Hole spins in germanium quantum dots are attractive for several reasons. The Ge/SiGe heterostructures used to confine them exhibit low charge noise and disorder, and the strong spin-orbit interaction of holes allows all-electrical manipulation of the qubit, eliminating the need for bulky microwave antennas. The same anisotropic physics that enables fast electrical driving, however, also complicates readout. Standard spin readout relies on Pauli spin blockade, in which a two-spin state is converted into a charge state that a nearby sensor can detect. In hole systems, site-dependent spin anisotropies and short relaxation times conspire to corrupt this conversion, and the highest readout fidelity previously reported for holes in germanium hovered around 94 percent, well short of the 99 percent threshold that fault-tolerant quantum computing demands.

The Zurich team, led by Eoin Kelly and Patrick Harvey-Collard, fabricated a linear array of six quantum dots on a strained germanium quantum well, with two single-hole dots coupled to a radiofrequency charge sensor. The sensor is read out through an inductor-capacitor matching circuit that uses a superconducting niobium nitride nanowire inductor fabricated on a separate chip, enabling fast and sensitive reflectometry measurements. A grounded screening layer in the first gate stack protects the radiofrequency excitation from dissipating in the heterostructure, though it introduces a parasitic capacitance of roughly six picofarads that limits the measurement bandwidth, a trade-off the authors note could be relaxed in future devices with less dissipative heterostructures.

At the heart of the work is a careful analysis of spin-to-charge conversion, the step that translates an invisible spin orientation into a measurable charge configuration. Because the two quantum dots have different, highly anisotropic g-tensors, the energy landscape contains two distinct singlet-triplet anticrossings whose positions and sizes shift dramatically with magnetic field orientation. The researchers identified three distinct conversion regimes depending on whether the voltage ramp used during readout is adiabatic or diabatic with respect to each anticrossing: rapid adiabatic passage, slow adiabatic passage, and a regime they call super-slow adiabatic passage, which is adiabatic with respect to both. In the super-slow regime, the down-down spin state maps uniquely onto the only non-blockaded charge state, while the other three two-spin states remain blockaded, a configuration the authors describe as a one-versus-three readout.

Choosing the right regime proved essential. At their operating field orientation, no ramp speed existed that would be simultaneously adiabatic for one anticrossing and diabatic for the other, because the two anticrossings overlap in detuning space. This ruled out the variable-ramp-rate double-ramp tricks used in other systems and made the super-slow regime the natural choice. The team then mapped out how the ramp time required to reach this regime varies with magnetic field direction, finding changes of orders of magnitude across angles. They modeled the data first with the Landau-Zener formula and, where that approximation broke down, with a full time-dependent simulation of a five-state Hamiltonian that incorporated the measured g-tensors of both qubits and an extracted spin-flip tunnelling term. Operating near an in-plane angle of 194 degrees minimized the required ramp time and, with it, the exposure to spin decay during conversion.

Even with the mapping optimized, decay processes lurk at every stage. The blockaded spin states inside the Pauli spin blockade window were found to relax surprisingly quickly, with characteristic times of only about twenty microseconds at low magnetic field, and the team observed that these decay times fall approximately exponentially as the field strength increases, in contrast to the power-law behavior often seen in spin qubits. This makes low magnetic fields doubly attractive: they benefit qubit coherence, as previously known, and they also extend the window available for readout. Without any lifetime-enhancing tricks, the authors calculate that this relaxation alone would cap the readout fidelity at roughly 64 percent under their measurement conditions, which is precisely why hole readout in germanium has been so challenging.

To beat this limit, the researchers employed a double-latched readout scheme. After spin-to-charge conversion, a fast pulse transfers the blocked spin states to a latched charge configuration in which one dot’s electron tunnels to a reservoir on a timescale of about ninety nanoseconds, far faster than the spin decay, while the unblocked singlet state remains locked in place. This converts the fragile spin information into a robust charge difference and enhances the measurement contrast, since the two outcomes now differ by a full unit of charge rather than a redistribution within the double dot. The metastable charge state that encodes the down-down state would still decay through co-tunnelling on a timescale of about two hundred microseconds, which would contribute an estimated 11.4 percent error during a fifty-microsecond integration, so the team added a second latching step at a larger detuning that suppresses co-tunnelling and extends the lifetime beyond two hundred microseconds, cutting that error to under one percent.

With all of these pieces in place, the team measured Rabi oscillations of each qubit while using the other spin as an ancilla, operating at a modest twenty-millitesla field to maximize the blockade lifetime. From ten thousand single shots per point, thresholded at the midpoint between two fitted Gaussian distributions, they extracted an average SPAM fidelity of 97.0 percent, with Rabi visibilities of about 94 percent for each qubit. The error budget is now dominated by two culprits rather than by fundamental physics: imperfect initialization of the down-down state, contributing about 1.7 percent, and a limited signal-to-noise ratio of 2.235, contributing between one and 1.3 percent. Both, the authors argue, are engineering limitations that can be improved, which suggests that SPAM fidelities above 99 percent should be within reach on this platform.

Beyond the headline number, the study delivers something arguably more valuable: a complete, quantitative accounting of where readout errors come from in a hole-spin system, from Landau-Zener physics at anticrossings to spin relaxation in the blockade window to charge decay during latching. The tune-up strategy relies only on uniform magnetic field and voltage conditions and should therefore extend to larger qubit arrays as well as smaller devices. The authors also note that latched readout is not fundamentally slower than conventional blockade readout, since the signal is triggered deterministically at the start of the integration window and is enhanced by the larger charge response of the sensor. As germanium quantum dot processors scale up, this kind of error-by-error optimization of the measurement chain may prove just as important as improvements in gate operations, because a quantum computer is only as good as its ability to read out its own answers.

Subject of Research: Error mechanisms and high-fidelity single-shot readout of hole spin qubits in germanium double quantum dots

Article Title: Identification and mitigation of errors in hole spin qubit readout

Article References: Kelly, E. G., Massai, L., Hetényi, B., Pita-Vidal, M., Orekhov, A., Carlsson, C., Seidler, I., Tsoukalas, K., Sommer, L., Aldeghi, M., Bedell, S. W., Paredes, S., Schupp, F. J., Mergenthaler, M., Fuhrer, A., Salis, G., & Harvey-Collard, P. (2026). Identification and mitigation of errors in hole spin qubit readout. Nature Electronics. https://doi.org/10.1038/s41928-026-01705-1

Image Credits: AI Generated

DOI: 10.1038/s41928-026-01705-1

Keywords: spin qubits, germanium, quantum dots, Pauli spin blockade, spin-to-charge conversion, readout fidelity, SPAM, latching readout, spin-orbit interaction, Landau-Zener, radiofrequency reflectometry, quantum computing