Published Papers – Computational Chemical Science Center

Elucidating the water–anatase TiO2(101) interface structure using infrared signatures and molecular dynamics

Christopher R. O’Connor, Marcos F. Calegari Andrade, Annabella Selloni and Greg A. Kimmel, J. Chem. Phys. 159, 104707 (2023)

Abstract

The structure and dynamics of water on solid surfaces critically affect the chemistry of materials in ambient and aqueous environments. Here, we investigate the hydrogen bonding network of water adsorbed on the majority (101) surface of anatase TiO₂, a widely used photocatalyst, using polarization- and azimuth-resolved infrared spectroscopy combined with neural network potential molecular dynamics simulations. Our results show that one monolayer of water saturates the undercoordinated titanium (Ti_5c) sites, forming one-dimensional chains of molecule hydrogen bonded to surface undercoordinated bridging oxygen (O_2c) atoms. As the coverage increases, water adsorption on O_2c sites leads to significant restructuring of the water monolayer and the formation of a two-dimensional hydrogen bond network characterized by tightly bound pairs of water molecules on adjacent Ti_5c and O_2c sites. This structural motif likely persists at ambient conditions, influencing the reactions occurring there. The results reported here provide critical details of the structure of the water–anatase (101) interface that were previously hypothesized but unconfirmed experimentally.

URL: https://doi.org/10.1063/5.0161895

Modeling the aqueous interface of amorphous TiO2 using deep potential molecular dynamics

Zhutian Ding and Annabella Selloni, J. Phys. Chem, 159, 024706 (2023)

Abstract

Amorphous titanium dioxide (a-TiO₂) is widely used as a coating material in applications such as electrochemistry and self-cleaning surfaces where its interface with water has a central role. However, little is known about the structures of the a-TiO₂ surface and aqueous interface, particularly at the microscopic level. In this work, we construct a model of the a-TiO₂ surface via a cut-melt-and-quench procedure based on molecular dynamics simulations with deep neural network potentials (DPs) trained on density functional theory data. After interfacing the a-TiO₂ surface with water, we investigate the structure and dynamics of the resulting system using a combination of DP-based molecular dynamics (DPMD) and ab initio molecular dynamics (AIMD) simulations. Both AIMD and DPMD simulations reveal that the distribution of water on the a-TiO₂ surface lacks distinct layers normally found at the aqueous interface of crystalline TiO₂, leading to an ∼10 times faster diffusion of water at the interface. Bridging hydroxyls (Ti₂–ObH) resulting from water dissociation decay several times more slowly than terminal hydroxyls (Ti–OwH) due to fast Ti–OwH₂ → Ti–OwH proton exchange events. These results provide a basis for a detailed understanding of the properties of a-TiO₂ in electrochemical environments. Moreover, the procedure of generating the a-TiO₂-interface employed here is generally applicable to studying the aqueous interfaces of amorphous metal oxides.

URL: https://doi.org/10.1063/5.0157188

Acid–Base Chemistry of a Model IrO2 Catalytic Interface

Abhinav S. Raman and Annabella Selloni, J. Phys. Chem, Lett. 2023, 14, 35, 7787-7794 (2023)

Abstract

Iridium oxide (IrO₂) is one of the most efficient catalytic materials for the oxygen evolution reaction (OER), yet the atomic scale structure of its aqueous interface is largely unknown. Herein, the hydration structure, proton transfer mechanisms, and acid-base properties of the rutile IrO₂(110)-water interface are investigated using ab initio based deep neural-network potentials and enhanced sampling simulations. The proton affinities of the different surface sites are characterized by calculating their acid dissociation constants, which yield a point of zero charge in agreement with experiments. A large fraction (≈80%) of adsorbed water dissociation is observed, together with a short lifetime (≈0.5 ns) of the resulting terminal hydroxy groups, due to rapid proton exchanges between adsorbed H₂O and adjacent OH species. This rapid surface proton transfer supports the suggestion that the rate-determining step in the OER may not involve proton transfer across the double layer into solution, as indicated by recent experiments

URL: https://doi.org/10.1021/acs.jpclett.3c02001

Probing pH-Dependent Dehydration Dynamics of Mg and Ca Cations in Aqueous Solutions with Multi-Level Quantum Mechanics/Molecular Dynamics Simulations

J.-N. Boyn and E. A. Carter, J. Am. Chem. Soc., 145, 37, 20462 (2023)

Abstract

The dehydration of aqueous calcium and magnesium cations is the most fundamental process controlling their reactivity in chemical and biological phenomena, such as the formation of ionic solids or passing through ion channels. It holds particular relevance in light of recent advancements in the development of carbon capture techniques that rely on mineralization for long-term carbon storage. Specifically, dehydration of Ca2+ and Mg2+ is a key step in proposed carbon capture processes aiming to exploit the relatively high concentration of dissolved carbon dioxide in seawater via the formation of carbonate minerals from solvated Ca2+ and Mg2+ cations for sequestration and storage. Nevertheless, atomic-scale understanding of the dehydration of aqueous Ca2+ and Mg2+ cations remains limited. Here, we utilize rare event sampling via density functional theory molecular dynamics and embedded wavefunction theory calculations to elucidate the dehydration dynamics of aqueous Ca2+ and Mg2+. Emphasis is placed on the investigation of the effect pH has on the stability of the different coordination environments. Our results reveal significant differences in the dehydration dynamics of the two cations and provide insight into how they may be modulated by pH changes.

URL: https://doi.org/10.1021/jacs.3c06182

Significance of energy conservation in coupled-trajectory approaches to non-adiabatic dynamics

Evaristo Villaseco Arribas, Lea M. Ibele, David Lauvergnat, Neepa T. Maitra, Federica Agostini, accepted in Journal of Chemical Theory and Computation (2023)

Abstract

Through approximating electron-nuclear correlation terms in the exact factorization approach, trajectory-based methods have been derived and successfully applied to the dynamics of a variety of light-induced molecular processes, capturing quantum (de)coherence effects rigorously. These terms account for the coupling among the trajectories, recovering the non-local nature of quantum nuclear dynamics which is completely overlooked in traditional independent-trajectory algorithms. Nevertheless, some of the approximations introduced in the derivation of some of these methods do not conserve the total energy. We analyze energy conservation in the coupled- trajectory mixed quantum-classical (CTMQC) algorithm and explore the performance of a modified algorithm, CTMQC-E where some of the terms are redefined to restore energy conservation. A set of molecular models is used as test, namely 2-cis-penta-2,4-dienimium cation, bis(methylene) adamantyl radical cation, butatriene cation, uracil radical cation, and neutral pyrazine.

URL: https://chemrxiv.org/engage/chemrxiv/article-details/64ccb98f4a3f7d0c0d91f42a

Different Flavors of Exact-Factorization-Based Mixed Quantum-Classical Methods for Multistate Dynamics

Evaristo Villaseco Arribas, Patricia Vindel-Zandbergen, Saswata Roy, Neepa T. Maitra, accepted in Physical Chemistry Chemical Physics as “accepted manuscript” (2023)

Abstract

The exact factorization approach has led to the development of new mixed quantum-classical methods for simulating coupled electron-ion dynamics. We compare their performance for dynamics when more than two electronic states are occupied at a given time and analyze: (1) the use of coupled versus auxiliary trajectories in evaluating the electron-nuclear correlation terms, (2) the approximation of using these terms within surface-hopping and Ehrenfest frameworks, and (3) the relevance of the exact conditions of zero population transfer away from nonadiabatic coupling regions and total energy conservation. Dynamics through the three-state conical intersection in the uracil radical cation as well as polaritonic models in one dimension are studied.

URL: https://doi.org/10.1039/D3CP03464J

Molecular Rotations, Multiscale Order, Hyperuniformity, and Signatures of Metastability during the Compression/Decompression Cycles of Amorphous Ices

Maud Formanek, Salvatore Torquato, Roberto Car, and Fausto Martelli, The Journal of Physical Chemistry B 2023, 127 (17), 3946-3957 (2023)

Abstract

We model, via large-scale molecular dynamics simulations, the isothermal compression of low-density amorphous ice (LDA) to generate high-density amorphous ice (HDA) and the corresponding decompression extending to negative pressures to recover the low-density amorphous phase (LDA_HDA). Both LDA and HDA are nearly hyperuniform and are characterized by a dynamical HBN, showing that amorphous ices are nonstatic materials and implying that nearly hyperuniformity can be accommodated in dynamical networks. In correspondence with both the LDA-to-HDA and the HDA-to-LDA_HDA phase transitions, the (partial) activation of rotational degrees of freedom activates a cascade effect that induces a drastic change in the connectivity and a pervasive reorganization of the HBN topology which, ultimately, break the samples’ hyperuniform character. Key to this effect is the rapid rate at which changes occur, and not their magnitude. The inspection of structural properties from the short- to the long-range shows that signatures of metastability are present at all length-scales, hence providing further solid evidence in support of the liquid–liquid critical point scenario. LDA and LDA_HDA differ in terms of HBN and structural properties, implying that they are distinct low-density glasses. Our work unveils the role of molecular rotations in the phase transitions between amorphous ices and shows how the unfreezing of rotational degrees of freedom generates a cascade effect that propagates over multiple length-scales. Our findings greatly improve our basic understanding of water and amorphous ices and can potentially impact the field of molecular network-forming materials at large.

URL: https://doi.org/10.1021/acs.jpcb.3c00611

DeePMD-kit v2: A software package for Deep Potential models

Jinzhe Zeng, Duo Zhang, Denghui Lu, Pinghui Mo, Zeyu Li, Yixiao Chen, Marián Rynik, Li’ang Huang, Ziyao Li, Shaochen Shi, Yingze Wang, Haotian Ye, Ping Tuo, Jiabin Yang, Ye Ding, Yifan Li, Davide Tisi, Qiyu Zeng, Han Bao, Yu Xia, Jiameng Huang, Koki Muraoka, Yibo Wang, Junhan Chang, Fengbo Yuan, Sigbjørn Løland Bore, Chun Cai, Yinnian Lin, Bo Wang, Jiayan Xu, Jia-Xin Zhu, Chenxing Luo, Yuzhi Zhang, Rhys E. A. Goodall, Wenshuo Liang, Anurag Kumar Singh, Sikai Yao, Jingchao Zhang, Renata Wentzcovitch, Jiequn Han, Jie Liu, Weile Jia, Darrin M. York, Weinan E, Roberto Car, Linfeng Zhang, Han Wang, the Journal of Chemical Physics (Vol.159, Issue 5) (2023)

Abstract

DeePMD-kit is a powerful open-source software package that facilitates molecular dynamics simulations using machine learning potentials (MLP) known as Deep Potential (DP) models. This package, which was released in 2017, has been widely used in the fields of physics, chemistry, biology, and material science for studying atomistic systems. The current version of DeePMD-kit offers numerous advanced features such as DeepPot-SE, attention-based and hybrid descriptors, the ability to fit tensile properties, type embedding, model deviation, Deep Potential – Range Correction (DPRc), Deep Potential Long Range (DPLR), GPU support for customized operators, model compression, non-von Neumann molecular dynamics (NVNMD), and improved usability, including documentation, compiled binary packages, graphical user interfaces (GUI), and application programming interfaces (API). This article presents an overview of the current major version of the DeePMD-kit package, highlighting its features and technical details. Additionally, the article benchmarks the accuracy and efficiency of different models and discusses ongoing developments.

URL: https://doi.org/10.1063/5.0155600

Thermal Conductivity of Water at Extreme Conditions

Cunzhi Zhang, Marcello Puligheddu, Linfeng Zhang, Roberto Car and Giulia Galli, J. Phys. Chem. B, 127, 31, 7011-7017 (2023)

Abstract

Measuring the thermal conductivity (κ) of water at extreme conditions is a challenging task and few experimental data are available. We predict κ for temperatures and pressures relevant to the conditions of the Earth mantle, between 1,000 and 2,000 K and up to 22 GPa. We employ close to equilibrium molecular dynamics simulations and a deep neural network potential fitted to density functional theory data. We then interpret our results by computing the equation of state of water on a fine grid of points and using a simple model for κ. We find that the thermal conductivity is weakly dependent on temperature and monotonically increases with pressure with an approximate square-root behavior. In addition we show how the increase of κ at high pressure, relative to ambient conditions, is related to the corresponding increase in the sound velocity. Although the relationships between the thermal conductivity, pressure and sound velocity established here are not rigorous, they are sufficiently accurate to allow for a robust estimate of the thermal conductivity of water in a broad range of temperatures and pressures, where experiments are still difficult to perform.

URL: https://doi.org/10.1021/acs.jpcb.3c02972

Why dissolving salt in water decreases its dielectric permittivity

Chunyi Zhang, Shuwen Yue, Athanassios Z. Panagiotopoulos, Michael L. Klein and Xifan Wu, Phys. Rev. Lett. 131, 076801 (2023)

Abstract

The dielectric permittivity of salt water decreases on dissolving more salt. For nearly a century, this phenomenon has been explained by invoking saturation in the dielectric response of the solvent water molecules. Herein, we employ an advanced deep neural network (DNN), built using data from density functional theory, to study the dielectric permittivity of sodium chloride solutions. Notably, the decrease in the dielectric permittivity as a function of concentration, computed using the DNN approach, agrees well with experiments. Detailed analysis of the computations reveals that the dominant effect, caused by the intrusion of ionic hydration shells into the solvent hydrogen-bond network, is the disruption of dipolar correlations among water molecules. Accordingly, the observed decrease in the dielectric permittivity is mostly due to increasing suppression of the collective response of solvent waters.

