back to table · edit · history · where entries came from · files · download
6911 bytes, as of the version from 2026-11-20 09:32 (current). Recorded here, not run.
# Mathematical construction and checked scope
This is build-session evidence, not an independent scientific verdict. All ten exact stored polynomials were checked; an independent actor must recheck the revision before the controller offers it. Canonical instructions: ../20261120T092423Z-build-skills/manifest.json, bundle SHA-256 b5c3c7985b7c9e178b4899d083a9e3bc4284989dff76f05e5b0e52835d47eed2. The complete newly reachable live bundle matches all archived resource hashes (skill-alignment.json).
## Sources and normalization
Hartmann arXiv:1101.4601v1 Section 4.1 defines the original Fermat pencil sum(X_j^4)-4*t*product(X_j)=0, its group quotient and minimal resolution. Definition 4.3 and Proposition 4.4 exclude t^4=1 and infinity. Theorems 4.8 and 4.11 give the primitive integral transcendental marking T_0=<4>+U, orthogonal to the generic algebraic rank-19 lattice; equation (10) is [P*h-e+2*P^2*f]. Example 4.27, Theorem 4.29 and Proposition 4.30 give the Frobenius branch at t>1, positive imaginary and tending to i*infinity. The isolated inconsistent vertex in Theorem 4.37 is not used. The projective ratio cancels a common residue scale; no absolute period claim is made.
Dembele et al. arXiv:1906.07384v3 equation (2.5) gives rho=256*R/(R+64)^2 with R=(eta(tau)/eta(2*tau))^24 and eta nome exp(2*pi*i*tau). Thus t^4=(R+128+4096/R)/256. Yang arXiv:2004.00820v1 equations (4.29), (4.33), (5.3) and (5.15)-(5.17) give the theta lambda and period comparison; Yang's t=psi^-4 is the inverse fourth power of the present base coefficient. The theta nome exp(pi*i*tau) is the square root of the eta nome. These identities and analytic continuation on the real positive branch, rather than a bare modular equality, fix P(t_r)=i*sqrt(r).
The lattice Z+Z*tau at tau=i*sqrt(r) has multiplier ring Z[sqrt(-r)]: a multiplier must equal a+b*tau by applying it to 1, and conversely its product with tau is -b*r+a*tau. Its discriminant is -4*r; square factors change conductors and are not removed. The associated primitive form is (1,0,r), so j(tau) is a root of the Hilbert class polynomial H_{-4r}. The PARI polclass manual states that inv=0 is the classical j polynomial and its CRT algorithm uses algebraic modular equations, rather than floating point polynomial recognition.
The canonical level-two eta invariant in the PARI ellmodulareqn manual is m=2^12*(eta(2*tau)/eta(tau))^24=4096/R. The returned type flag 0 means canonical; the actual polynomial returned by PARI 2.17.1 is (m+16)^3-m*j. Hence j=(m+16)^3/m. The exact relations used by the producer are m^h H((m+16)^3/m)=0 and 256*x^4*m-(m^2+128*m+4096)=0. No pole is introduced at the actual m>0.
## Why the selected factors are exact
The CM theorem and the exact modular equations make the analytic m and t roots of the elimination polynomials. The producer factors over QQ and evaluates EVERY distinct factor on a rigorous positive-embedding ball, retaining exactly one and excluding all others. The compatible irreducible factor is made monic. At each row an exact rational bracket has opposite endpoint signs, and a derivative enclosure excludes zero, so it contains one simple real root. Irreducibility and the CM/modular implication identify it with the minimal polynomial; a residual containing zero alone would not prove annihilation. Resultants are not assumed irreducible. The polynomial may have other conjugate roots; no common marked-period or cycle claim is made about all of them.
At r=1, H_-4(j)=j-1728 and the canonical equation becomes (m-8)^2*(m+64)=0. Positivity forces m=8, hence R_1=512 and t_1^4=81/32 exactly. The positive product R=q^-1*product(1+q^n)^-24 increases as q decreases. Thus R_r>=512 for every r>=1. Since R+128+4096/R increases for R>64, every t_r^4>=81/32>1, independently confirming the smooth positive branch throughout the domain.
## Bounds, alternate formula and controls
The producer uses at least 600 bits and q=exp(-2*pi*sqrt(r)). Summing log(1+q^n) through N leaves a nonnegative tail at most q^(N+1)/(1-q), since log(1+s)<=s. Arb encloses this tail and every later operation including positive roots. Minimum measured relative accuracy is 590 bits.
The checker imports no producer function. At 900 bits it independently sums theta_3=1+2 sum(q^(n*n)) and theta_2=2*q^(1/4) sum(q^(n*(n+1))), with q=exp(-pi*sqrt(r)) and N=24. Their positive tails are bounded respectively by 2*q^((N+1)^2)/(1-q^(2*N+3)) and 2*q^(1/4)*q^((N+1)*(N+2))/(1-q^(2*N+4)). The resulting lambda gives t^4=(lambda-2)^4/(16*(1-lambda)*lambda^2).
The alternate exact elimination clears H(256*(1-lambda+lambda^2)^3/(lambda^2*(1-lambda)^2)) and eliminates lambda against 16*(1-lambda)*lambda^2*x^4-(lambda-2)^4. Every wrong factor is excluded on the independent t ball, and the unique monic factor equals the stored polynomial exactly. All ten root brackets at 100 significant digits have opposite exact endpoint signs and a nonzero derivative enclosure. The theta minimum relative accuracy is 891 bits. An inverse elliptic-AGM calculation on the ENTIRE 100-digit root bracket overlaps sqrt(r), with absolute radius less than 2^-300. AGM at finite iteration encloses its limit between the arithmetic and geometric iterates, including their ball uncertainties.
Known-answer controls include H_-4=x-1728, H_-8=x-8000, lambda(i)=1/2 and M_1=x^4-81/32. Deliberately altered constant coefficients, nonmonic polynomials, wrong index, negative root, doubled theta nome, NaN, huge balls and the wrong lattice sign are rejected. Nonzero Gram pairings are also checked, so an always-zero pairing function is not accepted. See check_evidence.py and the actual offline/stored JSON outputs. The checker shares PARI class polynomials, Sage factorization, Arb and the mathematical source identities with the producer; this limitation is explicit.
## Geometric certificate and limitations
In the marked integral basis, gamma=(0,1,-2*r) is nonzero and primitive because its second coordinate is 1. The exact Gram matrix [[4,0,0],[0,0,1],[0,1,0]] gives gamma^2=-4*r and gamma.Omega=2*(P^2+r)=0 at P=i*sqrt(r). Since gamma is real integral, orthogonality also holds to the conjugate period; it is a (1,1) class. Hartmann's primitive embedding and Lefschetz (1,1) give an additional divisor class outside the generic algebraic lattice. Its norm is bounded by 4 through 40 in this finite roster. This is not an effective curve equation, a bound on degree in a polarization, an integral saturation or a classification of all special fibres. The exact rows are in lattice-certificate.json. All selected positive roots exceed one, which also verifies smoothness.
The auxiliary root brackets target 100 significant digits; the searchable outputs are exact rational polynomials, so no padded decimal precision is claimed. There is no quantile data, relation-finder discovery claim or completion claim for the broad wanted research direction.