There is a nice necessary and sufficient criterion for a polynomial $p_n$ of degree $n$ with real coefficients to have all zeros real and simple. Namely, it says: $p_n$ has $n$ (distinct) real roots $x_{1},\dots,x_{n}$ iff the Hankel matrix $$S_{n}=\begin{pmatrix} s_{0} & s_{1} & \dots & s_{n-1}\\ s_{1} & s_{2} & \dots & s_{n}\\ \vdots & \vdots & \ddots & \vdots\\ s_{n-1} & s_{n} & \dots & s_{2n-2} \end{pmatrix}, \quad s_{k}:=\sum_{j=1}^{n}x_{j}^{k},$$ is positive definite; see, for example here.

My first question asks if there is some "infinite analogue" of the above equivalence where a polynomial is replaced by an entire function.

More concretely, let me assume the following. Let $f$ is a given entire function with real Taylor coefficients. Next, let $\{p_{n}\}_{n\geq1}$ be a sequence of polynomials (with increasing degree) approximating $f$. For example, we may assume $p_{n}\to f$, for $n\to\infty$, uniformly on compact subsets of $\mathbb{C}$. (One can take $p_n$ to be the truncated Taylor expansion of $f$, for instance.) Assume further that $x_{1}(n),\dots,x_{n}(n)$ are zeros of $p_n$ and that there exist limits $$\hskip100pt \lim_{n\to\infty}\frac{1}{n}\sum_{j=1}^{n}\left(x_{j}(n)\right)^{k}=\alpha_{k},\hskip100pt (*)$$ for all $k\in\mathbb{N}$.

The second question is: if the Hankel matrix $A$, where $A_{i,j}:=\alpha_{i+j}$, $i,j\geq0$, is associated with a positive moment functional, i.e., all the principal sections of $A$ are positive definite, can one say something about zeros of $f$? Are they real and simple, for example?

The third question: Let me add one last related question. Assume that we have the sequence of polynomials $p_{n}$ as above, so the limits in $(*)$ hold, but we have no limit function $f$. The sequence $\{p_{n}\}$ is not necessarily convergent. Is it true that, if $A$ has all the principal sections positive definite, then the sequence $\{p_{n}\}$ is asymptotically real-rooted, that is, all the limit points of zeros are real?

A slight reformulation of the question as well as an addition of some other assumptions (for example, on the order of $f$, the rate of the convergence in the limits $(*)$, etc.) is possible.


It seems that the magic words are Laguerre-Polya Class. (the link is to an article by Tom Craven and George Csordas, which has many references.)

  • $\begingroup$ Thank you for the relevant keyword and the paper! I checked just quickly results of the paper and they are interesting (Section 2, in particular). Nevertheless, most part of my post still remain unanswered. $\endgroup$
    – Twi
    Sep 13 '16 at 13:28

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.