URL: https://doi.org/10.1103/PhysRevLett.131.076801

A first-principles machine-learning force field for heterogeneous ice nucleation on microcline feldspar

Pablo M. Piaggi, Annabella Selloni, Athanassios Z. Panagiotopoulos, Roberto Car and Pablo G. Debenedetti, Faraday Discussions (2023)

Abstract

The formation of ice in the atmosphere affects precipitation and cloud properties, and plays a key role in the climate of our planet. Although ice can form directly from liquid water at deeply supercooled conditions, the presence of foreign particles can aid ice formation at much warmer temperatures. Over the past decade, experiments have highlighted the remarkable efficiency of feldspar minerals as ice nuclei compared to other particles present in the atmosphere. However, the exact mechanism of ice formation on feldspar surfaces has yet to be fully understood. Here, we develop a first-principles machine-learning model for the potential energy surface aimed at studying ice nucleation at microcline feldspar surfaces. The model is able to reproduce with high-fidelity the energies and forces derived from density-functional theory (DFT) based on the SCAN exchange and correlation functional. Our training set includes configurations of bulk supercooled water, hexagonal and cubic ice, microcline, and fully-hydroxylated feldspar surfaces exposed to vacuum, liquid water, and ice. We apply the machine-learning force field to study different fully-hydroxylated terminations of the (100), (010), and (001) surfaces of microcline exposed to vacuum. Our calculations suggest that terminations that do not minimize the number of broken bonds are preferred in vacuum. We also study the structure of supercooled liquid water in contact with microcline surfaces, and find that water density correlations extend up to around 10 Å from the surfaces. Finally, we show that the force field maintains a high accuracy during the simulation of ice formation at microcline surfaces, even for large systems of around 30,000 atoms. Future work will be directed towards the calculation of nucleation free energy barriers and rates using the force field developed herein, and understanding the role of different microcline surfaces on ice nucleation.

URL: https://doi.org/10.1039/D3FD00100H

Hybrid Auxiliary Field Quantum Monte Carlo for Molecular Systems

Yixiao Chen, Linfeng Zhang, Weinan E. and Roberto Car, J. Chem. Theory Comput. 2 19, 14, 4484-4493 (2023)

Abstract

We propose a quantum Monte Carlo approach to solve the many-body Schrödinger equation for the electronic ground state. The method combines optimization from variational Monte Carlo and propagation from auxiliary field quantum Monte Carlo in a way that significantly alleviates the sign problem. In application to molecular systems, we obtain highly accurate results for configurations dominated by either dynamic or static electronic correlation.

URL: https://doi.org/10.1021/acs.jctc.3c00038

DeePKS+ ABACUS as a Bridge between Expensive Quantum Mechanical Models and Machine Learning Potentials

W. Li, Q. Ou, Y. Chen, Y. Cao, R. Liu, C. Zhang, D. Zheng, C. Cai, X. Wu, H. Wang, M. Chen, L. Zhang, The Journal of Physical Chemistry A, 126, 9154-9164 (2022)

Abstract

Recently, the development of machine learning (ML) potentials has made it possible to perform large-scale and long-time molecular simulations with the accuracy of quantum mechanical (QM) models. However, for high-level QM methods, such as density functional theory (DFT) at the meta-GGA level and/or with exact exchange, quantum Monte Carlo, etc., generating a sufficient amount of data for training a ML potential has remained computationally challenging due to their high cost. In this work, we demonstrate that this issue can be largely alleviated with Deep Kohn-Sham (DeePKS), a ML-based DFT model. DeePKS employs a computationally efficient neural network-based functional model to construct a correction term added upon a cheap DFT model. Upon training, DeePKS offers closely-matched energies and forces compared with high-level QM method, but the number of training data required is orders of magnitude less than that required for training a reliable ML potential. As such, DeePKS can serve as a bridge between expensive QM models and ML potentials: one can generate a decent amount of high-accuracy QM data to train a DeePKS model, and then use the DeePKS model to label a much larger amount of configurations to train a ML potential. This scheme for periodic systems is implemented in a DFT package ABACUS, which is open-source and ready for use in various applications.

URL: https://doi.org/10.1021/acs.jpca.2c05000

Characterizing Structure-dependent TiS2/Water Interfaces Using Deep-Neural-Network-Assisted Molecular Dynamics

L. Li, M.F. Calegari Andrade, R. Car, A. Selloni, E.A. Carter, J Phys Chem C 2023, 127, 9750-9758 (2023)

Abstract

As a promising layered electrode material, TiS₂-based capacitive deionization (CDI) devices for water desalination have attracted significant attention. However, TiS₂/H₂O interfacial features, potentially important for device optimization, remain unidentified. Using Deep Potential Molecular Dynamics (DPMD), we characterized distinct aqueous interfaces introduced by four TiS₂ terminations expected to be present as water intercalates into TiS₂, namely, Armchair, Zigzag, Zigzag-L, and Zigzag-R. First, we assessed important representative physical properties of the system to validate the deep potentials (DPs). DPMD simulations agree well with experiments and first-principles simulations, suggesting the DPs are accurate and reliable. Subsequent simulations of these TiS₂/water interfaces revealed how TiS₂ surface termination influences the structure of interfacial water. This effect is most evident in the first and second water layers close to the TiS₂ surface, and more pronounced when spontaneous dissociative adsorption of water occurs. The extent of water dissociation on each surface was evaluated using enhanced sampling. Zigzag-L is the only interface where proton transfer from adsorbed water to TiS₂ surface S atoms is thermodynamically and kinetically favored. The coexistence of surface four-fold-coordinated Ti (Ti_4c) and one-fold-coordinated S (S_1c) is found to be essential to making proton transfer feasible on the Zigzag-L surface. Furthermore, remaining unprotonated S_1c atoms can act as good proton acceptors after water dissociation. Thus, TiS₂ with Zigzag-L termination may be a surface to avoid in CDI device construction, given that pH fluctuations adversely affect performance. This work provides new understanding of TiS₂/H₂O interfacial features that could aid future design and optimization of TiS₂-based CDI devices for water desalination.

URL: https://doi.org/10.1021/acs.jpcc.2c08581

Dynamics of Aqueous Electrolyte Solutions: Challenges for Simulations

A. Z. Panagiotopoulos and S. Yue, J. Phys. Chem. B 127: 430-37 (2023)

Abstract

This Perspective article focuses on recent simulation work on the dynamics of aqueous electrolytes. It is well-established that full-charge, nonpolarizable models for water and ions generally predict solution dynamics that are too slow in comparison to experiments. Models with reduced (scaled) charges do better for solution diffusivities and viscosities but encounter issues describing other dynamic phenomena such as nucleation rates of crystals from solution. Polarizable models show promise, especially when appropriately parametrized, but may still miss important physical effects such as charge transfer. First-principles calculations are starting to emerge for these properties that are in principle able to capture polarization, charge transfer, and chemical transformations in solution. While direct ab initio simulations are still too slow for simulations of large systems over long time scales, machine-learning models trained on appropriate first-principles data show significant promise for accurate and transferable modeling of electrolyte solution dynamics.

URL: https://doi.org/10.1021/acs.jpcb.2c07477

Water Dissociation at the Water-Rutile TiO2(110) Interface from Ab-initio based Deep Neural Network Simulations

B. Wen, M. F. Calegari Andrade, L.M. Liu, A. Selloni, Proc. Nat. Acad. Sci., 120, e2212250120 (2023)

Abstract

The interaction of water with TiO₂ surfaces is of crucial importance in various scientific fields and applications, from photocatalysis for hydrogen production and the photooxidation of organic pollutants to self-cleaning surfaces and bio-medical devices. In particular, the equilibrium fraction of water dissociation at the TiO₂–water interface has a critical role in the surface chemistry of TiO₂, but is difficult to determine both experimentally and computationally. Among TiO₂ surfaces, rutile TiO₂(110) is of special interest as the most abundant surface of TiO₂’s stable rutile phase. While surface-science studies have provided detailed information on the interaction of rutile TiO₂(110) with gas-phase water, much less is known about the TiO₂(110)–water interface, which is more relevant to many applications. In this work, we characterize the structure of the aqueous TiO₂(110) interface using nanosecond timescale molecular dynamics simulations with ab initio-based deep neural network potentials that accurately describe water/TiO₂(110) interactions over a wide range of water coverages. Simulations on TiO₂(110) slab models of increasing thickness provide insight into the dynamic equilibrium between molecular and dissociated adsorbed water at the interface and allow us to obtain a reliable estimate of the equilibrium fraction of water dissociation. We find a dissociation fraction of 22 ± 6% with an associated average hydroxyl lifetime of 7.6 ± 1.8 ns. These quantities are both much larger than corresponding estimates for the aqueous anatase TiO₂(101) interface, consistent with the higher water photooxidation activity that is observed for rutile relative to anatase.

URL: https://doi.org/10.1073/pnas.2212250120

Energy-Conserving Coupled-Trajectory Mixed-Quantum Classical Dynamics

Villaseco Arribas and N. T. Maitra, J. Chem. Phys. 158, 161105 (2023)

Abstract

The coupled-trajectory mixed quantum–classical method (CTMQC), derived from the exact factorization approach, has successfully predicted photo-chemical dynamics in a number of interesting molecules, capturing population transfer and decoherence from first principles. However, due to the approximations made, CTMQC does not guarantee energy conservation. We propose a modified algorithm, CTMQC-E, which redefines the integrated force in the coupled-trajectory term so to restore energy conservation, and demonstrate its accuracy on scattering in Tully’s extended coupling region model and photoisomerization in a retinal chromophore model.

URL: https://doi.org/10.1063/5.0149116

Characterization of Hole States at the Zn-doped Hematite/Water Interface from Ab Initio Simulations

Zachary K. Goldsmith, Zhutian Ding, and Annabella Selloni, ACS Catalysis 2023, 13, 8, 5298-5306 (2023)

Abstract

Hole states at the surface of hematite (α-Fe₂O₃) are highly influential in the material’s performance as a photoanode for the oxygen evolution reaction. Zn-doping of hematite is known to both lower the overpotential for oxygen evolution and introduce hole carriers near the surface. In this work, hole states at the aqueous interface of hematite (0001) were characterized using density functional theory-based ab initio molecular dynamics (AIMD) together with hybrid density functional theory (DFT) calculations of the electronic structure. PBE0 with 12% exact exchange calculations of Zn-doped hematite (0001) slabs in vacuum revealed a hole state within the band gap of hematite, which was spatially localized on a Fe–O moiety in an adjacent layer of the slab. AIMD of the (0001) slab in contact with water was propagated at the PBE+D3 and PBE+U+D3 levels of theory, with hybrid PBE0 calculations performed on snapshots every 200 fs. Under both protocols we observed the fluctuation of the hole state energy within the band gap and the localization of the hole at the aqueous interface. Zn doping had an overall marginal effect on the interfacial hydration structure and hydrogen bonding dynamics. These calculations showed that Zn doping introduces surface-local hole states in the band gap at energies close to the O₂/H₂O redox level, providing atomistic insights into the lower overpotential observed for Zn-doped hematite and more broadly the potential role of surface-local hole states in driving water oxidation.

URL: https://pubs.acs.org/doi/10.1021/acscatal.3c00357

Melting curves of ice polymorphs in the vicinity of the liquid-liquid critical point

Pablo M. Piaggi, Thomas E. Gartner III, Roberto Car, and Pablo G. Debenedetti, J. Chem. Phys. 159, 054502 (2023)

Abstract

The possible existence of a liquid-liquid critical point in deeply supercooled water has been a subject of debate in part due to the challenges associated with providing definitive experimental evidence. Pioneering work by Mishima and Stanley [Nature 392, 164 (1998) and Phys. Rev. Lett. 85, 334 (2000)] sought to shed light on this problem by studying the melting curves of different ice polymorphs and their metastable continuation in the vicinity of the expected location of the liquid-liquid transition and its associated critical point. Based on the continuous or discontinuous changes in slope of the melting curves, Mishima suggested that the liquid-liquid critical point lies between the melting curves of ice III and ice V. Here, we explore this conjecture using molecular dynamics simulations with a purely-predictive machine learning model based on ab initio quantum-mechanical calculations. We study the melting curves o ices III, IV, V, VI, and XIII using this model and find that the melting lines of all the studied ice polymorphs are supercritical and do not intersect the liquid-liquid transition locus. We also find a pronounced, yet continuous, change in slope of the melting lines upon crossing of the locus of maximum compressibility of the liquid. Finally, we analyze critically the literature in light of our findings and conclude that the scenario in which melting curves are supercritical is favored by the most recent computational and experimental evidence. Thus, although the preponderance of experimental and computational evidence is consistent with the existence of a second critical point in water, the behavior of the melting lines of ice polymorphs does not provide strong evidence in support of this viewpoint, according to our calculations.

URL: https://doi.org/10.1063/5.0159288

First-principles-based Machine Learning Models for Phase Behavior and Transport Properties of CO2

Reha Mathur, Maria Carolina Muniz, Shuwen Yue, Roberto Car and Athanassios Z. Panagiotopoulos, J. Phys. Chem. B, 2023, 127, 20, 4562–4569 (2023)

Abstract

In this work, we construct distinct first-principles-based machine-learning models of CO₂, reproducing the potential energy surface of the PBE-D3, BLYP-D3, SCAN and SCAN-rvv10 approximations of density functional theory. We employ the Deep Potential methodology to develop the models and consequently achieve a significant computational efficiency over ab initio molecular dynamics (AIMD) that allows for larger system sizes and time scales to be explored. Although our models are trained only with liquid phase configurations, they are able to simulate a stable interfacial system and predict vapor-liquid equilibrium properties, in good agreement with results from the literature. Because of the computational efficiency of the models, we are also able to obtain transport properties, such as viscosity and diffusion coefficients. We find that the SCAN-based model underpredicts experimental liquid densities, while the SCAN-rvv10-based model shows improvement but still exhibits a temperature shift, which remains approximately constant for all properties investigated in this work. We find that the BLYP-D3-based model generally performs better for liquid phase and vapor-liquid equilibrium properties, but the PBE-D3-based model is better suited for predicting transport properties.

