Molecular Dynamics Simulations Of Liquid Research Articles

Full molecular dynamics simulations of molecular liquids for single-beam spectrally controlled two-dimensional Raman spectroscopy.

Single-beam spectrally controlled (SBSC) two-dimensional (2D) Raman spectroscopy is a unique 2D vibrational measurement technique utilizing trains of short pulses that are generated from a single broadband pulse by pulse shaping. This approach overcomes the difficulty of 2D Raman spectroscopy in dealing with small-signal extraction and avoids complicated low-order cascading effects, thus providing a new possibility for measuring the intramolecular and intermolecular modes of molecular liquids using fifth-order 2D Raman spectroscopy. Recently, for quantitatively investigating the mode-mode coupling mechanism, Hurwitz et al. [Opt. Express 28, 3803 (2020)] have developed a new pulse design for this measurement to separate the contributions of the fifth- and third-order polarizations, which are often overlapped in the original single-beam measurements. Here, we describe a method for simulating these original measurements and the new 2D Raman measurements on the basis of a second-order response function approach. We carry out full molecular dynamics simulations for carbon tetrachloride and liquid water using an equilibrium-nonequilibrium hybrid algorithm, with the aim of explaining the key features of the SBSC 2D Raman spectroscopic method from a theoretical point of view. The predicted signal profiles and intensities provide valuable information that can be applied to 2D spectroscopy experiments, allowing them to be carried out more efficiently.

Structure and Dynamics of Supercooled Liquid Ge2Sb2Te5 from Machine‐Learning‐Driven Simulations

Studies of supercooled liquid phase‐change materials are important for the development of phase‐change memory and neuromorphic computing devices. Herein, a machine‐learning (ML)‐based interatomic potential for Ge2Sb2Te5 (GST) to conduct large‐scale molecular dynamics simulations of liquid and supercooled liquid GST is used. A pronounced effect of the thermostat parameters on the simulation results is demonstrated, and it is shown how using a Langevin thermostat with optimized damping values can lead to excellent agreement with reference ab initio molecular dynamics (AIMD) simulations. Structural and dynamical analyses are presented, including the studies of radial and angular distributions, homopolar bonds, and the temperature‐dependent diffusivity. Herein, the usefulness of ML‐driven molecular dynamics for further studies of supercooled liquid GST, with length and timescales far exceeding those that are accessible to AIMD is demonstrated.

Density functional and classical simulations of liquid and glassy selenium

Molecular dynamics simulations of liquid and glassy selenium have been carried out using density functional (400--773 K, 600 atoms) and classical force field (290--500 K, 5488 atoms) methods. Structural features (structure factors, pair distribution functions, bond lengths, bond and dihedral angles, cavities) and dynamical properties (diffusion coefficients, power spectra, sound velocity, collective excitations, bond lifetimes) agree well with experimental data where available. The structures are predominantly chainlike, with a small fraction of rings with a range of sizes, and large cavity volumes lead to flexible chains. It is striking that the density functional simulations show very few ${\mathrm{Se}}_{8}$ rings at 600 K and below.

Accurate calculation of zero point energy from molecular dynamics simulations of liquids and their mixtures.

The two-phase thermodynamic (2PT) method is used to compute the zero point energy (ZPE) of several liquids and their mixtures. The 2PT method uses the density of states (DoS), which is computed from the velocity autocorrelation (VAC) function obtained from a short classical molecular dynamics trajectory. By partitioning the VAC and the DoS of a fluid into solid and gaslike components, quantum mechanical corrections to thermodynamical properties can be computed. The ZPE is obtained by combining the partition function of the quantum harmonic oscillator with the vibrational part of the solidlike DoS. The resulting ZPE is found to be in excellent agreement with both experimental and ab initio results. Solvent effects such as hydrogen bonding and polarization can be included by the utilization of ab initio density functional theory based molecular dynamics simulations. It is found that these effects significantly influence the DoS of water molecules. The obtained results demonstrate that the 2PT model is a powerful method for efficient ZPE calculations, in particular, to account for solvent effects and polarization.

