The expression $\int_{-\infty}^\infty e^{i k x} dx$ is sometimes identified as the Dirac delta function. This identification is said "formal" or "symbolic", and some physics texts say that the theory of distribution makes this identification rigorous. So I read some texts on distributions and I think now I understand what delta function is as a distribution. However, I still cannot give any meaninig on the expression as a distribution. Is it possible to give a well defined meaning as a mathematical object to the expression?

  • $\begingroup$ I'm not sure but I believe: its the inverse Fourier transform of 1, but of course writing it formally as this integral doesn't make sense, so you define as a limit of an approximating sequence in $\mathcal D'$. Check en.wikipedia.org/wiki/Oscillatory_integral $\endgroup$ – Calvin Khor Aug 8 '18 at 15:46
  • 1
    $\begingroup$ @CalvinKhor So you introduce two ways. i) The actually "intended" by the expression may be the distributional Fourier transform of 1. ii) Intepreting the improper integral as a special kind of limit introduced in the Wikipedia page makes the expression rigorous. Am I right? $\endgroup$ – lyrin Aug 8 '18 at 16:35
  • $\begingroup$ That's my impression but I'm not certain as I've only read about this in passing. $\endgroup$ – Calvin Khor Aug 8 '18 at 16:36

The way to give meaning to an expression as a distribution is to see its effect on a test function. If $\phi(x)$ is a test function (a member of the Schwartz space $\mathscr S(\mathbb R)$), then the action of the formal expression $T = \int_{-\infty}^\infty e^{ikx}\; dx$ on $\phi$ is obtained by integration (and switching the order of the integrals): $$ T(\phi) = \int_{-\infty}^\infty dx \int_{-\infty}^\infty dk \; \phi(k) e^{ikx} = \int_{-\infty}^\infty dx \; \hat{\phi}(-x/(2\pi)) = 2 \pi \phi(0)$$

  • $\begingroup$ Isn't it still "formal"? There is a distribtion $T$ defined as $\left< T, \phi \right> = \int_{-\infty}^\infty dx \int_{-\infty}^\infty dk \phi(k) e^{ikx}$. Do you mean I have to identify any occurrence of $\int_{-\infty}^\infty e^{ikx} dx$ in an integral with the action of $T$ even if the integral itself is not well defined? For example, $\int_{-\infty}^\infty dk \int_{-\infty}^\infty dx \phi(k) e^{ikx}$ is sometimes found in physics texts but it is meaningless if interpreted as it is. $\endgroup$ – lyrin Aug 8 '18 at 16:07

$$ \langle \int_{-R}^{R}e^{ikx}dk,\varphi\rangle=\int_{-\infty}^{\infty}\varphi(x)\int_{-R}^{R}e^{ikx}dkdx \\ = \int_{-R}^{R}\int_{-\infty}^{\infty}\varphi(x)e^{ikx}dx dk $$ As $R\rightarrow\infty$ the above tends to $2\pi\varphi(0)$. So the truncated integral converges in the distribution sense to a scalar times the Delta function at $0$: $$ \int_{-R}^{R}e^{ikx}dk \rightarrow 2\pi\delta_{0}. $$ The improper integral converges in the dual. It's an element of $\mathscr{S}^*$.


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.