URL: https://doi.org/10.1021/acs.jpcb.3c00610

A Deep Potential model for liquid-vapor equilibrium and cavitation rates of water

Ignacio Sanchez-Burgos, Maria Carolina Muniz, Jorge R. Espinosa, and Athanassios Z. Panagiotopoulos, J. Chem. Phys. 158, 184504 (2023)

Abstract

Computational studies of liquid water and its phase transition into vapor have traditionally been performed using classical water models. Here we utilize the Deep Potential methodology — a machine learning approach — to study this ubiquitous phase transition, starting from the phase diagram in the liquid-vapor coexistence regime. The machine learning model is trained on ab initio energies and forces based on the SCAN density functional which has been previously shown to reproduce solid phases and other properties of water. Here, we compute the surface tension, saturation pressure and enthalpy of vaporization for a range of temperatures spanning from 300 to 600 K, and evaluate the Deep Potential model performance against experimental results and the semi-empirical TIP4P/2005 classical model. Moreover, by employing the seeding technique, we evaluate the free energy barrier and nucleation rate at negative pressures for the isotherm of 296.4 K. We find that the nucleation rates obtained from the Deep Potential model deviate from those computed for the TIP4P/2005 water model, due to an underestimation in the surface tension from the Deep Potential model. From analysis of the seeding simulations, we also evaluate the Tolman length for the Deep Potential water model, which is (0.091 ± 0.008) nm at 296.4 K. Lastly, we identify that water molecules display a preferential orientation in the liquid-vapor interface, in which H atoms tend to point towards the vapor phase to maximize the enthalpic gain of interfacial molecules. We find that this behaviour is more pronounced for planar interfaces than for the curved interfaces in bubbles. This work represents the first application of Deep Potential models to the study of liquid-vapor coexistence and water cavitation.

URL: https://doi.org/10.1063/5.0144500

Liquid-liquid transition in water from first principles

Thomas E. Gartner III, Pablo M. Piaggi, Roberto Car, Athanassios Z. Panagiotopoulos and Pablo G. Debenedetti, Phys. Rev. Lett. 129, 255702 (2022)

Abstract

A long-standing question in water research is the possibility that supercooled liquid water can undergo a liquid-liquid phase transition (LLT) into high- and low-density liquids. We used several complementary molecular simulation techniques to evaluate the possibility of an LLT in an ab initio neural network model of water trained on density functional theory calculations with the SCAN exchange correlation functional. We conclusively show the existence of a first-order LLT and an associated critical point in the SCAN description of water, representing the first definitive computational evidence for an LLT in water from first principles potentials of binary oxides and chlorides, indicating a fundamental challenge for future seawater electrode materials design.

URL: https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.129.255702

Modeling the Electronic Absorption Spectra of the Indocarbocyanine Cy3

Mohammed Sorour, Andrew Marcus and Spiridoula Matsika, Molecules 2022, 27(13), 4062 (2022)

Abstract

Accurate modeling of optical spectra requires careful treatment of the molecular structures and vibronic, environmental, and thermal contributions. The accuracy of the computational methods used to simulate absorption spectra is limited by their ability to account for all the factors that affect the spectral shapes and energetics. The ensemble-based approaches are widely used to model the absorption spectra of molecules in the condensed-phase, and their performance is system dependent. The Franck–Condon approach is suitable for simulating high resolution spectra of rigid systems, and its accuracy is limited mainly by the harmonic approximation. In this work, the absorption spectrum of the widely used cyanine Cy3 is simulated using the ensemble approach via classical and quantum sampling, as well as, the Franck–Condon approach. The factors limiting the ensemble approaches, including the sampling and force field effects, are tested, while the vertical and adiabatic harmonic approximations of the Franck–Condon approach are also systematically examined. Our results show that all the vertical methods, including the ensemble approach, are not suitable to model the absorption spectrum of Cy3, and recommend the adiabatic methods as suitable approaches for the modeling of spectra with strong vibronic contributions. We find that the thermal effects, the low frequency modes, and the simultaneous vibrational excitations have prominent contributions to the Cy3 spectrum. The inclusion of the solvent stabilizes the energetics significantly, while its negligible effect on the spectral shapes aligns well with the experimental observations.

URL: https://doi.org/10.3390/molecules27134062

Oxygen-Chlorine Chemisorption Scaling for Seawater Electrolysis on Transition Metals: The Role of Redox

Robert B. Wexler and Emily A. Carter, Advanced Theory and Simulations/Early Review/Research Article, 2200592 (2022)

Abstract

To clarify what controls species oxidation selectivity in seawater electrolysis, density functional theory (DFT) is used to identify chemisorption enthalpy trends and scaling relations for the simplest relevant adsorbates (O, Cl, and H) on relevant surfaces of 3d transition metals, as well as Pd and Pt, in face-centered-cubic and, if different, their ground-state crystal structures. Approximations are tested for electron exchange-correlation (XC) and van der Waals interactions to assess their ability to reproduce experimental adsorption enthalpies of H and O on Pt(111). The vdW-uncorrected generalized gradient approximation to XC of Perdew, Burke, and Ernzerhof (PBE) agrees most closely with experiments. Using DFT-PBE thereafter, it is determined that the O chemisorption enthalpy on this wide range of transition-metal surfaces is proportional to the sum of first and second atomic ionization energies, akin to a Born–Haber cycle for a redox reaction, indicating that metal redox activity controls O chemisorption strength. Then it is shown that the O and Cl chemisorption enthalpies are strongly correlated, suggesting that the transition metals considered will oxidize unselectively water and Cl^–. This strong correlation appears also for crystal reduction potentials of binary oxides and chlorides, indicating a fundamental challenge for future seawater electrode materials design.

URL: https://doi.org/10.1002/adts.202200592

Modeling the Solvation and Acidity of Carboxylic Acids Using an Ab Initio Deep Neural Network Potential

Abhinav S. Raman and Annabella Selloni, J. Phys. Chem. A 2022, 126, 40, 7283–7290 (2022)

Abstract

Formic and acetic acid constitute the simplest of carboxylic acids, yet they exhibit fascinating chemistry in the condensed phase such as proton transfer and dimerization. The go-to method of choice for modeling these rare events have been accurate but expensive ab initio molecular dynamics simulations. In this study, we present a deep neural network potential trained using accurate ab initio data that can be used in tandem with enhanced-sampling methods to perform an efficient exploration of the free-energy surface of aqueous solutions of weak carboxylic acids. In particular, we show that our model captures proton dissociation and provides a good estimate of the pK_a, as well as the dimerization of formic and acetic acid. This provides a suitable starting point for applications in different research areas where computational efficiency coupled with the accuracy of ab initio methods is required.

URL: https://doi.org/10.1021/acs.jpca.2c06252

A Tensor Network Path Integral Study of Dynamics in B850 LH2 Ring with Atomistically Derived Vibrations

Amartya Bose and Peter L. Walters, J. Chem. Theory Comput. 2022, 18, 7, 4095–4108 (2022)

Abstract

The recently introduced multisite tensor network path integral (MS-TNPI) allows simulation of extended quantum systems coupled to dissipative media. We use MS-TNPI to simulate the exciton transport and the absorption spectrum of a B850 bacteriochlorophyll (BChl) ring. The MS-TNPI network is extended to account for the ring topology of the B850 system. Accurate molecular-dynamics-based description of the molecular vibrations and the protein scaffold is incorporated through the framework of Feynman–Vernon influence functional. To relate the present work with the excitonic picture, an exploration of the absorption spectrum is done by simulating it using approximate and topologically consistent transition dipole moment vectors. Comparison of these numerically exact MS-TNPI absorption spectra are shown with second-order cumulant approximations. The effect of temperature on both the exact and the approximate spectra is also explored.

URL: https://doi.org/10.1021/acs.jctc.2c00163

Effect of Temperature Gradient on Quantum Transport

Amartya Bose and Peter L. Waters, Phys. Chem. Chem. Phys., 2022,24, 22431-22436 (2022)

Abstract

The recently introduced multisite tensor network path integral (MS-TNPI) method [Bose and Walters, J. Chem. Phys., 2022, 156, 24101.] for simulation of quantum dynamics of extended systems has been shown to be effective in studying one-dimensional systems. Quantum transport in these systems are typically studied at a constant temperature. However, temperature seems to be a very obvious parameter that can be spatially changed to control the quantum transport. Here, MS-TNPI is used to study “non-equilibrium” effects of an externally imposed temperature gradient on the quantum transport in one-dimensional extended quantum systems.

URL: https://doi.org/10.1039/D2CP03030F

Phase diagram of the TIP4P/Ice water model by enhanced sampling simulations

Sigbjørn L Bore, Pablo M Piaggi, Roberto Car, Francesco Paesani, J. Chem. Phys. 157, 054504 (2022)

Abstract

We studied the phase diagram for the TIP4P/Ice water model using enhanced sampling molecular dynamics simulations. Our approach is based on the calculation of ice-liquid free energy differences from biased coexistence simulations that sample reversibly the melting and growth of layers of ice. We computed a total of 19 melting points for five different ice polymorphs which are in excellent agreement with the melting lines obtained from the integration of the Clausius-Clapeyron equation. For proton-ordered and fully proton-disordered ice phases, the results are in very good agreement with previous calculations based on thermodynamic integration. For the partially-proton-disordered ice III, we find a large increase in stability that is in line with previous observations using direct coexistence simulations for the TIP4P/2005 model. This issue highlights the robustness of the approach employed here for ice polymorphs with diverse degrees of proton disorder. Our approach is general and can be applied to the calculation of other complex phase diagrams.

URL: https://doi.org/10.1063/5.0097463

Exact Factorization Adventures: A Promising Approach for Non-Bound State

Evaristo Villaseco Arribas, Federica Agostini, and Neepa T. Maitra, Molecules (for a special issue on Molecular Quantum Dynamics Beyond Bound States), Molecules 2022, 27(13), 4002 (2022)

Abstract

Modeling the dynamics of non-bound states in molecules requires an accurate description of how electronic motion affects nuclear motion and vice-versa. The exact factorization (XF) approach offers a unique perspective, in that it provides potentials that act on the nuclear subsystem or electronic subsystem, which contain the effects of the coupling to the other subsystem in an exact way. We briefly review the various applications of the XF idea in different realms, and how features of these potentials aid in the interpretation of two different laser-driven dissociation mechanisms. We present a detailed study of the different ways the coupling terms in recently-developed XF-based mixed quantum-classical approximations are evaluated, where either truly coupled trajectories, or auxiliary trajectories that mimic the coupling are used, and discuss their effect in both a surface-hopping framework as well as the rigorously-derived coupled-trajectory mixed quantum-classical approach.

URL: https://www.mdpi.com/1420-3049/27/13/4002

Many-Body Effects in the X-ray Absorption Spectra of Liquid Water

Fujie Tang, Zhenglu Li, Chunyi Zhang, Steven G. Louie, Roberto Car, Diana Y. Qiu and Xifan Wu, Proc. Natl. Acad. Sci. U.S.A., 2022, 119, e2201258119 (2022)

Abstract

X-ray absorption spectroscopy (XAS) is a powerful experimental technique to probe the local order in materials with core electron excitations. Experimental interpretation requires supporting theoretical calculations. For water, these calculations are very demanding and, to date, could only be done with major approximations that limited the accuracy of the calculated spectra. This prompted an intense debate on whether a substantial revision of the standard picture of tetrahedrally bonded water was necessary to improve the agreement of theory and experiment. Here, we report a new first-principles calculation of the XAS of water that avoids the approximations of prior work thanks to recent advances in electron excitation theory. The calculated XAS spectra, and their variation with changes of temperature and/or with isotope substitution, are in excellent quantitative agreement with experiments. The approach requires accurate quasi-particle wavefunctions beyond density functional theory approximations, accounts for the dynamics of quasi-particles and includes dynamic screening as well as renormalization effects due to the continuum of valence-level excitations. The three features observed in the experimental spectra are unambiguously attributed to excitonic effects. The pre-edge feature is associated to a bound intramolecular exciton, the main-edge feature is associated to an exciton localized within the coordination shell of the excited molecule, while the post-edge one is delocalized over more distant neighbors, as expected for a resonant state. The three features probe the local order at short, intermediate, and longer range relative to the excited molecule. The calculated spectra are fully consistent with a standard tetrahedral picture of water.

URL: https://www.pnas.org/doi/full/10.1073/pnas.2201258119

Anomalies and Local Structure of Liquid Water from Boiling to the Supercooled Regime as Predicted by the Many-Body MB-pol Model

Thomas E. Gartner III, Kelly M. Hunter, Eleftherios Lambros, Alessandro Caruso, Marc Riera, Gregory R. Medders, Athanassios Z. Panagiotopoulos, Pablo G. Debenedetti and Francesco Paesani, J. Phys. Chem. Lett. 2022, 13, XXX, 3652–3658 (2022)

Abstract

For the last 50 years, researchers have sought molecular models that can accurately reproduce water’s microscopic structure and thermophysical properties across broad ranges of its complex phase diagram. Herein, molecular dynamics simulations with the many-body MB-pol model are performed to monitor the thermodynamic response functions and local structure of liquid water from the boiling point down to deeply supercooled temperatures at ambient pressure. The isothermal compressibility and isobaric heat capacity show maxima near 223 K, in excellent agreement with recent experiments, and the liquid density exhibits a minimum at ~208 K. A local tetrahedral arrangement, where each water molecule accepts and donates two hydrogen bonds, is found to be the most probable hydrogen-bonding topology at all temperatures. This work suggests that MB-pol may provide predictive capability for studies of liquid water’s physical properties across broad ranges of thermodynamic states, including the so-called water’s “no man’s land” which is difficult to probe experimentally.

