NAG-002737 · Historical whole-proof gap
Why a model profile cannot be treated piecewise
Records without a chapter binding
Remaining historical proof gap
Selected results establish the residual degree and Gauss points, but not equality with the actual mixed model measure at the original field scope, including transported coordinates.
Historical statement · exact TeX
\begin{proposition}[Why a model profile cannot be treated piecewise]
\label{f9tm:modelobstruction}
A continuous function on $\mathbf R$ which is affine on a finite
polyhedral subdivision and is $C^1$ is affine. Consequently, for
$m>0$ it cannot also have the two different tail slopes $-m$ and $0$
in \eqref{eq:tails}. Thus the positive-degree finite piecewise-affine
toric model profiles do not lie in the smooth fixed-torus-profile
class above.
For the product model $(\mathbf P^1_{K^\circ})^d$ with the standard
model metrics on $L_i$, the profiles are
$\phi_i^{\rm mod}(t)=m_i\max(0,-t)$. The Chambert--Loir model-measure
formula in this explicit reduced model gives
\begin{equation}\label{eq:modelmeasure}
\mu_{\rm CL}^{\rm mod}=D\,\delta_{\zeta_0},\qquad
\int u_F\,d\mu_{\rm CL}^{\rm mod}=D F(0)
\quad\bigl(F\in C(\overline{\mathbf R}^{,d})\bigr),
\end{equation}
where $\zeta_0$ is the unit-radius Gauss point of $T^\an$.
In contrast, discarding the corners and using the ordinary Hessians
on the affine pieces gives the zero density. It fails to reproduce
\eqref{eq:modelmeasure}, even for a compactly supported smooth radial
test equal to one near the origin.
\end{proposition}