3

I am trying to figure out, whether the following statement holds: Let $(f_k)_{k\in \mathbb{N}} =: \mathcal{F} \subseteq C(K)$ a sequence of continous functions, where $K\subseteq \mathbb{R}^n$ compact. Assume $\mathcal{F}$ is equicontinuous and the sequence $f_k$ converges to $f$ with respect to the $L_1$ norm. Does this imply that $f_k \rightarrow f$ uniformly on $K$?

I feel like I might be able to apply some version of Arzelà–Ascoli, but I am not entriely sure about the details.

mathxxx
  • 137
  • 1
    This is clearly false without continuity of $f$. – Kavi Rama Murthy Feb 04 '25 at 10:03
  • Thanks for your comment. Do you have any explanation why? I thought maybe one could argue by using that $L_1$ convergence gives the existence of a pointwise convergent subsequent to $f$ (at least almost everywhere), which should again imply uniform convergence of the respective subsequence. I am not sure if this helps in any way or if we do need continuity of $f$. – mathxxx Feb 04 '25 at 10:30
  • Are you asking whether the sequence converges uniformly or whether a subsequence converges uniformly? – Dean Miller Feb 04 '25 at 10:49
  • I am asking whether the whole sequence converges uniformly. In case this holds only for a subsequence it would still be interesting for me to know. I think if it holds for a subsequence, this probably implies continuity of f (?), so it could help maybe. – mathxxx Feb 04 '25 at 10:56
  • $f_n\equiv 0, f(x)=0$ for $x \neq 0$, $f(0)=1$. – Kavi Rama Murthy Feb 04 '25 at 11:15
  • I see your point. Now I'm confused. Let's additionally assume that $\mathcal{F}$ is uniformly bounded, which is obviously the case in your example. Then, by Arzela-Ascoli, as far as I know, it should hold, that $f_k$ has some uniformly converging subsequence to some limit $\bar{f}$. Since uniform convergence implies $L_1$ convergence, it should hold $\bar{f} = f$ on $K$. Furthermore, since $\bar{f}$ is the limit of a uniformly converging subsequence of continuous functions, $f$ should be continuous. What am I missing at this point? Since the example you constructed is not continuous. – mathxxx Feb 04 '25 at 11:31
  • $\overline f=f$ a.e. on $K$, so $f$ is a.e. equal to some function $g$ such that the subsequence converges to $g$ uniformly. $g$ is continuous, but $f$ need not be. – Kavi Rama Murthy Feb 04 '25 at 11:46
  • I see. I am wondering which type of properties could prevent $f$ from not being continuous. Actually I know that $f_k$ are probability density functions, so they have integral 1 and this also holds for $f$ (simple proof). I feel like this might exclude such examples as the one you created, right? – mathxxx Feb 04 '25 at 11:50

1 Answers1

1

We have the following result. The main idea of the proof comes from this answer.

Let $E\subseteq \mathbb{R}^{d}$ be a non-empty Borel set. Assume there is an increasing sequence $(K_{n})_{n\in\mathbb{N}}$ of non-empty compact subsets of $\mathbb{R}^{d}$ such that $E = \bigcup_{n\in\mathbb{N}} K_{n}$ and that every compact subset of $E$ is contained in a member of the sequence $(K_{n})_{n\in\mathbb{N}}$. (This condition is automatically satisfied if $E$ is a non-empty open or closed subset of $\mathbb{R}^{d}$.) Let $(f_{n})_{n\in\mathbb{N}}$ be an equicontinuous sequence in $C(E)\cap L_{p}(E)$ and $f\in L_{p}(E)$. Suppose the sequence $(f_{n})_{n\in\mathbb{N}}$ converges to $f$ in $L_{p}(E)$. Then the following hold.

  1. There exists a $g\in C(E)$ such that $(f_{n})_{n\in\mathbb{N}}$ converges to $g$ uniformly on compact subsets of $E$ and $f = g$ almost everywhere.

  2. If $f\in C(E)$ and $|U\cap E| > 0$ for all open $U\subseteq \mathbb{R}^{d}$ with $U\cap E \neq \emptyset$ (where $|A|$ is the Lebesgue measure of a measurable subset $A$), then $(f_{n})_{n\in\mathbb{N}}$ converges to $f$ uniformly on compact subsets of $E$.

