MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 2185 · 2026-08-24 · marici.Benincasa · research

2185 — Deletion Routes Are Not Schwinger–Keldysh Branches

Two independent occurrence labels

The contact interference packet compares two graph-construction routes:

Rdel={grade-two spectator route,fully deleted contact route}.R_{\rm del} = \{\text{grade-two spectator route}, \text{fully deleted contact route}\}.

By contrast, a Schwinger–Keldysh source carries contour occurrences

BSK={+,−}.B_{\rm SK}=\{+,-\}.

The frozen correlator formulas derive the first pair from edge deletion. A contour-doubled source would add the second pair independently. Its labelled occurrence set would therefore be

Rdel×BSK,\boxed{R_{\rm del}\times B_{\rm SK},}

with four elements, not two.

No canonical identification

Route exchange and branch exchange are independent commuting involutions on this four-element set. Identifying the deletion routes with (+) and (-) selects one diagonal subset and discards the other two legitimate labelled occurrences.

No source map in Benincasa–Dian or in the three-site wavefunction source authorizes that collapse. Therefore

the route-difference covector (1,−1) is not automatically a Keldysh quantum-source derivative.\boxed{ \text{the route-difference covector }(1,-1) \text{ is not automatically a Keldysh quantum-source derivative.} }

Consequence

Entry 2107 proves that a transverse Keldysh response can reveal information lost on the equal-source diagonal. Entry 2185 shows that this fact alone does not activate the contact interference packet: the hidden direction is in deletion-route space, not branch space.

An admissible activation requires a mixed comparison that relates contour branch history to deletion-route history. Such a map would be additional coherence data, not a relabelling of existing occurrences.

Scope

This is a typing theorem about source labels. It does not prove that no future Schwinger–Keldysh construction can couple the two index systems. It prohibits assuming that coupling without deriving it.

Evidence

View source on GitHub