Embedding human knowledge in material screening pipeline as filters to identify novel synthesizable inorganic materials.

How might one embed a chemist's knowledge into an automated materials-discovery pipeline? In generative design for inorganic crystalline materials, generating candidate compounds is no longer a bottleneck - there are now synthetic datasets of millions of compounds. However, weeding out unsynthesizable or difficult to synthesize compounds remains an outstanding challenge. Post-generation "filters" have been proposed as a means of embedding human domain knowledge, either in the form of scientific laws or rules of thumb. Examples include charge neutrality, electronegativity balance, and energy above hull. Some filters are "hard" and some are "soft" - for example, it is difficult to envision creating a stable compound while violating the rule of charge neutrality; however, several compounds break the Hume-Rothery rules. It is therefore natural to wonder: can one compile a comprehensive list of "filters" that embed domain knowledge, adopt a principled approach to classifying them as either non-conditional or conditional "filters," and envision a software environment to implement combinations of these in a systematic manner? In this commentary we explore such questions, "filters" for screening of novel inorganic compounds for synthesizability.

In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science.

Exceptional molecules and materials with one or more extraordinary properties are both technologically valuable and fundamentally interesting, because they often involve new physical phenomena or new compositions that defy expectations. Historically, exceptionality has been achieved through serendipity, but recently, machine learning (ML) and automated experimentation have been widely proposed to accelerate target identification and synthesis planning. In this Perspective, we argue that the data-driven methods commonly used today are well-suited for optimization but not for the realization of new exceptional materials or molecules. Finding such outliers should be possible using ML, but only by shifting away from using traditional ML approaches that tweak the composition, crystal structure, or reaction pathway. We highlight case studies of high-Tc oxide superconductors and superhard materials to demonstrate the challenges of ML-guided discovery and discuss the limitations of automation for this task. We then provide six recommendations for the development of ML methods capable of exceptional materials discovery: (i) Avoid the tyranny of the middle and focus on extrema; (ii) When data are limited, qualitative predictions that provide direction are more valuable than interpolative accuracy; (iii) Sample what can be made and how to make it and defer optimization; (iv) Create room (and look) for the unexpected while pursuing your goal; (v) Try to fill-in-the-blanks of input and output space; (vi) Do not confuse human understanding with model interpretability. We conclude with a description of how these recommendations can be integrated into automated discovery workflows, which should enable the discovery of exceptional molecules and materials.

Predicting Antimicrobial Activity of Conjugated Oligoelectrolyte Molecules via Machine Learning.

New antibiotics are needed to battle growing antibiotic resistance, but the development process from hit, to lead, and ultimately to a useful drug takes decades. Although progress in molecular property prediction using machine-learning methods has opened up new pathways for aiding the antibiotics development process, many existing solutions rely on large data sets and finding structural similarities to existing antibiotics. Challenges remain in modeling unconventional antibiotic classes that are drawing increasing research attention. In response, we developed an antimicrobial activity prediction model for conjugated oligoelectrolyte molecules, a new class of antibiotics that lacks extensive prior structure-activity relationship studies. Our approach enables us to predict the minimum inhibitory concentration for E. coli K12, with 21 molecular descriptors selected by recursive elimination from a set of 5305 descriptors. This predictive model achieves an R2 of 0.65 with no prior knowledge of the underlying mechanism. We find the molecular representation optimum for the domain is the key to good predictions of antimicrobial activity. In the case of conjugated oligoelectrolytes, a representation reflecting the three-dimensional shape of the molecules is most critical. Although it is demonstrated with a specific example of conjugated oligoelectrolytes, our proposed approach for creating the predictive model can be readily adapted to other novel antibiotic candidate domains.

Phase transition-induced band edge engineering of BiVO4 to split pure water under visible light.

Through phase transition-induced band edge engineering by dual doping with In and Mo, a new greenish BiVO4 (Bi1-XInXV1-XMoXO4) is developed that has a larger band gap energy than the usual yellow scheelite monoclinic BiVO4 as well as a higher (more negative) conduction band than H(+)/H2 potential [0 VRHE (reversible hydrogen electrode) at pH 7]. Hence, it can extract H2 from pure water by visible light-driven overall water splitting without using any sacrificial reagents. The density functional theory calculation indicates that In(3+)/Mo(6+) dual doping triggers partial phase transformation from pure monoclinic BiVO4 to a mixture of monoclinic BiVO4 and tetragonal BiVO4, which sequentially leads to unit cell volume growth, compressive lattice strain increase, conduction band edge uplift, and band gap widening.

