default(realprecision,160);
f = mfDelta(); mf = mfinit(f,0);
P = mfperiodpol(mf,f,1);
imag(polcoef(P,10))/36 /* real-scale for (N,k,i)=(1,12,1) */The 100 significant digits are obtained from PARI's direct period-polynomial evaluator at 160 and 220 decimal working digits, with 32 additional working bits, using the exact leading coefficient of the primitive shape. The two evaluations define an agreement interval. This is heuristic agreement, not a general certification of PARI's numerical algorithms [5]. The signed scales agree with every nonzero even coefficient; exact zero components follow from the defining real Fourier coefficients and path.
Exact Sage modular symbols [1] reconstruct all 18 primitive polynomials and match the published shapes. Independent dimension formulas [3] agree in all 21 selected spaces, including empty spaces. Separate Sage and PARI newform constructions agree through the Sturm bounds on all 33 real embedded forms; the retained indices also match the companion roster. For each rational form, the first 600 Fourier coefficients agree exactly. An integer eta-product convolution checks the first 100 coefficients of Delta.
A separate split Mellin computation uses Sage Fourier coefficients, 450 terms at 150 decimal working digits and 600 terms at 190, with ball arithmetic and a Fourier-tail majorant from the normalized-newform coefficient bound [4], Theorem 2.6. The split point is $1/\sqrt N$. The exact Fricke eigenvalue is computed in Sage after dividing its modular-symbol operator by $N^{k/2-1}$ [2], and independently checked with PARI. Every even coefficient, including forced zeros, is tested. All nonzero coefficient-derived scales lie inside the recorded 100-digit decimal intervals, with worst residual less than 0.47 units of the stored last place. The narrowest independent scale ball has more than 150 significant digits.
Three nonzero exponential-integral controls, Delta's primitive shape, and deliberately wrong signs and doubled scales test the checker. Nonfinite and excessively wide balls are rejected. The exact roster, executable generator and checks, numerical evidence, argument-selection rationale and coordinated skill manifest are attached. These calculations supply build evidence; independent scientific review is a separate step.