I am reading Optimal Transport for Applied mathematicians by Filippo Santambrogio and in the first chapter he describes weak convergence and weak-* convergence in terms of a Banach space and its dual. He then goes on to (by an abuse of notation, apparently) extend this concept to signed measures on a separable, locally compact metric space $X$ when he states the Riesz representation theorem by saying:

Let $\mathcal{X} = C_0(X)$ be the space of continuous function on X vanishing at infinity, i.e. $f \in C_0(x) \iff f \in C(X)$, and for every $\epsilon > 0$, there exists a compact subset $K \subset X$ such that $|f| < \epsilon$ on $X$ \ $K$. Let us endow this space with the sup norm since $C_0(X) \subset C_b(X)$ (this last space being the space of bounded continuous functions on $X$). Note that $C_0(X)$ is a Banach space and thaat it is a closed subset of $C_b(X)$. Then every element of $\mathcal{X}'$ (the dual of $\mathcal{X}$) is represented in a unique way as an element of $\mathcal{M}(X)$: for all $\xi \in \mathcal{X}'$ there exists a unique $\lambda \in \mathcal{M}(X)$ such that $\langle \xi, \phi \rangle = \int \phi d \lambda$ for every $\phi \in \mathcal{X}$; moreover, $\mathcal{X}'$ is isomorphic to $\mathcal{M}(X)$ endowed with the norm $||\lambda|| := |\lambda|(X)$.

For signed measures of $\mathcal{M}(X)$, we should call weak-* convergence the convergence in the duality with $C_0(X)$. Yet another interesting notion of convergence is that in duality with $C_b(X)$. We will call it (by abuse of notation) weak convergence and denote it through the symbol $\mu_n \rightharpoonup \mu$ iff for every $\phi \in C_b(X)$ we have $\int \phi d\mu_n \rightarrow \int \phi d \mu$.

My question is as follows: is weak-* convergence (i.e. weak convergence in duality with $C_0(X)$) the same as weak convergence in duality with $C_b(X)$, except for the fact that in the former condition we require $\int \phi d\mu_n \rightarrow \int \phi d \mu$ for all $\phi \in C_0(X)$ as opposed to $C_b(X)$?


Firstly, we remark that $\mathcal{M}(X)$, as a set, is the collection of all finite signed Radon measures (assuming that we are working with the scalar field $\mathbb{R}$). That is, $\mu\in\mathcal{M}(X)$ if $\mu$ is a finite signed measure, outer-regular in the sense that for each Borel subset $A\subseteq X$, $\mu(A)=\inf\{\mu(U)\mid A\subseteq U\mbox{ and }U\mbox{ is open}\}$, and inner-regular for open set in the sense that for each open set $U\subseteq X$, $\mu(U)=\sup\{\mu(K)\mid K\subseteq U\mbox{ and }K\mbox{ is compact.}\}$. Note that if $X$ is second countable, then every finite signed measure is automatically Radon. (see Folland, Real Analysis, Theorem 7.8).

For your question, the two modes of convergence are not the same. Obviously, given a sequence $(\mu_{n})$ in $\mathcal{M}(X)$ and $\mu\in\mathcal{M}(X)$, $\mu_{n}\rightharpoonup\mu\Rightarrow\mu_{n}\rightarrow\mu$ in weak*-topology. The converse does not hold in general. The following is a counter-example.

Let $X=\mathbb{R}$, equipped with the usual topology. For each $n\in\mathbb{N}$, let $\mu_{n}(A)=\lambda(A\cap[n,n+1])$, where $\lambda$ is the usual Lebesgue measure, then $\mu_{n}\in\mathcal{M}(X)$. Let $\mu=0$. Clearly $\mu_{n}\rightarrow\mu$ in weak*-topology. However, $\mu_{n}\not\rightharpoonup\mu$ because the constant function $f=1$ is an element in $C_{b}(X)$ and $\int fd\mu_{n}=1\neq\int fd\mu$.


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.