Variability is a problem for the scalability of semiconductor quantum devices. The parameter space is large, and the operating range is small. Our statistical tuning algorithm searches for specific electron transport features in gate-defined quantum dot devices with a gate voltage space of up to eight dimensions. Starting from the full range of each gate voltage, our machine learning algorithm can tune each device to optimal performance in a median time of under 70 minutes. This performance surpassed our best human benchmark (although both human and machine performance can be improved). The algorithm is approximately 180 times faster than an automated random search of the parameter space, and is suitable for different material systems and device architectures. Our results yield a quantitative measurement of device variability, from one device to another and after thermal cycling. Our machine learning algorithm can be extended to higher dimensions and other technologies.

Hybrid superconductor-semiconductor structures attract increasing attention owing to a variety of potential applications in quantum computing devices. They can serve to the realization of topological superconducting systems, as well as gate-tunable superconducting quantum bits. Here we combine a SiGe/Ge/SiGe quantum-well heterostructure hosting high-mobility two-dimensional holes and aluminum superconducting leads to realize prototypical hybrid devices, such as Josephson field-effect transistors (JoFETs) and superconducting quantum interference devices (SQUIDs). We observe gate-controlled supercurrent transport with Ge channels as long as one micrometer and 1 arXiv:1810.05012v2 [cond-mat.mes-hall] 23 Oct 2018 estimate the induced superconducting gap from tunnel spectroscopy measurements in superconducting point-contact devices. Transmission electron microscopy reveals the diffusion of Ge into the aluminum contacts, whereas no aluminum is detected in the Ge channel.Modern quantum nanoelectronics takes increasing advantage of newly synthesized hybrid superconductor-semiconductor (S-Sm) interfaces. 1 One of the main motivations is the search for Majorana zero modes that are predicted to appear in a topological superconductor. 2-4 A Josephson field effect transistor (JoFET) is one of the basic devices. It consists of a gatetunable semiconductor channel allowing Cooper-pair exchange between two superconducting contacts mediated by the superconducting proximity effect. 5 Gate control on the Josephson coupling has eventually led to the realization of electrically tunable transmon quantum bits, now often referred to as gatemons. 6-8 Many of the reported experimental realizations of hybrid S-Sm devices rely on bottomup fabrication starting from semiconductor nanowires or carbon nanotubes. 9-16 Recently, new hybrid S-Sm devices were demonstrated using top-down fabrication processes based on two-dimensional systems made of graphene, 17 InAs, 18,19 GaAs, 20 InGaAs 21 or Ge/SiGe. 22,23Top-down nanoscale devices offer significant advantages in terms of complexity and scalability. Those based on p-type SiGe heterostructures are readily compatible with silicon technology, 24 and, thanks to their intrinsically strong spin-orbit coupling, they are an attractive candidate for the development of topological superconducting systems. 22,[25][26][27][28][29][30][31][32] In this work, we present proof-of-concept S-Sm devices in which the semiconducting element consists of an undoped SiGe heterostucture embedding a strained Ge quantum-well (QW). A high-mobility two-dimensional hole gas (2DHG) is electrostatically accumulated in the QW by means of a surface gate electrode. (Hole mobilities as high as 5×10 5 cm 2 /Vs were reported for similar heterostructures. 12,22,33,34 ) The superconducting proximity effect induces gate-tunable superconductivity in the 2DHG enabling JoFET operation. This functionality is exploited for the realization of gate-controlled superconducting quantum interference

RésuméWe report on magneto-transport measurements in InAs nanowires under large magnetic field (up to 55T), providing a direct spectroscopy of the 1D electronic band structure. Large modulations of the magneto-conductance mediated by an accurate control of the Fermi energy reveal the Landau fragmentation, carrying the fingerprints of the confined InAs material. Our numerical simulations of the magnetic band structure consistently support the experimental results and reveal key parameters of the electronic confinement.

We report experimental evidence of ballistic hole transport in one-dimensional quantum wires gate-defined in a strained SiGe/Ge/SiGe quantum well. At zero magnetic field, we observe conductance plateaus at integer multiples of 2 e/ h. At finite magnetic field, the splitting of these plateaus by Zeeman effect reveals largely anisotropic g-factors with absolute values below 1 in the quantum-well plane, and exceeding 10 out-of-plane. This g-factor anisotropy is consistent with a heavy-hole character of the propagating valence-band states, which is in line with a predominant confinement in the growth direction. Remarkably, we observe quantized ballistic conductance in device channels up to 600 nm long. These findings mark an important step toward the realization of novel devices for applications in quantum spintronics.

Deep reinforcement learning is an emerging machine-learning approach that can teach a computer to learn from their actions and rewards similar to the way humans learn from experience. It offers many advantages in automating decision processes to navigate large parameter spaces. This paper proposes an approach to the efficient measurement of quantum devices based on deep reinforcement learning. We focus on double quantum dot devices, demonstrating the fully automatic identification of specific transport features called bias triangles. Measurements targeting these features are difficult to automate, since bias triangles are found in otherwise featureless regions of the parameter space. Our algorithm identifies bias triangles in a mean time of <30 min, and sometimes as little as 1 min. This approach, based on dueling deep Q-networks, can be adapted to a broad range of devices and target transport features. This is a crucial demonstration of the utility of deep reinforcement learning for decision making in the measurement and operation of quantum devices.

Quantum devices with a large number of gate electrodes allow for precise control of device parameters. This capability is hard to fully exploit due to the complex dependence of these parameters on applied gate voltages. We experimentally demonstrate an algorithm capable of fine-tuning several device parameters at once. The algorithm acquires a measurement and assigns it a score using a variational auto-encoder. Gate voltage settings are set to optimize this score in real-time in an unsupervised fashion. We report fine-tuning times of a double quantum dot device within approximately 40 min.

Fault-tolerant spin-based quantum computers will require fast and accurate qubit readout. This can be achieved using radio-frequency reflectometry given sufficient sensitivity to the change in quantum capacitance associated with the qubit states. Here, we demonstrate a 23-fold improvement in capacitance sensitivity by supplementing a cryogenic semiconductor amplifier with a SQUID preamplifier. The SQUID amplifier operates at a frequency near 200 MHz and achieves a noise temperature below 600 mK when integrated into a reflectometry circuit, which is within a factor 120 of the quantum limit. It enables a record sensitivity to capacitance of 0.07 aF/ √ Hz. The setup is used to acquire charge stability diagrams of a gate-defined double quantum dot in a short time with a signal-to-noise ration of about 38 in 1 µs of integration time.

No abstract