Control-Noise Trade-offs in Industrially Integrated ${}^{28}\mathrm{Si}/\mathrm{SiGe}$ Spin-1/2 Qubits

Leon C. Camenzind; Ik Kyeong Jin; Yi-Hsien Wu; Akito Noiri; Kenta Takeda; Takashi Nakajima; Takashi Kobayashi; Giordano Scappucci; Seigo Tarucha

AI summary · arXiv: 2610.41890v1 · quant-ph

Control-Noise Trade-offs in Industrially Integrated ${}^{28}\mathrm{Si}/\mathrm{SiGe}$ Spin-1/2 Qubits

Leon C. Camenzind, Ik Kyeong Jin, Yi-Hsien Wu, Akito Noiri, Kenta Takeda, Takashi Nakajima, Takashi Kobayashi, Giordano Scappucci,

and 1 more author

Seigo Tarucha

Original paper Submitted to arXiv on 30 Oct. 2026 28 pages, 4 main figures and 5 extended data figures License: NONEXCLUSIVE-DISTRIB 1.0 Also in: fr

How to read this summary? Illustrated guide

The essentials

Industrially fabricated silicon spin qubits achieve near-99.99% primitive gate fidelity, while micromagnet placement trades reduced charge-noise sensitivity for less efficient control.

Question

How does industrial fabrication and dense device integration affect control, coherence, noise, and qubit variability in silicon spin qubits? The study focuses on whether micromagnet placement in BEOL routing changes the balance between noise suppression and electrically driven spin resonance.

Method

The authors compare three-qubit devices made with Intel’s 300 mm process against a five-qubit RIKEN reference, using heterostructures with approximately 800 ppm residual 29Si. They measure Ramsey, Hahn-echo, and CPMG coherence, track frequency-noise spectra and inter-qubit correlations, and use randomized benchmarking to characterize primitive π/2 gates across Rabi frequencies.

Findings

Result Intel qubits show longer coherence, including median Hahn-echo coherence of about 200 μs versus about 70 μs for RIKEN, and 3.88 ± 0.09 ms versus 1.49 ± 0.31 ms under 1024-pulse CPMG.
See the passage in the paper
Under Hahn echo, the Intel (RIKEN) qubits exhibit median coherence times $T_{2}^{\mathrm{echo}}\approx 200~\mu\mathrm{s}$ ($\approx 70~\mu\mathrm{s}$), consistent with reduced charge-noise susceptibility arising from the weaker longitudinal magnetic-field gradient in the industrial devices.
Authors’ interpretation Intel devices exhibit weaker qubit-frequency noise and inter-qubit correlations, which the authors attribute to lower charge-noise susceptibility and a greater contribution from uncorrelated nuclear-spin noise.
See the passage in the paper
The reduced correlation magnitudes therefore support the interpretation that uncorrelated residual nuclear-spin noise dominates the low-frequency spectrum, masking weaker charge-noise correlations.
Result The Intel devices reach primitive π/2-gate fidelities near 99.99%, compared with 99.999% for the RIKEN reference.
See the passage in the paper
Despite the weaker control gradient, the Intel devices achieve primitive $\pi/2$-gate fidelities near $99.99\%$, while the RIKEN reference reaches $99.999\%$.
Authors’ interpretation Intel’s weaker transverse gradient requires higher microwave drive and is associated with drive-dependent errors becoming relevant at lower Rabi frequencies.
See the passage in the paper
These results show that the weaker $\nabla B_{\perp}$ in the present Intel integration moves the onset of drive-dependent gate errors to lower Rabi frequencies, thereby limiting the peak single-qubit fidelity.
Result The measured Intel devices have a narrower spread in required source microwave powers and in some driven-control quality factors, indicating reduced qubit-to-qubit variability in those characteristics.
See the passage in the paper
The Intel devices also show a narrower spread in several measured control-related operating parameters.