Relationship between Soret coefficient and potential energy distribution in silicate liquids: A molecular dynamics study

We present the results of force field molecular dynamics simulations of calcium aluminosilicate, sodium aluminosilicate, calcium silicate, and sodium silicate liquids under temperature gradient conditions. Non-equilibrium molecular dynamics simulations of these liquids under temperature gradient conditions reproduced the trend reported for Soret coefficients, with network-forming cations having larger coefficients compared with network-modifying ones. We found a linear relationship between the Soret coefficients and the variance of the potential energy distribution of each cation, which corresponds to the activation energy of diffusion. The Soret coefficients of sodium aluminosilicate liquids show a somewhat different behavior compared with that of the other liquids. This behavior could be explained by charge compensation of fourfold-coordinated aluminum species by sodium cations. Further studies of a larger variety of systems will reveal the full details of the relationship between Soret coefficients and microscopic structure/energy distribution, and enable us to estimate the Soret coefficients.

LiquidLib: A comprehensive toolbox for analyzing classical and ab initio molecular dynamics simulations of liquids and liquid-like matter with applications to neutron scattering experiments

Neutron scattering is a powerful experimental technique for characterizing the structure and dynamics of materials on the atomic or molecular scale. However, the interpretation of experimental data from neutron scattering is oftentimes not trivial, partly because scattering methods probe ensemble-averaged information in the reciprocal space. Therefore, computer simulations, such as classical and ab initio molecular dynamics, are frequently used to unravel the time-dependent atomistic configurations that can reproduce the scattering patterns and thus assist in the understanding of the microscopic origin of certain properties of materials. LiquidLib is a post-processing package for analyzing the trajectory of atomistic simulations of liquids and liquid-like matter with application to neutron scattering experiments. From an atomistic simulation, LiquidLib provides the computation of various statistical quantities including the pair distribution function, the weighted and unweighted structure factors, the mean squared displacement, the non-Gaussian parameter, the four-point correlation function, the velocity auto correlation function, the self and collective van Hove correlation functions, the self and collective intermediate scattering functions, and the bond orientational order parameter. LiquidLib analyzes atomistic trajectories generated from packages such as LAMMPS, GROMACS, and VASP. It also offers an extendable platform to conveniently integrate new quantities into the library and integrate simulation trajectories of other file formats for analysis. Weighting the quantities by element-specific neutron-scattering lengths provides results directly comparable to neutron scattering measurements. Lastly, LiquidLib is independent of dimensionality, which allows analysis of trajectories in two, three, and higher dimensions. The code is beginning to find worldwide use. Program summaryProgram Title: LiquidLibProgram Files doi:http://dx.doi.org/10.17632/tyggwp7656.1Licensing provisions: GNU Lesser General Public License v3.0Programming language: C++Supplementary material: User manual of the package.External routines/libraries: BOOST (optional), GSL (optional), xdrfilelib (optional, available at ftp.gromacs.org/pub/contrib/xdrfile-1.1.1.tar.gz)Nature of problem: LiquidLib is a post-processing package for analyzing the trajectory of atomistic simulations of liquids and liquid-like matter with application to neutron scattering experiments.Solution method: We provide a collection of executables to compute statistical quantities relevant to the structure and dynamics of liquids and liquid-like matter from the simulation trajectories. Furthermore, the option to weigh the quantities by element-specific neutron-scattering lengths provides results comparable to neutron scattering measurements. The quantities computable by LiquidLib include the pair distribution function, the weighted and unweighted structure factors, the mean squared displacement, the non-Gaussian parameter, the four-point correlation function, the velocity auto correlation function, the self and collective van Hove correlation functions, the self and collective intermediate scattering functions, and the bond orientational order parameter. LiquidLib can directly read trajectories of file types .xyz, .xtc, .trr, .atom, .dump, .gro, and XDATCAR.Additional comments: Each computable quantity in LiquidLib is defined as a derived class, and has its own respective executable. Thus, LiquidLib is extendable to compute other quantities not already offered in the current version. Similarly, LiquidLib is extendable to handle simulation trajectory files with formats not already offered in the current version.Restrictions: LiquidLib reads in the complete simulation trajectory into the memory at the start of runtime. Thus, if the trajectory size is larger than the available memory of the computer, LiquidLib will fail to execute properly. In this case, the user must reduce the size of the trajectory to be read in before LiquidLib can analyze the trajectory successfully.Unusual features: We provide the option to weigh the computed quantities by the element-specific respective coherent or incoherent scattering lengths. Therefore, LiquidLib allows direct comparisons of the computed quantities to neutron scattering experiments. In addition, the dimension of the simulation trajectories can be any positive integer.

