This following integral is not convergent

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

but I know for example that

$$ \int_0^\infty dx \, x^c e^{-b x} = b^{-1-c}\, \Gamma(1+c) $$ where $\Gamma$ is the Euler Gamma function.

Therefore calling $c=ia$ and $i\omega = -b$ one gets

$$ \int_0^\infty dx \, x^{ia} e^{i\omega x} = (-i \omega)^{1-ia} \Gamma(1+ia) = e^{-i \pi/2} \omega^{1-ia} \, \Gamma(1+ia) $$

which looks finite to me, so can you tell me where I've made a mistake or an abuse?

  • $\begingroup$ When you write not convergent, do you mean not absolutely integrable or that the improper integral is not convergent? (i.e. the there is no limit to the integrals restricted on $[0,n]$ when $n\to \infty$) $\endgroup$ – LL 3.14 Jun 15 '20 at 19:14

The very short answer is "analytic continuation": we can begin with definitions that are only well-defined on some subset $A \subseteq \mathbb{C}$, find an analytical relation that holds within $A$, and then if one side of this relation is valid on a larger set $B \supseteq A$, use that to define an extension into $B$. I'll provide a Gamma function example, since it complements your question.

We can define, for $\Re(z) > 0$, $\Gamma(z)$ as $$\Gamma(z) = \int_0^\infty e^{-t} t^{z-1} \; dt$$ Using this relation, we can show that $$\Gamma(z+1) = z \Gamma(z) \implies \Gamma(z) = \frac{\Gamma(z+1)}{z}$$ Now note that while our original definition of $\Gamma$ is only valid for $\Re(z) > 0$, the right-hand side of our new relation defines an analytical (strictly, meromorphic) function on $\Re(z) > -1, z \neq 0$.

This lets us analytically continue our original equation for $\Gamma$ into a larger subset of $\mathbb{C}$. Moreover, we can repeat this trick to extend it to all $z \in \mathbb{C} \setminus \lbrace 0, -1, -2, \dots \rbrace$. We started with a definition of an analytic function on $\Re(z) > 0$, and ended up defining an analytic function on almost all of $\mathbb{C}$.

This is more or less your question: by analytically continuing our expression, we can "sensibly" assign a value to our original non-convergent expression.

A similar example is how $\zeta(-1) = -\frac{1}{12}$. The starting point for our definition of $\zeta(z)$ does not converge for $z = -1$, but we can analytically continue it to a larger function which is finite at $-1$, where it takes the value $-\frac{1}{12}$. But of course, this does not mean that $$\sum_{n=1}^\infty n = -\frac{1}{12}$$

For a final example, consider how the series $$f(x) = 1 + x + x^2 + \dots$$ converges for $\lvert x \rvert < 1$. On this domain, we can show that $$f(x) = \frac{1}{1-x}$$ This is valid on a larger domain, and we can evaluate, say, $$f(2) = -1$$ but we can't back-substitute this into our original expression $$1 + 2 + 2^2 + \dots \neq -1$$ Instead, the (roughly) correct statement is that any analytical function that agrees with our original series on $\lvert x \rvert < 1$ and is valid on a larger domain (for avoidance of certain cases, let's say it's "suitably nice", say path connected) containing $2$, will have $f(2) = -1$.


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.