Ten-percent solar-to-fuel conversion with nonprecious materials.

Direct solar-to-fuels conversion can be achieved by coupling a photovoltaic device with water-splitting catalysts. We demonstrate that a solar-to-fuels efficiency (SFE) > 10% can be achieved with nonprecious, low-cost, and commercially ready materials. We present a systems design of a modular photovoltaic (PV)-electrochemical device comprising a crystalline silicon PV minimodule and low-cost hydrogen-evolution reaction and oxygen-evolution reaction catalysts, without power electronics. This approach allows for facile optimization en route to addressing lower-cost devices relying on crystalline silicon at high SFEs for direct solar-to-fuels conversion.

Methylammonium Bismuth Iodide as a Lead-Free, Stable Hybrid Organic-Inorganic Solar Absorber.

Methylammonium lead halide (MAPbX3 ) perovskites exhibit exceptional carrier transport properties. But their commercial deployment as solar absorbers is currently limited by their intrinsic instability in the presence of humidity and their lead content. Guided by our theoretical predictions, we explored the potential of methylammonium bismuth iodide (MBI) as a solar absorber through detailed materials characterization. We synthesized phase-pure MBI by solution and vapor processing. In contrast to MAPbX3, MBI is air stable, forming a surface layer that does not increase the recombination rate. We found that MBI luminesces at room temperature, with the vapor-processed films exhibiting superior photoluminescence (PL) decay times that are promising for photovoltaic applications. The thermodynamic, electronic, and structural features of MBI that are amenable to these properties are also present in other hybrid ternary bismuth halide compounds. Through MBI, we demonstrate a lead-free and stable alternative to MAPbX3 that has a similar electronic structure and nanosecond lifetimes.

Open-circuit voltage deficit, radiative sub-bandgap states, and prospects in quantum dot solar cells.

Quantum dot photovoltaics (QDPV) offer the potential for low-cost solar cells. To develop strategies for continued improvement in QDPVs, a better understanding of the factors that limit their performance is essential. Here, we study carrier recombination processes that limit the power conversion efficiency of PbS QDPVs. We demonstrate the presence of radiative sub-bandgap states and sub-bandgap state filling in operating devices by using photoluminescence (PL) and electroluminescence (EL) spectroscopy. These sub-bandgap states are most likely the origin of the high open-circuit-voltage (VOC) deficit and relatively limited carrier collection that have thus far been observed in QDPVs. Combining these results with our perspectives on recent progress in QDPV, we conclude that eliminating sub-bandgap states in PbS QD films has the potential to show a greater gain than may be attainable by optimization of interfaces between QDs and other materials. We suggest possible future directions that could guide the design of high-performance QDPVs.

The realistic energy yield potential of GaAs-on-Si tandem solar cells: a theoretical case study.

Si based tandem solar cells represent an alternative to traditional compound III-V multijunction cells as a promising way to achieve high efficiencies. A theoretical study on the energy yield of GaAs on Si (GaAs/Si) tandem solar cells is performed to assess their energy yield potential under realistic illumination conditions with varying spectrum. We find that the yield of a 4-terminal contact scheme with thick top cell is more than 15% higher than for a 2-terminal scheme. Furthermore, we quantify the main losses that occur for this type of solar cell under varying spectra. Apart from current mismatch, we find that a significant power loss can be attributed to low irradiance seen by the sub-cells. The study shows that despite non-optimal bandgap combination, GaAs/Si tandem solar cells have the potential to surpass 30% energy conversion efficiency.

Light-induced water oxidation at silicon electrodes functionalized with a cobalt oxygen-evolving catalyst.

Integrating a silicon solar cell with a recently developed cobalt-based water-splitting catalyst (Co-Pi) yields a robust, monolithic, photo-assisted anode for the solar fuels process of water splitting to O(2) at neutral pH. Deposition of the Co-Pi catalyst on the Indium Tin Oxide (ITO)-passivated p-side of a np-Si junction enables the majority of the voltage generated by the solar cell to be utilized for driving the water-splitting reaction. Operation under neutral pH conditions fosters enhanced stability of the anode as compared to operation under alkaline conditions (pH 14) for which long-term stability is much more problematic. This demonstration of a simple, robust construct for photo-assisted water splitting is an important step towards the development of inexpensive direct solar-to-fuel energy conversion technologies.

Using scalable computer vision to automate high-throughput semiconductor characterization.

