Progress towards machine learning reaction rate constants.

Quantum and classical reaction rate constant calculations come at the cost of exploring potential energy surfaces. Due to the "curse of dimensionality", their evaluation quickly becomes unfeasible as the system size grows. Machine learning algorithms can accelerate the calculation of reaction rate constants by predicting them using low cost input features. In this perspective, we briefly introduce supervised machine learning algorithms in the context of reaction rate constant prediction. We discuss existing and recently created kinetic datasets and input feature representations as well as the use and design of machine learning algorithms to predict reaction rate constants or quantities required for their computation. Amongst these, we first describe the use of machine learning to predict activation, reaction, solvation and dissociation energies. We then look at the use of machine learning to predict reactive force field parameters, reaction rate constants as well as to help accelerate the search for minimum energy paths. Lastly, we provide an outlook on areas which have yet to be explored so as to improve and evaluate the use of machine learning algorithms for chemical reaction rate constants.

Machine Learning Quantum Reaction Rate Constants.

The ab initio calculation of exact quantum reaction rate constants comes at a high cost due to the required dynamics of reactants on multidimensional potential energy surfaces. In turn, this impedes the rapid design of the kinetics for large sets of coupled reactions. In an effort to overcome this hurdle, a deep neural network (DNN) was trained to predict the logarithm of quantum reaction rate constants multiplied by their reactant partition function-rate products. The training dataset was generated in-house and contains ∼1.5 million quantum reaction rate constants for single, double, symmetric and asymmetric one-dimensional potentials computed over a broad range of reactant masses and temperatures. The DNN was able to predict the logarithm of the rate product with a relative error of 1.1%. Furthermore, when comparing the difference between the DNN prediction and classical transition state theory at temperatures below 300 K a relative percent error of 31% was found with respect to the exact difference. Systems beyond the test set were also studied, these included the H + H2 reaction, the diffusion of hydrogen on Ni(100), the Menshutkin reaction of pyridine with CH3Br in the gas phase, the reaction of formalcyanohydrin with HS- in water and the F + HCl reaction. For these reactions, the DNN predictions were accurate at high temperatures and in good agreement with the exact rates at lower temperatures. This work shows that one can take advantage of a DNN to gain insight on reactivity in the quantum regime.

Reaction Dynamics of Cyanohydrins with Hydrosulfide in Water.

We studied the reaction dynamics of a proposed prebiotic reaction theoretically. The chemical process involves acetone cyanohydrin or formalcyanohydrin reacting with hydrosulfide in an aqueous environment. Rate constants and populations of reactant and product bimolecular geometric orientations for the reactions were obtained by using density functional theory for the energies, transition-state theory for the rates, and matrix exponentiation as well as the hybrid tau-leaping algorithm for the population dynamics. The role of including the solvent explicitly versus implicitly was also investigated. We found that adding explicit water or hydrogen sulfide molecules lowers the activation energy barrier and leads to a more efficient reaction pathway. In particular, hydrogen sulfide was a better proton donor. Finally, we studied the role of including more than one reactant and product bimolecular orientation geometry in the dynamics. Including all initial pathways, reactant to reactant, product to product, reactant to product, and product to reactant led to a larger reaction rate constant compared to the single minimum energy pathway approach. Overall, we found that most reactions which involve formalcyanohydrin occur more rapidly or at the same speed as reactions which involve acetone cyanohydrin at room temperature.

Atomistic study of energy funneling in the light-harvesting complex of green sulfur bacteria.

Phototrophic organisms such as plants, photosynthetic bacteria, and algae use microscopic complexes of pigment molecules to absorb sunlight. Within the light-harvesting complexes, which frequently have several functional and structural subunits, the energy is transferred in the form of molecular excitations with very high efficiency. Green sulfur bacteria are considered to be among the most efficient light-harvesting organisms. Despite multiple experimental and theoretical studies of these bacteria, the physical origin of the efficient and robust energy transfer in their light-harvesting complexes is not well understood. To study excitation dynamics at the systems level, we introduce an atomistic model that mimics a complete light-harvesting apparatus of green sulfur bacteria. The model contains approximately 4000 pigment molecules and comprises a double wall roll for the chlorosome, a baseplate, and six Fenna-Matthews-Olson trimer complexes. We show that the fast relaxation within functional subunits combined with the transfer between collective excited states of pigments can result in robust energy funneling to the initial excitation conditions and temperature changes. Moreover, the same mechanism describes the coexistence of multiple time scales of excitation dynamics frequently observed in ultrafast optical experiments. While our findings support the hypothesis of supertransfer, the model reveals energy transport through multiple channels on different length scales.

Atomistic study of the long-lived quantum coherences in the Fenna-Matthews-Olson complex.

A remarkable amount of theoretical research has been carried out to elucidate the physical origins of the recently observed long-lived quantum coherence in the electronic energy transfer process in biological photosynthetic systems. Although successful in many respects, several widely used descriptions only include an effective treatment of the protein-chromophore interactions. In this work, by combining an all-atom molecular dynamics simulation, time-dependent density functional theory, and open quantum system approaches, we successfully simulate the dynamics of the electronic energy transfer of the Fenna-Matthews-Olson pigment-protein complex. The resulting characteristic beating of populations and quantum coherences is in good agreement with the experimental results and the hierarchy equation of motion approach. The experimental absorption, linear, and circular dichroism spectra and dephasing rates are recovered at two different temperatures. In addition, we provide an extension of our method to include zero-point fluctuations of the vibrational environment. This work thus presents, to our knowledge, one of the first steps to explain the role of excitonic quantum coherence in photosynthetic light-harvesting complexes based on their atomistic and molecular description.

On the alternatives for bath correlators and spectral densities from mixed quantum-classical simulations.

We investigate on the procedure of extracting a "spectral density" from mixed QM/MM calculations and employing it in open quantum systems models. In particular, we study the connection between the energy gap correlation function extracted from ground state QM/MM and the bath spectral density used as input in open quantum system approaches. We introduce a simple model which can give intuition on when the ground state QM/MM propagation will give the correct energy gap. We also discuss the role of higher order correlators of the energy-gap fluctuations which can provide useful information on the bath. Further, various semiclassical corrections to the spectral density, are applied and investigated. Finally, we apply our considerations to the photosynthetic Fenna-Matthews-Olson complex. For this system, our results suggest the use of the Harmonic prefactor for the spectral density rather than the Standard one, which was employed in the simulations of the system carried out to date.

Exciton transport in thin-film cyanine dye J-aggregates.

We present a theoretical model for the study of exciton dynamics in J-aggregated monolayers of fluorescent dyes. The excitonic evolution is described by a Monte-Carlo wave function approach which allows for a unified description of the quantum (ballistic) and classical (diffusive) propagation of an exciton on a lattice in different parameter regimes. The transition between the ballistic and diffusive regime is controlled by static and dynamic disorder. As an example, the model is applied to three cyanine dye J-aggregates: TC, TDBC, and U3. Each of the molecule-specific structure and excitation parameters are estimated using time-dependent density functional theory. The exciton diffusion coefficients are calculated and analyzed for different degrees of film disorder and are correlated to the physical properties and the structural arrangement of molecules in the aggregates. Further, exciton transport is anisotropic and dependent on the initial exciton energy. The upper-bound estimation of the exciton diffusion length in the TDBC thin-film J-aggregate is of the order of hundreds of nanometers, which is in good qualitative agreement with the diffusion length estimated from experiments.

Low-cost prediction of molecular and transition state partition functions via machine learning.

We have generated an open-source dataset of over 30 000 organic chemistry gas phase partition functions. With this data, a machine learning deep neural network estimator was trained to predict partition functions of unknown organic chemistry gas phase transition states. This estimator only relies on reactant and product geometries and partition functions. A second machine learning deep neural network was trained to predict partition functions of chemical species from their geometry. Our models accurately predict the logarithm of test set partition functions with a maximum mean absolute error of 2.7%. Thus, this approach provides a means to reduce the cost of computing reaction rate constants ab initio. The models were also used to compute transition state theory reaction rate constant prefactors and the results were in quantitative agreement with the corresponding ab initio calculations with an accuracy of 98.3% on the log scale.

First principles semiclassical calculations of vibrational eigenfunctions.

Vibrational eigenfunctions are calculated on-the-fly using semiclassical methods in conjunction with ab initio density functional theory classical trajectories. Various semiclassical approximations based on the time-dependent representation of the eigenfunctions are tested on an analytical potential describing the chemisorption of CO on Cu(100). Then, first principles semiclassical vibrational eigenfunctions are calculated for the CO(2) molecule and its accuracy evaluated. The multiple coherent states initial value representations semiclassical method recently developed by us has shown with only six ab initio trajectories to evaluate eigenvalues and eigenfunctions at the accuracy level of thousands trajectory semiclassical initial value representation simulations.

Absence of Selection for Quantum Coherence in the Fenna-Matthews-Olson Complex: A Combined Evolutionary and Excitonic Study.

We present a study on the evolution of the Fenna-Matthews-Olson bacterial photosynthetic pigment-protein complex. This protein complex functions as an antenna. It transports absorbed photons-excitons-to a reaction center where photosynthetic reactions initiate. The efficiency of exciton transport is therefore fundamental for the photosynthetic bacterium's survival. We have reconstructed an ancestor of the complex to establish whether coherence in the exciton transport was selected for or optimized over time. We have also investigated the role of optimizing free energy variation upon folding in evolution. We studied whether mutations which connect the ancestor to current day species were stabilizing or destabilizing from a thermodynamic viewpoint. From this study, we established that most of these mutations were thermodynamically neutral. Furthermore, we did not see a large change in exciton transport efficiency or coherence, and thus our results predict that exciton coherence was not specifically selected for.

Machine learning exciton dynamics.

Obtaining the exciton dynamics of large photosynthetic complexes by using mixed quantum mechanics/molecular mechanics (QM/MM) is computationally demanding. We propose a machine learning technique, multi-layer perceptrons, as a tool to reduce the time required to compute excited state energies. With this approach we predict time-dependent density functional theory (TDDFT) excited state energies of bacteriochlorophylls in the Fenna-Matthews-Olson (FMO) complex. Additionally we compute spectral densities and exciton populations from the predictions. Different methods to determine multi-layer perceptron training sets are introduced, leading to several initial data selections. In addition, we compute spectral densities and exciton populations. Once multi-layer perceptrons are trained, predicting excited state energies was found to be significantly faster than the corresponding QM/MM calculations. We showed that multi-layer perceptrons can successfully reproduce the energies of QM/MM calculations to a high degree of accuracy with prediction errors contained within 0.01 eV (0.5%). Spectral densities and exciton dynamics are also in agreement with the TDDFT results. The acceleration and accurate prediction of dynamics strongly encourage the combination of machine learning techniques with ab initio methods.

Influence of Force Fields and Quantum Chemistry Approach on Spectral Densities of BChl a in Solution and in FMO Proteins.

Studies on light-harvesting (LH) systems have attracted much attention after the finding of long-lived quantum coherences in the exciton dynamics of the Fenna-Matthews-Olson (FMO) complex. In this complex, excitation energy transfer occurs between the bacteriochlorophyll a (BChl a) pigments. Two quantum mechanics/molecular mechanics (QM/MM) studies, each with a different force-field and quantum chemistry approach, reported different excitation energy distributions for the FMO complex. To understand the reasons for these differences in the predicted excitation energies, we have carried out a comparative study between the simulations using the CHARMM and AMBER force field and the Zerner intermediate neglect of differential orbital (ZINDO)/S and time-dependent density functional theory (TDDFT) quantum chemistry methods. The calculations using the CHARMM force field together with ZINDO/S or TDDFT always show a wider spread in the energy distribution compared to those using the AMBER force field. High- or low-energy tails in these energy distributions result in larger values for the spectral density at low frequencies. A detailed study on individual BChl a molecules in solution shows that without the environment, the density of states is the same for both force field sets. Including the environmental point charges, however, the excitation energy distribution gets broader and, depending on the applied methods, also asymmetric. The excitation energy distribution predicted using TDDFT together with the AMBER force field shows a symmetric, Gaussian-like distribution.

Electromagnetic study of the chlorosome antenna complex of Chlorobium tepidum.

Green sulfur bacteria are an iconic example of nature's adaptation: thriving in environments of extremely low photon density, the bacterium ranks itself among the most efficient natural light-harvesting organisms. The photosynthetic antenna complex of this bacterium is a self-assembled nanostructure, ≈60 × 150 nm, made of bacteriochlorophyll molecules. We study the system from a computational nanoscience perspective by using electrodynamic modeling with the goal of understanding its role as a nanoantenna. Three different nanostructures, built from two molecular packing moieties, are considered: a structure built of concentric cylinders of aggregated bacteriochlorophyll d monomers, a single cylinder of bacteriochlorophyll c monomers, and a model for the entire chlorosome. The theoretical model captures both coherent and incoherent components of exciton transfer. The model is employed to extract optical spectra, concentration and depolarization of electromagnetic fields within the chlorosome, and fluxes of energy transfer for the structures. The second model nanostructure shows the largest field enhancement. Further, field enhancement is found to be more sensitive to dynamic noise rather than structural disorder. Field depolarization, however, is similar for all structures. This indicates that the directionality of transfer is robust to structural variations, while on the other hand, the intensity of transfer can be tuned by structural variations.