Geometry optimization made simple with translation and rotation coordinates

Investigation of magnetic ionic liquids for selective and rapid extraction of gallic acid from complex samples using experimental, statistical modeling and density functional theory studies

Given the high antioxidant capacity of gallic acid (GA), there is a great deal of interest in the development of rapid, selective, simple, and easily accessible analytical methods for its determination from complex samples. Consequently, the present study aimed to develop an ultrasonic assisted magnetic ionic liquid-based dispersive liquid microextraction (UA-MIL-DLLME) method for the extraction of GA from various samples prior to its spectrophotometric detection. The method's key variables were optimized through statistical analysis. Four magnetic liquids (MILs) were prepared and tested to extract the GA-Se complex formed in aqueous solution. Both experimental studies and theoretical calculations demonstrated that the most suitable MIL for the phase separation of the relevant complex is [P6,6,6,14][Mn(hfacac)3]. The developed UA-MIL-DLLME method exhibited a wide linear range (5-400 ng mL-1), a remarkable enhancement factor (133), and a low limit of detection (1.6 ng mL-1). Additionally, high extraction recovery (97 ± 1%) with a low relative standard deviation (1.9%) was achieved. The extraction time for the UA-MIL-DLLME method was 8 min. The precision of the method was evaluated through repeatability and reproducibility studies. Finally, the UA-MIL-DLLME method was successfully applied to the extraction of the GA from complex samples using a reference method.

Constructing Accurate and Efficient General-Purpose Atomistic Machine Learning Model with Transferable Accuracy for Quantum Chemistry

Density functional theory (DFT) has been a cornerstone in computational science, providing powerful insights into structure-property relationships for molecules and materials through first-principles quantum-mechanical (QM) calculations. However, the advent of atomistic machine learning (ML) is reshaping the landscape by enabling large-scale dynamics simulations and high-throughput screening at DFT-equivalent accuracy with drastically reduced computational cost. Yet, the development of general-purpose atomistic ML models as surrogates for QM calculations faces several challenges, particularly in terms of model capacity, data efficiency, and transferability across chemically diverse systems. This work introduces a novel extension of the polarizable atom interaction neural network (namely, XPaiNN) to address these challenges. Two distinct training strategies have been employed, one direct-learning and the other Δ-ML on top of a semiempirical QM method. These methodologies have been implemented within the same framework, allowing for a detailed comparison of their results. The XPaiNN models, in particular the one using Δ-ML, not only demonstrate competitive performance on standard benchmarks, but also demonstrate the effectiveness against other ML models and QM methods on comprehensive downstream tasks, including noncovalent interactions, reaction energetics, barrier heights, geometry optimization and reaction thermodynamics, etc. This work represents a significant step forward in the pursuit of accurate and efficient atomistic ML models of general-purpose, capable of handling complex chemical systems with transferable accuracy.

Analytical Evaluation of Ground State Gradients in Quantum Electrodynamics Coupled Cluster Theory

Analytical gradients of potential energy surfaces play a central role in quantum chemistry, allowing for molecular geometry optimizations and molecular dynamics simulations. In strong coupling conditions, potential energy surfaces can account for strong interactions between matter and the quantized electromagnetic field. In this paper, we derive expressions for the ground state analytical gradients in quantum electrodynamics coupled cluster theory. We also present a Cholesky-based implementation for the coupled cluster singles and doubles model. We report timings to show the performance of the implementation and present optimized geometries to highlight cavity-induced molecular orientation effects in strong coupling conditions.

Semiclassical Nonadiabatic Molecular Dynamics Using Linearized Pair-Density Functional Theory

Nonadiabatic molecular dynamics is an effective method for modeling nonradiative decay in electronically excited molecules. Its accuracy depends strongly on the quality of the potential energy surfaces, and its affordability for long direct-dynamic simulations with adequate ensemble averaging depends strongly on the cost of the required electronic structure calculations. Linearized pair-density functional theory (L-PDFT) is a recently developed post-self-consistent-field multireference method that can model potential energy surfaces with an accuracy similar to expensive multireference perturbation theories but at a computational cost similar to the underlying multiconfiguration self-consistent field method. Here, we integrate the SHARC dynamics and PySCF electronic structure code to utilize L-PDFT for electronically nonadiabatic calculations and use the combined programs to study the photoisomerization reaction of cis-azomethane. We show that L-PDFT is able to successfully simulate the photoisomerization without crashes, and it yields results similar to the more expensive extended multistate complete active space second-order perturbation theory. This shows that L-PDFT can model internal conversion, and it demonstrates its promise for broader photodynamics applications.

Assessing the Partial Hessian Approximation in QM/MM-Based Vibrational Analysis

The partial Hessian approximation is often used in vibrational analysis of quantum mechanics/molecular mechanics (QM/MM) systems because calculating the full Hessian matrix is computationally impractical. This approach aligns with the core concept of QM/MM, which focuses on the QM subsystem. Thus, using the partial Hessian approximation implies that the main interest is in the local vibrational modes of the QM subsystem. Here, we investigate the accuracy and applicability of the partial Hessian vibrational analysis (PHVA) approach as it is typically used within QM/MM, i.e., only the Hessian belonging to the QM subsystem is computed. We focus on solute-solvent systems with small, rigid solutes. To separate two of the major sources of errors, we perform two separate analyses. First, we study the effects of the partial Hessian approximation on local normal modes, harmonic frequencies, and harmonic IR and Raman intensities by comparing them to those obtained using full Hessians, where both partial and full Hessians are calculated at the QM level. Then, we quantify the errors introduced by QM/MM used with the PHVA by comparing normal modes, frequencies, and intensities obtained using partial Hessians calculated using a QM/MM-type embedding approach to those obtained using partial Hessians calculated at the QM level. Another aspect of the PHVA is the appearance of normal modes resembling the translation and rotation of the QM subsystem. These pseudotranslational and pseudorotational modes should be removed as they are collective vibrations of the atoms in the QM subsystem relative to a frozen MM subsystem and, thus, not well-described. We show that projecting out translation and rotation, usually done for systems in isolation, can adversely affect other normal modes. Instead, the pseudotranslational and pseudorotational modes can be identified and removed.

Impact of Protonation Sites on Collision-Induced Dissociation-MS/MS Using CIDMD Quantum Chemistry Modeling

Protonation is the most frequent adduct found in positive electrospray ionization collision-induced mass spectra (CID-MS/MS). In a parallel report Lee, J. J. Chem. Inf. Model. 2024, 10.1021/acs.jcim.4c00760, we developed a quantum chemistry framework to predict mass spectra by collision-induced dissociation molecular dynamics (CIDMD). As different protonation sites affect fragmentation pathways of a given molecule, the accuracy of predicting tandem mass spectra by CIDMD ultimately depends on the choice of its protomers. To investigate the impact of molecular protonation sites on MS/MS spectra, we compared CIDMD-predicted spectra to all available experimental MS/MS spectra by similarity matching. We probed 10 molecules with a total of 43 protomers, the largest study to date, including organic acids (sorbic acid, citramalic acid, itaconic acid, mesaconic acid, citraconic acid, and taurine) as well as aromatic amines including uracil, aniline, bufotenine, and psilocin. We demonstrated how different protomers can converge different fragmentation pathways to the same fragment ions but also may explain the presence of different fragment ions in experimental MS/MS spectra. For the first time, we used in silico MS/MS predictions to test the impact of solvents on proton affinities, comparing the gas phase and a mixture of acetonitrile/water (1:1). We also extended applications of in silico MS/MS predictions to investigate the impact of protonation sites on the energy barriers of isomerization between protomers via proton transfer. Despite our initial hypothesis that the thermodynamically most stable protomer should give the best match to the experiment, we found only weak inverse relationships between the calculated proton affinities and corresponding entropy similarities of experimental and CIDMD-predicted MS/MS spectra. CIDMD-predicted mechanistic details of fragmentation reaction pathways revealed a clear preference for specific protomer forms for several molecules. Overall, however, proton affinity was not a good predictor corresponding to the predicted CIDMD spectra. For example, for uracil, only one protomer predicted all experimental MS/MS fragment ions, but this protomer had neither the highest proton affinity nor the best MS/MS match score. Instead of proton affinity, the transfer of protons during the electrospray process from the initial protonation site (i.e., mobile proton model) better explains the differences between the thermodynamic rationale and experimental data. Protomers that undergo fragmentation with lower energy barriers have greater contributions to experimental MS/MS spectra than their thermodynamic Boltzmann populations would suggest. Hence, in silico predictions still need to calculate MS/MS spectra for multiple protomers, as the extent of distributions cannot be readily predicted.

