Band-gap screening requires a calculation policy because agreement and cost vary with chemistry and computational protocol. The observation target is one literature gap compiled for each benchmark record across optical and fundamental measurements, temperatures, and sample conditions. A depth-two decision tree using band-edge orbital character reduced mean absolute deviation from this compiled target by 35.5 meV (95% confidence interval, 19.9-52.4 meV) relative to uniform mBJ while assigning HSE06 to 18.7% of 461 compounds. No equal-count random policy matched the router among 50,000 draws (p = 2.0 × 10-5). Complete retraining under global target shifts up to ±100 meV and under 25-100 meV record- or source-correlated jitter retained a positive allocation gain. Under leave-element-out transfer, the compound-level gain remained 24 meV (95% confidence interval, 10-38 meV) and was concentrated in chalcogenides and oxides. At the same HSE06 count, gap- and composition-based learned rules matched the orbital router. The framework therefore links allocation, admission, observation-target sensitivity, and chemical transfer in one deployable screening rule.
Structures built for or generated from Molecular Dynamics simulations can often have problems such as wrong chiralities, bad parameter assignment, steric clashes, and/or distorted geometries. It is important to be able to detect such problems in a rapid and automated way since such issues can prevent subsequent simulations from running successfully. Here we detail a structure check procedure which focuses on checking for some of the more common structural issues, including atomic overlaps, distorted bonds, and ring intersections, as well as the strategies for parallelizing the check procedure so it can rapidly detect issues in structures with millions of atoms.
Thermodynamic quantities in the hydrated state provide essential reference information for understanding hydrothermal biomass decomposition. However, accurately estimating thermodynamic quantities in the hydrated state using computational methods remains challenging. In this study, we develop a general computational framework to evaluate accurate solution-phase thermodynamics in the standard state for compounds involved in the decomposition of xylose and glucose in water by combining gas-phase free-energy calculations with machine-learning prediction of hydration free energies. Gas-phase thermodynamic quantities are evaluated by a composite method designed to reproduce coupled-cluster theory with singles, doubles, and perturbative triples in the complete basis set limit (CCSD(T)/CBS) accuracy. Hydration free energies are predicted using molecular descriptors and machine learning. The resulting approach enables efficient evaluation of large reaction networks and is demonstrated for a dataset comprising 84 molecules and 73 reactions. Comparison with experimentally-derived free-energy differences for sugar isomerization indicates that the method achieves an accuracy within 3.7 kcal/mol in aqueous solution. The computed thermodynamic data further identify furfural and 5-hydroxymethylfurfural as thermodynamically favorable products in the xylose and glucose systems, respectively. This framework provides a practical route to accurate standard-state solution-phase thermodynamics by combining high-accuracy gas-phase thermochemistry with machine-learning prediction of hydration free energies.
Density Functional Theory calculations were performed to design atomistic models of ruthenium surfaces interfaced with partially dehydrogenated lithium-based layers (LiNH2, Li2NH, and Li3N). The structural, energetic, and magnetic properties of Li2NH and Li3N bulks and surfaces were systematically evaluated by using the GGA-PBE and dispersion-corrected PBE-dDsC functionals. The surface-termination analysis showed that the (100) orientations of these compounds minimize the lattice mismatch with Ru(0001), favoring interface formation. Based on these results, a multilayer model, Ru(0001)/Li3N(001)/Li2NH(001)/LiNH2(001), was developed to describe the progressive dehydrogenation of LiNH2 in contact with Ru. The partially dehydrogenated interface showed a greater thermodynamic stability than fully hydrogenated systems. Energetic and magnetic analyses indicated that progressive dehydrogenation enhances interfacial stability, allowing the suggestion of mechanistic assumptions for ammonia decomposition. Overall, this study offers an atomistic perspective on Li-N-H/Ru interfaces and a basis for designing improved catalysts for ammonia decomposition and hydrogen storage.
Inspired by the successful synthesis of zig-zag C2h-N6 (Nature 2025, 642, 356-360), we systematically explored the structure, stability, infrared spectroscopy, chemical bonding and weak interaction of novel X6 (X = N, P, As, Sb, Bi) molecules using relativistic DFT and CCSD(T) method. The results of global-minimum structural searches indicate the N6 prefers a trimer-like C3 geometry, while P6/As6 tends to form the bag-like C2v structure, and Sb6/Bi6 prefers the triangular prism D3h structure energetically instead. Simulated infrared spectra reveal the trimer-like N6 exhibiting an isolated high-frequency triple-bond stretching band far separated from the low-frequency single-bond skeletal vibration regimes of heavier pnictogen hexamers. Canonical molecular orbital analyses reveal a non-monotonic periodic evolution of HOMO-LUMO electronic band gaps, with N6, Sb6 and Bi6 possessing substantially wider frontier orbital energy separation relative to narrow-gap bag-like P6 and As6 species. Multi-modal wavefunction analytical tools including AdNDP, QTAIM, ELF, IRI and RDG/NCI resolve continuous periodic evolution of covalent bonding motifs from discrete diatomic triple bonds in N6, mixed σ and π bonding in P6 and As6, to pure localized σ single-bonded frameworks devoid of π orbital contributions in Sb6 and Bi6. Quantified dissociation energy values for two representative uni-molecular fragmentation pathways demonstrate monotonically increasing resistance to full decomposition from N6 to Bi6. Additional SOC-CASSCF energy level calculations further quantify the progressive intensification of spin-orbit coupling relativistic effects down group 15, which exert negligible electronic perturbation on N6 and P6, moderate orbital splitting for As6, and decisive rearrangement of frontier orbital manifolds for Sb6 and Bi6 that dictates geometric stability preferences of heavy hexamer prism. This work systematically reveals the periodic evolution law of group VA X6 molecules, providing key theoretical insights for the design and synthesis of novel pnictogen clusters.
We describe TStrail, a transition state (TS) conformational sampling tool that automatically generates TS conformers without fixing the bond forming/breaking distance of an initial TS structure. Conventionally, the conformations of TS structures were searched with the fixed length of active bonds, making it easier to re-optimize to the TS but potentially limiting the comprehensiveness of TS conformers when the active bond lengths varied among conformers. By collectively performing pathway searches among various reactant conformations, TStrail was able to generate TS conformers with different active bond lengths. Furthermore, applications to more complex systems demonstrated that TStrail provides a practical tool for systematically analyzing challenging TS conformers, particularly for large, flexible, ionic, and structurally complex TSs, thereby enabling deeper mechanistic insights through the identification of diverse TS conformers and reaction pathways.
Recently, [JCP 161, 114114 (2024)] successfully derived numerically stable and converging (to the time-dependent complete active space self-consistent-field) time-dependent orthogonal coupled cluster (CC) methods, namely, time-dependent orbital optimized CC (TD-ooCC) methods for fermion-mixtures with arbitrary fermion kinds and numbers, via the time-dependent bivariational principle and setting constraints in the ansatz. In this work, we extend the TD-ooCC methods to eighteen non-bivariational time-dependent orbital-quasi-optimized coupled-cluster (TD-oqoCC-NB) methods. In TD-oqoCC-NB methods, the constraints in the ansatz are also used to avoid overparameterization in the complete active space limit and eliminate the numerical instability in the finite truncation case. The equations of motion are not determined by the time-dependent bivariational principle. Instead, all equations ensuring the convergence to the complete active space limit can be selected flexibly. The applications to electron dynamics and vibrational dynamics are also discussed. In particular, we find that all NB methods violate the Ehrenfest theorem even when they are gauge-invariant, which distinguishes them from TD-ooCC methods. The stationary condition of EOMs of each NB methods also provides novel CC methods for time-independent electronic structure calculations.
The microsolvation of pyridine (C5H5N) by water molecules was investigated systematically through a comprehensive exploration of the conformational landscape of the PYR(H2O) n clusters ( n = 1 - 10 ) at the MN12SX-D3/def2-TZVP level of density functional theory, selected between 26 functionals, including dispersion corrections, by benchmarking to the DLPNO-CCSD(T1)/CBS. A total of 326 stationary points were characterized, ranging from 3 conformers at n = 1 to 55 at n = 9 . For small cluster sizes ( n = 1 - 2 ), the global minimum is dominated by a single O-H ⋯ N hydrogen bond between water and the pyridine nitrogen, in agreement with microwave spectroscopy data. From n = 3 onwards, cooperative O-H ⋯ O inter-water networks emerge and the energy landscape flattens dramatically, yielding dense manifolds of quasi-degenerate conformers. For n ≥ 7 , the water subcluster progressively adopts closed-cage topologies characteristic of large water clusters. Temperature-dependent Boltzmann populations ( 0 - 500 K) reveal that the global minimum is the sole populated species only for n ≤ 2 , while a broad conformational distribution governs larger clusters at ambient temperature. The incremental clustering energies converge toward a bulk-like plateau of approximately -10 kcal . mol - 1 beyond n = 5 , and the incremental Gibbs free energy remains positive throughout the series, demonstrating that stepwise gas-phase microsolvation is entropically disfavored at every hydration step. Quantum Theory of Atoms in Molecules (QTAIM) analysis of the global minima identifies five types of non-covalent interactions: O-H ⋯ N, O-H ⋯ O, C-H ⋯ O, O ⋯ C, and O-H ⋯ π contacts, whose diversity and number grow systematically with cluster size, reflecting the progressive transition from a simple donor-acceptor complex to a three-dimensional hydrogen-bond network that partially encapsulates the aromatic ring. These results provide a unified picture of the sequential microsolvation of pyridine and establish a benchmark dataset for the development of force fields and solvation models for nitrogen-containing heterocycles.
ABSTRACT In this work, we performed crystal structure searches for rubidium and cesium carbonates (RbCO and CsCO) in the pressure range of 0–100 GPa using evolutionary algorithms based on the density functional theory. As a result, two new stable high‐pressure polymorphs, and , were predicted for both carbonates. The MCO‐ (M = Rb, Cs) phase is isostructural with the high‐pressure KCO‐ phase, whereas the MCO‐ phase has no known structural analogs and represents a novel phase for alkali metal carbonates. A common sequence of phase transitions was established for both compounds: . For RbCO, the transition pressures are 4.9 GPa and 23.4 GPa, and for CsCO, they are 5.2 GPa and 35.5 GPa. The phonon calculations confirmed the dynamic stability of all predicted phases. It was also shown that these high‐pressure phases cannot be quenched to ambient pressure. Within the quasi‐harmonic approximation, P–T phase diagrams of RbCO and CsCO were constructed for the first time, revealing a weak temperature dependence of the phase boundaries. The obtained results elucidate a general trend in the phase transitions of alkali metal carbonates: under high pressure, the cation sublattices of all alkali carbonates tend to adopt an AlB‐type configuration, with the transition pressure increasing systematically with the ionic radius of the cation.
ABSTRACT The growing demand for sustainable hydrogen production has intensified research into photocatalytic water splitting driven by solar energy. Graphitic carbon nitride (g‐C 3 N 4 ), a metal‐free polymeric semiconductor with visible‐light activity, tunable electronic structure, low cost, and environmental compatibility, has emerged as a promising photocatalyst. Nevertheless, pristine g‐C 3 N 4 is limited by insufficient visible‐light absorption, sluggish charge transport, rapid electron–hole recombination, and low surface reactivity. Heterojunction engineering has therefore become a key strategy to overcome these intrinsic limitations and enhance photocatalytic efficiency. Distinct from conventional experimental or performance‐oriented reviews, this work presents a computationally driven and methodology‐oriented review that explicitly addresses how to model g‐C 3 N 4 ‐based heterojunction photocatalysts using density functional theory (DFT). Rather than merely summarizing reported efficiencies, this review provides a step‐by‐step conceptual and practical framework that enables readers to rationally construct, analyze, and optimize heterojunction photocatalysts at the atomic scale. The review begins with the structural and electronic fundamentals of g‐C 3 N 4 , followed by a mechanistic overview of photocatalytic water splitting, including light absorption, charge generation, interfacial charge separation, and surface redox reactions. Various heterojunction architectures: Type I, Type II, Z‐scheme, S‐scheme, p–n junctions, Schottky interfaces, and multicomponent systems are systematically discussed, with particular emphasis on interfacial charge‐transfer mechanisms revealed by first‐principles calculations. A key contribution of this review is the consolidation of DFT‐based modeling protocols and descriptors essential for evaluating and enhancing photocatalytic performance. These include structural stability (lattice mismatch, phonon dispersion, and ab initio molecular dynamics), electronic properties (band structure, density of states, and band‐edge alignment), charge‐transfer characteristics (work function, charge density difference, planar‐averaged charge density, and Bader charge analysis), optical properties (absorption spectra, dielectric function, optical band gap, and solar‐to‐hydrogen efficiency), and carrier transport descriptors (effective mass and carrier mobility). In addition, DFT‐based reaction pathway analysis is discussed to elucidate hydrogen and oxygen evolution mechanisms at heterojunction interfaces. Importantly, this review highlights computational strategies to enhance photocatalytic activity, including strain engineering, external electric fields, defect and dopant engineering, and interface optimization, providing clear guidance on how these approaches can be implemented and interpreted within a DFT framework. Challenges related to computational cost, finite‐size effects, and realistic interface construction are critically evaluated, along with practical mitigation strategies. Emerging directions such as beyond‐DFT methods (GW and TDDFT), machine learning, and high‐throughput screening are also discussed as powerful tools for accelerating heterojunction discovery. By integrating model construction principles, computational descriptors, and activity‐enhancement strategies into a unified roadmap, this review serves as a practical guide for researchers seeking to design, model, and optimize g‐C 3 N 4 ‐based heterojunction photocatalysts using DFT, thereby bridging the gap between theoretical modeling and experimental realization of efficient, stable, and scalable systems for solar‐driven hydrogen production.
We introduce a hybrid data-driven coupled-cluster (DDCC) framework that enables efficient coupled-cluster singles and doubles (CCSD)-level geometry optimizations through direct prediction of wavefunction parameters. In conventional CCSD, each optimization step requires solving the nonlinear t -amplitude equations and the linear Λ -equations, both scaling as O ( N 6 ) , making geometry optimizations computationally demanding. The Λ data-driven CC ( Λ DDCC) scheme replaces these iterative solvers with neural network predictions of the t and λ amplitudes, thereby removing the dominant computational bottleneck. The model utilizes a physically grounded representation of electron correlation, combining excitation-based descriptors with a molecular orbital (MO) framework defined through orbital centroids. This design enables accurate recovery of the CC wave function parameters, which in turn yields reliable nuclear gradients throughout the optimization process. Across a diverse set of thermally perturbed geometries and out-of-sample molecules, Λ DDCC reproduces CCSD geometries with near-quantitative agreement, while reducing computational cost by over an order of magnitude. These results establish Λ DDCC as an effective and transferable strategy for accelerating high-level electronic structure calculations, providing a practical route toward routine CCSD-quality geometry optimization in larger and more complex molecular systems.
In atomic layer deposition (ALD), precursor adsorption plays a critical role in determining growth characteristics. This is particularly important in area-selective ALD (AS-ALD), where adsorption on the non-growth surface (NGS) leads to selectivity loss. In this context, both the ability of a precursor to access reactive surface groups on the NGS and the strength of its binding are key parameters governing the selectivity in experiment. Adsorption on the NGS is strongly influenced by the precursor's substituents, which control Lewis acidity, steric demand, and propensity for dimerization. We applied energy decomposition analysis for extended systems (pEDA) to a set of commonly used aluminum precursors to quantitatively assess how these factors affect adsorption energetics on SiO2 as the NGS, with trimethoxypropylsilane (TMPS) as a small-molecule inhibitor (SMI). Using fluoride ion affinity (FIA) as a metric for Lewis acidity, we find that this largely governs the trends in precursor bonding to both the surface and the SMIs. Notable exceptions are precursors bearing branched alkyl substituents for which steric repulsion dominates over Lewis acidity. Lewis acidity furthermore determines whether adsorption between the SMI layer remains energetically favorable. In this regime, steric effects-Pauli repulsion and deformation of the SMI layer-substantially weaken the adsorption. Dimerization generally reduces adsorption energies. This weakening effect is less pronounced, however, when adsorption takes place within the inhibitor layer. Based on these insights on bonding energetics, we recommend targeted modifications of existing aluminum precursors and propose systematic strategies for precursor design aimed at minimizing interactions with the NGS.
Although popular throughout the 1990s and early 2000s, interest in Al2H2 suddenly diminished, leaving an unresolved chasm between theory and experiment for over 20 years. Our computations reconcile the two, unveiling evidence for the unidentified but earlier observed Al2H2 butterfly global minimum in matrix IR. We confirm the experimental IR assignments of the cyclic and monobridged structures. The critical ingredient in our case is the b2 symmetry Al-H stretching frequency predicted here to be 1096 cm-1. This feature was observed at 1090 cm-1 by Andrews et al., but never identified. Andrews et al. also assigned a feature at 1647 cm-1 to the linear structure of Al2H2. However, this is unlikely, as we do not find the linear structure to be an equilibrium geometry. CCSD(T) geometries and energies for four neutral Al2H2 isomers in this work are computed at the highest levels of theory among theoretical investigations in the literature. For final energetic predictions, basis sets as large as augmented quadruple zeta were used, as were theoretical methods as advanced as CCSDT(Q). We also report Al2H2 VPT2 fundamental frequencies for the first time. In addition, two anionic structures are investigated at the same levels of theory, the butterfly ground state and monobridged local minimum, which have never been characterized. We conclude that these anion structures could be identified in future mass-selected gas-phase spectroscopy experiments.
Predicting reaction minimum-energy paths (MEPs) and locating transition states (TSs) on the potential energy surface (PES) are challenging problems. The high scaling of electronic structure methods and the computational expense of higher-energy derivative calculations limit the exploration of reaction mechanisms to smaller molecular systems. Therefore, tracing the MEPs along the intrinsic reaction coordinate is formidable. In this work, we have benchmarked a rapid method, the Feature-vector Force-driven Reaction Coordinate Exploration (FFoRCE) approach, for generating MEPs connecting the reactant to the product on the PES. This two-step approach involves a reaction-coordinate navigator that generates the initial path from reactant to product, followed by local relaxations of each structure along the sampled path. The reaction coordinate navigator transforms the coordinate space into a set of features and assigns each structure along the reaction coordinate path a unique value. The reaction coordinate navigator generates an initial path iteratively, followed by a symmetry function-constrained geometry optimization. In this process, the MEP structures are generated by minimizing the total energy along the TS. In this work, the performance of the FFoRCE approach is benchmarked on 10 elementary gas-phase organic reactions by tracing MEPs and locating TSs. Furthermore, we compared their geometries, activation energies, and vibrational frequencies with those from whole-molecule calculations. A fair agreement between them shows that the FFoRCE approach reproduces the MEP and locates the TS as accurately as DFT methods, and it is promising for future applications to complex reactions.
Ultrafast excited-state dynamics of two Co(III) complexes, bearing different alkyl groups (Co1: Methyl, Co2: Ethyl) at the nitrogen of the imidazole ligand, are investigated using trajectory surface hopping based on a linear vibronic coupling model. Despite nearly identical MLCT absorption spectra, experimental singlet lifetimes differ markedly (1.45 vs. 4.3 ps). Relaxation occurs via rapid internal conversion within a strongly vibronically coupled singlet manifold, accompanied by intersystem crossing to long-lived 3 MC states. The prolonged lifetime of Co2 originates from enhanced vibronic coupling and larger structural distortions, which trap population in the singlet manifold rather than from spin-orbit effects. Minimum-energy crossing points reveal low barriers for nonradiative decay, rationalizing emission quenching. These results demonstrate how subtle structural differences, by modulating vibronic coupling, can substantially influence excited-state lifetimes in these complexes.
On-water catalysis is widely exploited, yet the microscopic origin of rate enhancements at aqueous interfaces remains debated. In the spirit of the second-generation Car-Parrinello method, we use a modified Langevin-type scheme to accelerate density-functional based tight-binding molecular dynamics in order to investigate the free energy of two Diels-Alder reactions on-water: the cycloaddition of cyclopentadiene with ethyl cinnamate and that with ethyl thionocinnamate. The only difference between the two studied molecules is that the carbonyl oxygen of the cinnamate is a thiocarbonyl sulfur in the thionocinnamate; hence, the first is a strong and the second a weak hydrogen bond acceptor. The oxygen-containing dienophile reacts through an almost synchronous transition state with a barrier of ~13 kcal/mol, whereas the sulfur analogue follows an asynchronous pathway with a barrier of ~24 kcal/mol because a transient SC contact competes with CC bond formation. We find no increase in the number or strength of hydrogen bonds at the transition state. Instead, the catalytic role of the interface is to maintain an electronically productive hydrogen-bond environment that stabilizes charge reorganization along the reaction coordinate. These results indicate that, for the present on-water reactions, hydrogen-bond quality and charge-transfer capability are more decisive than the hydrogen-bond count alone.