Benchmarking and Optimizing Atomistic Forcefields with Density Measurements

Predictive biophysical models of drug binding and selectivity could accelerate drug discovery. However, a limited understanding of molecular-scale phenomena currently prevents biophysical models from impacting drug design. To assess the quality of the GAFF small molecule forcefield, we have performed nearly 1000 molecular dynamics simulations of liquids containing drug-like atom types. We compare our simulations to an extensive database of physical properties from NIST's Thermodynamics Research Center. We also evaluate GAFF's performance using density measurements of neat liquids and binary mixtures performed on an in-house automation system.

A unified framework for force-based and energy-based adaptive resolution simulations

Adaptive resolution schemes enable molecular dynamics simulations of liquids and soft matter employing two different resolution levels concurrently in the same setup. These methods are based on a position-dependent interpolation of either forces or potential energy functions. While force-based methods generally lead to non-conservative forces, energy-based ones include undesired force terms proportional to the gradient of the interpolation function. In this work we establish a so far missing bridge between these formalisms making use of the generalized Langevin equation, thereby providing a unifying framework to traditionally juxtaposed approaches to adaptive simulations.

Dynamic properties of liquid and undercooled aluminum

First-principles molecular dynamics simulations of liquid and undercooled aluminum have been performed to study the evolution of dynamic properties across the melting point. Single-atom as well collective dynamic properties are determined and are related to the structural evolution of the liquid phase. The temperature dependence of the self-diffusion coefficient is computed from both the mean square displacement and the velocity autocorrelation function. Self-diffusion coefficients follow an Arrhenius law with the single activation energy of 250 meV, consistent with experimental results for the liquid phase. Moreover, we show that the Arrhenius relation can also be applied to the undercooled state at T = 875 K. Finally, a direct calculation of the shear viscosity from the transverse current correlation function shows that the Stokes–Einstein relation can be applied over the temperature range investigated.

Ab initio investigation of a possible liquid–liquid phase transition in MgSiO3 at megabar pressures

We perform density functional molecular dynamics simulations of liquid and solid MgSiO3 in the pressure range of 120–1600 GPa and for temperatures up to 20,000 K in order to provide new insight into the nature of the liquid–liquid phase transition that was recently predicted on the basis of decaying laser shock wave experiments [Phys. Rev. Lett. 108 (2012) 065701]. However, our simulations did not show any signature of a phase transition in the liquid phase. We derive the equation of state for the liquid and solid phases and compute the shock Hugoniot curves. We discuss different thermodynamic functions and by explore alternative interpretations of the experimental findings.

Bayesian uncertainty quantification and propagation in molecular dynamics simulations: A high performance computing framework

We present a Bayesian probabilistic framework for quantifying and propagating the uncertainties in the parameters of force fields employed in molecular dynamics (MD) simulations. We propose a highly parallel implementation of the transitional Markov chain Monte Carlo for populating the posterior probability distribution of the MD force-field parameters. Efficient scheduling algorithms are proposed to handle the MD model runs and to distribute the computations in clusters with heterogeneous architectures. Furthermore, adaptive surrogate models are proposed in order to reduce the computational cost associated with the large number of MD model runs. The effectiveness and computational efficiency of the proposed Bayesian framework is demonstrated in MD simulations of liquid and gaseous argon.

Simulating liquid and amorphous silicon dioxide using real-space pseudopotentials

We present ab initio molecular dynamics simulations of liquid and amorphous silicon dioxide. The interatomic forces in our simulations are calculated using real-space pseudopotentials, which were constructed using density-functional theory. Our simulations are carried out using Born-Oppenheimer molecular dynamics (i.e., the electronic structure problem is solved by performing fully self-consistent calculations for each time step). Using a subspace filtering iteration technique, we avoid solving the Kohn-Sham eigenvalue with ``standard'' diagonalization methods. We consider systems with up to 192 atoms (64 SiO${}_{2}$ units) in a periodic supercell for simulations over 20 ps. The liquid and amorphous ensembles are formed by thermally quenching random configurations of silicon and oxygen atoms. We compare our liquid and amorphous simulations with previously performed Car-Parrinello molecular dynamic simulations and with experiment. In particular, we examine the possible formation of two-membered rings, which were not observed in previous simulations using quantum forces. We attribute this difference to a ``biased'' initial configuration, which inhibits the formation of two-membered rings. We also compare the structural properties of our simulated amorphous systems with neutron diffraction measurements and find good agreement.