URL: https://pubs.acs.org/doi/full/10.1021/acs.jpclett.2c00567

Pathways for Electron Transfer at MgO-Water Interfaces from Ab-Initio Molecular Dynamics

Zhutian Ding, Zachary K Goldsmith and Annabella Selloni, J. Am. Chem. Soc. 144, 2002-2009 (2022)

Abstract

The nature of electron transfer across metal oxide−water interfaces depends significantly on the band gap of the oxide and its band edge energies relative to the potentials of relevant aqueous redox couples. Here we focus on the water interface with MgO, a prototypical wide band gap oxide whose conduction band edge is close in energy to that of water. We investigate the behavior of an excess electron at and out of equilibrium near the interface using ab initio molecular dynamics based on hybrid density functional theory. Our simulations show that under equilibrium conditions the excess electron (donated by an Al impurity in MgO) localizes to a midgap defect state comparable in energy and shape to a hydrated electron in bulk water. To characterize the electron transfer from the conduction band of MgO to interfacial product states, we dope near-equilibrium configurations of the pristine MgO−water system with Al and run short trajectories of these instantaneously out-of-equilibrium systems. We observe two distinct products associated with the excess electron: a surface-localized electron (e− ) and an aqueous hydrogen radical (H•). The H• pathway exhibits a much higher activation barrier despite being more exoergic, making e− the kinetic product. Our characterization of the pathways on the basis of Marcus theory is consistent with the poor observed utility of MgO for water radiolysis. Moreover, we anticipate that the computational framework employed here will be broadly applicable to assessing electron transfer mechanisms at aqueous, photocatalytic interfaces.

URL: https://pubs.acs.org/doi/10.1021/jacs.1c13250?ref=pdf

Highly accurate and constrained density functional obtained with differentiable programming

Sebastian Dick and Marivi Fernandez-Serra, Phys Rev B 104, L161109 (2021)

Abstract

Using an end-to-end differentiable implementation of the Kohn-Sham self-consistent field equations, we obtain a highly accurate neural network–based exchange and correlation (XC) functional of the electronic density. The functional is optimized using information on both energy and density while exact constraints are enforced through an appropriate neural network architecture. We evaluate our model against different families of XC approximations and show that at the meta-GGA level our functional exhibits unprecedented accuracy for both energy and density predictions. For nonempirical functionals, there is a strong linear correlation between energy and density errors. We use this correlation to define an XC functional quality metric that includes both energy and density errors, leading to an improved way to rank different approximations.

URL: https://doi.org/10.1103/PhysRevB.104.L161109

A deep potential model with long-range electrostatic interactions [Editor’s pick]

L Zhang, H Wang, MC Muniz, AZ Panagiotopoulos, R. Car, W. E, J. Chem. Phys. 156, 124107 (2022)

Abstract

Machine learning models for the potential energy of multi-atomic systems, such as the deep potential (DP) model, make possible molecular simulations with the accuracy of quantum mechanical density functional theory, at a cost only moderately higher than that of empirical force fields. However, the majority of these models lack explicit long-range interactions and fail to describe properties that derive from the Coulombic tail of the forces. To overcome this limitation we extend the DP model by approximating the long-range electrostatic interaction between ions (nuclei+core electrons) and valence electrons with that of distributions of spherical Gaussian charges located at ionic and electronic sites. The latter are rigorously defined in terms of the centers of the maximally localized Wannier distributions, whose dependence on the local atomic environment is modeled accurately by a deep neural network. In the deep potential long-range (DPLR) model, the electrostatic energy of the Gaussian charge system is added to short-range interactions that are represented as in the standard DP model. The resulting potential energy surface is smooth and possesses analytical forces and virial. Missing effects in the standard DP scheme are recovered, improving on accuracy and predictive power. By including long-range electrostatics, DPLR correctly extrapolates to large systems the potential energy surface learned from quantum mechanical calculations on smaller systems. We illustrate the approach with three examples, the potential energy profile of the water dimer, the free energy of interaction of a water molecule with a liquid water slab, and the phonon dispersion curves of the NaCl crystal.

URL: https://doi.org/10.1063/5.0083669

Using Neural Networks force fields to ascertain the quality of ab initio simulations of liquid water

Alberto Torres, Luana S. Pedroza, Marivi Fernandez-Serra,and Alexandre R.Rocha, submitted to J. Phys. Chem. C (2021)

Abstract

Accurately simulating the properties of bulk water, despite the apparent simplicity of the molecule, is still a challenge. In order to fully understand and reproduce its complex phase diagram, it is necessary to perform simulations at the ab initio level, including quantum mechanical effects both for electrons and nuclei. This comes at a high computational cost, given that the structural and dynamical properties tend to require long timescales and large simulation cells. In this work, we evaluate the errors that density functional theory (DFT)-based simulations routinely incur into due time- and size-scale limitations. These errors are evaluated using neural-network-trained force fields that are accurate at the level of DFT methods. We compare different exchange and correlation potentials for properties of bulk water that require large timescales. We show that structural properties are less dependent on the system size and that dynamical properties such as the diffusion coefficient have a strong dependence on the simulation size and timescale. Our results facilitate comparisons of DFT-based simulation results with experiments and offer a path to discriminate between model and convergence errors in these simulations.

URL: https://doi.org/10.1021/acs.jpcb.1c04372

Homogeneous ice nucleation in an ab-initio machine learning model of water

Pablo M. Piaggi, Jack Weiss, Athanassios Z. Panagiotopoulos, Pablo G. Debenedetti, Roberto Car, PNAS, 119 (33) e2207294119 (2022)

Abstract

Molecular simulations have provided valuable insight into the microscopic mechanisms underlying homogeneous ice nucleation. While empirical models have been used extensively to study this phenomenon, simulations based on first-principles calculations have so far proven prohibitively expensive. Here, we circumvent this difficulty by using an efficient machine learning model trained on density-functional theory (DFT) energies and forces. We compute nucleation rates at atmospheric pressure, over a broad range of supercoolings, using the seeding technique and systems of up to hundreds of thousands of atoms simulated with ab initio accuracy. The key quantity provided by the seeding technique is the size of the critical cluster (i.e., a size such that the cluster has equal probabilities of growing or melting at the given supersaturation) which is used together with the equations of classical nucleation theory to compute nucleation rates. We find that nucleation rates for our model at moderate supercoolings are in good agreement with experimental measurements within the error of our calculation. We also study the impact of properties such as the thermodynamic driving force, interfacial free energy, and stacking disorder on the calculated rates.

URL: https://doi.org/10.1073/pnas.2207294119

Viscosity in water from first-principles and deep-neural-network simulations

Cesare Malosso, Linfeng Zhang, Roberto Car, Stefano Baroni, Davide Tisi, npj Computational Materials, volume 8, Article number: 139 (2022)

Abstract

We report on an extensive study of the viscosity of liquid water at near-ambient conditions, performed within the Green-Kubo theory of linear response and equilibrium ab initio molecular dynamics (AIMD), based on density-functional theory (DFT). In order to cope with the long simulation times necessary to achieve an acceptable statistical accuracy just above melting, our ab initio approach is enhanced with deep-neural-network potentials (NNP) force fields, trained and validated on extensive DFT data. This approach is first validated against AIMD results for the viscosity, obtained by using the PBE exchange-correlation functional and paying careful attention to crucial aspects of the statistical data analysis that are often overlooked. We then train a second NNP to a dataset generated from the SCAN-DFT functional, which is known to describe significantly better than PBE the H-bonding features of different phases of water. Close to melting, the viscosity depends very sensitively on temperature. Once the error resulting from the imperfect prediction of the melting line is offset by referring the simulated temperature to the theoretical melting one, our SCAN DFT predictions of the shear viscosity of water are in very good agreement with experiments.

URL: https://www.nature.com/articles/s41524-022-00830-7

Dissolving salt is not equivalent to applying a pressure on water

Chunyi Zhang, Shuwen Yue, Athanassios Z. Panagiotopoulos, Michael L. Klein, and Xifan Wu, Nature Communications 13, Article number: 822 (2022)

Abstract

Salt water is ubiquitous, playing crucial roles in geological and physiological processes. Despite centuries of investigations, whether or not water’s structure is drastically changed by dissolved ions is still debated. Based on density functional theory, we employ machine learning based molecular dynamics to model sodium chloride, potassium chloride, and sodium bromide solutions at different concentrations. The resulting computed reciprocal-space structure factors agree quantitatively with corresponding neutron diffraction data. Here we provide clear evidence that the ions in salt water do not distort the structure of water in the same way as neat water responds to elevated pressure. Rather, the computed structural changes are restricted to the ionic first solvation shells intruding into the hydrogen bond network, beyond which the oxygen radial-distribution function does not undergo major change relative to neat water. Our findings suggest that the widely cited pressure-like effect on the solvent in Hofmeister series ionic solutions should be carefully revisited.

URL:https://doi.org/10.1038/s41467-022-28538-8

A pairwise connected tensor network representation of path integrals

Amartya Bose, Phys. Rev. B 105, 024309 (2022)

Abstract

It has been recently shown how the tensorial nature of real-time path integrals involving the Feynman-Vernon influence functional can be utilized using matrix product states, taking advantage of the finite length of the non-Markovian memory. Tensor networks promise to provide a new, unified language to express the structure of path integral. Here, a generalized tensor network is derived and implemented specifically incorporating the pairwise interaction structure of the influence functional, allowing for a compact representation and efficient evaluation. This pairwise connected tensor network path integral (PCTNPI) is illustrated through applications to typical spin-boson problems and explorations of the differences caused by the exact form of the spectral density. The storage requirements and performance are compared with iterative quasi-adiabatic propagator path integral and iterative blip-summed path integral. Finally, the viability of using PCTNPI for simulating multistate problems is demonstrated taking advantage of the compressed representation.

URL:https://doi.org/10.1103/PhysRevB.105.024309

A Computational Study of RNA Tetraloop Thermodynamics, Including Misfolded States

Gül H. Zerze, Pablo M. Piaggi, and Pablo G. Debenedetti
J. Phys. Chem. B 125, 50, 13685-13695 (2021)

Abstract

An important characteristic of RNA folding is the adoption of alternative configurations of similar stability, often referred to as misfolded configurations. These configurations are considered to compete with correctly folded configurations, although their rigorous thermodynamic and structural characterization remains elusive. Tetraloop motifs found in large ribozymes are ideal systems for an atomistically detailed computational quantification of folding free energy landscapes and the structural characterization of their constituent free energy basins, including nonnative states. In this work, we studied a group of closely related 10-mer tetraloops using a combined parallel tempering and metadynamics technique that allows a reliable sampling of the free energy landscapes, requiring only knowledge that the stem folds into a canonical A-RNA configuration. We isolated and analyzed unfolded, folded, and misfolded populations that correspond to different free energy basins. We identified a distinct misfolded state that has a stability very close to that of the correctly folded state. This misfolded state contains a predominant population that shares the same structural features across all tetraloops studied here and lacks the noncanonical A-G base pair in its loop portion. Further analysis performed with biased trajectories showed that although this competitive misfolded state is not an essential intermediate, it is visited in most of the transitions from unfolded to correctly folded states. Moreover, the tetraloops can transition from this misfolded state to the correctly folded state without requiring extensive unfolding.

URL:https://pubs.acs.org/doi/abs/10.1021/acs.jpcb.1c08038

Heat transport in liquid water from first-principles and deep neural network simulations

Davide Tisi, Linfeng Zhang, Riccardo Bertossa, Han Wang, Roberto Car, and Stefano Baroni, Phys. Rev. B 104, 224202 – Published 13 (2021)

Abstract

We compute the thermal conductivity of water within linear response theory from equilibrium molecular dynamics simulations, by adopting two different approaches. In one, the potential energy surface (PES) is derived on the fly from the electronic ground state of density functional theory (DFT) and the corresponding analytical expression is used for the energy flux. In the other, the PES is represented by a deep neural network (DNN) trained on DFT data, whereby the PES has an explicit local decomposition and the energy flux takes a particularly simple expression. By virtue of a gauge invariance principle, established by Marcolongo, Umari, and Baroni, the two approaches should be equivalent if the PES were reproduced accurately by the DNN model. We test this hypothesis by calculating the thermal conductivity, at the GGA (PBE) level of theory, using the direct formulation and its DNN proxy, finding that both approaches yield the same conductivity, in excess of the experimental value by approximately 60%. Besides being numerically much more efficient than its direct DFT counterpart, the DNN scheme has the advantage of being easily applicable to more sophisticated DFT approximations, such as meta-GGA and hybrid functionals, for which it would be hard to derive analytically the expression of the energy flux. We find in this way that a DNN model, trained on meta-GGA (SCAN) data, reduces the deviation from experiment of the predicted thermal conductivity by about 50%, leaving the question open as to whether the residual error is due to deficiencies of the functional, to a neglect of nuclear quantum effects in the atomic dynamics, or, likely, to a combination of the two.

URL:https://doi.org/10.1103/PhysRevB.104.224202

Hydrogen Bonds and H3O+ Formation at the Water Interface with Formic Acid Covered Anatase TiO2

Bo Wen, and Annabella Selloni, J Phys Chem Lett., 12, 6840-6846 (2021)

Abstract

Carboxylic acid-modified TiO₂ surfaces in aqueous environment are of widespread interest, yet atomic-scale understanding of their structure is limited. We here investigate formic acid (FA) on anatase TiO₂ (101) (A-101) in contact with water using density functional theory (DFT) and ab initio molecular dynamics (AIMD). Isolated FA molecules adsorbed in a deprotonated bridging bidentate (BD) form on A-101 are found to remain stable at the interface with water, with the acid proton transferred to a surface oxygen to form a surface bridging hydroxyl (O_brH). With increasing FA coverage, adsorbed monolayers of only BD and successively of alternating monodentate (MD) and BD species give rise to a higher concentration of surface O_brH’s. Simulations of these adsorbed monolayers in water environment show that some protons are released from the surface O_brH’s to water resulting in a negatively charged surface with nearby solvated H₃O⁺ ions. These results provide insight into the complex acid–base equilibrium between an oxide surface, adsorbates and water and can also help obtain a better understanding of the wetting properties of chemically modified TiO₂ surfaces.

