RIKEN Achieves 99.9% Fidelity Across Five Silicon Qubits
A crucial challenge in scaling silicon spin qubits is that adding more qubits doesn’t just increase computational power; it complicates control. Each additional qubit competes for a shared communication line, leading to interference between neighbors. A team at RIKEN has developed a method to significantly reduce this “overhearing,” and enable accurate parallel control of five qubits simultaneously. Imagine a single loudspeaker instructing a room full of listeners, each responding only to their unique tone. This frequency-selective driving relies on preventing anyone from reacting to another’s instructions.
Scope: Demonstration and Scale
The RIKEN team demonstrated parallel single-qubit control using a five-qubit silicon spin-qubit array, all managed via a single shared microwave line. Their success lies not only in building the device but also in identifying and largely mitigating the primary source of error. These experimental results, published in Nature Physics, were reported by RIKEN.
Key figures from the report include:
- Five silicon spin qubits controlled by one shared microwave drive line.
- Individual single-qubit gate fidelities exceeding 99.99%, with the best nearing 99.999%.
- Fidelity maintained above 99.99% when driving up to three qubits concurrently.
- Approximately 99.9% fidelity achieved across all five qubits driven simultaneously, approaching fault-tolerant thresholds.
- The primary error source identified was the AC Stark shift, compensated using a pairwise phase-calibration protocol.
The Challenge of Parallel Control in Silicon Scaling
Silicon spin qubits are favorable for their compatibility with existing chip fabrication processes, which are already fundamental to modern processors. Information is stored in the spin of a single trapped electron, and this spin is manipulated with microwave pulses, individually tuned for each qubit.
As more qubits are integrated, the problem intensifies. Each qubit requires distinct addressing, which leads to increased wiring complexity and frequency congestion. When one qubit is driven through a shared line, its neighbors are also affected by the electromagnetic field. This unintended exposure causes their energy levels to shift due to the AC Stark effect, subtly altering the phase of their quantum state and degrading the accuracy of operations.
Thus, the challenge isn’t individual qubit malfunction, but rather crosstalk, the unintended interaction between qubits.
How Calibration Mitigates Noise
A straightforward, but impractical, solution would be to calibrate every possible combination of simultaneously operating qubits. For five qubits, this is a combinatorial challenge, and for fifty, it’s unfeasible. The RIKEN team circumvented this by calibrating qubits in pairs, measuring the unwanted phase shift between each pair, and using these pairwise corrections to compensate during parallel operations.
This pairwise approach works because the dominant Stark-shift interference occurs between pairs. Correcting at this level addresses most of the problem without an explosion in the number of required calibrations, forming the scalability argument for this method. As first author Yi-Hsien Wu explained, the method relies on pairwise calibrations rather than unique procedures for every combination.
Achieving fidelities above 99.99% with three qubits and 99.9% with five, all on a shared-line silicon control system, is a significant accomplishment.
The Unresolved Limit
It’s important to acknowledge a remaining limitation. Driving five qubits simultaneously requires more microwave power, which, in turn, leads to drive-induced decoherence. This is a different issue from the AC Stark shift. Though the phase-calibration technique effectively corrects for the Stark shift, it can’t address coherence loss, as this represents the quantum state of the qubit decaying, not merely a phase drift. This explains why the fidelity decreases from over 99.99% with three qubits to approximately 99.9% with five.
Co-first author Leon Camenzind views this as a limitation that can be addressed through improved device design. The results demonstrate five qubits operating with measured fidelities at a specified scale, acknowledging a known bottleneck that remains.
The Trend Towards Offloading Control
RIKEN’s work aligns with a broader industry trend to consolidate control functions onto shared hardware, moving away from dedicated infrastructure for each qubit. For instance, Alice & Bob, working with a different qubit type, have offloaded error-correction decoding to GPUs using NVIDIA’s NVQLink. This reflects a similar drive for efficiency at a different layer of the quantum stack.
The RIKEN research operates at the physical drive line, tackling the question of how many qubits a single microwave channel can effectively serve before crosstalk becomes insurmountable.
Currently, the answer is five qubits, achieving 99.9% fidelity, with future improvements dependent on quieter devices rather than more intricate calibration. Project manager Seigo Tarucha emphasized that reliable parallel control is a critical step toward developing a fault-tolerant silicon quantum machine, a significant, yet understated, long-term goal. True scaling is less about headline numbers and more about maintaining manageable calibration complexity as qubit counts increase.
For further insights into how control systems intersect with real-world security needs, explore QNews’s coverage on the race to develop quantum-safe cryptography: Developers Cut Quantum-Safe Bitcoin Transaction Cost Below $70. Additionally, for those tracking migration timelines, the global PQC deadline picture is already in motion.