Hi I am trying to solve this double integral $$ I:=\int_0^\infty \int_0^\infty \frac{\log x \log y}{\sqrt {xy}}\cos(x+y)\,dx\,dy=(\gamma+2\log 2)\pi^2. $$ Thank you.

The constant in the result is given by $\gamma\approx .577$, and is known as the Euler-Mascheroni constant. I was thinking to write $$ I=\Re \bigg[\int_0^\infty \int_0^\infty \frac{\log x \log y}{\sqrt{xy}}\, e^{i(x+y)}\, dx\, dy\bigg] $$ and using Leibniz's rule for differentiation under the integral sign to write $$ I(\eta, \xi)=\Re\bigg[ \int_0^\infty \int_0^\infty \ \frac{\log (\eta x)\log(\xi y)}{\sqrt{xy}} e^{i(x+y)}dx\,dy. \bigg]\\ $$ After taking the derivatives it became obvious that I need to try another method since the x,y constants cancel out. How can we solve this integral I? Thanks.


Using the identity


The integral can be written $$ I=\int_0^\infty \int_0^\infty \frac{\log x \log y}{\sqrt {xy}}\left(\cos(x)\cos(y)-\sin(x)\sin(y)\right)\,dx\,dy $$

Now by splitting the integrals

$$\int_0^\infty \int_0^\infty \frac{\log x \log y}{\sqrt {xy}}\cos(x)\cos(y)\,dx\,dy-\int_0^\infty \int_0^\infty \frac{\log x \log y}{\sqrt {xy}}\sin(x)\sin(y)\,dx\,dy $$

Notice by symmetry of the integrals we have

$$\left(\int^\infty_0 \frac{\log x }{\sqrt {x}}\cos(x)\,dx \right)^2-\left(\int^\infty_0 \frac{\log x }{\sqrt {x}}\sin(x)\,dx \right)^2 $$

Both inegrals are solvable by using the mellin transforms

$$\int^\infty_0 x^{s-1}\sin(x)\,dx = \Gamma (s) \sin\left( \frac{\pi s}{2} \right)$$

$$\int^\infty_0 x^{s-1}\cos(x)\,dx = \Gamma (s) \cos\left( \frac{\pi s}{2} \right)$$

By differentiation under the integral sign and using $s=\frac{1}{2}$.

$$\int^\infty_0 \frac{\log x }{\sqrt {x}}\cos(x)\,dx =-\frac{1}{2} \sqrt{\frac{π}{2}} \left(2 \gamma +π+\log(16) \right) $$

$$\int^\infty_0 \frac{\log x }{\sqrt {x}}\sin(x)\,dx=\frac{1}{2} \sqrt{\frac{π}{2}} (-2 \gamma +π- \log(16)) $$

Collecting the results together we have

$$I=(\gamma+2\log 2)\pi^2$$



$$\int_0^{\infty} dx \, x^{\alpha} e^{i x}$$

We know from Cauchy's theorem that this integral is equal to (when it converges)

$$i \, e^{i \pi \alpha/2} \int_0^{\infty} du \, u^{\alpha} \, e^{-u} = i \, e^{i \pi \alpha/2} \, \Gamma(\alpha+1)$$

Differentiating both sides with respect to $\alpha$, we get

$$\int_0^{\infty} dx \, x^{\alpha} e^{i x}\, \log{x} = \Gamma(\alpha+1) e^{i \pi \alpha/2} \left [i \, \psi(\alpha+1)-\frac{\pi}{2} \right ] $$

Square both sides:

$$\begin{align}\int_0^{\infty} dx \, x^{\alpha} e^{i x}\, \log{x} \int_0^{\infty} dy \, y^{\alpha} e^{i y}\, \log{y} &= \Gamma(\alpha+1)^2 e^{i \pi \alpha} \left [\frac{\pi^2 }{4}-\psi(\alpha+1)^2-i \pi \psi(\alpha+1) \right ] \end{align}$$

Now plug in $\alpha=-1/2$ and consolidate; use the fact that $\Gamma(1/2)=\sqrt{\pi}$ and $\psi(1/2)=-\gamma-2 \log{2}$:

$$\int_0^{\infty} dx \, \int_0^{\infty} dy \frac{\log{x} \log{y}}{\sqrt{x y}} e^{i (x+y)} = -i \pi \left [\frac{\pi^2}{16} - (\gamma+2 \log{2})^2 + i \pi (\gamma+2 \log{2}) \right ]$$

Take the real part of both sides, and get

$$\int_0^{\infty} dx \, \int_0^{\infty} dy \frac{\log{x} \log{y}}{\sqrt{x y}} \cos{(x+y)} = \pi^2 (\gamma+2 \log{2}) $$

as was to be shown.

