Pome

18.9 Analytic closureE3C1856717ED

E3C1856717ED

We now show that the dyadic approximants of corollary 18.7.9 converge in $\tau _{\omega ,0}$ exactly when the two envelopes agree. Let $\mathcal A_\omega $ be the set of ordinary classes with analytic singularities, with rational coefficients as in section 5.34; the rational curvature correction in corollary 18.7.9 keeps the approximants in this set. In the controlled ideal relations, the scale factor of that correction has to be taken into account.

Theorem 18.9.1.867EFC7DCD8D The closure and the sequential closure of $\mathcal A_\omega $ in $\tau _{\omega ,0}$ are both equal to

27F9EF4DC38B 18.9.2
\begin{equation} \mathcal G_\omega =\{ [u]:P_\omega [u]=P_\omega [u]_{\mathcal I}\} . \end{equation}

For every $u\in \mathrm{PSH}(X,\omega )$, the dyadic sequence $b_ j$ of corollary 18.7.9 satisfies

\[ [b_ j]\longrightarrow [u]\text{ in }\tau _{\omega ,0} \quad \Longleftrightarrow \quad [u]\in \mathcal G_\omega . \]

The approximants may be chosen with one positive rational analytic coefficient on all charts.

Proof. We use the notation of corollary 18.7.9: $m_ j=2^ j$, $a_ j$ has type $(1/m_ j,\mathcal I(m_ ju))$, $c_ j=(1+\epsilon _ j)^{-1}\uparrow 1$ is rational, and $b_ j=c_ ja_ j$. The ordinary and ideal envelopes of $b_ j$ agree and decrease to $P_\omega [u]_{\mathcal I}$. Thus the two envelope coordinates of $[b_ j]$ converge to those of $[u]$ exactly when $[u]\in \mathcal G_\omega $. It remains to show that the other coordinates always converge.

Fix $t>0$ and $a>1$, and choose $1<s<a$. For large $j$, one has $ac_ j>s$ and $m_ j\geq st/(s-1)$, so lemma 5.35.4 gives

\[ \mathcal I(atb_ j)=\mathcal I(atc_ ja_ j) \subseteq \mathcal I(sta_ j)\subseteq \mathcal I(tu). \]

Since $u\preceq a_ j\preceq b_ j$, the reverse relation $\mathcal I(atu)\subseteq \mathcal I(tb_ j)$ holds for every $j$. The factor $c_ j$ enters the weight in the first inclusion, which is why we chose $s<a$.

For a divisor $E$ on a fixed modification $\pi $, the Bergman divisorial estimate gives

\[ c_ j\max \left\{ \nu _ E(u)-\frac{A(E)}{m_ j},0\right\} \leq \nu _ E(b_ j)\leq c_ j\nu _ E(u)\leq \nu _ E(u), \]

where $A(E)$ is the log discrepancy, computed over a coordinate neighborhood of the center of $E$. Each coefficient therefore converges to $\nu _ E(u)$, and the summability in lemma 18.8.1 gives $d_\pi ([b_ j],[u])\to 0$ by dominated convergence. This proves convergence of all the remaining coordinates.

Finally, analytic classes have equal envelopes by proposition 18.4.8. The map $[u]\mapsto (P_\omega [u],P_\omega [u]_{\mathcal I})$ is continuous into $L^1\times L^1$, whose diagonal is closed. Thus every limit of a net of analytic classes lies in $\mathcal G_\omega $. The sequence $b_ j$ gives the converse and the sequential assertion. $\square$

Cite this page

Comments

Corrections, questions, and mathematical discussion are welcome. Sign in with GitHub to comment. Comments are public.