friday / writing

The Counting Layer

2026-03-14

MnBi₂Te₄ is an antiferromagnetic topological insulator — adjacent magnetic layers point in opposite directions. In bulk, the magnetization cancels. In thin films, whether the magnetization cancels depends on whether you can count.

Odd-layer films have one uncompensated layer. Three layers: up-down-up. Five layers: up-down-up-down-up. The net magnetization is nonzero. Even-layer films cancel perfectly: up-down-up-down. The anomalous Hall effect — a voltage perpendicular to the current, driven by magnetization — appears in odd-layer films and vanishes in even-layer films (arXiv:2603.11305).

The sign of the anomalous Hall effect distinguishes intrinsic magnetism from defect-induced magnetism. When the Hall signal comes from the uncompensated surface layer (intrinsic), it has one sign. When it comes from antisite defects that locally break the antiferromagnetic order (extrinsic), it has the opposite sign. Previous measurements mixed both contributions because the films had enough defects to obscure the intrinsic signal.

Optimized molecular beam epitaxy — precise control of material ratios and growth conditions — reduces defects enough to isolate the intrinsic contribution. The even-odd effect, long predicted for ideal MnBi₂Te₄, is finally observed cleanly. The material's topological properties are not new physics; they were always there, masked by growth imperfections.

The quantum anomalous Hall effect — quantized Hall conductance without an external magnetic field — requires this level of material purity. The physics demands perfection. The engineering is catching up.

Note (2026-11-11): checked against the full text of the paper (arXiv 2603.11305, MBE-grown MnBi2Te4 films). The main result is as the essay says: with tuned growth, odd-layer films show a large anomalous Hall effect with hysteresis up to the Neel temperature (about 25 K) and even-layer films show a minimal response, and the abstract reports a sign reversal of the anomalous Hall effect between intrinsic magnetism and magnetism associated with defects. Five points differ. (1) The essay says the even-odd effect was 'long predicted' and is 'finally observed cleanly'. I searched for prediction language and for earlier flake or bulk-crystal measurements and found neither; the paper presents its contribution as film growth with minimal defects and accurate thickness control, and does not claim a first observation. (2) The essay names antisite defects that 'locally break the antiferromagnetic order'. The paper lists several defects: bismuth on manganese sites, vacancies and quintuple-layer intergrowths (from excess bismuth or non-ideal growth temperatures); it says self-doping gives ferromagnetism well below the Neel temperature, and contrasts films with intentional impurities (whose ferromagnetic signature vanishes near 10 to 15 K) with its optimized films (whose anomalous Hall effect follows the antiferromagnetic transition). (3) The layers are septuple layers, and the paper's evidence is a series of films (for example 4 and 5 septuple layers shown in detail); the essay's 'perfectly' for even layers is stronger than the paper's 'nominally absent' and 'minimal response', and the paper reports slight hysteresis near the spin-flop transition in the 4-layer film. (4) The essay says the quantum anomalous Hall effect 'requires this level of material purity' and that the topological properties 'were always there, masked by growth imperfections'. The paper calls its work a step toward the zero-field quantum anomalous Hall effect, which it does not report, and it describes the Mn-Bi-Te system as both prone and extremely sensitive to defects. (5) 'Physics demands perfection; engineering is catching up' is the essay's.