Theorem Stable diagonal selection of one residue class
✓ theorem proved

Stable diagonal selection of one residue class

Let \(R\) be a commutative ring, \(X\) a finite index set, and \(L\subseteq\prod_{x\in X}E_x\) a submodule. Let \(\chi_a:X\to R\), indexed by \(a\in T\), be stable weights: \(\chi_a\in\operatorname{Stab}(L)\).

Fix a prime element \(p\) and \(x_0\in X\). For its residue class \(C=\{z:x_0\sim_p z\}\), there exist \(d\in R\) and \(g\in\operatorname{Stab}(L)\) such that \[ p\nmid d,\qquad g(z)=d\ (z\in C),\qquad g(z)=0\ (z\notin C). \]

Here primality means that \(p\) is nonzero and noninvertible, and \(p\mid ab\) implies \(p\mid a\) or \(p\mid b\). These are the properties used in the finite products below.

Proof

Step 1. Choose a test for each outside index

For \(y\notin C\), the residue relation fails. Choose \(a_y\in T\) with \[ p\nmid\chi_{a_y}(x_0)-\chi_{a_y}(y). \] Define the weight \[ h_y(z)=\chi_{a_y}(z)-\chi_{a_y}(y). \] It is stable, being the difference of a stable weight and a constant weight. It satisfies \(h_y(y)=0\) and \(p\nmid h_y(x_0)\). The source sets \(h_y=0\) on the unused indices \(y\in C\), giving its full test family.

Step 2. Prove that the tests survive on the whole class

For \(z\in C\), the residue relation gives \[ p\mid\chi_{a_y}(z)-\chi_{a_y}(x_0) =h_y(z)-h_y(x_0). \] If \(p\mid h_y(z)\), subtracting the displayed divisible difference would imply \(p\mid h_y(x_0)\), contrary to Step 1. Hence \[ p\nmid h_y(z)\qquad(z\in C,\ y\notin C). \]

Step 3. Multiply the separating tests

Set \[ f=\prod_{y\notin C}h_y. \] This finite product belongs to the stabilizer algebra.

If \(z\notin C\), the product includes the factor \(h_z\), which vanishes at \(z\). Thus \(f(z)=0\). If \(z\in C\), every factor \(h_y(z)\) is not divisible by \(p\). Primality, applied successively to the finite product, gives \(p\nmid f(z)\). The empty product is \(1\), also not divisible by a prime.

Step 4. Make the value uniform across the class

Apply the uniform diagonal operator theorem to \(f\) and \(C\). It constructs \[ d=\prod_{z\in C}f(z),\qquad g=d-\prod_{z\in C}(f(z)-f). \] The weight \(g\) is stable, equals \(d\) on \(C\), and vanishes outside \(C\).

Step 3 and primality again imply \(p\nmid d\). This proves the theorem. The source's class_projector performs Steps 2–4 for a general family of separating tests; character_class_projector constructs those tests from the residue relation in Step 1.

Step 5. Calculate a two-coordinate example

Let \(R=\mathbb R[\alpha]\) and let \(L=\{(u,v):u-v\in\alpha R\}\). Use the two coefficient characters \(a\mapsto a\) and \(a\mapsto a(-\alpha)\), and the prime \(p=\alpha-1\).

For the first coordinate, the test \(a=\alpha\) gives stable weights \[ h=(2\alpha,0). \] It is zero at the other coordinate, and \(p\nmid2\alpha\), since evaluation at \(\alpha=1\) gives \(2\). Thus \[ (u,v)\longmapsto(2\alpha\,u,0) \] preserves \(L\) and selects the first coordinate with a scalar not divisible by \(p\). This is the one-element-class instance of the construction.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.FiniteCharacterLocalization.class_projector: IntegralPairings.lean:1730.

OAI.KLInvariance.FiniteCharacterLocalization.character_class_projector: IntegralPairings.lean:1794.

Used by