Uniform convergence of this particular limit functional

122 Views Asked by At

Let $\{h_n(x)\}_{n\in\mathbb{N}}$ a family of infinitely differentiable functions, defined on the entire real line, vanishing outside $[a,b]$. Let $F(x)$ be any function defined on $\mathbb{R}$ with the property that $F$ and all its derivatives are $O(|x|^{-N})$ as $|x|\to\infty$, for every $N$.

By hypothesis, the following limit:

$$\lim_{n\to\infty}\int_{x=a}^b h_n(x)F(x)\mathrm{d}x$$

exists for any $F$ of the type specified above.

Is it true or false that $\lim_{n\to\infty}\int_{x=a}^b h_n(x)F(x+\tau)\mathrm{d}x$ converge uniformly w.r.t. $\tau\in[\tau_1,\tau_2]\subset\mathbb{R}$?

I tried with Cauchy criterion, but without success... If the answer is "no", with which hypothesis the answer could become "yes"?

1

There are 1 best solutions below

1
On

As said in the comments of your post, the hypothesis on $F$ toward infinity is useless. We can choose $F$ to be any smooth function defined over $\mathbb{R}$. But since only the values of $F$ on a neighborhood of $[a,b]$ are important, we can assume that $F$ has a compact support if needed.

I can answer positively to your question if we add the hypothesis that the sequence $\|h_n\|_{L^1} = \int_a^b |h_n(x)| dx$ is bounded. In this case, by uniform continuity of $F$ (it can be assumed compactly supported thus Heine theorem holds), $\tau \mapsto (x \mapsto F(x + \tau))$ is continuous for the $L^\infty$ norm (just play with $\varepsilon$-$\delta$ and the definition of uniform continuity). If we set for all $n$ and $\tau$, $$ I_n(\tau) = \int_a^b h_n(x)F(x + \tau) dx $$ and $I(\tau) = \lim_{n \rightarrow +\infty} I_n(\tau)$ (which exists by hypothesis), we have for all $n,\tau,t$, $$ |I_n(t) - I_n(\tau)| \leqslant \left|\int_a^b h_n(x)(F(x + \tau) - F(x + t))\right| \leqslant \|h_n\|_{L^1}\|x \mapsto F(x + \tau) - F(x + t)\|_{L^\infty}. $$ By continuity of $\tau \mapsto (x \mapsto F(x + \tau))$ and boundedness of $(\|h_n\|_{L^1})$, we deduce that the $I_n$ are uniformly continuous, thus its limit $I$ is continuous, which is the wanted result.

In the general case, the question is harder. For example, if you take $h_n(x) = n^k\cos(2\pi(b - a)nx)$ on some $[a + \varepsilon,b - \varepsilon]$ that you extend into a smooth function supported in $[a,b]$, you can show that your hypothesis holds for any smooth $F$ and any $k \geqslant 0$ (look at the convergence speed of Fourrier coefficients) but if $k > 0$, the hypothesis I added is not verified.

Try to see if this particular case is a counter-example, it may help you to find a counter-example or a proof in the general case.