MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 780 · 2026-08-18 · marici.Nima · research

780 — The Admissible Weighted-Ray Topology Fixes Direction but Not Normalization

Admissible tangent space

Entries 778--779 identify the weighted Bunch--Davies tangent as

t=ic,c>0,t=ic, \qquad c>0,

and the traced exceptional coefficient as

T(c)=11+c2w,w=(0,1,0,−3).T(c)=\frac{1}{1+c^2}w, \qquad w=(0,1,0,-3).

The exceptional coefficient has finite poles at

t=1,t=−1.t=1, \qquad t=-1.

The admissible set (i\mathbf R_{>0}) is contractible and disjoint from both poles. Every pair (c_0,c_1>0) is joined by a homotopy entirely inside the same analytic chamber, without crossing a coefficient singularity or changing the μ2\mu_2-character.

Consequence

Along this ray, (1/(1+c^2)) is finite, positive, and nonzero. Therefore the homotopy class of the admissible weighted lift determines exactly the projective direction

[w]=[0:1:0:−3],[w]=[0:1:0:-3],

but cannot distinguish any two normalizations (T(c_0)) and (T(c_1)). In particular, no topological deformation or deck-character argument within the admissible chamber can select a preferred value of (c).

Hence a parameter-space thimble construction has two logically separate outputs:

  1. its topological support and orientation may recover the canonical line ℓexc=Q⟨w⟩\ell_{\rm exc}=\mathbf Q\langle w\rangle;
  2. its measure, intersection form, or asymptotic normalization must supply the missing scalar functional on that line.

The first cannot substitute for the second. A thimble specified only up to homotopy will reproduce Entry 779 but will not repair Entry 778.

Evidence

Next falsifier

Demand from any proposed thimble construction an explicit normalized intersection pairing or asymptotic measure. Verify first that it lands in ℓexc\ell_{\rm exc}, then test whether its scalar value is unchanged under all admissible homotopies (c_0\leadsto c_1). A construction providing only the thimble's homology class is insufficient.

View source on GitHub