  • 1
    $\begingroup$ +1, Of course, assuming that the considered integral does converge and we have the right to use Fubini's theorem. $\endgroup$ – Omran Kouba May 12 '14 at 8:39
  • $\begingroup$ @OmranKouba: Yes, I should have at least mentioned this. Very good point. Thankfully, it does. $\endgroup$ – Ron Gordon May 12 '14 at 8:41
  • $\begingroup$ Do you have proof that the the integral converges? YOu say "when it converges" $\endgroup$ – Jeff Faraci May 13 '14 at 20:17
  • $\begingroup$ @Integrals: it converges for certain $\alpha$ by Cauchy's theorem, i.e., $\operatorname{Re}{\alpha} \gt -1$. $\endgroup$ – Ron Gordon May 13 '14 at 20:22

The integral is $$I=\Re\left(\left(\int_0^{\infty} \frac{\ln x}{\sqrt{x}}e^{ix}\,dx\right)^2\right)$$ Evaluating the definite integral first: $$J=\int_0^{\infty} \frac{\ln x}{\sqrt{x}}e^{ix}\,dx$$ Use the substitution $\sqrt{x}=e^{i\pi/4}t$ to obtain: $$J=2e^{i\pi/4}\int_0^{\infty} e^{-t^2}\ln(t^2e^{i\pi/2})\,dt=2e^{i\pi/4}\int_0^{\infty} \left(2\ln te^{-t^2}+\frac{i\pi}{2}e^{-t^2}\right)\,dt$$ Using the following results: $$\int_0^{\infty} e^{-t^2}\ln t\,dt=-\frac{\sqrt{\pi}}{4}(\gamma+2\ln2)$$ $$\int_0^{\infty} e^{-t^2}\,dt=\frac{\sqrt{\pi}}{2}$$ ....we obtain: $$J=2e^{i\pi/4}\left(-\frac{\sqrt{\pi}}{2}(\gamma+2\ln 2)+\frac{i\pi}{2}\frac{\sqrt{\pi}}{2}\right)$$ $$\Rightarrow e^{i\pi/4}\sqrt{\pi}\left(\frac{i\pi}{2}-(\gamma+2\ln 2)\right)$$ Squaring $J$ and taking the real part, $$I=\pi^2(\gamma+2\ln 2)$$

  • $\begingroup$ +1 and checked off as answer since you do not use any special functions! Thanks Pranav. However, do you have proof of $$ \int_0^{\infty} e^{-t^2}\ln t\,dt=-\frac{\sqrt{\pi}}{4}(\gamma+2\ln2) $$ $\endgroup$ – Jeff Faraci May 12 '14 at 15:36
  • $\begingroup$ Honestly, I am not sure about its proof. I found it in the list of integral representations for the constant here: mathworld.wolfram.com/Euler-MascheroniConstant.html . I did found a method to prove it though but it relies on gamma and digamma functions which I haven't yet learned. I first evaluated the integral $\int_0^{\infty} x^{n}e^{-x^2}\,dx$ using wolfram alpha. It came out to be $\frac{\Gamma((n+1)/2)}{2}$. Differentiate with respect to $n$ and substitute $n=0$ to obtain the definite integral. I have no idea about how much work it takes to do the above steps by hand. $\endgroup$ – Pranav Arora May 12 '14 at 16:02
  • $\begingroup$ Now I feel that it looks exactly like Ron Gordon's method, I suggest marking his solution as the answer. $\endgroup$ – Pranav Arora May 12 '14 at 16:04
  • $\begingroup$ @integrals, the solution must use a special function to get the euler constant . $\endgroup$ – Zaid Alyafeai May 12 '14 at 22:24
  • $\begingroup$ @ZaidAlyafeai Good point! Thank you Pranav, but your solution is still OK. However, i marked the other solution. $\endgroup$ – Jeff Faraci May 13 '14 at 14:13

