Rapid Data-Efficient Optimization of Perovskite Nanocrystal Syntheses through Machine Learning Algorithm Fusion

Applying machine learning and genetic algorithms accelerated for optimizing ethanol production

Corn straws can produce bioethanol via simultaneous saccharification and co-fermentation (SSCF). However, identifying optimal combinations of operating parameters from numerous possibilities through a cost-effective strategy to improve SSCF efficiency and yield remains challenging. The eXtreme Gradient Boost (XGB) and deep neural network (DNN) models were constructed to accurately predict ethanol yield from only five input variables, achieving >83 % accuracy. Subsequently, the XGB and the DNN models were merged with the genetic algorithm (GA) as the new optimization strategies. Experimental validation showed that the new strategy optimize the efficiency and yield of the SSCF ethanol production system quickly and accurately. Moreover, the potential optimization mechanism was investigated through the comprehensive interpretability analysis for XGB and the microbial ecology analysis. Enzyme Solution Volume (61.7 %) dominated, followed by time (12.9 %), substrate concentration (10.4 %), temperature (7.7 %), and inoculum volume (7.3 %). This efficient and accurate algorithm design strategy can significantly reduce the time required to optimize biochemical systems.

Antisolvent controls the shape and size of anisotropic lead halide perovskite nanocrystals

Colloidal lead halide perovskite nanocrystals have potential for lighting applications due to their optical properties. Precise control of the nanocrystal dimensions and composition is a prerequisite for establishing practical applications. However, the rapid nature of their synthesis precludes a detailed understanding of the synthetic pathways, thereby limiting the optimisation. Here, we deduce the formation mechanisms of anisotropic lead halide perovskite nanocrystals, 1D nanorods and 2D nanoplatelets, by combining in situ X-ray scattering and photoluminescence spectroscopy. In both cases, emissive prolate nanoclusters form when the two precursor solutions are mixed. The ensuing antisolvent addition induces the divergent anisotropy: The intermediate nanoclusters are driven into a dense hexagonal mesophase, fusing to form nanorods. Contrastingly, nanoplatelets grow freely dispersed from dissolving nanoclusters, stacking subsequently in lamellar superstructures. Shape and size control of the nanocrystals are determined primarily by the antisolvent's dipole moment and Hansen hydrogen bonding parameter. Exploiting the interplay of antisolvent and organic ligands could enable more complex nanocrystal geometries in the future.

Active Learning of Ligands That Enhance Perovskite Nanocrystal Luminescence

Ligands play a critical role in the optical properties and chemical stability of colloidal nanocrystals (NCs), but identifying ligands that can enhance NC properties is daunting, given the high dimensionality of chemical space. Here, we use machine learning (ML) and robotic screening to accelerate the discovery of ligands that enhance the photoluminescence quantum yield (PLQY) of CsPbBr3 perovskite NCs. We developed a ML model designed to predict the relative PL enhancement of perovskite NCs when coordinated with a ligand selected from a pool of 29,904 candidate molecules. Ligand candidates were selected using an active learning (AL) approach that accounted for uncertainty quantified by twin regressors. After eight experimental iterations of batch AL (corresponding to 21 initial and 72 model-recommended ligands), the uncertainty of the model decreased, demonstrating an increased confidence in the model predictions. Feature importance and counterfactual analyses of model predictions illustrate the potential use of ligand field strength in designing PL-enhancing ligands. Our versatile AL framework can be readily adapted to screen the effect of ligands on a wide range of colloidal nanomaterials.

Artificial Intelligence-Based, Wavelet-Aided Prediction of Long-Term Outdoor Performance of Perovskite Solar Cells

The commercial development of perovskite solar cells (PSCs) has been significantly delayed by the constraint of performing time-consuming degradation studies under real outdoor conditions. These are necessary steps to determine the device lifetime, an area where PSCs traditionally suffer. In this work, we demonstrate that the outdoor degradation behavior of PSCs can be predicted by employing accelerated indoor stability analyses. The prediction was possible using a swift and accurate pipeline of machine learning algorithms and mathematical decompositions. By training the algorithms with different indoor stability data sets, we can determine the most relevant stress factors, thereby shedding light on the outdoor degradation pathways. Our methodology is not specific to PSCs and can be extended to other PV technologies where degradation and its mechanisms are crucial elements of their widespread adoption.

Unveiling dominant recombination loss in perovskite solar cells with a XGBoost-based machine learning approach

Remarkable and intelligent perovskite solar cells (PSCs) have attracted substantial attention from researchers and are undergoing rapid advancements in photovoltaic technology. These developments aim to create highly efficient energy devices with fewer dominant recombination losses within the realm of third-generation solar cells. Diverse machine learning (ML) algorithms implemented, addressing dominant losses due to recombination in PSCs, focusing on grain boundaries (GBs), interfaces, and band-to-band recombination. The extreme gradient boosting (XGBoost) classifier effectively predicts the recombination losses. Our model trained with 7-fold cross-validation to ensure generalizability and robustness. Leveraging Optuna and shapley additive explanations (SHAP) for hyperparameter optimization and investigate the influence of features on target variables, achieved 85% accuracy on over 2 million simulated data, respectively. Because of the input parameters (light intensity and open-circuit voltage), the performance evaluation measures for the dominant losses caused by the recombination predicted by proposed model were superior to those of state-of-the-art models.

