Fast machine-learning online optimization of ultra-cold-atom experiments

Fundamental Limits of Feedback Cooling Ultracold Atomic Gases

We investigate the fundamental viability of cooling ultracold atomic gases with quantum feedback control. Our Letter shows that the trade-off between the resolution and destructiveness of optical imaging techniques imposes constraints on the efficacy of feedback cooling, and that rapid rethermalization is necessary for cooling thermal gases. We construct a simple model to determine the limits to feedback cooling set by the visibility of density fluctuations, measurement-induced heating, and three-body atomic recombination. We demonstrate that feedback control can rapidly cool high-temperature thermal clouds in quasi-2D geometries to degenerate temperatures with minimal atom loss compared to traditional evaporation. Our analysis confirms the feasibility of feedback cooling ultracold atomic gases, providing a pathway to new regimes of cooling not achievable with current approaches.

Physics-constrained machine learning for electrodynamics without gauge ambiguity based on Fourier transformed Maxwell's equations

We utilize a Fourier transformation-based representation of Maxwell's equations to develop physics-constrained neural networks for electrodynamics without gauge ambiguity, which we label the Fourier-Helmholtz-Maxwell neural operator method. In this approach, both of Gauss's laws and Faraday's law are built in as hard constraints, as well as the longitudinal component of Ampère-Maxwell in Fourier space, assuming the continuity equation. An encoder-decoder network acts as a solution operator for the transverse components of the Fourier transformed vector potential,A ^ ⊥ ( k , t ) , whose two degrees of freedom are used to predict the electromagnetic fields. This method was tested on two electron beam simulations. Among the models investigated, it was found that a U-Net architecture exhibited the best performance as it trained quicker, was more accurate and generalized better than the other architectures examined. We demonstrate that our approach is useful for solving Maxwell's equations for the electromagnetic fields generated by intense relativistic charged particle beams and that it generalizes well to unseen test data, while being orders of magnitude quicker than conventional simulations. We show that the model can be re-trained to make highly accurate predictions in as few as 20 epochs on a previously unseen data set.

Bose-Einstein Condensation by Polarization Gradient Laser Cooling

Attempts to create quantum degenerate gases without evaporative cooling have been pursued since the early days of laser cooling, with the consensus that polarization gradient cooling (PGC, also known as "optical molasses") alone cannot reach condensation. In the present work, we report that simple PGC can generate a small Bose-Einstein condensate (BEC) inside a corrugated micrometer-sized optical dipole trap. The experimental parameters enabling BEC creation were found by machine learning, which increased the atom number by a factor of 5 and decreased the temperature by a factor of 2.5, corresponding to almost 2 orders of magnitude gain in phase space density. When the trapping light is slightly misaligned through a microscopic objective lens, a BEC of ∼250 ^{87}Rb atoms is formed inside a local dimple within 40 ms of PGC after MOT loading.

Polarization self-compensation in a laser-driven interferometric fiber optic gyroscope with high long-term stability

We present a laser-driven interferometric fiber optic gyroscope (IFOG) with polarization self-compensation to achieve high scale-factor stability, sensitivity, and long-term stability. Coherent light with 200kHz linewidth is employed to keep the scale factor stable. The optical scheme ensures polarization reciprocity as well as the optimal working point for good sensitivity. Furthermore, a hybrid machine learning loop (MLL) method, combining the advantages of PID fast response and artificial neural network (ANN) dynamic search, can control a liquid crystal rotator (LCR) to dynamically compensate for slow drift induced by polarization coupling. In open environment, when the sensitivity is 0.005 ∘/h, the bias instability (BI) is significantly optimized from 0.6723°/h at 60s (PID) to 0.3869°/h at 200s (MLL), which is close to the Sagnac interferometric limit (SIL). Such IFOG can meet the real-time and robust requirements for inertial navigation systems in long-term measurement.

Performance metrics to unleash the power of self-driving labs in chemistry and materials science

With the rise of self-driving labs (SDLs) and automated experimentation across chemical and materials sciences, there is a considerable challenge in designing the best autonomous lab for a given problem based on published studies alone. Determining what digital and physical features are germane to a specific study is a critical aspect of SDL design that needs to be approached quantitatively. Even when controlling for features such as dimensionality, every experimental space has unique requirements and challenges that influence the design of the optimal physical platform and algorithm. Metrics such as optimization rate are therefore not necessarily indicative of the capabilities of an SDL across different studies. In this perspective, we highlight some of the critical metrics for quantifying performance in SDLs to better guide researchers in implementing the most suitable strategies. We then provide a brief review of the existing literature under the lens of quantified performance as well as heuristic recommendations for platform and experimental space pairings.

Machine-learning iterative optimization for all polarization-maintaining linear cavity Er:fiber laser

All polarization-maintaining (PM) linear cavity mode-locked fiber lasers are promising ultrafast laser sources due to their compactness and environmental robustness. Here, we demonstrate a linear cavity fiber laser with all-PM configuration experimentally and investigate the mode-locking formation of the laser using a machine-learning iterative optimization method based on the Gaussian process. The optimization algorithm can converge rapidly after only 30 runs. Using the optimized parameters, we simulate the generation of mode-locked pulses from noise. The output spectrum and pulse energy are highly consistent with the experiment. Furthermore, we describe the intracavity dynamic evolution under group velocity mismatch. We then show that the pulse trapping induced by cross-phase modulation leads to the overcompensated time synchronization between the orthogonally polarized components.

Automated Machine Learning Strategies for Multi-Parameter Optimisation of a Caesium-Based Portable Zero-Field Magnetometer

Machine learning (ML) is an effective tool to interrogate complex systems to find optimal parameters more efficiently than through manual methods. This efficiency is particularly important for systems with complex dynamics between multiple parameters and a subsequent high number of parameter configurations, where an exhaustive optimisation search would be impractical. Here we present a number of automated machine learning strategies utilised for optimisation of a single-beam caesium (Cs) spin exchange relaxation free (SERF) optically pumped magnetometer (OPM). The sensitivity of the OPM (T/Hz), is optimised through direct measurement of the noise floor, and indirectly through measurement of the on-resonance demodulated gradient (mV/nT) of the zero-field resonance. Both methods provide a viable strategy for the optimisation of sensitivity through effective control of the OPM's operational parameters. Ultimately, this machine learning approach increased the optimal sensitivity from 500 fT/Hz to <109fT/Hz. The flexibility and efficiency of the ML approaches can be utilised to benchmark SERF OPM sensor hardware improvements, such as cell geometry, alkali species and sensor topologies.

Self-driving laboratories: A paradigm shift in nanomedicine development

Nanomedicines have transformed promising therapeutic agents into clinically approved medicines with optimal safety and efficacy profiles. This is exemplified by the mRNA vaccines against COVID-19, which were made possible by lipid nanoparticle technology. Despite the success of nanomedicines to date, their design remains far from trivial, in part due to the complexity associated with their preclinical development. Herein, we propose a nanomedicine materials acceleration platform (NanoMAP) to streamline the preclinical development of these formulations. NanoMAP combines high-throughput experimentation with state-of-the-art advances in artificial intelligence (including active learning and few-shot learning) as well as a web-based application for data sharing. The deployment of NanoMAP requires interdisciplinary collaboration between leading figures in drug delivery and artificial intelligence to enable this data-driven design approach. The proposed approach will not only expedite the development of next-generation nanomedicines but also encourage participation of the pharmaceutical science community in a large data curation initiative.

A comprehensive deep learning method for empirical spectral prediction and its quantitative validation of nano-structured dimers

Nanophotonics exploits the best of photonics and nanotechnology which has transformed optics in recent years by allowing subwavelength structures to enhance light-matter interactions. Despite these breakthroughs, design, fabrication, and characterization of such exotic devices have remained through iterative processes which are often computationally costly, memory-intensive, and time-consuming. In contrast, deep learning approaches have recently shown excellent performance as practical computational tools, providing an alternate avenue for speeding up such nanophotonics simulations. This study presents a DNN framework for transmission, reflection, and absorption spectra predictions by grasping the hidden correlation between the independent nanostructure properties and their corresponding optical responses. The proposed DNN framework is shown to require a sufficient amount of training data to achieve an accurate approximation of the optical performance derived from computational models. The fully trained framework can outperform a traditional EM solution using on the COMSOL Multiphysics approach in terms of computational cost by three orders of magnitude. Furthermore, employing deep learning methodologies, the proposed DNN framework makes an effort to optimise design elements that influence the geometrical dimensions of the nanostructure, offering insight into the universal transmission, reflection, and absorption spectra predictions at the nanoscale. This paradigm improves the viability of complicated nanostructure design and analysis, and it has a lot of potential applications involving exotic light-matter interactions between nanostructures and electromagnetic fields. In terms of computational times, the designed algorithm is more than 700 times faster as compared to conventional FEM method (when manual meshing is used). Hence, this approach paves the way for fast yet universal methods for the characterization and analysis of the optical response of nanophotonic systems.

QCforever: A Quantum Chemistry Wrapper for Everyone to Use in Black-Box Optimization

To obtain observable physical or molecular properties such as ionization potential and fluorescent wavelength with quantum chemical (QC) computation, multi-step computation manipulated by a human is required. Hence, automating the multi-step computational process and making it a black box that can be handled by anybody are important for effective database construction and fast realistic material design through the framework of black-box optimization where machine learning algorithms are introduced as a predictor. Here, we propose a Python library, QCforever, to automate the computation of some molecular properties and chemical phenomena induced by molecules. This tool just requires a molecule file for providing its observable properties, automating the computation process of molecular properties (for ionization potential, fluorescence, etc.) and output analysis for providing their multi-values for evaluating a molecule. Incorporating the tool in black-box optimization, we can explore molecules that have properties we desired within the limitation of QC computation.