Predicting Collision-Induced-Dissociation Tandem Mass Spectra (CID-MS/MS) Using Ab Initio Molecular Dynamics

Compound identification is at the center of metabolomics, usually by comparing experimental mass spectra against library spectra. However, most compounds are not commercially available to generate library spectra. Hence, for such compounds, MS/MS spectra need to be predicted. Machine learning and heuristic models have largely failed except for lipids. Here, quantum chemistry software can be used to predict mass spectra. However, quantum chemistry predictions for collision induced dissociation (CID) mass spectra in LC-MS/MS are rare. We present the CIDMD (Collision-Induced Dissociation via Molecular Dynamics) framework to model CID-based MS/MS spectra. It uses first-principles molecular dynamics (MD) to simulate the physical process of molecular collisions in CID tandem mass spectrometry. First, molecular ions are constructed at specific protonation sites. Using density functional theory, these protonated ions are targeted by argon collider gas atoms at user-specified velocities. Subsequent bond breakages are simulated over time for at least 1,000 fs. Each simulation is repeated multiple times from various collisional directions. Fragmentations are accumulated over those repeated collisions to generate CIDMD in silico mass spectra. Twelve small metabolites (<205 Da) were selected to test the accuracy of this framework in comparison to experimental MS/MS spectra. When testing different protomers, collider velocities, number of simulations, simulation time and impact factor b cutoffs, we yielded 261 predicted mass spectra. These in silico spectra resulted in entropy similarity scores of an average 624 ± 189 for all 261 spectra compared to their corresponding experimental spectra, which improved to 828 ± 77 when using optimal parameters of the most probable protomers for 12 molecules. With increasing molecular mass, higher velocities achieved better results. Similarly, different protomers showed large differences in fragmentation; hence, with increasing numbers of protomers and tautomers, the average CIDMD prediction accuracy decreased. Mechanistic details showed that specific fragment ions can be produced from different protomers via multiple fragmentation pathways. We propose that CIDMD is a suitable tool to predict mass spectra of small metabolites like produced by the gut microbiome.

Two P, Ten P, White P, Red P: Mechanistic Exploration of the Oligomerization of Red Phosphorus from Diphosphorus with the Ab Initio Nanoreactor

Phosphorus is critical to humans on many fronts, yet we do not have a mechanistic understanding of some of its most basic transformations and reactions─namely the oligomerization of white phosphorus to red. With heat or under ultraviolet (UV) exposure, it has been experimentally demonstrated that white phosphorus dissociates into diphosphorus units which readily form red phosphorus. However, the mechanism of this process is unknown. The ab initio nanoreactor approach was used to explore the potential energy surface of phosphorus clusters. Density functional theory and metadynamics simulations were used to characterize potential reaction pathways. A mechanism for oligomerization is proposed to take place via diphosphorus additions at π-bonds and weak σ-bonds through three membered ring intermediates. Downhill paths through P6 and P8 clusters eventually result in P10 clusters that can oligomerize into red phosphorus chains. The initial, rate limiting step for this process has an energy barrier of 24.2 kcal/mol.

The orientation dependence of cavity-modified chemistry

Recent theoretical studies have explored how ultra-strong light-matter coupling can be used as a handle to control chemical transformations. Ab initio cavity quantum electrodynamics calculations demonstrate that large changes to reaction energies or barrier heights can be realized by coupling electronic degrees of freedom to vacuum fluctuations associated with an optical cavity mode, provided that large enough coupling strengths can be achieved. In many cases, the cavity effects display a pronounced orientational dependence. Here, we highlight the critical role that geometry relaxation can play in such studies. As an example, we consider a recent work [Pavošević et al., Nat. Commun. 14, 2766 (2023)] that explored the influence of an optical cavity on Diels-Alder cycloaddition reactions and reported large changes to reaction enthalpies and barrier heights, as well as the observation that changes in orientation can inhibit the reaction or select for one reaction product or another. Those calculations used fixed molecular geometries optimized in the absence of the cavity and fixed relative orientations of the molecules and the cavity mode polarization axis. Here, we show that when given a chance to relax in the presence of the cavity, the molecular species reorient in a way that eliminates the orientational dependence. Moreover, in this case, we find that qualitatively different conclusions regarding the impact of the cavity on the thermodynamics of the reaction can be drawn from calculations that consider relaxed vs unrelaxed molecular structures.

Computational Screening of Putative Catalyst Transition Metal Complexes as Guests in a Ga4L612- Nanocage

Metal-organic cages form well-defined microenvironments that can enhance the catalytic proficiency of encapsulated transition metal complexes (TMCs). We introduce a screening protocol to efficiently identify TMCs that are promising candidates for encapsulation in the Ga4L612- nanocage. We obtain TMCs from the Cambridge Structural Database with geometric and electronic characteristics amenable to encapsulation and mine the text of associated manuscripts to curate TMCs with documented catalytic functionality. By docking candidate TMCs inside the nanocage cavity and carrying out electronic structure calculations, we identify a subset of successfully optimized candidates (TMC-34) and observe that encapsulated guests occupy an average of 60% of the cavity volume, in line with previous observations. Notably, some guests occupy as much as 72% of the cavity as a result of linker rotation. Encapsulation has a universal effect on the electrostatic potential (ESP), systematically decreasing the ESP at the metal center of each TMC in the TMC-34 data set, while minimally altering TMC metal partial charges. Collectively these observations support geometry-based screening of potential guests and suggest that encapsulation in Ga4L612- cages could electrostatically stabilize diverse cationic or electropositive intermediates. We highlight candidate guests with associated known reactivity and solubility most amenable for encapsulation in experimental follow-up studies.

Enantiomer Differentiation by Interaction-Specific Prediction of Residual Dipolar Couplings in Spherical-like Molecules

Residual Dipolar Couplings (RDCs) are averaged dipolar couplings between nuclear spins of atoms in a molecule that can be measured by nuclear magnetic resonance (NMR) spectroscopy upon partial alignment by a chiral alignment medium. The estimation of differences in alignment of enantiomers may, in principle, enable the determination of absolute configuration. Here, we use molecular dynamics (MD) simulations to mimic the alignment of chiral molecules (i.e., isopinocampheol, quinuclidin-3-ol, borneol, and camphor) to the chiral poly-γ-benzyl-L-glutamate (PBLG) polymer to predict RDCs in silico and compare calculated and experimentally measured residual dipolar couplings for the four enantiomeric pairs. The aim is to validate the computational scheme for the prediction of RDCs in chiral molecules and understand the interaction leading to the alignment in more detail. We determine the indispensable importance of hydrogen bonds between a chiral molecule and the alignment medium on the overall quality of the simulated alignment and interaction poses toward high agreement with experiments. A good correlation with experimental data is found for camphor and isopinocampheol, while the correlation for quinuclidin-3-ol and borneol is lower. We attribute this observation to the high difficulty of the RDC prediction for rather almost spherical molecules. The study reveals that the prediction of alignment with small enantiomeric differences is possible with an MD-based approach; however, extended simulation times (e.g., 50-100 μs) are required to sufficiently reduce the statistical uncertainty. This may be further used for the determination of the relative, as well as absolute, configuration of chiral molecules.

A Polarizable QM/MM Model That Combines the State-Averaged CASSCF and AMOEBA Force Field for Photoreactions in Proteins

This study presents the polarizable quantum mechanics/molecular mechanics (QM/MM) embedding of the state-averaged complete active space self-consistent field (SA-CASSCF) in the atomic multipole optimized energetics for biomolecular applications (AMOEBA) force field for the purpose of studying photoreactions in protein environments. We describe two extensions of our previous work that combine SA-CASSCF with AMOEBA water models, allowing it to be generalized to AMOEBA models for proteins and other macromolecules. First, we discuss how our QM/MM model accounts for the discrepancy between the direct and polarization electric fields that arises in the AMOEBA description of intramolecular polarization. A second improvement is the incorporation of link atom schemes to treat instances in which the QM/MM boundary goes through covalent bonds. A single-link atom scheme and double-link atom scheme are considered in this work, and we will discuss how electrostatic interaction, van der Waals interaction, and various kinds of valence terms are treated across the boundary. To test the accuracy of the link atom scheme, we will compare QM/MM with full QM calculations and study how the errors in ground state properties, excited state properties, and excitation energies change when tuning the parameters in the link atom scheme. We will also test the new SA-CASSCF/AMOEBA method on an elementary reaction step in NanoLuc, an artificial bioluminescence luciferase. We will show how the reaction mechanism is different when calculated in the gas phase, in polarizable continuum medium (PCM), versus in protein AMOEBA models.

The Open Force Field Initiative: Open Software and Open Science for Molecular Modeling

Force fields are a key component of physics-based molecular modeling, describing the energies and forces in a molecular system as a function of the positions of the atoms and molecules involved. Here, we provide a review and scientific status report on the work of the Open Force Field (OpenFF) Initiative, which focuses on the science, infrastructure and data required to build the next generation of biomolecular force fields. We introduce the OpenFF Initiative and the related OpenFF Consortium, describe its approach to force field development and software, and discuss accomplishments to date as well as future plans. OpenFF releases both software and data under open and permissive licensing agreements to enable rapid application, validation, extension, and modification of its force fields and software tools. We discuss lessons learned to date in this new approach to force field development. We also highlight ways that other force field researchers can get involved, as well as some recent successes of outside researchers taking advantage of OpenFF tools and data.

Functionality optimization for effective singlet fission coupling screening in the full-dimensional molecular and intermolecular coordinate space

In computational chemistry, accurately predicting molecular configurations that exhibit specific properties remains a critical challenge. Its intricacies become especially evident in the study of molecular aggregates, where the light-induced functionality is tied to highly structure-dependent electronic couplings between molecules. Here, we present an efficient strategy for the targeted screening of the structural space employing a "functionality optimization" technique, in which a chosen descriptor, constrained by the ground state energy expression, is optimized. The chosen algorithmic differentiation (AD) framework allows one to automatically obtain gradients without its tedious implementation. We demonstrate the effectiveness of the approach by identifying perylene bisimide (PBI) dimer motifs with enhanced effective SF coupling. Our findings reveal that certain structural modifications of the PBI monomer, such as helical twisting and bending as well as slipped-rotated packing arrangements, can significantly increase the effective SF coupling.

Performant automatic differentiation of local coupled cluster theories: Response properties and ab initio molecular dynamics

In this work, we introduce a differentiable implementation of the local natural orbital coupled cluster (LNO-CC) method within the automatic differentiation framework of the PySCFAD package. The implementation is comprehensively tuned for enhanced performance, which enables the calculation of first-order static response properties on medium-sized molecular systems using coupled cluster theory with single, double, and perturbative triple excitations [CCSD(T)]. We evaluate the accuracy of our method by benchmarking it against the canonical CCSD(T) reference for nuclear gradients, dipole moments, and geometry optimizations. In addition, we demonstrate the possibility of property calculations for chemically interesting systems through the computation of bond orders and Mössbauer spectroscopy parameters for a [NiFe]-hydrogenase active site model, along with the simulation of infrared spectra via ab initio LNO-CC molecular dynamics for a protonated water hexamer.

XMECP: Reaching State-of-the-Art MECP Optimization in Multiscale Complex Systems

The Python-based program, XMECP, is developed for realizing robust, efficient, and state-of-the-art minimum energy crossing point (MECP) optimization in multiscale complex systems. This article introduces the basic capabilities of the XMECP program by theoretically investigating the MECP mechanism of several example systems including (1) the photosensitization mechanism of benzophenone, (2) photoinduced proton-coupled electron transfer in the cytosine-guanine base pair in DNA, (3) the spin-flip process in oxygen activation catalyzed by an iron-containing 2-oxoglutarate-dependent oxygenase (Fe/2OGX), and (4) the photochemical pathway of flavoprotein adjusted by the intensity of an external electric field. MECPs related to multistate reaction and multistate reactivity in large-scale complex biochemical systems can be well-treated by workflows suggested by the XMECP program. The branching plane updating the MECP optimization algorithm is strongly recommended as it provides derivative coupling vector (DCV) with explicit calculation and can equivalently evaluate contributions from non-QM residues to DCV, which can be nonadiabatic coupling or spin-orbit coupling in different cases. In the discussed QM/MM examples, we also found that the influence on the QM region by DCV can occur through noncovalent interactions and decay with distance. In the example of DNA base pairs, the nonadiabatic coupling occurs across the π-π stacking structure formed in the double-helix system. In contrast to general intuition, in the example of Fe/2OGX, the central ferrous and oxygen part contribute little to the spin-orbit coupling; however, a nearby arginine residue, which is treated by molecular mechanics in the QM/MM method, contributes significantly via two hydrogen bonds formed with α-ketoglutarate (α-KG). This indicates that the arginine residue plays a significant role in oxygen activation, driving the initial triplet state toward the productive quintet state, which is more than the previous knowledge that the arginine residue can bind α-KG at the reaction site by hydrogen bonds.

Which Options Exist for NISQ-Friendly Linear Response Formulations?

Linear response (LR) theory is a powerful tool in classic quantum chemistry crucial to understanding photoinduced processes in chemistry and biology. However, performing simulations for large systems and in the case of strong electron correlation remains challenging. Quantum computers are poised to facilitate the simulation of such systems, and recently, a quantum linear response formulation (qLR) was introduced [Kumar et al., J. Chem. Theory Comput. 2023, 19, 9136-9150]. To apply qLR to near-term quantum computers beyond a minimal basis set, we here introduce a resource-efficient qLR theory, using a truncated active-space version of the multiconfigurational self-consistent field LR ansatz. Therein, we investigate eight different near-term qLR formalisms that utilize novel operator transformations that allow the qLR equations to be performed on near-term hardware. Simulating excited state potential energy curves and absorption spectra for various test cases, we identify two promising candidates, dubbed "proj LRSD" and "all-proj LRSD".

Analytic Nuclear Gradients for Complete Active Space Linearized Pair-Density Functional Theory

Accurately modeling photochemical reactions is difficult due to the presence of conical intersections and locally avoided crossings, as well as the inherently multiconfigurational character of excited states. As such, one needs a multistate method that incorporates state interaction in order to accurately model the potential energy surface at all nuclear coordinates. The recently developed linearized pair-density functional theory (L-PDFT) is a multistate extension of multiconfiguration PDFT, and it has been shown to be a cost-effective post-MCSCF method (as compared to more traditional and expensive multireference many-body perturbation methods or multireference configuration interaction methods) that can accurately model potential energy surfaces in regions of strong nuclear-electronic coupling in addition to accurately predicting Franck-Condon vertical excitations. In this paper, we report the derivation of analytic gradients for L-PDFT and their implementation in the PySCF-forge software, and we illustrate the utility of these gradients for predicting ground- and excited-state equilibrium geometries and adiabatic excitation energies for formaldehyde, s-trans-butadiene, phenol, and cytosine.

Rotational spectroscopy of methyl tert-butyl ether with a new Ka band chirped-pulse Fourier transform microwave spectrometer

Chirped-pulse Fourier transform microwave (CP-FTMW) spectroscopy is a powerful tool for performing broadband gas-phase rotational spectroscopy, and its applications include discovery of new molecules, complex mixture analysis, and exploration of fundamental molecular physics. Here we report the development of a new Ka band (26.5-40 GHz) CP-FTMW spectrometer that is equipped with a pulsed supersonic expansion source and a heated reservoir for low-volatility samples. The spectrometer is built around a 150 W traveling wave tube amplifier and has an instantaneous bandwidth that covers the entire Ka band spectral range. To test the performance of the spectrometer, the rotational spectrum of methyl tert-butyl ether (MTBE), a former gasoline additive and environmental pollutant, has been measured for the first time in this spectral range. Over 1000 spectroscopic transitions have been measured and assigned to the vibrational ground state and a newly-identified torsionally excited state; all transitions were fit using the XIAM program to a root-mean-square deviation of 22 kHz. The spectrum displays internal rotation splitting, nominally forbidden transitions, and an intriguing axis-switching effect between the ground and torsionally excited state that is a consequence of MTBE's extreme near-prolate nature. Finally, the sensitivity of the spectrometer enabled detection of all singly-substituted 13C and 18O isotopologues in natural abundance. This set of isotopic spectra allowed for a partial r0 structure involving the heavy atoms to be derived, resolving a structural discrepancy in the literature between previous microwave and electron diffraction measurements.

Massively scalable workflows for quantum chemistry: BigChem and ChemCloud

Electronic structure theory, i.e., quantum chemistry, is the fundamental building block for many problems in computational chemistry. We present a new distributed computing framework (BigChem), which allows for an efficient solution of many quantum chemistry problems in parallel. BigChem is designed to be easily composable and leverages industry-standard middleware (e.g., Celery, RabbitMQ, and Redis) for distributed approaches to large scale problems. BigChem can harness any collection of worker nodes, including ones on cloud providers (such as AWS or Azure), local clusters, or supercomputer centers (and any mixture of these). BigChem builds upon MolSSI packages, such as QCEngine to standardize the operation of numerous computational chemistry programs, demonstrated here with Psi4, xtb, geomeTRIC, and TeraChem. BigChem delivers full utilization of compute resources at scale, offers a programable canvas for designing sophisticated quantum chemistry workflows, and is fault tolerant to node failures and network disruptions. We demonstrate linear scalability of BigChem running computational chemistry workloads on up to 125 GPUs. Finally, we present ChemCloud, a web API to BigChem and successor to TeraChem Cloud. ChemCloud delivers scalable and secure access to BigChem over the Internet.

Reweighting from Molecular Mechanics Force Fields to the ANI-2x Neural Network Potential

To achieve chemical accuracy in free energy calculations, it is necessary to accurately describe the system's potential energy surface and efficiently sample configurations from its Boltzmann distribution. While neural network potentials (NNPs) have shown significantly higher accuracy than classical molecular mechanics (MM) force fields, they have a limited range of applicability and are considerably slower than MM potentials, often by orders of magnitude. To address this challenge, Rufa et al. [Rufa et al. bioRxiv 2020, 10.1101/2020.07.29.227959.] suggested a two-stage approach that uses a fast and established MM alchemical energy protocol, followed by reweighting the results using NNPs, known as endstate correction or indirect free energy calculation. This study systematically investigates the accuracy and robustness of reweighting from an MM reference to a neural network target potential (ANI-2x) for an established data set in vacuum, using single-step free-energy perturbation (FEP) and nonequilibrium (NEQ) switching simulation. We assess the influence of longer switching lengths and the impact of slow degrees of freedom on outliers in the work distribution and compare the results to those of multistate equilibrium free energy simulations. Our results demonstrate that free energy calculations between NNPs and MM potentials should be preferably performed using NEQ switching simulations to obtain accurate free energy estimates. NEQ switching simulations between the MM potentials and NNPs are efficient, robust, and trivial to implement.

New physical insights into the supporting subspace factorization of XMS-CASPT2 and generalization to multiple spin states via spin-free formulation

This paper introduces a spin-free formulation of the supporting subspace factorization [C. Song and T. J. Martínez, J. Chem. Phys. 149, 044108 (2018)], enabling a reduction in the computational scaling of the extended multi-state complete active space second-order perturbation (XMS-CASPT2) method for arbitrary spins. Compared to the original formulation that is defined in the spin orbitals and is limited to singlet states, the spin-free formulation in this work treats different spin states equivalently, thus naturally generalizing the idea beyond singlet states. In addition, we will present a new way of deriving the supporting subspace factorization with the purpose of understanding its physical interpretation. In this new derivation, we separate the sources that make CASPT2 difficult into the "same-site interactions" and "inter-site interactions." We will first show how the Kronecker sum can be used to remove the same-site interactions in the absence of inter-site interactions, leading to MP2 energy in dressed orbitals. We will then show how the inter-site interactions can be exactly recovered using Löwdin partition, where the supporting subspace concept will naturally arise. The new spin-free formulation maintains the main advantage of the supporting subspace factorization, i.e., allowing XMS-CASPT2 energies to be computed using highly optimized MP2 energy codes and Fock build codes, thus reducing the scaling of XMS-CASPT2 to the same scaling as MP2. We will present and discuss results that benchmark the accuracy and performance of the new method. To demonstrate how the new method can be useful in studying real photochemical systems, the supporting subspace XMS-CASPT2 is applied to a photoreaction sensitive to magnetic field effects. The new spin-free formulation makes it possible to calculate the doublet and quartet states required in this particular photoreaction mechanism.

Systematic QM/MM Study for Predicting 31P NMR Chemical Shifts of Adenosine Nucleotides in Solution and Stages of ATP Hydrolysis in a Protein Environment

NMR (nuclear magnetic resonance) spectroscopy allows for important atomistic insights into the structure and dynamics of biological macromolecules; however, reliable assignments of experimental spectra are often difficult. Herein, quantum mechanical/molecular mechanical (QM/MM) calculations can provide crucial support. A major problem for the simulations is that experimental NMR signals are time-averaged over much longer time scales, and since computed chemical shifts are highly sensitive to local changes in the electronic and structural environment, sufficiently large averages over representative structural ensembles are essential. This entails high computational demands for reliable simulations. For NMR measurements in biological systems, a nucleus of major interest is 31P since it is both highly present (e.g., in nucleic acids) and easily observable. The focus of our present study is to develop a robust and computationally cost-efficient framework for simulating 31P NMR chemical shifts of nucleotides. We apply this scheme to study the different stages of the ATP hydrolysis reaction catalyzed by p97. Our methodology is based on MM molecular dynamics (MM-MD) sampling, followed by QM/MM structure optimizations and NMR calculations. Overall, our study is one of the most comprehensive QM-based 31P studies in a protein environment and the first to provide computed NMR chemical shifts for multiple nucleotide states in a protein environment. This study sheds light on a process that is challenging to probe experimentally and aims to bridge the gap between measured and calculated NMR spectroscopic properties.