$\newcommand{\+}{^{\dagger}} \newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\down}{\downarrow} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\isdiv}{\,\left.\right\vert\,} \newcommand{\ket}[1]{\left\vert #1\right\rangle} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large a}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert} \newcommand{\wt}[1]{\widetilde{#1}}$ $\ds{I\equiv\int_{0}^{\infty}\int_{0}^{\infty} {\ln\pars{x}\ln\pars{y} \over \root{xy}}\,\cos\pars{x + y}\,\dd x\,\dd y =\bracks{\gamma + 2\ln\pars{2}}\pi^{2}:\ {\large ?}}$

\begin{align} I&=\Re\int_{0}^{\infty}\int_{0}^{\infty} {\ln\pars{x}\ln\pars{y} \over \root{xy}}\,\expo{\ic\pars{x + y}}\,\dd x\,\dd y =\Re\braces{\bracks{\color{#c00000}{\int_{0}^{\infty} {\ln\pars{x} \over \root{x}}\,\expo{\ic x}\,\dd x}}^{2}} \end{align}

\begin{align} &\color{#c00000}{\int_{0}^{\infty} {\ln\pars{x} \over \root{x}}\,\expo{\ic x}\,\dd x} =\lim_{\mu \to -1/2}\partiald{}{\mu}\ \overbrace{\int_{0}^{\infty}x^{\mu}\expo{\ic x}\,\dd x} ^{\ds{t\ \equiv\ -\ic x\ \imp\ x\ =\ \ic t}}\ \\[3mm]&=\lim_{\mu \to -1/2}\partiald{}{\mu} \int_{0}^{-\ic\infty}\expo{\ic\pi\mu/2}t^{\mu}\expo{-t}\,\ic\,\dd t \\[3mm]&=\ic\lim_{\mu \to -1/2}\partiald{}{\mu}\braces{\expo{\ic\pi\mu/2}\bracks{% \int_{0}^{\infty}t^{\mu}\expo{-t}\,\dd t -\overbrace{\left.\lim_{R \to \infty}\int_{-\pi/2}^{0}z^{\mu}\expo{-z}\,\dd z\, \right\vert_{z\ \equiv\ R\expo{\ic\theta}}}^{\ds{=\ 0}}}} \\[3mm]&=\ic\lim_{\mu \to -1/2}\partiald{}{\mu} \bracks{\expo{\ic\pi\mu/2}\Gamma\pars{\mu + 1}} \end{align} where $\ds{\Gamma\pars{z}}$ is the Gamma Function ${\bf\mbox{6.1.1}}$.

\begin{align} I&=\color{#c00000}{\int_{0}^{\infty}{% \ln\pars{x} \over \root{x}}\,\expo{\ic x}\,\dd x} =\ic\lim_{\mu \to -1/2} \bracks{\expo{\ic\pi\mu/2}\,{\ic\pi \over 2}\,\Gamma\pars{\mu + 1} +\expo{\ic\pi\mu/2}\Gamma\pars{\mu + 1}\Psi\pars{\mu + 1}} \end{align} where $\ds{\Psi\pars{z}}$ is the Digamma Function ${\bf\mbox{6.3.1}}$.

\begin{align} I&=\color{#c00000}{\int_{0}^{\infty}{% \ln\pars{x} \over \root{x}}\,\expo{\ic x}\,\dd x} =\ic\expo{-\ic\pi/4}\Gamma\pars{\half} \bracks{{\ic\pi \over 2} + \Psi\pars{\half}} \\[3mm]&=\root{\pi \over 2}\pars{1 + \ic}\bracks{{\ic\pi \over 2} - \gamma - 2\ln\pars{2}} \end{align} $\ds{\gamma}$ is the Euler-Mascheroni Constant ${\bf\mbox{6.1.3}}$ and we used the identities $\ds{\Gamma\pars{\half} = \root{\pi}}$ and $\ds{\Psi\pars{\half}=-\gamma - 2\ln\pars{2}}$.

\begin{align} I&=\Re\int_{0}^{\infty}\int_{0}^{\infty} {\ln\pars{x}\ln\pars{y} \over \root{xy}}\,\expo{\ic\pars{x + y}}\,\dd x\,\dd y \\[3mm]&=\Re\pars{\braces{\root{\pi \over 2}\pars{1 + \ic}\bracks{{\ic\pi \over 2} - \gamma - 2\ln\pars{2}}}^{2}} \\[3mm]&=\Re\pars{{\pi \over 2}\,2\ic\braces{\bracks{\gamma + 2\ln\pars{2}}^{2} - {\pi^{2} \over 4} - \ic\pi\bracks{\gamma + 2\ln\pars{2}}}} \end{align}

$$\color{#00f}{\large% I\equiv\int_{0}^{\infty}\int_{0}^{\infty} {\ln\pars{x}\ln\pars{y} \over \root{xy}}\,\cos\pars{x + y}\,\dd x\,\dd y =\bracks{\gamma + 2\ln\pars{2}}\pi^{2}} $$


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.