A Generalized Framework of Multifidelity Max-Value Entropy Search through Joint Entropy

Bayesian optimization (BO) is a popular method for expensive black-box optimization problems; however, querying the objective function at every iteration can be a bottleneck that hinders efficient search capabilities. In this regard, multifidelity Bayesian optimization (MFBO) aims to accelerate BO by incorporating lower-fidelity observations available with a lower sampling cost. In our previous work, we proposed an information-theoretic approach to MFBO, referred to as multifidelity max-value entropy search (MF-MES), which inherits practical effectiveness and computational simplicity of the well-known max-value entropy search (MES) for the single-fidelity BO. However, the applicability of MF-MES is still limited to the case that a single observation is sequentially obtained. In this letter, we generalize MF-MES so that information gain can be evaluated even when multiple observations are simultaneously obtained. This generalization enables MF-MES to address two practical problem settings: synchronous parallelization and trace-aware querying. We show that the acquisition functions for these extensions inherit the simplicity of MF-MES without introducing additional assumptions. We also provide computational techniques for entropy evaluation and posterior sampling in the acquisition functions, which can be commonly used for all variants of MF-MES. The effectiveness of MF-MES is demonstrated using benchmark functions and real-world applications such as materials science data and hyperparameter tuning of machine-learning algorithms.

Quantum machine learning for chemistry and physics

Machine learning (ML) has emerged as a formidable force for identifying hidden but pertinent patterns within a given data set with the objective of subsequent generation of automated predictive behavior. In recent years, it is safe to conclude that ML and its close cousin, deep learning (DL), have ushered in unprecedented developments in all areas of physical sciences, especially chemistry. Not only classical variants of ML, even those trainable on near-term quantum hardwares have been developed with promising outcomes. Such algorithms have revolutionized materials design and performance of photovoltaics, electronic structure calculations of ground and excited states of correlated matter, computation of force-fields and potential energy surfaces informing chemical reaction dynamics, reactivity inspired rational strategies of drug designing and even classification of phases of matter with accurate identification of emergent criticality. In this review we shall explicate a subset of such topics and delineate the contributions made by both classical and quantum computing enhanced machine learning algorithms over the past few years. We shall not only present a brief overview of the well-known techniques but also highlight their learning strategies using statistical physical insight. The objective of the review is not only to foster exposition of the aforesaid techniques but also to empower and promote cross-pollination among future research in all areas of chemistry which can benefit from ML and in turn can potentially accelerate the growth of such algorithms.

Gaussian process regression for absorption spectra analysis of molecular dimers

A common task is the determination of system parameters from spectroscopy, where one compares the experimental spectrum with calculated spectra, that depend on the desired parameters. Here we discuss an approach based on a machine learning technique, where the parameters for the numerical calculations are chosen from Gaussian Process Regression (GPR). This approach does not only quickly converge to an optimal parameter set, but in addition provides information about the complete parameter space, which allows for example to identify extended parameter regions where numerical spectra are consistent with the experimental one. We consider as example dimers of organic molecules and aim at extracting in particular the interaction between the monomers, and their mutual orientation. We find that indeed the GPR gives reliable results which are in agreement with direct calculations of these parameters using quantum chemical methods.

Robotic search for optimal cell culture in regenerative medicine

Induced differentiation is one of the most experience- and skill-dependent experimental processes in regenerative medicine, and establishing optimal conditions often takes years. We developed a robotic AI system with a batch Bayesian optimization algorithm that autonomously induces the differentiation of induced pluripotent stem cell-derived retinal pigment epithelial (iPSC-RPE) cells. From 200 million possible parameter combinations, the system performed cell culture in 143 different conditions in 111 days, resulting in 88% better iPSC-RPE production than that obtained by the pre-optimized culture in terms of the pigmentation scores. Our work demonstrates that the use of autonomous robotic AI systems drastically accelerates systematic and unbiased exploration of experimental search space, suggesting immense use in medicine and research.

Dynamic compensation of stray electric fields in an ion trap using machine learning and adaptive algorithm

Surface ion traps are among the most promising technologies for scaling up quantum computing machines, but their complicated multi-electrode geometry can make some tasks, including compensation for stray electric fields, challenging both at the level of modeling and of practical implementation. Here we demonstrate the compensation of stray electric fields using a gradient descent algorithm and a machine learning technique, which trained a deep learning network. We show automated dynamical compensation tested against induced electric charging from UV laser light hitting the chip trap surface. The results show improvement in compensation using gradient descent and the machine learner over manual compensation. This improvement is inferred from an increase of the fluorescence rate of 78% and 96% respectively, for a trapped [Formula: see text]Yb[Formula: see text] ion driven by a laser tuned to [Formula: see text] MHz of the [Formula: see text]S[Formula: see text]P[Formula: see text] Doppler cooling transition at 369.5 nm.

