Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2405.04524 · Physical review. B./Physical review. B

Neural network based deep learning analysis of semiconductor quantum dot qubits for automated control

Jacob Taylor, S. Das Sarma

Published Jan 3, 2025• 16 citations
arXivHTMLPDF

Abstract

In this paper, we introduce a methodology that employs machine learning, specifically convolutional neural networks (CNNs), to discern the disorder landscape in the parameters of the disordered extended Hubbard model underlying the semiconductor quantum dot spin qubit architectures. Our approach, which we demonstrate with numerically simulated devices, takes advantage of currently used experimentally obtainable charge stability diagrams from neighboring quantum dot pairs, enabling the CNN to accurately identify disorder in each parameter of the extended Hubbard model. Remarkably, our CNN can process site-specific disorder in Hubbard parameters, including variations in hopping constants, on-site potentials (gate voltages), and both intrasite and intersite Coulomb terms. This advancement facilitates the prediction of spatially dependent disorder across all parameters simultaneously with high accuracy (${R}^{2}>0.994$) and fewer parameter constraints, marking a significant improvement over previous methods that were focused only on analyzing on-site potentials at low coupling. Furthermore, our approach allows for the tuning of five or more quantum dots at a time, effectively addressing the often-overlooked issue of crosstalk. Not only does our method streamline the tuning process, potentially enabling fully automated adjustments, but it also introduces a ``no trust'' verification method to rigorously validate the neural network's predictions. Ultimately, this paper aims to lay the groundwork for generalizing our method to tackle a broad spectrum of physical problems.

DOI
10.1103/physrevb.111.035301

Related papers

Predicting spin-orbit coupling in hole spin qubit arrays with vision-transformer-based neural networks on a generalized Hubbard model

Jacob R. Taylor, Katharina Laubscher, S. Das Sarma

We introduce a neural-network-based machine learning method to predict the effective spin-orbit coupling (SOC) strength in hole quantum dot arrays from standard charge stability diagrams. Specifically, we study a $2\times 2$ Ge hole quantum dot array described by a generalized spin-orbit coupled Hubbard model that incorporates random site- and bond-dependent disorder in all system parameters, including onsite potentials, Coulomb interaction strengths, interdot tunneling amplitudes, as well as the direction and angle of the SOC-induced spin rotations accompanying interdot tunneling. We train the neural network on numerically simulated charge stability diagrams from nearest-neighbor pairs of quantum dots for different chemical potentials and out-of-plane magnetic fields, and show that this enables us to predict the SOC-induced spin-flip tunneling amplitudes -- and, thus, the effective SOC strength -- with high fidelity ($R^2\approx 0.94$) even when all other Hubbard model parameters are unknown. Furthermore, our neural network can also predict the other Hubbard model parameters with high fidelity, demonstrating that neural-network-based approaches can be a powerful tool for the automated characterization of hole spin qubit arrays.

arXiv:2604.05052•Physical review. B./Physical review. B•Oct 27, 2026•1 citation

Large Scale Optimization of Disordered Hubbard Models through Tensor and Neural Networks

Jacob R. Taylor, S. Das Sarma

We theoretically demonstrate a practical method for tuning randomly disordered 2D quantum-dot grids underlying spin qubit platforms using vision-based neural networks trained on tensor-network generated charge-stability data. We show that a simulatable local $3\times 3$ window already contains sufficient information to tune the central dot within a much larger array, thereby validating a sliding-window approach in which one tunes a local region and then translates that window across the lattice to calibrate a larger device. This avoids the computationally intractable necessity for obtaining the ground states for large systems with exponentially large Hilbert space. For the experimentally relevant case where only the on-site disorder is unknown, the neural network predicts the relevant parameters with very high fidelity in the $3\times 3$ setting [$R^2 >0.99$], and after fine tuning on only a small number of larger-device samples, it retains high accuracy for the central dot of a $5\times 5$ plaquette [$R^2\approx 0.98$]. When all the dots parameters are treated as unknown, prediction of the on-site disorder remains robust [$R^2>0.9$ for both $3\times 3$ and $5\times 5$], although the remaining parameters are substantially more difficult to infer from the same charge-stability data. This shows that the most practically important disorder parameter for tuning can still be inferred reliably even in the fully disordered setting for the computationally difficult 5x5 arrays.

