A resistance peak in a twisted graphene device appears to depend strongly on how the device is brought into an ordered electronic state. In a new laboratory preprint, different gate-voltage paths and in-plane magnetic fields were associated with different outcomes: some preparations led to a pronounced peak at zero in-plane field, while others did not.
The result comes from low-temperature transport measurements on one micron-scale Hall bar made from two Bernal-stacked bilayers twisted by approximately 1.34 degrees. The experiment tracked longitudinal resistance, Hall resistance and differential resistance while varying gate voltage, magnetic-field direction and the route used to reach the target state. The readings used a typical alternating-current bias of 1.4 nanoamperes, and the field-zero calibration had an estimated uncertainty of 0.1 millitesla.
A narrow feature with a selective home
Most resistance traces as the in-plane field was swept were featureless. Strong peaks at zero in-plane field clustered inside the high-filling halo, above a filling factor of 3.5, while peaks near the halo boundaries were much weaker. The feature was confined to a millitesla-scale window.
Its response depended on field orientation. The peak was isotropic within the two in-plane directions tested, but it was unaffected by an out-of-plane field of up to 10 millitesla.
The route into the state mattered
The peak was not a fixed property that appeared every time the device was measured. Over long periods it was generally not repeatable, and the sample could jump abruptly between resistance states during a single scan. The authors describe this as metastable behavior, meaning that the device could remain in a temporary state before switching to another one.
The clearest comparison came from 20 repeated approaches into the ordered region. Along one gate-voltage trajectory, called path alpha, the device never reached a state that later produced the in-plane-field peak. Along path beta, it frequently did, although the peak height varied greatly from cycle to cycle.
The resistance measured immediately after the gate-voltage approach provided a practical clue. States with higher arrival resistance were the ones that later showed the peak, and the arrival resistance closely tracked the eventual peak height. The paper notes a small mismatch in some cases, so the first reading was a useful marker rather than a perfect predictor.
Small fields were linked to different branches
The magnetic field applied during the gate sweep was also associated with which resistance branch the device entered. After the filling passed 3.4, the traces split according to the in-plane field used during preparation. With the measured in-plane component set to zero, the high-resistance state occurred in approximately half of the approaches. At 0.6 millitesla in that direction, the device consistently entered the low-resistance state.
The pattern was more detailed than a simple field-strength effect. For positive out-of-plane fields, the large-peak state appeared mostly at positive in-plane fields, and the preference reversed when the out-of-plane field changed sign. The strongest selection occurred near plus or minus 0.3 millitesla, depending on that sign. The state disappeared during the ramp when the in-plane field reached roughly 0.4 millitesla in magnitude, while it persisted to out-of-plane fields six times larger.
A much larger preparation field brought back the contrast even along path alpha, which otherwise did not produce the peak state. Above an approach field of approximately 175 millitesla, the on-peak resistance rose to 11.5 kilohms and the off-peak resistance to 5.5 kilohms.
What the authors think is happening
The authors associate the peak with a pattern of magnetic domains and suggest that this pattern carries the device’s preparation history. They identify the domains as regions with different valley polarization, based in part on the fact that the peak appears where transport shows an anomalous Hall effect associated with spontaneous breaking of time-reversal symmetry.
They propose that spin is part of the link. In their interpretation, Kane-Mele-type spin-orbit interaction ties valley polarization to spin pointing out of the graphene plane, while the field-direction result is consistent with a role for spin’s Zeeman coupling. A current-driven precession of domain walls is offered as one possible explanation for the extra resistance.
A promising clue, not a finished explanation
The measurements support an association between preparation history and the later resistance state, but they do not directly image magnetic domains in this device. The history-dependent evidence comes from the reported Hall bar, so the paper does not establish that the behavior is general across devices.
The metastability itself made systematic current-dependence measurements difficult. Supplementary measurements found a sharp differential-resistance feature near a direct-current bias of 1 nanoampere. The feature shifted to about 1.5 nanoamperes at an in-plane field magnitude of 0.6 millitesla, was sharpest at 20 millikelvin and became very weak by 200 millikelvin.
The authors therefore stop short of assigning a unique microscopic cause to the peak. They explicitly present their explanation as a speculative scenario, leaving the relationship between the field history, frozen domain configuration and resistance unresolved.
Paper data and sources
Original title: Metastable magnetic domains and the anomalous $B_\parallel=0$ resistance peak in twisted double bilayer graphene
Authors: Zhenxiang Gao, Christopher Coleman, Silvia Folk et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text