Experiment-Driven Atomistic Materials Modeling: A Case Study Combining X-Ray Photoelectron Spectroscopy and Machine Learning Potentials to Infer the Structure of Oxygen-Rich Amorphous Carbon.

An important yet challenging aspect of atomistic materials modeling is reconciling experimental and computational results. Conventional approaches involve generating numerous configurations through molecular dynamics or Monte Carlo structure optimization and selecting the one with the closest match to experiment. However, this inefficient process is not guaranteed to succeed. We introduce a general method to combine atomistic machine learning (ML) with experimental observables that produces atomistic structures compatible with experiment by design. We use this approach in combination with grand-canonical Monte Carlo within a modified Hamiltonian formalism, to generate configurations that agree with experimental data and are chemically sound (low in energy). We apply our approach to understand the atomistic structure of oxygenated amorphous carbon (a-COx), an intriguing carbon-based material, to answer the question of how much oxygen can be added to carbon before it fully decomposes into CO and CO2. Utilizing an ML-based X-ray photoelectron spectroscopy (XPS) model trained from GW and density functional theory (DFT) data, in conjunction with an ML interatomic potential, we identify a-COx structures compliant with experimental XPS predictions that are also energetically favorable with respect to DFT. Employing a network analysis, we accurately deconvolve the XPS spectrum into motif contributions, both revealing the inaccuracies inherent to experimental XPS interpretation and granting us atomistic insight into the structure of a-COx. This method generalizes to multiple experimental observables and allows for the elucidation of the atomistic structure of materials directly from experimental data, thereby enabling experiment-driven materials modeling with a degree of realism previously out of reach.

Gaussian approximation potentials: Theory, software implementation and application examples.

Gaussian Approximation Potentials (GAPs) are a class of Machine Learned Interatomic Potentials routinely used to model materials and molecular systems on the atomic scale. The software implementation provides the means for both fitting models using ab initio data and using the resulting potentials in atomic simulations. Details of the GAP theory, algorithms and software are presented, together with detailed usage examples to help new and existing users. We review some recent developments to the GAP framework, including Message Passing Interface parallelisation of the fitting code enabling its use on thousands of central processing unit cores and compression of descriptors to eliminate the poor scaling with the number of different chemical elements.

Tensor-Reduced Atomic Density Representations.

Density-based representations of atomic environments that are invariant under Euclidean symmetries have become a widely used tool in the machine learning of interatomic potentials, broader data-driven atomistic modeling, and the visualization and analysis of material datasets. The standard mechanism used to incorporate chemical element information is to create separate densities for each element and form tensor products between them. This leads to a steep scaling in the size of the representation as the number of elements increases. Graph neural networks, which do not explicitly use density representations, escape this scaling by mapping the chemical element information into a fixed dimensional space in a learnable way. By exploiting symmetry, we recast this approach as tensor factorization of the standard neighbour-density-based descriptors and, using a new notation, identify connections to existing compression algorithms. In doing so, we form compact tensor-reduced representation of the local atomic environment whose size does not depend on the number of chemical elements, is systematically convergable, and therefore remains applicable to a wide range of data analysis and regression tasks.

A general-purpose machine learning Pt interatomic potential for an accurate description of bulk, surfaces, and nanoparticles.

A Gaussian approximation machine learning interatomic potential for platinum is presented. It has been trained on density-functional theory (DFT) data computed for bulk, surfaces, and nanostructured platinum, in particular nanoparticles. Across the range of tested properties, which include bulk elasticity, surface energetics, and nanoparticle stability, this potential shows excellent transferability and agreement with DFT, providing state-of-the-art accuracy at a low computational cost. We showcase the possibilities for modeling of Pt systems enabled by this potential with two examples: the pressure-temperature phase diagram of Pt calculated using nested sampling and a study of the spontaneous crystallization of a large Pt nanoparticle based on classical dynamics simulations over several nanoseconds.

Sleep quality is mediated by physical activity level in adolescents.

BACKGROUND: Sleep is essential for the adolescent's health and well-being. Despite existing evidence of the positive relationship between physical activity and quality of sleep, some other factors could mediate this association. The purpose of the present study was to clarify the interaction between the level of physical activity and sleep in adolescents depending on their gender. METHODS: A total of 12,459 subjects 11 to 19 years old (5073 male and 5016 female) reported data regarding their quality of sleep and their level of physical activity. RESULTS: Better quality of sleep was reported by males regardless of the level of physical activity (d=0.25, P<0.001). Better quality of sleep was reported by active subjects (P<0.05), and it improved in both sexes as the level of physical activity was higher (P<0.001). CONCLUSIONS: Male adolescents have better sleep quality than females regardless of their competitive level. The higher the adolescents' physical activity level the higher the quality of sleep.

