1
$\begingroup$

Let $C(\mathbb R^d)$ be the space of continuous functions on $\mathbb R^d$, and $C_{lip}(\mathbb R^d)\subset C(\mathbb R^d)$ be the subspace of Lipschitz functions. We endow $C_{lip}(\mathbb R^d)$ with the following topology: $(f_n)_{n\ge 1} \subset C_{lip}(\mathbb R^d)$ converges to $f\in C_{lip}(\mathbb R^d)$ iff

$$\lim_{n\to\infty} \left\{\left|\int_{\mathbb R^d}(f_n-f)(x)u(x)dx\right| + \left|\int_{\mathbb R^d}\nabla(f_n-f)(x)\cdot w(x)dx\right|\right\} = 0,$$

for all $u:\mathbb R^d\to\mathbb R$ and $w:\mathbb R^d\to\mathbb R^d$ satisfying

$$\int_{\mathbb R^d}|u(x)|(1+|x|)dx<\infty \quad\mbox{and}\quad \int_{\mathbb R^d}|w(x)|dx<\infty.$$

Is this topology metrizable?

PS: The motivation is to apply Baire's theorem. Any answers or comments are highly appreciated.

PS2: Thank Iosif Pinelis for the answer. I wish to ask a further question: Consider the completion $\overline{C}_{lip}(\mathbb R^d)$ of $C_{lip}(\mathbb R^d)$ w.r.t. this metric. Could we show that any linear continuous function $T: \overline{C}_{lip}(\mathbb R^d)\to\mathbb R$ must be of the form

$$T(f)=\int_{\mathbb R^d}f(x)u(x)dx+\int_{\mathbb R^d}\nabla f(x)w(x)dx?$$

$\endgroup$
3
$\begingroup$

I think the answer is no, no topology with this convergence is metrizable, for the following reason.

The space of the Lipschitz functions on $\mathbb{R}^r$, with the norm $\|f\|_{\text{Lip}}:=\big\|{f\over1+|x|}\big\|_\infty+\|\nabla f\|_\infty$ has a pre-dual Banach space $X$ (see Dual space of the completion of the space of Lipschitz functions). As said there, the convergence you are considering is the corresponding weak* convergence, that is, the point-wise convergence of functionals on $X$. edit here is a nice online reference.

It is well known that the weak* topology of the dual of an infinite dimensional Banach space $X$ is never metrizable (although its trace on bounded sets is). But is there any distance $d$ on $X^*$, whose metric topology $\tau_d$ has the same convergence of the weak* topology $\tau_{w^*}$?

The answer is no: assume by contradiction $d$ is such a distance. Then $(X^*,\tau_d)\to (X^*,\tau_{w^*})$ is sequentially continuous, hence continuous, that is $\tau_d\subset \tau_{w^*}$. On the other hand, the evaluation on elements of $X$ are continuous functionals on $(X,\tau_d)$ so $\tau_d\supset \tau_{w^*}$ because $\tau_{w^*}$ is the weaker topology that makes them continuous. So $\tau_d=\tau_{w^*}$, a contradiction, because as said $\tau_{w^*}$ is not metrizable.

$\endgroup$
  • $\begingroup$ What i'm saying is that the op defines a topologic, non metrizable convergence. It is the convergence of a w* topology and of no metric topology. $\endgroup$ – Pietro Majer Nov 23 '19 at 4:19
  • $\begingroup$ I think not really "iff": the convergence $d(f_n,f)\to0$ is equivalent to the OP's condition only on bounded sets, in general it is weaker and depends on the particolar chosen dense set. $\endgroup$ – Pietro Majer Nov 23 '19 at 4:27
  • $\begingroup$ And yes, of course the question should have been more precisely phrased: a) Is this convergence associated to some topology? And b) Is this convergence associated to some metric topology?. $\endgroup$ – Pietro Majer Nov 23 '19 at 4:36
  • $\begingroup$ Thanks for the answer. Indeed, what I am saying is that, this convergence defines the collection of open sets, and then the collection of open sets defines a topology. $\endgroup$ – Neymar Nov 23 '19 at 12:23
  • 1
    $\begingroup$ A convergence does not identify a topology; though here a natural choice is the w* topology. It is not metrizable, but it is metrizable on bounded sets. $\endgroup$ – Pietro Majer Nov 23 '19 at 12:28
2
$\begingroup$

According to the answer by Pietro Majer, the convergence in question is incompatible with a metrizable topology.

On the hand, as Pietro Majer commented, the $w^*$ topology, which is compatible with this convergence, is metrizable on bounded sets. Let us provide details on this comment.
Let $$\||u\||:=\int_{\mathbb R^d}|u(x)|(1+|x|)dx\quad\quad\mbox{and}\quad \|w\|:=\int_{\mathbb R^d}|w(x)|dx<\infty.$$ Let $U$ be the normed space of all $u$ with $\||u\||<\infty$, and let $W$ be defined similarly, so that $W=L^1$. These two normed spaces are separable. Let $\{u_j\colon j\in\mathbb N\}$ and $\{w_j\colon j\in\mathbb N\}$ be corresponding countable dense sets in $U$ and $W$. Let $$d_{u,w}(g,f):=\left|\int_{\mathbb R^d}(g-f)(x)u(x)dx\right| + \left|\int_{\mathbb R^d}\nabla(g-f)(x)\cdot w(x)dx\right|. $$

Then it is easy to see that, for any sequence $(f_n)$ in $C_{lip}(\mathbb R^d)$ bounded with respect to the norm $C_{lip}(\mathbb R^d)\ni g\mapsto\sup_x|g(x)|+\sup_{x\ne y}\frac{|g(x)-g(y)|}{|x-y|}$, we have $f_n\to f$ iff $d_{u_j,w_j}(f_n,f)\to0$ for each natural $j$ iff $d(f_n,f)\to0$, where $d$ is the metric defined by $$d(g,f):=\sum_{j=1}^\infty\frac1{2^j}\frac{d_{u_j,w_j}(g,f)}{1+d_{u_j,w_j}(g,f)}. $$

$\endgroup$
  • $\begingroup$ I was thinking about this, but I thought I saw a flaw. What is the approximation argument that would ensure $f_n$ goes to $f$ if the distances $d_j(f_n,f)$ go to zero? $\endgroup$ – Pierre PC Nov 22 '19 at 16:47
  • $\begingroup$ Thanks for the reply. I've a related question (detailed above). Do you think the answer is also yes? $\endgroup$ – Neymar Nov 22 '19 at 17:43
  • $\begingroup$ @PierrePC : I have added a detail concerning your question. $\endgroup$ – Iosif Pinelis Nov 22 '19 at 19:07
  • 1
    $\begingroup$ @IosifPinelis Thank you, but I still don't see it... To do it the way I think of it, I need the estimate $|f_n-f|=O(1+|x|)$ uniformly in $n$. I get that $\langle f,u\rangle$ is uniquely determined by $\langle f,u_j\rangle$, but I cannot seem to see why it would suffice. Is there a hidden Banach-Steinhaus somewhere? $\endgroup$ – Pierre PC Nov 22 '19 at 19:54
  • 1
    $\begingroup$ @IosifPinelis I think the question is clarified a bit if we put it in a more abstract setting: For a dual of a separable infinite dimensional Banach space $X$, is there a distance on $X^*$ that induces the point-wise convergence of bounded functionals ---that is, the same convergence of the w*-topology? I think the answer is always no (see my answer) $\endgroup$ – Pietro Majer Nov 24 '19 at 10:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.