Density-scaling and the Prigogine–Defay ratio in liquids

The term "strongly correlating liquids" refers to materials exhibiting near proportionality of fluctuations in the potential energy and the virial pressure, as seen in molecular dynamics simulations of liquids whose interactions are comprised primarily of van der Waals forces. Recently it was proposed that the Prigogine-Defay ratio, Π, of strongly correlating liquids should fall close to unity. We verify this prediction herein by showing that the degree to which relaxation times are a function T/ρ(γ), the ratio of temperature to density with the latter raised to a material constant (a property inherent to strongly correlating liquids) is reflected in values of Π closer to unity. We also show that the dynamics of strongly correlating liquids are governed more by density than by temperature. Thus, while Π may never strictly equal 1 for the glass transition, it is approximately unity for many materials, and thus can serve as a predictor of other dynamic behavior. For example, Π ≫ 1 is indicative of additional control parameters besides T/ρ(γ).

Thermal conduction in molecular materials using coarse grain dynamics: Role of mass diffusion and quantum corrections for molecular dynamics simulations

We use a mesodynamical method, denoted dynamics with implicit degrees of freedom (DID), to characterize thermal transport in a model molecular crystal below and above its melting temperature. DID represents groups of atoms (molecules in this case) using mesoparticles and the thermal role of the intramolecular degrees of freedom (DoFs) are described implicitly using their specific heat. We focus on the role of these intramolecular DoFs on thermal transport. We find that thermal conductivity is independent of intramolecular specific heat for solid samples and a linear relationship between the two quantities in liquid samples with the coefficient of proportionality being the mass diffusivity of the mesoparticles. As the temperature of the liquids is increased, thermal conductivity exhibits an increased sensitivity with respect to the specific heat of the internal DoFs due to the enhanced molecular mobility. Based on these results, we propose a simple method to incorporate quantum corrections to thermal conductivity obtained from nonequilibrium molecular dynamics simulations of molecular liquids. Our results also provide insight into the development of thermally accurate coarse grain models of soft materials.

Dynamics of liquid and undercooled silicon: Anab initiomolecular dynamics study

First-principles molecular dynamics simulations of liquid and undercooled silicon have been performed in order to study the evolution of dynamic properties across the melting point. The calculated dynamic structure factors show collective excitations at small wave vectors above the melting point, in good agreement with recent experimental data, while the transverse current correlation functions do not display propagating shear modes. Upon undercooling, our results evidence the occurrence of well-defined shear modes that result from an enhancement of local tetrahedral arrangement associated to strong covalent bonding suggesting that propagating shear waves might be sustained by a concerted motion of tetrahedra.

Development of suitable interatomic potentials for simulation of liquid and amorphous Cu–Zr alloys

We present a new semi-empirical potential suitable for molecular dynamics simulations of liquid and amorphous Cu–Zr alloys. To provide input data for developing the potential, new experimental measurements of the structure factors for amorphous Cu64.5Zr35.5 alloy were performed. In this work, we propose a new method to include diffraction data in the potential development procedure, which also includes fitting to first-principles and liquid density and enthalpy of mixing data. To refine the new potential, we used first-principles and liquid enthalpy of mixing data published earlier combined with the densities of liquid Cu64.5Zr35.5 measured over a range of temperatures. We show that the potential predicts a liquid-to-glass transition temperature that agrees reasonably well with experimental data. Finally, we compare the new potential with two previously developed semi-empirical potentials for Cu–Zr alloys and examine their comparative and contrasting descriptions of structure and properties for Cu64.5Zr35.5 liquids and glasses.

Properties and 3D Structure of Liquid Water: A Perspective from a High-Rank Multipolar Electrostatic Potential.

We propose a new rigid, nonpolarizable high-rank multipolar potential for the simulation of liquid water. The electrostatic interaction is represented by spherical tensor multipole moments on oxygen and hydrogen, up to hexadecupole. The Quantum Chemical Topology (QCT) method yields the atomic multipole moments from a MP2/aug-cc-p-VTZ electron density of a single water molecule in the gas phase. These moments reproduce the experimental molecular dipole and quadrupole moment within less than 1%. Given its high-rank multipole moments, used in conjunction with a consistent high-rank multipolar Ewald summation, the QCT potential is ideal to assess the performance of exhaustive "gas phase" electrostatics in molecular dynamics simulations of liquids. The current article explores the performance of this potential at 17 temperatures between -35 °C (238 K) and 90 °C (363 K) and at 7 pressures between 1 and 10 000 atm. The well-known maximum in the liquid's density at 4 °C is reproduced at 6 °C. Six bulk properties are calculated and found to deviate from experiment in a homogeneous manner, that is, without serious outliers, compared to several other potentials. Spatial distribution functions (i.e., gOO(r,Ω)) and the (more common) radial distribution functions are used to analyze the local water structure. At the lone pair side of a central water, neighboring waters form a continuous horseshoe-like distribution, with substantial narrowing in the central part. The latter feature is unique to the QCT potential. Under high pressure, the local structure undergoes dramatic rearrangement and results in the collapse of second shell neighbors into the interstitial region of the first shell, which is in close agreement with experiment. Our results also corroborate the suggestion that the local hydrogen-bonded network remains largely intact even under such conditions.

