Locating critical points on multi-dimensional surfaces by genetic algorithm: test cases including normal and perturbed argon clusters

Why an integrated approach between search algorithms and chemical intuition is necessary?

Though search algorithms are appropriate tools for identifying low-energy isomers, fixing several constraints seems to be a fundamental prerequisite to successfully running any structural search program. This causes some potential setbacks as far as identifying all possible isomers, close to the lowest-energy isomer, for any elemental composition. The number of explored candidates, the choice of method, basis set, and availability of CPU time needed to analyze the various initial test structures become necessary restrictions in resolving the issues of structural isomerism reasonably. While one could arrive at new structures through chemical intuition, reproducing or achieving those exact same structures requires increasing the number of variables in any given program, which causes further constraints in exploring the potential energy surface in a reasonable amount of time. Thus, it is emphasized here that an integrated approach between search algorithms and chemical intuition is necessary by taking the C₁₂O₂Mg₂ system as an example. Our initial search through the AUTOMATON program yielded 1450 different geometries. However, through chemical intuition, we found eighteen new geometries within 40.0 kcal mol^-1 at the PBE0-D3/def2-TZVP level. These results indirectly emphasize that an integrated approach between search algorithms and chemical intuition is necessary to further our knowledge in chemical space for any given elemental composition.

Machine Learning to Predict Diels-Alder Reaction Barriers from the Reactant State Electron Density

Reaction barriers are key to our understanding of chemical reactivity and catalysis. Certain reactions are so seminal in chemistry that countless variants, with or without catalysts, have been studied, and their barriers have been computed or measured experimentally. This wealth of data represents a perfect opportunity to leverage machine learning models, which could quickly predict barriers without explicit calculations or measurement. Here, we show that the topological descriptors of the quantum mechanical charge density in the reactant state constitute a set that is both rigorous and continuous and can be used effectively for the prediction of reaction barrier energies to a high degree of accuracy. We demonstrate this on the Diels-Alder reaction, highly important in biology and medicinal chemistry, and as such, studied extensively. This reaction exhibits a range of barriers as large as 270 kJ/mol. While we trained our single-objective supervised (labeled) regression algorithms on simpler Diels-Alder reactions in solution, they predict reaction barriers also in significantly more complicated contexts, such a Diels-Alder reaction catalyzed by an artificial enzyme and its evolved variants, in agreement with experimental changes in k_cat. We expect this tool to apply broadly to a variety of reactions in solution or in the presence of a catalyst, for screening and circumventing heavily involved computations or experiments.

Si2C5H2 isomers - search algorithms versus chemical intuition

The pros and cons of using search algorithms alone in identifying new geometries have been discussed by using the Si₂C₅H₂ elemental composition as an example. Within 30 kcal mol^-1 at the CCSD(T)/def2-TZVP//PBE0/def2-TZVP level of theory, the coalescence kick and cuckoo methods postulate merely four isomers (1, 3, 6, and 7) for Si₂C₅H₂ (O. Yañez et. al., Chem. Commun., 2017, 53, 12112). On the contrary, chemical intuition yields fourteen (2, 4, 5, and 8-18) new isomers within the same energy range at the B3LYP/6-311++G(2d,2p) level of theory. Based on the relative energies of the first eleven isomers of Si₂C₅H₂ (1, C_2v, 0.00; 2, C_s, 21.39; 3, C_s, 21.95; 4, C_s, 22.76; 5, C_s, 24.74; 6, C_s, 25.34; 7, C_s, 25.64; 8, C_s, 25.79; 9, C_s, 27.20; 10, C_2v, 28.59; and 11, C_2v, 29.16 kcal mol^-1) calculated at the CCSD(T)/cc-pVTZ level of theory, it is evident that the search algorithms had missed at least seven isomers in the same energy range. The relative energy gaps of isomers 12-18 fall in the range of 30-40 kcal mol^-1 at the latter level of theory. Consequentially, this scenario triggers a speculation going forward with search algorithms alone in the search of all new isomers. While one cannot underestimate the power of these algorithms, the role of chemical intuition may not be completely neglected. Retrospectively, the fourteen new isomers found by chemical intuition may help in writing better search algorithms. All eighteen isomers - including the most stable isomer with a planar tetracoordinate carbon atom 1- remain elusive in the laboratory to date. Thus, structural and spectroscopic parameters have been presented here, which may possibly aid the future experimental studies.

Coulomb explosion in dicationic noble gas clusters: a genetic algorithm-based approach to critical size estimation for the suppression of Coulomb explosion and prediction of dissociation channels

We present a genetic algorithm based investigation of structural fragmentation in dicationic noble gas clusters, Ar(n)(+2), Kr(n)(+2), and Xe(n)(+2), where n denotes the size of the cluster. Dications are predicted to be stable above a threshold size of the cluster when positive charges are assumed to remain localized on two noble gas atoms and the Lennard-Jones potential along with bare Coulomb and ion-induced dipole interactions are taken into account for describing the potential energy surface. Our cutoff values are close to those obtained experimentally [P. Scheier and T. D. Mark, J. Chem. Phys. 11, 3056 (1987)] and theoretically [J. G. Gay and B. J. Berne, Phys. Rev. Lett. 49, 194 (1982)]. When the charges are allowed to be equally distributed over four noble gas atoms in the cluster and the nonpolarization interaction terms are allowed to remain unchanged, our method successfully identifies the size threshold for stability as well as the nature of the channels of dissociation as function of cluster size. In Ar(n)(2+), for example, fissionlike fragmentation is predicted for n=55 while for n=43, the predicted outcome is nonfission fragmentation in complete agreement with earlier work [Golberg et al., J. Chem. Phys. 100, 8277 (1994)].

Determination, prediction, and understanding of structures, using the energy landscapes of chemical systems – Part II

In the past decade, new theoretical approaches have been developed to determine, predict and understand the struc-ture of chemical compounds. The central element of these methods has been the investigation of the energy landscape of chemical systems. Applications range from extended crystalline and amorphous compounds over clusters and molecular crystals to proteins. In this review, we are going to give an introduction to energy landscapes and methods for their investigation, together with a number of examples. These include structure prediction of extended and mo-lecular crystals, structure prediction and folding of proteins, structure analysis of zeolites, and structure determination of crystals from powder diffraction data.

Mindless Chemistry

Applications of an automated stochastic search procedure for locating all possible minima with a given composition are illustrated by the pentatomic molecules BCNOS, CAlSiPS, C(4)B(-), C(4)Al(-), and CBe(4)(2-), as well as by C(6)Be, the C(6)Be(2-) dianion, and C(6)H(2). All previously identified minima were reproduced, and many new structures, often with nonintuitive geometries, were found.

A fast annealing evolutionary algorithm for global optimization

By combining the aspect of population in genetic algorithms (GAs) and the simulated annealing algorithm (SAA), a novel algorithm, called fast annealing evolutionary algorithm (FAEA), is proposed. The algorithm is similar to the annealing evolutionary algorithm (AEA), and a very fast annealing technique is adopted for the annealing procedure. By an application of the algorithm to the optimization of test functions and a comparison of the algorithm with other stochastic optimization methods, it is shown that the algorithm is a highly efficient optimization method. It was also applied in optimization of Lennard-Jones clusters and compared with other methods in this study. The results indicate that the algorithm is a good tool for the energy minimization problem.