Proposition 6.1: residual correction spaces #
positive-deformations.tex (prop:residual-correction-spaces). In characteristic two,
under the brick hypothesis, call a periodic component residual if ⊕_m coker D_m(C) has a
nonzero class supported on it. Then each residual component is ℤΔ/⟨τ^{s_i}ρ⟩ with Δ of
type D or E, there is a path p_i : τ v_i → v_i of degree d_i = 2s_i + 2, and
⊕_{m≥2} coker D_m(C) = ⊕_{i∈I} k[p_i].
The spanning by at most one row per periodic component is residual_cokerSpan. Here:
- the mesh derivative commutes with the projection to a union of components of a quiver
(
D_projA), so the image ofDis stable under it (projM_mem_imageD); - after the localization of Proposition 5.1 to the periodic vertices (where periodic
components are unions of connected components), a family supported on one periodic
component lies in the image of
Das soon as it does modulo families supported elsewhere (mem_imageD_of_supported); - hence rows in distinct periodic components with nonzero classes are linearly independent
(
rows_independent).
Projection to a union of components #
The arrow family kept on the arrows with target in P.
Instances For
The mesh family kept on the vertices in P.
Instances For
The mesh derivative commutes with the projection to a union of components.
Families supported on one periodic component #
Membership in a periodic component is preserved along arrows between periodic vertices.
A family supported on one periodic component lies in the image of D_m as soon as
it does modulo a family supported off that component.
Rows in distinct periodic components with nonzero classes are linearly independent.
Rows and the cokernel #
The families of degree m + 1 are spanned by the rows.
The span of the rows of S of degree m + 1, in the families.
Instances For
A family of degree m + 1 is a combination of the rows of a spanning set, modulo
the image of D_m.
A periodic component is residual if ⊕_{m≥2} coker D_m(C) has a nonzero class
supported on it.
Instances For
Proposition 6.1 #
Proposition 6.1 (prop:residual-correction-spaces). In characteristic two, for a
finite-dimensional scalar mesh category on a finite translation quiver with at most one arrow
between ordered vertices, satisfying the brick hypothesis, there is a set S of rows p_i
at periodic vertices, at most one in each periodic component, such that:
- the classes
[p_i]span⊕_{m≥2} coker D_m(C), are nonzero, and are linearly independent (so⊕_{m≥2} coker D_m(C) = ⊕_i k[p_i]); - a periodic component is residual exactly when it carries a row of
S(soSis indexed by the residual componentsI); - the component of each
p_iisℤΔ/⟨τ^{s_i}ρ⟩withΔof typeDorE,ρfixing the centre, andp_ihas degreed_i = 2s_i + 2(StarTrace.DEModel; in particulard_iis the only degree of a nonzero class on the component); - the rows are killed by two over every coefficient ring (used for Lemma B.2).