By @josebrox
# Uniform complement guards and a reference-aligned normalizer The recovered 193-variable quartic and the v75 construction records are unchanged. Their complete counts remain 193 for a quartic, 25 in the general construction, and 11 under ordinary rational abc, with each construction's retained arithmetic premises and fixed-parameter qualification. The full v76 remote regression workflow passed at commit `a715d6dd1afd11288aad2dc165d86c1ea03a1b3f`. The new [complete TeX note](../papers/pole_parameters/uniform_complement_guard_20261117_v77.tex) proves a uniform guard theorem and a constant normalizer source transfer. Neither proves a complete nine- or twenty-one-variable definition. ## Every real point character has a positive complement Keep the actual v75 tests, full auxiliary points, point and difference guards, physical contact map, balanced product W and reciprocal trace V. At either real embedding every elliptic abscissa satisfies `x>-5`. For `u>=10`, the first four balanced tests obey `0<1/(x+u)<2/u`; the two comparison tests lie in `[0,1/u)`. Consequently, at every one of the 64 point characters, ```text 1 < W < (1+4/u^3)^6 2/(1+4/u^3)^6 < V < 2 0 < eta=28u/((1+u)^2 V^2) < 147/242 < 1. ``` The literal denominator `1+u/x` may be negative when x is negative. Its nonvanishing follows from the positive cleared denominator `x+u`. No positivity assumption on arbitrary full auxiliary abscissae is used. The existing physical choice `z>a+10` forces `u>10` for both contact sets. The exact lower bounds are `1199520/98251` for seven contacts and `15120/1093` for nine contacts. Thus the bound is compatible with the original order of completeness choices and uses no new coordinate. If the selected complement is square, the identity ```text c^2=1-d => P_d(1,c)=[2d(1-d)]^2 ``` supplies the native pole witness `(1,c,2d(1-d))`. With `d=eta`, the new complement and terminal ordinate are units at every character. One finite quadratic root replaces the two free pole parameters. The sound source has nine total coordinates and finite rank 128. Its soundness requires ordinary rational abc and the retained source premises, with a sufficiently large fixed odd defining shift. The missing assertion remains the existence of actual H, U and rational physical z with `V(H,U,z)^2-28w(z)` rationally square, at one fixed sufficient defining shift for every required integer input. Positive real values and the local preparation do not prove this rational square. ## A constant normalizer aligned at the reference trace The replacement ```text w*=16u/(4+7u)^2, b*=4w*/V^2, eta*=7b* ``` keeps the actual tests, detector and physical u unchanged. Retain all old guards and add `4+7u!=0`. Outside the existing exceptional primes two and seven, positive odd valuations of u retain their valuation in w*, negative odd valuations become positive, and all even cases stay even. Fresh primes in the new denominator contribute even valuations. The local extraction and square-denominator preparation therefore transfer. At two and seven, arbitrarily deep existing preparation uses the square leading term `w*~16/(49u)`; the added depths are fixed before the same prime-pattern theorem is applied. For `u>=10`, the full auxiliary real estimate improves to `0<eta*<294/1369<1` at every character. The unchanged intrinsic-pole theorem gives a complete eleven-variable source variant under the same premises. No new degree or coefficient-height bound is claimed for its changed circuit, and it does not replace the best compiled v75 record. At the formal reference value `V=2`, its complement is exactly `((4-7u)/(4+7u))^2`. On the chosen real completeness domain `u>=10`, every finite guarded actual tuple has `V<2`. Thus this identity supplies no actual witness. The corresponding nine-coordinate criterion still requires ```text S^2 = V^2 - 448u/(4+7u)^2 ``` at the actual trace and physical slope. This global assertion remains unproved, as does the original trace-square assertion. ## Other routes checked in this continuation The point-role audit returns to the existing v66 single coupling `U_S=Z+H`, with `b=(a-1)h`. It fixes h once the roots are fixed; varying the Lucas parameter changes the trace scale and therefore the prime-pattern coefficients. A fixed trace scale bounds the root indices, so it does not leave an infinite adjustment preserving that scale. No new complete point-coordinate reduction follows. The fixed-slope search also retains the existing restrictions on rational, elliptic and abelian witness families. These restrictions do not exclude all individual rational points or nonabelian families. The [v76 checkpoint](results_20261117_v76.md) retains the exact physical counterexamples and their scope. The lost prior modulo-53 nonlift fixture is still not recovered; the available modulo-53 certificates are squarefreeness certificates, and are not used as a lift obstruction. ## Reproduction and precise verification scope ```sh python3 -B scripts/verify_v77.py --report /tmp/v77-verification.json ``` The [exact package](../computation/pole_parameters/uniform_complement_guard_v77) has separate standard-library producer and independent checker. They verify the polynomial identities, exact integer margin, harmonic sums and rational endpoint bounds; the producer also binds the previously audited source metadata and guards. The universal inequalities and arithmetic source transfer have a separate mathematical review. The gate rebuilds the certificates in an isolated tree and compares their bytes, then checks the manuscript and render receipt bindings. It does not claim to reprove all imported arithmetic foundations or to find a rational point on either actual square cover. The [criterion records](../data/criteria_20261117_v77.json) keep both target completeness flags false. They record the positive guard and normalizer theorems separately from the unresolved global realization.