4
$\begingroup$

Consider an upper semicontinuous function $\phi: \Omega \to (-\infty, \infty]$, in the sense that $\phi = \phi^*$, where $\phi^*$ denotes the upper semicontinuous regularization $$ \phi^*(z) = \varlimsup_{\zeta \to z} \phi(\zeta) $$ and $\Omega$ is some domain in $\mathbb{C}^n$, let's say strictly pseudoconvex if it helps.

Forming the usual Perron-Bremermann envelope of $\phi$, i.e. $$ U(z) = \sup\{ u(z) : u \in \mathcal{PSH}(\Omega), u \le \phi \}, $$ it is clear (assuming that $U$ is locally bounded) that $U^* \le \phi^* = \phi$, so $U^*$ is a member in the defining family for $U$, and thus $U \ge U^*$ and consequently, $U = U^*$. In particular, $U$ is already upper semi-continuous, and thus plurisubharmonic without having to take the upper semicontinuous regularization.

Now, let us consider the following variation, where we take the envelope only using upper bounded functions: $$ V(z) = \sup\{ u(z) : u \in \mathcal{PSH}(\Omega), u \le \phi, \sup_\Omega u < +\infty \}. $$ Note that in general, $V < U$ (take for example $\phi$ as the Poisson kernel with a pole at $z=1$ on the unit disc: then $U = \phi$, but $V = 0$).

Question: Is it true that $V$ is upper semicontinuous and thus plurisubharmonic? The problem being, of course, that $V$ is in general not upper bounded, so it doesn't follow immediately that $V^*$ is a member of the defining family for $V$.

$\endgroup$
2
  • $\begingroup$ Upper semicontinuous functions take their values in $[-\infty,\infty)$, not $(-\infty,\infty]$... $\endgroup$
    – user111
    Commented Apr 4, 2021 at 13:11
  • $\begingroup$ @user111 That's the most common convention, for sure, but I specifically want to allow things like $\phi = -\log|f|$ where $f$ is holomorphic. (Allowing $\phi$ to take the value $-\infty$ is ok.) $\endgroup$
    – mrf
    Commented Apr 5, 2021 at 21:24

1 Answer 1

1
$\begingroup$

It is not true, even for strictly pseudoconvex domains. Consider for $1>a>0$ the function $$ u_a(z,w) = (-\log(1-|w|^2))^a+ \max\{-1, \sum_{k=1}^{\infty} \frac{1}{2^k}\log|z-\frac{1}{2^k}|\} $$ defined on the unit ball $B$ in $\mathbb{C}^2$, and denote its restriction to $\partial B$ by $\phi$. Then $\phi: \partial B \rightarrow [-1, \infty]$ with $\phi^* = \phi$, and there exists a harmonic extension $h_\phi: B \rightarrow [-1, \infty]$ since $u_a$ is dominated by the plurisuperharmonic function $$v_b(z,w) = (-\log(|z|^2))^b,$$ with $1>b > a$. With $h_\phi$ as dominating function, we have $$V\leq u_a$$ by the comparison principle (this requires a little bit of work), but $u_a - \epsilon v_b \leq V$ for all $\epsilon >0$. Hence $V=u_a$ outside $\{z=0\}$, and in particular, $V$ is not continuous on $\Omega$.

This constitutes a counterexample since on strictly pseudoconvex domains with $h_\phi$ lower semicontinuous on $\bar \Omega$, $V$ is always lower semicontinuous, so upper semicontinuity would imply that $V$ above is continuous. To see that $V$ is lower semicontinuous, note that we for any element $u$ in the defining family may associate an upper semicontinuous function $u^*\mid_{\partial \Omega}$ on the boundary. By Katětov–Tong insertion, there exists a continuous function on the boundary between $u^*\mid_{\partial \Omega}$ and $\phi$, which one may extend to a maximal, continuous plurisubharmonic function $\tilde u$ such that $u \leq \tilde u \leq h_\phi$. This shows that $V$ is lower continuous, as it may be written as an envelope of a family of continuous functions.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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