Correction to "Structure and Pore Size Distribution in Nanoporous Carbon".

[This corrects the article DOI: 10.1021/acs.chemmater.1c03279.].

Accurate Computational Prediction of Core-Electron Binding Energies in Carbon-Based Materials: A Machine-Learning Model Combining Density-Functional Theory and GW.

We present a quantitatively accurate machine-learning (ML) model for the computational prediction of core-electron binding energies, from which X-ray photoelectron spectroscopy (XPS) spectra can be readily obtained. Our model combines density functional theory (DFT) with GW and uses kernel ridge regression for the ML predictions. We apply the new approach to disordered materials and small molecules containing carbon, hydrogen, and oxygen and obtain qualitative and quantitative agreement with experiment, resolving spectral features within 0.1 eV of reference experimental spectra. The method only requires the user to provide a structural model for the material under study to obtain an XPS prediction within seconds. Our new tool is freely available online through the XPS Prediction Server.

Reassignment of magic numbers for icosahedral Au clusters: 310, 564, 928 and 1426.

Icosahedral Au clusters with three and four shells of atoms are found to deviate significantly from the commonly assumed Mackay structures. By introducing additional atoms in the surface shell and creating a vacancy in the center of the cluster, the calculated energy per atom can be lowered significantly, according to several different descriptions of the interatomic interaction. Analogous icosahedral structures with five and six shells of atoms are generated using the same structural motifs and are similarly found to be more stable than Mackay icosahedra. The lowest energy per atom obtained here is for clusters containing 310, 564, 928 and 1426 atoms, as compared with the commonly assumed magic numbers of 309, 561, 923 and 1415. Some of the vertices in the optimized clusters have a hexagonal ring of atoms, rather than a pentagon, with the vertex atom missing. An inner shell atom in some cases moves outwards by more than an Ångström into the surface shell at such a vertex site. This feature, as well as the wide distribution of nearest-neighbor distances in the surface layer, can strongly influence the properties of icosahedral clusters, for example catalytic activity. The structural optimization is initially carried out using the GOUST method with atomic forces estimated with the EMT empirical potential function, but the atomic coordinates are then refined by minimization using electron density functional theory (DFT) or Gaussian approximation potential (GAP). A single energy barrier is found to separate the Mackay icosahedron from a lower energy structure where a string of atoms moves outwards in a concerted manner from the center so as to create a central vacancy while placing an additional atom in the surface shell.

Addressing Dynamics at Catalytic Heterogeneous Interfaces with DFT-MD: Anomalous Temperature Distributions from Commonly Used Thermostats.

Density functional theory-based molecular dynamics (DFT-MD) has been widely used for studying the chemistry of heterogeneous interfacial systems under operational conditions. We report frequently overlooked errors in thermostated or constant-temperature DFT-MD simulations applied to study (electro)catalytic chemistry. Our results demonstrate that commonly used thermostats such as Nosé-Hoover, Berendsen, and simple velocity-rescaling methods fail to provide a reliable temperature description for systems considered. Instead, nonconstant temperatures and large temperature gradients within the different parts of the system are observed. The errors are not a "feature" of any particular code but are present in several ab initio molecular dynamics implementations. This uneven temperature distribution, due to inadequate thermostatting, is well-known in the classical MD community, where it is ascribed to the failure in kinetic energy equipartition among different degrees of freedom in heterogeneous systems (Harvey et al. J. Comput. Chem. 1998, 726-740) and termed the flying ice cube effect. We provide tantamount evidence that interfacial systems are susceptible to substantial flying ice cube effects and demonstrate that the traditional Nosé-Hoover and Berendsen thermostats should be applied with care when simulating, for example, catalytic properties or structures of solvated interfaces and supported clusters. We conclude that the flying ice cube effect in these systems can be conveniently avoided using Langevin dynamics.

Exercise Addiction Stability and Health Effects. A 6-Month Follow-up Postcompetition Study in Amateur Endurance Cyclists.

