Let $\mathcal{M}^{+}(\mathbb{R}_{+})$ be space of non-negative Radon measures on $\mathbb{R}_{+}$ with bounded total variation and define the metric $\rho$ on $\mathcal{M}^{+} (\mathbb{R}_{+})$ as $$ \rho(\mu,\nu)= \sup \left \{ \int_{\mathbb{R}_{+}} \psi d (\mu - \nu) ~|~ \psi \in C^{1}(\mathbb{R}_{+}), \|\psi \|_{\infty} \le 1 , \|\partial_{x} \psi \|_{\infty} \le 1 \right \} .$$ How to prove $\mathcal{M}^{+}(\mathbb{R}_{+})$ is complete w.r.t. $\rho$. I know that $$ \lim_{n \to \infty} \rho(\mu_{n},\mu) = 0 \iff \mu_{n} \to \mu~ \text{narrowly for}~ n \to \infty.$$ But how above equivalence can help us to prove the completeness ?

New contributor
Manoj Kumar is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
  • $\begingroup$ I think your "metric" is ill defined. Since each $\mu \in M$ may be unbounded, in the integral you may have the difference $\infty-\infty$. You need some condition which implies that this does not happen. $\endgroup$ – Dieter Kadelka Feb 15 at 9:24
  • $\begingroup$ @DieterKadelka You are right, I edited my question. $\endgroup$ – Manoj Kumar Feb 15 at 9:37

Assuming that the above equivalence is valid, all you have to show that $\lim_{m,n \to \infty} \rho(\mu_m,\mu_n) = 0$ implies uniform tightness of the sequence $(\mu_n)$. So let $\epsilon > 0$ be arbitrary. Then there is $n_0 \in \mathbb{N}$ with $\rho(\mu_m,\mu_n) \leq \epsilon$ for $m,n \geq n_0$, in particular $\mu_n(\mathbb{R}_+) - \mu_m(\mathbb{R}_+) < \epsilon$. Since $\mu_{n_0}$ is tight, there is $t_0 \in \mathbb{R}_+$ with $\mu_{n_0}((t_0,\infty)) \leq \epsilon$. Now let $\psi \geq 0$ be any function with the properties above and with $\psi|{[0,t_0]} \equiv 0$, $\psi|{[t_0+2,\infty]} \equiv 1$ and $\psi$ increasing. Such function exists. Then $$\mu_n((t_0+2,\infty)) \leq \int \psi~d\mu_n \leq \int \psi~d\mu_{n_0} + \epsilon \leq 2\epsilon.$$ It follows that $(\mu_n)$ is uniformly tight, hence there are limit points $\mu$ with respect to the narrow topology. But then $\mu_{n_k} \to \mu$ weakly for some subsequence $(\mu_{n_k})$, hence $\rho(\mu_{n_k},\mu) \to 0$ by the above equivalence. Since $(\mu_n)$ is Cauchy neceesarily $\mu$ is unique and $\lim_{n \to \infty} \rho(\mu,\mu_n) = 0$.

  • $\begingroup$ It is well known that $\mathcal{M}^+$ is Polish. Proving that a concrete metric is complete almost always is done as in my answer. $\endgroup$ – Dieter Kadelka Feb 15 at 10:50
  • $\begingroup$ Thanks for the answer, I just want to know what do you mean by- $\psi$ with the properties above in your answer? Also, will the existence of $\psi$ be shown through the standard argument of mollifiers? $\endgroup$ – Manoj Kumar Feb 15 at 13:52
  • $\begingroup$ The properties of $\psi$ in your definition of $\rho$, such a $\psi \in C^1$ etc. $\endgroup$ – Dieter Kadelka Feb 15 at 14:29
  • $\begingroup$ Concerning mollifier: You can do it by mollifier and also directly. Note that the interval $[t_0,t_0+2]$ has length $2$. $\endgroup$ – Dieter Kadelka Feb 15 at 14:57

Your Answer

Manoj Kumar is a new contributor. Be nice, and check out our Code of Conduct.

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.