Limitations

The Intel-versus-RIKEN comparison combines differences in fabrication context, device integration, and heterostructure details, so the observed contrast cannot be assigned solely to industrial fabrication.
See the passage in the paper
Although the heterostructures are nominally similar and isotopically enriched to the same residual 29Si concentration, the devices differ both in fabrication process and in the integration of qubits, micromagnets, and control wiring.
Reported microwave powers are nominal source-calibrated values rather than direct measurements of power delivered to each qubit.
See the passage in the paper
A full in-situ calibration of the power delivered to each qubit is outside the scope of this work.
The measurements do not distinguish whether the weaker Ramsey response to microwave irradiation reflects frequency-range differences or irradiation-time-dependent effects.
See the passage in the paper
The measurements do not distinguish between these two effects.

Why it matters

The authors argue that industrially fabricated 300 mm silicon spin qubits can achieve very high gate fidelity while offering improved reproducibility, but dense routing and micromagnet placement create a control–noise trade-off. Scaling will require co-designing the routing architecture and magnet placement to preserve transverse gradients for efficient control while managing charge-noise sensitivity.

Written by AI from the paper. Passages are checked word for word against the text we analysed; the rest may contain errors.

AI-generated keywords: Silicon spin qubits industrial fabrication qubit coherence noise correlations gate fidelity

Read further

Comprehensive Summary
Layman's Summary
Blog article
Original abstract

The paper examines how industrial-scale fabrication and dense device integration affect control performance and noise in silicon spin qubits. The authors compare three-qubit devices fabricated using Intel’s industry-compatible 300 mm process with a high-performance five-qubit reference device fabricated at RIKEN. Both use isotopically enriched silicon–silicon-germanium heterostructures containing about 800 parts per million residual silicon-29. However, they differ in layer stacks, gate layouts, well depth, and micromagnet placement. The study asks how these design choices affect qubit coherence, noise correlations, control efficiency, and single-qubit gate fidelity. The researchers combine device and micromagnet characterization with coherence measurements, noise-spectrum analysis, inter-qubit correlation studies, and randomized benchmarking. They measure Ramsey, Hahn-echo, and CPMG coherence, along with qubit-frequency fluctuations tracked using repeated Ramsey and CPMG sequences. For control, they use electrically driven spin resonance with micromagnets and measure primitive pi/2-gate fidelity across a range of drive rates. The reported microwave powers are nominal source-calibrated values, not direct measurements of the power delivered to each qubit. In the Intel devices, the micromagnets are integrated into the contact-interlayer dielectric alongside the back-end routing. They are farther from the quantum well than in the RIKEN device, a difference associated with weaker magnetic-field gradients. Simulations estimate transverse driving gradients of about 0.4–0.5 millitesla per nanometre for Intel, compared with about 1.15 for RIKEN. Consistent with less efficient electrical spin resonance, Intel qubits required a median nominal microwave power 7.7 decibels higher to reach a 1 MHz Rabi frequency. The narrower range of source powers required across the Intel array, together with some driven-control quality factors, suggests less qubit-to-qubit variability in these measurements. The weaker longitudinal gradient in the Intel devices reduces the conversion of electric-field fluctuations into qubit-frequency noise. Intel qubits generally showed longer Ramsey coherence, particularly at longer measurement integration times, and a median Hahn-echo coherence of about 200 microseconds, compared with about 70 microseconds for RIKEN. With 1,024 CPMG pulses, the two measured Intel qubits reached a median coherence of 3.88 milliseconds, compared with 1.49 milliseconds across five RIKEN qubits. Both platforms showed broadly similar, near-one-half scaling of CPMG coherence with pulse number, consistent with 1/f-like noise at higher frequencies. Sensor-referred measurements indicated comparable, and slightly higher, local charge noise in Intel. This supports the interpretation that Intel’s lower qubit-frequency noise reflects reduced susceptibility rather than a quieter charge environment. The low-frequency noise differed between the platforms: RIKEN showed stronger correlations between nearby qubits and features attributed to local oxide two-level fluctuators. Intel showed much weaker inter-qubit correlations and excess low-frequency noise attributed to residual nuclear spins, including silicon-29. The authors consider this attribution consistent with occupation-dependent coherence and noise measurements. For gate performance, Intel achieved primitive pi/2 fidelities near 99.99%, while RIKEN reached about 99.999%. In both platforms, errors initially decreased as gates became faster but rose at higher drive. The increase began at lower Rabi frequencies in Intel, consistent with its weaker driving gradient and greater microwave-power requirements. Measurements suggest that increased frequency noise under microwave irradiation may contribute to this loss; however, the measured microwave-induced detuning shifts alone were too small to explain it. The authors conclude that very high fidelity is achievable in industrially fabricated devices, but micromagnet placement and routing must be co-designed. Weaker gradients can improve resilience to charge noise while reducing control efficiency.