Deep learning enhanced Rydberg multifrequency microwave recognition

Recognition of multifrequency microwave (MW) electric fields is challenging because of the complex interference of multifrequency fields in practical applications. Rydberg atom-based measurements for multifrequency MW electric fields is promising in MW radar and MW communications. However, Rydberg atoms are sensitive not only to the MW signal but also to noise from atomic collisions and the environment, meaning that solution of the governing Lindblad master equation of light-atom interactions is complicated by the inclusion of noise and high-order terms. Here, we solve these problems by combining Rydberg atoms with deep learning model, demonstrating that this model uses the sensitivity of the Rydberg atoms while also reducing the impact of noise without solving the master equation. As a proof-of-principle demonstration, the deep learning enhanced Rydberg receiver allows direct decoding of the frequency-division multiplexed signal. This type of sensing technology is expected to benefit Rydberg-based MW fields sensing and communication.

From Platform to Knowledge Graph: Evolution of Laboratory Automation

High-fidelity computer-aided experimentation is becoming more accessible with the development of computing power and artificial intelligence tools. The advancement of experimental hardware also empowers researchers to reach a level of accuracy that was not possible in the past. Marching toward the next generation of self-driving laboratories, the orchestration of both resources lies at the focal point of autonomous discovery in chemical science. To achieve such a goal, algorithmically accessible data representations and standardized communication protocols are indispensable. In this perspective, we recategorize the recently introduced approach based on Materials Acceleration Platforms into five functional components and discuss recent case studies that focus on the data representation and exchange scheme between different components. Emerging technologies for interoperable data representation and multi-agent systems are also discussed with their recent applications in chemical automation. We hypothesize that knowledge graph technology, orchestrating semantic web technologies and multi-agent systems, will be the driving force to bring data to knowledge, evolving our way of automating the laboratory.

Quantum Reservoir Computing for Speckle Disorder Potentials

Quantum reservoir computing is a machine learning approach designed to exploit the dynamics of quantum systems with memory to process information. As an advantage, it presents the possibility to benefit from the quantum resources provided by the reservoir combined with a simple and fast training strategy. In this work, this technique is introduced with a quantum reservoir of spins and it is applied to find the ground state energy of an additional quantum system. The quantum reservoir computer is trained with a linear model to predict the lowest energy of a particle in the presence of different speckle disorder potentials. The performance of the task is analyzed with a focus on the observable quantities extracted from the reservoir and it is shown to be enhanced when two-qubit correlations are employed.

Online Machine Learning for Accelerating Molecular Dynamics Modeling of Cells

We developed a biomechanics-informed online learning framework to learn the dynamics with ground truth generated with multiscale modeling simulation. It was built on Summit-like supercomputers, which were also used to benchmark and validate our framework on one physiologically significant modeling of deformable biological cells. We generalized the century-old equation of Jeffery orbits to a new equation of motion with additional parameters to account for the flow conditions and the cell deformability. Using simulation data at particle-based resolutions for flowing cells and the learned parameters from our framework, we validated the new equation by the motions, mostly rotations, of a human platelet in shear blood flow at various shear stresses and platelet deformability. Our online framework, which surrogates redundant computations in the conventional multiscale modeling by solutions of our learned equation, accelerates the conventional modeling by three orders of magnitude without visible loss of accuracy.

Maximized atom number for a grating magneto-optical trap via machine-learning assisted parameter optimization

We present a parameter set for obtaining the maximum number of atoms in a grating magneto-optical trap (gMOT) by employing a machine learning algorithm. In the multi-dimensional parameter space, which imposes a challenge for global optimization, the atom number is efficiently modeled via Bayesian optimization with the evaluation of the trap performance given by a Monte-Carlo simulation. Modeling gMOTs for six representative atomic species - ⁷Li, ²³Na, ⁸⁷Rb, ⁸⁸Sr, ¹³³Cs, ¹⁷⁴Yb - allows us to discover that the optimal grating reflectivity is consistently higher than a simple estimation based on balanced optical molasses. Our algorithm also yields the optimal diffraction angle which is independent of the beam waist. The validity of the optimal parameter set for the case of ⁸⁷Rb is experimentally verified using a set of grating chips with different reflectivities and diffraction angles.

