The large-degree claim #
Sendov.U_le_Ut removed the degree from the bound; what is left is the single inequality
Ut α < 1 on 0 ≤ α ≤ 17, and then R n α < 1 for every n ≥ 101.
Ut is a rational function of α times (α/(3+α)) ^ 48, so multiplying by
D = 865200 · (3+α) ^ 49 · γ ^ 4, γ = 300 + 47α - α²,
which is positive on [0,17], turns the claim into positivity of the single polynomial
F = D (1 - Ut) of degree 58. The two closed forms used are
c 101 α = γ / (100 (3+α)),
T1 101 α = 1600000 (100-2α)² (3+α)³ / (721 γ⁴).
A Bernstein certificate on [0,17] settles it: all 59 coefficients of F in the basis
α^m (17-α)^(58-m) are positive, so no subdivision of the α-range is needed. (The
informal write-up splits at α = 16 and estimates the two pieces separately; that is not
necessary once the sharp Beta constant is used, which is what leaves the margin — Ut peaks
at α = 17 with value 0.9229.)
How the identity is checked #
D and D · Ut are written as coefficient lists (Sendov.endD, Sendov.endN), using the
polynomial arithmetic of Sendov.FiniteRange.Certificate, so that F = endD - endN is a
list the kernel can compute, and the Bernstein certificate Sendov.endB is a decidable
equality of lists (Sendov.pev_pos_of_bern). The only identity proved by ring is
pev endN α = Ut α · pev endD α (Sendov.endN_eq): each of the six terms of Ut cancels
against the matching factor of D separately, and (3+α) ^ 48 and γ ^ 4 are generalized
away before field_simp, so nothing of degree 48 is ever expanded. Upstream, this identity
and the certificate were closed by ring under a twentyfold heartbeat budget.
Main statements #
Sendov.Ut_lt_one:Ut α < 1on0 ≤ α ≤ 17;Sendov.large_degree:R n α < 1forn ≥ 101.
The denominator D = 865200 (3+α)^49 γ⁴.
Equations
- Sendov.endD = Sendov.pscale 865200 (Sendov.pmul (Sendov.ppow [3, 1] 49) (Sendov.ppow Sendov.gam 4))
Instances For
The Bernstein coefficients of 17 ^ 58 * F on [0, 17].
Equations
- One or more equations did not get rendered due to their size.