arXiv:2604.18711•Apr 20, 2026•0 citations

Machine learning techniques for state recognition and auto-tuning in quantum dots

Sandesh S. Kalantre, Justyna P. Zwolak, Stephen Ragole +4 more

Abstract Recent progress in building large-scale quantum devices for exploring quantum computing and simulation has relied upon effective tools for achieving and maintaining good experimental parameters, i.e., tuning up devices. In many cases, including quantum dot-based architectures, the parameter space grows substantially with the number of qubits, and may become a limit to scalability. Fortunately, machine learning techniques for pattern recognition and image classification, using so-called deep neural networks, have shown surprising successes for computer-aided understanding of complex systems. We propose a new paradigm for fully automated experimental initialization through a closed-loop system relying on machine learning and optimization techniques. We use deep convolutional neural networks to characterize states and charge configurations of semiconductor quantum dot arrays when only measurements of a current−voltage characteristic of transport are available. For simplicity, we model a semiconductor nanowire connected to leads and capacitively coupled to depletion gates using the Thomas−Fermi approximation and Coulomb blockade physics. We then generate labeled training data for the neural networks, and find at least 90 % accuracy for charge and state identification for single and double dots. Using these characterization networks, we can then optimize the parameter space to achieve a desired configuration of the array, a technique we call “auto-tuning”. Finally, we show how such techniques can be implemented in an experimental setting by applying our approach to an experimental dataset, and outline further problems in this domain, from using charge sensing data to extensions to full one- and two-dimensional arrays, that can be tackled with machine learning.

arXiv:1712.04914•npj Quantum Information•Jan 14, 2019•122 citations

Machine Learning for Charge State Characterization of Isolated Double Quantum Dots

Hyma Harish Vallabhapurapu, Marco Candido, Krishna Choudhary +7 more

Scaling semiconductor quantum dot arrays toward fault-tolerant quantum computing requires efficient tuneup of spin qubits, a process that depends on the analysis of charge stability maps (CSMs) and remains largely manual. While machine learning has been widely applied to CSM analysis in reservoir-coupled devices, automated tuning in the increasingly important isolated-mode regime has received limited attention. In isolated-mode CSMs, charge transitions appear as near-vertical lines, making them well suited to compact, task-specific models. We present two convolutional neural networks with fewer than one million parameters, trained on CSMs collected from 32 silicon metal-oxide-semiconductor (SiMOS) double-quantum-dot devices measured at approximately 1 K using an automated cryogenic probing system. Sixteen devices were used for training and sixteen were held out to evaluate cross-device generalization against hand-labeled ground truth. CSMClassifier identifies charge instability and sensor artifacts, achieving 94% macro-averaged accuracy across three quality classes on 2,407 held-out images. ChargeLineNet localizes charge-transition lines and determines electron occupancy, achieving 95.3% exact line-count accuracy on 1,131 held-out images. Combined into a single pipeline, the models correctly determine electron occupancy for 93.8% of clean held-out images. Pre-training on synthetic images substantially improves label efficiency. Fine-tuning the pre-trained model on limited experimental data maintains over 90% accuracy, whereas training from scratch degrades significantly under the same conditions. Together, the two models occupy only 6.5 MB and process images in less than 60 ms on standard laboratory hardware, demonstrating a practical path toward scalable, automated characterization and tuneup of quantum-dot devices.

arXiv:2607.20871•Jul 23, 2026•0 citations

Automatic charge state tuning of 300 mm silicon quantum dots using neural network segmentation of charge stability diagram

Peter Samaha, Amine Torki, Ysaline Renaud +4 more

Abstract Tuning of gate-defined semiconductor quantum dots (QDs) is a major bottleneck for scaling spin-qubit technologies. We present a deep learning driven, semantic-segmentation pipeline that performs charge auto-tuning by locating transition lines in full charge stability diagrams (CSDs) and returns gate voltage targets for the single charge regime. We assemble and manually annotate a large, heterogeneous dataset of 1015 experimental CSDs measured from fully depleted silicon-on-insulator (FD-SOI) QD devices, spanning nine design geometries, multiple wafers, and fabrication runs. A U-Net style convolutional neural network with a MobileNetV2 encoder is trained and validated through five-fold group cross-validation. Our model achieves an overall offline tuning success of 80.0% in locating the single-charge regime, with peak performance exceeding 88% for some designs. We analyze dominant failure modes and propose targeted mitigations. Finally, wide-range diagram segmentation also naturally enables scalable physics-based feature extraction that can feed back to fabrication and design workflows and outline a roadmap for real-time integration in a cryogenic wafer prober. Overall, our results show that neural network based wide-diagram segmentation is a practical step toward automated, high-throughput charge tuning for silicon QD qubits.

