Automated in situ optimization and disorder mitigation in a quantum device
Jacob Benestad, Torbjørn Rasmussen, Bertram Brovang, Oswin Krause, Saeed Fallahi, Geoffrey C. Gardner, Michael J. Manfra, Charles M. Marcus, Jeroen Danon, Ferdinand Kuemmeth, Anasua Chatterjee, Evert van Nieuwenburg · 2024
Read Automated in situ optimization and disorder mitigation in a quantum device by Jacob Benestad, Torbjørn Rasmussen, Bertram Brovang, Oswin Krause, Saeed Fallahi, Geoffrey C. Gardner, Michael J. Manfra, Charles M. Marcus, Jeroen Danon, Ferdinand Kuemmeth, Anasua Chatterjee, Evert van Nieuwenburg (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-04997.txt.
Introduction
Jacob Benestad and a large team of co-authors work on tuning a tiny quantum device in a two thousand twenty-four paper. The device uses nine small gates set over a narrow channel. By changing the gate voltages, staff can shape the flow of charge through the channel. Doing this by hand is slow and hard.
The team let a search method tune the gates on its own, guided by a score for how clean the flow steps look. In computer tests, the tuning fixed much of the harm from random flaws in the device. They then ran the same method on a real device. The flow steps became much cleaner there too.
Catalog record
| CREATOR | Jacob Benestad, Torbjørn Rasmussen, Bertram Brovang, Oswin Krause, Saeed Fallahi, Geoffrey C. Gardner, Michael J. Manfra, Charles M. Marcus, Jeroen Danon, Ferdinand Kuemmeth, Anasua Chatterjee, Evert van Nieuwenburg |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-04997
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Automated in situ optimization and disorder mitigation in a quantum device Preprint: APS/123-QED Jacob Benestad, Torbjørn Rasmussen, Bertram Brovang, Oswin Krause, Saeed Fallahi, Geoffrey C. Gardner, Michael J. Manfra, Charles M. Marcus, Jeroen Danon, Ferdinand Kuemmeth, Anasua Chatterjee, Evert van Nieuwenburg Abstract We investigate automated in situ optimization of the potential landscape in a quantum point contact device, using a 3 × 3 3\times 3 gate array patterned atop the constriction. Optimization is performed using the covariance matrix adaptation evolutionary strategy, for which we introduce a metric for how “step-like” the conductance is as the channel becomes constricted. We first perform the optimization of the gate voltages in a tight-binding simulation and show how such in situ tuning can be used to mitigate a random disorder potential. The optimization is then performed in a physical device in experiment, where we also observe a marked improvement in the quantization of the conductance resulting from the optimization procedure. Introduction Machine learning has garnered increasing attention within quantum and condensed matter physics in recent years, finding applications as a new and useful tool for scientific discovery Neupert et al. 2022 ; Dawid et al. 2023 ; Gebhart et al. 2023 . One such application is the facilitation of quantum experiments by partly or fully automating the experimental workflow in cases where tuning of several parameters is necessary Ares 2021 ; Zwolak and Taylor 2023 . Significant work has already been done towards characterization of multi-quantum-dot devices Nguyen et al. 2021 ; Zubchenko et al. 2024 ; Kalantre et al. 2019 ; Krause et al. 2022a ; Krause et al. 2022b ; Koch et al. 2023 ; Taylor and Das Sarma 2024 ; van Driel et al. 2024 and automated tuning of such systems to a desired operating configuration Moon et al. 2020 ; Durrer et al. 2020 ; Ziegler et al. 2023 . One common challenge in tuning up quantum devices is posed by material imperfections and other types of unpredictable disorder, which often hinder straightforward navigation through a large parameter space. An important envisioned application of machine learning is thus to characterize or mitigate the disorder in quantum devices, yielding optimized device tunings that would be difficult to obtain manually Thamm and Rosenow 2023 ; Craig et al. 2024 ; Chatzikyriakou et al. 2022 ; Benestad et al. 2024 . In this work, we implement an automated evolutionary strategy for in situ optimization of a quantum device. We demonstrate how our tuning protocol can both enhance device functionality and mitigate the effects of disorder. As an illustrative proof-of-principle, we apply the same optimization algorithm to a numerically simulated device as well as to a real device in experiment, where in both cases we choose the simplest approach resulting in efficient optimization. We focus on a quantum device for which most of the physics is well understood, namely a quantum point contact (QPC) defined within a GaAs-based two-dimensional electron gas, where the potential landscape in the constriction can be modified via a 3\times 3 array of electrostatic top-gates. QPCs have proved to be an invaluable resource for development of semiconductor quantum technologies. The quantized conductance demonstrated in such devices Wharam et al. 1988 ; van Wees et al. 1988 ; Büttiker 1990 is not only a fundamental hallmark of quantum physics, but has also found important applications in quantum sensing: the sensitivity to nearby charges when operated between two conductance plateaus makes QPCs well suited for detection of single-electron tunneling events in quantum-dot systems Field et al. 1993 ; Elzerman et al. 2004 ; Petta et al. 2005 ; Reilly et al. 2007 ; Cassidy et al. 2007 ; Gustavsson et al. 2009 ; Barthel et al. 2009 ; Bäuerle et al. 2018 , which is an important ingredient in readout of spin qubits Vandersypen et al. 2017 ; Burkard et al. 2023 . Other uses for QPCs are for instance to build interferometers for probing anyonic statistics Bishara et al. 2009 ; Rosenow et al. 2016 ; Bartolomei et al. 2020 ; Nakamura et al. 2023 , or to make switches to toggle between tunneling and ballistic transport Smith 1995 . To optimize QPC functionality, i.e., the appearance of stable plateaus with integer quantized conductance, separated by sharp steps, we use the “covariance matrix adaptation evolution strategy” (CMA-ES) Hansen 2016 algorithm. Since this algorithm is (i) gradient-free, (ii) extremely simple to implement, and (iii) relatively insensitive to device-specific details influencing the loss landscape for optimization, it combines several desirable characteristics in the context of device tuning where measurements include noise and the optimization landscapes are likely non-trivial. Via the nine gate voltages of the 3\times 3 “pixel”-gate array we have direct control over the potential landscape, and can hence correct for possible microscopic disorder at length scales comparable to the pixel-gate dimensions. We first implement the optimization routine on a simulated device (using KWANT Groth et al. 2014 ), where we have full control over the properties of the device and disorder, and can experiment with possible functionality metrics. The resulting optimized gate configuration, both with and without random disorder, yields clear quantized conductance plateaus at values consistent with the physics of QPC devices, even as the optimization algorithm remains agnostic about the underlying disorder distribution, and about where the conductance plateaus should occur. More importantly, we then employ the same numerical optimization routine directly on a real device, and also demonstrate in-situ mitigation of device-specific disorder in an experiment. Device to optimize It is well known that ballistic electron transport through nanowires or 2DEG split-gate structures occurs in quantized conductance steps of 2e^{2}/h . For a sufficient splitting of the energy subbands, this becomes visible in experiments as conductance plateaus with sharp jumps between them when the transverse confinement is changed van Wees et al. 1988 ; Wharam et al. 1988 . Except in the ideal case at zero temperature, these tell-tale steps in the quantized conductance curve will to some extent be smoothed out and distorted due to factors such as temperature, scattering off the confining potential, and disorder Thomas Ihn 2010 . A question one might thus ask is if it is possible to construct a tunable potential landscape to improve the staircase signature of the quantized conductance. We consider the following system for optimization of the quantized conductance curve: a 3\times 3 array of square voltage-controlled gates, analogous to pixels in a camera sensor, patterned between a split-gate structure of two outer gates. Together, these 11 gates lie some distance z above a 2DEG in an AlGaAs/GaAS heterostructure, and the electric potential from the applied voltages shapes the potential landscape in the 2DEG. A sketch of the gate pattern is shown in Fig. 1 (a). We denote the pixel gate voltages as V_{\eta\nu} (with \eta,\nu\in\{-1,0,1\} denoting the pixel position), while the split-gate voltages are named V_{\text{SG}1} and V_{\text{SG}2} . Having several tunable gates results in many choices for what gate voltages to sweep over to constrict the channel to produce a conductance staircase. We will denote the parameter for constricting the QPC transport channel as V_{\text{QPC}} , which is the independent variable of the conductance curve. Different choices will result in different behavior; the results of sweeping the