OBJECTIVE: To study the longitudinal stability of exercise addiction and its health effects in apparently healthy amateur endurance cyclists from pre- to 6-month post-competition. METHODS: In total, 330 (30 women) adult cyclists were divided into 4 groups based on scores on the Exercise Addiction Inventory at both periods: nonrisk (n=262, 79.1%), transient (n=35, 10.6%), emerging (n=14, 4.2%) and persistent (n=20, 6.1%). RESULTS: The prevalence of high-risk exercise addiction was reduced postcompetition (16.7% vs 10.3%, P = 0.017). Of the cyclists with a high precompetition risk of exercise addiction, 63.6% (35/55) had a transient addiction associated with favorable effects on mental quality of life (effect size [ES]=0.52, 95% confidence interval: [0.20, 0.86]) and sleep quality (ES=-0.50 [-0.89, -0.12]) and avoided the worsening of depression symptom severity compared to the remaining groups (ES range=0.51-0.65). The 5.1% (14/275) of cyclists with a precompetition low risk of exercise addiction presented emerging exercise addiction that was associated with a worsened mental quality of life compared to the remaining groups (ES ranged 0.59-0.91), sleep quality compared to the nonrisk (ES=-0.56 [-0.02, -1.10]) and transient (ES=-0.72 [-1.36, -0.08]) groups and anxiety symptom severity compared to the persistent group (ES=0.51 [1.20,-0.19]). CONCLUSIONS: Exercise addiction had a marked transitory component at 6-month postcompetition with associated health benefits in amateur endurance cyclists.

X-ray Spectroscopy Fingerprints of Pristine and Functionalized Graphene.

In this work, we demonstrate how to identify and characterize the atomic structure of pristine and functionalized graphene materials from a combination of computational simulation of X-ray spectra, on the one hand, and computer-aided interpretation of experimental spectra, on the other. Despite the enormous scientific and industrial interest, the precise structure of these 2D materials remains under debate. As we show in this study, a wide range of model structures from pristine to heavily oxidized graphene can be studied and understood with the same approach. We move systematically from pristine to highly oxidized and defective computational models, and we compare the simulation results with experimental data. Comparison with experiments is valuable also the other way around; this method allows us to verify that the simulated models are close to the real samples, which in turn makes simulated structures amenable to several computational experiments. Our results provide ab initio semiquantitative information and a new platform for extended insight into the structure and chemical composition of graphene-based materials.

Connection between the physicochemical characteristics of amorphous carbon thin films and their electrochemical properties.

Connecting a material's surface chemistry with its electrocatalytic performance is one of the major questions in analytical electrochemistry. This is especially important in many sensor applications where analytes from complex media need to be measured. Unfortunately, today this connection is still largely missing except perhaps for the most simple ideal model systems. Here we present an approach that can be used to obtain insights about this missing connection and apply it to the case of carbon nanomaterials. In this paper we show that by combining advanced computational techniques augmented by machine learning methods with x-ray absorption spectroscopy (XAS) and electrochemical measurements, it is possible to obtain a deeper understanding of the correlation between local surface chemistry and electrochemical performance. As a test case we show how by computationally assessing the growth of amorphous carbon (a-C) thin films at the atomic level, we can create computational structural motifs that may in turn be used to deconvolute the XAS data from the real samples resulting in local chemical information. Then, by carrying out electrochemical measurements on the same samples from which x-ray spectra were measured and that were further characterized computationally, it is possible to gain insight into the interplay between the local surface chemistry and electrochemical performance. To demonstrate this methodology, we proceed as follows: after assessing the basic electrochemical properties of a-C films, we investigate the effect of short HNO3treatment on the sensitivity of these electrodes towards an inner sphere redox probe dopamine to gain knowledge about the influence of altered surface chemistry to observed electrochemical performance. These results pave the way towards a more general assessment of electrocatalysis in different systems and provide the first steps towards data driven tailoring of electrode surfaces to gain optimal performance in a given application.

A general-purpose machine-learning force field for bulk and nanostructured phosphorus.

Elemental phosphorus is attracting growing interest across fundamental and applied fields of research. However, atomistic simulations of phosphorus have remained an outstanding challenge. Here, we show that a universally applicable force field for phosphorus can be created by machine learning (ML) from a suitably chosen ensemble of quantum-mechanical results. Our model is fitted to density-functional theory plus many-body dispersion (DFT + MBD) data; its accuracy is demonstrated for the exfoliation of black and violet phosphorus (yielding monolayers of "phosphorene" and "hittorfene"); its transferability is shown for the transition between the molecular and network liquid phases. An application to a phosphorene nanoribbon on an experimentally relevant length scale exemplifies the power of accurate and flexible ML-driven force fields for next-generation materials modelling. The methodology promises new insights into phosphorus as well as other structurally complex, e.g., layered solids that are relevant in diverse areas of chemistry, physics, and materials science.

