Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section3.Provider.Base.PlateauLandmarks

PlateauLandmarks #

Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.

The landmark specializations of the annealed plateau bound #

ABK26, records two consequences of the unconditional plateau display e.basecase.shom.m. For every m below the base landmark m**,

nu Id <= shom_m <= (1 + gamma^2) nu Id (e.basecase.homogenization),

and for every m below the plateau landmark m*,

nu Id <= shom_m <= 2 nu Id (e.basecase.diffusivity).

The proof of p.base.case says these "follow immediately from e.basecase.shom.m. by using the definitions of m** and m*", and that is exactly the route taken here. This file delivers only the diffusivity conjuncts of the two displays; their homogenization-error conjuncts are the business of other modules.

Route #

Write

A_m := c+ (2 nu^2)^{-1} gamma^{-1} (1 + 4 gamma) 3^{2 gamma m}

for the plateau amplitude of annealedPlateau, so that sigmaBar_m <= nu (1 + A_m) unconditionally. Both landmark characterizations have the quotient form 3^{2 gamma m} <= nu^2 w / (2 (log 3) c+), with weight w = gamma^3 at m** and w = gamma at m*. Substituting the quotient bound into A_m cancels c+ and nu^2 and leaves

A_m <= (1 + 4 gamma) w gamma^{-1} / (4 log 3) <= w gamma^{-1},

because gamma <= 1/4 gives 1 + 4 gamma <= 2 while log 3 > 1 gives 4 log 3 > 4. That single estimate is plateau_amplitude_le; the two branches are its w = gamma^3 instance (w gamma^{-1} = gamma^2) and its w = gamma instance (w gamma^{-1} = 1).

Main results #

References #

The numeric input #

The common landmark collapse #

The two landmark displays #

ABK26, e.basecase.homogenization, diffusivity conjunct.

For every m <= m**, the annealed running diffusivity of the coefficient cutoff satisfies nu <= sigmaBar_m <= (1 + gamma^2) nu.

The lower leg is the unconditional plateau's own lower leg. For the upper leg, e.mstarstar gives 3 ^ (2 gamma m) <= nu^2 gamma^3 / (2 (log 3) c+), so the plateau amplitude is at most gamma^3 gamma⁻¹ = gamma^2.

ABK26, e.basecase.diffusivity, diffusivity conjunct.

For every m <= m*, the annealed running diffusivity of the coefficient cutoff satisfies nu <= sigmaBar_m <= 2 nu.

The m* window is the wider of the two landmark windows: e.mstar gives only 3 ^ (2 gamma m) <= nu^2 gamma / (2 (log 3) c+), so the plateau amplitude is bounded by gamma gamma⁻¹ = 1, which is exactly what the printed constant 2 accommodates.