Vanishing Deformations Remove the Reflection Obstruction
Result
Entry 387 does more than remove coefficient ambiguity. It also decides the reflection bit isolated in Entry 141, conditional only on existence of the underlying connector.
Let denote the space of admissible connectors with the generic, Cartier, lower-Cech, and two endpoint faces frozen. If nonempty, its components form a torsor under the degree-zero relative deformation group. Entries 384--387 compute the relevant fine-graded group as
Consequently
This is the precise dichotomy left by the calculations.
Reflection
Physical reflection preserves the frozen target faces: Entry 140 proves the target purity square has sign , while the orientation twist of Entries 136 and 143 is already included in the road object. Therefore, if , then is another element of the same connector space.
The difference
is a relative endpoint deformation, hence represents an element of . Since that group vanishes,
Thus the endpoint defect class from Entry 141 is forced to be
whenever the connector exists.
Entry 141 proves that the conductor Bockstein
is an isomorphism. Therefore
The previously possible obstruction does not survive the literal fine grading.
What is and is not proved
This does not construct a connector. It proves:
- if no nonequivariant connector exists, the realization fails already at the mixed-variance existence step;
- if one exists, it is unique up to the admissible relative homotopy;
- its reflection defect is necessarily zero; and
- it admits the unique loaded equivariant component predicted by Entry 141.
So the remaining frontier is no longer “existence plus a parity choice.” It is only existence.
Consequence for the next computation
The rank-nine strict carrier family from Entry 136 should not be searched for a reflection-preferred point. All those representatives describe the same unpointed roof before the admissible support and grading constraints. After the faces are frozen and fine grading is imposed, the connector fibre has at most one component.
The next test should therefore be a solvability test for
in the normal-Cech enhanced mixed-variance complex. No parity variable and no free coefficient should be included. A solution proves the unique equivariant connector; inconsistency proves a genuine existence obstruction.
Evidence
research/voevodsky/check_d03_zero_torsor_reflection_gate.py checks the zero-torsor and Bockstein implication. Its mathematical inputs are:
- Entry 387: ;
- Entry 140: target reflection naturality has sign ;
- Entry 141: the parity-to-loaded-obstruction Bockstein is an isomorphism; and
- Entry 136: front/back AW representatives differ by an equivariant homotopy and cannot themselves toggle parity.
Outcome contract
{
"claim": "The fine-graded endpoint-relative deformation group is zero. Hence the admissible connector space is either empty or a singleton up to relative homotopy. If a connector exists, its reflected connector has zero difference class, so p_partial,Q=0 and the Entry-141 loaded Bockstein obstruction vanishes.",
"status": "proved_conditional_reflection_obstruction_zero",
"closed": [
"endpoint coefficient deformation",
"reflection parity conditional on existence",
"loaded Z/2 obstruction conditional on existence"
],
"not_closed": [
"existence of the mixed-variance connector",
"construction of the normal-Cech enhanced AW collar",
"full D3 realization",
"full primal trace"
],
"next_experiment": "Solve the frozen connector equation in the literal normal-Cech mixed-variance mapping complex with no coefficient or parity parameters."
}