High-throughput materials synthesis methods, crucial for discovering novel functional materials, face a bottleneck in property characterization. These high-throughput synthesis tools produce 104 samples per hour using ink-based deposition while most characterization methods are either slow (conventional rates of 101 samples per hour) or rigid (e.g., designed for standard thin films), resulting in a bottleneck. To address this, we propose automated characterization (autocharacterization) tools that leverage adaptive computer vision for an 85x faster throughput compared to non-automated workflows. Our tools include a generalizable composition mapping tool and two scalable autocharacterization algorithms that: (1) autonomously compute the band gaps of 200 compositions in 6 minutes, and (2) autonomously compute the environmental stability of 200 compositions in 20 minutes, achieving 98.5% and 96.9% accuracy, respectively, when benchmarked against domain expert manual evaluation. These tools, demonstrated on the formamidinium (FA) and methylammonium (MA) mixed-cation perovskite system FA1-xMAxPbI3, 0 ≤ x ≤ 1, significantly accelerate the characterization process, synchronizing it closer to the rate of high-throughput synthesis.

Origins of structural hole traps in hydrogenated amorphous silicon.

The inherently disordered nature of hydrogenated amorphous silicon (a-Si:H) obscures the influence of atomic features on the trapping of holes. To address this, we have created a set of over two thousand ab initio structures of a-Si:H and explored the influence of geometric factors on the occurrence of deep hole traps using density-functional theory. Statistical analysis of the relative contribution of various structures to the trap distribution shows that floating bonds and ionization-induced displacements correlate most strongly with hole traps in our ensemble.

Predicting Synthesizability using Machine Learning on Databases of Existing Inorganic Materials.

Defining the metric for synthesizability and predicting new compounds that can be experimentally realized in the realm of data-driven research is a pressing problem in contemporary materials science. The increasing computational power and advancements in machine learning (ML) algorithms provide a new avenue to solve the synthesizability challenge. In this work, using the Inorganic Crystal Structure Database (ICSD) and the Materials Project (MP) database, we represent crystal structures in Fourier-transformed crystal properties (FTCP) representation and use a deep learning model to predict synthesizability in the form of a synthesizability score (SC). Such an SC model, as a synthesizability filter for new materials, enables an efficient and accurate classification to identify promising material candidates. The SC prediction model achieved 82.6/80.6% (precision/recall) overall accuracy in predicting ternary crystal materials. We also trained the SC model by only considering compounds uploaded on the MP before 2015 as the training set and testing on multiple sets of materials uploaded after 2015. In the post-2019 test set, we obtain a high 88.60% true positive rate accuracy, coupled with 9.81% precision, indicating that newly added materials remain unexplored and have high synthesis potential. Further, we provide a list of 100 materials predicted to be synthesizable from this post-2019 dataset (highest SC) for future studies, and our SC model, as a validation filter, is beneficial for future material screening and discovery.

Insulator-to-metal transition in selenium-hyperdoped silicon: observation and origin.

Hyperdoping has emerged as a promising method for designing semiconductors with unique optical and electronic properties, although such properties currently lack a clear microscopic explanation. Combining computational and experimental evidence, we probe the origin of sub-band-gap optical absorption and metallicity in Se-hyperdoped Si. We show that sub-band-gap absorption arises from direct defect-to-conduction-band transitions rather than free carrier absorption. Density functional theory predicts the Se-induced insulator-to-metal transition arises from merging of defect and conduction bands, at a concentration in excellent agreement with experiment. Quantum Monte Carlo calculations confirm the critical concentration, demonstrate that correlation is important to describing the transition accurately, and suggest that it is a classic impurity-driven Mott transition.

High photocurrent in silicon photoanodes catalyzed by iron oxide thin films for water oxidation.

Silicon splits: The application of silicon to water oxidation is limited due to unfavorable interface properties. However, these can be circumvented by using a high-performance silicon photoanode with a catalytically active iron oxide thin film (see picture). This approach results in photocurrents as high as 17 mA cm(-2) under 1 sun and zero overpotential conditions.

Environmental Stability of Crystals: A Greedy Screening.

