Answers are grounded in this paper and cite their sources.
Second edition (November 2026). It replaces the first edition (June 2026, doi:10.5281/zenodo.20670713), which proposed a spin loop in the Angelica GaN-on-Si stack (strain → Rashba coupling → spin-orbit-torque switching → topological edge states from an Fe₃GaTe₂ magnet). Recomputed with the series' engines, that loop is withdrawn, and the loop the series can actually close is built: a loop of memory. (1) The spin loop. The 2DEG sits 2 µm above the grading; across −350 to +150 MPa its Rashba field moves by 0.9 %. A surface magnet reaches the 2DEG through the barrier with a weight below 10⁻⁴⁰, so no exchange splitting or topological state follows. The measured Rashba coefficient of GaN 2DEGs is 14 times below the value used, and spins diffuse about 0.25–0.35 µm. (2) A disturb matrix of the series' stores in one gate: Ce (Styx) survives every operation of the Eu short-term memory (Janus), and every Ce write empties the Eu; the Hypnos ferroelectric copy is fragile at the coercive field its copy requires. (3) One gate, two stores: the Ce holes bias the Eu spacer (0.011–0.11 MV/cm) and add to the read; consolidated cells capture and forget more slowly (metaplasticity). (4) The loop closed: days of learning, nights of replay, light consolidation into Ce. In one gate the replay necessarily reads both stores and runs away to a single attractor unless the Ce read is weak; with the Ce in its own transistor, switched out of the replay, the loop never runs away and keeps about a third of the important patterns. (5) Without erasure the long-term store holds its memory in part-per-million differences and 1 % cell spread erases it; a nightly 3 % erase makes it a palimpsest that survives realistic cells. A night's light costs 13 µJ–1.3 mJ for a 200-neuron array. The Ce bias alone turns the loop obsessive above a critical value of about 0.07–0.13 MV/cm, which moves with the consolidation step, the Eu self-field and a day-to-night hold — the question left to Raphael. Files: paper in English, Traditional Chinese and Japanese (the English version governs); a primer in each language; code and sources (code_and_sources.zip) reproducing every number and figure. Computational throughout; nothing is measured. Part of the Angelica Paper Series. Licence: text and figures CC BY 4.0; code Apache-2.0. Defensive publication: disclosed as prior art, free for anyone to use and not to be patented by others. This edition replaces the first edition's licences (CC BY-NC-ND 4.0; BUSL 1.1).
No takes yet. Share an insight, caveat, or question.
Tung Ning Liu (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: