Interpolation Methods for Molecular Potential Energy Surface Construction

Characterization of iron(III) in aqueous and alkaline environments with ab initio and ReaxFF potentials

The iron(III) complexes [Fe(H2O)n(OH)m]3-m (n + m = 5, 6, m ≤ 3) and corresponding proton transfer reactions are studied with total energy calculations, the nudged elastic band (NEB) method, and molecular dynamics (MD) simulations using ab initio and a modification of reactive force field potentials, the ReaxFF-AQ potentials, based on the implementation according to Böhm et al. [J. Phys. Chem. C 120, 10849-10856 (2016)]. Applying ab initio potentials, the energies for the reactions [Fe(H2O)n(OH)m]3-m + H2O → [Fe(H2O)n-1(OH)m+1]2-m + H3O+ in a gaseous environment are in good agreement with comparable theoretical results. In an aqueous (aq) or alkaline environment, with the aid of NEB computations, respective minimum energy paths with energy barriers of up to 14.6 kcal/mol and a collective transfer of protons are modeled. Within MD simulations at room temperature, a permanent transfer of protons around the iron(III) ion is observed. The information gained concerning the geometrical and energetic properties of water and the [Fe(H2O)n(OH)m]3-m complexes from the ab initio computations has been used as reference data to optimize parameters for the O-H-Fe interaction within the ReaxFF-AQ approach. For the optimized ReaxFF-AQ parameter set, the statistical properties of the basic water model, such as the radial distribution functions and the proton hopping functions, are evaluated. For the [Fe(H2O)n(OH)m]3-m complexes, it was found that while geometrical and energetic properties are in good agreement with the ab initio data for gaseous environment, the statistical properties as obtained from the MD simulations are only partly in accordance with the ab initio results for the iron(III) complexes in aqueous or alkaline environments.

Surrogate Modeling of the Relative Entropy for Inverse Design Using Smolyak Sparse Grids

Relative entropy minimization, a statistical-mechanics approach for finding potential energy functions that produce target structural ensembles, has proven to be a powerful strategy for the inverse design of nanoparticle self-assembly. For a given target structure, the gradient of the relative entropy with respect to the adjustable parameters of the potential energy function is computed by performing a simulation, and then these parameters are updated using iterative gradient-based optimization. Small parameter updates per iteration and many iterations can be required for numerical stability, but this incurs considerable computational expense because a new simulation must be performed to reevaluate the gradient at each iteration. Here, we investigate the use of surrogate modeling to decouple the process of minimizing the relative entropy from the computationally demanding process of determining its gradient. We approximate the relative-entropy gradient using Chebyshev polynomial interpolation on Smolyak sparse grids. Our approach potentially increases the robustness and computational efficiency of using the relative entropy for inverse design, primarily for physically informed potential energy functions that have a small number of adjustable parameters.

Employing neural density functionals to generate potential energy surfaces

CONTEXT

With the union of machine learning (ML) and quantum chemistry, amid the debate between machine-learned functionals and human-designed functionals in density functional theory (DFT), this paper aims to demonstrate the generation of potential energy surfaces using computations with machine-learned density functional approximation (ML-DFA). A recent research trend is the application of ML in quantum sciences in the design of density functionals such as DeepMind's Deep Learning model (DeepMind21, DM21). Though science reported the state-of-the-art performance of DM21, the opportunity to utilize DeepMind's pretrained DM21 neural networks in computations in quantum chemistry has not yet been tapped. So far in the literature, the Deep Learning density functionals (DM21) have not been applied to generate potential energy surfaces. While the superior accuracy of DM21 has been reported, there is still a scarcity of publications that apply DM21 in calculations in the field. In this context, for the first time in literature, neural density functionals inferring 2D potential energy surfaces (ML-DFA-PES) based on machine-learned DFA-based computational method is contributed in this paper. This paper reports the ML-DFA-generated PES for C4H8, H2O, H2, and H2+ by employing a pretrained DM21m TensorFlow model with cc-pVDZ basis set. In addition, we also analyze the long-range behavior of DM21 based PES to investigate the ability to describe a system at long ranges. Furthermore, we compare PES diagrams from DM21 with popular DFT functionals (b3lyp/ PW6B95) and CCSD(T).

METHODS

In this method, 2D potential energy surfaces are obtained using a method that relies upon the neural network's ability to accurately learn the mapping between 3D electron density and exchange-correlation potential. By inserting Deep Learning inference in DFT with a pretrained neural network, self-consistent field (SCF) energy at different geometries along the coordinates of interest is computed, and then, potential energy surfaces are plotted. In this method, first, the electron density is computed mathematically, and this computed 3D electron density is used as a ML feature vector to predict the exchange correlation potential as a ML inference computed by a forward pass of pre-trained DM21 TensorFlow computational graph, followed by the computation of self-consistent field energy at multiple geometries, and then, SCF energies at different bond lengths/angles are plotted as 2D PES. We implement this in a python source code using frameworks such as PySCF and DM21. This paper contributes this implementation in open source. The source code and DM21-DFA-based PES are contributed at https://sites.google.com/view/MLfunctionals-DeepMind-PES .

Mapping spin contamination-free potential energy surfaces using restricted open-shell methods with Grassmannians

The Lagrange-based Grassmann interpolation (G-Int) method has been extended for open-shell systems using restricted open-shell (RO) methods. The performance of this method was assessed in constructing potential energy surfaces (PESs) for vanadium(II) oxide, benzyl radical, and methanesulfenyl chloride radical cation. The density matrices generated by G-Int when used as initial guesses for self-consistent field (SCF) calculations, exhibit superior performance compared to other traditional SCF initial guess schemes, such as SADMO, GWH, and CORE. Additionally, the energy obtained from the G-Int scheme satisfies the variational principle and outperforms the direct energy-based Lagrange interpolation approach. In the case of methanesulfenyl chloride radical cation, a unique example with a flat PES at the end region along the H-C-S-Cl dihedral angle, the use of an equally-spaced grid sampling leads to significant oscillations near the end of the interval due to the effects of Runge's phenomenon. Introducing an unequally-spaced grid sampling based on a scaled Gauss-Chebyshev quadrature effectively mitigated the Runge's phenomenon, making it suitable for combining with G-Int in constructing PESs for general applications. Thus, G-Int provides an efficient and robust strategy for building spin contamination-free PESs with consistent accuracy.

Generating accurate density matrices on the tangent space of a Grassmann manifold

Interpolating a density matrix from a set of known density matrices is not a trivial task. This is because a linear combination of density matrices does not necessarily correspond to another density matrix. In this Communication, density matrices are examined as objects of a Grassmann manifold. Although this manifold is not a vector space, its tangent space is a vector space. As a result, one can map the density matrices on this manifold to their corresponding vectors in the tangent space and then perform interpolations on that tangent space. The resulting interpolated vector can be mapped back to the Grassmann manifold, which can then be utilized (1) as an optimal initial guess for a self-consistent field (SCF) calculation or (2) to derive energy directly without time-consuming SCF iterations. Such a promising approach is denoted as Grassmann interpolation (G-Int). The hydrogen molecule has been used to illustrate that the described interpolated method in this work preserves the essential attributes of a density matrix. For phosphorus mononitride and ferrocene, it was demonstrated numerically that reference points for the definition of the corresponding tangent spaces can be chosen arbitrarily. In addition, the interpolated density matrices provide a superior and essentially converged initial guess for an SCF calculation to make the SCF procedure itself unnecessary. Finally, this accurate, efficient, robust, and systematically improved G-Int strategy has been used for the first time to generate highly accurate potential energy surfaces with fine details for the difficult case, ferrocene.

Portable Models for Entropy Effects on Kinetic Selectivity

Differences in entropies of competing transition states can direct kinetic selectivity. Understanding and modeling such entropy differences at the molecular level is complicated by the fact that entropy is statistical in nature; i.e., it depends on multiple vibrational states of transition structures, the existence of multiple dynamically accessible pathways past these transition structures, and contributions from multiple transition structures differing in conformation/configuration. The difficulties associated with modeling each of these contributors are discussed here, along with possible solutions, all with an eye toward the development of portable qualitative models of use to experimentalists aiming to design reactions that make use of entropy to control kinetic selectivity.

Mass spectrometric stochastic dynamic 3D structural analysis of mixture of steroids in solution - Experimental and theoretical study

There is explored, herein, functional relation: Experimental mass spectrometric phenomenon, obeying a certain scientific law ⇔ 3D molecular conformations and electronic structures of analytes obtained for quantum chemical theories. The paper answers to questions: (a) What evidence claims these actual relations among measurable and theoretical parameters, experimental factors and molecular properties; (b) how the provided evidence is collected and used; and (c) how empirical proof relates to assign and explain mass spectrometric phenomena of steroids afforded by our innovative stochastic dynamic mass spectrometric formula, D″_SD = 2.6388.10^-17.(<I²>-<I>²), quantum chemical 3D conformations, electronic structures and energetics of molecules, respectively. The paper address issue concerning empirical evidence at very high-to-exact level of assignment of 3D molecular conformations of steroids to experimental mass spectrometric fragment ions, accounting precisely for (i) effect of protonation; (ii) intramolecular rearrangement for A-D rings of steroidal skeleton and proton transfer effect, if any; in addition to (iii) examination of enantiomers of steroids in mixture with different stereochemistry, (R) and (S), of a set of six atoms of the molecular backbone of hydrocortisone (1), deoxycorticosterone (2), progesterone (3) and methyltestosterone (4), respectively. Results from testosterone (5) are discussed, as well. There are used ultra-high resolution atmospheric pressure chemical ionization mass spectrometric data on analytes (1)-(4) at ng.(mL)^-1 concentration levels in mixtures in solution obtained for positive operation mode. High accuracy static and molecular dynamic quantum chemical computations and chemometrics are also utilized. Experimental 3D structural parameters of steroids obtained for stochastic dynamic diffusion theory are correlated with available crystallographic data.

Adaptive Fitting of Potential Energy Surfaces of Small to Medium-Sized Molecules in Sum-of-Product Form: Application to Vibrational Spectroscopy

A semi-automatic sampling and fitting procedure for generating sum-of-product (Born-Oppenheimer) potential energy surfaces based on a high-dimensional model representation is presented. The adaptive sampling procedure and subsequent fitting relies on energies only and can be used for re-fitting existing analytic potential energy surfaces in sum-of-product form or for direct fits from ab initio computa- tions. The method is tested by fitting ground electronic state potential energy surfaces for small to medium sized semi-rigid molecules, i.e., HFCO, HONO, and HCOOH, based upon ab initio computations at the CCSD(T)-F12/cc-pVTZ-F12 or MP2/aug-cc-pVTZ levels of theory. Vibrational eigenstates are computed using block improved relaxation in the Heidelberg MCTDH package and compared to available experimental and theoretical data. The new potential energy surfaces are compared to the best ones currently available for these molecules, in terms of accuracy, including of resulting vibrational states, required numbers of sampling points, and number of fitting parameters. The present procedure leads to compact expansions and scales well with the number of dimensions for simple potentials such as single or double wells.