Diagonal operators preserving a coordinate submodule
Let \(R\) be a commutative ring, \(X\) an index set, and \(E_x\) an \(R\)-module for each \(x\). A scalar weight function \(f:X\to R\) defines the diagonal operator \[ D_f:\prod_{x\in X}E_x\to\prod_{x\in X}E_x,\qquad (D_fm)_x=f(x)m_x. \] It is \(R\)-linear by the module laws and commutativity of \(R\).
For a submodule \(L\subseteq\prod_xE_x\), its diagonal stabilizer is \[ \operatorname{Stab}(L) =\{f\in R^X:D_f(L)\subseteq L\}. \] This is the source's stabilizer subalgebra. Membership means that applying the coordinate weights to any vector already in \(L\) produces another vector in \(L\).
Algebra operations on stable weights
Pointwise addition and multiplication satisfy \[ D_{f+g}=D_f+D_g,\qquad D_{fg}=D_fD_g. \] Thus stable weights are closed under addition and multiplication. The zero and constant-one functions give the zero and identity operators. A constant function with value \(r\) acts by scalar multiplication by \(r\), preserving every submodule.
Negation and subtraction follow from the same formulas. Therefore every polynomial expression in stable weights and constant functions remains in \(\operatorname{Stab}(L)\). This is the algebra used to construct the class projectors.
Scaled block projections
A class projector in this construction is a stable weight \(g\) that equals one common scalar \(d\) on a selected class \(C\) and is zero outside \(C\). Its diagonal operator sends a vector to its \(C\)-coordinates multiplied by \(d\).
It satisfies \(D_g^2=dD_g\). In a localization where \(d\) is invertible, dividing by \(d\) gives an idempotent projection. The class-projector theorem constructs \(d\) outside the chosen prime, which is the condition needed for a prime-local denominator.
The congruence-pair example
Take \(R=\mathbb R[\alpha]\) and \[ L=\{(u,v)\in R^2:u-v\in\alpha R\}. \] A weight pair \((a,b)\) stabilizes \(L\) exactly when \(a-b\in\alpha R\).
Necessity follows by applying it to \((1,1)\in L\). For sufficiency, \[ au-bv=a(u-v)+(a-b)v\in\alpha R. \] Thus the stabilizer condition has an explicit congruence interpretation.
Coefficient-action weights
For a word coordinate image over its symmetric coefficient algebra \(A\), the weight function associated to a coefficient \(a\) is \[ x\longmapsto\chi_x(a). \] It is stable because right multiplication by \(a\) on a word vector satisfies \[ \pi_x(m\cdot a)=\chi_x(a)\pi_x(m). \] Consequently the diagonal operator on an image tuple is the projection of another actual word vector. The finite-coordinate construction uses these stable weights as its scalar tests.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.FiniteCharacterLocalization.diagonal: IntegralPairings.lean:1664.
OAI.KLInvariance.FiniteCharacterLocalization.stabilizer: IntegralPairings.lean:1669.
OAI.KLInvariance.FiniteCharacterLocalization.diagonal_mem: IntegralPairings.lean:1694.