives $$I = \frac{1}{2\sqrt{a}} \left[\int_0^\infty u^{-1/2} e^{-u}\ln u \, du - \ln a \int_0^\infty u^{-1/2}e^{-u}\ du \right]$$

The first integral is just $\Gamma'(1/2)=\sqrt{\pi}(-\gamma-2\ln 2)$, the second integral is $\Gamma(1/2)=\sqrt{\pi}$.

$\endgroup$
3
  • $\begingroup$ I tried doing so research, but do you know how to prove $\frac{d}{dx}[\Gamma(1/2)] = (\pi)^\frac{1}{2}(-\gamma-2\ln(2))$ Or a link to somewhere with the proof? $\endgroup$ – Tom Himler Oct 31 '17 at 16:35
  • $\begingroup$ The simplest way is possibly via the digamma function $\psi(z) = \Gamma'(z)/\Gamma(z)$. The digamma function satisfies $$\psi(1+z) = -\gamma + \int_0^1 \frac{1-x^z}{1-x} dx $$ from which you can calculate $\psi(1/2)$ easily. You can learn more about these functions by consulting relevant books, online resources sometimes only list formula but no proofs. $\endgroup$ – pisco Oct 31 '17 at 17:46
  • $\begingroup$ Thanks, I really appreciate it. $\endgroup$ – Tom Himler Nov 1 '17 at 3:06

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.