LatticeTwo #
Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.
Distance-two lattice paths: geometry and separated selection #
ABK26's abstract percolation machinery (l.percolation.bound.general) steps
along paths whose consecutive sites are at lattice distance at most one. The
bad-cluster lemma l.percolation.bad.clusters however joins base cubes □, □'
with dist(□,□') ≤ 1, i.e. lattice sup distance at most two in units of
the cube side 3^{m-h}. A 2-adjacent chain is not a 1-adjacent lattice
path, so the abstract layer as proved does not literally cover 2-clusters.
This module is the first half of that re-run: the purely combinatorial geometry of distance-two paths. It is a parallel layer under new names; no proved declaration is modified.
Where the step size enters #
Exactly three places, and each costs only a dimension-only factor:
- Truncation. A distance-two path truncated at its first exit from
z + □_nis confined to2 · latDist z (x i) ≤ 3 ^ n + 3(the distance-one layer has3 ^ n + 1). - Shell selection. The distance to the starting site now jumps by up to
two per step, so a first passage of the level
m * sproves in[m*s, m*s + 1]rather than exactly onm * s; consecutive selected sites are therefores - 1-separated instead ofs-separated. Taking the steps = 3 ^ (k+1) + 5restores the separation3 ^ (k+1) + 4needed below. - Range of independence. The wider localisation boxes need the wider site
separation
3 ^ (k+1) + 4(against3 ^ (k+1) + 2) to keep the boxes at sup distance more than3 ^ k;three_pow_lt_latDist_of_mem_locBox₂supplies it.
The number 3 ^ (h - 3) of selected sites — the quantity that drives the whole
induction — is unchanged, because the arithmetic
2 · 3 ^ (h-3) · (3 ^ (k+1) + 5) ≤ 3 ^ (k+h) still holds. This is why the
gate constant of the distance-two layer is the same 16 d exp(40 (1-b)^{-1}) as
the distance-one one.
Main definitions #
IsLatticePath₂: the distance-two path predicate.locBox₂: the localisation box of a truncated distance-two path.
Main results #
exists_truncated_path₂: truncation at the first exit from a cube.card_locBox₂_le:#(locBox₂ n z) ≤ 3 ^ (d * (n + 2)).three_pow_lt_latDist_of_mem_locBox₂: the independence range.exists_separated_subpaths₂: the separated-subpath extraction, with the same count3 ^ (h - 3)as the distance-one layer.
References #
- ABK26,
l.percolation.bad.clusters, (the 2-connected component) andl.percolation.bound.general, (the abstract layer).
Distance-two paths #
The distance-two path predicate: consecutive sites are at sup distance at most two.
Equations
- Algsuperdiff.Section3.Provider.Percolation.IsLatticePath₂ x N = ∀ i < N, Algsuperdiff.Section3.Provider.Percolation.latDist (x i) (x (i + 1)) ≤ 2
Instances For
A distance-two path may be shortened.
A distance-two path may be restarted at any of its sites.
Along a distance-two path the sup distance to the starting site grows by at most two per step.
Truncation at the first exit from a cube #
Truncation of a distance-two path at its first exit from z + □_n: from a
path started at z and ending outside z + □_n we extract an initial segment
which still ends outside z + □_n and all of whose sites satisfy
2 * latDist z (x i) ≤ 3 ^ n + 3.
The distance-one layer confines the truncated path to 3 ^ n + 1; the extra 2
is the single place where the step size enters the localisation.
The localisation box of a distance-two path #
Half-width, rounded up, of the distance-two localisation box at scale n.
Equations
- Algsuperdiff.Section3.Provider.Percolation.locRadius₂ n = (3 ^ n + 3) / 2
Instances For
The finite set of lattice sites y with 2 * latDist z y ≤ 3 ^ n + 3: the
region in which a distance-two path started at z and truncated at its first
exit from z + □_n is confined.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Distance-two localisation boxes of two sites at sup distance at least 3 ^ (k + 1) + 4 are at sup distance strictly greater than 3 ^ k. This is the
separation required by the finite-range independence hypothesis of ABK26; the
distance-one layer needs only 3 ^ (k + 1) + 2.
Shell selection along a distance-two path #
Shell selection for distance-two paths. Along a distance-two path whose
final site is at sup distance at least M * s from the starting site there are
indices i 0 ≤ ⋯ ≤ i M (bounded by the length of the path) with i 0 = 0 such
that x (i m) lies in the shell [m * s, m * s + 1] around x 0.
Distinct shell-selected sites of a distance-two path are at sup distance at
least s - 1, stated in the addition-free form s ≤ latDist + 1.
The separated-subpath extraction for distance-two paths.
Let x be a distance-two path started at z which ends outside z + □_{k+h},
with h ≥ 3, and suppose a property P holds at every site of the path. Then
there are 3 ^ (h - 3) sites of the path — the same count as in the
distance-one layer — all lying in locBox₂ (k + h) z, pairwise at sup distance
at least 3 ^ (k + 1) + 4, each of them the starting site of a distance-two
subpath which leaves its own □_k-cube and along which P still holds.