Journal of Chemical Theory and Computation最新文献

The interplay of quantum and classical simulation and the delicate divide between them is in the focus of massively parallelized tensor network state (TNS) algorithms designed for high performance computing (HPC). In this contribution, we present novel algorithmic solutions together with implementation details to extend current limits of TNS algorithms on HPC infrastructure building on state-of-the-art hardware and software technologies. Benchmark results obtained via large-scale density matrix renormalization group (DMRG) simulations on single node multiGPU NVIDIA A100 system are presented for selected strongly correlated molecular systems addressing problems on Hilbert space dimensions up to 4.17 × 10³⁵.

Existing machine learning models attempt to predict the energies of large molecules by training small molecules, but eventually fail to retain high accuracy as the errors increase with system size. Through an orbital pairwise decomposition of the correlation energy, a pretrained neural network model on hundred-scale data containing small molecules is demonstrated to be sufficiently transferable for accurately predicting large systems, including molecules and crystals. Our model introduces a residual connection to explicitly learn the pairwise energy corrections, and employs various low-rank retraining techniques to modestly adjust the learned network parameters. We demonstrate that with as few as only one larger molecule retraining the base model originally trained on only small molecules of (H₂O)₆, the MP2 correlation energy of the large liquid water (H₂O)₆₄ in a periodic supercell can be predicted at chemical accuracy. Similar performance is observed for large protonated clusters and periodic poly glycine chains. A demonstrative application is presented to predict the energy ordering of symmetrically inequivalent sublattices for distinct hydrogen orientations in the ice XV phase. Our work represents an important step forward in the quest for cost-effective, highly accurate and transferable neural network models in quantum chemistry, bridging the electronic structure patterns between small and large systems.

Organic diradical dications, due to reduced intermolecular interactions, exhibit a greater tendency to adopt high spin states in the solid phase compared to their neutral diradical counterparts. This characteristic makes them promising candidates for applications involving organic electronics. We present a theoretical study of a recently synthesized sulfur-based diradical dication, a unique system exhibiting a robust triplet ground state. Using a number of density functional theory (DFT)-based methods (e.g., standard broken-symmetry DFT, constrained DFT, spin-flip TDDFT) and wave function-based multireference CASSCF+NEVPT2 methods, we investigate its magnetic properties and explore the influence of chalcogen substitution on magnetic exchange coupling. An active space scanning method was adopted to overcome the difficulties in choosing the correct active space for multireference calculation. Our findings highlight the critical role of multireference methods in accurately capturing the magnetic behavior of highly π-conjugated systems. The study reveals a surprising variation in magnetic properties among sulfur, selenium, and tellurium-based diradical dications despite being elements of the same group. These results offer valuable insights into the design and tuning of magnetic properties in organic diradical dications.

We investigate the convergence of quasiparticle energies for periodic systems to the thermodynamic limit using increasingly large simulation cells corresponding to increasingly dense integration meshes in reciprocal space. The quasiparticle energies are computed at the level of equation-of-motion coupled-cluster theory for ionization (IP-EOM-CC) and electron attachment processes (EA-EOM-CC). By introducing an electronic correlation structure factor, the expected asymptotic convergence rates for systems with different dimensionality are formally derived. We rigorously test these derivations through numerical simulations for trans-polyacetylene using IP/EA-EOM-CCSD and the G₀W₀@HF approximation, which confirm the predicted convergence behavior. Our findings provide a solid foundation for efficient schemes to correct finite-size errors in IP/EA-EOM-CCSD calculations.

We explore the melting of a lattice DNA in the presence of atmospheric disorder, which mimics the crowded environment inside the cell nucleus, using Monte Carlo simulations. The disorder is modeled by randomly retaining lattice sites with probability p while diluting the rest, rendering them unavailable to the DNA. By varying the disorder over a wide range from p = 1 (zero disorder) up to the percolation critical point p_c = 0.3116, we show the melting temperature (T_m) to increase nearly linearly with disorder up to p ≈ 0.6, while strong nonlinearity enters for p ≲ 0.6. Associated changes in the bubble statistics have been investigated, showing a substantial change in the bubble size exponent at corresponding melting points for p ≤ 0.5. Based on these findings, two distinct disorder regimes showing weak and strong effects on melting have been identified. For simulations, we use the pruned and enriched Rosenbluth method in conjunction with a depth-first implementation of the Leath algorithm to generate the underlying disorder.

We extend the traditional conceptual density functional theory (CDFT) to conceptual density matrix functional theory (CDMFT) by replacing the external potential v(r) by the one-electron integral h_rs in the energy functional. This approach provides a new path for investigating intrinsic bond properties such as bond reactivity. The derivation of the Fukui matrix, i.e., derivative of the density matrix P with respect to the number of electrons N, is elucidated, and the result is illustrated in a case study on H₂O. The matrix is shown to play a crucial role in quantifying changes of bond strength for electron removal or addition processes via the bond order derivative ${\left(\frac{\partial B}{\partial N}\right)}^{-}$. Using the Mayer bond order and different atoms-in-molecules partitioning methods, we show that as a first-order response quantity, the bond order derivative agrees well with the finite difference bond order changes. The bond order derivative (bond Fukui function) is a bond reactivity descriptor. We demonstrate this by predicting the regioselectivity of a classical electrophilic addition reaction (the bromination of alkenes) and predicting the initial electron-driven bond cleavage in mass spectrometry. Specifically, the bond order derivative captures all of the major signals from the experimental mass spectra for a series of small molecules with a variety of functional groups.

