Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $K$ be a compact Hausdorff space. Denote by $ca_r(K)$ the set of all countably additive, signed Borel measures which are regular and of bounded variation. Let $(\mu_n)_{n\in\mathbb{N}}\subset ca_r(K)$ be a bounded sequence satisfying $\mu_n\geq 0$ for all $n\in\mathbb{N}$. Let $f:K\rightarrow\mathbb{R}^n$ be continuous and bounded and assume that the limit $z:=\lim_{n\rightarrow\infty}\int fd\mu_n\in\mathbb{R}^n$ exists. Is there a $\mu\in ca_r(K)$, $\mu\geq 0$, such that $z=\int fd\mu$? This question is related to this.

share|cite|improve this question
You want $\mu$ to work for this particular $f$ or all $f\in C(K)$? – Davide Giraudo Sep 19 '12 at 15:26
up vote 2 down vote accepted

Yes. Let the limit be $L$. Let $\min \{f(x): x \in K\} = f(a)$, $\max \{f(x): x \in K\} = f(b)$. Now $ \|\mu_n\| f(a) \le \int f \ d\mu_n \le \|\mu_n\| f(b)$, so there is $\lambda_n \in [0,1]$ such that $\int f \ d\mu_n = \|\mu_n\| (\lambda_n f(a) + (1-\lambda_n) f(b)$. Taking a subsequence (and noting that $\|\mu_n\|$ is bounded), we can assume $\|\mu_n\|$ and $\lambda_n$ converge, and then $\mu = \lim_{n} \|\mu_n\| (\lambda_n \delta(a) + (1-\lambda_n) \delta(b)$ (where $\delta(a)$ and $\delta(b)$ are the unit point masses at $a$ and $b$) satisfies the requirements.

share|cite|improve this answer
Thanks for the very nice answer! I realized what the problems are and therefore changed the question and wrote down the full question I had in mind... – Andy Teich Sep 19 '12 at 15:50

It suffices to handle multidimensional $f$. To simultaneously handle $f$ and $g$ you could just look at the function $h=(f,g)\in\mathbb{R}^{n+1}$.

Taking a subsequence, we may assume that $\mu_n(K)\to m$ for some $0\leq m<\infty$. If $m=0$, we let $\mu=0$ otherwise we use the following argument.

Assume that $m>0$. Define $\nu_n=\mu_n/\mu_n(K)$ so that $\nu_n$ are probability measures. For every $n$, the value of $\int f(x)\,\nu_n(dx)$ belongs to the convex hull of $f(K)$, which is a compact subset of $\mathbb{R}^n$. The limit $z/m$ also belongs to this set. That is, we can write $z/m=\sum_{i=1}^N \lambda_i\, f(k_i)$ for some weights $\lambda_i>0$ with $\sum_{i=1}^N \lambda_i=1$.

Therefore $z=\int f(x)\,\mu(dx)$ where $\mu=m\sum_{i=1}^N \lambda_i \delta_{k_i}.$

This is just a modification of Robert Israel's answer.

share|cite|improve this answer
What is if we drop boundedness of $f$? I realized that in my case, $f$ is only bounded in one component, and not in the others... – Andy Teich Sep 20 '12 at 8:25
@AndyTeich So $f$ is not continuous? Then the result is false, even for bounded $f$. – Byron Schmuland Sep 20 '12 at 12:06
I meant the case when $f$ is unbounded but continuous. But then $f(K)$ is still compact, which I had overseen... – Andy Teich Sep 20 '12 at 14:12

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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