Cocrystal engineering have been widely applied in pharmaceutical, chemistry and material fields. However, how to effectively choose coformer has been a challenging task on experiments. Here we develop a graph neural network (GNN) based deep learning framework to quickly predict formation of the cocrystal. In order to capture main driving force to crystallization from 6819 positive and 1052 negative samples reported by experiments, a feasible GNN framework is explored to integrate important prior knowledge into end-to-end learning on the molecular graph. The model is strongly validated against seven competitive models and three challenging independent test sets involving pharmaceutical cocrystals, π–π cocrystals and energetic cocrystals, exhibiting superior performance with accuracy higher than 96%, confirming its robustness and generalization. Furthermore, one new energetic cocrystal predicted is successfully synthesized, showcasing high potential of the model in practice. All the data and source codes are available at https://github.com/Saoge123/ccgnet for aiding cocrystal community.

The harvesting of ‘hot’ triplet excitons through high-lying reverse intersystem crossing mechanism has emerged as a hot research issue in the field of organic light-emitting diodes. However, if high-lying reverse intersystem crossing materials lack the capability to convert ‘cold’ T₁ excitons into singlet ones, the actual maximum exciton utilization efficiency would generally deviate from 100%. Herein, through comparative studies on two naphthalimide-based compounds CzNI and TPANI, we revealed that the ‘cold’ T₁ excitons in high-lying reverse intersystem crossing materials can be utilized effectively through the triplet-triplet annihilation-mediated high-lying reverse intersystem crossing process if they possess certain triplet-triplet upconversion capability. Especially, quite effective triplet-triplet annihilation-mediated high-lying reverse intersystem crossing can be triggered by endowing the high-lying reverse intersystem crossing process with a ³ππ*→¹nπ* character. By taking advantage of the permanent orthogonal orbital transition effect of ³ππ*→¹nπ*, spin–orbit coupling matrix elements of ca. 10 cm⁻¹ can be acquired, and hence ultra-fast mediated high-lying reverse intersystem crossing process with rate constant over 10⁹ s⁻¹ can be realized.

Four novel iridium(III) complexes bearing biphenyl (7a–7c) or fluorenyl (7d) modified benzothiazole cyclometallate ligands are synthesized. In comparison with the yellow parent complex, bis(2-phenylbenzothiozolato-N,C^2′) iridium(III) (acetylacetonate) [(pbt)₂Ir(acac)] (λ_PLmax = 557 nm, φ_PL = 0.26), 7a–7d show 20–43 nm bathochromic shifted orange or red phosphorescence in solution, with maximum photoluminescence (PL) quantum yield of 0.62, and PL lifetime of 1.8–2.0 μs. Meanwhile, the resulting complexes also exhibit intense orange or red phosphorescence of λ_PLmax = 588–611 nm in solid films. The complex 7c with two tert-butyl substituents possesses the highest phosphorescent efficiency both in dilute solution and thin solid films, therefore may be a prospective candidate for both doping and host emitting electrophosphorescent material. Furthermore, despite the observation of severe oxygen quenching for 7a–7d in solution, 7a and 7c even show efficient emission intensity quenching by oxygen in their solid state due to the existence of void channels in crystals; consequently, they are promising molecular oxygen sensor reagents. Electrochemical measurement and DFT calculation results suggest that all these chelates own declined LUMOs of 0.1 eV relative to that of (pbt)₂Ir(acac) owing to the contribution of the phenyl substituents; whereas only 7d shows a more destabilized HOMO (∼0.1 eV) compared with the parent chelate.

Convolution neural network (CNN)-based detectors have shown great performance on ship detections of synthetic aperture radar (SAR) images. However, the performance of current models has not been satisfactory enough for detecting multiscale ships and small-size ones in front of complex backgrounds. To address the problem, we propose a novel SAR ship detector based on CNN, which consist of three subnetworks: the Fusion Feature Extractor Network (FFEN), Region Proposal Network (RPN), and Refine Detection Network (RDN). Instead of using a single feature map, we fuse feature maps in bottom–up and top–down ways and generate proposals from each fused feature map in FFEN. Furthermore, we further merge features generated by the region-of-interest (RoI) pooling layer in RDN. Based on the feature representation strategy, the CNN framework constructed can significantly enhance the location and semantics information for the multiscale ships, in particular for the small ships. On the other hand, the residual block is introduced to increase the network depth, through which the detection precision could be further improved. The public SAR ship dataset (SSDD) and China Gaofen-3 satellite SAR image are used to validate the proposed method. Our method shows excellent performance for detecting the multiscale and small-size ships with respect to some competitive models and exhibits high potential in practical application.

High-throughput design offers a promising way to expedite the de novo design of novel energetic molecules, but achieving this goal necessitates accurate methods for property prediction and efficient schemes for molecular screening. Two approaches for generating energetic molecules are proposed, based on a generative model and a fragment docking scheme, respectively. A high-throughput computation (HTC) workflow based on quantum chemistry is developed for energetic molecule design. Machine learning models are established for predicting crystal density, enthalpy of formation, R–NO₂ bond dissociation energy, detonation velocity, detonation pressure, detonation heat, detonation volume and detonation temperature, yielding coefficients of determination (R²) of 0.928, 0.948, 0.984, 0.989, 0.986, 0.986, 0.990 and 0.995, respectively. Thereby, an easy-to-use platform named Energetic Materials Studio (EM-Studio) integrates all the methods and models. Therein, five modules, EM-Generator, EM-QC, EM-DB, EM-ML and EM-Visualizer, work for molecule generation, HTC-aided molecule design, data management, machine learning prediction, and human–computer interaction, respectively. The effectiveness and capabilities of EM-Studio in HTC- and AI-aided molecular design are demonstrated through two cases of fused-ring energetic molecules.