A proposed semiconductor qubit could switch between a fast-control mode and a longer-lived mode by moving a single hole between two layers, according to a new numerical preprint. The calculation describes a bilayer Ge device with SiGeSn barriers in which the hole is heavy-hole-like in one layer and light-hole-like in the other. The authors say that combination could allow rapid manipulation followed by quieter idling at a charge-noise sweet spot.
The strongest operating point in the model was not the point where the heavy- and light-hole characters were mixed most evenly. A lower-voltage sweet spot, with about 25% heavy-hole weight, was judged preferable to the 50% mixed state at a gate voltage of 602.7 millivolts because it had smaller modeled sensitivity to gate-voltage noise and longer calculated dephasing time. Dephasing is the loss of a qubit’s phase relationship, which makes its state harder to preserve.
Two layers, two kinds of control
The device is built around two strained Ge wells. The narrower well is 6 nanometres wide and is designed to favor heavy-hole character; the wider well is 20 nanometres wide and can host light-hole character. Both wells are modeled with 0.25% tensile strain. In the proposed operation, changing the plunger-gate voltage shifts the hole between the layers and changes which character dominates the ground state.
In the calculations, the ground state is light-hole-like at a gate voltage of 300 millivolts and heavy-hole-like at 700 millivolts. Across the transition, the in-plane magnetic response varies nonmonotonically and reaches a peak at 602.7 millivolts, while the out-of-plane g-factor remains around 2. In this calculation, the g-factor is the measure of the qubit’s response to a magnetic field.
The calculation links that peak to a resonance between subbands, or quantized energy levels, under in-plane confinement. At 602.7 millivolts, the modeled state is fully hybridized, with a 50% heavy-hole fraction, and the relevant heavy-hole/light-hole subband gap is about 2.2 millielectronvolts. The proposed mechanism therefore depends on a narrow region where the two types of state strongly mix.
Fast gates are possible on paper
Near the resonance, the model uses an orbital splitting of 54.6 microelectronvolts to estimate vertical shuttling in roughly 100 picoseconds. Including the rest of the hopping operation gives an estimated single-qubit gate time of 3.70 to 8.64 nanoseconds for an in-plane magnetic field of 10 to 25 millitesla. The authors describe this as a rough estimate that depends on tunnel coupling and on how precisely voltage pulses can be shaped.
The broader calculation predicts Rabi frequencies on the order of 100 megahertz at a sweet spot, comparable to the light-hole regime, alongside coherence on the order of 100 microseconds. The reported coherence is more than ten times longer than in the faster-control comparison used in the study. In practical terms, the model is aiming to keep electrical manipulation quick without giving up as much time for the qubit to retain its state.
The charge-noise calculation treats nearby fluctuators collectively as one global, quasistatic fluctuation of the plunger-gate voltage rather than modeling individual fluctuators. Under that restricted noise model, the calculated dephasing-time curve has two peaks: 217.6 microseconds at the lower-voltage peak and 4.7 microseconds at the right-hand peak. That large difference is why the lower-voltage sweet spot, rather than the 50% mixed resonance itself, becomes the preferred operating point in the paper.
A promising design still needs a device
These results come from one modeled quantum-dot design, not measurements from a fabricated qubit. The calculation uses a finite-difference solution of the six-band hole Hamiltonian expanded in the lowest 150 heavy-hole and 150 light-hole states, with convergence reported within 1%. The study therefore estimates how the proposed device might behave; it does not demonstrate its operation.
The material description is another important uncertainty. The authors note that established Luttinger parameters for SiGeSn alloys are missing, limiting more accurate modeling. They estimate that realizing the design would require about ±1% control of the silicon concentration and an in-plane strain between roughly 0.20% and 0.29%.
The noise numbers should be read as comparisons within the chosen model, not as guaranteed device lifetimes. The authors caution that the absolute dephasing prediction can change with noise amplitude and with the quality or purity of the heterostructure. Their analysis represents nearby fluctuators as a collective gate-voltage disturbance, so it does not capture every possible detail of charge noise.
The document is an arXiv version 1 preprint dated 26 August 2026. The proposal still needs experimental testing of whether the layers and strain can be made with enough control, and whether the predicted switching, fast electrical driving, noise sweet spots, coherence and shuttling performance survive in a real device.
Paper data and sources
Original title: Switchable heavy-hole/light-hole spin qubit
Authors: Zoltán György, Dmitry Miserev, Jelena Klinovaja, Daniel Loss
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text