Discovering materials that are environmentally stable and also exhibit the necessary collection of properties required for a particular application is a perennial challenge in materials science. Herein, we present an algorithm to rapidly screen materials for their thermodynamic stability in a given environment, using a greedy approach. The performance was tested against the standard energy above the hull stability metric for inert conditions. Using data of 126 320 crystals, the greedy algorithm was shown to estimate the driving force for decomposition with a mean absolute error of 39.5 meV/atom, giving it sufficient resolution to identify stable materials. To demonstrate the utility outside of a vacuum, the in-oxygen stability of 39 654 materials was tested. The enthalpy of oxidation was found to be largely exothermic. Further analysis showed that 1438 of these materials fall into the range required for self-passivation based on the Pilling-Bedworth ratio.

A Machine Learning and Computer Vision Approach to Rapidly Optimize Multiscale Droplet Generation.

Generating droplets from a continuous stream of fluid requires precise tuning of a device to find optimized control parameter conditions. It is analytically intractable to compute the necessary control parameter values of a droplet-generating device that produces optimized droplets. Furthermore, as the length scale of the fluid flow changes, the formation physics and optimized conditions that induce flow decomposition into droplets also change. Hence, a single proportional integral derivative controller is too inflexible to optimize devices of different length scales or different control parameters, while classification machine learning techniques take days to train and require millions of droplet images. Therefore, the question is posed, can a single method be created that universally optimizes multiple length-scale droplets using only a few data points and is faster than previous approaches? In this paper, a Bayesian optimization and computer vision feedback loop is designed to quickly and reliably discover the control parameter values that generate optimized droplets within different length-scale devices. This method is demonstrated to converge on optimum parameter values using 60 images in only 2.3 h, 30× faster than previous approaches. Model implementation is demonstrated for two different length-scale devices: a milliscale inkjet device and a microfluidics device.

Extracting film thickness and optical constants from spectrophotometric data by evolutionary optimization.

In this paper, we propose a simple and elegant method to extract the thickness and the optical constants of various films from the reflectance and transmittance spectra in the wavelength range of 350 - 1000 nm. The underlying inverse problem is posed here as an optimization problem. To find unique solutions to this problem, we adopt an evolutionary optimization approach that drives a population of candidate solutions towards the global optimum. An ensemble of Tauc-Lorentz Oscillators (TLOs) and an ensemble of Gaussian Oscillators (GOs), are leveraged to compute the reflectance and transmittance spectra for different candidate thickness values and refractive index profiles. This model-based optimization is solved using two efficient evolutionary algorithms (EAs), namely genetic algorithm (GA) and covariance matrix adaptation evolution strategy (CMAES), such that the resulting spectra simultaneously fit all the given data points in the admissible wavelength range. Numerical results validate the effectiveness of the proposed approach in estimating the optical parameters of interest.

Discovery of temperature-induced stability reversal in perovskites using high-throughput robotic learning.

Stability of perovskite-based photovoltaics remains a topic requiring further attention. Cation engineering influences perovskite stability, with the present-day understanding of the impact of cations based on accelerated ageing tests at higher-than-operating temperatures (e.g. 140°C). By coupling high-throughput experimentation with machine learning, we discover a weak correlation between high/low-temperature stability with a stability-reversal behavior. At high ageing temperatures, increasing organic cation (e.g. methylammonium) or decreasing inorganic cation (e.g. cesium) in multi-cation perovskites has detrimental impact on photo/thermal-stability; but below 100°C, the impact is reversed. The underlying mechanism is revealed by calculating the kinetic activation energy in perovskite decomposition. We further identify that incorporating at least 10 mol.% MA and up to 5 mol.% Cs/Rb to maximize the device stability at device-operating temperature (<100°C). We close by demonstrating the methylammonium-containing perovskite solar cells showing negligible efficiency loss compared to its initial efficiency after 1800 hours of working under illumination at 30°C.

The Impact of COVID-19-Related Measures on the Solar Resource in Areas with High Levels of Air Pollution.

Restrictions enacted to reduce the spreading of COVID-19 have resulted in notably clearer skies around the world. In this study, we confirm that reduced levels of air pollution correlate with unusually high levels of clear-sky insolation in Delhi, India. Restrictions here were announced on March 19th, with the nation going into lockdown on March 24th. Comparing insolation data before and after these dates with insolation from previous years (2017 to 2019), we observe an 8.3% ± 1.7% higher irradiance than usual in late March and a 5.9% ± 1.6% higher one in April, while we find no significant differences in values from previous years in February or early March. Using results from a previous study, we calculated the expected increase in insolation based on measured PM2.5 concentration levels. Measurements and calculations agree within confidence intervals, suggesting that reduced pollution levels are a major cause for the observed increase in insolation.

Author Correction: How machine learning can help select capping layers to suppress perovskite degradation.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.