Machine Learning Interatomic Potentials as Emerging Tools for Materials Science.

Atomic-scale modeling and understanding of materials have made remarkable progress, but they are still fundamentally limited by the large computational cost of explicit electronic-structure methods such as density-functional theory. This Progress Report shows how machine learning (ML) is currently enabling a new degree of realism in materials modeling: by "learning" electronic-structure data, ML-based interatomic potentials give access to atomistic simulations that reach similar accuracy levels but are orders of magnitude faster. A brief introduction to the new tools is given, and then, applications to some select problems in materials science are highlighted: phase-change materials for memory devices; nanoparticle catalysts; and carbon-based electrodes for chemical sensing, supercapacitors, and batteries. It is hoped that the present work will inspire the development and wider use of ML-based interatomic potentials in diverse areas of materials research.

Reactivity of Amorphous Carbon Surfaces: Rationalizing the Role of Structural Motifs in Functionalization Using Machine Learning.

Systematic atomistic studies of surface reactivity for amorphous materials have not been possible in the past because of the complexity of these materials and the lack of the computer power necessary to draw representative statistics. With the emergence and popularization of machine learning (ML) approaches in materials science, systematic (and accurate) studies of the surface chemistry of disordered materials are now coming within reach. In this paper, we show how the reactivity of amorphous carbon (a-C) surfaces can be systematically quantified and understood by a combination of ML interatomic potentials, ML clustering techniques, and density functional theory calculations. This methodology allows us to process large amounts of atomic data to classify carbon atomic motifs on the basis of their geometry and quantify their reactivity toward hydrogen- and oxygen-containing functionalities. For instance, we identify subdivisions of sp and sp2 motifs with markedly different reactivities. We therefore draw a comprehensive, both qualitative and quantitative, picture of the surface chemistry of a-C and its reactivity toward -H, -O, -OH, and -COOH. While this paper focuses on a-C surfaces, the presented methodology opens up a new systematic and general way to study the surface chemistry of amorphous and disordered materials.

Growth Mechanism and Origin of High sp

We study the deposition of tetrahedral amorphous carbon (ta-C) films from molecular dynamics simulations based on a machine-learned interatomic potential trained from density-functional theory data. For the first time, the high sp^{3} fractions in excess of 85% observed experimentally are reproduced by means of computational simulation, and the deposition energy dependence of the film's characteristics is also accurately described. High confidence in the potential and direct access to the atomic interactions allow us to infer the microscopic growth mechanism in this material. While the widespread view is that ta-C grows by "subplantation," we show that the so-called "peening" model is actually the dominant mechanism responsible for the high sp^{3} content. We show that pressure waves lead to bond rearrangement away from the impact site of the incident ion, and high sp^{3} fractions arise from a delicate balance of transitions between three- and fourfold coordinated carbon atoms. These results open the door for a microscopic understanding of carbon nanostructure formation with an unprecedented level of predictive power.

Redox Potentials from Ab Initio Molecular Dynamics and Explicit Entropy Calculations: Application to Transition Metals in Aqueous Solution.

We present a complete methodology to consistently estimate redox potentials strictly from first-principles, without any experimental input. The methodology is based on (i) ab initio molecular dynamics (MD) simulations, (ii) all-atom explicit solvation, (iii) the two-phase thermodynamic (2PT) model, and (iv) the use of electrostatic potentials as references for the absolute electrochemical scale. We apply the approach presented to compute reduction potentials of the following redox couples: Cr2+/3+, V2+/3+, Ru(NH3)62+/3+, Sn2+/4+, Cu1+/2+, FcMeOH0/1+, and Fe2+/3+ (in aqueous solution) and Fc0/1+ (in acetonitrile). We argue that fully quantum-mechanical simulations are required to correctly model the intricate dynamical effects of the charged complexes on the surrounding solvent molecules within the solvation shell. Using the proposed methodology allows for a computationally efficient and statistically stable approach to compute free energy differences, yielding excellent agreement between our computed redox potentials and the experimental references. The root-mean-square deviation with respect to experiment for the aqueous test set and the two exchange-correlation density functionals used, PBE and PBE with van der Waals corrections, are 0.659 and 0.457 V, respectively. At this level of theory, depending on the functional employed, its ability to correctly describe each particular molecular complex seems to be the factor limiting the accuracy of the calculations.

