Finite rank-one separation of the damped arctangent kernel #
This leaf contains only the finite algebra behind the later mean estimate. It
splits the nonzero-frequency truncated Perron integrand into two rank-one
rectangular terms and records the coefficient-energy bounds needed to feed
rankOneRectangularWeightedPrimitiveMean_le. It makes no final mean claim.
This module is intentionally not imported by a canonical facade.
The logarithm in the product kernel separates into its two rectangular coordinates. Positivity hypotheses are explicit because this is the form used on positive integer rectangles.
The original rectangular double sum with the truncated Perron integrand.
Equations
- AnalyticNumberTheory.LargeSieve.rectangularKernelCharacterSum a b y t Ma Mb Na Nb q χ = ∑ m ∈ Finset.Icc (Ma + 1) (Ma + ↑Na), ∑ n ∈ Finset.Icc (Mb + 1) (Mb + ↑Nb), a m * b n * ↑χ ↑(m * n) * ↑(AnalyticNumberTheory.LargeSieve.truncatedPerronIntegrand (Real.log (y / ↑(m * n))) t)
Instances For
Four one-dimensional rectangular Dirichlet sums reproduce the original
kernel-weighted double sum exactly. Its quantifier order and interval format
match rankOneRectangularWeightedPrimitiveMean_le directly.
Half-step geometry bound for the left logarithmic coordinate. The lower
half-step hypothesis is necessary: an upper bound on y alone cannot control
|log y|.
Half-step specialization of the left sine-energy estimate on a rectangle.
This is in exactly the Icc (start+1) (start+length) format consumed by the
rank-one primitive mean theorem.
Positive-support specialization of the right sine-energy estimate on the same rectangle format.