Mechanistic and Machine Learning Modeling of Microwave Heating Process in Domestic Ovens: A Review

The domestic microwave oven has been popularly used at home in heating foods for its rapid heating rate and high power efficiency. However, non-uniform heating by microwave is the major drawback that can lead to severe food safety and quality issues. In order to alleviate this problem, modeling of microwave heating process in domestic ovens has been employed to simulate and understand the complicated interactions between microwaves and food products. This paper extensively reviews the mechanistic models with different geometric dimensions and physics/kinetics that simulated the microwave heating process. The model implementation and validation strategies related to the model accuracy and efficiency are also discussed. With the emergence of the machine learning technique, this paper also discusses the recent development of hybrid models that integrate machine learning with mechanistic models in improving microwave heating performance. Besides, pure machine learning models using only experimental data as input are also covered. Further research is needed to improve the model accuracy, efficiency, and ease of use to enable the industrial application of the models in the development of microwave systems and food products.

Unsupervised machine learning of topological phase transitions from experimental data

Identifying phase transitions is one of the key challenges in quantum many-body physics. Recently, machine learning methods have been shown to be an alternative way of localising phase boundaries from noisy and imperfect data without the knowledge of the order parameter. Here, we apply different unsupervised machine learning techniques, including anomaly detection and influence functions, to experimental data from ultracold atoms. In this way, we obtain the topological phase diagram of the Haldane model in a completely unbiased fashion. We show that these methods can successfully be applied to experimental data at finite temperatures and to the data of Floquet systems when post-processing the data to a single micromotion phase. Our work provides a benchmark for the unsupervised detection of new exotic phases in complex many-body systems.

Olympus: a benchmarking framework for noisy optimization and experiment planning

Research challenges encountered across science, engineering, and economics can frequently be formulated as optimization tasks. In chemistry and materials science, recent growth in laboratory digitization and automation has sparked interest in optimization-guided autonomous discovery and closed-loop experimentation. Experiment planning strategies based on off-the-shelf optimization algorithms can be employed in fully autonomous research platforms to achieve desired experimentation goals with the minimum number of trials. However, the experiment planning strategy that is most suitable to a scientific discovery task is a priori unknown while rigorous comparisons of different strategies are highly time and resource demanding. As optimization algorithms are typically benchmarked on low-dimensional synthetic functions, it is unclear how their performance would translate to noisy, higher-dimensional experimental tasks encountered in chemistry and materials science. We introduce Olympus, a software package that provides a consistent and easy-to-use framework for benchmarking optimization algorithms against realistic experiments emulated via probabilistic deep-learning models. Olympus includes a collection of experimentally derived benchmark sets from chemistry and materials science and a suite of experiment planning strategies that can be easily accessed via a user-friendly Python interface. Furthermore, Olympus facilitates the integration, testing, and sharing of custom algorithms and user-defined datasets. In brief, Olympus mitigates the barriers associated with benchmarking optimization algorithms on realistic experimental scenarios, promoting data sharing and the creation of a standard framework for evaluating the performance of experiment planning strategies.

High-field solution state DNP using cross-correlations

At the magnetic fields of common NMR instruments, electron Zeeman frequencies are too high for efficient electron-nuclear dipolar cross-relaxation to occur in solution. The rate of that process fades with the electron Zeeman frequency as ω^-2 - in the absence of isotropic hyperfine couplings, liquid state dynamic nuclear polarisation (DNP) in high-field magnets is therefore impractical. However, contact coupling and dipolar cross-relaxation are not the only mechanisms that can move electron magnetisation to nuclei in liquids: multiple cross-correlated (CC) relaxation processes also exist, involving various combinations of interaction tensor anisotropies. The rates of some of those processes have more favourable high-field behaviour than dipolar cross-relaxation, but due to the difficulty of their numerical - and particularly analytical - treatment, they remain largely uncharted. In this communication, we report analytical evaluation of every rotationally driven relaxation process in liquid state for 1e1n and 2e1n spin systems, as well as numerical optimisations of the steady-state DNP with respect to spin Hamiltonian parameters. A previously unreported cross-correlated DNP (CCDNP) mechanism was identified for the 2e1n system, involving multiple relaxation interference effects and inter-electron exchange coupling. Using simulations, we found realistic spin Hamiltonian parameters that yield stronger nuclear polarisation at high magnetic fields than dipolar cross-relaxation.

Quantifying the separability of data classes in neural networks

We introduce the Generalized Discrimination Value (GDV) that measures, in a non-invasive manner, how well different data classes separate in each given layer of an artificial neural network. It turns out that, at the end of the training period, the GDV in each given layer L attains a highly reproducible value, irrespective of the initialization of the network's connection weights. In the case of multi-layer perceptrons trained with error backpropagation, we find that classification of highly complex data sets requires a temporal reduction of class separability, marked by a characteristic 'energy barrier' in the initial part of the GDV(L) curve. Even more surprisingly, for a given data set, the GDV(L) is running through a fixed 'master curve', independently from the total number of network layers. Finally, due to its invariance with respect to dimensionality, the GDV may serve as a useful tool to compare the internal representational dynamics of artificial neural networks with different architectures for neural architecture search or network compression; or even with brain activity in order to decide between different candidate models of brain function.

