Example 18.14.7.A467F501617F Let $M=\mathbb P^2\times \mathbb P^1$, with projections $p,r$, and let $\sigma _ j$ be the standard Fubini–Study form on $\mathbb P^ j$. In the positive closed background
\[ \Theta =2p^*\sigma _2+r^*\sigma _1, \]
there is a logarithmic potential $\Psi $ of a smooth prime divisor containing the fiber over $E=[0:0:1]$. Its fiber infimum for the zero obstacle is finite and continuous off $E$ and tends to $+\infty $ at $E$. Nevertheless,
\[ P_\Theta [\Psi ]=P_\Theta [\Psi ]_{\mathcal I}=\Psi +q\circ p, \]
where $q\in \mathrm{PSH}(\mathbb P^2,\sigma _2)$ is bounded, $0\leq q\leq \log 4$ and $q(E)=\log 4$. Every ordinary representative of this model has zero extended mass, so both singularity volumes vanish. The two envelopes also agree below every smooth obstacle.
Proof.
Use homogeneous coordinates $[y_0:y_1:y_2]$ and $[z_0:z_1]$, and take
\[ H=\{ y_0z_0+y_1z_1=0\} ,\qquad \Psi =\log \frac{|y_0z_0+y_1z_1|^2}{(|y_0|^2+|y_1|^2+|y_2|^2)(|z_0|^2+|z_1|^2)}. \]
Outside $E$, the divisor is the graph $[z_0:z_1]=[-y_1:y_0]$. Its closure contains the whole fiber over $E$, so $H$ is prime. It is smooth outside that fiber by the graph description, and smooth along it because the derivatives in the $y_0$ and $y_1$ directions are $z_0$ and $z_1$, which cannot both vanish. Its product metric has curvature $p^*\sigma _2+r^*\sigma _1$, hence
\[ \Theta +\mathrm{dd}^{\mathrm{c}}\Psi =p^*\sigma _2+[H]. \]
Writing $S=|y_0|^2+|y_1|^2$, the Cauchy–Schwarz inequality gives the fiber obstacle
\[ G(y)=-\max _{z\in \mathbb P^1}\Psi (y,z) =\log \frac{S+|y_2|^2}{S},\qquad G(E)=+\infty . \]
It has no finite continuous extension at $E$.
We claim that the largest $\sigma _2$-plurisubharmonic potential below $G$ is
\[ q(y)=\begin{cases} \displaystyle \log \frac{4|y_2|^2}{S+|y_2|^2},& S\leq |y_2|^2,\\[6pt] \displaystyle \log \frac{S+|y_2|^2}{S},& S\geq |y_2|^2. \end{cases} \]
To see that $q$ is $\sigma _2$-plurisubharmonic, work on $y_2\ne 0$ with $x=(y_0/y_2,y_1/y_2)$ and $t=\log |x|^2$. Adding the Fubini–Study potential $L(x)=\log (1+|x|^2)$ gives
\[ q+L=\begin{cases} \log 4,& t\leq 0,\\ 2\log (1+\mathrm e^ t)-t,& t\geq 0. \end{cases} \]
This function and its first derivative are continuous at $t=0$. For $t>0$, its first two derivatives are $(\mathrm e^ t-1)/(\mathrm e^ t+1)\geq 0$ and $2\mathrm e^ t/(1+\mathrm e^ t)^2>0$. It is therefore a convex nondecreasing function of $t$, hence plurisubharmonic in $x$, across the unit sphere as well. The expression is constant near $x=0$, and $q=G$ is smooth near $y_2=0$, so $q$ is globally $\sigma _2$-plurisubharmonic and bounded.
If $b\in \mathrm{PSH}(\mathbb P^2,\sigma _2)$ lies below $G$, then $b+L\leq \log 4$ on $|x|=1$, and the maximum principle on the unit ball gives $b\leq q$ inside it. Outside the ball, the obstacle already gives $b\leq G=q$. For $0<R=|x|^2\leq 1$, the inequality $q\leq G$ reads $4R\leq (1+R)^2$, and at $E$ it is automatic. The formula for $q$ also gives the asserted bounds, which proves the claim.
For every ordinary or ideal competitor $v$ for $\Psi $, the ordinary order or theorem 5.37.3 gives $\nu (v,H)\geq 1$. Divisorial removal and descent along the fibers, lemma 5.38.1, proposition 5.43.5, give $v=\Psi +b\circ p$ with $b\in \mathrm{PSH}(\mathbb P^2,\sigma _2)$. The bound $v\leq 0$ gives $b\leq G$ off $E$, hence everywhere with the extended value at $E$. Thus $v\leq \Psi +q\circ p$. Conversely, this upper bound is a competitor for both envelopes, since it is nonpositive and differs boundedly from $\Psi $. This proves the envelope identities directly.
On $M\setminus H$, bounded local truncations of $b$ have products pulled back from a two-dimensional base, so their third products vanish by lemma 5.28.4, and plurifine locality gives zero extended mass. Divisorial removal puts every ordinary representative of the model in this form, which proves both volume assertions. Finally, since the model is ordinarily equivalent to $\Psi $, proposition 16.28.62 gives $\Psi \in \mathcal E_{\mathrm{env}}(M,\Theta ,P_\Theta [\Psi ]_{\mathcal I})$, and proposition 18.12.2 gives equality below every smooth obstacle.
$\square$
Comments
Corrections, questions, and mathematical discussion are welcome. Sign in with GitHub to comment. Comments are public.