Explicitly correlated MRCI-F12 potential energy surfaces for methane fit with several permutation invariant schemes and full-dimensional vibrational calculations

Using nested tensor train contracted basis functions with group theoretical techniques to compute (ro)-vibrational spectra of molecules with non-Abelian groups

In this paper, we use nested tensor-train contractions to compute vibrational and ro-vibrational energy levels of molecules with five and six atoms. At each step, we fully exploit symmetry by using symmetry adapted basis functions obtained from an irreducible tensor method. Contracted basis functions are determined by diagonalizing reduced dimensional Hamiltonian matrices. The size of matrices of eigenvectors, used to account for coupling between groups of coordinates, is reduced by discarding rows and columns. The size of the matrices that must be diagonalized is thus substantially reduced, making it possible to use direct eigensolvers, even for molecules with five and six atoms. The symmetry-adapted contracted vibrational basis functions have been used to compute J = 0 energy levels of the CH3CN (C3v) and J > 0 levels of CH4.

A controlled study of the effect of deviations from symmetry of the potential energy surface (PES) on the accuracy of the vibrational spectrum computed with collocation

Symmetry, in particular permutational symmetry, of a potential energy surface (PES) is a useful property in quantum chemical calculations. It facilitates, in particular, state labelling and identification of degenerate states. In many practically important applications, however, these issues are unimportant. The imposition of exact symmetry and the perception that it is necessary create additional methodological requirements narrowing or complicating algorithmic choices that are thereby biased against methods and codes that by default do not incorporate symmetry, including most off-the-shelf machine learning methods that cannot be directly used if exact symmetry is demanded. By introducing symmetric and unsymmetric errors into the PES of H2CO in a controlled way and computing the vibrational spectrum with collocation using symmetric and nonsymmetric collocation point sets, we show that when the deviations from an ideal PES are random, imposition of exact symmetry does not bring any practical advantages. Moreover, a calculation ignoring symmetry may be more accurate. We also compare machine-learned PESs with and without symmetrization and demonstrate that there is no advantage of imposing exact symmetry for the accuracy of the vibrational spectrum.

Accurate fundamental invariant-neural network representation of ab initio potential energy surfaces

Highly accurate potential energy surfaces are critically important for chemical reaction dynamics. The large number of degrees of freedom and the intricate symmetry adaption pose a big challenge to accurately representing potential energy surfaces (PESs) for polyatomic reactions. Recently, our group has made substantial progress in this direction by developing the fundamental invariant-neural network (FI-NN) approach. Here, we review these advances, demonstrating that the FI-NN approach can represent highly accurate, global, full-dimensional PESs for reactive systems with even more than 10 atoms. These multi-channel reactions typically involve many intermediates, transition states, and products. The complexity and ruggedness of this potential energy landscape present even greater challenges for full-dimensional PES representation. These PESs exhibit a high level of complexity, molecular size, and accuracy of fit. Dynamics simulations based on these PESs have unveiled intriguing and novel reaction mechanisms, providing deep insights into the intricate dynamics involved in combustion, atmospheric, and organic chemistry.

Neural Network with Optimal Neuron Activation Functions Based on Additive Gaussian Process Regression

Feed-forward neural networks (NNs) are a staple machine learning method widely used in many areas of science and technology, including physical chemistry, computational chemistry, and materials informatics. While even a single-hidden-layer NN is a universal approximator, its expressive power is limited by the use of simple neuron activation functions (such as sigmoid functions) that are typically the same for all neurons. More flexible neuron activation functions would allow the use of fewer neurons and layers and thereby save computational cost and improve expressive power. We show that additive Gaussian process regression (GPR) can be used to construct optimal neuron activation functions that are individual to each neuron. An approach is also introduced that avoids nonlinear fitting of neural network parameters by defining them with rules. The resulting method combines the advantage of robustness of a linear regression with the higher expressive power of an NN. We demonstrate the approach by fitting the potential energy surfaces of the water molecule and formaldehyde. Without requiring any nonlinear optimization, the additive-GPR-based approach outperforms a conventional NN in the high-accuracy regime, where a conventional NN suffers more from overfitting.

Neural network potentials for chemistry: concepts, applications and prospects

Artificial Neural Networks (NN) are already heavily involved in methods and applications for frequent tasks in the field of computational chemistry such as representation of potential energy surfaces (PES) and spectroscopic predictions. This perspective provides an overview of the foundations of neural network-based full-dimensional potential energy surfaces, their architectures, underlying concepts, their representation and applications to chemical systems. Methods for data generation and training procedures for PES construction are discussed and means for error assessment and refinement through transfer learning are presented. A selection of recent results illustrates the latest improvements regarding accuracy of PES representations and system size limitations in dynamics simulations, but also NN application enabling direct prediction of physical results without dynamics simulations. The aim is to provide an overview for the current state-of-the-art NN approaches in computational chemistry and also to point out the current challenges in enhancing reliability and applicability of NN methods on a larger scale.

Novel methodology for systematically constructing global effective models from ab initio-based surfaces: A new insight into high-resolution molecular spectra analysis

In this paper, a novel methodology is presented for the construction of ab initio effective rotation-vibration spectroscopic models from potential energy and dipole moment surfaces. Non-empirical effective Hamiltonians are obtained via the block-diagonalization of selected variationally computed eigenvector matrices. For the first time, the derivation of an effective dipole moment is carried out in a systematic way. This general approach can be implemented quite easily in most of the variational computer codes and turns out to be a clear alternative to the rather involved Van Vleck perturbation method. Symmetry is exploited at all stages to translate first-principles calculations into a set of spectroscopic parameters to be further refined on experiment. We demonstrate on H₂CO, PH₃, CH₄, C₂H₄, and SF₆ that the proposed effective model can provide crucial information to spectroscopists within a very short time compared to empirical spectroscopic models. This approach brings a new insight into high-resolution spectrum analysis of polyatomic molecules and will be also of great help in the modeling of hot atmospheres where completeness is important.

Computational vibrational spectroscopy of molecule-surface interactions: what is still difficult and what can be done about it

Interactions of molecules with solid surfaces are responsible for key functionalities for a range of currently actively pursued technologies, including heterogeneous catalysis for synthesis or decomposition of molecules, sensitization, surface...

Interpolation Methods for Molecular Potential Energy Surface Construction

The concept of a potential energy surface (PES) is one of the most important concepts in modern chemistry. A PES represents the relationship between the chemical system's energy and its geometry (i.e., atom positions) and can provide useful information about the system's chemical properties and reactivity. Construction of accurate PESs with high-level theoretical methodologies, such as density functional theory, is still challenging due to a steep increase in the computational cost with the increase of the system size. Thus, over the past few decades, many different mathematical approaches have been applied to the problem of the cost-efficient PES construction. This article serves as a short overview of interpolative methods for the PES construction, including global polynomial interpolation, trigonometric interpolation, modified Shepard interpolation, interpolative moving least-squares, and the automated PES construction derived from these.

The Low-Lying Electronic States of NO2: Potential Energy and Dipole Surfaces, Bound States, and Electronic Absorption Spectrum

Nitrogen dioxide, NO₂, is a free radical composed of the two most abundant elements in Earth's atmosphere, nitrogen and oxygen, and is relevant to atmospheric and combustion chemistry. The electronic structure of even its lowest-lying states is remarkably complex, with various conical intersections and Renner-Teller pairings, giving rise to complex and perturbed vibronic states. Here we report some analysis of the 18 molecular states of doublet spin-multiplicity formed by combining ground-state N(⁴S_u) and O(³P_g) atoms. Three-dimensional potential energy surfaces were fit at the MRCI(Q)-F12/VTZ-F12 level, describing the lowest four (X̃, Ã, B̃, and C̃) electronic states. A properties-based diabatization procedure was applied to accommodate the intersections, producing energies in a quasidiabatic representation and yielding couplings that were also fit into surfaces. The low-lying vibrational levels on the ground X̃ state were computed and compared with experimental measurements. Compared to experiment, the lowest 125 calculated vibrational levels (up to 8500 cm^-1 above the zero-point energy) have a root-mean-squared error of 16.5 cm^-1. In addition, dipole moments for each of the lowest four electronic states-and the transition dipoles between them-were also computed and fit. With the coupled energy and dipole surfaces, the electronic spectrum was calculated in absolute intensity and compared with experimental measurements. Detailed structure in the experimental spectrum was successfully reproduced, and the total integrated intensity matches experiment to an accuracy of ∼1.5% with no empirical adjustments.

Exploring the anharmonic vibrational structure of carbon dioxide trimers

Our previously developed mbCO2 potential [O. Sode and J. N. Cherry, J. Comput. Chem. 38, 2763 (2017)] is used to describe the vibrational structure of the intermolecular motions of the CO₂ trimers: barrel-shaped and cyclic trimers. Anharmonic corrections are accounted for using the vibrational self-consistent field theory, vibrational second-order Møller-Plesset perturbation (VMP2) theory, and vibrational configuration interaction (VCI) methods and compared with experimental observations. For the cyclic structure, we revise the assignments of two previously observed experimental peaks based on our VCI and VMP2 results. We note that the experimental band observed near 13 cm^-1 is the out-of-phase out-of-plane degenerate motion with E″ symmetry, while the peak observed at 18 cm^-1 likely corresponds to the symmetric out-of-plane torsion A″ vibration. Since the VCI treatment of the vibrational motions accounts for vibrational mixing and delocalization, overtones and combination bands were also observed and quantified in the intermolecular regions of the two trimer isomers.

Enabling complete multichannel nonadiabatic dynamics: A global representation of the two-channel coupled, 1,21A and 13A states of NH3 using neural networks

Global coupled three-state two-channel potential energy and property/interaction (dipole and spin-orbit coupling) surfaces for the dissociation of NH₃(Ã) into NH + H₂ and NH₂ + H are reported. The permutational invariant polynomial-neural network approach is used to simultaneously fit and diabatize the electronic Hamiltonian by fitting the energies, energy gradients, and derivative couplings of the two coupled lowest-lying singlet states as well as fitting the energy and energy gradients of the lowest-lying triplet state. The key issue in fitting property matrix elements in the diabatic basis is that the diabatic surfaces must be smooth, that is, the diabatization must remove spikes in the original adiabatic property surfaces attributable to the switch of electronic wavefunctions at the conical intersection seam. Here, we employ the fit potential energy matrix to transform properties in the adiabatic representation to a quasi-diabatic representation and remove the discontinuity near the conical intersection seam. The property matrix elements can then be fit with smooth neural network functions. The coupled potential energy surfaces along with the dipole and spin-orbit coupling surfaces will enable more accurate and complete treatment of optical transitions, as well as nonadiabatic internal conversion and intersystem crossing.

Spectroscopy and Scattering Studies Using Interpolated Ab Initio Potentials

The Born-Oppenheimer potential energy surface (PES) has come a long way since its introduction in the 1920s, both conceptually and in predictive power for practical applications. Nevertheless, nearly 100 years later-despite astonishing advances in computational power-the state-of-the-art first-principles prediction of observables related to spectroscopy and scattering dynamics is surprisingly limited. For example, the water dimer, (H₂O)₂, with only six nuclei and 20 electrons, still presents a formidable challenge for full-dimensional variational calculations of bound states and is considered out of reach for rigorous scattering calculations. The extremely poor scaling of the most rigorous quantum methods is fundamental; however, recent progress in development of approximate methodologies has opened the door to fairly routine high-quality predictions, unthinkable 20 years ago. In this review, in relation to the workflow of spectroscopy and/or scattering studies, we summarize progress and challenges in the component areas of electronic structure calculations, PES fitting, and quantum dynamical calculations.

Neural Network Potential Energy Surfaces for Small Molecules and Reactions

We review progress in neural network (NN)-based methods for the construction of interatomic potentials from discrete samples (such as ab initio energies) for applications in classical and quantum dynamics including reaction dynamics and computational spectroscopy. The main focus is on methods for building molecular potential energy surfaces (PES) in internal coordinates that explicitly include all many-body contributions, even though some of the methods we review limit the degree of coupling, due either to a desire to limit computational cost or to limited data. Explicit and direct treatment of all many-body contributions is only practical for sufficiently small molecules, which are therefore our primary focus. This includes small molecules on surfaces. We consider direct, single NN PES fitting as well as more complex methods that impose structure (such as a multibody representation) on the PES function, either through the architecture of one NN or by using multiple NNs. We show how NNs are effective in building representations with low-dimensional functions including dimensionality reduction. We consider NN-based approaches to build PESs in the sums-of-product form important for quantum dynamics, ways to treat symmetry, and issues related to sampling data distributions and the relation between PES errors and errors in observables. We highlight combinations of NNs with other ideas such as permutationally invariant polynomials or sums of environment-dependent atomic contributions, which have recently emerged as powerful tools for building highly accurate PESs for relatively large molecular and reactive systems.

Introductory lecture: advances in ion spectroscopy: from astrophysics to biology

This introduction provides a historical context for the development of ion spectroscopy over the past half century by following the evolution of experimental methods to the present state-of-the-art. Rather than attempt a comprehensive review, we focus on how early work on small ions, carried out with fluorescence, direct absorption, and photoelectron spectroscopy, evolved into powerful technologies that can now address complex chemical problems ranging from catalysis to biophysics. One of these developments is the incorporation of cooling and temperature control to enable the general application of "messenger tagging" vibrational spectroscopy, first carried out using ionized supersonic jets and then with buffer gas cooling in radiofrequency ion traps. Some key advances in the application of time-resolved pump-probe techniques to follow ultrafast dynamics are also discussed, as are significant benchmarks in the refinement of ion mobility to allow spectroscopic investigation of large biopolymers with well-defined shapes. We close with a few remarks on challenges and opportunities to explore molecular level mechanics that drive macroscopic behavior.

Understanding the anharmonic vibrational structure of the carbon dioxide dimer

Understanding the vibrational structure of the CO₂ system is important to confirm the potential energy surface and interactions in such van der Waals complexes. In this work, we use our previously developed mbCO2 potential function to explore the vibrational structure of the CO₂ monomer and dimer. The potential function has been trained to reproduce the potential energies at the CCSD(T)-F12b/aug-cc-pVTZ level of electronic structure theory. The harmonic approximation, as well as anharmonic corrections using vibrational structure theories such as vibrational self-consistent field, vibrational second-order Møller-Plesset perturbation, and vibrational configuration interaction (VCI), is applied to address the vibrational motions. We compare the vibrational results using the mbCO2 potential function with traditional electronic structure theory results and to experimental frequencies. The anharmonic results for the monomer most closely match the experimental data to within 3 cm^-1, including the Fermi dyad frequencies. The intermolecular and intramolecular dimer frequencies were treated separately and show good agreement with the most recent theoretical and experimental results from the literature. The VCI treatment of the dimer vibrational motions accounts for vibrational mixing and delocalization, such that we observe the dimer Fermi resonance phenomena, both in the intramolecular and intermolecular regions.

Inverse Multiquadratic Functions as the Basis for the Rectangular Collocation Method to Solve the Vibrational Schrödinger Equation

We explore the use of inverse multiquadratic (IMQ) functions as basis functions when solving the vibrational Schrödinger equation with the rectangular collocation method. The quality of the vibrational spectrum of formaldehyde (in six dimensions) is compared to that obtained using Gaussian basis functions when using different numbers of width-optimized IMQ functions. The effects of the ratio of the number of collocation points to the number of basis functions and of the choice of the IMQ exponent are studied. We show that the IMQ basis can be used with parameters where the IMQ function is not integrable. We find that the quality of the spectrum with IMQ basis functions is somewhat lower that that with a Gaussian basis when the basis size is large, and for a range of IMQ exponents. The IMQ functions are; however, advantageous when a small number of functions is used or with a small number of collocation points (e.g., when using square collocation).

Highly excited vibrational levels of methane up to 10 300 cm-1: Comparative study of variational methods

In this work, we report calculated vibrational energy levels of the methane molecule up to 10 300 cm^-1. Two potential energy surfaces constructed in quite different coordinate systems with different analytical representations are employed in order to evaluate the uncertainty of vibrational predictions. To calculate methane energy levels, we used two independent techniques of the variational method. One method uses an exact kinetic energy operator in internal curvilinear coordinates. Another one uses an expansion of Eckart-Watson nuclear motion Hamiltonian in rectilinear normal coordinates. In the Icosad range (up to five vibrational quanta bands-below 7800 cm^-1), the RMS standard deviations between calculated and observed energy levels were 0.22 cm^-1 and 0.41 cm^-1 for these two quite different approaches. For experimentally well-known 3v₃ sub-levels, the calculation accuracy is estimated to be ∼1 cm^-1. In the Triacontad range (7660-9188 cm^-1), the average error of the calculation is about 0.5 cm^-1. The accuracy and convergence issues for higher energy ranges are discussed.