Design of coaxial coils using hybrid machine learning

A coil system to generate a uniform field is urgently needed in quantum experiments. However, general coil configurations based on the analytical method have not considered practical restrictions, such as the region for coil placement due to holes in the center of the magnetic shield, which could not be directly applied in most of the quantum experiments. In this paper, we develop a coil design method for quantum experiments using hybrid machine learning. The algorithm part consists of a machine learner based on an artificial neural network and a differential evolution (DE) learner. The cooperation of both learners demonstrates its higher efficiency than a single DE learner and robustness in the coil optimization problem compared with analytical proposals. With the help of a DE learner, in numerical simulation, a machine learner can successfully design coaxial coil systems that generate fields whose relative inhomogeneity in a 25 mm-long central region is ∼10^-6 under constraints. In addition, for experiments, a coil system with 0.069% inhomogeneity of the field, designed by a machine learner, is constructed, which is mainly limited by machining the precision of the circuit board. Benefitting from machine learning's high-dimension optimization capabilities, our coil design method is convenient and has potential for various quantum experiments.

The Raspberry Pi auto-aligner: Machine learning for automated alignment of laser beams

We present a novel solution to automated beam alignment optimization. This device is based on a Raspberry Pi computer, stepper motors, commercial optomechanics and electronic devices, and the open-source machine learning algorithm M-LOOP. We provide schematic drawings for the custom hardware necessary to operate the device and discuss diagnostic techniques to determine the performance. The beam auto-aligning device has been used to improve the alignment of a laser beam into a single-mode optical fiber from manually optimized fiber alignment, with an iteration time of typically 20 minutes. We present example data of one such measurement to illustrate device performance.

Autonomous intelligent agents for accelerated materials discovery

We present an end-to-end computational system for autonomous materials discovery. The system aims for cost-effective optimization in large, high-dimensional search spaces of materials by adopting a sequential, agent-based approach to deciding which experiments to carry out. In choosing next experiments, agents can make use of past knowledge, surrogate models, logic, thermodynamic or other physical constructs, heuristic rules, and different exploration-exploitation strategies. We show a series of examples for (i) how the discovery campaigns for finding materials satisfying a relative stability objective can be simulated to design new agents, and (ii) how those agents can be deployed in real discovery campaigns to control experiments run externally, such as the cloud-based density functional theory simulations in this work. In a sample set of 16 campaigns covering a range of binary and ternary chemistries including metal oxides, phosphides, sulfides and alloys, this autonomous platform found 383 new stable or nearly stable materials with no intervention by the researchers.

ChemOS: An orchestration software to democratize autonomous discovery

The current Edisonian approach to discovery requires up to two decades of fundamental and applied research for materials technologies to reach the market. Such a slow and capital-intensive turnaround calls for disruptive strategies to expedite innovation. Self-driving laboratories have the potential to provide the means to revolutionize experimentation by empowering automation with artificial intelligence to enable autonomous discovery. However, the lack of adequate software solutions significantly impedes the development of self-driving laboratories. In this paper, we make progress towards addressing this challenge, and we propose and develop an implementation of ChemOS; a portable, modular and versatile software package which supplies the structured layers necessary for the deployment and operation of self-driving laboratories. ChemOS facilitates the integration of automated equipment, and it enables remote control of automated laboratories. ChemOS can operate at various degrees of autonomy; from fully unsupervised experimentation to actively including inputs and feedbacks from researchers into the experimentation loop. The flexibility of ChemOS provides a broad range of functionality as demonstrated on five applications, which were executed on different automated equipment, highlighting various aspects of the software package.

Excitonic Wave Function Reconstruction from Near-Field Spectra Using Machine Learning Techniques

A general problem in quantum mechanics is the reconstruction of eigenstate wave functions from measured data. In the case of molecular aggregates, information about excitonic eigenstates is vitally important to understand their optical and transport properties. Here we show that from spatially resolved near field spectra it is possible to reconstruct the underlying delocalized aggregate eigenfunctions. Although this high-dimensional nonlinear problem defies standard numerical or analytical approaches, we have found that it can be solved using a convolutional neural network. For both one-dimensional and two-dimensional aggregates we find that the reconstruction is robust to various types of disorder and noise.

Non-standard trajectories found by machine learning for evaporative cooling of 87Rb atoms

We present a machine-learning experiment involving evaporative cooling of gaseous ⁸⁷Rb atoms. The evaporation trajectory was optimized to maximize the number of atoms cooled down to a Bose-Einstein condensate using Bayesian optimization. After 300 trials within 3 hours, Bayesian optimization discovered trajectories that achieved atom numbers comparable with those of manual tuning by a human expert. Analysis of the machine-learned trajectories revealed minimum requirements for successful evaporative cooling. We found that the manually obtained curve and the machine-learned trajectories were quite similar in terms of evaporation efficiency, although the manual and machine-learned evaporation ramps were significantly different.

