Answers are grounded in this paper and cite their sources.
Paper 6 of the Angelica Paper Series. AI data centres are moving to an 800 V DC bus. Using the Angelica engine, with nothing measured, this paper asks how the 650 V-class Angelica GaN-on-Si transistor should serve that bus. Device: With a three-step graded field plate, the 6 µm stack blocks 600–800 V at gate–drain lengths of 6–15 µm. Arrangement: No single device reaches 800 V with margin, so the bus must be split. With each device blocking 400 V, leg loss falls by 34 % (two in series), 54 % (input-series output-parallel) and 67 % (three-level flying capacitor). A 1200 V SiC two-level leg loses 60.7 W against 11.8 W for the three-level GaN leg. Base: At 800 V the base compound matters only through its DC resistivity. Gd₂O₃ ranks first, with (Gd,Sm)₂O₃ as the alternative. Power stage: A 10 kW stage reaches 99.75 % (hard switching) and 99.88 % (zero-voltage switching), counting GaN loss only. Moving the die to diamond changes almost nothing, because the package, not the die, holds the heat. Dynamics: The stack shows no current collapse while the channel layer leaks at least ten times more than the buffer. Benchmark: The intrinsic cell’s R_on·Q_oss is 3.8–7× lower than commercial GaN and SiC parts, before layout losses. The record contains the paper in English, Traditional Chinese and Japanese, a primer in all three languages, and the full code and sources. 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.
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: