Home » Quantum Computing » RIKEN Keeps Silicon Spin Qubits Accurate Running Five at Once

RIKEN Keeps Silicon Spin Qubits Accurate Running Five at Once

RIKEN kept quantum chip accuracy high while running five silicon qubits together. The catch: this device still can’t entangle two qubits well enough to compute anything with them yet.

Macro view of an iridescent silicon microchip wafer, representing the silicon spin qubits in RIKEN's quantum chip research

One of RIKEN’s silicon spin qubits now gets it right about 99,999 times out of every 100,000 tries. That is close to the best any single qubit’s control has ever measured. The hard part was never one qubit, though. It was keeping that number anywhere close when five qubits run at the same time. A team in Japan just showed a way to do it, and they built it out of silicon, the same base material in the chip running your phone.

Nature Physics published the peer-reviewed result on October 15, 2026. The paper comes from RIKEN’s Center for Emergent Matter Science and Center for Quantum Computing, working with researchers at National Taiwan University, ETH Zurich, and TU Delft’s QuTech institute. The team drove five silicon spin qubits through one shared microwave control line and held single-qubit accuracy near 99.9 percent, even with all five firing together. This happened on real hardware, not in a simulation. It chips away at a problem that has quietly capped how large silicon-based quantum chips can grow.

The Problem: Qubits That Interfere With Each Other

Silicon spin qubits store information in the spin of a single electron sitting inside a tiny silicon structure called a quantum dot. Engineers like silicon because the same factories that already mass-produce ordinary computer chips can, in principle, build it too. The catch shows up when you try to control several of these qubits at once.

Send microwave pulses down a shared control line to drive several qubits together. Each pulse nudges the energy levels of every nearby qubit, an effect called the AC Stark shift. That shift throws off the timing, or phase, of each qubit’s quantum state. Before this work, researchers say driving qubits in parallel typically dragged fidelity down from above 99.9 percent to around 99.5 percent. That drop sounds small, but a real quantum computer needs to run this operation millions of times in a row without a single uncorrected slip.

How the Team Fixed the Crosstalk

RIKEN’s Yi-Hsien Wu and Leon Camenzind, the paper’s lead authors, built a calibration method that corrects for the AC Stark shift without the usual cost of testing every possible group of qubits. Here’s roughly how it works:

  1. Measure how each pair of qubits shifts the others’ phase when driven together, instead of testing every possible group of three, four, or five at once.
  2. Use those pairwise measurements to calculate a correction for each qubit’s control pulse, canceling out most of the unwanted phase shift before it happens.
  3. Apply the correction during the actual run, then check the result with randomized benchmarking, a standard way of measuring how often a quantum gate gets the right answer.

That pairwise approach scales much better than testing every combination. For today’s five-qubit device, pairwise calibration needs 20 measurements instead of 26. Push that same math to ten qubits, and pairwise calibration needs about 90 measurements versus more than a thousand for every combination, per Quantum Computing Report’s reading of the paper. That gap only widens as chips get bigger, which is exactly why a shortcut like this matters for scaling.

Setup Single-Qubit Gate Fidelity What It Means
One qubit alone 99.993% to 99.998% Near the best individual results silicon spin qubits have shown
Two or three qubits at once Above 99.99% Barely any drop from running qubits together
All five qubits at once Roughly 99.9% Close to the threshold error-correction codes are expected to need
Two-qubit entangling gate (same device) Below 97% Still far short, a separate problem the paper flags but doesn’t solve

Source: Wu, Camenzind, et al., “Simultaneous High-Fidelity Single-Qubit Gates in a Spin Qubit Array,” Nature Physics and arXiv preprint, and Quantum Computing Report’s coverage of the paper, October 2026.

Why Silicon Spin Qubits Matter for Scaling Up

Idle silicon spin qubits actually lose their quantum information faster than driven ones, the paper notes. Qubits sitting still hold their state for around 10 microseconds, while driven qubits last roughly ten times longer, about 100 microseconds. That’s backwards from what you might expect, and it means a practical error-correction cycle can’t run qubits one at a time. It needs them moving together, which is exactly what this calibration method makes more accurate.

Silicon also carries a manufacturing edge other qubit types don’t share as directly. Diraq and Equinix are already planning an 8-qubit silicon spin system inside a commercial data center, betting that fabs built for ordinary chips can eventually stamp out quantum processors too. A shared control line that still delivers high fidelity trims how much specialized wiring a bigger chip would need, one of the quieter costs between today’s lab demos and a useful machine.

Why It Matters to You

Nothing about this result changes any product you can buy today. RIKEN’s team worked with five qubits in a university lab, not a commercial processor. The next milestone, actually entangling those qubits with high fidelity instead of just controlling them one at a time, hasn’t been shown yet.

The honest long-term payoff sits where most quantum computing promises do: faster drug and materials discovery, better logistics and grid planning, and eventually, decades out, a real risk to today’s encryption standards. None of that is available yet, and this result doesn’t change that timeline dramatically. What it does is remove one practical obstacle, making shared control hardware a little more realistic on a qubit type that already fits existing chip factories. That is the kind of unglamorous progress that tends to matter more over ten years than it does this week.

Where the Evidence Stops

  • This demonstration covers single-qubit gates only. The same device’s two-qubit entangling gate, the operation real algorithms actually need, stays below 97 percent fidelity. The cause is a separate problem called spin-valley mixing, which the paper does not fix.
  • The remaining errors in the five-qubit result come from something the calibration can’t touch: decoherence caused by the extra microwave power it takes to drive five qubits at once. RIKEN’s Leon Camenzind called this “a limitation” in coverage of the paper and said better device design could help.
  • The paper describes the five-qubit fidelity slightly inconsistently. The text and a figure caption put it at about 99.9 percent, but one qubit in the data table comes in a bit under that line, a small discrepancy worth knowing before treating the number as a clean threshold crossing.
  • The team’s path to ten qubits sharing one control line is a projection based on this device’s numbers, not something they’ve actually built and tested.
  • No outside lab has independently reproduced this result. Nature Physics’ peer review is the only outside check so far.

Scaling a quantum computer has never really been one breakthrough away. It is hundreds of small, grinding fixes like this one. Each one closes a gap that would otherwise cap how big a chip can get. Watch for two things next. First, whether RIKEN or a competitor can make this calibration trick work on an entangling gate, not just single-qubit control. Second, whether anyone pushes a shared control line past five qubits without the same accuracy hit. That second number is the one that will tell you whether this scales, or just works once in a single lab.

What is a silicon spin qubit?

It’s a quantum bit built from the spin, or magnetic orientation, of a single electron trapped in a tiny silicon structure. Researchers like it because it can be made with manufacturing techniques closely related to the ones that already build ordinary computer chips.

Does this mean quantum computers are close to breaking encryption?

No. This result improves control over a handful of qubits in a lab, not a working computer. Breaking modern encryption would need millions of highly accurate, error-corrected qubits working together, something no group, including RIKEN’s, has built.

Sources

Related Reading