I am trying to understand in which metric spaces the metric is jointly measurable.

There exist a metric space $(X,d)$ for which the Borel $\sigma$-algebra, does not coincide with the product Borel $\sigma$-algebra, that is $\mathcal{B}(X)\otimes \mathcal{B}(X) \subsetneq B(X\times X)$. Every construction, I have encountered, of such a (non-separable) metric space one considers a metric space of cardinality strictly greater than the continuum $\mathfrak{c}$. For in such a metric space the diagonal $I=\{(x,x):x\in X\}\in B(X\times X)$, is not an element of $\mathcal{B}(X)\otimes \mathcal{B}(X)$ (c.f. Nedoma's pathology). This also proves that $d$ is not jointly measurable, since $d^{-1}(\{0\})=I\not\in\mathcal{B}(X)\otimes \mathcal{B}(X)$.

For metric spaces $(X,d)$ with $\text{card}(X)\leq \aleph_0$, we obviously have that $d$ is jointly measurable, but the area of interest is : What happends when the cardinality is $\aleph_0 < \text{card}(X) \leq \mathfrak{c}$?

More specifically, I have been wondering about the following questions: Let $(X,d)$ be a metric space,

  1. Is the metric $d$ jointly-measurable, if $\aleph_0 < \text{card}(X) \leq \mathfrak{c}$?
  2. If not, is there equivalence between joint measurability of $d$ and separability of $X$?
  3. If not, do you know property of $(X,d)$, weaker than separability, that ensures joint-measurability of $d$?

1 Answer 1


One can probably not go too far from separability, there must always exist a sub-$\sigma$-algebra that looks like the Borel $\sigma$-algebra of a separable metric space.

If $\mathcal{A}$ is a family of subsets of a set $X$ and $A\in\sigma(A)$, then there exists a countable family $\mathcal{C}\subseteq\mathcal{A}$ such that $A\in\sigma(\mathcal{C})$. In particular, if the diagonal $D\subseteq X\times X$ is in the $\sigma$-algebra generated by measurable rectangles, then the countable family of sets $\mathcal{C}$ must have the property that for each $x,y\in X$ with $x\neq y$, there is $C\in\mathcal{C}$ such that $x\in C$ and $y\notin C$ or $x\notin C$ and $y\in C$.

Let $\langle C_n\rangle$ be a sequence that lists all elements of $\mathcal{C}$. Then the Marczewski-funtion $g:X\to\mathbb{R}$ given by $$g(x)=\sum_n 1/3^n 1_{C_n}(x)$$ has the property that $\sigma(\mathcal{C})$ consists exactly of inverse images of Borel sets in $\mathbb{R}$. Since $\mathcal{C}$ separates points, it is injective. Then $d:X\times X\to\mathbb{R}$ given by $d(x,y)=|g(x)-g(y)|$ turns $X$ into a separable metric space with Borel $\sigma$-algebra $\sigma(\mathcal{C})$.

However, separability is not necessary. Endow $\omega_1$ with the discrete metric, so it becomes a non-separable metric space with Borel $\sigma$-algebra $2^{\omega_1}$. In this case, the metric is still jointly measurable, for one can show that $2^{\omega_1}\otimes2^{\omega_1}=2^{\omega_1\times\omega_1}$. The proof can be found in:

V. Rao, On discrete Borel spaces and projective sets Bull. Amer. Math. Soc. Volume 75, Number 3 (1969), 614-617.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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