Skip to content

[P2 exact/probability] Matching root as a homological balance point: prove p_N→p_c before fitting its shift #276

Description

@LightChainr

Trigger

The digital-Alexander theorem gives the finite matching function a new exact meaning:

M_N(p)=E_p[q]
      =E_p[r]-1
      =P_2(p)-P_0(p),

where

r = rank im[H1(occupied black complex) -> H1(T^2)],
P_j = Prob(r=j).

Therefore every finite matching-polynomial/root estimate is solving

boxed:
P_2(p_N)=P_0(p_N)

or equivalently

E_{p_N}[r]=1.

This is a much stronger interpretation than “the finite matching polynomial happens to have a root near p_c”. It is the point at which trivial ambient topology and full ambient topology have equal probability.

Call p_N the homological balance point.

Main theorem target

Prove, for the declared square-site torus sequence,

p_N -> p_c(square-site)

without using CFT, a guessed correction exponent, or numerical knowledge of p_c.

The proof should use only monotonicity / sharpness / torus crossing-homology arguments plus the exact 4/8 matching duality.

This would give the finite matching root a rigorous probability-theory role independently of the operator-identification program.

Phase A — finite-volume monotonicity and uniqueness

The ambient rank is monotone under adding occupied sites:

omega subset omega'  =>  r(omega) <= r(omega').

Hence

M_N(p)=E_p[r]-1

is monotone increasing in p.

Strengthen this to strict monotonicity on (0,1) for every nondegenerate torus by either:

  1. an explicit rank-pivotal configuration at every site class / a Russo formula; or
  2. the already-established matching Russo identity.

With

M_N(0)=-1,
M_N(1)=+1,

this gives a unique balance root p_N.

Record the exact relation

M_N'(p)=d/dp E_p[r]

as total expected rank-birth influence. This is the natural topological version of the pivotal mass.

Phase B — prove localization around p_c

For every fixed epsilon>0, target

P_0(p_c-epsilon) -> 1,
P_2(p_c+epsilon) -> 1

on growing square tori.

Then

M_N(p_c-epsilon) -> -1,
M_N(p_c+epsilon) -> +1,

and monotonicity immediately implies

p_c-epsilon < p_N < p_c+epsilon

for sufficiently large N.

This is the entire convergence theorem.

The probabilistic input should be stated cleanly in terms of standard subcritical exponential decay and supercritical torus wrapping / uniqueness, or an RSW + sharpness formulation. Do not invoke a finite-size scaling ansatz to prove convergence.

Relation to homological-percolation literature

Duncan--Kahle--Schweinhart, arXiv:2011.11903, prove sharp transitions and threshold convergence for ambient-homology percolation on growing tori, including self-dual middle-dimensional cases.

Our square-site problem is not literally their plaquette model, but after PR #271 the observable is exactly of the same ambient-image type:

rank im[H1(P)->H1(T^2)].

Use their proof architecture as a template and identify precisely which steps transfer to site percolation with the 4/8 digital complex.

The point is not to reprove the full percolation phase transition. It is to deduce convergence of this specific median-like homology balance statistic from known sharpness facts.

Phase C — define rank-pivotal events geometrically

The derivative

d/dp E[r]

has a more canonical local meaning than a generic wrapping pivotal.

For a site v, define

Delta_v r = r(omega with v=1)-r(omega with v=0).

Classify the possible rank births:

0 -> 1,
1 -> 2,
possibly 0 -> 2 on finite degenerate geometries if allowed.

and retain the new ambient homology line/subspace created at the insertion.

This gives a homology-marked pivotal measure whose unmarked total is exactly M_N'(p).

It should become the preferred local bridge to the global matching observable, because it differentiates the exact quantity r itself rather than a proxy wrapping event.

A later spin-4 projection can ask which continuum sector carries rank birth.

Phase D — only after convergence, study the shift

Once p_N -> p_c is established, write

p_N-p_c = - M_N(p_c)/M_N'(xi_N)

for some intermediate xi_N, or use the local implicit-function expansion when justified.

This factorization is conceptually useful:

numerator   = finite-size violation of the critical homology balance identity,
denominator = rank-birth pivotal susceptibility.

It separates two pieces that the current root fits mix together.

Then compare fixed mechanisms for the numerator:

singlet spin-4 lattice defect / Q4-Jordan,
ordinary analytic correction,
other topology-sector correction,

while the denominator is controlled by the thermal/pivotal exponent.

Do not fit a free root-shift exponent before deriving these two ingredients.

Phase E — continuum balance at criticality

Arguin, arXiv:hep-th/0111193, gives the critical FK torus relation

Z_Q(cross)=Q Z_Q(trivial).

At Q=1 this becomes

pi_1(rank2)=pi_1(rank0)

for the continuum critical theory.

Thus the finite balance root has a natural continuum target: it is the value of p where the lattice rank contrast most directly enforces the same topological equality that holds exactly in the Q=1 continuum limit.

Issue #275 asks which irrelevant singlet field controls the residual at fixed critical p. This issue asks the complementary probability-theory question: where must the zero of that residual converge?

High-risk stronger conjecture

The matching-polynomial root may be best regarded as a topological phenomenological renormalization condition:

E[r]=1.

It is analogous to fixing a dimensionless crossing probability at a universal value, except the target value follows from Alexander duality and the exact torus homology relation rather than being chosen empirically.

If so, root sequences on different torus moduli should share the same thermal scaling field but have shape-dependent irrelevant shifts. This could make the matching root a particularly clean coordinate for finite-size RG.

Falsification / boundary

  • Convergence does not imply a closed form for p_c.
  • The proof for honest square-cell tori should not silently cover short-period degenerate quotients outside PR Prove the digital Alexander rank bridge #271's theorem scope.
  • Homological balance is not automatically the optimal numerical estimator of p_c; this is a structural statement first.
  • A rate claim requires quantitative near-critical input beyond the qualitative convergence theorem.

Literature anchors

  • Duncan--Kahle--Schweinhart, arXiv:2011.11903 — homological percolation on tori, sharp transition and threshold convergence.
  • Arguin, arXiv:hep-th/0111193 — critical FK homology-sector probabilities on the torus.
  • Duncan--Schweinhart, arXiv:2207.08339 / 2406.08043 — ambient-homology duality and topological random-cluster structure.
  • Garban--Pete--Schramm, arXiv:1008.1378 and near-critical follow-ups — pivotal/near-critical scaling as the later quantitative input for the shift.

Related: #3, #47, #54, #114, #118, #123, #269, #275, PR #270, PR #271.

Metadata

Metadata

Assignees

No one assigned

    Labels

    priority:P2Deferred research or on-demand support; no default new compute allocation.

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions