Yongchul G. Chung

Selective oxidation of hydrocarbons requires highly selective catalysts toward desired reaction products. In this study, SiO2-supported Ti catalysts were synthesized featuring isolated Ti oxide species (0.2 Ti atoms/nm2) that serve as active Lewis acid sites for the selective oxidation of cyclohexene with H2O2. To investigate the effect of the surface microenvironment on catalytic performance, post-modification was performed using alcohols with different terminal groups (octyl (O), triglycol (TG), and fluorous (F)). FT-IR spectroscopy confirms that the deposited organic groups were successfully installed on the silica support. Furthermore, the change in the silanol peak (3730 cm–1) confirms that the introduced organic groups effectively minimize moisture contact on the catalyst surface, creating a hydrophobic/hydrophilic environment near the Ti active sites. In the cyclohexene epoxidation reaction (60 °C), the surface-modified catalysts increase the epoxide yield by an average of 25% compared to the unmodified catalyst. Notably, Ti-SiO2-O exhibits the highest direct pathway selectivity of 93.1%, representing a relative increase of 12% compared to the Ti-SiO2 catalyst (83.4%). Raman and XPS results confirm that the reaction proceeds via a titanium hydroperoxide (Ti-OOH) intermediate. Furthermore, density functional theory (DFT) calculations elucidate that the hydrophobic octyl groups provide an optimal balance between cyclohexene adsorption and product desorption. Conversely, the hydrophilic nature of the TG surface and the electron-rich characteristics of the fluorous chains result in the homolytic activation of H2O2. These findings demonstrate that tailoring the surface microenvironment via post-synthetic modification serves as a vital strategy for maximizing the efficiency of catalysts for selective oxidation reactions.

Digital reticular chemistry relies on accurate crystal structures to power computational screening, data‐driven discovery, and structure‐property analysis, yet recent studies reveal that more than half of the top‐performing candidates in major computational screening campaigns are chemically invalid. In experimental MOF databases, structural errors arise when disordered or incomplete structural models are incorrectly converted into fully specified simulation inputs. In hypothetical MOF database, structures are complete by construction but may encode chemically implausible oxidation states, coordination environments, or charge distributions. We term these erroneous structural models “structural demons.” This mini‐review asks three questions: where these errors enter, how we find them, and how we prevent them. On the prevention side, the key steps are keeping diffraction data and synthesis details together from the start, using consistent curation when structures enter a database, and filtering topology choices before structure generation. Connecting these steps can keep many bad structures out of downstream databases and reduce the need to fix them later.

Solid-state electrolytes (SSEs) are attractive for next-generation lithium-ion batteries due to improved safety and stability, but their low room-temperature ionic conductivity hinders practical application. Experimental synthesis and testing of new SSEs remain time-consuming and resource-intensive. Machine learning offers an accelerated route for SSE discovery; however, composition-only models neglect structural factors important for ion transport, while graph neural networks are challenged by the scarcity of structure-labeled conductivity data and the prevalence of crystallographic disorder in crystal structures (CIFs). Here, we train two complementary predictors on the same room-temperature, structure-labeled dataset (n = 499). A gradient-boosted tree regressor model using stoichiometric descriptors alone achieves a test MAE of 1.108 in log(S/cm); adding geometric descriptors (combined MAE = 1.172) does not lower the test error but reveals complementary structural information through Shapley Additive exPlanations, which shows that stoichiometric descriptors, particularly the oxygen ratio, dominate feature importance (seven of the top ten features), with three geometric descriptors (density, Lmax, and Lmin) also contributing meaningfully. In parallel, we fine-tune large language models (LLMs) using compact text prompts derived from CIF metadata (formula with optional symmetry and disorder tags), avoiding direct use of raw atomic coordinates. Notably, while Mistral-7B achieves the lowest absolute error [MAE = 0.798 in log(S/cm)], Qwen3-8B demonstrates the best overall ranking performance (SRCC = 0.849) using formula and disorder information, eliminating the need for numerical feature extraction from CIF files. Together, these results show that global geometric descriptors improve tree-based predictions and enable interpretable structure–property analysis, while LLMs provide a competitive low-preprocessing alternative for rapid SSE screening.

A multiscale framework integrating molecular modeling, PVSA process optimization, and techno-economic analyses evaluates CALF-20 derivatives for biogas upgrading, and identifies CALF-20 as the best with 9.35 kWh kg ⁻¹ energy consumption.

The presence of defects in metal-organic frameworks (MOFs) significantly impacts their adsorption and catalytic properties, yet noninvasive quantitative defect characterization remains challenging. In this study, we introduce a computational approach to estimate missing-cluster defect concentrations in UiO-66, a prototypical zirconium MOF, based on nitrogen adsorption isotherms at 77 K. By combining grand canonical Monte Carlo (GCMC) simulations and a statistical modeling framework, we construct composite models from unit-cell or 2 × 2 × 2 supercell pristine and defective models to reproduce the isotherms of larger test structures. We show that choosing 2 × 2 × 2 supercell models as the basis models outperformed unit-cell models by capturing a broader range of pore environments, enabling reliable defect prediction across a wide defect concentration range. This nondestructive, simulation-based method provides a generalizable platform for defect analysis in MOFs and other porous materials with well-defined local motifs.

The computational discovery and design of new crystalline materials, particularly metal-organic frameworks (MOFs), heavily rely on high-quality, "computation-ready" structural data. However, recent studies have revealed significant error rates within existing MOF databases, posing a critical data problem that hinders efficient high-throughput computational screening. While rule-based algorithms like MOSAEC, MOFChecker, and the Chen and Manz method (Chen-Manz) have been developed to address this, they often suffer from inherent limitations and misclassification of structures. To overcome this challenge, we developed MOFClassifier, a novel machine learning approach built upon a positive-unlabeled crystal graph convolutional neural network (PU-CGCNN) model. MOFClassifier learns intricate patterns from perfect crystal structures to predict a "crystal-likeness score" (CLscore), effectively classifying MOFs as computation-ready. Our model achieves an ROC value of 0.979 (previous best value 0.912) and, importantly, can identify subtle structural and chemical errors that are undetectable by current rule-based methods. By accurately recovering previously misclassified false-negative structures, the MOFClassifier reduces the risk of overlooking promising material candidates in large-scale computational screening campaigns. This user-friendly tool is freely available and has been integrated into the preparation workflow for the updated CoRE MOF DB 2025 v1.0, contributing to the accelerated computational discovery of MOF materials.