NAG-003194 · Historical whole-proof gap
A verifiable model class and coordinate calibration
Records without a chapter binding
Remaining historical proof gap
The sealed ledger records a gap in the proof of the historical whole statement.
Historical statement · exact TeX
\begin{theorem}[A verifiable model class and coordinate calibration]\label{f9tf:model}
Suppose a model line $\mathcal A$ is supplied with a finite family of global sections generating it as an invertible sheaf. Its actual bounded-frame norm has, in every ordinary analytic local frame, a potential locally equal to a maximum of finitely many logarithms of analytic units. In particular it is continuous and belongs to $\PP$.
Consequently, for $i=1,2$, suppose a root model presentation of $L_i$ of exponent $e_i\geq1$ is supplied with an actual model isomorphism
\begin{equation}\label{eq:modelquotient}
\mathcal M_i\simeq\mathcal A_i\otimes\mathcal B_i^{-1},
\end{equation}
where both $\mathcal A_i$ and $\mathcal B_i$ have supplied finite generating sections. Then its metric satisfies the hypotheses of Theorems~\ref{f9tf:first}--\ref{f9tf:two}, on all native tests. The statement does not require a reduced special fibre or assert positivity of the quotient metric. A local variant requires the resulting norm identities on ordinary analytic neighborhoods; generation on a list of formal domains alone is not claimed sufficient.
For a normalization example in every $n\geq2$, restrict the standard model metrics from $(\mathbf P_R^1)^n$ to the generic torus, and write $t_k=-\log|z_k|$ and
$w_i=\max(0,t_i)$. These are the potentials of the model frames $X_{1,i}$. If $i\ne j$, and $\psi\in C_c^\infty(\R^n)$, put
\[
\alpha= t^*\!\left(\psi\bigwedge_{k\ne i,j}(d't_k\wedge d''t_k)\right).
\]
Then
\begin{equation}\label{eq:coordinate}
R(w_i,w_j)(\alpha)=
\int_{\R^{n-2}}\psi(t_1,\ldots,t_n)\big|_{t_i=t_j=0}
\prod_{k\ne i,j}dt_k.
\end{equation}
For $i=j$, $R(w_i,w_i)=0$ on all native tests. Root metrics divide \eqref{eq:coordinate} by $e_i e_j$.
The dual model metric with potential $-w_i$ is not positive: for a nonnegative coordinate test
$\beta=t^*(\psi\bigwedge_{k\ne i}d't_k\wedge d''t_k)$ with
$\int\psi|_{t_i=0}\prod_{k\ne i}dt_k=1$, one has $T_{-w_i}(\beta)=-1$.
\end{theorem}