3

Problem

Let $\text{Ai}: \mathbb{R} \to \mathbb{R}$ with $$\text{Ai}(x) = \frac{1}{\pi} \Re \int_{0}^{\infty} \omega e^{ -\frac{t^3}{3} + i x \omega t} dt$$ be the Airy function with $\omega := e^{\frac{i\pi}{6}}$. Prove that $\text{Ai}''(x) = x\text{Ai}(x)$.

My approach

We have $\omega = e^{\frac{i\pi}{6}} = \cos\frac{\pi}{6} + i\sin\frac{\pi}{6} = \frac{\sqrt{3}}{2} + \frac{1}{2} i$ and $-\frac{t^3}{3} + i x \omega t = -\frac{t^3}{3} - \frac{xt}{2} + \frac{xt\sqrt{3}}{2}i$. We also have $\omega^3 = e^{\frac{i\pi}{2}} = i$. Let $f: \mathbb{R}_0^+\times\mathbb{R}\to\mathbb{C}$ with:

\begin{align} &f(t,x) = \omega e^{ -\frac{t^3}{3} + i x \omega t}\\ &\frac{\partial}{\partial x}f(t,x) = i \omega^2 t e^{ -\frac{t^3}{3} + i x \omega t} &&\left| \frac{\partial}{\partial x}f(t,x) \right| = t e^{-\frac{t^3}{3} - \frac{xt}{2}}\\ &\frac{\partial^2}{\partial^2 x} f(t,x) = - i t^2 e^{ -\frac{t^3}{3} + i x \omega t} &&\left| \frac{\partial^2}{\partial^2 x}f(t,x) \right| = t^2 e^{-\frac{t^3}{3} - \frac{xt}{2}} \end{align}

I need to show the following equation in order to progress: $$ \frac{\partial^2}{\partial^2 x} \int_0^\infty f(t,x) dt = \frac{\partial}{\partial x} \int_0^\infty \frac{\partial}{\partial x} f(t,x) dt = \int_0^\infty \frac{\partial^2}{\partial^2 x} f(t,x) dt $$

For this I need to show dominated convergence for $\frac{\partial}{\partial x}f(t,x)$ and $\frac{\partial^2}{\partial^2 x}f(t,x)$, a.i. I need to find a function $g_1:\mathbb{R}_0^+ \to \mathbb{R}$ with $\left| \frac{\partial}{\partial x}f(t,x) \right| \leq g_1(t)$ for all $t \in \mathbb{R}_0^+$ and $x \in \mathbb{R}$. I also need to find a function $g_2:\mathbb{R}_0^+ \to \mathbb{R}$ with $\left| \frac{\partial^2}{\partial^2 x}f(t,x) \right| \leq g_2(t)$ for all $t \in \mathbb{R}_0^+$ and $x \in \mathbb{R}$. However, $x\in\mathbb{R}$ can be arbitrarely small making $\frac{\partial}{\partial x}f(t,x)$ and $\frac{\partial^2}{\partial^2 x}f(t,x)$ arbitrarely large at some point. I am unable to find functions $g_1(t)$, $g_2(t)$ to show dominated convergence. What am I missing?

Note

This problem was already discussed here, but the one particular point which I am interested in was handwaved.

  • 1
    I noticed that you referred to one of my postings and that you said I handwaved one argument. In a sense I did since I had posted in the past a but a common general situations about analytic functions coming from integral transformations here – Mittens Apr 08 '22 at 01:39

1 Answers1

2

The goal here is to change the order of integration and differentiation.

Let $a(x) = \int_0^\infty \omega e^{ -\frac{t^3}{3} } g(x,t) dt$, where $g(x,t) = e^{i\omega xt}$.

You want to show that $a'(x) = \int_0^\infty \omega e^{ -\frac{t^3}{3} } { \partial g(x,t) \over \partial x } dt$, so look at the estimate $| {a(x+h) - a(h) \over h} - \int_0^\infty \omega e^{ -\frac{t^3}{3} } { \partial g(x,t) \over \partial x } dt | \le \int_0^\infty | \omega| e^{ -\frac{t^3}{3} }| {g(x+h,t)-g(x,t) \over h} - i\omega t g(x,t)|dt $.

If we can show that $\int_0^\infty | \omega| e^{ -\frac{t^3}{3} }| {g(x+h,t)-g(x,t) \over h} - i\omega t g(x,t)|dt \to 0$ then we see that $a$ is diferentiable and the derivative given as above.

Use the mean value theorem to show $| {g(x+h,t)-g(x,t) \over h} - i\omega t g(x,t)| = | { e^{i\omega ht } -1 \over h } - i \omega t| \le |\omega|(1+t)$ and $| \omega| e^{ -\frac{t^3}{3} } |\omega|(1+t)$ is an integrable upper bound. Hence the DCT applies.

Rinse & repeat for the second derivative.

copper.hat
  • 178,207
  • Isn't $\left| \omega \right| = 1$? – Wilfred Montoya Apr 03 '22 at 22:41
  • 2
    It is a constant , whether it is $1$ or $42$ is immaterial. My understanding was that showing a dominating function was the point here, as Oliver has addressed the equality in the linked post? – copper.hat Apr 03 '22 at 22:43
  • What is $\left| \omega \right| e^{\frac{t^3}{3}} \left| \omega \right| (1+t)$ an upper bound of? Certainly it's not an upper bound of $\frac{\partial}{\partial x}f(t,x)$ and $\frac{\partial^2}{\partial^2 x}f(t,x)$, which I wanted to proof. I am confused. – Wilfred Montoya Apr 03 '22 at 22:50
  • 1
    It is a bound on ${f (x+h,t) -f(x,t) \over h} - {\partial f(x,t) \over \partial x}$. It shows that the derivative of the integral is the integral of the derivative, which, as far as I can tell from above, is where you are having difficulty. – copper.hat Apr 03 '22 at 22:54
  • I thought my main objective was to find a bound on $\left| \frac{\partial f(t,x)}{\partial x} \right|$ of the form $g(t)$. You are presenting a different approach which leaves me with a lot of questions, e.g. aren't $\lim_{h\to 0} \frac{f(t,x+h) - f(t,x)}{h}$ and $\frac{\partial f(t,x)}{\partial x}$ the same by definition? How does putting a bound on $\frac{f(t,x+h) - f(t,x)}{h} - \frac{\partial f(t,x)}{\partial x}$ help me to show domininant convergence? What is wrong with my original approach? – Wilfred Montoya Apr 03 '22 at 23:08
  • 1
    Finding that bound is not sufficient to show that you can interchange integration & differentiation. Think about how you would do that. – copper.hat Apr 03 '22 at 23:39