arXiv:2604.13662•Machine Learning Science and Technology•Jun 12, 2026•2 citations

On Scalable Integrated Charge State Tuning for Semiconductor Quantum Dot Devices

Fabian Hader

Semiconductor quantum computing leverages quantum dots as qubits, where achieving high qubit fidelity is one essential requirement for scalability. The formation of qubits and associated sensor dots necessitates meticulous iterative voltage adjustments, a tuning process that is ordinarily executed manually by researchers. While various automated approaches have been proposed, only a few address scalability in the context of a system with a million qubits or consider shifting key data analysis tasks into the cryostat. The latter is necessary to alleviate wiring constraints and to minimize data transmission to room temperature electronics. This thesis explores scalable solutions for automated in-cryostat data analysis, identifying charge transition detection as a pivotal task. A simulation model is developed to generate realistic charge stability diagrams with known ground truth. The model’s output is benchmarked against experimental data to verify fidelity and diversity. Utilizing this model, a comprehensive evaluation of charge transition detection methods is conducted, comparing classical algorithms and a range of machine learning approaches. In consideration of energy-efficient integration, this study includes an initial investigation into reducing the complexity of convolutional neural networks. Subsequent to their optimization using simulated data, the approaches are benchmarked on both simulated and experimental data from diverse qubit samples. When applied to contemporary data qualities, evidence is presented that classical approaches are inadequate in terms of robustness and accuracy while numerous machine learning models demonstrate compelling performance. Among the latter, convolutional neural networks like U-Net are identified to be particularly eligible due to their advantageous structure for efficient hardware integration. Although the U-Net architecture already demonstrates a strong potential for network size reduction under current data quality constraints, a cryogenic hardware integration remains a significant challenge. Consequently, a subsequent study explores if enhanced sensor dot tuning and sensor compensation reduce the complexity of charge transition analysis and if they enable the utilization of ray-based detection methodologies. The findings indicate that particularly sensor compensation has a substantial impact on the data quality and allows a streamlining of the analytical process. However, achieving robust ray-based analysis remains feasible only if consistently reaching the highest observed measurement quality and utilizing machine learning. Even in this case, classical methods and a proposed averaging and thresholding circuit fall short of the requisite detection accuracy. Only with further improvements toward optimal data quality and additional enhancements to the proposed circuit, an integrated solution without machine learning appears attainable. Finally, the energy efficiency of scaled-down neural networks is assessed using state-of-the art hardware accelerators. The findings suggest that the cryogenic integration of machine learning-based charge transition detection is feasible, thus propelling the development of scalable and energy-efficient quantum control architectures.