We present an extension of our canonical energy decomposition analysis (EDA) from the gas phase to a Solution-Phase approach (SP-EDA), that is, to interactions between molecular fragments in the condensed phase, simulated using the Conductor-like Screening Model (COSMO). To this end, we have introduced different solvation states that serve to simulate particular physical phenomena that can be associated with the step from separate solvated fragments to the solvated complex of these fragments. In addition to the well-known deformation strain of the fragments, induced by complex formation, one can distinguish the desolvation of the contact points via which the fragments bind, the response of the solvent to the deformation of the fragments as well as the response of the solvent to the formation of the complex, and finally the complex formation between the partially desolvated fragments. We provide definitions of how to model these phenomena using a continuum model, such as COSMO. Finally, we analyze, for representative model systems, how the bonding and, in particular, the EDA terms of electrostatic attraction, Pauli repulsion, and stabilizing orbital interactions behave across the above steps associated with interactions in solution. An important practical conclusion is that neglecting the effect of the partial desolvation at the binding sites has, in most cases, only minor consequences for the values and trends in the EDA terms. This finding can be used to simplify investigations using SP-EDA.
We present a relativistic density-functional and explainable machine learning study of the vertical electron affinity (EA) of 14 homologous actinide(IV) carboxylate complexes [M(L)3]+ (M = Th-Lr; L = propionate, acrylate). All values were obtained at the MN12-L level via the def2-TZVPPD basis set for H, C, and O and a Stuttgart small-core relativistic effective-core potential, together with its associated valence basis, for the actinide. For every (metal, ligand) pair, all physically accessible An(IV) multiplicities were considered (52 calculations in total), and the ground state was assigned to the multiplicity with the lowest total electronic energy. The EA, evaluated as the negative of the lower α- and β-LUMO eigenvalues (the "true LUMO"), increases monotonically across the series from 5.53 eV at Th(5f0) to 10.80 eV at Lr(5f13) for the propionates (5.95 → 10.58 eV for the acrylates), with one outlier: at Th, the LUMO is forced into the 7 s/7p-ligand manifold because no 5f acceptor is available. Squared coefficient population analysis of the corrected ground-state wavefunctions confirms that for Pa-No, the LUMO is predominantly metal-5f (59%-82% on the metal), whereas at Lr, the highest-energy 5f spin orbital mixes strongly with the carboxylate π* manifold (5f ≈20%-28%, O ≈42%-50%). A nine-feature explainable machine-learning model trained on the corrected dataset reproduces the EA series under leave-one-metal-out cross-validation with R2 = +0.70 and MAE = 0.56 eV; permutation and SHapley additive exPlanations (SHAP) analyses identify the 5f-electron count and the atomic spin-orbit constant as the single dominant predictor cluster (Pearson r = +0.94 for the 5f count). The cross-ligand transferability is excellent (R2 = 0.85-0.88, MAE = 0.35-0.43 eV). The combined DFT-ML analysis provides a quantitative reference for An (IV) carboxylate EAs and isolates the 5f-shell occupation as the single physical lever governing their reduction propensity.
Blind docking is a method for predicting a binding mode of a ligand with a protein without any prior information about a binding site. Some tools allow this type of docking experiment directly, others, including some established tools, require binding site information being passed as an input. In this latter case, one can use cavity prediction tools and use the results of their prediction as an input in these docking calculations. However, it is still unclear if the results of these predictions can be reliably used in protein-ligand docking and what is the best technical way to pass this information to the docking algorithm. In this study we estimated the applicability of the binding pocket prediction tools in docking experiments to address this gap in knowledge. We use four different computational tools for cavity prediction and use the best predicted cavities represented in different ways to run GOLD docking calculations. Analysis of subsequent use in docking highlights that Fpocket and CAVIAR are the best performing cavity prediction tools in this context. Further analysis shows that accurate binding site input does not guarantee accurate binding pose predictions and, even with the predicted cavities, the more restrained the input is, the more reliable the docking results are.
