Vibrational spectroscopy for glycine adsorbed on silicon clusters: Harmonic and anharmonic calculations for models of the Si(100)-2×1 surface

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.

Using Collocation to Solve the Schrödinger Equation

We review the collocation approach to the solution of the Schrödinger equation and its uses in applications. Interrelations between collocation and other methods are highlighted. We also stress advantages and disadvantages of the rectangular collocation formulation. Using collocation makes it possible to use any, e.g. optimized, coordinates and basis functions, including nonintegrable basis functions, and provides a straightforward way of dealing with singularities in the potential. In addition, we stress that using collocation facilitates tuning the shape of basis functions and the placement of points, both of which can be done with machine-learning methods. Applications to electronic and vibrational problems are reviewed focusing on calculations for molecules on surfaces for which there are few variational calculations. Collocation has advantages when potential energy surfaces are unavailable, in particular, for molecule-surface systems, and for systems for which standard direct product quadrature grids, often used with variational methods, are costly.

Machine learning in computational chemistry: interplay between (non)linearity, basis sets, and dimensionality

Machine learning (ML) based methods and tools have now firmly established themselves in physical chemistry and in particular in theoretical and computational chemistry and in materials chemistry. The generality of popular ML techniques such as neural networks or kernel methods (Gaussian process and kernel ridge regression and their flavors) permitted their application to diverse problems from prediction of properties of functional materials (catalysts, solid state ionic conductors, etc.) from descriptors to the building of interatomic potentials (where ML is currently routinely used in applications) and electron density functionals. These ML techniques are assumed to have superior expressive power of nonlinear methods, and are often used "as is", with concepts such as "non-parametric" or "deep learning" used without a clear justification for their need or advantage over simpler and more robust alternatives. In this Perspective, we highlight some interrelations between popular ML techniques and traditional linear regressions and basis expansions and demonstrate that in certain regimes (such as a very high dimensionality) these approximations might collapse. We also discuss ways to recover the expressive power of a nonlinear approach and to help select hyperparameters with the help of high-dimensional model representation and to obtain elements of insight while preserving the generality of the method.

Tracing the Primordial Chemical Life of Glycine: A Review from Quantum Chemical Simulations

Glycine (Gly), NH2CH2COOH, is the simplest amino acid. Although it has not been directly detected in the interstellar gas-phase medium, it has been identified in comets and meteorites, and its synthesis in these environments has been simulated in terrestrial laboratory experiments. Likewise, condensation of Gly to form peptides in scenarios resembling those present in a primordial Earth has been demonstrated experimentally. Thus, Gly is a paradigmatic system for biomolecular building blocks to investigate how they can be synthesized in astrophysical environments, transported and delivered by fragments of asteroids (meteorites, once they land on Earth) and comets (interplanetary dust particles that land on Earth) to the primitive Earth, and there react to form biopolymers as a step towards the emergence of life. Quantum chemical investigations addressing these Gly-related events have been performed, providing fundamental atomic-scale information and quantitative energetic data. However, they are spread in the literature and difficult to harmonize in a consistent way due to different computational chemistry methodologies and model systems. This review aims to collect the work done so far to characterize, at a quantum mechanical level, the chemical life of Gly, i.e., from its synthesis in the interstellar medium up to its polymerization on Earth.

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...

Advances, challenges and perspectives of quantum chemical approaches in molecular spectroscopy of the condensed phase

The purpose of this review is to demonstrate advances, challenges and perspectives of quantum chemical approaches in molecular spectroscopy of the condensed phase. Molecular spectroscopy, particularly vibrational spectroscopy and electronic spectroscopy, has been used extensively for a wide range of areas of chemical sciences and materials science as well as nano- and biosciences because it provides valuable information about structure, functions, and reactions of molecules. In the meantime, quantum chemical approaches play crucial roles in the spectral analysis. They also yield important knowledge about molecular and electronic structures as well as electronic transitions. The combination of spectroscopic approaches and quantum chemical calculations is a powerful tool for science, in general. Thus, our article, which treats various spectroscopy and quantum chemical approaches, should have strong implications in the wider scientific community. This review covers a wide area of molecular spectroscopy from far-ultraviolet (FUV, 120-200 nm) to far-infrared (FIR, 400-10 cm-1)/terahertz and Raman spectroscopy. As quantum chemical approaches, we introduce several anharmonic approaches such as vibrational self-consistent field (VSCF) and the combination of periodic harmonic calculations with anharmonic corrections based on finite models, grid-based techniques like the Numerov approach, the Cartesian coordinate tensor transfer (CCT) method, Symmetry-Adapted Cluster Configuration-Interaction (SAC-CI), and the ZINDO (Semi-empirical calculations at Zerner's Intermediate Neglect of Differential Overlap). One can use anharmonic approaches and grid-based approaches for both infrared (IR) and near-infrared (NIR) spectroscopy, while CCT methods are employed for Raman, Raman optical activity (ROA), FIR/terahertz and low-frequency Raman spectroscopy. Therefore, this review overviews cross relations between molecular spectroscopy and quantum chemical approaches, and provides various kinds of close-reality advanced spectral simulation for condensed phases.

Current and future research directions in computer-aided near-infrared spectroscopy: A perspective

The present review aims to draw a perspective on the vibrational spectroscopy combined with the tools of computational chemistry. This includes an overview of the accomplishments made so far, the assessment of the present development trends and the prospects for continuing these advances. State-of-the-art methods, current challenges and the expected future advances are evaluated from the point-of-view of the practical application in vibrational spectroscopy. A special attention is given to near-infrared (NIR) spectroscopy, which occupies a distinct position among the techniques of vibrational spectroscopy. As the result of intrinsically complex spectra, reliance on the anharmonicity as well as keen interest given to complex materials, NIR spectroscopy may particularly benefit from computational chemistry. The present key limitations hindering development of NIR spectroscopy are identified; these constitute primarily the limit in the treatable system size and the inability to effectively include chemical matrix effects. Given the expanding role of NIR spectroscopy in science and industry, lifting these limitations would directly enhance the general potential of this technique.

Theoretical model of the interaction of glycine with hydrogenated amorphous carbon (HAC)

A theoretical model of hydrogenated amorphous carbon (HAC) is developed and applied to study the interaction of glycine with HAC surfaces at astronomical temperatures. Two models with different H content are tried for the HAC surface. The theory is applied at the Density Functional Theory (DFT) level, including a semiempirical dispersion correlation potential, d-DFT or Grimme DFT-D2. The level of theory is tested on glycine adsorption on a Si(001) surface. Crystalline glycine is also studied in its two stable phases, α and β, and the metastable γ phase. For the adsorption on Si or HAC surfaces, molecular glycine is introduced in the neutral and zwitterionic forms, and the most stable configurations are searched. All theoretical predictions are checked against experimental observations. HAC films are prepared by plasma enhanced vapor deposition at room temperature. Glycine is deposited at 20 K into a high vacuum, cold temperature chamber, to simulate astronomical conditions. Adsorption takes place through the acidic group COO(-) and when several glycine molecules are present, they form H-bond chains among them. Comparison between experiments and predictions suggests that a possible way to improve the theoretical model would require the introduction of aliphatic chains or a polycyclic aromatic core. The lack of previous models to study the interaction of amino-acids with HAC surfaces provides a motivation for this work.

Hybrid MP2/MP4 potential surfaces in VSCF calculations of IR spectra: applications for organic molecules

This study introduces an improved hybrid MP2/MP4 ab initio potential for vibrational spectroscopy calculations which is very accurate, yet without high computational demands. The method uses harmonic vibrational calculations with the MP4(SDQ) potential to construct an improved MP2 potential by coordinate scaling. This improved MP2 potential is used for the anharmonic VSCF calculation. The method was tested spectroscopically for four molecules: butane, acetone, ethylene and glycine. Very good agreement with experiment was found. For most of the systems, the more accurate harmonic treatment considerably improved the MP2 anharmonic results.

Towards Accurate Spectroscopic Identification of Species at Catalytic Surfaces: Anharmonic Vibrations of Formate on AuPt

We present a calculation of vibrational frequencies of formate on the AuPt(111) surface alloy including full anharmonicity and coupling of all six intramolecular degrees of freedom. This species is a key intermediate in methanol oxidation on this material. We use a modified version of the method of Manzhos and Carrington to compute the spectrum directly from a small number (<10,000) of DFT single-point energies, bypassing the construction of a potential energy surface. This is the first such calculation for a 4-atomic species at a surface. The spectrum is obtained using rectangular collocation and a small basis set of parameterized Hermite functions. The achievable accuracy of the order of several cm-1 corresponds to the typical experimental resolution. Using normal coordinates makes the equations simple and general and easily applicable to other systems. This calculation is doable on a PC. We predict that anharmonicity and coupling lower the fundamental frequencies by dozens of cm-1, which could affect species assignment.

Reliable structural, thermodynamic, and spectroscopic properties of organic molecules adsorbed on silicon surfaces from computational modeling: the case of glycine@Si(100)

Chemisorption of glycine on Si(100) has been studied by an integrated computational strategy based on perturbative anharmonic computations employing geometries and harmonic force fields evaluated by hybrid density functionals coupled to purposely tailored basis sets. It is shown that such a strategy allows the prediction of spectroscopic properties of isolated and chemisorbed molecules with comparable accuracy, paving the route toward a detailed analysis of surface-induced changes of glycine vibrational spectra.