•Jan 1, 2025•0 citations
In this paper, we introduce a methodology that employs machine learning, specifically convolutional neural networks (CNNs), to discern the disorder landscape in the parameters of the disordered extended Hubbard model underlying the semiconductor quantum dot spin qubit architectures. Our approach, which we demonstrate with numerically simulated devices, takes advantage of currently used experimentally obtainable charge stability diagrams from neighboring quantum dot pairs, enabling the CNN to accurately identify disorder in each parameter of the extended Hubbard model. Remarkably, our CNN can process site-specific disorder in Hubbard parameters, including variations in hopping constants, on-site potentials (gate voltages), and both intrasite and intersite Coulomb terms. This advancement facilitates the prediction of spatially dependent disorder across all parameters simultaneously with high accuracy (R2>0.994{R}^{2}>0.994R2>0.994) and fewer parameter constraints, marking a significant improvement over previous methods that were focused only on analyzing on-site potentials at low coupling. Furthermore, our approach allows for the tuning of five or more quantum dots at a time, effectively addressing the often-overlooked issue of crosstalk. Not only does our method streamline the tuning process, potentially enabling fully automated adjustments, but it also introduces a ``no trust'' verification method to rigorously validate the neural network's predictions. Ultimately, this paper aims to lay the groundwork for generalizing our method to tackle a broad spectrum of physical problems.
We introduce a neural-network-based machine learning method to predict the effective spin-orbit coupling (SOC) strength in hole quantum dot arrays from standard charge stability diagrams. Specifically, we study a 2×22\times 22×2 Ge hole quantum dot array described by a generalized spin-orbit coupled Hubbard model that incorporates random site- and bond-dependent disorder in all system parameters, including onsite potentials, Coulomb interaction strengths, interdot tunneling amplitudes, as well as the direction and angle of the SOC-induced spin rotations accompanying interdot tunneling. We train the neural network on numerically simulated charge stability diagrams from nearest-neighbor pairs of quantum dots for different chemical potentials and out-of-plane magnetic fields, and show that this enables us to predict the SOC-induced spin-flip tunneling amplitudes -- and, thus, the effective SOC strength -- with high fidelity (R2≈0.94R^2\approx 0.94R2≈0.94) even when all other Hubbard model parameters are unknown. Furthermore, our neural network can also predict the other Hubbard model parameters with high fidelity, demonstrating that neural-network-based approaches can be a powerful tool for the automated characterization of hole spin qubit arrays.
We theoretically demonstrate a practical method for tuning randomly disordered 2D quantum-dot grids underlying spin qubit platforms using vision-based neural networks trained on tensor-network generated charge-stability data. We show that a simulatable local 3×33\times 33×3 window already contains sufficient information to tune the central dot within a much larger array, thereby validating a sliding-window approach in which one tunes a local region and then translates that window across the lattice to calibrate a larger device. This avoids the computationally intractable necessity for obtaining the ground states for large systems with exponentially large Hilbert space. For the experimentally relevant case where only the on-site disorder is unknown, the neural network predicts the relevant parameters with very high fidelity in the 3×33\times 33×3 setting [R2>0.99R^2 >0.99R2>0.99], and after fine tuning on only a small number of larger-device samples, it retains high accuracy for the central dot of a 5×55\times 55×5 plaquette [R2≈0.98R^2\approx 0.98R2≈0.98]. When all the dots parameters are treated as unknown, prediction of the on-site disorder remains robust [R2>0.9R^2>0.9R2>0.9 for both 3×33\times 33×3 and 5×55\times 55×5], although the remaining parameters are substantially more difficult to infer from the same charge-stability data. This shows that the most practically important disorder parameter for tuning can still be inferred reliably even in the fully disordered setting for the computationally difficult 5x5 arrays.
Abstract Recent progress in building large-scale quantum devices for exploring quantum computing and simulation has relied upon effective tools for achieving and maintaining good experimental parameters, i.e., tuning up devices. In many cases, including quantum dot-based architectures, the parameter space grows substantially with the number of qubits, and may become a limit to scalability. Fortunately, machine learning techniques for pattern recognition and image classification, using so-called deep neural networks, have shown surprising successes for computer-aided understanding of complex systems. We propose a new paradigm for fully automated experimental initialization through a closed-loop system relying on machine learning and optimization techniques. We use deep convolutional neural networks to characterize states and charge configurations of semiconductor quantum dot arrays when only measurements of a current−voltage characteristic of transport are available. For simplicity, we model a semiconductor nanowire connected to leads and capacitively coupled to depletion gates using the Thomas−Fermi approximation and Coulomb blockade physics. We then generate labeled training data for the neural networks, and find at least 90 % accuracy for charge and state identification for single and double dots. Using these characterization networks, we can then optimize the parameter space to achieve a desired configuration of the array, a technique we call “auto-tuning”. Finally, we show how such techniques can be implemented in an experimental setting by applying our approach to an experimental dataset, and outline further problems in this domain, from using charge sensing data to extensions to full one- and two-dimensional arrays, that can be tackled with machine learning.