What Changes When Spin Qubits Move to an Industrial Process?

Building a useful quantum computer will require more than making individual qubits work well. Devices must also be fabricated reliably and packed with the wiring and components needed to control them. In “Control-Noise Trade-offs in Industrially Integrated ${}^{28}\mathrm{Si}/\mathrm{SiGe}$ Spin-1/2 Qubits,” Leon C. Camenzind, Ik Kyeong Jin, Yi-Hsien Wu, Akito Noiri, Kenta Takeda, Takashi Nakajima et al. investigate how industrial-scale fabrication and device integration affect silicon spin-qubit control and noise.

The researchers compare three-qubit devices made using Intel’s industry-compatible 300 mm process with a high-performance five-qubit reference device fabricated at RIKEN. Both platforms use isotopically enriched silicon–silicon-germanium materials, with about 800 parts per million of residual silicon-29. But their layer stacks, gate layouts, quantum-well depths, and micromagnet placements differ. Those engineering choices provide a way to examine the balance between quiet qubits and efficient control.

A Trade-off Built into the Device

Spin qubits are controlled electrically with help from micromagnets, which create magnetic-field gradients. The gradients let electric signals rotate a qubit’s spin, but they also influence how strongly electric-field fluctuations shift its frequency. This can make a qubit more vulnerable to charge noise.

In Intel’s devices, micromagnets sit in the contact-interlayer dielectric alongside back-end routing. They are farther from the quantum well than the micromagnets in the RIKEN device. Simulations estimate transverse driving gradients of roughly 0.4–0.5 millitesla per nanometre for Intel, compared with about 1.15 for RIKEN.

The practical effect is less efficient electrical spin resonance in the Intel devices. Reaching a 1 MHz Rabi frequency required a median nominal microwave power 7.7 decibels higher than in the RIKEN device. These are source-calibrated power values, not direct measurements of the power delivered to each qubit. Even so, the result highlights a design challenge: a more integrated layout can place constraints on the magnetic gradients that make control easy.

Longer Coherence, Despite Comparable Charge Noise

The weaker longitudinal gradient in the Intel devices reduces the conversion of electric-field fluctuations into qubit-frequency noise. That difference appears in the coherence measurements. The researchers measured Ramsey, Hahn-echo, and CPMG coherence, as well as frequency fluctuations using repeated Ramsey and CPMG sequences.

Intel qubits generally showed longer Ramsey coherence, especially at longer integration times, and a median Hahn-echo coherence of about 200 microseconds. The RIKEN device reached about 70 microseconds. With 1,024 CPMG pulses, the two measured Intel qubits reached a median coherence of 3.88 milliseconds, compared with 1.49 milliseconds across five RIKEN qubits.

Both platforms showed broadly similar, near-one-half scaling of CPMG coherence with pulse number. This pattern is consistent with noise that has a 1/f-like spectrum at higher frequencies. Yet sensor-referred measurements found comparable—and slightly higher—local charge noise in Intel. The longer coherence is therefore not evidence of a quieter charge environment. Instead, it supports the interpretation that Intel’s qubits are less sensitive to charge fluctuations because of their weaker gradients.