Doping as a means to probe the potential dependence of dopamine adsorption on carbon-based surfaces: A first-principles study.

In this work, we study the adsorption characteristics of dopamine (DA), ascorbic acid (AA), and dopaminequinone (DAox) on carbonaceous electrodes. Our goal is to obtain a better understanding of the adsorption behavior of these analytes in order to promote the development of new carbon-based electrode materials for sensitive and selective detection of dopamine in vivo. Here we employ density functional theory-based simulations to reach a level of detail that cannot be achieved experimentally. To get a broader understanding of carbonaceous surfaces with different morphological characteristics, we compare three materials: graphene, diamond, and amorphous carbon (a-C). Effects of solvation on adsorption characteristics are taken into account via a continuum solvent model. Potential changes that take place during electrochemical measurements, such as cyclic voltammetry, can also alter the adsorption behavior. In this study, we have utilized doping as an indirect method to simulate these changes by shifting the work function of the electrode material. We demonstrate that sp2- and sp3-rich materials, as well as a-C, respond markedly different to doping. Also the adsorption behavior of the molecules studied here differs depending on the surface material and the change in the surface potential. In all cases, adsorption is spontaneous, but covalent bonding is not detected in vacuum. The aqueous medium has a large effect on the adsorption behavior of DAox, which reaches its highest adsorption energy on diamond when the potential is shifted to more negative values. In all cases, inclusion of the solvent enhances the charge transfer between the slab and DAox. Largest differences in adsorption energy between DA and AA are obtained on graphene. Gaining better understanding of the behavior of the different forms of carbon when used as electrode materials provides a means to rationalize the observed complex phenomena taking place at the electrodes during electrochemical oxidation/reduction of these biomolecules.

Spanish version of Pregnancy Symptoms Inventory: transcultural adaptation and reliability.

AIM: To transcultural adapt and analyze the reliability of Spanish version of Pregnancy Symptoms Inventory (PSI) and assess the prevalence of pregnancy symptoms in Spanish pregnant women. MATERIALS AND METHODS: A subsample of 120 healthy pregnant women answered the PSI twice and a sample of 280 report the prevalence and limitation of pregnancy symptoms. The reliability was examined by means of percent agreement and weighted Kappa coefficients. The prevalence of pregnancy symptoms was evaluated by the frequency of answers. RESULTS: Perfect and perfect-acceptable agreement was observed in 82% and 96% of the pregnant women, respectively. Weighted Kappa coefficients ranged from 0.589 to 0.889, indicating a good reliability. The most frequent symptoms perceived by Spanish pregnant women were urinary frequency, poor sleep, increased vaginal discharge and tiredness. CONCLUSION: Spanish Pregnancy Symptoms Inventory is a brief, conceptually equivalent and satisfactory reliable tool that allows an early assessment of the wide range of pregnancy symptoms in the health care practices.

Accurate schemes for calculation of thermodynamic properties of liquid mixtures from molecular dynamics simulations.

We explore different schemes for improved accuracy of entropy calculations in aqueous liquid mixtures from molecular dynamics (MD) simulations. We build upon the two-phase thermodynamic (2PT) model of Lin et al. [J. Chem. Phys. 119, 11792 (2003)] and explore new ways to obtain the partition between the gas-like and solid-like parts of the density of states, as well as the effect of the chosen ideal "combinatorial" entropy of mixing, both of which have a large impact on the results. We also propose a first-order correction to the issue of kinetic energy transfer between degrees of freedom (DoF). This problem arises when the effective temperatures of translational, rotational, and vibrational DoF are not equal, either due to poor equilibration or reduced system size/time sampling, which are typical problems for ab initio MD. The new scheme enables improved convergence of the results with respect to configurational sampling, by up to one order of magnitude, for short MD runs. To ensure a meaningful assessment, we perform MD simulations of liquid mixtures of water with several other molecules of varying sizes: methanol, acetonitrile, N, N-dimethylformamide, and n-butanol. Our analysis shows that results in excellent agreement with experiment can be obtained with little computational effort for some systems. However, the ability of the 2PT method to succeed in these calculations is strongly influenced by the choice of force field, the fluidicity (hard-sphere) formalism employed to obtain the solid/gas partition, and the assumed combinatorial entropy of mixing. We tested two popular force fields, GAFF and OPLS with SPC/E water. For the mixtures studied, the GAFF force field seems to perform as a slightly better "all-around" force field when compared to OPLS+SPC/E.