Validating the performance of exchange-correlation functionals is vital to ensure the reliability of density functional theory (DFT) calculations. Typically, these validations involve benchmarking data sets. Currently, such data sets are usually assembled in an unprincipled manner, suffering from uncontrolled chemical bias, and limiting the transferability of benchmarking results to a broader chemical space. In this work, a data-efficient solution based on active learning is explored to address this issue. Focusing─as a proof of principle─on pericyclic reactions, we start from the BH9 benchmarking data set and design a chemical reaction space around this initial data set by combinatorially combining reaction templates and substituents. Next, a surrogate model is trained to predict the standard deviation of the activation energies computed across a selection of 20 distinct DFT functionals. With this model, the designed chemical reaction space is explored, enabling the identification of challenging regions, i.e., regions with large DFT functional divergence, for which representative reactions are subsequently acquired as additional training points. Remarkably, it turns out that the function mapping the molecular structure to functional divergence is readily learnable; convergence is reached upon the acquisition of fewer than 100 reactions. With our final updated model, a more challenging─and arguably more representative─pericyclic benchmarking data set is curated, and we demonstrate that the functional performance has changed significantly compared to the original BH9 subset.

Accurately calculating the diradical character (y₀) of molecular systems remains a significant challenge due to the scarcity of experimental data and the inherent multireference nature of the electronic structure. In this study, various quantum mechanical approaches, including broken symmetry density functional theory (BS-DFT), spin-flip time-dependent density functional theory (SF-TDDFT), mixed-reference spin-flip time-dependent density functional theory (MRSF-TDDFT), complete active space self-consistent field (CASSCF), complete active space second-order perturbation theory (CASPT2), and multiconfigurational pair-density functional theory (MCPDFT), are employed to compute the singlet–triplet energy gaps (E_ST) and y₀ values in Thiele, Chichibabin, and Müller analogous diradicals. By systematically comparing the results from these computational methods, we identify optimally tuned long-range corrected functional CAM-B3LYP in the BS-DFT framework as a most efficient method for accurately and affordably predicting both E_ST and y₀ values. Additionally, our results demonstrate that (i) MRSF-TDDFT performs much better than SF-TDDFT; (ii) the MCPDFT method is robust in determining E_ST with minimal dependence on the choice of active space. These findings provide insight into the electronic structure and diradical character of the investigated molecules and highlight effective computational strategies for future studies in this domain. Thus, this work not only advances our understanding of diradical systems but also offers practical guidelines for their computational investigation.

In this review we discuss the development of methodology for calculating the temperature dependence and thermodynamic activation parameters for chemical reactions in solution and in enzymes, from computer simulations. We outline how this is done by combining the empirical valence bond method with molecular dynamics free energy simulations. In favorable cases it turns out that such simulations can even capture temperature optima for the catalytic rate. The approach turns out be very useful both for addressing questions regarding the roles of enthalpic and entropic effects in specific enzymes and also for attacking evolutionary problems regarding enzyme adaptation to different temperature regimes. In the latter case, we focus on cold-adaptation of enzymes from psychrophilic species and show how computer simulations have revealed the basic mechanisms behind such adaptation. Understanding these mechanisms also opens up the possibility of designing the temperature dependence, and we highlight a recent example of this.

Despite its importance in understanding biology and computer-aided drug discovery, the accurate prediction of protein ionization states remains a formidable challenge. Physics-based approaches struggle to capture the small, competing contributions in the complex protein environment, while machine learning (ML) is hampered by the scarcity of experimental data. Here, we report the development of pK_a ML (KaML) models based on decision trees and graph attention networks (GAT), exploiting physicochemical understanding and a new experiment pK_a database (PKAD-3) enriched with highly shifted pK_a’s. KaML-CBtree significantly outperforms the current state of the art in predicting pK_a values and ionization states across all six titratable amino acids, notably achieving accurate predictions for deprotonated cysteines and lysines─a blind spot in previous models. The superior performance of KaMLs is achieved in part through several innovations, including the separate treatment of acid and base, data augmentation using AlphaFold structures, and model pretraining on a theoretical pK_a database. We also introduce the classification of protonation states as a metric for evaluating pK_a prediction models. A meta-feature analysis suggests a possible reason for the lightweight tree model to outperform the more complex deep learning GAT. We release an end-to-end pK_a predictor based on KaML-CBtree and the new PKAD-3 database, which facilitates a variety of applications and provides the foundation for further advances in protein electrostatic research.