5

I'm wondering if a closed subspace of $L^p [a,b](1\leq p<\infty)$ consisting of continuous functions must be finite-dimensional?

I already know it's true for $L^2([a,b]): $ this question

Mittens
  • 46,352
  • 2
    Welcome to [math.se] SE. Take a [tour]. You'll find that simple "Here's the statement of my question, solve it for me" posts will be poorly received. What is better is for you to add context (with an [edit]): What you understand about the problem, what you've tried so far, etc.; something both to show you are part of the learning experience and to help us guide you to the appropriate help. You can consult this link for further guidance. – Another User Oct 15 '22 at 15:20

2 Answers2

5

The answer to the OP's question is yes, a closed subspace $M$ of $L_p([a,b])$ consisting of continuous functions is finite dimensional. Since $M\subset\mathcal{C}([a,b])$, then $M\subset L_\infty$ and the result follows from a Theorem by Grothendieck (see, Rudin, W., Functional Analysis, McGraw Hill, Chapter 5.)

Theorem: Suppose $(X,\mathscr{B},\mu)$ is a finite measure space, $M$ closed in $L_p(\mu)$ for some $0< p<\infty$ and $M\subset L_\infty(\mu)$. Then $M$ is finite dimensional.

The theorem above follows from the $L_2(\mu)$ case as we show below.

Sketch of the proof $1\leq p<\infty$: Without loss of generality assume $\mu(X)=1$. The case $p=2$ is the subject of this posting.

Notice that for all $1\leq r<\infty$, $L_\infty(\mu)\subset L_r(\mu)$ and that the inclusion map $\iota_r:f\mapsto f$ is injective and a continuous operator since $\|f\|_r=0$ iff $f=0$ $\mu$-almost surely, and $$\|\iota_r (f)\|_r=\|f\|_r\leq \|f\|_\infty$$

If $M$ closed in $L_p(\mu)$ for some $1\leq p<\infty$, then $M$ is also closed in $L_\infty(\mu)$: suppose $(f_n:n\in\mathbb{N})\subset M$ converges to some $f\in L_\infty(\mu)$, i.e. $\|f_n-f\|_\infty\xrightarrow{n\rightarrow\infty}0$, then $$\|f_n-f\|_p\leq\|f_n-f\|_\infty\xrightarrow{n\rightarrow\infty}0$$ Since $M$ is closed in $L_p(\mu)$, it follows that $f\in M$. An application of the open map theorem yields that $\iota^{-1}_p:(M,\|\;\|_p)\rightarrow (M,\|\;\|_\infty)$ is also bounded. Thus, there is a constant $K>0$ such that $$\|f\|_p\leq \|f\|_\infty\leq K\|f\|_p,\qquad f\in M$$

  • If $1\leq p\leq 2$, then for any $f\in M$ $$\|f\|_2\leq\|f\|_\infty\leq K\|f\|_p\leq K\|f\|_2$$ Thus $M$ is also closed in $L_2(\mu)$, and then the result in the posting cited above can be used.

  • If $2<p<\infty$ then \begin{align} \|f\|_p\leq\|f\|^{1-2/p}_\infty\|f\|^{p/2}_2 \end{align} It follows that $$\|f\|_\infty\leq K\|f\|_p\leq K\|f\|^{1-2/p}_\infty\|f\|^{p/2}_2$$ and so, $$\|f\|_2\leq \|f\|_\infty\leq K^{p/2}\|f\|_2$$ Consequently $M$ is closed in $L_2(\mu)$.

  • If $0<p<1$, the space $L_p(\mu)$ equipped with translation invariant metric $d_p(f, g)=\|f-g\|^p_p=\int|f-g|^p\,d\mu$ is complete metrizable linear space (an $F$-space). It is easy to prove that a set in $B\subset (L_p,d_p)$ is bounded iff $\sup\{\|f\|_p: f\in B\}<\infty$. It follows that a linear operator $P:(L_\infty(\mu),\|\;\|_\infty)\rightarrow (L_p(\mu),d_p)$ is bounded iff there is a constant $k>0$ such that $\|Pf\|_p\leq k\|f\|_\infty$. Conversely, a linear operator $R:(L_p(\mu),d_p)\rightarrow(L_\infty(\mu),\|\;\|_\infty)$ is bounded iff there is a contant $K>0$ such that $\|Rf\|_\infty\leq K\|f\|_p$. The proof of the Theorem for this case as we can see, becomes similar to that of $1\leq p<2$.

Mittens
  • 46,352
  • This proof is true for $1\leq p<2$.However If $2< p$,we can't conclude that M is closed in $L^2$ since $\lVert f\rVert_2 \leq K\lVert f\rVert_p$ is always true,and a closed space in $L^p,p>2$ doesn't necessarily closed in $L^2$ – Shuliang Li Oct 15 '22 at 17:26
  • Nice reduction to $p=2.$ !! (+) – Ryszard Szwarc Oct 16 '22 at 17:38
  • On $L^p(0,\infty)$ it is easy as one may consider the closed linear span of the functions $f_n(x)=f(x-n),$ where $f(x)=\min{x,1-x}.$ for $0\le x\le 1$ and $0$ otherwise. Concerning $L^p((0,1)$ I have found this I am not certain it is correct. – Ryszard Szwarc Oct 17 '22 at 14:11
  • @RyszardSzwarc: Thanks for your comment. In fact, as I just found out the problem is solved by a Theorem by Grothendieck for all $0<p<\infty$, that is in any probability space $(X,\mathscr{B},\mu)$, for any $0<p<\infty$, if $M\subset L_p(\mu)\cap L_\infty(\mu)$ and $M$ closed in $L_p(\mu)$, then $M$ is finite dimensional. I suspect you already about this. – Mittens Oct 17 '22 at 16:35
  • I was not aware of the Grothendieck result. It's a nice way of yours to handle the case $p>2.$ – Ryszard Szwarc Oct 17 '22 at 17:40
  • Very nice. We discussed a baby version of that theorem of Grothendieck recently, here. – Giuseppe Negro Oct 17 '22 at 18:01
  • @GiuseppeNegro: long time no see... yes, I recall know your posting! – Mittens Oct 17 '22 at 18:04
  • @OliverDíaz A nice proof!Thank you! – Shuliang Li Oct 19 '22 at 01:55
  • Dear @OliverDíaz The property of being not dense can be achieved in an artificial way by replacing $X$ with the direct sum $X\oplus X.$ – Ryszard Szwarc Nov 04 '22 at 17:52
  • @OliverDíaz I meant for example $\ell^1\subset c_0$ is dense and not closed, but $\ell^1\oplus {0}\subseteq c_0\oplus c_0$ is not closed and not dense. Or I misunderstood something – Ryszard Szwarc Nov 04 '22 at 18:12
4

The identity map $Id:(X ,\|\cdot\|_\infty) \to (X,\|\cdot\|_p)$ is linear, continuous, and bijective, hence invertible. This implies there is $c>0$ such that $\|x\|_\infty \le c \|x\|_p$ for all $x\in X$.

But this implies $\|x\|_2 \le \sqrt{b-a} \|x\|_\infty \le c \sqrt{b-a} \|x\|_p$ for all $x\in X$. Hence, $X$ is a closed subspace of $L^2$, and hence finite-dimensional.

daw
  • 54,637
  • 2
  • 44
  • 85