Impact of isoelectronic substitution on the excited state processes in polycyclic aromatic hydrocarbons: a joint experimental and theoretical study of 4a,8a-azaboranaphthalene

Substituting CC with the isoelectronic BN units is a promising approach to modify the optoelectronic properties of polycyclic aromatic hydrocarbons. While computational studies have already addressed trends in the electronic structure of the various isosteres, experimental data are still scarce. Here, the excited state spectroscopy and dynamics of 4a,8a-azaboranaphthalene were studied by picosecond time-resolved photoionization in a supersonic jet and analyzed with the aid of XMS-CASPT2 and time-dependent DFT calculations. A resonance-enhanced multiphoton ionization spectrum (REMPI) reveals the S1 origin at  = 33 830 ± 12 cm-1. Several vibrational bands were resolved and assigned by comparison with the computations. A [1+1] photoelectron spectrum via the S1 origin yielded an adiabatic ionization energy of 8.27 eV. Selected vibrational bands were subsequently investigated by pump-probe photoionization. While the origin as well as several low-lying vibronic states exhibit lifetimes in the ns-range, a monoexponential decay is observed at higher excitation energies, ranging from 400 ps at +1710 cm-1 to 13 ps at +3360 cm-1. The deactivation is attributed to an internal conversion of the optically excited S1 state via a barrier that gives access to a conical intersection (CI) to the S0 state. The doping significantly changes the energetic ordering of CIs and lowers the corresponding energy barrier for the associated deactivation pathway, as revealed by nudged elastic band (NEB) calculations.

Machine Learning Many-Body Green's Functions for Molecular Excitation Spectra

We present a machine learning (ML) framework for predicting Green's functions of molecular systems, from which photoemission spectra and quasiparticle energies at quantum many-body level can be obtained. Kernel ridge regression is adopted to predict self-energy matrix elements on compact imaginary frequency grids from static and dynamical mean-field electronic features, which gives direct access to real-frequency many-body Green's functions through analytic continuation and Dyson's equation. Feature and self-energy matrices are represented in a symmetry-adapted intrinsic atomic orbital plus projected atomic orbital basis to enforce rotational invariance. We demonstrate good transferability and high data efficiency of the proposed ML method across molecular sizes and chemical species by showing accurate predictions of density of states (DOS) and quasiparticle energies at the level of many-body perturbation theory (GW) or full configuration interaction. For the ML model trained on 48 out of 1995 molecules randomly sampled from the QM7 and QM9 data sets, we report the mean absolute errors of ML-predicted highest occupied and lowest unoccupied molecular orbital energies to be 0.13 and 0.10 eV, respectively, compared to GW@PBE0. We further showcase the capability of this method by applying the same ML model to predict DOS for significantly larger organic molecules with up to 44 heavy atoms.

Molecular π-Orbital Construction for Non-Planar Conjugated Systems

We extend the π-orbital space (PiOS) method introduced for planar π-conjugated molecular systems [J. Chem. Theory Comput. 2019, 15, 1679] to also allow constructing efficient π-orbital active spaces for non-planar π-conjugated systems. We demonstrate the performance of this method with multiconfigurational and multireference calculations on prototypical non-planar π-systems: cycloacenes, short carbon nanotubes, various conformations of the 2,2-bipyridine anion, and C20 fullerenes.

Classification of Hemilabile Ligands Using Machine Learning

Hemilabile ligands have the capacity to partially disengage from a metal center, providing a strategy to balance stability and reactivity in catalysis, but they are not straightforward to identify. We identify ligands in the Cambridge Structural Database that have been crystallized with distinct denticities and are thus identifiable as hemilabile ligands. We implement a semi-supervised learning approach using a label-spreading algorithm to augment a small negative set that is supported by heuristic rules of ligand and metal co-occurrence. We show that a heuristic based on coordinating atom identity alone is not sufficient to identify whether a ligand is hemilabile, and our trained machine-learning classification models are instead needed to predict whether a bi-, tri-, or tetradentate ligand is hemilabile with high accuracy and precision. Feature importance analysis of our models shows that the second, third, and fourth coordination spheres all play important roles in ligand hemilability.

Integrating Chemical Information into Reinforcement Learning for Enhanced Molecular Geometry Optimization

Geometry optimization is a crucial step in computational chemistry, and the efficiency of optimization algorithms plays a pivotal role in reducing computational costs. In this study, we introduce a novel reinforcement-learning-based optimizer that surpasses traditional methods in terms of efficiency. What sets our model apart is its ability to incorporate chemical information into the optimization process. By exploring different state representations that integrate gradients, displacements, primitive type labels, and additional chemical information from the SchNet model, our reinforcement learning optimizer achieves exceptional results. It demonstrates an average reduction of about 50% or more in optimization steps compared to the conventional optimization algorithms that we examined when dealing with challenging initial geometries. Moreover, the reinforcement learning optimizer exhibits promising transferability across various levels of theory, emphasizing its versatility and potential for enhancing molecular geometry optimization. This research highlights the significance of leveraging reinforcement learning algorithms to harness chemical knowledge, paving the way for future advancements in computational chemistry.

Accuracy of One- and Two-Photon Intensities with the Extended Polarizable Density Embedding Model

The recently developed extended polarizable density embedding (PDE-X) model is evaluated for the spectroscopic properties of organic chromophores solvated in water, including both one- and two-photon absorption properties. The PDE-X embedding model systematically improves vertical excitation energies over the preceding polarizable density embedding model (PDE). PDE-X shows more modest improvements over existing embedding models for oscillator strengths and two-photon absorption cross-sections, which are more sensitive properties. We argue that the origin of these discrepancies is related to the description of polarization effects, suggesting directions for future development of the embedding model.

Geometry Optimization: A Comparison of Different Open-Source Geometry Optimizers

Based on a series of energy minimizations with starting structures obtained from the Baker test set of 30 organic molecules, a comparison is made between various open-source geometry optimization codes that are interfaced with the open-source QUantum Interaction Computational Kernel (QUICK) program for gradient and energy calculations. The findings demonstrate how the choice of the coordinate system influences the optimization process to reach an equilibrium structure. With fewer steps, internal coordinates outperform Cartesian coordinates, while the choice of the initial Hessian and Hessian update method in quasi-Newton approaches made by different optimization algorithms also contributes to the rate of convergence. Furthermore, an available open-source machine learning method based on Gaussian process regression (GPR) was evaluated for energy minimizations over surrogate potential energy surfaces with both Cartesian and internal coordinates with internal coordinates outperforming Cartesian. Overall, geomeTRIC and DL-FIND with their default optimization method as well as with the GPR-based model using Hartree-Fock theory with the 6-31G** basis set needed a comparable number of geometry optimization steps to the approach of Baker using a unit matrix as the initial Hessian to reach the optimized geometry. On the other hand, the Berny and Sella offerings in ASE outperformed the other algorithms. Based on this, we recommend using the file-based approaches, ASE/Berny and ASE/Sella, for large-scale optimization efforts, while if using a single executable is preferable, we now distribute QUICK integrated with DL-FIND.

On the Simulation of Thermal Isomerization of Molecular Photoswitches in Biological Systems

Molecular photoswitches offer precise, reversible photocontrol over biomolecular functions and are promising light-regulated drug candidates with minimal side effects. Quantifying thermal isomerization rates of photoswitches in their target biomolecules is essential for fine-tuning their light-controlled drug activity. However, the effects of protein binding on isomerization kinetics remain poorly understood, and simulations are crucial for filling this gap. Challenges in the simulation include describing multireference electronic structures near transition states, disentangling competing reaction pathways, and sampling protein-ligand interactions. To overcome these challenges, we used multiscale simulations to characterize the thermal isomerization of photostatins (PSTs), which are light-regulated microtubule inhibitors for potential cancer phototherapy. We employed a new ab initio multireference electronic structure method in a quantum mechanics/molecular mechanics setting and combined it with enhanced sampling techniques to characterize the cis to trans free-energy profiles of three PSTs in a vacuum, aqueous solution, and tubulin dimer. The significant advantage of our novel approach is the efficient treatment of the multireference character in PSTs' electronic wavefunction throughout the conformational sampling of protein-ligand interactions along their isomerization pathways. We also benchmarked our calculations using high-level ab initio multireference electronic structure methods and explored the competing isomerization pathways. Notably, calculations in a vacuum and implicit solvent models cannot predict the order of the PSTs' thermal half-lives in the aqueous solution observed in the experiment. Only by explicitly treating the solvent molecules can the correct order of isomerization kinetics be reproduced. Protein binding perturbs free-energy barriers due to hydrogen bonding between PSTs and nearby polar residues. Our work generates comprehensive, high-quality benchmark data and offers guidance for selecting computational methods to study the thermal isomerization of photoswitches. Ab initio multireference free-energy calculations in explicit molecular environments are crucial for predicting the effects of substituents on the thermal half-lives of photoswitches in biological systems.

A quantum chemical interaction energy dataset for accurately modeling protein-ligand interactions

Fast and accurate calculation of intermolecular interaction energies is desirable for understanding many chemical and biological processes, including the binding of small molecules to proteins. The Splinter ["Symmetry-adapted perturbation theory (SAPT0) protein-ligand interaction"] dataset has been created to facilitate the development and improvement of methods for performing such calculations. Molecular fragments representing commonly found substructures in proteins and small-molecule ligands were paired into >9000 unique dimers, assembled into numerous configurations using an approach designed to adequately cover the breadth of the dimers' potential energy surfaces while enhancing sampling in favorable regions. ~1.5 million configurations of these dimers were randomly generated, and a structurally diverse subset of these were minimized to obtain an additional ~80 thousand local and global minima. For all >1.6 million configurations, SAPT0 calculations were performed with two basis sets to complete the dataset. It is expected that Splinter will be a useful benchmark dataset for training and testing various methods for the calculation of intermolecular interaction energies.

Exploring Ligand-to-Metal Charge-Transfer States in the Photo-Ferrioxalate System Using Excited-State Specific Optimization

The photo-ferrioxalate system (PFS), [Fe(III)(C2O4)]3-, more than an exact chemical actinometer, has been extensively applied in wastewater and environment treatment. Despite many experimental efforts to improve clarity, important aspects of the mechanism of ferrioxalate photolysis are still under debate. In this paper, we employ the recently developed WΓ-CASSCF to investigate the ligand-to-metal charge-transfer states that are key to ferrioxalate photolysis. This investigation provides a qualitative picture of these states and key potential energy surface features related to the photolysis. Our theoretical results are consistent with the prompt charge-transfer picture seen in recent experiments and clarify some features that are not visible in experiments. Two ligand-to-metal charge-transfer states contribute to the photolysis of ferrioxalate, and the avoided crossing barrier between them is low compared with the initial photoexcitation energy. Our data also clarify that one Fe-O bond cleaves first, followed by the C-C bond and the other Fe-O bond.

The E3 state of FeMoco: one hydride, two hydrides or dihydrogen?

Hydrides are present in the reduced states of the iron-molybdenum cofactor (FeMoco) of Mo nitrogenase and are believed to play a key mechanistic role in the dinitrogen reduction reaction catalyzed by the enzyme. Two hydrides are present in the E4 state according to 1H ENDOR and there is likely a single hydride in the E2 redox state. The 2-hydride E4 state has been experimentally observed to bind N2 and it has been speculated that E3 may bind N2 as well. However, the E3 state has not been directly observed and very little is known about its molecular and electronic structure or reactivity. In recent computational studies, we have explored the energy surfaces of the E2 and E4 by QM/MM modelling, and found that the most stable hydride isomers contain bridging or partially bridging hydrides with an open protonated belt sulfide-bridge. In this work we systematically explore the energy surface of the E3 redox state, comparing single hydride and two-hydride isomers with varying coordination and bridging vs. terminal sulfhydryl groups. We also include a model featuring a triply protonated carbide. The results are only mildly dependent on the QM-region size. The three most stable E3 isomers at the r2SCAN level of theory have in common: an open belt sulfide-bridge (terminal sulfhydryl group on Fe6) and either 2 bridging hydrides (between Fe2 and Fe6), 1 bridging-1-terminal hydride (around Fe2 and Fe6) or a dihydrogen ligand bound at the Fe2 site. Analyzing the functional dependency of the results, we find that functionals previously found to predict accurate structures of spin-coupled Fe/Mo dimers and FeMoco (TPSSh, B97-D3, r2SCAN, and B3LYP*) are in generally good agreement about the stability of these 3 E3 isomers. However, B3LYP*, similar to its parent B3LYP method, predicts a triply protonated carbide isomer as the most stable isomer, an unlikely scenario in view of the lack of experimental evidence for carbide protonation occurring in reduced FeMoco states. Distinguishing further between the 3 hydride isomers is difficult and this flexible coordination nature of hydrides suggests that multiple hydride isomers could be present during experimental conditions. N2 binding was explored and resulted in geometries with 2 bridging hydrides and N2 bound to either Fe2 or Fe6 with a local low-spin state on the Fe. N2 binding is predicted to be mildly endothermic, similar to the E2 state, and it seems unlikely that the E3 state is capable of binding N2.

First principles reaction discovery: from the Schrodinger equation to experimental prediction for methane pyrolysis

Our recent success in exploiting graphical processing units (GPUs) to accelerate quantum chemistry computations led to the development of the ab initio nanoreactor, a computational framework for automatic reaction discovery and kinetic model construction. In this work, we apply the ab initio nanoreactor to methane pyrolysis, from automatic reaction discovery to path refinement and kinetic modeling. Elementary reactions occurring during methane pyrolysis are revealed using GPU-accelerated ab initio molecular dynamics simulations. Subsequently, these reaction paths are refined at a higher level of theory with optimized reactant, product, and transition state geometries. Reaction rate coefficients are calculated by transition state theory based on the optimized reaction paths. The discovered reactions lead to a kinetic model with 53 species and 134 reactions, which is validated against experimental data and simulations using literature kinetic models. We highlight the advantage of leveraging local brute force and Monte Carlo sensitivity analysis approaches for efficient identification of important reactions. Both sensitivity approaches can further improve the accuracy of the methane pyrolysis kinetic model. The results in this work demonstrate the power of the ab initio nanoreactor framework for computationally affordable systematic reaction discovery and accurate kinetic modeling.

Computational Discovery of Stable Metal-Organic Frameworks for Methane-to-Methanol Catalysis

The challenge of direct partial oxidation of methane to methanol has motivated the targeted search of metal-organic frameworks (MOFs) as a promising class of materials for this transformation because of their site-isolated metals with tunable ligand environments. Thousands of MOFs have been synthesized, yet relatively few have been screened for their promise in methane conversion. We developed a high-throughput virtual screening workflow that identifies MOFs from a diverse space of experimental MOFs that have not been studied for catalysis, yet are thermally stable, synthesizable, and have promising unsaturated metal sites for C-H activation via a terminal metal-oxo species. We carried out density functional theory calculations of the radical rebound mechanism for methane-to-methanol conversion on models of the secondary building units (SBUs) from 87 selected MOFs. While we showed that oxo formation favorability decreases with increasing 3d filling, consistent with prior work, previously observed scaling relations between oxo formation and hydrogen atom transfer (HAT) are disrupted by the greater diversity in our MOF set. Accordingly, we focused on Mn MOFs, which favor oxo intermediates without disfavoring HAT or leading to high methanol release energies─a key feature for methane hydroxylation activity. We identified three Mn MOFs comprising unsaturated Mn centers bound to weak-field carboxylate ligands in planar or bent geometries with promising methane-to-methanol kinetics and thermodynamics. The energetic spans of these MOFs are indicative of promising turnover frequencies for methane to methanol that warrant further experimental catalytic studies.

Development and Benchmarking of Open Force Field 2.0.0: The Sage Small Molecule Force Field

We introduce the Open Force Field (OpenFF) 2.0.0 small molecule force field for drug-like molecules, code-named Sage, which builds upon our previous iteration, Parsley. OpenFF force fields are based on direct chemical perception, which generalizes easily to highly diverse sets of chemistries based on substructure queries. Like the previous OpenFF iterations, the Sage generation of OpenFF force fields was validated in protein-ligand simulations to be compatible with AMBER biopolymer force fields. In this work, we detail the methodology used to develop this force field, as well as the innovations and improvements introduced since the release of Parsley 1.0.0. One particularly significant feature of Sage is a set of improved Lennard-Jones (LJ) parameters retrained against condensed phase mixture data, the first refit of LJ parameters in the OpenFF small molecule force field line. Sage also includes valence parameters refit to a larger database of quantum chemical calculations than previous versions, as well as improvements in how this fitting is performed. Force field benchmarks show improvements in general metrics of performance against quantum chemistry reference data such as root-mean-square deviations (RMSD) of optimized conformer geometries, torsion fingerprint deviations (TFD), and improved relative conformer energetics (ΔΔE). We present a variety of benchmarks for these metrics against our previous force fields as well as in some cases other small molecule force fields. Sage also demonstrates improved performance in estimating physical properties, including comparison against experimental data from various thermodynamic databases for small molecule properties such as ΔHmix, ρ(x), ΔGsolv, and ΔGtrans. Additionally, we benchmarked against protein-ligand binding free energies (ΔGbind), where Sage yields results statistically similar to previous force fields. All the data is made publicly available along with complete details on how to reproduce the training results at https://github.com/openforcefield/openff-sage.

Linearized Pair-Density Functional Theory

Multiconfiguration pair-density functional theory (MC-PDFT) is a post-SCF multireference method that has been successful at computing ground- and excited-state energies. However, MC-PDFT is a single-state method in which the final MC-PDFT energies do not come from diagonalization of a model-space Hamiltonian matrix, and this can lead to inaccurate topologies of potential energy surfaces near locally avoided crossings and conical intersections. Therefore, in order to perform physically correct ab initio molecular dynamics with electronically excited states or to treat Jahn-Teller instabilities, it is necessary to develop a PDFT method that recovers the correct topology throughout the entire nuclear configuration space. Here we construct an effective Hamiltonian operator, called the linearized PDFT (L-PDFT) Hamiltonian, by expanding the MC-PDFT energy expression to first order in a Taylor series of the wave function density. Diagonalization of the L-PDFT Hamiltonian gives the correct potential energy surface topology near conical intersections and locally avoided crossings for a variety of challenging cases including phenol, methylamine, and the spiro cation. Furthermore, L-PDFT outperforms MC-PDFT and previous multistate PDFT methods for predicting vertical excitations from a variety of representative organic chromophores.

Drude Polarizable Lipid Force Field with Explicit Treatment of Long-Range Dispersion: Parametrization and Validation for Saturated and Monounsaturated Zwitterionic Lipids

Accurate empirical force fields of lipid molecules are a critical component of molecular dynamics simulation studies aimed at investigating properties of monolayers, bilayers, micelles, vesicles, and liposomes, as well as heterogeneous systems, such as protein-membrane complexes, bacterial cell walls, and more. While the majority of lipid force field-based simulations have been performed using pairwise-additive nonpolarizable models, advances have been made in the development of the polarizable force field based on the classical Drude oscillator model. In the present study, we undertake further optimization of the Drude lipid force field, termed Drude2023, including improved treatment of the phosphate and glycerol linker region of PC and PE headgroups, additional optimization of the alkene group in monounsaturated lipids, and inclusion of long-range Lennard-Jones interactions using the particle-mesh Ewald method. Initial optimization targeted quantum mechanical (QM) data on small model compounds representative of the linker region. Subsequent optimization targeted QM data on larger model compounds, experimental data, and dihedral potentials of mean force from the CHARMM36 additive lipid force field using a parameter reweighting protocol. The use of both experimental and QM target data during the reweighting protocol is shown to produce physically reasonable parameters that reproduce a collection of experimental observables. Target data for optimization included surface area/lipid for DPPC, DSPC, DMPC, and DLPC bilayers and nuclear magnetic resonance (NMR) order parameters for DPPC bilayers. Validation data include prediction of membrane thickness, scattering form factors, electrostatic potential profiles, compressibility moduli, surface area per lipid, water permeability, NMR T1 relaxation times, diffusion constants, and monolayer surface tensions for a variety of saturated and unsaturated lipid mono- and bilayers. Overall, the agreement with experimental data is quite good, though the results are less satisfactory for the NMR T1 relaxation times for carbons near the ester groups. Notable improvements compared to the additive C36 force field were obtained for membrane dipole potentials, lipid diffusion coefficients, and water permeability with the exception of monounsaturated lipid bilayers. It is anticipated that the optimized polarizable Drude2023 force field will help generate more accurate molecular simulations of pure bilayers and heterogeneous systems containing membranes, advancing our understanding of the role of electronic polarization in these systems.

Fractional Charge Density Functional Theory and Its Application to the Electro-inductive Effect

We employed the chemical potential equalization principle to demonstrate that fractional electrons are involved in the electro-inductive effect as well as the vibrational Stark effect. By the chemical potential model, we were able to deduce that the frontier molecular orbitals of immobilized molecules can provide valuable insight into these effects. To further understand and quantify these findings, we introduced fractional charge density functional theory (FC-DFT), a canonical ensemble approach for open systems. This method allows for the calculation of electronic energies, nuclear gradients, and the Hessian matrix of fractional electronic systems. To correct the spurious delocalization error commonly found in approximate density functionals for small systems, we imposed the Perdew-Parr-Levy-Balduz (PPLB) condition through linear interpolation of two adjacent integer points (LI-FC-DFT). Although this approach is relatively simple in terms of molecular modeling, the results obtained through LI-FC-DFT calculations predict the same trend seen in experimental reactivity and the frequency change of immobilized molecules.

