Pome

18.45.13 A rank-one selection with nonclosed sampling8461E7D0D299

Example 18.45.14.9600E351F18B On $(\mathbb P^1)^2$ take $L=\mathcal O(4,4)$, $w=0$, and $Q=\{ (2+t,2-t):-1/2\leq t\leq 1/2\} $. Then $\Lambda =\mathbb Z(1,-1)$, $\mathrm dq_\Lambda =\mathrm dt$, and

\[ V_ m=\operatorname {span}\{ z_1^{2m+j}z_2^{2m-j}e^ m: j\in \mathbb Z,\ |j|\leq m/2\} , \qquad N_ m=2\lfloor m/2\rfloor +1. \]

The limit $\mu _ Q$ has mass one: it is the average, over $-1/2\leq t\leq 1/2$, of the Haar probability measures on the real tori $\{ |z_1|^2=(2+t)/(2-t),\ |z_2|^2=(2-t)/(2+t)\} $; indeed, these equations describe the inverse gradient map. The moment shell is a compact three-dimensional part of the hypersurface $|z_1z_2|=1$. Writing $c(p)=p\log (p/4)+(4-p)\log ((4-p)/4)$, we have $F^*(2+t,2-t)=2c(2+t)$, which gives the envelope explicitly.

The quotient of two adjacent selected monomials is $z_1/z_2$. In homogeneous coordinates $z_ j=v_ j/u_ j$, the corresponding map $[v_1u_2:u_1v_2]$ to $\mathbb P^1$ is undefined at $(0,0)$ and $(\infty ,\infty )$. Thus the limit arises from radial concentration rather than from a globally defined submersion as in section 18.36.

Let $\omega _0$ be the sum of the two Fubini–Study forms, and set $f=c|z_1|^2/(1+|z_1|^2)$ with a constant $c\ne 0$. Then $\omega =\mathrm e^ f\omega _0$ is a torus-invariant Hermitian form, and $\mathrm d\omega =\mathrm e^ f\mathrm df\wedge \omega _0\ne 0$ wherever $\mathrm df\ne 0$. Sampling with $\nu =\omega ^2/\int _ X\omega ^2$ gives the same limit of mass one, while the curvature remains $\theta =4\omega _0$.

The assumption $b\in \mathbb Z^ n$ excludes a parity obstruction: if $Q=\{ b\} $ with $b$ half-integral, then $mQ$ contains no lattice point for odd $m$. Under our assumptions, theorem 18.45.7 shows that $V_ m\neq 0$ for all sufficiently large $m$. Polytopes touching the boundary of $P$, weights or sampling measures that are not torus-invariant, and general multiplier-ideal families require additional arguments. In particular, the growth condition defining $E_ Q$ does not by itself identify the ideal envelope of section 18.32.

Cite this page

Comments

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