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

In my grad quantum/E&M classes I had to do intuition-bending regularization of integrals that didn't seem mathematically justified (but got full credit and were repeated in the solutions) like the following:

$$\int_{0}^{\infty}\sin x \;\mathrm{d}x = \lim_{\alpha \to 0^+} \int_{0}^{\infty} e^{-\alpha x} \sin x \;\mathrm{d}x = \lim_{\alpha \to 0^+}\frac{1}{\alpha^2+1} = 1$$

which is unreasonable as $\int_{a}^{b}\sin x \;\mathrm{d}x = \cos(a)-\cos(b)$, or in our case $\lim_{b \to \infty} 1 - \cos(b) = \mathrm{undefined}$ as the limit doesn't converge.

Is there any formal mathematics behind this line of thought claiming divergent things converge with regularization? (E.g., similar to the infinite sums 1+1+1+... = -1/2 or 1+2+3+4+... = -1/12)?

share|cite|improve this question
Usually Abel or Borel is called into play for "evaluating" divergent oscillatory integrals. Something similar can be done for $\int_0^\infty \log\,t\sin\,t\mathrm dt=-\gamma$. – J. M. Nov 28 '11 at 18:34
up vote 14 down vote accepted

Regularization is a way to make a divergent integral into a parametric family of finite integrals. Say, we need to evaluate $\int f(x) \mathrm{d} x$. We introduce (and this is a craft in itself) $f_\alpha(x)$, such that $\lim_{\alpha \to 0} f_\alpha(x) = f(x)$, such that $\int f_\alpha(x) \mathrm{d} x$ exists. We then perform, unjustified, interchanging of integration and taking the limit: $$ \int f(x) \mathrm{d} x = \int \lim_{\alpha \to 0} f_\alpha(x) \mathrm{d} x \stackrel{?}{=} \lim_{\alpha \to 0} \int f_\alpha(x) \mathrm{d} x $$

In your particular case, integral can be defined using Abel summations, since $$ \int_0^\infty \sin(x) \mathrm{d} x = \sum_{n=0}^\infty \int_{\pi n}^{\pi (n+1)} \sin(x) \mathrm{d} x = \sum_{n=0}^\infty 2 (-1)^n \stackrel{\text{Abel}}{=} 1 $$ Different regularization of divergent series do not need to give the same result, however, under certain conditions they do.

Coming back to your integral, its regularization is done, just like you did, by taking advantage of the hypergeometric nature of the integrand, and representing the integral as limit of Mellin convolutions (i.e. $\int_0^\infty h_\alpha(x) \sin(x) \mathrm{d} x$, where $h_\alpha(x)$ is another hypergeometric function, that makes the integral convergent, and such that $\lim_{\alpha \downarrow 0} h_\alpha(x) = 1$): $$ \lim_{\alpha \downarrow 0} \int_0^\infty \frac{\sin(x)}{x^\alpha}\mathrm{d} x = \lim_{\alpha \downarrow 0} \left( \Gamma(1-\alpha) \cos\left( \frac{\pi \alpha}{2} \right) \right) = 1 $$ $$ \lim_{\alpha \downarrow 0} \int_0^\infty \mathrm{e}^{-\alpha x} \sin(x) \mathrm{d} x = \lim_{\alpha \downarrow 0} \frac{1}{\alpha^2 + 1} = 1 $$ $$ \lim_{\alpha \downarrow 0} \int_0^\infty \frac{\sin(x)}{1+\alpha x} \mathrm{d} x = \lim_{\alpha \downarrow 0} \left( \frac{2 \sin \left(\frac{1}{\alpha }\right) \text{Ci}\left(\frac{1}{\alpha }\right)+\cos \left(\frac{1}{\alpha }\right) \left(\pi -2 \text{Si}\left(\frac{1}{\alpha }\right)\right)}{2 \alpha } \right) = 1 $$

share|cite|improve this answer

\begin{eqnarray*} \int_{0}^{\infty}\sin\left(x\right)\,{\rm d}x & = & \Im\int_{-\infty}^{\infty}\Theta\left(x\right){\rm e}^{{\rm i}x}\,{\rm d}x = \Im\int_{-\infty}^{\infty} \left({\rm i}\int_{-\infty}^{\infty}{{\rm d}k \over 2\pi}\,{{\rm e}^{-{\rm i}kx} \over k + {\rm i}0^{+}}\right)\,{\rm e}^{{\rm i}x}\,{\rm d}x \\[3mm] & = & \Re\int_{-\infty}^{\infty}{{\rm d}k \over k + {\rm i}0^{+}} \int_{-\infty}^{\infty}{{\rm d}x \over 2\pi}\,{\rm e}^{-{\rm i}\left(k - 1\right)x} = \Re\int_{-\infty}^{\infty}{{\rm d}k \over k + {\rm i}0^{+}}\,\delta\left(k - 1\right) \\[3mm] & = & \Re\left(1 \over 1 + {\rm i}0^{+}\right) = \Re\left\lbrack 1 - {\rm i}\,\pi\,\delta\left(1\right)\right\rbrack = 1 \end{eqnarray*}

Why does it work ?: See $\tt @Sasha$ answer.

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.