A laboratory preprint reports that YPtBi-based thin-film stacks retained a sizable negative effective spin-orbit-torque response while the YPtBi layer remained predominantly disordered through 400 °C. At 450 °C, X-ray diffraction showed changes consistent with crystallization onset. The pattern directs attention to chemical mixing at the upper YPtBi interface, rather than bulk crystallization, as the more relevant design target. The authors present that relationship as an association, not a demonstrated mechanism.
The question behind the experiment
The work focuses on whether a technologically useful spin-orbit-torque response can survive without bulk YPtBi crystallization, and whether the upper YPtBi/W interface controls the response. Its working hypothesis places the decisive variable in a chemically mixed, electronically active upper interface rather than in long-range order throughout the YPtBi layer.
To examine that question, the researchers built multilayers on Si/SiO2. The nominal design included a 100-Å YPtBi layer, an 8-Å tungsten layer and a 12-Å CoFeB layer beneath an MgO-based cap. The films were deposited by physical vapor deposition under cold-cap and hot-cap process families, with YPtBi deposition conditions spanning room temperature to 450 °C.
The stack was examined in several ways. Anomalous Hall and second-harmonic Hall measurements assessed transport and torque response; X-ray diffraction and cross-sectional transmission electron microscopy examined structure; X-ray reflectivity measured thickness, roughness and density; electron-energy-loss spectroscopy assessed interface composition; and sheet-resistance mapping was used for current partitioning.
Disorder did not erase the signal
X-ray diffraction found no strong YPtBi diffraction peak from room temperature through 400 °C. At 450 °C, it showed qualitative diffraction and texture changes consistent with crystallization onset. The lower-temperature films therefore formed the reported disordered window, although X-ray diffraction alone could not distinguish a fully amorphous film from nanocrystallites below its detection limit.
Local imaging gave a similar but more nuanced picture. TEM and fast-Fourier-transform views showed predominantly diffuse contrast in the room-temperature and 350 °C samples. At 400 °C, the film remained predominantly disordered, but limited larger grains appeared in an amorphous or weakly nanocrystalline matrix.
Despite that structural state, the measured magnetic and electrical response persisted. Anomalous Hall loops remained well defined, while second-harmonic signals remained sizable and negative under the paper’s sign convention from room temperature to 400 °C in both process families.
That persistence was not confined to a single thermal route: the torque signal remained present across the reported disordered range rather than appearing only alongside the diffraction and texture changes seen at 450 °C.
Resistance alone did not offer a clean explanation. The effective spin Hall angle did not vary monotonically with resistance, arguing against a purely conductivity-driven reading of the data. The paper describes this comparison as a trend in the reported figures, not a result backed by formal statistical testing.
The boundary became the clue
Cross-sectional images showed that the YPtBi/W/Co-containing boundary was visibly broadened, with the broadened zone localized near the interface. That pattern was consistent with interfacial rather than homogeneous modification of the YPtBi layer.
Composition measurements made the boundary more specific. The nominally abrupt YPtBi/W boundary became compositionally graded, with tungsten penetrating the upper YPtBi surface and mixed YPtBi(W) and W(Pt) material forming across the region.
The authors then used a two-source spin-conversion model to estimate how much of the measured response might come from the mixed region. For relevant Pt-W interlayers, they estimated compositions of roughly 50–65% W and 35–50% Pt, an effective thickness of about 8 Å and conductivity of approximately 5×10^5–8×10^5 S/m. The model assigned this interlayer a small positive effective spin Hall angle of 0.005–0.015, a contribution the authors said was insufficient to explain the large negative response of the full stack.
Across the reported measurements, higher integrated W concentration at the upper YPtBi interface tended to accompany a more negative effective response. The relationship showed scatter, however, and was not treated as definitive proof of an interface-doping mechanism. The authors therefore place more weight on upper-interface chemistry than on bulk structural ordering when interpreting the effective response.
A mechanism still needs testing
The authors’ conclusion is a change in design emphasis. They argue that engineering the top interface of disordered YPtBi may be a more relevant route than enforcing bulk crystallization for the targeted process. In that view, the important materials variable is the chemical state of the upper boundary, not simply whether the underlying YPtBi has formed a long-range crystal.
The explanation remains deliberately limited. The proposed electronic-structure mechanism was not tested directly, and the authors identify first-principles calculations as a needed way to examine whether the interfacial chemical changes affect the relevant electronic states.
The study’s strongest result is therefore a materials association: a sizable negative response coexisted with predominantly disordered YPtBi, while structural broadening and chemical grading appeared near the upper interface. That combination supports further testing of interface engineering, but it does not establish that W incorporation itself produces the observed response.
The authors report no conflicts of interest. They say the manuscript data can be obtained from the first author upon reasonable request, while no funding statement is reported in the supplied document.
Paper data and sources
Original title: Amorphous and Nanocrystalline Topological Semimetal YPtBi/W/CoFeB Heterostructures for BEOL-Compatible Spin-Orbit Torque Devices
Authors: Quang Le, Brian R. York, Cherngye Hwang et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text