No smaller signed-turning constant works for finite laws
ProvedCubicTurning.no_smaller_signed_turning_constantFor , use Euclidean modulus and set , with real division by zero defined to be . Let mean or the same two inequalities with exchanged. Put , if holds and otherwise, and if holds and otherwise. Every threshold inequality is non-strict. For complex put . A length-three chain means precisely , , and . Being forward for an axis and consecutive displacements , means , , and ; zero displacements and zero projections are allowed. A finite balanced forward law with slots for consists of real weights for with , length-three chains for this same triple, and axes forward for those chains, with the complex equality . Its mean energy is . A slot is not prescribed to terminate at a particular labelled root, weights may be zero, and all slots, including zero-weight slots, must satisfy their chain and forward conditions. No separate common-start field is required. For the normalization is impossible because the empty sum is , so there is no such law. For real , define , , and , with the nonnegative real square root. Put and . For every real , there exists a real with such that for every natural and every finite balanced forward law for with slots, . The inequality is strict. The same works for all cardinalities and laws; no sign condition on is present, and is excluded. Zero weights are allowed, and the case is vacuous because no normalized law exists.
import Mathlib import Definitions.Def_CubicTurning_finite_balanced import Definitions.Def_CubicTurning_foundations import Definitions.Def_DerivativeChainPrivate_foundations set_option autoImplicit false set_option maxHeartbeats 1200000 open scoped BigOperators open ComplexConjugate
theorem CubicTurning.no_smaller_signed_turning_constant (K : ℝ) (hK : K<2/3) :
∃ t : ℝ, 0<t ∧ t≤1 ∧
∀ (n : ℕ) (L : FiniteBalancedForwardLaw n (sharpCentral t) (sharpOuterPlus t) (sharpOuterMinus t)),
K*variance (sharpCentral t) (sharpOuterPlus t) (sharpOuterMinus t)<
∑ i,L.weights i*halfAngleEnergy (L.axes i)
(L.paths i 1-L.paths i 0) (L.paths i 2-L.paths i 1) := by sorry