# Research: primitive even modular periods Checked definitions: PARI's mfperiodpol(mf,f,1) is the even part of the upward integral of (X-tau)^(k-2) f(tau) from 0 to i infinity. Since k is even, this equals the family's (tau-X) convention. Paşol–Popa Section 1 defines the same identity-coset component; Proposition 5.11(a,c) supplies rationality up to a scalar and its imaginary phase for rational real coefficients. Their Equation (5.7) relates moments to the uncompleted arithmetic L-value with (2*pi)^(-s)*Gamma(s), without a conductor factor. Thus the coefficient of X^(w-m) is binom(w,m)*i^(1-m)*m!/(2*pi)^(m+1)*L(f,m+1), w=k-2. Normalization: P_f^+ is primitive integral with positive leading coefficient; Omega_f^+ is signed in r_f^+=i*Omega_f^+*P_f^+. The path presentation is exactly i*Omega_f^+, not an absolute value. Reality comes from the Fourier coefficients and path, not from dropping numerical noise. Delta's primitive shape has leading coefficient 36, so Omega_Delta^+=L(Delta,1)/(72*pi). Live discovery: catalogue.json caches all eight pages (761 tables). Exact title, primitive-period and Manin/Shimura searches found no existing family. The already_here candidates are not duplicate verdicts. already_asked finds proposal #349 itself. lookup-examples.json preserves a positive T323 control and related source-period probes; their absence does not establish uniqueness. All four suggested tags (modular form, number theory, period, special values) were checked by exact live registry lookups and apply to the defined quantity. Origin #262 supports explicitly normalized modular periods; #303 and #309 are context rather than direct origins. Initial history must cite #349/#262. Related tables: T695 holds the exact primitive shape, T696 the separate odd shape; T323/T683 hold critical L-values in the same arithmetic convention; T326/T684 hold Petersson self-products. The current T683/T684 higher-level roster is in other weights and is not an oracle for these weight-12 values. Only full coset-valued period data justifies a higher-level Haberland pairing; no norm equality from an identity-coset polynomial is asserted. The odd scalar sibling has no T-number yet and receives no guessed link. Reusable exact roster/shape code was fetched from the website's T695 attachment, SHA-256 a6fc713059fcc42db1e7e4f62f41fade07d68f41e31e70581109e0bc055c7e87, and isolated as forms.py. It is a computation component to be checked, not independent evidence. The independently constructed PARI forms must match all Sage embedded coefficients through the Sturm bounds. Numerical generation uses PARI mfperiodpol at 160/220 decimal working digits; the independent Mellin route uses Sage's exact Fourier coefficients and its exact Fricke eigenvalues. Truncation bounds and numerical controls are recorded by checks.py. Source pages are cached in pari-modular.html, pasol-popa.html, sage-modsym.html and pari-lfun.html; URLs and hashes are in source-hashes.json. Sage's rational_period_mapping documentation describes an exact rational vector-space map with the same kernel as the analytic period map. PARI's generic L-function accuracy disclaimer prevents inferring proven bounds from generic evaluations. The stated table rigour remains heuristic (agreement-checked). Instruction evidence: manifest.json is the unchanged coordinated archive manifest; its resources were verified and its valid local/live bundles match. The core SHA-256 is b79b7fbc9fa3ce63f2d642cccdac12b0d214757ab01db66c03a2d84264a58e5f. All six task-relevant references were read. No skill or pipeline source changed. Scientific and argument-selection validation by a fresh critic remain required. Checked computation results: all 36 entries pass the independent Mellin comparisons; all 18 shapes match live T695 exactly; all 33 embedded-form indices and all 21 space dimensions agree. Every rational form has 600 exact Fourier coefficients crosschecked between Sage and PARI. Stored decimal intervals contain the independently derived scales for every nonzero even coefficient; worst residual is below 0.47 stored ULP. The narrowest independent scale ball has 520 bits of relative accuracy. The largest upper-integral Fourier-tail majorant is below 4.65e-377 (before its displayed rounding). The final declared rigour is still heuristic (agreement-checked). New inspected source passages: Bringmann–Han–Heim–Kane, Vanishing properties of Fourier coefficients of holomorphic eta-quotients, Theorem 2.6 on page 4, states |a_n|<=d(n)*n^((k-1)/2) for normalized newforms. The checker derives the looser bound 2*n^(k/2) using d(n)<=2*sqrt(n), and sums a positive geometric majorant after the incomplete-gamma expansion. Stein's dimension chapter, Propositions 6.1 and 6.4, supplies the independent dimension formulas. Sage's atkin_lehner_operator documentation explicitly includes the factor N^(k/2-1); the checker divides by it before using the Fricke sign. The exact Sage docstring and N=1,3,4 controls are saved in probe-fricke.log. Execution provenance: checks5.log is the completed run, with per-form checkpoints. The six level-one Fourier/Mellin computations are reused from checks4.log because the subsequent correction only normalizes the higher-level Atkin-Lehner operator. resume-checkpoints.json binds those results to unchanged generator/forms hashes and archives the prior checker hash; checks4-source.py preserves that checker. All indices and shape comparisons were rerun. Earlier failed runtime probes remain as artifacts. The successful dry-run.log executes the existing dry_run.py and check.py.