Let $X$ be a compact metrizable topological space and let $f$ be a bounded, real valued, Borel function on $X$. Denoting by $P(X)$ the collection of all probability measures on $X$, consider the subset $$ Z(f):= \left\{μ∈P(X): \int_X fdμ=0\right\}. $$

Question: What is the most general condition one should require of $f$ to ensure that $Z(f)$ is closed in the weak* topology of $P(X)$ (seen within the dual Banach space of $C(X)$).

For obvious reasons $Z(f)$ is closed when $f$ is continuous, but it is also closed if, say, $f$ is the pointwise limit of an increasing sequence $\{f_n\}_n$ of non-negative continuous functions (hence lower semi-continuous). This is because $$ Z(f) = \bigcap_nZ(f_n), $$ so $Z(f)$ is closed.

On the other hand, if the Borel subset $A⊆X$ is not open, there exists a sequence $\{x_n\}_n$ in the complement of $A$, converging to a point $p$ in $A$. In this case the Dirac measures $δ_{x_n}$ all lie in $Z(1_A)$ (by $1_A$ I mean the characteristic function of $A$), and they converge weak*ly to $δ_p$, which isn't in $Z(1_A)$. Therefore $Z(1_A)$ is not closed!

  • 1
    $\begingroup$ Another trivial example is when $f$ is any strictly positive (resp., strictly negative) Borel function, for then we have $Z(f) = \emptyset$. This shows that the answer can't really have to do with the regularity of $f$. $\endgroup$ – Nate Eldredge Jul 13 '18 at 7:05
  • $\begingroup$ @NateEldredge, you are right. The answer will likely be a combination of positivity and regularity. Incidentally, the case I am mostly interested in involves a function $f$ of the form $g-h$, where both $g$ and $h$ are non-negative, $g$ is continuous, and $h$ is upper-continuous. $\endgroup$ – Ruy Jul 13 '18 at 11:50
  • $\begingroup$ if one assumes $f$ to be non-negative, then I'd say that $Z(f)$ is closed if and only if the lower semicontinuous envelope of $f$ (i.e. the maximal lsc function $\leq f$) has the same zeros of $f$. Not sure if this helps for the case $g-h$ mentioned above $\endgroup$ – Nicola Gigli Jul 13 '18 at 13:44

It is a fact, well-known and elementary, that a linear functional on a locally convex space is continuous if and only if its kernel is closed.

| cite | improve this answer | |
  • 1
    $\begingroup$ This is true, but I don't see how it helps answer the question? $P(X)$ isn't a vector space. $\endgroup$ – Nate Eldredge Jul 13 '18 at 7:02

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.