In this work, we report the discovery of energy cocrystals using an efficient iterative workflow combining an evolutionary algorithm and a machine learning potential (MLP). The compound 2,4,6,8,10,12-Hexanitro-2,4,6,8,10,12-hexaazaisowurtzitane (CL-20) has attracted significant attention owing to its higher energy density than traditional energetic materials. However, the higher sensitivity has limited its applications. An important way to reduce its sensitivity involves cocrystal engineering with traditional explosives. Many cocrystal structures are expected to be composed of these two components. We developed an efficient iterative workflow to explore the phase space of CL-20 and 1,3,5,7-tetranitro-1,3,5,7-tetraazacyclooctane (HMX) cocrystals using an evolutionary algorithm and an MLP. The algorithm was based on the Universal Structure Predictor: Evolutionary Xtallography (USPEX) software, and the MLP was the reactive force field with neural networks (ReaxFF-nn) model. A set of high-density cocrystal structures was produced through this workflow; these structures were further checked via first-principles geometry optimizations. After careful screening, we identified several high-density cocrystal structures with densities of up to 1.937 g/cm³ and HMX: CL-20 ratios of 1:1 and 1:2. The influence of hydrogen bonds on the formation of high-density cocrystals was also discussed, and a roughly linear relationship was found between energy and density.

The growth of zinc dendrites limits the practical application of zinc ion batteries, which can be effectively suppressed by surface doping. Herein, the density functional theory combined with symbolic regression algorithm had been used to study the growth of zinc nuclei on 22 single atom-doped surfaces. The results indicate that the doping surfaces with convex structure can suppress the zinc dendrite growth because of the weak adsorption energy and low diffusion activation energy of zinc atoms. Moreover, the diffusion activation energy and the orientation of zinc nucleation depend on the adsorption energy of the first zinc atom. The larger the adsorption energy, the greater the diffusion barrier of zinc atoms and the greater the tendency for vertical growth of zinc nuclei. Therefore, the symbolic regression algorithm was utilized to identify the relationship between the adsorption energy of the first zinc atom and the properties of the doped atom. It was found that the radius and d-band center of doped atoms are key factors affecting the adsorption energy of the first zinc atom, and the doped atom with large atom radius and low d-band center can inhibit the zinc dendrite growth. Finally, the Al, Ag, Cd, In, Sn, Au, Hg, Tl, and Bi atoms are screened out to be the promising doping single atoms that can suppress the zinc dendrite growth.

Light-induced segregation limits the practical application of mixed halide perovskites in solar cells. Herein, halide segregation is evaluated by a data-driven approach with constructing a bandgap database of 53,361 mixed ABX₃ [where A = Cs, formamidinium (FA) or methylammonium (MA); B = Pb or Sn; X = Br, Cl, or I] perovskites. A transfer learning strategy was employed to fine-tune the parameters of a Graph Neural Network model using experimental and density functional theory (DFT)-calculated bandgaps. This approach accelerated the construction of a unique database, distinguishing it from others primarily focused on ABX₃ perovskite element substitution. The database is characterized by continuously varying compositions and accurate bandgaps. It was utilized to calculate the free energy of 20,688 mixed iodine-bromine perovskites and generate corresponding phase diagrams for predicting their light-induced segregation behavior. It is found that the bandgap increases with decreasing ionic radii at the A-site and X-site. This composition-dependent bandgap difference drives halide segregation. Moreover, using a higher Cs content at the A-site, rather than MA, reduces this bandgap difference, enhancing photostability. The proposed data-driven strategy can facilitate the targeted design of novel perovskites with mixed compositions and the investigation of halide perovskite segregation.

To address the issues with molecular representation of copolymerized polyimides (PIs) and the mini dataset of PI powders. We constructed an interpretable machine learning (ML) model for PI films using the weighted-additive Morgan Fingerprints with Frequency descriptors and developed an interpretable transfer learning model for PI powders. To enhance Thermal Stability (Temperature at 5% weight loss) of PI films and powders, it is recommended to add conjugated functional groups to diamines, control phenyl ring side chains, and reduce pyridine and hydroxyl groups; select copolyimides (co-PIs); ensure that anhydride is directly connected to the benzene ring in dianhydrides, avoiding aliphatic cycles. It is noteworthy that the close alignment between experimental results and model predictions serves to confirm the model is a reliable prediction tool. It is hoped that this polymer informatics approach will provide further implementation for practical applications of other functional materials.

Lithium-ion battery (LIB) health prognosis is essential for ensuring the safety of electric vehicles while they are in use. However, conventional approaches for accurate health state forecasting face challenges due to the complex interplay of battery degradation mechanisms and the significant variability in operating conditions during cycling. In this study, we propose a data-driven method composed of convolutional neural networks (CNNs) and bidirectional long short-term memory (BiLSTM) to accurately predict the state of health and remaining useful life of LIBs. The model is trained using a well-established open-source electrochemical impedance spectroscopy (EIS) database. This database includes over 20,000 EIS spectra from commercial LIBs, collected under various states of health, states of charge and temperatures. The CNN-BiLSTM model surpasses the previous state-of-the-art Gaussian process method in current capacity estimation and remaining useful life prediction. Furthermore, we showcase the model’s capability to forecast the capacity degradation trajectory of a cell using its early-cycle EIS data. Our research demonstrates the versatility of the battery forecasting method by integrating EIS with machine learning, and emphasizes the value of implementing the EIS-based artificial approach in a battery management system.