Different Sources of Low-Frequency Noise

The platforms also differed in how noise was shared between neighboring qubits. RIKEN showed stronger inter-qubit correlations and features attributed to local oxide two-level fluctuators. Intel showed much weaker correlations, alongside excess low-frequency noise attributed to residual nuclear spins, including silicon-29. The researchers considered that explanation consistent with measurements of coherence and noise under different occupation conditions.

This distinction matters for scaling. Correlated disturbances can affect several nearby qubits at once, while local noise may be more isolated. But the Intel results also show that isotopic enrichment does not remove every nuclear-spin contribution: even a small residual concentration can remain relevant at low frequencies.

High Fidelity, with a Control-Speed Limit

For gate tests, the researchers used electrically driven spin resonance and randomized benchmarking to measure primitive pi/2-gate fidelity across a range of drive rates. Intel reached fidelities near 99.99%, while RIKEN reached about 99.999%. Both are very high, though the reference device performed better in these measurements.

In both platforms, errors initially fell as gates became faster, then rose at higher drive. That increase began at lower Rabi frequencies in Intel, consistent with its weaker driving gradient and greater microwave-power requirements. Measurements suggest that microwave irradiation may increase frequency noise, although the observed microwave-induced detuning shifts were too small on their own to explain the loss in fidelity.

The narrower range of source powers needed across the Intel array, together with some driven-control quality factors, points to less qubit-to-qubit variability in these measurements. This is a useful counterpoint to the higher power requirement: control may be less efficient overall, while appearing more consistent across the tested devices.

Designing for Scale

The central message is not that one platform wins on every measure. Industrially fabricated devices can deliver millisecond-scale coherence and primitive gate fidelities near 99.99%. At the same time, micromagnet placement and routing shape both noise sensitivity and control efficiency. Weaker gradients can protect qubits from charge-related frequency shifts, but they demand more drive power and can limit how quickly high-quality gates are achieved.

For future devices, the challenge is to co-design micromagnets, wiring, and the quantum-well structure rather than optimize them separately. The study shows why integration is not merely a packaging problem: the physical layout becomes part of the qubit’s control and noise environment.

Created on 03 Nov. 2026

Was this synthesis useful?

Score: 0

Ask questions about this paper to our AI assistant

How to chat with this paper? Illustrated guide

Compare with the PDF

You can also chat with multiple papers at once here.

AI assistant instructions?

Tools for this paper

Explain it differently 3 free rewrites left today

Save to a collection Group papers in collections to compare them, chat with them and export them. Create a free account Log in

Export BibTeX · RIS (Zotero, Mendeley, EndNote) · Obsidian (Markdown) · Notion (Markdown)

Advanced tools Premium

For researchers who follow a field closely: citation alerts, journal-club slides, version comparison and Zotero export. Discover Premium

Zotero

Citation alerts Citation alerts guide

Who works on it? The people and institutions who publish on its topic

Weekly letter Presented in "Quantum physics this week" of 2 November 2026. Quantum physics this week: Making noise, measurement and matter work for quantum devices

Journal club Premium+ Slides to present it Propose for the journal club

Versions What changed between versions

Talk about it in public Premium+ Press kit: release, questions, hype check Press kit guide

Disclaimer: The AI-based summarization tool and virtual assistant provided on this website may not always provide accurate and complete summaries or responses. We encourage you to carefully review and evaluate the generated content to ensure its quality and relevance to your needs.

Control-Noise Trade-offs in Industrially Integrated ${}^{28}\mathrm{Si}/\mathrm{SiGe}$ Spin-1/2 Qubits

The essentials

Question

Method

Findings

Limitations

Why it matters

Read further

Definitions

What Changes When Spin Qubits Move to an Industrial Process?

A Trade-off Built into the Device

Longer Coherence, Despite Comparable Charge Noise

Different Sources of Low-Frequency Noise

High Fidelity, with a Control-Speed Limit

Designing for Scale

Ask questions about this paper to our AI assistant

Tools for this paper

Advanced tools Premium

Papers around this one

Closest papers

Prior work it builds on

Go further

Cite this paper (BibTeX)