Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

How can I solve this problem? $$ \int_{-\infty}^{+\infty}\frac1{1+x^2}\left(\frac{\mathrm d^n}{\mathrm dx^n}e^{-x^2}\right)\mathrm dx $$

share|cite|improve this question
From the Rodrigues formula for the Hermite polynomials, we have $$\dfrac{\mathrm d^n}{\mathrm dx^n}\exp(-x^2)=(-1)^n \exp(-x^2) H_n(x)$$ If $n$ is odd, $H_n(x)$ is odd, so the integral is zero for odd $n$. – J. M. Dec 22 '11 at 13:58
With some Mathematica assistance, I obtained $$\int_{-\infty}^\infty \frac{\exp(-x^2)}{1+x^2}H_{2n}(x)\mathrm dx=e\sqrt{\pi}(-1)^n \frac{(2n)!}{n!}\sum_{k=0}^n (-1)^k \binom{n}{k} E_{k+\frac12}(1)$$ where $E_p(z)$ is the generalized exponential integral. The generalized exponential integral satisfies a recursion relation, so further simplification is very much possible. – J. M. Dec 22 '11 at 14:31
My goal of solving this integral is to find $a_{n}$ for fourier-hermit series expansion of $f(x)=\frac{1}{1+x^2}=\displaystyle\sum_{n=0}^{\infty}a_{n}H_{n}(x)$, where $a_{n}$ is $$\frac{1}{2^nn!\sqrt{\pi}}\int_{-\infty}^{+\infty}e^{-x^{2}}f(x)H_{n}(x)dx$$. – gex Dec 22 '11 at 14:36
You should have said that earlier, you know. I hope that your $a_n$ are zero for odd $n$ is obvious to you. – J. M. Dec 22 '11 at 14:37
You've seen my second comment, no? – J. M. Dec 22 '11 at 15:09

Write $f(x) \leftrightarrow F(\omega)$ to mean that $F(\omega)$ is the Fourier transform of $f(x)$, where $$F(\omega)) = \int_{-\infty}^\infty f(x) \exp(-i\omega x)\mathrm dx.$$ Then we have $$\begin{align*} \frac{1}{1+x^2} &\leftrightarrow \pi \exp(-\vert \omega \vert )\\ \exp(-x^2) &\leftrightarrow \sqrt{\pi}\exp(- \omega^2/4),\\ \frac{\mathrm d^n}{\mathrm dx^n}\exp(-x^2) &\leftrightarrow (i\omega)^n\sqrt{\pi}\exp(-\omega^2/4) \end{align*} $$ The inner-product version of Parseval's theorem is $\displaystyle \int_{-\infty}^\infty f^*(x) g(x)\mathrm dx = \frac{1}{2\pi}\int_{-\infty}^\infty F^*(\omega) G(\omega)\mathrm d\omega$ and so we have $$\begin{align*} \int_{-\infty}^\infty \frac{1}{1+x^2} \frac{\mathrm d^n}{\mathrm dx^n}\exp(-x^2) \mathrm dx &= \frac{1}{2\pi}\int_{-\infty}^\infty \pi\exp(-\vert \omega \vert) (i\omega)^n \sqrt{\pi} \exp(-\omega^2/4) \mathrm d\omega \end{align*}$$ where the integrand on the right is an odd function of $\omega$ when $n$ is odd and so the integral is $0$. Otherwise, convert to an integral on $\mathbb R^+$ to get $$\begin{align*} a_{2n} &= \frac{\sqrt{\pi}}{2}\int_{-\infty}^\infty \exp(-\vert \omega \vert) (i\omega)^{2n} \exp(-\omega^2/4) \mathrm d\omega\\ &= (-1)^n\sqrt{\pi}\int_0^\infty \omega^{2n}\exp(-\omega) \exp(-\omega^2/4) \mathrm d\omega\\ &= (-1)^n\sqrt{\pi}b_{2n} \end{align*}$$ where $\displaystyle b_n = \int_0^\infty \omega^{n}\exp(-\omega) \exp(-\omega^2/4) \mathrm d\omega$ can be integrated by parts to get $$\begin{align*} b_{n} &= \int_0^\infty \omega^{n-1}\exp(-\omega) \cdot \omega \exp(-\omega^2/4) \mathrm d\omega\\ &= \left[ -2\omega^{n-1}\exp(-\omega) \exp(-\omega^2/4)\biggr|_0^\infty \right . \\ &\quad \quad\quad \quad\quad \left . + \int_0^\infty 2[(n-1)\omega^{n-2}\exp(-\omega) - \omega^{n-1}\exp(-\omega)]\exp(-\omega^2/4) \mathrm d\omega\right ]\\ &= -2b_{n-1} + 2(n-1)b_{n-2} \end{align*}$$ or, with a slight change in indices $$b_{n+1} = -2b_n +2nb_{n-1}.$$ Note that $b_0 = e\sqrt{\pi}\text{erfc}(1)$ as noted here while $b_1 = 2(1-b_0)$. Thus, the recurrence relation can be used to find $b_{2n}$ and hence $a_{2n}$.

share|cite|improve this answer

It appears that your integral for even $n = 2 k$ has the exponential generating function (egf) $$F(t) = \frac{\pi (1 - \text{erf}(1/\sqrt{4t+1}))}{\sqrt{4t+1}} e^{1/(4t+1)}$$ That is $$\int_{-\infty}^{\infty} \frac{1}{1+x^2} \frac{d^{2k}}{dx^{2k}} e^{-x^2} \ dx = F^{(k)}(0) $$

Each of these values will be of the form $a_k \sqrt{\pi} + b_k \pi e (\text{erf}(1) - 1)$ where $a_k$ and $b_k$ are integers. The egf for $a_k$ is $$A(t) = \frac{\sqrt{\pi} (\text{erf}(1) - \text{erf}(1/\sqrt{4t+1}))}{\sqrt{4t+1}} e^{1/(4t+1)}$$ and the egf for $b_k$ is $$B(t) = \frac{-1}{\sqrt{4t+1}} e^{-1 + 1/(4t+1)}$$

share|cite|improve this answer

$$ \int_{-\infty}^{+\infty}\frac1{1+x^2}\left(\frac{\mathrm d^n}{\mathrm dx^n}e^{-x^2}\right)\mathrm dx = \,U\left(1+\frac{n}{2},\frac{3}{2},1\right)\cos\left(\frac{\pi\,n}{2}\right)\,n!\,\sqrt{\pi}, $$ where $U(a,b,z)$ is the confluent hypergeometric function of the second kind.

The proof is quite long. I started with applying the Fourier transform to $e^{-x^2}$ so that the derivative $\frac{\mathrm d^n}{\mathrm dx^n}$turns into a multiplication by $(-i\, t)^n$. The inverse transform gives a general form for n-th derivative in terms of hypergeometric functions.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.