Quantum at the Scale of Industry: Silicon Spin Qubits Surpassing 99% Fidelity in Fabrication-Ready Architectures
- hashtagworld
- Oct 18
- 4 min read
The future of quantum technologies is no longer defined by laboratory precision alone but by the engineering stability of industrial production. As we approach an era where information is processed at the atomic scale, a decisive question emerges: can quantum systems maintain their performance when transferred from carefully controlled research environments to full-scale industrial fabrication?
A study published in Nature in 2025 “Industry-compatible silicon spin-qubit unit cells exceeding 99% fidelity” provides a direct and compelling answer. Conducted through a collaboration between UNSW Sydney, Diraq, and imec, the research demonstrates that silicon-based spin qubits, fabricated on a 300-millimeter CMOS process line, can operate with an error rate below one percent. This achievement marks not only a scientific milestone but also the beginning of the industrial era of quantum computation.
Silicon Spin Qubits: The Architecture Where Quantum Meets Industry
At the core of quantum information processing lies the qubit, a quantum counterpart of the classical bit. In silicon spin qubits, the qubit is encoded in the spin state of a single electron confined within a quantum dot. The spin can be “up” or “down,” but unlike a classical bit, it can also exist in a quantum superposition of both states simultaneously allowing computation across a continuous probability space.
The choice of silicon is both strategic and physical. When isotopically purified, silicon provides a nearly nuclear-spin-free environment, dramatically extending qubit coherence times. Furthermore, this architecture is intrinsically compatible with conventional CMOS technology, enabling the fabrication of quantum processors using existing semiconductor manufacturing infrastructure. This compatibility is not merely an engineering convenience it is the key to scalability, bridging the precision of quantum mechanics with the reproducibility of modern fabrication.
From Laboratory Precision to Industrial Production
The primary goal of the study was to verify whether silicon spin qubits could retain their quantum performance within industry-standard fabrication processes. Devices were produced in imec’s 300 mm CMOS line in Belgium using planar MOS structures, each containing a double quantum dot coupled to a single-electron transistor (SET) charge detector.
The fabrication process followed standard semiconductor procedures: thermally grown high-quality Si/SiO₂ interfaces, polysilicon gate electrodes, and optimized doping profiles designed to minimize mechanical strain at cryogenic temperatures. The completed devices were cooled to 10 millikelvin using a dilution refrigerator. Experiments were conducted under magnetic fields between 0.66 and 0.7 Tesla, with microwave excitation near 18.6 GHz.
Electron spins were manipulated via Pauli spin blockade and microwave-based electron spin resonance (ESR) pulses. Readout was achieved through charge sensing in the SET, while error characterization was performed using Gate Set Tomography (GST) a high-precision method that decomposes the total error into its physical components. This approach allowed the researchers to identify not only the magnitude but also the specific sources of decoherence within the system.
Experimental Results: Surpassing the Threshold of Fault Tolerance
The findings reveal a level of performance once thought unattainable outside laboratory conditions. Across four tested devices (labeled A through D), the reported fidelities were:
Single-qubit gates: 99.4% to 99.9%
Two-qubit (CZ) gates: 99.0% to 99.6%
State preparation and measurement (SPAM): 99.9%
These results exceed the surface code threshold required for fault-tolerant quantum computation, meaning the devices are sufficiently reliable to support quantum error correction protocols.
Coherence times were equally remarkable: the spin lifetime T_1 reached 9.5 seconds, the Ramsey coherence time T_2^* was measured at 40.6 microseconds, and the Hahn echo coherence T_2^{Hahn} extended to 1.9 milliseconds. These values represent record performance for qubits fabricated through an industrial process, confirming that large-scale manufacturing does not degrade quantum quality when properly optimized.
Noise Mechanisms and Quantum Material Realities
A detailed error analysis identified the dominant source of decoherence as the hyperfine interaction with residual ^{29}\text{Si} nuclei, which constitute approximately 4.7% of natural silicon. These nuclear spins generate stochastic magnetic field fluctuations that induce dephasing in the electron spin.
Secondary contributions stem from spin–orbit coupling and electric-field-induced Stark shifts, both of which introduce sensitivity to charge noise. The study predicts that further isotopic purification, reducing ^{29}\text{Si} concentration from 400 ppm to below 50 ppm, would push fidelities beyond 99.9%.
Importantly, all four devices displayed nearly identical performance metrics, validating the reproducibility and uniformity of the 300 mm CMOS process. This uniformity is essential for scalable quantum systems, where thousands of identical qubits must operate coherently under a shared control architecture.
Industrial Implications: The Age of Fabricated Quantum Systems
This work demonstrates that quantum computation can be realized within the same technological ecosystem that has powered the classical computing revolution. The ability to produce high-fidelity qubits on standard semiconductor process lines bridges the gap between experimental prototypes and manufacturable products.
Such compatibility enables the future integration of quantum and classical electronics on a single chip. Control circuitry, error correction logic, and qubit arrays can coexist within one silicon architecture, paving the way for truly hybrid quantum–classical processors. Moreover, the inherent automation of CMOS manufacturing processes makes it possible to embed calibration and error-tracking mechanisms directly into the fabrication workflow transforming quantum devices into repeatable, measurable, and scalable systems.
Challenges and Path Forward
Despite its success, the study acknowledges several open challenges. The devices still require ultra-low temperatures (~10 mK) for stable operation, and the need for ~0.7 T magnetic fields poses integration constraints for conventional CMOS components. However, these limitations are not structural; they are engineering challenges with defined trajectories toward resolution.
Research directions now include low-field spin resonance, electrically driven spin manipulation without magnetic bias, and on-chip cryogenic control electronics. Combined with deeper isotopic purification and active noise suppression, these approaches will allow future silicon quantum processors to reach unprecedented reliability.
Conclusion: Quantum Becomes Manufacturable
This study marks a defining moment in the evolution of quantum computing. By achieving over 99% fidelity in silicon spin qubits produced on a 300 mm industrial line, it proves that quantum information processing can be manufactured, not merely demonstrated.
From this point forward, progress will hinge less on isolated experimental breakthroughs and more on scalable engineering, automation, and integration. As quantum processors begin to emerge from the same fabrication facilities that once defined classical computing, the boundary between physics and industry dissolves.
Quantum computation has crossed its most critical threshold it is no longer a fragile laboratory phenomenon.
It has become a manufacturable reality.
References
Steinacker, P. et al. Industry-compatible silicon spin-qubit unit cells exceeding 99% fidelity. Nature, Vol. 646, October 2, 2025.
UNSW Quantum Engineering Group; Diraq Pty Ltd.; imec Semiconductor Research, Leuven, Belgium : UNSW Quantum Engineering Group - https://www.unsw.edu.au/engineering/quantum
Diraq Quantum Computing - https://www.diraq.com
imec Semiconductor Research - https://www.imec-int.com
Zenodo Dataset: https://doi.org/10.5281/zenodo.15571656
Comments