0
$\begingroup$

Given $C \geq 1$ and $\epsilon > 0$, is there a number $N = N(C,\, \epsilon)$ such that the following holds:

For every set $S \subseteq S^1$ of cardinality $C$, there is a function $f: S^1 \to \mathbb{R}$ such that:

1). $f(z) = \sum_{n = -N}^N a_n z^n$ (i.e. $f$ is a Laurent polynomial of in degrees $[-N,\, N]$).

2). $f \geq 0$ on $S^1$

3). $f(z) \geq 1$ for each $z \in S$

4). $\|f\|_{L^1} < \epsilon$

$\endgroup$
1
  • $\begingroup$ Why not do it one point at a time? $\endgroup$ Nov 5, 2014 at 1:05

1 Answer 1

2
$\begingroup$

So it suffices to do this for $S=\{1\}$. If you can find a degree $N$ Laurent polynomial doing the job for $\{1\}$ with $L^1$ distance $\epsilon$, $f_1$ say, then $f_\zeta(z):=f_1(\bar \zeta z)$ is another Laurent polynomial of the same degree doing the job for $\{\zeta\}$ with the same $\epsilon$. Suppose you can find $N(1,\epsilon)$ for each $\epsilon>0$, then there is a trivial bound $N(C,\epsilon)\le N(1,\epsilon/C)$: just run the argument above to get an $f_\zeta$ for each $\zeta$ in $C$ and add them together. The sum has the right lower bounds, and the $L^1$ bound is trivial.

Finally how to get an $f_1$ with $L^1$ norm less than $\epsilon$. No doubt there are much better ways than I'm about to suggest, but anyway: take a continuous function $g$ of $L^1$ norm less than $2\epsilon/3$ that is bounded below by $\epsilon/3$ and satisfies $g(1)>1+\epsilon/3$. Using the Féjér kernel, obtain a trigonometric polynomial of some degree approximating $g$ uniformly within $\epsilon/3$. This is your $f_1$. Et voilà!

$\endgroup$

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.