Closed-Loop Error-Correction Learning Accelerates Experimental Discovery of Thermoelectric Materials

The exploration of thermoelectric materials is challenging considering the large materials space, combined with added exponential degrees of freedom coming from doping and the diversity of synthetic pathways. Here, historical data is incorporated, and is updated using experimental feedback by employing error-correction learning (ECL). This is achieved by learning from prior datasets and then adapting the model to differences in synthesis and characterization that are otherwise difficult to parameterize. This strategy is thus applied to discovering thermoelectric materials, where synthesis is prioritized at temperatures <300 °C. A previously unexplored chemical family of thermoelectric materials, PbSe:SnSb, is documented, finding that the best candidate in this chemical family, 2 wt% SnSb doped PbSe, exhibits a power factor more than 2× that of PbSe. The investigations herein reveal that a closed-loop experimentation strategy reduces the required number of experiments to find an optimized material by a factor as high as 3× compared to high-throughput searches powered by state-of-the-art machine-learning (ML) models. It is also observed that this improvement is dependent on the accuracy of the ML model in a manner that exhibits diminishing returns: once a certain accuracy is reached, factors that are instead associated with experimental pathways begin to dominate trends.

Utilizing Data-Driven Optimization to Automate the Parametrization of Kinetic Monte Carlo Models

Kinetic Monte Carlo (kMC) simulations are a popular tool to investigate the dynamic behavior of stochastic systems. However, one major limitation is their relatively high computational costs. In the last three decades, significant effort has been put into developing methodologies to make kMC more efficient, resulting in an enhanced runtime efficiency. Nevertheless, kMC models remain computationally expensive. This is in particular an issue in complex systems with several unknown input parameters where often most of the simulation time is required for finding a suitable parametrization. A potential route for automating the parametrization of kinetic Monte Carlo models arises from coupling kMC with a data-driven approach. In this work, we equip kinetic Monte Carlo simulations with a feedback loop consisting of Gaussian Processes (GPs) and Bayesian optimization (BO) to enable a systematic and data-efficient input parametrization. We utilize the results from fast-converging kMC simulations to construct a database for training a cheap-to-evaluate surrogate model based on Gaussian processes. Combining the surrogate model with a system-specific acquisition function enables us to apply Bayesian optimization for the guided prediction of suitable input parameters. Thus, the amount of trial simulation runs can be considerably reduced facilitating an efficient utilization of arbitrary kMC models. We showcase the effectiveness of our methodology for a physical process of growing industrial relevance: the space-charge layer formation in solid-state electrolytes as it occurs in all-solid-state batteries. Our data-driven approach requires only 1-2 iterations to reconstruct the input parameters from different baseline simulations within the training data set. Moreover, we show that the methodology is even capable of accurately extrapolating into regions outside the training data set which are computationally expensive for direct kMC simulation. Concluding, we demonstrate the high accuracy of the underlying surrogate model via a full parameter space investigation eventually making the original kMC simulation obsolete.

Introduction to new horizons in materials for energy conversion, optics and electronics

In conjunction with the Emerging Investigator Forum celebrating the 120^th anniversary of Southeast University, we herein present a collection of articles focused on the energy conversion, optics, and electronics applications of (nano)materials.

Optical anisotropy of CsPbBr3 perovskite nanoplatelets

The two-dimensional CsPbBr3 nanoplatelets have a quantum well electronic structure with a band gap tunable with sample thicknesses in discreet steps based upon the number of monolayers. The polarized optical properties of CsPbBr3 nanoplatelets are studied using fluorescence anisotropy and polarized transient absorption spectroscopies. Polarized spectroscopy shows that they have absorption and emission transitions which are strongly plane-polarized. In particular, photoluminescence excitation and transient absorption measurements reveal a band-edge polarization approaching 0.1, the limit of isotropic two-dimensional ensembles. The degree of anisotropy is found to depend on the thickness of the nanoplatelets: multiple measurements show a progressive decrease in optical anisotropy from 2 to 5 monolayer thick nanoplatelets. In turn, larger cuboidal CsPbBr3 nanocrystals, are found to have consistently positive anisotropy which may be attributed to symmetry breaking from ideal perovskite cubes. Optical measurements of anisotropy are described with respect to the theoretical framework developed to describe exciton fine structure in these materials. The observed planar absorption and emission are close to predicted values at thinner nanoplatelet sizes and follow the predicted trend in anisotropy with thickness, but with larger anisotropy than theoretical predictions. Dominant planar emission, albeit confined to the thinnest nanoplatelets, is a valuable attribute for enhanced efficiency of light-emitting devices.