7

Given that $\int_{-\infty}^\infty e^{-x^2} \, dx=\sqrt{\pi}$. Find the value of $$\int_{-\infty}^\infty\int_{-\infty}^\infty e^{-(x^2+xy+y^2)} \, dx\,dy$$

I don't understand how I find this double integral by using the given data. Please help.

StubbornAtom
  • 17,052
A.D
  • 6,400
  • 1
  • 20
  • 43
  • 7
    Hints: (i) use a change of variables to find the value of $\int_{-\infty}^\infty e^{-(x+a)^2},dx$. (ii) write $x^2 + xy + y^2$ as $(x+y/2)^2 + 3y^2/4$ and then use the answer of (i) while doing the inner integral. – Dilip Sarwate May 07 '13 at 18:30
  • A more general version of the integral:https://math.stackexchange.com/questions/653159/if-a-is-positive-definite-then-int-mathbbrn-mathrme-langle-ax-x?noredirect=1&lq=1. – StubbornAtom Jan 17 '18 at 18:20

3 Answers3

13

For fun, I want to point out that much more can be said. The spectral theorem for real symmetric matrices tells us that real symmetric matrices are orthogonally diagonalizable. Thus, if $A$ is some symmetric matrix, then there exists an orthogonal matrix $U$ and diagonal matrix $D$ such that $A$ can be written as $A=U^{-1} DU$. Hence $x^TAx=x^TU^TDUx=(Ux)^TD(Ux)$. The Jacobian det of the transformation $x\mapsto Ux$ is simply $1$, so we have a change of variables $u=Ux$:

$$\begin{align} \int_{{\bf R}^n}\exp\left(-x^TAx\right)\,dV & =\int_{{\bf R}^n}\exp\left(-(\lambda_1u_1^2+\cdots+\lambda_nu_n^2)\right)\,dV \\[6pt] & =\prod_{i=1}^n\int_{-\infty}^{+\infty}\exp\left(-\lambda_i u_i^2\right)du_i \\[6pt] & =\prod_{i=1}^n\left[\frac{1}{\sqrt{\lambda_i}}\int_{-\infty}^{+\infty}e^{-u^2} \, du\right] \\[6pt] & =\sqrt{\frac{\pi^n}{\det A}}.\end{align}$$

Note that we don't even have to compute $U$ or $D$. The above calculation is the generalized version of the "completing the squares" approach when $n=2$ (which is invoked elsewhere in this thread).

This formula is in fact the basis for the Feynman path integral formulation of the functional determinant from quantum field theory; since technically the integral diverges we need to compare them instead of looking at individual ones outright. Wikipedia has some more details.

anon
  • 151,657
  • Neat observation; certainly more insightful than my answer :-) – robjohn May 07 '13 at 18:54
  • @robjohn Well, the observation clearly does not originate with me. What's even more amazing is the fact that Feynman path integrals and zeta regularization agree with each other on functional determinants (see link at the end of this answer). Why this is, though, is a mystery to me, and is the subject of this question of mine. – anon May 07 '13 at 19:03
  • The reason your integral signs were so tiny and otherwise formatted as if they were inline rather than displayed is that you used "array" rather than "align". I edited. – Michael Hardy Jul 06 '20 at 21:58
13

$$ \begin{align} \int_{-\infty}^\infty\int_{-\infty}^\infty e^{-(x^2+xy+y^2)}\,\mathrm{d}x\,\mathrm{d}y &=\int_{-\infty}^\infty\int_{-\infty}^\infty e^{-((x+y/2)^2+3y^2/4)}\,\mathrm{d}x\,\mathrm{d}y\\ &=\int_{-\infty}^\infty\int_{-\infty}^\infty e^{-(u^2+3y^2/4)}\,\mathrm{d}u\,\mathrm{d}y\\ &=\int_{-\infty}^\infty e^{-u^2}\,\mathrm{d}u \int_{-\infty}^\infty e^{-3y^2/4}\,\mathrm{d}y\\ &=\int_{-\infty}^\infty e^{-u^2}\,\mathrm{d}u \sqrt{\frac43}\int_{-\infty}^\infty e^{-v^2}\,\mathrm{d}v\\ &=\sqrt\pi\sqrt{\frac43}\sqrt\pi\\ &=\pi\sqrt{\frac43}\\ &=\frac{2\pi}{\sqrt3} \end{align} $$

robjohn
  • 345,667
2

Integrate in polar coordinates

$$ \begin{align} &\int_{-\infty}^\infty\int_{-\infty}^\infty e^{-(x^2+xy+y^2)}\,\mathrm{d}x\,\mathrm{d}y\\ =&\int_{0}^{2\pi} \int_{0}^\infty e^{-r^2(1+\frac12\sin2\theta)}rdr d\theta = \int_{0}^{2\pi} \frac{d\theta}{2+\sin2\theta} =\frac{2\pi}{\sqrt3} \end{align} $$

Quanto
  • 97,352