Glassy Phase of Optimal Quantum Control

We study the problem of preparing a quantum many-body system from an initial to a target state by optimizing the fidelity over the family of bang-bang protocols. We present compelling numerical evidence for a universal spin-glasslike transition controlled by the protocol time duration. The glassy critical point is marked by a proliferation of protocols with close-to-optimal fidelity and with a true optimum that appears exponentially difficult to locate. Using a machine learning (ML) inspired framework based on the manifold learning algorithm t-distributed stochastic neighbor embedding, we are able to visualize the geometry of the high-dimensional control landscape in an effective low-dimensional representation. Across the transition, the control landscape features an exponential number of clusters separated by extensive barriers, which bears a strong resemblance with replica symmetry breaking in spin glasses and random satisfiability problems. We further show that the quantum control landscape maps onto a disorder-free classical Ising model with frustrated nonlocal, multibody interactions. Our work highlights an intricate but unexpected connection between optimal quantum control and spin glass physics, and shows how tools from ML can be used to visualize and understand glassy optimization landscapes.

Approaching the adiabatic timescale with machine learning

The control and manipulation of quantum systems without excitation are challenging, due to the complexities in fully modeling such systems accurately and the difficulties in controlling these inherently fragile systems experimentally. For example, while protocols to decompress Bose-Einstein condensates (BECs) faster than the adiabatic timescale (without excitation or loss) have been well developed theoretically, experimental implementations of these protocols have yet to reach speeds faster than the adiabatic timescale. In this work, we experimentally demonstrate an alternative approach based on a machine-learning algorithm which makes progress toward this goal. The algorithm is given control of the coupled decompression and transport of a metastable helium condensate, with its performance determined after each experimental iteration by measuring the excitations of the resultant BEC. After each iteration the algorithm adjusts its internal model of the system to create an improved control output for the next iteration. Given sufficient control over the decompression, the algorithm converges to a solution that sets the current speed record in relation to the adiabatic timescale, beating out other experimental realizations based on theoretical approaches. This method presents a feasible approach for implementing fast-state preparations or transformations in other quantum systems, without requiring a solution to a theoretical model of the system. Implications for fundamental physics and cooling are discussed.

Remote optimization of an ultracold atoms experiment by experts and citizen scientists

The emerging field of gamified citizen science continually probes the fault line between human and artificial intelligence. A better understanding of citizen scientists’ search strategies may lead to cognitive insights and provide inspiration for algorithmic improvements. Our project remotely engages both the general public and experts in the real-time optimization of an experimental laboratory setting. In this citizen science project the game and data acquisition are designed as a social science experiment aimed at extracting the collective search behavior of the players. A further understanding of these human skills will be a crucial challenge in the coming years, as hybrid intelligence solutions are pursued in corporate and research environments.

We introduce a remote interface to control and optimize the experimental production of Bose–Einstein condensates (BECs) and find improved solutions using two distinct implementations. First, a team of theoreticians used a remote version of their dressed chopped random basis optimization algorithm (RedCRAB), and second, a gamified interface allowed 600 citizen scientists from around the world to participate in real-time optimization. Quantitative studies of player search behavior demonstrated that they collectively engage in a combination of local and global searches. This form of multiagent adaptive search prevents premature convergence by the explorative behavior of low-performing players while high-performing players locally refine their solutions. In addition, many successful citizen science games have relied on a problem representation that directly engaged the visual or experiential intuition of the players. Here we demonstrate that citizen scientists can also be successful in an entirely abstract problem visualization. This is encouraging because a much wider range of challenges could potentially be opened to gamification in the future.

Generation of Bose-Einstein Condensates' Ground State Through Machine Learning

We show that both single-component and two-component Bose-Einstein condensates’ (BECs) ground states can be simulated by a deep convolutional neural network. We trained the neural network via inputting the parameters in the dimensionless Gross-Pitaevskii equation (GPE) and outputting the ground-state wave function. After the training, the neural network generates ground-state wave functions with high precision. We benchmark the neural network for either inputting different coupling strength in the GPE or inputting an arbitrary potential under the infinite double walls trapping potential, and it is found that the ground state wave function generated by the neural network gives the relative chemical potential error magnitude below 10⁻³. Furthermore, the neural network trained with random potentials shows prediction ability on other types of potentials. Therefore, the BEC ground states, which are continuous wave functions, can be represented by deep convolutional neural networks.

