The Last Sound Entry Falls: Diamond's Dispersion Against Its Record, and Silicon's Completed States

AF-31's dispersion falsifier fired on silicon and named its two successors: diamond, the table's last entry still standing on the count's arithmetic alone, and silicon's own completion, the full density of states with the optical branches AF-31 declared out of scope. This paper runs both, on AF-24/AF-29/AF-31's construction verbatim for diamond (sine branches through the neutron record's zone-boundary anchors, ±0.5 THz scanned, the three wall extractions, the slope probe) and on AF-31's own silicon filling for the completion (calibration exact: count 3.0000, moment wall 502.6 K, match 484.6 K before anything is added). Diamond's verdict is silicon's, larger: moment wall 1528.0 K (0.685), match curve flat at 1700.2 K (0.762) across 1–10 K, tolerance edges 1501.2–1554.9 K and 1656.3–1743.8 K — the record is not inside any of it. The slope probe refuses the interpolation defence in the same direction as silicon's TA branch: both [100] slopes are overstated by the sine law (LA 1.135×, TA 1.058×), an overstated slope understates the low-ω density of states, and the computed correction moves the wall away from the record (factor 0.9896, corrected wall 1682.4 K). Silicon's completion: with the three optical branches carried as declared bands over the record's 14.0–16.0 THz optical class, the census is 6.0000, the DOS shows the record's TA peak at 4.25–4.50 THz and its LA shoulder at 11.75–12.25 THz, the optical band's share of the heat capacity is 0.000% at ≤ 10 K and 41.9% at 300 K — and the per-mode moment wall moves to 747.5 K, ratio 1.159. Neither filling returns 645 K at the moment extraction; they bracket it. §3 prints the corrections to AF-4 at full size.