URL:https://pubs.acs.org/doi/abs/10.1021/acs.jpclett.1c01886

Nuclear quantum effects on the quasiparticle properties of the chloride anion aqueous solution within the GW approximation

Fujie Tang, Jianhang Xu, Diana Y. Qiu, and Xifan Wu,, Phys. Rev. B 104, 035117 (2021)

Abstract

Photoelectron spectroscopy experiments in ionic solutions reveal important electronic structure information, in which the interaction between hydrated ions and water solvent can be inferred. Based on many-body perturbation theory with GW approximation, we theoretically compute the quasiparticle electronic structure of chloride anion solution, which is modeled by path-integral ab initio molecular dynamics simulation by taking account the nuclear quantum effects (NQEs). The electronic levels of hydrated anion as well as water are determined and compared with the recent experimental photoelectron spectra. It is found that NQEs improve the agreement between theoretical prediction and experiment because NQEs effectively weaken the hybridization of the between the Cl⁻ anion and water. Our study indicates that NQEs plays a small but non-negligible role in predicting the electronic structure of the aqueous solvation of ions of the Hofmeister series.

URL:https://doi.org/10.1103/PhysRevB.104.035117

Exact-Factorization-Based Surface-Hopping for Multi-State Dynamics

Patricia Vindel-Zandbergen, Spiridoula Matsika and Neepa T. Maitra, . Phys. Chem. Lett. 2022, 13, 7, 1785–1790 (2022)

Abstract

A surface-hopping algorithm recently derived from the exact factorization approach, SHXF, [Ha, Lee, Min, J. Phys. Chem. Lett. 9, 1097 (2018)] introduces an additional term in the electronic equation of surface-hopping, which couples electronic states through the quantum momentum. This term not only provides a first-principles description of decoherence but here we show it is crucial to accurately capture non-adiabatic dynamics when more than two states are occupied at any given time. Using a vibronic coupling model of the uracil cation, we show that the lack of this term in traditional surface-hopping methods, including those with decoherence-corrections, leads to failure to predict the dynamics through a three-state intersection, while SHXF performs similarly to the multi-configuration time dependent Hartree quantum dynamics benchmark.

URL: https://doi.org/10.1021/acs.jpclett.1c04132

A Multisite Decomposition of the Tensor Network Path Integrals

Amartya Bose and Peter L. Walters, accepted by J. Chem. Phys. 156, 024101 (2022)

Abstract

Tensor network decompositions of path integrals for simulating open quantum systems have recently been proven to be useful. In this work, we extend the tensor network path integral (TNPI) framework to efficiently simulate extended systems coupled with local vibrational and phononic modes. The Feynman-Vernon influence functional is a very popular approach used to account for the effect of a bath on the dynamics of the system. In order to facilitate the incorporation of the influence functional into a multisite framework (MS-TNPI), we combine a matrix product state decomposition of the reduced density tensor of the system along the sites with a corresponding tensor network representation of the time axis to construct an efficient 2D tensor network. The 2D MS-TNPI network, when finally contracted, yields the time-dependent reduced density tensor of the extended system as a matrix product state. The decomposition and algorithm presented are independent of the nature of the system Hamiltonian. We also outline an iteration scheme to take the simulation beyond the non-Markovian memory length introduced by the dissipative baths. Applications to spin chains coupled to local harmonic baths is presented; we consider interactions defined by the Ising, XXZ and the Heisenberg models. We demonstrate that the presence of dissipative environments can often dissipate the entanglement between the sites as measured by the bond dimension of the reduced density matrix product state. The MS-TNPI method would be useful for studying a variety of extended quantum systems coupled with vibrational baths or phononic modes.

URL:https://aip.scitation.org/doi/abs/10.1063/5.0073234

Modeling the Ultrafast Electron Attachment Dynamics of Solvated Uracil

Cate S. Anstöter, Mark DelloStritto, Michael L. Klein, and Spiridoula Matsika, J. Phys. Chem. A 2021, 125, 32, 6995–7003 (2021)

Abstract

Electron attachment to DNA by low energy electrons can lead to DNA damage, so a fundamental understanding of how electrons interact with the components of nucleic acids in solution is an open challenge. In solution, low energy electrons can generate presolvated electrons, epre, which are efficiently scavanged by pyrimidine nucleobases to form transient negative ions, able to relax to either stable valence bound anions or undergo dissociative electron detachment or transfer to other parts of DNA/RNA leading to strand breakages. In order to understand the initial electron attachment dynamics, this paper presents a joint molecular dynamics and high-level electronic structure study into the behavior of the electronic states of the solvated uracil anion. Both the valence and non-valence epre states of the solvated uracil system are studied, and the effect of the solvent environment and the geometric structure of the uracil core are uncoupled to gain insight into the physical origin of the stabilization of the solvated uracil anion. Solvent reorganization is found to play a dominant role followed by relaxation of the uracil core.

URL:https://pubs.acs.org/doi/abs/10.1021/acs.jpca.1c05288

Modeling liquid water by climbing up Jacob’s ladder in density functional theory facilitated by using deep neural network potentials

Chunyi Zhang, Fujie Tang, Mohan Chen, Jianhang Xu, Linfeng Zhang, Diana Y. Qiu, John P. Perdew, Michael L. Klein, and Xifan Wu, J. Phys. Chem. B 2021, 125, 11444−11456 (2021)

Abstract

Within the framework of Kohn-Sham density functional theory (DFT), the ability to provide good predictions of water properties by employing a strongly constrained and appropriately normed (SCAN) functional has been extensively demonstrated in recent years. Here, we further advance the modeling of water by building a more accurate model on the fourth rung of Jacob’s ladder with the hybrid functional, SCAN0. In particular, we carry out both classical and Feynman path-integral molecular dynamics calculations of water with the SCAN0 functional and the isobaric-isothermal ensemble. In order to generate the equilibrated structure of water, a deep neural network potential is trained from the atomic potential energy surface based on ab initio data obtained from SCAN0 DFT calculations. For the electronic properties of water, a separate deep neural network potential is trained using the Deep Wannier method based on the maximally localized Wannier functions of the equilibrated trajectory at the SCAN0 level. The structural, dynamic, and electric properties of water were analyzed. The hydrogen-bond structures, density, infrared spectra, diffusion coefficients, and dielectric constants of water, in the electronic ground state, are computed using a large simulation box and long simulation time. For the properties involving electronic excitations, we apply the GW approximation within many-body perturbation theory to calculate the quasiparticle density of states and bandgap of water. Compared to the SCAN functional, mixing exact exchange mitigates the self-interaction error in the meta-generalized-gradient approximation and further softens liquid water towards the experimental direction. For most of the water properties, the SCAN0 functional shows a systematic improvement over the SCAN functional. However, some important discrepancies remain. The H-bond network predicted by the SCAN0 functional is still slightly overstructured compared to the experimental results.

URL:https://https://doi.org/10.1021/acs.jpcb.1c03884

Hydration structure of flat and stepped MgO surfaces

Zhutian Ding, and Annabella Selloni, J. Chem. Phys. 154, 114708 (2021)

Abstract

We investigate the solvation structure of flat and stepped MgO(001) in neutral liquid water using ab initio molecular dynamics based on a hybrid density functional with dispersion corrections. Our simulations show that the MgO surface is covered by a densely packed layer of mixed intact and dissociated adsorbed water molecules in a planar arrangement with strong intermolecular H-bonds. The water dissociation fractions in this layer are >20% and >30% on the flat and stepped surfaces, respectively. Slightly above the first water layer, we observe metastable OH groups perpendicular to the interface, similar to those reported in low temperature studies of water monolayers on MgO. These species receive hydrogen bonds from four nearby water molecules in the first layer and have their hydrophobic H end directed toward bulk water, while their associated protons are bound to surface oxygens. The formation of these OH species is attributed to the strong basicity of the MgO surface and can be relevant for understanding various phenomena from morphology evolution and growth of (nano)crystalline MgO particles to heterogeneous catalysis.

URL: https://doi.org/10.1063/5.0044700

Phase Diagram of a Deep Potential Water Model

Linfeng Zhang, Han Wang, Roberto Car, and Weinan E, Phys. Rev. Lett. 126, 236001 (2021)

Abstract

Using the Deep Potential methodology, we construct a model that reproduces accurately the potential energy surface of the SCAN approximation of density functional theory for water, from low temperature and pressure to about 2400 K and 50 GPa, excluding the vapor stability region. The computational efficiency of the model makes it possible to predict its phase diagram using molecular dynamics. Satisfactory overall agreement with experimental results is obtained. The fluid phases, molecular and ionic, and all the stable ice polymorphs, ordered and disordered, are predicted correctly, with the exception of ice III and XV that are stable in experiments, but metastable in the model. The evolution of the atomic dynamics upon heating, as ice VII transforms first into ice VII′′ and then into an ionic fluid, reveals that molecular dissociation and breaking of the ice rules coexist with strong covalent fluctuations, explaining why only partial ionization was inferred in experiments.

URL:https://link.aps.org/doi/10.1103/PhysRevLett.126.236001

Vapor–liquid equilibrium of water with the MB-pol many-body potential

Maria Carolina Muniz, Thomas E. Gartner III, Marc Riera, Christopher Knight, Shuwen Yue, Francesco Paesani, and Athanassios Z. Panagiotopoulos, J. Chem. Phys. 154, 211103 (2021)

Abstract

Among the many existing molecular models of water, the MB-pol many-body potential has emerged as a remarkably accurate model, capable of reproducing thermodynamic, structural, and dynamic properties across water’s solid, liquid, and vapor phases. In this work, we assessed the performance of MB-pol with respect to an important set of properties related to vapor–liquid coexistence and interfacial behavior. Through direct coexistence classical molecular dynamics simulations at temperatures of 400 K < T < 600 K, we calculated properties such as equilibrium coexistence densities, vapor–liquid interfacial tension, vapor pressure, and enthalpy of vaporization and compared the MB-pol results to experimental data. We also compared rigid vs fully flexible variants of the MB-pol model and evaluated system size effects for the properties studied. We found that the MB-pol model predictions are in good agreement with experimental data, even for temperatures approaching the vapor–liquid critical point; this agreement was largely insensitive to system sizes or the rigid vs flexible treatment of the intramolecular degrees of freedom. These results attest to the chemical accuracy of MB-pol and its high degree of transferability, thus enabling MB-pol’s application across a large swath of water’s phase diagram.

URL:https://aip.scitation.org/doi/10.1063/5.0050068?via=site

A study of the decoherence correction derived from the exact factorization approach for non-adiabatic dynamics

Patricia Vindel-Zandbergen, Lea M. Ibele, Jong-Kwon Ha, Seung Kyu Min, Basile F.E. Curchod, Neepa T. Maitra, J. Chem. Theory Comput. 2021, 17, 3852−3862 (2021)

Abstract

We present a detailed study of the decoherence correction to surface-hopping that was recently derived from the exact factorization approach. Ab initio multiple spawning calculations that use the same initial conditions and same electronic structure method are used as a reference for three molecules: ethylene, methaniminium cation, and fulvene, for which non-adiabatic dynamics follows a photo-excitation. A comparison with the Granucci-Persico energy-based decoherence correction, and the augmented fewest-switches surface-hopping scheme shows that the three decoherence-corrected methods operate on individual trajectories in a qualitatively different way, but results averaged over trajectories are similar for these systems.

URL:https://pubs.acs.org/doi/10.1021/acs.jctc.1c00346

Manifestations of metastable criticality in the long-range structure of model water glasses

Thomas E. Gartner III, Salvatore Torquato, Roberto Car, and Pablo G. Debenedetti, Nature Communications, 12, 3398 (2021)

Abstract

Water’s metastable phase behavior has, for decades, been a source of interest to researchers across a broad range of the physical sciences. Much attention has been devoted to water’s polyamorphism (multiple amorphous solid phases) and to the hypothesized metastable liquid-liquid transition and associated critical point (LLCP). However, the possible relationship between these phenomena remains incompletely understood. Using molecular dynamics simulations of the realistic TIP4P/2005 model, we found a striking signature of the LLCP in the structure of water glasses, manifested as a pronounced increase in long-range density fluctuations in the vicinity of the critical pressure associated with this model’s liquid-liquid transition. By contrast, such long-range density fluctuations were absent in glasses of two model systems that lack an LLCP. We also characterized the effect of applied pressure on the departure from equilibrium upon glass formation, as quantified by the ‘non-equilibrium index’, and found that water-like systems exhibited a strong pressure dependence in this metric, whereas simple liquids did not. These results reflect a surprising relationship between the metastable equilibrium phenomenon of the LLCP and the non-equilibrium long-range structure of glassy water, with implications for our understanding of water phase behavior, and more broadly of glass physics. Furthermore, our computational approach suggests a possible experimental route to probing the existence of the LLCP in water and other fluids.

URL: https://doi.org/10.1038/s41467-021-23639-2

Effects of applied voltage on water at a gold electrode interface from ab initio molecular dynamics

Zachary K. Goldsmith, Marcos F. Calegari Andrade, Annabella Selloni, Chem. Sci. (2021)

Abstract

Electrode-water interfaces under voltage bias demonstrate anomalous electrostatic and structural properties that are influential in their catalytic and technological applications. Mean-field and empirical models of the electrical double layer (EDL) that forms in response to an applied potential do not capture the heterogeneity that polarizable, liquid-phase water molecules engender. To illustrate the inhomogeneous nature of the electrochemical interface, Born-Oppenheimer ab initio molecular dynamics calculations of electrified Au(111) slabs interfaced with liquid water were performed using a combined explicit-implicit solvent approach. The excess charges localized on the model electrode were held constant and the electrode potentials were computed at frequent simulation times. The electrode potential in each trajectory fluctuated with changes in the atomic structure, and the trajectory-averaged potentials converged and yielded a physically reasonable differential capacitance for the system. The effects of the average applied voltages, both positive and negative, on the structural, hydrogen bonding, dynamical, and vibrational properties of water were characterized and compared to literature where applicable. Controlled-potential simulations of the interfacial solvent dynamics provide a framework for further investigation of more complex or reactive species in the EDL and broadly for understanding electrochemical interfaces in situ.

URL: https://pubs.rsc.org/en/content/articlelanding/2021/SC/D1SC00354B#!divAbstract

Phase equilibrium of water with hexagonal and cubic ice using the SCAN functional

PM Piaggi, AZ Panagiotopoulos, PG Debenedetti, and R Car, J. Chem. Theory Comput. 17, 5, 3065–3077 (2021)

Abstract

Machine learning models are rapidly becoming widely used to simulate complex physicochemical phenomena with ab initio accuracy. Here, we use one such model as well as direct density functional theory (DFT) calculations to investigate the phase equilibrium of water, hexagonal ice (Ih), and cubic ice (Ic), with an eye towards studying ice nucleation. The machine learning model is based on deep neural networks and has been trained on DFT data obtained using the SCAN exchange and correlation functional. We use this model to drive enhanced sampling simulations aimed at calculating a number of complex properties that are out of reach of DFT-driven simulations and then employ an appropriate reweighting procedure to compute the corresponding properties for the SCAN functional. This approach allows us to calculate the melting temperature of both ice polymorphs, the driving force for nucleation, the heat of fusion, the densities at the melting temperature, the relative stability of ice Ih and Ic, and other properties. We find a correct qualitative prediction of all properties of interest. In some cases, quantitative agreement with experiment is better than for state-of-the-art semiempirical potentials for water. Our results also show that SCAN correctly predicts that ice Ih is more stable than ice Ic.

URL: https://doi.org/10.1021/acs.jctc.1c00041

Enhancing the formation of ionic defects to study the ice Ih/XI transition with molecular dynamics simulations

PM Piaggi and R Car, Mol. Phys. e1916634 (2021)

Abstract

Ice Ih, the common form of ice in the biosphere, contains proton disorder. Its proton-ordered counterpart, ice XI, is thermodynamically stable below 72 K. However, even below this temperature the formation of ice XI is kinetically hindered and experimentally it is obtained by doping ice with KOH. Doping creates ionic defects that promote the migration of protons and the associated change in proton configuration. In this article, we mimic the effect of doping in molecular dynamics simulations using a bias potential that enhances the formation of ionic defects. The recombination of the ions thus formed proceeds through fast migration of the hydroxide and results in the jump of protons along a hydrogen bond loop. This provides a physical and expedite way to change the proton configuration, and to accelerate diffusion in proton configuration space. A key ingredient of this approach is a machine learning potential trained with density functional theory data and capable of modeling molecular dissociation. We exemplify the usefulness of this idea by studying the order-disorder transition using an appropriate order parameter to distinguish the proton environments in ice Ih and XI. We calculate the changes in free energy, enthalpy, and entropy associated with the transition. Our estimated entropy agrees with experiment within the error bars of our calculation.

URL: https://doi.org/10.1080/00268976.2021.1916634

Quantum phase transitions in long-range interacting hyperuniform spin chains in a transverse field

Amartya Bose and Salvatore Torquato, Phys. Rev. B 103, 014118 (2021)

Abstract

Hyperuniform states of matter are characterized by anomalous suppression of long-wavelength density fluctuations. While most of the interesting cases of disordered hyperuniformity are provided by complex many-body systems such as liquids or amorphous solids, classical spin chains with certain long-range interactions have been shown to demonstrate the same phenomenon. Such spin-chain systems are ideal models for exploring the effects of quantum mechanics on hyperuniformity. It is well-known that the transverse field Ising model shows a quantum phase transition (QPT) at zero temperature. Under the quantum effects of a transverse magnetic field, classical hyperuniform spin chains are expected to lose their hyperuniformity. High-precision simulations of these cases are complicated because of the presence of highly nontrivial long-range interactions. We perform an extensive analysis of these systems using density matrix renormalization group simulations to study the possibilities of phase transitions and the mechanism by which they lose hyperuniformity. Even for a spin chain of length 30, we see discontinuous changes in properties like the “τ order metric” of the ground state, the measure of hyperuniformity, and the second cumulant of the total magnetization along the x-direction, all suggestive of first-order QPTs. An interesting feature of the phase transitions in these disordered hyperuniform spin chains is that, depending on the parameter values, the presence of a transverse magnetic field may lead remarkably to an increase in the order of the ground state as measured by the “τ order metric,” even if hyperuniformity is lost. Therefore, it would be possible to design materials to target specific novel quantum behaviors in the presence of a transverse magnetic field. Our numerical investigations suggest that these spin chains can show no more than two QPTs. We further analyze the long-range interacting spin chains via the Jordan-Wigner mapping onto a system of spinless fermions, showing that under the pairwise-interaction approximation and a mean-field treatment, there can be at most two quantum phase transitions. Based on these numerical and theoretical explorations, we conjecture that for spin chains with long-range pair interactions that have convergent cosine transforms, there can be a maximum of two quantum phase transitions at zero temperature.

URL: https://journals.aps.org/prb/abstract/10.1103/PhysRevB.103.014118

When do short-range atomistic machine-learning models fall short?

Shuwen Yue*, Maria Carolina Muniz*, Marcos F. Calegari Andrade, Linfeng Zhang, Roberto Car, and Athanassios Z. Panagiotopoulos, J. Chem. Phys. 154, 034111 (2021)

Abstract

We explore the role of long-range interactions in atomistic machine-learning models by analyzing the effects on fitting accuracy, isolated cluster properties, and bulk thermodynamic properties. Such models have become increasingly popular in molecular simulations given their ability to learn highly complex and multi-dimensional interactions within a local environment; however, many of them fundamentally lack a description of explicit long-range interactions. In order to provide a well-defined benchmark system with precisely known pairwise interactions, we chose as the reference model a flexible version of the Extended Simple Point Charge (SPC/E) water model. Our analysis shows that while local representations are sufficient for predictions of the condensed liquid phase, the short-range nature of machine-learning models falls short in representing cluster and vapor phase properties. These findings provide an improved understanding of the role of long-range interactions in machine learning models and the regimes where they are necessary.

URL: https://aip.scitation.org/doi/10.1063/5.0031215

Importance of nuclear quantum effects on the hydration of chloride ion

Jianhang Xu, Zhaoru Sun, Chunyi Zhang, Mark DelloStritto, Deyu Lu, Michael L. Klein, and Xifan Wu, Physical Review Materials, 5, L012801 (2021)

Abstract

Path-integral ab initio molecular dynamics (PI-AIMD) calculations have been employed to probe the nature of chloride ion solvation in aqueous solution. Nuclear quantum effects (NQEs) are shown to weaken hydrogen bonding between the chloride anion and the solvation shell of water molecules. As a consequence, the disruptive effect of the anion on the solvent water structure is significantly reduced compared to what is found in the absence of NQEs. The chloride hydration structure obtained from PI-AIMD agrees well with information extracted from neutron scattering data. Inparticular, the observed satellite peak in the hydrogen-chloride-hydrogen triple angular distribution serves as a clear signature of NQEs. The present results suggest that NQEs are likely to play acrucial role in determining the structure of saline solutions.

URL: https://journals.aps.org/prmaterials/abstract/10.1103/PhysRevMaterials.5.L012801

DeePKS: A Comprehensive Data-Driven Approach toward Chemically Accurate Density Functional Theory

Yixiao Chen, Linfeng Zhang, Han Wang, and Weinan E, Journal of Chemical Theory and Computation, 2021, 17, 1, 170–181 (2020)

Abstract

We propose a general machine learning-based framework for building an accurate and widely applicable energy functional within the framework of generalized Kohn-Sham density functional theory. To this end, we develop a way of training self-consistent models that are capable of taking large datasets from different systems and different kinds of labels. We demonstrate that the functional that results from this training procedure gives chemically accurate predictions on energy, force, dipole, and electron density for a large class of molecules. It can be continuously improved when more and more data are available.

URL: https://doi.org/10.1021/acs.jctc.0c00872

Electron Trapping and Ion Leaching at the Li-Modified Quartz–Water Interface*

Zhutian Ding and Annabella Selloni, J. Phys. Chem. C 2020, 124, 49, 26741–26747 (2020)

Abstract

Electrons and ions at metal oxide-water interfaces have a critical role in many phenomena and applications, thus making their properties of considerable interest. We here investigate the behavior of a lithium impurity at the water interface with Li-doped SiO2 using hybrid density functional based ab initio molecular dynamics simulations. We find that the excess electron donated by the lithium dopant localizes on a surface silicon atom and is partially solvated at the aqueous interface. While the position of the excess electron does not change during our ∼60 ps simulation, the lithium cation diffuses from its initial position in the SiO2 subsurface toward the interface and eventually leaches out of the oxide surface and becomes solvated by four water molecules, forming an aqueous Li+-electron complex at the interface. The degree of interaction between the localized electron and the lithium ion controls not only the energy level of the excess electron but also the extent of structural distortion and hydrogen-bonding at the interface. The results of our study can be relevant for understanding the initial stages of glass corrosion, where alkali ions are leached from the surface region by interactions with water.

URL: https://pubs.acs.org/doi/abs/10.1021/acs.jpcc.0c07581

First-Principles Modeling of Sodium Ion and Water Intercalation into Titanium Disulfide Interlayers for Water Desalination

Lesheng Li, Shenzhen Xu and Emily A. Carter, Chem. Mater. 2020, 32, 24, 10678–10687 (2020)

Abstract

Recent experiments revealed the possibility of using titanium disulfide (TiS₂) as the cathode material in capacitive deionization (CDI) devices for water desalination. Although it performed stably up to 70 cycles with a salt removal capacity of 14 mg/g (corresponding to a sodium ion removal capacity of 35.8 mg/g) at high molar concentration (600 mM NaCl), the maximum capacity of TiS₂ as a CDI electrode was much lower (∼70 mAh/g) than as a supercapacitor (239 mAh/g). Understanding why the ion capacities of these two configurations of the same material differ will entail elucidating detailed charge/discharge mechanisms at the atomic scale. Here, we present density functional theory simulations of sodium intercalation into TiS₂ interlayers in order to gain such understanding. We systematically investigated TiS₂ stacking patterns and ion intercalation sites, energetics of the intercalated compounds, and phase transformation during sodium intercalation. The calculated structural evolution and capacitance–voltage curve agree quite well with previous measurements. We conclude that the different maximum capacities of TiS₂ measured in aqueous and dry environments originate from weaker interlayer interactions with respect to shear strain after ∼33% intercalation of Na⁺–H₂O pairs, which is detrimental to the mechanical stability of TiS₂. This study sheds light on the underlying mechanisms of ion intercalation into layered materials and contributes to understanding requirements for future design and optimization of CDI electrode materials for water desalination.

URL: https://pubs.acs.org/doi/10.1021/acs.chemmater.0c03891

Isotope effects in molecular structures and electronic properties of liquid water via deep potential molecular dynamics based on the SCAN functional

Jianhang Xu, Chunyi Zhang, Linfeng Zhang, Mohan Chen, Biswajit Santra, and Xifan Wu, Physical Review B 102, 24113 (2020)

Abstract

Feynman path-integral deep potential molecular dynamics (PI-DPMD) calculations have been employed to study both light (H2O) and heavy water (D2O) within the isothermalisobaric ensemble. In particular, the deep neural network is trained based on ab initio data obtained from the strongly constrained and appropriately normed (SCAN) exchange-correlation functional. Because of the lighter mass of hydrogen than deuteron, the properties of light water are more influ https://doi.org/10.1103/PhysRevB.102.214113enced by nuclear quantum effect than those of heavy water. Clear isotope effects are observed and analyzed in terms of hydrogen-bond structure and electronic properties of water that are closely associated with experimental observables. The molecular structures of both liquid H2O and D2O agree well with the data extracted from scattering experiments. The delicate isotope effects on radial distribution functions and angular distribution functions are well reproduced as well. Our approach demonstrates that deep neural network combined with SCAN functional based ab initio molecular dynamics provides an accurate theoretical tool for modeling water and its isotope effects.

URL: https://doi.org/10.1103/PhysRevB.102.214113

Signatures of a liquid-liquid transition in an ab-initio neural network model for water

Thomas E. Gartner III, Linfeng Zhang, Pablo M. Piaggi, Roberto Car, Athanassios Z. Panagiotopoulos, and Pablo G. Debenedetti , PNAS, 117 (42) 26040-26046 (2020)

Abstract

The possible existence of a metastable liquid–liquid transition (LLT) and a corresponding liquid–liquid critical point (LLCP) in supercooled liquid water remains a topic of much debate. An LLT has been rigorously proved in three empirically parametrized molecular models of water, and evidence consistent with an LLT has been reported for several other such models. In contrast, experimental proof of this phenomenon has been elusive due to rapid ice nucleation under deeply supercooled conditions. In this work, we combined density functional theory (DFT), machine learning, and molecular simulations to shed additional light on the possible existence of an LLT in water. We trained a deep neural network (DNN) model to represent the ab initio potential energy surface of water from DFT calculations using the Strongly Constrained and Appropriately Normed (SCAN) functional. We then used advanced sampling simulations in the multithermal–multibaric ensemble to efficiently explore the thermophysical properties of the DNN model. The simulation results are consistent with the existence of an LLCP, although they do not constitute a rigorous proof thereof. We fit the simulation data to a two-state equation of state to provide an estimate of the LLCP’s location. These combined results—obtained from a purely first-principles approach with no empirical parameters—are strongly suggestive of the existence of an LLT, bolstering the hypothesis that water can separate into two distinct liquid forms.

URL: https://doi.org/10.1073/pnas.2015440117

Hydrogen dynamics in supercritical water probed by neutron scattering and computer simulations

Carla Andreani, Giovanni Romanelli, Alexandra Parmentier, Roberto Senesi, Alexander I. Kolesnikov, Hsin-Yu Ko, Marcos Calegari Andrade, and Roberto Car, Hydrogen dynamics in supercritical water probed by neutron scattering and computer simulations, J. Phys. Chem. Lett. 11, 9461–9467 (2020)

Abstract

URL: https://dx.doi.org/10.1021/acs.jpclett.0c02547

Structure of disordered TiO2 phases from ab initio based deep neural network simulations

Marcos Calegari Andrade, M. F. & Selloni, A. Structure of disordered TiO2 phases from ab initio based deep neural network simulations, Phys. Rev. Mater. 4, 113803 (2020)

Abstract

URL: https://doi.org/10.1103/PhysRevMaterials.4.113803

Proton-transfer dynamics in ionized water chains using real-time Time Dependent Density Functional Theory

Vidushi Sharma and Marivi Fernández-Serra, Phys. Rev. Research 2, 043082 (2020)

Abstract

In density functional-theoretic studies of photoionized water-based systems, the role of charge localization in proton-transfer dynamics is not well understood. This is due to the inherent complexity in extracting the contributions of coupled electron-nuclear non-adiabatic dynamics in the presence of exchange and correlation functional errors. In this work, we address this problem by simulating a model system of ionized linear H-bonded water clusters using real-time Time Dependent Density Functional Theory (rt-TDDFT)-based Ehrenfest dynamics. Our aim is to understand how self-interaction error in semilocal exchange and correlation functionals affects the probability of proton transfer. In particular, we show that for H-bonded (H2O)+n chains (with n>3), the proton-transfer probability attains a maximum, becoming comparable to that predicted by hybrid functionals. This is because the formation of hemibonded-type geometries is largely suppressed in extended H-bonded structures. We also show how the degree of localization of the initial photo-hole is connected to the probability of a proton-transfer reaction, as well as to the separation between electronic and nuclear charge. These results are compared to those obtained with adiabatic dynamics where the initial wavefunction is allowed to relax to the ground state of the ion cluster, explaining why different functionals and dynamical approaches lead to quantitatively different results.

URL: https://doi.org/10.1103/PhysRevResearch.2.043082

Machine Learning a Highly Accurate Exchange and Correlation Functional of the Electronic Density

Sebastian Dick and Marivi Fernandez-Serra, Nat Commun 11, 3509 (2020)

Abstract

Density functional theory (DFT) is the standard formalism to study the electronic structure of matter at the atomic scale. In Kohn–Sham DFT simulations, the balance between accuracy and computational cost depends on the choice of exchange and correlation functional, which only exists in approximate form. Here, we propose a framework to create density functionals using supervised machine learning, termed NeuralXC. These machine-learned functionals are designed to lift the accuracy of baseline functionals towards that provided by more accurate methods while maintaining their efficiency. We show that the functionals learn a meaningful representation of the physical information contained in the training data, making them transferable across systems. A NeuralXC functional optimized for water outperforms other methods characterizing bond breaking and excels when comparing against experimental results. This work demonstrates that NeuralXC is a first step towards the design of a universal, highly accurate functional valid for both molecules and solids.

URL: https://doi.org/10.1038/s41467-020-17265-7.

Understanding the Interplay Between the Non-Valence and Valence State of the Uracil Anion Upon Mono-Hydration

Cate S. Anstöter and Spiridoula Matsika, J. Phys. Chem. A, 2020, 124, 44, 9237-9243 (2020)

Abstract

In this work we present an ab initio investigation into the effect of mono-hydration on the interaction of uracil with low energy electrons. Electron attachment and photodetachment experimental studies have previously shown dramatic changes in uracil upon solvation with even a single water molecule, due to an inversion of the character of the ground state of the anion. Here we explore the interplay between the non-valence and valence states of the uracil anion, as a function of geometry and site of solvation. Our model provides unambiguous interpretation of previous photoelectron studies, reproducing the binding energies and photoelectron images for bare uracil and a single isomer of the U•(H2O)1 cluster. The results of this study provide insight into how electrons may attach to hydrated nucleobases. These results lay the foundations for further investigations into the effect of micro-hydration on the electronic structure and dynamics of nucleobases.

URL: https://doi.org/10.1021/acs.jpca.0c07407

Ground state energy functional with Hartree-Fock efficiency and chemical accuracy

Yixiao Chen, Linfeng Zhang, Han Wang, and Weinan E, J. Phys. Chem. A 2020, 124, 35, 7155–7165 (2020)

Abstract

We introduce the Deep Post-Hartree-Fock (DeePHF) method, a machine learning based scheme for constructing accurate and transferable models for the ground-state energy of electronic structure problems. DeePHF predicts the energy difference between results of highly accurate models such as the coupled cluster method and low accuracy models such as the the Hartree-Fock (HF) method, using the ground-state electronic orbitals as the input. It preserves all the symmetries of the original high accuracy model. The added computational cost is less than that of the reference HF or DFT and scales linearly with respect to system size. We examine the performance of DeePHF on organic molecular systems using publicly available datasets and obtain the state-of-art performance, particularly on large datasets.

URL: https://doi.org/10.1021/acs.jpca.0c03886

Isotope effects in x-ray absorption spectra of liquid water

Chunyi Zhang, Linfeng Zhang, Jianhang Xu, Fujie Tang, Biswajit Santra, and Xifan Wu, Physical Review B 102, 115155 (2020)

Abstract

The isotope effects in x-ray absorption spectra of liquid water are studied by a many-body approach within electron-hole excitation theory. The molecular structures of both light and heavy water are modeled by path-integral molecular dynamics based on the advanced deep-learning technique. The neural network is trained on ab initio data obtained with SCAN density functional theory. The experimentally observed isotope effect in x-ray absorption spectra is reproduced semiquantitatively in theory. Compared to the spectrum in normal water, the blueshifted and less pronounced pre- and main-edge in heavy water reflect that the heavy water is more structured at short- and intermediate-range of the hydrogen-bond network. In contrast, the isotope effect on the spectrum is negligible at post-edge, which is consistent with the identical long-range ordering in both liquids as observed in the diffraction experiment.

URL: https://doi.org/10.1103/PhysRevB.102.115155

Hydration of NH+4 in water: bifurcated hydrogen bonding structures and fast rotational dynamics

Jianqing Guo, Liying Zhou, Andrea Zen, Angelos Michaelides, Xifan Wu, Enge Wang, Limei Xu, and Ji Chen, Phys. Rev. Lett. 125, 106001, (2020)

Abstract

Understanding the hydration and diffusion of ions in water at the molecular level is a topic of widespread importance. The ammonium ion (NH+4) is an exemplar system that has received attention for decades because of its complex hydration structure and relevance in industry. Here we report a study of the hydration and the rotational diffusion of NH+4 in water using ab initio molecular dynamics simulations and quantum Monte Carlo calculations. We find that the hydration structure of NH+4 features bifurcated hydrogen bonds, which leads to a rotational mechanism involving the simultaneous switching of a pair of bifurcated hydrogen bonds. The proposed hydration structure and rotational mechanism are supported by existing experimental measurements, and they also help to rationalize the measured fast rotation of NH+4 in water. This study highlights how subtle changes in the electronic structure of hydrogen bonds impacts the hydration structure, which consequently affects the dynamics of ions and molecules in hydrogen bonded systems.

URL: https://doi.org/10.1103/PhysRevLett.125.106001

Stabilization of Hydroxide Ion at Interface of Hydrophobic Monolayer on Water via Reduced Proton Transfer

Shanshan Yang, Mohan Chen, Yudan, Su, Jianhang Xu, Xifan Wu, and Chuanshan Tian, Phys. Rev. Lett. 125, 156803, (2020)

Abstract

We report a joint study, using sum-frequency vibrational spectroscopy (SFVS) and ab initio molecular dynamics (AIMD) simulations, respectively, on (sub-)monolayer hexane/water interface with varied vapor pressures of hexane and different pHs in water. We show clear evidence that hexane on water revises the interfacial water structure in a way that stabilizes the hyper-coordinated solvation structure and slows down the migration of hydroxide ion (OH−) relative to that in bulk water. This mechanism effectively attracts the OH− to the water-hydrophobic interface with respect to its counter-ion. The result illustrates the striking difference of proton transfer of hydrated OH− at the interface and in the bulk, which is responsible for the intrinsic charging effect at the hydrophobic interface.

URL: https://doi.org/10.1103/PhysRevLett.125.156803

86 PFLOPS Deep Potential Molecular Dynamics simulation of 100 million atoms with ab initio accuracy

D Lu, H Wang, M Chen, J Liu, L Lin, R Car, W. E, W Jia, L Zhang
Comp. Phys. Comm., vol 259, February 2021, 107624. (2021)

Abstract

We present the GPU version of DeePMD-kit, which, upon training a deep neural network model using ab initio data, can drive extremely large-scale molecular dynamics (MD) simulation with ab initio accuracy. Our tests show that the GPU version is 7 times faster than the CPU version with the same power consumption. The code can scale up to the entire Summit supercomputer. For a copper system of 113, 246, 208 atoms, the code can perform one nanosecond MD simulation per day, reaching a peak performance of 86 PFLOPS (43% of the peak). Such unprecedented ability to perform MD simulation with ab initio accuracy opens up the possibility of studying many important issues in materials and molecules, such as heterogeneous catalysis, electrochemical cells, irradiation damage, crack propagation, and biochemical reactions.

URL: https://doi.org/10.1016/j.cpc.2020.107624

Isotope effects in liquid water via deep potential molecular dynamics

Hsin-Yu Ko, Linfeng Zhang, Biswajit Santra, Han Wang, Weinan E, Robert A DiStasio Jr, Roberto Car
Molecular Physics, 11(22), 3269-3281 (2019)

Abstract

A comprehensive microscopic understanding of ambient liquid water is a major challenge for ab initio simulations as it simultaneously requires an accurate quantum mechanical description of the underlying potential energy surface (PES) as well as extensive sampling of configuration space. Due to the presence of light atoms (e.g. HH or DD), nuclear quantum fluctuations lead to observable changes in the structural properties of liquid water (e.g. isotope effects), and therefore provide yet another challenge for ab initio approaches. In this work, we demonstrate that the combination of dispersion-inclusive hybrid density functional theory (DFT), the Feynman discretised path-integral (PI) approach, and machine learning (ML) constitutes a versatile ab initio based framework that enables extensive sampling of both thermal and nuclear quantum fluctuations on a quite accurate underlying PES. In particular, we employ the recently developed deep potential molecular dynamics (DPMD) model – a neural-network representation of the ab initio PES – in conjunction with a PI approach based on the generalised Langevin equation (PIGLET) to investigate how isotope effects influence the structural properties of ambient liquid H2OH2O and D2OD2O. Through a detailed analysis of the interference differential cross sections as well as several radial and angular distribution functions, we demonstrate that this approach can furnish a semi-quantitative prediction of these subtle isotope effects.

URL: https://doi.org/10.1080/00268976.2019.1652366

Neural Canonical Transformation with Symplectic Flows

Shuo-Hui Li, Chen-Xiao Dong, Linfeng Zhang, and Lei Wang, Physical Review X 10, 021020 (2020)

Abstract

Canonical transformation plays a fundamental role in simplifying and solving classical Hamiltonian systems. Intriguingly, it has a natural correspondence to normalizing flows with a symplectic constraint. Building on this key insight, we design a neural canonical transformation approach to automatically identify independent slow collective variables in general physical systems and natural datasets. We present an efficient implementation of symplectic neural coordinate transformations and two ways to train the model based either on the Hamiltonian function or phase-space samples. The learned model maps physical variables onto an independent representation where collective modes with different frequencies are separated, which can be useful for various downstream tasks such as compression, prediction, control, and sampling. We demonstrate the ability of this method first by analyzing toy problems and then by applying it to real-world problems, such as identifying and interpolating slow collective modes of the alanine dipeptide molecule and MNIST database images.

URL: DOI: 10.1103/PhysRevX.10.021020

DP-GEN: A concurrent learning platform for the generation of reliable deep learning based potential energy models

Y. Zhang, H. Wang, W. Chen, J. Zeng, L. Zhang, H. Wang, W. E
Comp. Phys. Comm. (2020)

Abstract

In recent years, promising deep learning based interatomic potential energy surface (PES) models have been proposed that can potentially allow us to perform molecular dynamics simulations for large scale systems with quantum accuracy. However, making these models truly reliable and practically useful is still a very non-trivial task. A key component in this task is the generation of datasets used in model training. In this paper, we introduce the Deep Potential GENerator (DP-GEN), an open-source software platform that implements the recently proposed ”on-the-fly” learning procedure (Zhang et al. 2019) and is capable of generating uniformly accurate deep learning based PES models in a way that minimizes human intervention and the computational cost for data generation and model training. DP-GEN automatically and iteratively performs three steps: exploration, labeling, and training. It supports various popular packages for these three steps: LAMMPS for exploration, Quantum Espresso, VASP, CP2K, etc. for labeling, and DeePMD-kit for training. It also allows automatic job submission and result collection on different types of machines, such as high performance clusters and cloud machines, and is adaptive to different job management tools, including Slurm, PBS, and LSF. As a concrete example, we illustrate the details of the process for generating a general-purpose PES model for Cu using DP-GEN.

