5
$\begingroup$

Consider a continuous function $f: \mathbb{R} \rightarrow \mathbb{R}$, supported on $[-1,1]$, of positive type. Assume $f(0) = 1$; what is the "largest area" $\int f\,dx$ that can be achieved?

To be more precise, let $f: \mathbb{R} \rightarrow \mathbb{R}$ be a continuous function satisfying:

  • $f$ is supported on $[-1,1]$,
  • for all $x$, $0 \leq f(x) \leq 1 = f(0)$, and
  • $f$ has positive type: for any finite family of points $x_1 < \cdots < x_n$ in $\mathbb{R}$, the matrix $(f(x_i - x_j))_{ij}$ is positive semi-definite. (Equivalently--by Bochner's theorem--the Fourier transform is pointwise nonnegative.)

How large can $\int f(x)\,dx$ be?


Remarks

  1. One example of such a function is the "triangle" $t(x) = \max(1 - |x|,0)$. This achieves $\int t(x)\,dx = 1$. Is that the best one can do?
  2. There are functions $g$ of positive type (and satisfying the other requirements above) for which $g(x) > t(x)$ for some $x$. (However, I do not know of any for which $\int g(x)\,dx > 1$.)
  3. I asked this question the "Mathematics" stack exchange site, but didn't have any takers.
$\endgroup$
2
  • $\begingroup$ The Poisson summation formula would give $\sum_{n\in {\Bbb Z}} f(n) = \sum_{k\in {\Bbb Z}} {\hat f}(k)$. The LHS is $1$, and since ${\hat f}(k)\ge 0$ by assumption, it follows that ${\hat f}(0) \le 1$. $\endgroup$
    – Lucia
    Nov 6, 2014 at 14:19
  • $\begingroup$ Awesome! That's just right. If you turn it in to an "answer," I can accept it. $\endgroup$
    – acr
    Nov 6, 2014 at 14:38

1 Answer 1

8
$\begingroup$

As noted in my comment above, the Poisson summation formula would give $$ 1= f(0) = \sum_{n\in {\Bbb Z}} f(n) = \sum_{k\in {\Bbb Z}} {\hat f}(k) \ge {\hat f}(0), $$ since ${\hat f}(k) \ge 0$ by assumption. Since ${\hat f}(0) = \int_{-1}^{1} f(x) dx$, this proves the desired bound.

If you are worried about using the Poisson summation formula on just a continuous function, use the above argument with $f$ convolved with $\phi_{\epsilon}$ where $\phi_{\epsilon}$ is smooth (or just in $C^2$, say) and non-negative, supported on $(-\epsilon,\epsilon)$, has non-negative Fourier transform, and ${\hat \phi_{\epsilon}}(0)=1$. Now use the above argument with $f*\phi_{\epsilon}$, to obtain ${\hat f}(0)\le 1+O(\epsilon)$, and then let $\epsilon \to 0$.

$\endgroup$

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.