Understanding global infrared opacity and hot bands of greenhouse molecules with low vibrational modes from first-principles calculations: the case of CF4

Fluorine containing molecules have a particularly long atmospheric lifetime and their very big estimated global warming potentials are expected to rapidly increase in the future. This work is focused on the global theoretical prediction of infrared spectra of the tetrafluoromethane molecule that is considered as a potentially powerful greenhouse gas having the largest estimated lifetime of over 50 000 years in the atmosphere. The presence of relatively low vibrational frequencies makes the Boltzmann population of the excited levels important. Consequently, the "hot bands" corresponding to transitions among excited rovibrational states contribute significantly to the CF4 opacity in the infrared even at room temperature conditions but the existing laboratory data analyses are not sufficiently complete. In this work, we construct the first accurate and complete ab initio based line lists for CF4 in the range 0-4000 cm-1, containing rovibrational bands that are the most active in absorption. An efficient basis set compression method was applied to predict more than 700 new bands and subbands via variational nuclear motion calculations. We show that already at room temperature a quasi-continuum of overlapping weak lines appears in the CF4 infrared spectra due to the increasing density of bands and transitions. In order to converge the infrared opacity at room temperature, it was necessary to include a high rotational quantum number up to J = 80 resulting in 2 billion rovibrational transitions. In order to make the cross-section simulation faster, we have partitioned our data into two parts: (a) strong & medium line lists with lower energy levels for calculation of selective absorption features that can be used at various temperatures and (b) compressed "super-line" libraries of very weak transitions contributing to the quasi-continuum modelling. Comparisons with raw previously unassigned experimental spectra showed a very good accuracy for integrated absorbance in the entire range of the reported spectra predictions. The data obtained in this work will be made available through the TheoReTS information system (http://theorets.univ-reims.fr, http://theorets.tsu.ru) that contains ab initio born line lists and provides a user-friendly graphical interface for a fast simulation of the CF4 absorption cross-sections and radiance under various temperature conditions from 80 K to 400 K.

Structure-based sampling and self-correcting machine learning for accurate calculations of potential energy surfaces and vibrational levels

We present an efficient approach for generating highly accurate molecular potential energy surfaces (PESs) using self-correcting, kernel ridge regression (KRR) based machine learning (ML). We introduce structure-based sampling to automatically assign nuclear configurations from a pre-defined grid to the training and prediction sets, respectively. Accurate high-level ab initio energies are required only for the points in the training set, while the energies for the remaining points are provided by the ML model with negligible computational cost. The proposed sampling procedure is shown to be superior to random sampling and also eliminates the need for training several ML models. Self-correcting machine learning has been implemented such that each additional layer corrects errors from the previous layer. The performance of our approach is demonstrated in a case study on a published high-level ab initio PES of methyl chloride with 44 819 points. The ML model is trained on sets of different sizes and then used to predict the energies for tens of thousands of nuclear configurations within seconds. The resulting datasets are utilized in variational calculations of the vibrational energy levels of CH₃Cl. By using both structure-based sampling and self-correction, the size of the training set can be kept small (e.g., 10% of the points) without any significant loss of accuracy. In ab initio rovibrational spectroscopy, it is thus possible to reduce the number of computationally costly electronic structure calculations through structure-based sampling and self-correcting KRR-based machine learning by up to 90%.

Accuracy of Frequencies Obtained with the Aid of Explicitly Correlated Wave Function Based Methods

We asses the basis set convergence of harmonic frequencies using different explicitly correlated wave function based methods. All commonly available CCSD(T) variants as well as MP2-F12 and MP4(F12*) are considered, and a hierarchy of the different approaches is established. As for reaction and atomization energies, CCSD(F12*)(T*) is a close approximation to CCSD(F12)(T*) and clearly superior to the other tested approximations. The used scaling for the triples correction enhances the accuracy relative to CCSD(F12*)(T) especially for small basis sets and is very attractive since no additional computational costs are added. However, this scaling slightly breaks size consistency, and therefore we additionally study the accuracy of CCSD(F12*)(T*) and CCSD(F12*)(T) in the context of calculating anharmonic frequencies to check if this causes problems in the generation of the potential energy surface (PES). We find a fast basis set convergence for harmonic and anharmonic frequencies. Already in the cc-pVDZ-F12 basis, the RMSD to the CBS limit is only around 4-5 cm^-1.

Ab initio variational predictions for understanding highly congested spectra: rovibrational assignment of 108 new methane sub-bands in the icosad range (6280–7800 cm−1)

This work demonstrates for the first time how accurate first principles global calculations allow assigning complicated spectra of a molecule with more than 4 atoms.