2
$\begingroup$

Let $(E, d)$ be a metric space, $\mathcal C_b(E)$ the space of all real-valued bounded continuous functions on $E$, and $\mathcal P(E)$ the space of all Borel probability measures on $E$. For $f \in \mathcal C_b(E)$, we define $$ L_f:\mathcal P(E) \to \mathbb R, \mu \mapsto \int_E f\mathrm d \mu. $$

Let $\tau$ be the initial topology on $\mathcal P(E)$ that is induced by the collection $\{L_f : f \in \mathcal C_b(E)\}$. If $(\mu_d)_{d\in D}$ is a net in $\mathcal P(E)$ and $\mu \in \mathcal P(E)$ then $\mu_d \to \mu$ in $\tau$ IFF $L_f(\mu_d) \to L_f(\mu)$ for all $f \in \mathcal C_b(E)$. We have the famous result, i.e.,

Portmanteau theorem Let $(\mu_n)_{n\in \mathbb N}$ be a sequence in $\mathcal P(E)$ and $\mu \in \mathcal P(E)$. The following statements are equivalent:

  • $\mu_n \to \mu$ in $\tau$;
  • $\int f \mathrm d \mu_n \to \int f \mathrm d \mu$ for every real-valued bounded Lipschitz function $f$ on $E$;
  • $\lim \inf \int f \mathrm d \mu_n \ge \int f \mathrm d \mu$ for every real-valued lower semi-continuous function $f$ on $E$ that is bounded from below;
  • $\liminf \mu_{n}(U) \ge \mu(U)$ for every open subset $U$ of $E$;
  • $\lim \mu_{n}(A) = \mu(A)$ for every Borel subset $A$ of $E$ such that $\mu(\partial A) = 0$.

Of course, $\liminf$ and $\limsup$ are well-defined for nets.

Are there references that contain a version of Portmanteau theorem where $(\mu_n)_{n\in \mathbb N}$ is replaced by a net $(\mu_d)_{d\in D}$?

Thank you so much for your help!

$\endgroup$
2
  • 2
    $\begingroup$ Bogachev does that both in the second volume of "Measure Theory" and in "Weak Convergence of Measures." Though the additional generality is only important to study weak convergence on more general topological spaces. $\endgroup$ Commented Nov 28, 2022 at 22:47
  • $\begingroup$ @MichaelGreinecker Thank you so much for your references! Could you post it as an answer? It seems Theorem 6.1 in Parthasarathy's Probability measures on metric spaces also provides such a version. $\endgroup$
    – Analyst
    Commented Nov 28, 2022 at 22:56

0

You must log in to answer this question.