ALP-ttbar Toy MC v1.8 — Preregistration: All-Pixel Three-Dimensional Vertex Fit with Reduced Material for Displaced Light ALP → e⁺e⁻

Preregistration of version 1.8 of a small, numpy-only toy Monte Carlo study motivated by the displaced-dielectron extension (Alimena et al., arXiv:2610.38932) of the top-associated long-lived ALP strategy of Rygaard et al. (JHEP 10 (2023) 138). This plan was frozen on GitHub before any v1.8 simulation or fit was run; the PDF reproduces the frozen README unchanged (SHA-256 ff00323388042f94b839af1db98f3c3e3746e6a4e28cfc7962f42ab190fefc42) and adds two figures made only from analytic formulas and published v1.6/v1.7 numbers. v1.6 found that only reducing the material budget substantially reduced the 17 MeV loss in the transverse view (127× → 65× → 35× for material scale factors f = 1, 0.5, 0.25). v1.7 found that measuring z at every layer separated ALP pairs from 50 MeV on but gained only 7.8% at 17 MeV, which was attributed to bremsstrahlung-reduced momenta that scatter more than the r–z covariance assumes. v1.8 combines both: an all-pixel 3D two-hypothesis vertex fit at f = 1, 0.5 and 0.25, plus a diagnostic covariance that uses the true momentum at every material surface (a truth-level upper bound for a bremsstrahlung-aware r–z fit, not a realistic method). Seven predictions (P1–P7), a sanity gate and three validations (W1–W3) are fixed in advance. Low-material all-pixel trackers are pursued by next-generation concepts such as ALICE 3, ALICE ITS3 and the FCC-ee detectors; this toy only mimics such a tracker and makes no statement about their performance. Independent personal project, not affiliated with the authors of the referenced papers or with any collaboration. GitHub: https://github.com/wspaik5/ALP_ttbar_Toy_MC_v1.8_Preregistration_AllPixel_Reduced_Material