A near-monochrome collage of seven laboratory photographs cut into diagonal panels: a cryostat and vacuum chamber trailing cabling and steel flanges; an optical table with posts, a glowing lamp and a circuit board; a bright ring of light seen through a dark chamber viewport; a chamber interior with a finned copper disc, translation stages and stepper motors; a rack of long cryostat probe inserts with stacked baffles; a vacuum chamber wrapped in aluminium foil with pneumatic valves; and a chamber with a brilliant glowing plume beneath a circular electrode.

Interdiffusion at a 2D metal/semiconductor contact, and what it does to device performance

The short version

Two magnetic crystals stacked together are usually assumed to sit side by side without mixing. Here, simply cooling the device down and warming it up again moves atoms across that contact, closes the gap between the layers, and changes the strength of the magnetic coupling by around a factor of ten. Nothing about the fabrication changed; only the thermal history did.

MnPS3 is a compensated antiferromagnet: on paper its interface spins cancel and it should pin nothing. Stacked on the metallic ferromagnet Fe3GeTe2 (FGT) it produces 170 mT of exchange bias at 5 K, among the largest reported for any van der Waals heterostructure. Scanning nitrogen-vacancy magnetometry located the source: below 40 K MnPS3 undergoes a spin reorientation that leaves a small uncompensated moment, and the bias tracks that moment rather than the 78 K Néel temperature.

The second result is the one with engineering consequences. Cooling and rewarming the same device moves its bias between 25 and 132 mT, a swing approaching 1000% across devices, with no change to fabrication. Cross-sectional electron microscopy shows why: the van der Waals gap at this metal/semiconductor contact closes by about 9 nm over two thermal cycles, and elemental mapping shows atoms migrating across it. Interdiffusion at a nominally inert 2D interface is therefore a thermal-budget and reliability question, not only a growth one. The link is correlative, drawn from cycled and uncycled lamellae rather than from one interface watched continuously.

Three panels. Top left: an optical micrograph of the device, a blue FGT flake under a red-outlined MnPS3 flake and a dotted hBN outline, on gold contacts, with a 10 micrometre scale bar. Bottom left: a bar chart of exchange bias magnitude in millitesla against measurement number from 1 to 15, the bars ranging from about 25 to 132 millitesla with no monotonic trend. Right: six false-colour elemental maps in two rows, iron, germanium and tellurium for the Fe3GeTe2 layer above, manganese, phosphorus and sulfur for the MnPS3 layer below, each with dashed white lines marking the interface region and a 5 nm scale bar.

The device, fifteen consecutive cooling cycles on it, and the elemental maps across the interface. The bias magnitude changes by a factor of five between cycles on one device, and the elemental maps show the intermixed region that accompanies it.

Shared first authorship with Aravind Puthirath Balan. Published in Advanced Materials 36, 2403685 (2024), 10.1002/adma.202403685. Analysis code: EB-in-FGT-MPS-vdW-heterostructure.