A team of scientists has developed a physics-aware machine learning approach to bridge the ‘reality gap’ in solid-state quantum devices. This gap, caused by unpredictable material defects, hinders the optimization and scalability of these devices. The team’s approach combines a physical model, deep learning, Gaussian random field, and Bayesian inference to infer the disorder potential of a nanoscale electronic device. This could further our understanding of device variability, key to developing more complex quantum systems.
Bridging the Reality Gap in Quantum Devices with Machine Learning
The study focuses on the discrepancies between reality and simulation in the optimization and scalability of solid-state quantum devices. The team uses physics-aware machine learning to bridge this gap, specifically an approach that combines a physical model, deep learning, Gaussian random field, and Bayesian inference.
The Reality Gap in Quantum Devices
The reality gap refers to the difference between predictions and observed results. In the context of solid-state quantum devices, this gap is often widened by the unpredictable distribution of material defects. These defects can cause nominally identical devices to display different characteristics, hindering the scalability of promising qubit realizations. The team’s physics-aware machine learning approach aims to narrow this reality gap by revealing hidden features of the material imperfections affecting the devices.
Physics-Aware Machine Learning Approach
The team’s approach involves using simple measurements of current at different voltage settings to inform the simulator. This approach reduces the gap between simulations and experimental measurements, leading to more accurate predictions of nanoscale device properties. This understanding of device variability is key to developing more complex quantum systems. The team’s approach combines a physical model accelerated by deep learning, Bayesian inference informed by indirect experimental measurements, and dimensionality reduction of the 2D disorder potential using inducing points and random Fourier features.
The Role of Disorder Potentials
Disorder potentials, induced by randomly located donor ions, can be a significant source of variability in delta-doped semiconductor quantum-dot devices. Being able to observe these disorder potentials and provide a quantitative measure of their impact on transport properties could inform the growth of semiconductor platforms for quantum device fabrication. It could also benefit the operation of quantum devices, as particular gate-voltage configurations could be chosen to avoid the negative effects of steep gradients in the electrostatic environment.
The Inference Process
The inference process used in the study follows the philosophy of approximate Bayesian computation. It utilizes the deep learning approximation of the electrostatic model. The team developed a novel reparametrization to greatly reduce the dimensionality of the inference problem while selecting only the most informative regions of the disorder potential. This approach can be adapted to capture varying sources of disorder present in different device realizations by choosing an appropriate kernel for the Gaussian process.
Verification of the Approach
To verify the effectiveness of their approach, the team used the disorder potentials produced by the algorithm to predict the electron-transport regime of new measurements. These predictions showed good agreement with the experiment, indicating that the physics-aware method is effective. The physical model can determine the number of quantum dots at a given voltage location. Using posterior disorder samples within this model allows the team to predict the voltage locations of double quantum dots and verify these predictions with the experiment in different thermal cycles of the device. The results show that the team’s physics-aware machine learning provides a clear advantage over an uninformed model of the disorder potential when predicting the location of double-quantum-dot features in gate-voltage space.
One of the driving forces of human discovery is the difference between predictions and observed results; this is the reality gap. When playing “crazy golf,” for example, the ball may enter a tunnel and exit with a speed or direction that does not match our predictions. But with a few more shots, a crazy-golf simulator, and some machine learning, we might get better at predicting the ball’s movements and narrow the reality gap. Solid-state quantum devices throw up similar barriers to understanding: Nominally identical devices will often display different characteristics. Here, we take a machine learning approach to narrow the reality gap in such devices.
Differences between theory and experiment pervade all of science and are one of the driving forces of human discovery. Simulations often require fewer resources than real experiments but rarely capture the full complexity of a system, limiting their practical application. Narrowing the gap between a model and the real world is key for the control of complex systems using machine learning, especially when a machine learning model is trained on a simulation before being applied to real systems.
Being able to observe disorder potentials and provide a quantitative measure of the extent to which disorder impacts transport properties could inform the growth of semiconductor platforms for quantum device fabrication. It would also benefit the operation of quantum devices since particular gate-voltage configurations could be chosen to avoid the negative effects of steep gradients in the electrostatic environment.
Summary
Researchers have used a physics-aware machine learning approach to bridge the gap between reality and simulation in solid-state quantum devices, which are often hindered by unpredictable material defects. This method allows for the inference of disorder potential in nanoscale electronic devices, improving the accuracy of simulations and aiding the development of more complex quantum systems.
- A team of researchers, including D. L. Craig, H. Moon, F. Fedele, D. T. Lennon, B. van Straaten, F. Vigneau, L. C. Camenzind, D. M. Zumbühl, G. A. D. Briggs, M. A. Osborne, D. Sejdinovic, and N. Ares, have developed a physics-aware machine learning approach to bridge the ‘reality gap’ in solid-state quantum devices.
- The ‘reality gap’ refers to the discrepancies between simulation predictions and actual results, often caused by unpredictable material defects.
- The team’s approach combines a physical model, deep learning, Gaussian random field, and Bayesian inference to infer the disorder potential of a nanoscale electronic device from electron-transport data.
- This method has been validated by predicting the gate-voltage values required for a quantum-dot device in AlGaAs/GaAs to produce current features corresponding to a double-quantum-dot regime.
- The approach reduces the gap between simulations and experimental measurements, improving the understanding of device variability, which is crucial for developing more complex quantum systems.
- The team’s work could inform the growth of semiconductor platforms for quantum device fabrication and benefit the operation of quantum devices by allowing for more disorder-resilient device design.
- The researchers’ physics-aware machine learning approach could be applied to observe spatial features of any hidden 2D function that influences observations, such as different sources of electrostatic disorder in different material systems.
- The work is published in Physical Review X Titled: Bridging the Reality Gap in Quantum Devices with Physics-Aware Machine Learning
