The square witness at q = Θ(1/t) #
Theorem thm:squarelinear of papers/inflation-nontermination/paper/sections/15-cycles.tex.
The corrected Fourier density with m = 4 families and c = 5,
W = Re ∏_g f_g + 5 ∑_g (1 − Re f_g), is pushed forward along the parity read of
CycleWitness.lean. Its moments agree with those of the auxiliary density auxH on every
set of signs that is empty or meets at least two families, and every character that an
injectable, ancestral-product or diagonal prescription looks at has a latent boundary of
that kind. So each prescribed pushforward of the new witness equals the corresponding
pushforward of parWit, and the obligations are inherited from parWit_diag,
parWit_injectable and parWit_ai, which hold for every real q.
Four-factor telescoping #
The m = 4, c = 5 density Re(u₀u₁u₂u₃) + 5 ∑_g (1 − Re u_g) is at least 1/3
whenever the common modulus R satisfies 1 ≤ R and R² ≤ 16/15.
The corrected density with a general family type #
f_g(s) = ∏_i (1 + i√q s_{g,i}), the factor of the family g.
Equations
- TriangleInflation.Graph.SqWitnessAux.fam t q s g = ∏ i : Fin t, TriangleInflation.Graph.triAtom q (s (g, i))
Instances For
W(s) = Re ∏_g f_g + 5 ∑_g (1 − Re f_g).
Equations
- TriangleInflation.Graph.SqWitnessAux.sqW t q s = (∏ g : κ, TriangleInflation.Graph.SqWitnessAux.fam t q s g).re + 5 * ∑ g : κ, (1 - (TriangleInflation.Graph.SqWitnessAux.fam t q s g).re)
Instances For
On a spread set the family corrections cancel: the g-term of the moment vanishes.
ρ = 2^{−N} W, the witness density on the auxiliary signs.
Equations
- TriangleInflation.Graph.SqWitnessAux.sqDensity t q s = 1 / 2 ^ Fintype.card (κ × Fin t) * TriangleInflation.Graph.SqWitnessAux.sqW t q s
Instances For
The witness on a cycle #
The square-type witness: the corrected density pushed forward along the parity read.
Equations
Instances For
A pushforward whose characters all pull back to characters with spread latent boundary is
the same for the new witness and for parWit.
B2: the linear-in-t witnesses #
AUDIT-NOTES B2(i). With the corrected Fourier density (m families of t signs,
f_g = ∏_i (1 + i√q s_{g,i}), W = Re ∏_g f_g + c ∑_g (1 − Re f_g), m = 4, c = 5,
positive for tq ≤ 1/16), the square parity target with q = 1/(16t) passes every AI
prescription at order t. This is the source of the Ω(1/t) square lower bound, an order of
magnitude better in t than the q = 1/(4m²t²) of cycle_witness.