identity-notes.md

back to table · edit · history · where entries came from · files · download

2671 bytes, as of the version from 2026-11-17 11:01 (current). Recorded here, not run.

# Symbolic identity notes for T735

These are short mathematical derivations supplementing the inspected sources
and finite checks. They do not constitute the controller's independent critique.

For a finite field F, extend multiplicative characters by zero at 0 and fix a
nontrivial additive character e. In G(chi)G(psi), collect the terms x+y=z.
For z=0, the contribution is psi(-1) times the sum of (chi*psi)(x), which
vanishes when chi*psi is nontrivial. For z nonzero, write x=z*t and y=z*(1-t).
The summand factors as (chi*psi)(z)*e(z)*chi(t)*psi(1-t). Thus
G(chi)G(psi)=J(chi,psi)G(chi*psi). This proves the Gauss–Jacobi identity used
in Similar tables with its stated nontrivial-character hypotheses; agreement
over F7 is an additional implementation control, not its universal proof.

For an order-three character on F_p and a primitive root g modulo p, write
each nonzero residue uniquely as g^(j+3*t), 0<=j<3 and 0<=t<(p-1)/3.
The character is chi(g)^j, since chi(g)^3=1. Grouping these terms gives
G(chi)=eta_0+chi(g)*eta_1+chi(g)^2*eta_2, with the periods eta_j defined in
the table. A period polynomial supplies the unordered periods, not this
weighted ordering. This explains the T146 relation without a phase claim
inferred merely from polynomial roots.

Because c is prime to 3, lambda=1+2*omega is invertible modulo c. In the
different-additive sum substitute x=lambda*y. This gives
G_dif(c)=(lambda/c)_3*G_tr(c). More generally changing a denominator to u*c,
for a unit u, gives (u/c)_3*G_tr(c) by substituting x=u*y. The cubic character
sends -1 to 1, so a sign change cannot be used as a negative control here.
The local checker instead verifies this sign invariance and tests the
possibly nontrivial omega-unit factors. The sextic character need not have
the same sign invariance.

Conjugating the denominator and residue representatives conjugates the cubic
character. Changing the residue variable by its negative accounts for
conjugation of the additive exponential, using (-1/c)_3=1. Hence
G_tr(conj(c))=conj(G_tr(c)). If rational c is squarefree and primary, each
split rational prime contributes two ideal primes and each inert prime one.
Its primary sign implies mu_O(c)=sgn(c). The source cube identity therefore
gives G_tr(c)^3=abs(c)^3. Reality forces G_tr(c)=abs(c), an exact integer.
This branch rule is proved, rather than read from rounded decimals.

The remaining cube, magnitude, CRT and prime-power laws are the inspected
unconditional identities in DR Section 1.2 and DFDS Section 3.3, not any
GRH-dependent asymptotic assertion in those papers. Executed coverage and
known/wrong controls are in the attached numerical evidence.