Insights into the deviation from piecewise linearity in transition metal complexes from supervised machine learning models

Virtual high-throughput screening (VHTS) and machine learning (ML) with density functional theory (DFT) suffer from inaccuracies from the underlying density functional approximation (DFA). Many of these inaccuracies can be traced to the lack of derivative discontinuity that leads to a curvature in the energy with electron addition or removal. Over a dataset of nearly one thousand transition metal complexes typical of VHTS applications, we computed and analyzed the average curvature (i.e., deviation from piecewise linearity) for 23 density functional approximations spanning multiple rungs of "Jacob's ladder". While we observe the expected dependence of the curvatures on Hartree-Fock exchange, we note limited correlation of curvature values between different rungs of "Jacob's ladder". We train ML models (i.e., artificial neural networks or ANNs) to predict the curvature and the associated frontier orbital energies for each of these 23 functionals and then interpret differences in curvature among the different DFAs through analysis of the ML models. Notably, we observe spin to play a much more important role in determining the curvature of range-separated and double hybrids in comparison to semi-local functionals, explaining why curvature values are weakly correlated between these and other families of functionals. Over a space of 187.2k hypothetical compounds, we use our ANNs to pinpoint DFAs for which representative transition metal complexes have near-zero curvature with low uncertainty, demonstrating an approach to accelerate screening of complexes with targeted optical gaps.

Quantum Alchemy Based Bonding Trends and Their Link to Hammett's Equation and Pauling's Electronegativity Model

We present an intuitive and general analytical approximation estimating the energy of covalent single and double bonds between participating atoms in terms of their respective nuclear charges with just three parameters, [E_AB ≈ a - bZ_AZ_B + c(Z_A^7/3 + Z_B^7/3) ]. The functional form of our expression models an alchemical atomic energy decomposition between participating atoms A and B. After calibration, reasonably accurate bond dissociation energy estimates are obtained for hydrogen-saturated diatomics composed of p-block elements coming from the same row 2 ≤ n ≤ 4 in the periodic table. Corresponding changes in bond dissociation energies due to substitution of atom B by C can be obtained via simple formulas. While being of different functional form and origin, our model is as simple and accurate as Pauling's well-known electronegativity model. Analysis indicates that the model's response in covalent bonding to variation in nuclear charge is near-linear, which is consistent with Hammett's equation.

Active Learning Exploration of Transition-Metal Complexes to Discover Method-Insensitive and Synthetically Accessible Chromophores

Transition-metal chromophores with earth-abundant transition metals are an important design target for their applications in lighting and nontoxic bioimaging, but their design is challenged by the scarcity of complexes that simultaneously have well-defined ground states and optimal target absorption energies in the visible region. Machine learning (ML) accelerated discovery could overcome such challenges by enabling the screening of a larger space but is limited by the fidelity of the data used in ML model training, which is typically from a single approximate density functional. To address this limitation, we search for consensus in predictions among 23 density functional approximations across multiple rungs of "Jacob's ladder". To accelerate the discovery of complexes with absorption energies in the visible region while minimizing the effect of low-lying excited states, we use two-dimensional (2D)efficient global optimization to sample candidate low-spin chromophores from multimillion complex spaces. Despite the scarcity (i.e., ∼0.01%) of potential chromophores in this large chemical space, we identify candidates with high likelihood (i.e., >10%) of computational validation as the ML models improve during active learning, representing a 1000-fold acceleration in discovery. Absorption spectra of promising chromophores from time-dependent density functional theory verify that 2/3 of candidates have the desired excited-state properties. The observation that constituent ligands from our leads have demonstrated interesting optical properties in the literature exemplifies the effectiveness of our construction of a realistic design space and active learning approach.

Low-cost machine learning prediction of excited state properties of iridium-centered phosphors

Prediction of the excited state properties of photoactive iridium complexes challenges ab initio methods such as time-dependent density functional theory (TDDFT) both from the perspective of accuracy and of computational cost, complicating high-throughput virtual screening (HTVS). We instead leverage low-cost machine learning (ML) models and experimental data for 1380 iridium complexes to perform these prediction tasks. We find the best-performing and most transferable models to be those trained on electronic structure features from low-cost density functional tight binding calculations. Using artificial neural network (ANN) models, we predict the mean emission energy of phosphorescence, the excited state lifetime, and the emission spectral integral for iridium complexes with accuracy competitive with or superseding that of TDDFT. We conduct feature importance analysis to determine that high cyclometalating ligand ionization potential correlates to high mean emission energy, while high ancillary ligand ionization potential correlates to low lifetime and low spectral integral. As a demonstration of how our ML models can be used for HTVS and the acceleration of chemical discovery, we curate a set of novel hypothetical iridium complexes and use uncertainty-controlled predictions to identify promising ligands for the design of new phosphors while retaining confidence in the quality of the ANN predictions.

Accurate Hellmann-Feynman forces from density functional calculations with augmented Gaussian basis sets

The Hellmann-Feynman (HF) theorem provides a way to compute forces directly from the electron density, enabling efficient force calculations for large systems through machine learning (ML) models for the electron density. The main issue holding back the general acceptance of the HF approach for atom-centered basis sets is the well-known Pulay force which, if naively discarded, typically constitutes an error upward of 10 eV/Å in forces. In this work, we demonstrate that if a suitably augmented Gaussian basis set is used for density functional calculations, the Pulay force can be suppressed, and HF forces can be computed as accurately as analytical forces with state-of-the-art basis sets, allowing geometry optimization and molecular dynamics to be reliably performed with HF forces. Our results pave a clear path forward for the accurate and efficient simulation of large systems using ML densities and the HF theorem.

State averaged CASSCF in AMOEBA polarizable water model for simulating nonadiabatic molecular dynamics with nonequilibrium solvation effects

This paper presents a state-averaged complete active space self-consistent field (SA-CASSCF) in the atomic multipole optimized energetics for biomolecular application (AMOEBA) polarizable water model, which enables rigorous simulation of non-adiabatic molecular dynamics with nonequilibrium solvation effects. The molecular orbital and configuration interaction coefficients of the solute wavefunction, and the induced dipoles on solvent atoms, are solved by minimizing the state averaged energy variationally. In particular, by formulating AMOEBA water models and the polarizable continuum model (PCM) in a unified way, the algorithms developed for computing SA-CASSCF/PCM energies, analytical gradients, and non-adiabatic couplings in our previous work can be generalized to SA-CASSCF/AMOEBA by properly substituting a specific list of variables. Implementation of this method will be discussed with the emphasis on how the calculations of different terms are partitioned between the quantum chemistry and molecular mechanics codes. We will present and discuss results that demonstrate the accuracy and performance of the implementation. Next, we will discuss results that compare three solvent models that work with SA-CASSCF, i.e., PCM, fixed-charge force fields, and the newly implemented AMOEBA. Finally, the new SA-CASSCF/AMOEBA method has been interfaced with the ab initio multiple spawning method to carry out non-adiabatic molecular dynamics simulations. This method is demonstrated by simulating the photodynamics of the model retinal protonated Schiff base molecule in water.

A transferable recommender approach for selecting the best density functional approximations in chemical discovery

Approximate density functional theory has become indispensable owing to its balanced cost-accuracy trade-off, including in large-scale screening. To date, however, no density functional approximation (DFA) with universal accuracy has been identified, leading to uncertainty in the quality of data generated from density functional theory. With electron density fitting and Δ-learning, we build a DFA recommender that selects the DFA with the lowest expected error with respect to the gold standard (but cost-prohibitive) coupled cluster theory in a system-specific manner. We demonstrate this recommender approach on the evaluation of vertical spin splitting energies of transition metal complexes. Our recommender predicts top-performing DFAs and yields excellent accuracy (about 2 kcal mol-1) for chemical discovery, outperforming both individual Δ-learning models and the best conventional single-functional approach from a set of 48 DFAs. By demonstrating transferability to diverse synthesized compounds, our recommender potentially addresses the accuracy versus scope dilemma broadly encountered in computational chemistry.