We first show that every subsequence of $(f_{n})_{n\in\mathbb{N}}$ has a subsequence converging uniformly to some $g\in C(E)$ that is equal to $f$ almost everywhere. Let $(f_{n_{j}})_{j\in\mathbb{N}}$ be a subsequence of $(f_{n})_{n\in\mathbb{N}}$. By the Arzelà-Ascoli theorem there is a subsequence $(f_{n_{j_{k}}})_{k\in\mathbb{N}}$ of $(f_{n_{j}})_{j\in\mathbb{N}}$ which converges uniformly on compact subsets of $E$ to some $g\in C(E)$. In particular, the sequence $(f_{n_{j_{k}}})_{k\in\mathbb{N}}$ converges almost everywhere to $g$, and so does every subsequence of $(f_{n_{j_{k}}})_{k\in\mathbb{N}}$. On the other hand, since $(f_{n_{j_{k}}})_{k\in\mathbb{N}}$ converges to $f$ in $L_{p}(E)$, there is a subsequence of $(f_{n_{j_{k}}})_{k\in\mathbb{N}}$ converging to $f$ almost everywhere. Hence we have $g = f$ almost everywhere.

Since at most one member of $C(E)$ can be equal almost everywhere to $f$, we deduce from above that there is some $g\in C(E)$ which is equal to $f$ almost everywhere such that every subsequence of $(f_{n})_{n\in\mathbb{N}}$ has a subsequence converging to $g$ uniformly on compact subsets of $E$. By this result applied to the sequence $(f_{n})_{n\in\mathbb{N}}$ in $C(E)$, we conclude that $(f_{n})_{n\in\mathbb{N}}$ converges to $g$ uniformly on compact subsets of $E$. This shows the first part.

Now suppose that $f\in C(E)$ and that $|U\cap E| > 0$ for all open $U\subseteq\mathbb{R}^{d}$ with $U\cap E \neq \emptyset$. By the first part there is some $g\in C(E)$ such that $(f_{n})_{n\in\mathbb{N}}$ converges to $g$ uniformly on compact subsets of $E$ and $f = g$ almost everywhere. The set $A = \{x\in E : f(x) \neq g(x)\}$ is a relatively open subset of $E$ with $|A| = 0$, so by assumption we have $A = \emptyset$. Hence $f = g$ pointwise and the second part follows.

Dean Miller
  • 7,015
  • 1
  • 7
  • 34
  • $K$ non-empty doesn't seem to be the correct assumption and something like "the support of the restriction to $K$ of the Lebesgue measure is $K$" looks better. For example, if $K$ has an isolated point, you can modify the functions at that point to be anything and still have convergence in mean of order $p$ as well as equicontinuity. – P. P. Tuong Apr 06 '25 at 11:28
  • @P.P.Tuong Thanks for the suggestion. I am struggling a bit to see how the suggestion "the support of the restriction to $K$ of the Lebesgue measure is $K$" sharpens the result, because in this case you can still have the limit function differing from a continuous function on a set of measure zero. Convergence in $L^{1}$ does not appear to be enough to imply uniform convergence (or even pointwise convergence) at isolated points. It may be needed to assume that the limit function is continuous to deduce uniform convergence. – Dean Miller Apr 06 '25 at 11:49
  • The answer to OP's question is indeed negative, but I imagine what was really meant to be asked was

    "Let $(f_k){n\in\mathbb N}$ be an equicontinuous sequence of integrable functions on $\mathbb R^n$ whch converges in mean. Does $(f_k){n\in\mathbb N}$ converge uniformly on every compact subset of $\mathbb R^n$?"

    in which case the answer is positive since every compact subset of $\mathbb R^n$ is contained in a compact subset of $\mathbb R^n$ on which the restriction of the Lebesgue measure has full support.

    – P. P. Tuong Apr 06 '25 at 12:01
  • @P.P.Tuong Thanks for mentioning this. I just added a result of that form to the answer. – Dean Miller Apr 08 '25 at 09:03