ActiveFamily #
Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.
The active family ๐_R, its multiplicity bound, and the superposed competitor #
GlobalSuperposition.lean builds the multi-component competitor โฬ_p = โ_p + ฮฃ_๐ (โฬ_p^๐ โ โ_p) and proves the three sum_of_a_decomp clauses for its
field, with no finiteness of the component family. Its Scope section
names the gap that this module fills:
- the active family is finite, with a cardinality bounded by a dimensional
constant
M(d), uniformly in the Whitney cube.
The active family and the constant #
For a Whitney cube R writes
๐_R := {๐ : R โ ๐ฉ(๐)}
which is activeComponents m hn โ R below (the components of โ whose
touching collar contains R). Here it is
M(d) = 3 ยท 11^d ,
3 for the three layers nโ1, n, n+1 a cube touching a layer-n cube can
occupy (BoundaryDistance.whitneyLayer_index_close_of_cubeTouch, the adjacency
clause alone) and 11^d for the per-layer packing at the cubewise touch
relation of e.neighborhood.def
(Multiscale.CollarMass.card_le_eleven_pow_of_forall_cubeTouch), the two
generations of scale slack being's ยฑ2 depth window
(Multiscale.CollarMass.layerDepth_sub_le_two).
Inactive components and the cell slope #
The active family is what the superposed cell constants actually see. The
Whitney cubes touching one Whitney cube form a finite family
(finite_setOf_mem_whitneyPartition_cubeTouch), each of them activating at most
M(d) components; and a component that is inactive at the Whitney cube Q
carrying a cell T โ one whose touching collar misses Q โ leaves the
competitor's cell constant on T equal to the ambient p, so it contributes
nothing to the sum of superposedCompetitorCellSlope_eq_sum. That is
globalCompetitorCellSlope_eq_of_notMem_activeComponents, proved from the
predecessor's globalCompetitorSlope_eq_cellSlope. Together with the
multiplicity bound it caps at M(d) the number of summands that can be nonzero
on any single cell.
Scope: what this module does NOT do #
- No gradient estimate.'s summed bound
|โโฬ_p โ p| โค C(d)|p| ฮฃ_{๐ โ ๐_R} diam(๐)/ฯ_๐is read and not consumed: no diameter, no interpolation scaleฯ_๐and no constantC(d)occurs below. This module supplies the two inputs that estimate needs and nothing else โ the finite index set๐_Rover which the sum runs, and|๐_R| โค M(d)for the squared/Cauchy--Schwarz form. - No weak gradient, no
Hยนโ. The cell constants read below are those of the defined piecewise-constant fieldsuperposedCompetitorCellSlope, and nothing identifies them with a distributional gradient; no a.e. statement, noMemVectorL2, noH10Functionand no zero-trace statement occurs.e.hat.linear.1and thehF/hGbinders ofMultiscale.sum_of_a_decompremain untouched. - No continuity and no compact support. The counting below is confined to
the open root cube, and that is sharp for this route: the Whitney cubes
accumulate at
โโก_m, so the finiteness argument has no analogue there. As the predecessor recorded, compact support must come from the cutoff/truncation, not from the collar. - No disjointness of collars is asserted or used, exactly as in
GlobalSuperposition.lean; the overlap is counted, not removed. Merging overlapping components is not attempted โ rules it unsafe. - No three-layer window and no face-to-face property of
SW(โก_m)is used.
References #
- ABK26, (
๐ฒ(โก_m),e.W.n.def, the boundary distance), (โ,e.neighborhood.def), (the components), (the adjacency clause), (e.SW.def), (l.piecewise.affine.approx,e.hat.linear.properties), (the false disjointness), (Step 1, "well-defined and continuous"), (the gradient estimate โ not consumed), (โ_e). - Two ideas were taken: bound the touching cubes layerwise over
Icc (n-1) (n+1)and sum the per-layer packing; and choose one bad cube per active component, recovering the component from it for injectivity. Everything is restated at this repository's carriers (CubeTouch,whitneyNeighborhood,badComponent,KuhnCell,globalCompetitor) and reproved from this repository's proved inputs.
Main results #
card_le_three_mul_eleven_pow_of_forall_cubeTouch,finite_setOf_mem_whitneyPartition_cubeTouchโ the Whitney cubes touching one Whitney cube, counted and bounded.activeComponents,activeComponents_finite,card_le_multiplicityBound_of_forall_mem_activeComponentsโ the deliverable|๐_R| โค M(d) = 3ยท11^d, the bound beingmultiplicityBound d = 3ยท11^d.globalCompetitorCellSlope_eq_of_notMem_activeComponentsโ a component that is inactive at the Whitney cube of a cell leaves the competitor's slope on that cell equal top.
The Whitney cubes touching one Whitney cube #
The three-layer window of the touch relation. A Whitney cube touching a
layer-n cube occupies one of the layers n-1, n, n+1 โ the adjacency
clause, which consequence 2 needs no percolation input and no largeness of
kโ, only h_n โฅ 1.
Only finitely many Whitney cubes touch one Whitney cube. They are
confined to three layers, and each layer is a Finset.
The dimension-only count of the touching cubes: at most 3ยท11^d Whitney
cubes touch a fixed Whitney cube. 3 is the layer window and 11^d is the
proved per-layer packing at the cubewise touch relation
(Multiscale.card_le_eleven_pow_of_forall_cubeTouch), applicable because's
ยฑ2 depth window (Multiscale.layerDepth_sub_le_two) puts the three layers
within two generations of one another.
The active family ๐_R #
The multiplicity constant M(d), M(d) = 3ยท11^d: three layers times the
cubewise per-layer packing.
Equations
Instances For
The active family ๐_R = {๐ : R โ ๐ฉ(๐)}: the components of the bad family
whose touching collar contains the Whitney cube R. These are exactly the
components whose correction can be nonzero on R.
Equations
- One or more equations did not get rendered due to their size.
Instances For
An active component is recovered from any of its bad cubes touching R.
This is the maximality fact isolates, in the form that makes the counting
injective.
The chosen bad cube of an active component, touching R.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Finiteness of the active family โ the input GlobalSuperposition.lean
names as the operative obstruction to continuity and per-cell affineness.
The multiplicity bound, Finset form. Every finite family of components
active at R has at most M(d) members.
Inactive components contribute nothing #
A component inactive at the cell's Whitney cube has cell constant p โ
clause 3 of e.hat.linear.properties read at the active family.