Iterated singular convolutions on the four-dimensional torus #
This file packages the induction implicit in paper (5.4). The hard
analytic input is the uniform threefold-convolution estimate proved in
Continuum/SingularConv.lean. Every additional |z|⁻² factor costs only
its finite L¹ mass, so all convolution chains of length at least three
are uniformly bounded, with an exponential constant in the chain length.
This is the reusable chain form of blueprint node I-singconv. It is still only the translation-invariant analytic core of paper (5.3)--(5.4); the cell supports and rescaling factors belong to node R-51.
The n-fold convolution chain of the model kernel |z|⁻².
The values at lengths 0 and 1 are totalized as 0 and invSqKer.
For n ≥ 2, the recursion adds one kernel on the left.
Equations
- Anderson4D.singularChain 0 = fun (x : Anderson4D.T4) => 0
- Anderson4D.singularChain 1 = Anderson4D.invSqKer
- Anderson4D.singularChain n.succ.succ = fun (x : Anderson4D.T4) => ∫ (z : Anderson4D.T4), Anderson4D.invSqKer (x - z) * Anderson4D.singularChain (n + 1) z ∂Anderson4D.paperMeasure
Instances For
Every convolution chain is pointwise nonnegative.
The finite L¹ mass of the model singular kernel.
Equations
Instances For
Translation invariance of the L¹ mass, in the orientation used by
singularChain.
The shifted kernel in the left-convolution orientation is integrable.
Uniform chain bound (paper (5.4), analytic core).
There is one absolute constant C > 0 such that every convolution chain
of length at least three is bounded by Cⁿ, uniformly in its endpoint.
The base case is triple_conv_invSqKer_le; the induction uses only the
finite L¹ mass of one additional kernel.