Let $H_k(x)$ be (probabilists' or physicists', does not matter for this question) Hermite polynomials.

It is well-known that all the gaps between consecutive roots of $H_k(x)$ are at least a multiple of $1 / \sqrt{k}$ (see, e.g., Szego, "Orthogonal Polynomials", page 130).

Is the same known to be true if we consider the roots of $H_k(x)$ and $H_{k+1}(x)$ together? Equivalently, is it true that the zeros of $H_{k+1}(x)$ are roughly half-way between its respective local minima and maxima?

Looking at the numerical values for $k \leq 100$, the desired bound seems to be true.


The answers are yes and yes.

Consider the Hermite Gauss functions : $\psi_n(x)=e^{-x^2/2}H_n(x) $, we have two properties: $$\psi_n''(x)+(2n+1-x^2)\psi_n(x)=0 $$ and $$\psi_{n+1}=\psi_n'(x) +x\psi_n(x) $$ For $n$ large, we take a lenght $L>0$ which is very small (for example $L=1/\log{n}$). On any $[x_0-L/2,x_0+L/2]\subset ]- \sqrt{2n},\sqrt{2n}[$, $\psi_n$ is rountly the solution of $$\psi_n''(x)+(2n+1-x_0^2)\psi_n(x)=0 $$ and therefore $\psi_n(x)=A\sin(\omega_0 x+\alpha)$ on $[x_0-L/2,x_0+L/2]$ with $\omega_0=\sqrt{2n+1-x_0^2}$ and $\psi_{n+1}(x)=-\omega_0 A \cos(\omega_0 x+\alpha)+x_0A\sin(\omega_0 x+\alpha)$. In particular the zeros of $\psi_n$ are placed periodically every $\pi/\omega_0$ and the zeros of $\psi_{n+1}$ are just shifted by $\omega_0 (\pi/2 + \arctan(x_0/\omega_0))$

For $x_0=\pm \sqrt{2n}$, the $\sin$ has to be replaced by the Airy function with a scale $1/\sqrt{n}$. And the distribution of its zeros and the zeros of its derivative are the same (with the scale $1/\sqrt{n}$) https://en.wikipedia.org/wiki/Airy_function

| cite | improve this answer | |

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.