2
$\begingroup$

Let $n \in \mathbb N$, let $p:\mathbb R \to \mathbb R$ be a real polynomial, and let $\lambda_1< \lambda_2 <\dots < \lambda_n$. Now let $$ f \in \mathrm{span} \left \{ p(x - \lambda_k)e^{\lambda_k x} : k=1,\dots, n \right \}, \quad f \neq 0. $$ It is well-known that if $p =1$ then $f$ has at most $n-1$ distinct (real) zeros in $[0,1]$. What about the case where $p \neq 0$? Can I find assumptions on the $\lambda_j$'s so that $f$ has again at most $n-1$ distinct zeros in $[0,1]$? I'm in particular interested in the case where the $\lambda_k$'s are of the form $\lambda_k = C + \alpha k$ for some fixed $\alpha,C>0$.

Thank you for any help!

$\endgroup$
1
  • $\begingroup$ You certainly need some restrictions on $p$. If you want it to be true for an arbitrary polynomial $p$ with some restrictions on $\lambda$'s alone (which seems to me to be the formal meaning of what is written now), then you are doomed (just recall that $p$ can approximate absolutely any continuous function on any compact set with any precision). $\endgroup$
    – fedja
    Jan 28, 2023 at 21:12

1 Answer 1

2
$\begingroup$

To estimate the number of real zeros from above, you need some assumption on the degree of $p$ (otherwise no such estimate is possible, even for $n=1$, see @fedja's comment). Then there is a general theorem: If $P_j$ are real polynomials of degrees $d_j$ then $$P_1(x)e^{\lambda_1 x}+\ldots +P_n(x)e^{\lambda_nx}$$ has at most $d_1+\ldots+d_n+n-1$ real zeros.

See, for example, Polya Szego, Part V, Chap I, Sect. 5, probl. 75.

$\endgroup$
1
  • $\begingroup$ Thank you very much for the answer! Do you think one can gain anything in the setting where $P_j=P(\cdot - \lambda_j)$? Or does this seem unrealistic ... $\endgroup$
    – r_l
    Jan 29, 2023 at 9:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.