URL: https://www.sciencedirect.com/science/article/pii/S001046552030045X

Aqueous Solvation of the Chloride Ion Revisited with Density Functional Theory: Impact of Correlation and Exchange Approximations

Mark DelloStritto, Jianhang Xu, Xifan Wu and Michael L. Klein, Phys. Chem. Chem. Phys., 2020, 22, 10666-10675 (2020)

Abstract

The specificity of aqueous halide solvation is fundamental to a wide range of bulk and interfacial phenomena spanning from biology to materials science. Halide polarizability is thought to drive the ion specificity, and if so, it is essential to have an accurate description of the electronic properties of halide ions in water. To this end, the solvation of the chloride anion, Cl− has been reinvestigated with state-of-the-art density functional theory. Specifically, the PBE-D3, PBE0-D3, and SCAN functionals have been employed to probe the impact of correlation and exchange approximations. Anticipating the findings, adding exact exchange improves the electronic structure, but simultaneously significantly reduces the Cl− polarizability, resulting in an over-structured Cl–O radial distribution function (RDF) and longer water H-bond lifetimes to Cl−. SCAN does not yield as much improvement in the energetics of Cl− relative to bulk water, but does result in a smaller reduction of the polarizability and thus a less structured Cl–O RDF, which agrees better with experiment. Special consideration is therefore warranted in assessing the impact of exchange on the energy, charge density, and the charge density response when designing and testing hybrid functionals for aqueous halide solvation.

URL: https://doi.org/10.1039/C9CP06821J

Phase equilibrium of liquid water and hexagonal ice from enhanced sampling molecular dynamics simulations

Pablo M. Piaggi and Roberto Car, J. Chem. Phys. 152, 204116 (2020)

Abstract

We study the phase equilibrium between liquid water and ice Ih modeled by the TIP4P/Ice interatomic potential using enhanced sampling molecular dynamics simulations. Our approach is based on the calculation of ice Ih-liquid free energy differences from simulations that visit reversibly both phases. The reversible interconversion is achieved by introducing a static bias potential as a function of an order parameter. The order parameter was tailored to crystallize the hexagonal diamond structure of oxygen in ice Ih. We analyze the effect of the system size on the ice Ih-liquid free energy differences, and we obtain a melting temperature of 270 K in the thermodynamic limit. This result is in agreement with estimates from thermodynamic integration (272 K) and coexistence simulations (270 K). Since the order parameter does not include information about the coordinates of the protons, the spontaneously formed solid configurations contain proton disorder as expected for ice Ih.

URL: https://doi.org/10.1063/5.0011140

Raman spectrum and polarizability of liquid water from deep neural networks

Grace M. Sommers, Marcos F. Calegari Andrade, Linfeng Zhang, Han Wang and Roberto Car, R.
Phys. Chem. Chem. Phys. 2020, 22, 10592-10602 (2020)

Abstract

We introduce a scheme based on machine learning and deep neural networks to model the environmental dependence of the electronic polarizability in insulating materials. Application to liquid water shows that training the network with a relatively small number of molecular configurations is sufficient to predict the polarizability of arbitrary liquid configurations in close agreement with ab initio density functional theory calculations. In combination with a neural network representation of the interatomic potential energy surface, the scheme allows us to calculate the Raman spectra along 2-nanosecond classical trajectories at different temperatures for H2O and D2O. The vast gains in efficiency provided by the machine learning approach enable longer trajectories and larger system sizes relative to ab initio methods, reducing the statistical error and improving the resolution of the low-frequency Raman spectra. Decomposing the spectra into intramolecular and intermolecular contributions elucidates the mechanisms behind the temperature dependence of the low-frequency and stretch modes.

URL: DOI: 10.1039/D0CP01893G

Enabling Large-Scale Condensed Phase Hybrid Density Functional Theory Based Ab Initio Molecular Dynamics. I. Theory, Algorithm, and Performance

Hsin-Yu Ko, Junkeng Jia, Biswajit Santra, Xifan Wu, Roberto Car, and Robert A. DiStasio Jr., J. Chem. Theory Comput. 2020, 16, 6, 3757–3785 (2020)

Abstract

By including a fraction of exact exchange (EXX), hybrid functionals reduce the self-interaction error in semilocal density functional theory (DFT) and thereby furnish a more accurate and reliable description of the underlying electronic structure in systems throughout biology, chemistry, physics, and materials science. However, the high computational cost associated with the evaluation of all required EXX quantities has limited the applicability of hybrid DFT in the treatment of large molecules and complex condensed-phase materials. To overcome this limitation, we describe a linear-scaling approach that utilizes a local representation of the occupied orbitals (e.g., maximally localized Wannier functions (MLWFs)) to exploit the sparsity in the real-space evaluation of the quantum mechanical exchange interaction in finite-gap systems. In this work, we present a detailed description of the theoretical and algorithmic advances required to perform MLWF-based ab initio molecular dynamics (AIMD) simulations of large-scale condensed-phase systems of interest at the hybrid DFT level. We focus our theoretical discussion on the integration of this approach into the framework of Car-Parrinello AIMD, and highlight the central role played by the MLWF-product potential (i.e., the solution of Poisson’s equation for each corresponding MLWF-product density) in the evaluation of the EXX energy and wave function forces. We then provide a comprehensive description of the exx algorithm implemented in the open-source Quantum ESPRESSO program, which employs a hybrid MPI/OpenMP parallelization scheme to efficiently utilize the high-performance computing (HPC) resources available on current- and next-generation supercomputer architectures. This is followed by a critical assessment of the accuracy and parallel performance (e.g., strong and weak scaling) of this approach when AIMD simulations of liquid water are performed in the canonical (NVT) ensemble. With access to HPC resources, we demonstrate that exx enables hybrid DFT-based AIMD simulations of condensed-phase systems containing 500-1000 atoms (e.g., (H2O)256) with a wall time cost that is comparable to that of semilocal DFT. In doing so, exx takes us one step closer to routinely performing AIMD simulations of complex and large-scale condensed-phase systems for sufficiently long time scales at the hybrid DFT level of theory.

URL: https://dx.doi.org/10.1021/acs.jctc.9b01167

Free energy of proton transfer at the water–TiO2 interface from ab initio deep potential molecular dynamics

Marcos F. Calegari Andrade, Hsin-Yu Ko, Linfeng Zhan, Roberto Car, Annabella Selloni, Chemical Science, 11(9), 2335–2341 (2020)

Abstract

TiO2 is a widely used photocatalyst in science and technology and its interface with water is important in fields ranging from geochemistry to biomedicine. Yet, it is still unclear whether water adsorbs in molecular or dissociated form on TiO2 even for the case of well-defined crystalline surfaces. To address this issue, we simulated the TiO2–water interface using molecular dynamics with an ab initio-based deep neural network potential. Our simulations show a dynamical equilibrium of molecular and dissociative adsorption of water on TiO2. Water dissociates through a solvent-assisted concerted proton transfer to form a pair of short-lived hydroxyl groups on the TiO2 surface. Molecular adsorption of water is ΔF = 8.0 ± 0.9 kJ mol−1 lower in free energy than the dissociative adsorption, giving rise to a 5.6 ± 0.5% equilibrium water dissociation fraction at room temperature. Due to the relevance of surface hydroxyl groups to the surface chemistry of TiO2, our model might be key to understanding phenomena ranging from surface functionalization to photocatalytic mechanisms.

URL: https://doi.org/10.1039/c9sc05116c

Deep neural network for the dielectric response of insulators

Linfeng Zhang, Mohan Chen, Xifan Wu, Han Wang and Roberto Car, Phys. Rev. B, 102, 04112 (R) (2020)

Abstract

We introduce a deep neural network to model in a symmetry preserving way the environmental dependence of the centers of the electronic charge. The model learns from {} density functional theory, wherein the electronic centers are uniquely assigned by the maximally localized Wannier functions. When combined with the Deep Potential model of the atomic potential energy surface, the scheme predicts the dielectric response of insulators for trajectories inaccessible to direct {} simulation. The scheme is non-perturbative and can capture the response of a mutating chemical environment. We demonstrate the approach by calculating the infrared spectra of liquid water at standard conditions, and of ice under extreme pressure, when it transforms from a molecular to an ionic crystal.

URL: https://doi.org/10.1103/PhysRevB.102.041121

Learning from the Density to Correct Total Energy and Forces in First Principle Simulations

Sebastian Dick and Marivi Fernandez-Serra, J. Chem. Phys. 151, 144102 (2019)

Abstract

We propose a new molecular simulation framework that combines the transferability, robustness and chemical flexibility of an ab initio method with the accuracy and efficiency of a machine learned force field. The key to achieve this mix is to use a standard density functional theory (DFT) simulation as a pre-processor for the atomic and molecular information, obtaining a good quality electronic density. General, symmetry preserving, atom-centered electronic descriptors are then built from this density to train a neural network to correct the baseline DFT energies and forces. These electronic descriptors encode much more information than local atomic environments, allowing a simple neural network to reach the accuracy required for the problem of study at a negligible cost. The balance between accuracy and efficiency is determined by the baseline simulation. This is shown in results where high level quantum chemical accuracy is obtained for simulations of liquid water at standard DFT cost, or where high level DFT-accuracy is achieved in simulations with a low-level baseline DFT calculation, at a significantly reduced cost.

URL https://doi.org/10.1063/1.5114618

Effect of Functional and Electron Correlation on the Structure and Spectroscopy of the Al2O3(001)-H2O Interface

Mark J. DelloStritto, Stephan M. Piontek, Michael L. Klein, and Eric Borguet, J. Phys. Chem. Lett., 2019, 10, 9, 2031–2036 (2019)

Abstract

Oxide−water interfaces are ubiquitous, with many applications in industry and the environment, yet there is a great deal of controversy over their properties and microscopic structure. This controversy stems, in part, from the unique H-bond networks formed at different surface terminations and mineral compositions. Density functional theory simulations of these interfaces require an accurate description of both the oxide mineral and water in diverse H-bond environments. Thus, herein we simulate the Al2O3(001)−H2O interface using the PBE, PBE-TS, RPBE, SCAN, and HSE06-TS functionals to determine how calculated interfacial properties depend on the choice of functional. We find that the structure of the first few layers of water at the surface is determined by electron correlation in a way that cannot be approximated using semiemipirical van der Waals corrections. Of the functionals investigated, we find that SCAN yields the most accurate interfacial structure, dynamics, and sum frequency generation spectrum. Furthermore, SCAN leads to a reduction in the order of the 2D H-bond network of water at the alumina surface predicted by GGA functionals, leading to a significant decrease in the anisotropy of the diffusion coefficient at the surface. We emphasize the importance of using a functional which accurately describes electron correlation for more complex oxides, such as transition-metal oxides, where electron correlation may play an even greater role in determining the structure and dynamics of the oxide−water interface.

URL: https://doi.org/10.1021/acs.jpclett.9b00016

Active learning of uniformly accurate interatomic potentials for materials simulation

Linfeng Zhang, De-Ye Lin, Han Wang, Roberto Car, Weinan E., Phys. Rev. Mat. 3, 023804 (2019)

Abstract

An active learning procedure called deep potential generator (DP-GEN) is proposed for the construction of accurate and transferable machine learning-based models of the potential energy surface (PES) for the molecular modeling of materials. This procedure consists of three main components: exploration, generation of accurate reference data, and training. Application to the sample systems of Al, Mg, and Al-Mg alloys demonstrates that DP-GEN can produce uniformly accurate PES models with a minimal number of reference data.

URL: https://doi.org/10.1103/PhysRevMaterials.3.023804

Importance of van der Waals effects on the hydration of metal ions from the Hofmeister series

Liying Zhou, Jianhang Xu, Limei Xu and Xifan Wu, J. Chem. Phys. 150, 124505 (2019)

Abstract

The van der Waals (vdW) interaction plays a crucial role in the description of liquid water. Based on ab initio molecular dynamics simulations, including the non-local and fully self-consistent density-dependent implementation of the Tkatchenko-Scheffler dispersion correction, we systematically studied the aqueous solutions of metal ions (K+, Na+, and Ca2+) from the Hofmeister series. Similar to liquid water, the vdW interactions strengthen the attractions among water molecules in the long-range, leading to the hydrogen bond networks softened in all the ion solutions. However, the degree that the hydration structure is revised by the vdW interactions is distinct for different ions, depending on the strength of short-range interactions between the hydrated ion and surrounding water molecules. Such revisions by the vdW interactions are important for the understanding of biological functionalities of ion channels.

URL: https://doi.org/10.1063/1.5086939

Structure, Polarization, and Sum Frequency Generation Spectrum of Interfacial Water on Anatase TiO2

Marcos F. Calegari Andrade, Hsin-Yu Ko, Roberto Car, Annabella Selloni, J. Chem. Phys. Lett. 9, 23, 6716 – 6721 (2018)

Abstract

Amorphous TiO2 (a-TiO2) is widely used in many fields, ranging from photoelectrochemistry to bioengineering, hence detailed knowledge of its atomic structure is of scientific and technological interest. Here we use an ab initio-based deep neural network potential (DP) to simulate large scale atomic models of crystalline and disordered TiO2 with molecular dynamics. Our DP reproduces the structural properties of all 11 TiO2 crystalline phases, predicts the densities and structure factors of molten and amorphous TiO2 with only a few percent deviation from experiments, and describes the pressure dependence of the amorphous structure in agreement with recent observations. It can be extended to model additional structures and compositions and can be thus of great value in the study of TiO2-based nanomaterials.

URL: https://doi.org/10.1021/acs.jpclett.8b03103