Multiparameter optimisation of a magneto-optical trap using deep learning

Machine learning based on artificial neural networks has emerged as an efficient means to develop empirical models of complex systems. Cold atomic ensembles have become commonplace in laboratories around the world, however, many-body interactions give rise to complex dynamics that preclude precise analytic optimisation of the cooling and trapping process. Here, we implement a deep artificial neural network to optimise the magneto-optic cooling and trapping of neutral atomic ensembles. The solution identified by machine learning is radically different to the smoothly varying adiabatic solutions currently used. Despite this, the solutions outperform best known solutions producing higher optical densities.

Machine learning & artificial intelligence in the quantum domain: a review of recent progress

Quantum information technologies, on the one hand, and intelligent learning systems, on the other, are both emergent technologies that are likely to have a transformative impact on our society in the future. The respective underlying fields of basic research-quantum information versus machine learning (ML) and artificial intelligence (AI)-have their own specific questions and challenges, which have hitherto been investigated largely independently. However, in a growing body of recent work, researchers have been probing the question of the extent to which these fields can indeed learn and benefit from each other. Quantum ML explores the interaction between quantum computing and ML, investigating how results and techniques from one field can be used to solve the problems of the other. Recently we have witnessed significant breakthroughs in both directions of influence. For instance, quantum computing is finding a vital application in providing speed-ups for ML problems, critical in our 'big data' world. Conversely, ML already permeates many cutting-edge technologies and may become instrumental in advanced quantum technologies. Aside from quantum speed-up in data analysis, or classical ML optimization used in quantum experiments, quantum enhancements have also been (theoretically) demonstrated for interactive learning tasks, highlighting the potential of quantum-enhanced learning agents. Finally, works exploring the use of AI for the very design of quantum experiments and for performing parts of genuine research autonomously, have reported their first successes. Beyond the topics of mutual enhancement-exploring what ML/AI can do for quantum physics and vice versa-researchers have also broached the fundamental issue of quantum generalizations of learning and AI concepts. This deals with questions of the very meaning of learning and intelligence in a world that is fully described by quantum mechanics. In this review, we describe the main ideas, recent developments and progress in a broad spectrum of research investigating ML and AI in the quantum domain.

Experimental Demonstration of Shaken-Lattice Interferometry

We experimentally demonstrate a shaken-lattice interferometer. Atoms are trapped in the ground Bloch state of a red-detuned optical lattice. Using a closed-loop optimization protocol based on the dcrab algorithm, we phase-modulate (shake) the lattice to transform the atom momentum state. In this way, we implement an atom beam splitter and build five interferometers of varying interrogation times T_{I}. The sensitivity of shaken-lattice interferometry is shown to scale as T_{I}^{2}, consistent with simulation (2C. A. Weidner, H. Yu, R. Kosloff, and D. Z. Anderson, Phys. Rev. A 95, 043624 (2017).PLRAAN2469-992610.1103/PhysRevA.95.043624). Finally, we show that we can measure the sign of an applied signal and optimize the interferometer in the presence of a bias signal.

High-performance Raman memory with spatio-temporal reversal

A number of techniques exist to use an ensemble of atoms as a quantum memory for light. Many of these propose to use backward retrieval as a way to improve the storage and recall efficiency. We report on a demonstration of an off-resonant Raman memory that uses backward retrieval to achieve an efficiency of 65 ± 6% at a storage time of one pulse duration. The memory has a characteristic decay time of 60 μs, corresponding to a delay-bandwidth product of 160.

Plasmonic nanostructure design and characterization via Deep Learning

Nanophotonics, the field that merges photonics and nanotechnology, has in recent years revolutionized the field of optics by enabling the manipulation of light-matter interactions with subwavelength structures. However, despite the many advances in this field, the design, fabrication and characterization has remained widely an iterative process in which the designer guesses a structure and solves the Maxwell's equations for it. In contrast, the inverse problem, i.e., obtaining a geometry for a desired electromagnetic response, remains a challenging and time-consuming task within the boundaries of very specific assumptions. Here, we experimentally demonstrate that a novel Deep Neural Network trained with thousands of synthetic experiments is not only able to retrieve subwavelength dimensions from solely far-field measurements but is also capable of directly addressing the inverse problem. Our approach allows the rapid design and characterization of metasurface-based optical elements as well as optimal nanostructures for targeted chemicals and biomolecules, which are critical for sensing, imaging and integrated spectroscopy applications.

Rapid Bayesian optimisation for synthesis of short polymer fiber materials

The discovery of processes for the synthesis of new materials involves many decisions about process design, operation, and material properties. Experimentation is crucial but as complexity increases, exploration of variables can become impractical using traditional combinatorial approaches. We describe an iterative method which uses machine learning to optimise process development, incorporating multiple qualitative and quantitative objectives. We demonstrate the method with a novel fluid processing platform for synthesis of short polymer fibers, and show how the synthesis process can be efficiently directed to achieve material and process objectives.