Scaling semiconductor quantum dot arrays toward fault-tolerant quantum computing requires efficient tuneup of spin qubits, a process that depends on the analysis of charge stability maps (CSMs) and remains largely manual. While machine learning has been widely applied to CSM analysis in reservoir-coupled devices, automated tuning in the increasingly important isolated-mode regime has received limited attention. In isolated-mode CSMs, charge transitions appear as near-vertical lines, making them well suited to compact, task-specific models. We present two convolutional neural networks with fewer than one million parameters, trained on CSMs collected from 32 silicon metal-oxide-semiconductor (SiMOS) double-quantum-dot devices measured at approximately 1 K using an automated cryogenic probing system. Sixteen devices were used for training and sixteen were held out to evaluate cross-device generalization against hand-labeled ground truth. CSMClassifier identifies charge instability and sensor artifacts, achieving 94% macro-averaged accuracy across three quality classes on 2,407 held-out images. ChargeLineNet localizes charge-transition lines and determines electron occupancy, achieving 95.3% exact line-count accuracy on 1,131 held-out images. Combined into a single pipeline, the models correctly determine electron occupancy for 93.8% of clean held-out images. Pre-training on synthetic images substantially improves label efficiency. Fine-tuning the pre-trained model on limited experimental data maintains over 90% accuracy, whereas training from scratch degrades significantly under the same conditions. Together, the two models occupy only 6.5 MB and process images in less than 60 ms on standard laboratory hardware, demonstrating a practical path toward scalable, automated characterization and tuneup of quantum-dot devices.
Abstract Tuning of gate-defined semiconductor quantum dots (QDs) is a major bottleneck for scaling spin-qubit technologies. We present a deep learning driven, semantic-segmentation pipeline that performs charge auto-tuning by locating transition lines in full charge stability diagrams (CSDs) and returns gate voltage targets for the single charge regime. We assemble and manually annotate a large, heterogeneous dataset of 1015 experimental CSDs measured from fully depleted silicon-on-insulator (FD-SOI) QD devices, spanning nine design geometries, multiple wafers, and fabrication runs. A U-Net style convolutional neural network with a MobileNetV2 encoder is trained and validated through five-fold group cross-validation. Our model achieves an overall offline tuning success of 80.0% in locating the single-charge regime, with peak performance exceeding 88% for some designs. We analyze dominant failure modes and propose targeted mitigations. Finally, wide-range diagram segmentation also naturally enables scalable physics-based feature extraction that can feed back to fabrication and design workflows and outline a roadmap for real-time integration in a cryogenic wafer prober. Overall, our results show that neural network based wide-diagram segmentation is a practical step toward automated, high-throughput charge tuning for silicon QD qubits.
Semiconductor quantum computing leverages quantum dots as qubits, where achieving high qubit fidelity is one essential requirement for scalability. The formation of qubits and associated sensor dots necessitates meticulous iterative voltage adjustments, a tuning process that is ordinarily executed manually by researchers. While various automated approaches have been proposed, only a few address scalability in the context of a system with a million qubits or consider shifting key data analysis tasks into the cryostat. The latter is necessary to alleviate wiring constraints and to minimize data transmission to room temperature electronics. This thesis explores scalable solutions for automated in-cryostat data analysis, identifying charge transition detection as a pivotal task. A simulation model is developed to generate realistic charge stability diagrams with known ground truth. The model’s output is benchmarked against experimental data to verify fidelity and diversity. Utilizing this model, a comprehensive evaluation of charge transition detection methods is conducted, comparing classical algorithms and a range of machine learning approaches. In consideration of energy-efficient integration, this study includes an initial investigation into reducing the complexity of convolutional neural networks. Subsequent to their optimization using simulated data, the approaches are benchmarked on both simulated and experimental data from diverse qubit samples. When applied to contemporary data qualities, evidence is presented that classical approaches are inadequate in terms of robustness and accuracy while numerous machine learning models demonstrate compelling performance. Among the latter, convolutional neural networks like U-Net are identified to be particularly eligible due to their advantageous structure for efficient hardware integration. Although the U-Net architecture already demonstrates a strong potential for network size reduction under current data quality constraints, a cryogenic hardware integration remains a significant challenge. Consequently, a subsequent study explores if enhanced sensor dot tuning and sensor compensation reduce the complexity of charge transition analysis and if they enable the utilization of ray-based detection methodologies. The findings indicate that particularly sensor compensation has a substantial impact on the data quality and allows a streamlining of the analytical process. However, achieving robust ray-based analysis remains feasible only if consistently reaching the highest observed measurement quality and utilizing machine learning. Even in this case, classical methods and a proposed averaging and thresholding circuit fall short of the requisite detection accuracy. Only with further improvements toward optimal data quality and additional enhancements to the proposed circuit, an integrated solution without machine learning appears attainable. Finally, the energy efficiency of scaled-down neural networks is assessed using state-of-the art hardware accelerators. The findings suggest that the cryogenic integration of machine learning-based charge transition detection is feasible, thus propelling the development of scalable and energy-efficient quantum control architectures.