2
$\begingroup$

Let $(C, \|\cdot\|)$ be the Banach space of continuous paths $x: [0,1]\rightarrow\mathbb{R}^d$ starting at zero with sup-norm $\|\cdot\|$.

Let further $B\subset C$ be the subspace of $0$-started continuous paths of bounded variation, i.e.

$$B = \big\{x\in C \ \big| \ \|x\|_1:=\sup\nolimits_{(t_\nu)}{\textstyle\sum_\nu}|x_{t_\nu} - x_{t_{\nu-1}}|<\infty \ \text{ and } \ x(0)=0\big\}$$

where the above sup runs over all partitions $(t_\nu)$ of $[0,1]$.

The set $B$ can be endowed with both the (uniform) norm $\|\cdot\|$ and the (1-variation) norm $\|\cdot\|_1$.

Do you know if the Borel $\sigma$-algebras on $(B, \|\cdot\|)$ and $(B, \|\cdot\|_1)$ coincide?

Any references or hints are welcome.

Edit: As pointed out by Gerald Edgar, we narrowed $B$ down to paths started at zero to make the question more meaningful.

Edit 2: As Gerald Edgar pointed out in the comments to his answer below, the answer to this question is no, the Borel $\sigma$-algebras cannot coincide because there are more $\|\cdot\|_1$-open sets in $B$ than there are $\|\cdot\|$-Borel sets in $C$.

$\endgroup$

1 Answer 1

2
$\begingroup$

I think there is a problem with a mere semi-norm. The constant functions have variation distance $0$ from each other. Any variation-open set contains either all the constants, or none of them. Therefore, any variation-Borel set contains either all the constants, or none of them.

Choose $\mathbf u \in \mathbb R^n \setminus \{0\}$. Take the subset $\{0,\mathbf{u}\} \subseteq B$ of two constant functions. It is sup-norm closed, so it is a uniform-Borel set. But it is not a variation-Borel set.


Maybe a better question would restrict to the subset of $B$ with $x(0) = 0$. Then at least we have a norm.


rmcerafl pointed out a problem with the following. For now, I will leave it here for reference.

Let $C_0 = \{x \in C : x(0)=0\}$ with the uniform norm. Let $B_0 = \{x \in B : x(0)=0\}$ with the variation norm. Both of these are complete separable metric spaces. The inclusion map $i : B_0 \to C_0$ is continuous and injective. Then citing a result or two from descriptive set theory*, we get: $i(B_0)$ is a Borel subset of $C_0$ and $i : B_0 \to i(B_0)$ is a Borel-isomorphic map. This shows, in particular, the variation-Borel subsets of $B_0$ coincide with the uniform-Borel subsets of $i(B_0)$.


${}^*$ Proposition 8.3.5 and Theorem 8.3.7 in:

Cohn, Donald L., Measure theory, Boston, Basel, Stuttgart: Birkhäuser. IX, 373 p. DM 42.00 (1980). ZBL0436.28001.

$\endgroup$
5
  • $\begingroup$ Thank you, I edited the question accordingly. $\endgroup$
    – rmcerafl
    Jun 14 at 14:36
  • $\begingroup$ I don't think $(B_0, \|\cdot\|_1)$ is separable, see for instance Theorem 1.25 in: Friz and Victoir, Multidimensional Stochastic Processes as Rough Paths. $\endgroup$
    – rmcerafl
    Jun 14 at 15:19
  • 1
    $\begingroup$ I think you are right. References I consulted only discuss non-separability of BV and of NBV (=right-continuous BV), but not of continuous BV. If it is non-separable, I am inclinded to think that there are far too many Borel sets, then. $\endgroup$ Jun 14 at 16:12
  • 3
    $\begingroup$ There are $2^{\mathfrak c}$ variation-open sets in $B_0$, but only $\mathfrak c$ uniform-Borel sets in $C_0$. $\endgroup$ Jun 14 at 18:34
  • 1
    $\begingroup$ @GeraldEdgar. Nice! Except that your answer is in the comments while you've got a bunch of comments in your answer! $\endgroup$
    – Ruy
    Jun 15 at 20:08

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.