Efficiency of Rotational Operators for Geometric Manipulation of Chain Molecules

Geometric manipulation of molecules is an essential elementary component in computational modeling programs for molecular structure, stability, dynamics, and design. The computational complexity of transformation of internal coordinates to Cartesian coordinates was discussed before.1 The use of rotation matrices was found to be slightly more efficient than that of quaternion although quaternion operators have been widely advertised for rotational operations, especially in molecular dynamics simulations of liquids where the orientation is a dynamical variable.2 The discussion on computational efficiency is extended here to a more general case in which bond angles and sidechain torsion angles are allowed to vary. The algorithm of Thompson3 is derived again in terms of quaternions as well as rotation matrices, and an algorithm with optimal efficiency is described. The algorithm based on rotation matrices is again found to be slightly more efficient than that based on quaternions.

Modeling the structural, dynamical, and magnetic properties of liquidAl1−xMnx(x=0.14, 0.2, and 0.4): A first-principles investigation

We report the results of first-principles molecular dynamics simulations of liquid ${\mathrm{Al}}_{1\ensuremath{-}x}{\mathrm{Mn}}_{x}$ alloys at three different compositions. The local structure as defined by the Bhatia-Thornton partial structure factors is found to display significant changes at $x=0.4$. In addition, a structural analysis using three-dimensional pair-analysis techniques evidences a fivefold symmetry around $x=0.14$, in agreement with the experimental quasicrystal-forming range, and an increasing complexity of the Frank-Kasper polytetrahedral symmetry around Mn atoms at $x=0.4$. We also examine the time evolution of the configurations at the three compositions in terms of the mean-square displacements and self-diffusion coefficients. Finally, we show a strong interplay between the structural changes and the evolution of the magnetic properties of the Mn atoms as a function of composition.

Ab initiomolecular dynamics simulations on structural change of liquid eutectic alloyGe15Te85from573to1073K

We have performed ab initio molecular dynamics simulations of liquid ${\mathrm{Ge}}_{15}{\mathrm{Te}}_{85}$ alloy at seven different temperatures from $573\phantom{\rule{0.3em}{0ex}}\text{to}\phantom{\rule{0.3em}{0ex}}1073\phantom{\rule{0.3em}{0ex}}\mathrm{K}$. The structural and electronic properties of liquid ${\mathrm{Ge}}_{15}{\mathrm{Te}}_{85}$ alloy and their temperature dependence were studied. Comparing with available experimental data, the calculated pair-correlation functions are acceptable. In the low-temperature region from $573\phantom{\rule{0.3em}{0ex}}\text{to}\phantom{\rule{0.3em}{0ex}}733\phantom{\rule{0.3em}{0ex}}\mathrm{K}$, the total coordination number ${N}_{\mathit{\text{Total}}}$ increases gradually with increasing temperature in contrast to the behavior of a classical isotropic fluid; but in the high-temperature region from $833\phantom{\rule{0.3em}{0ex}}\text{to}\phantom{\rule{0.3em}{0ex}}1073\phantom{\rule{0.3em}{0ex}}\mathrm{K}$, it nearly remains a constant. The partial coordination number ${N}_{\mathrm{Ge}\mathrm{Te}}$ nearly remains a constant 4.7 in the whole temperature region from $573\phantom{\rule{0.3em}{0ex}}\text{to}\phantom{\rule{0.3em}{0ex}}1073\phantom{\rule{0.3em}{0ex}}\mathrm{K}$. The partial coordination number ${N}_{\mathrm{Te}\mathrm{Te}}$ first increases in the low-temperature region with increasing temperature and then nearly keeps a constant in the high-temperature region, showing the same temperature dependence as ${N}_{\mathit{\text{Total}}}$. The calculated angular limited triplet correlation functions of Te-Te-Te and Te-Ge-Te suggest that at low temperatures there exist Peierls-type distorted local structure around both Ge and Te atoms, i.e., the short-long and long-short bonds correlation, however, it is no more true at high temperatures. At low temperatures $(573--733\phantom{\rule{0.3em}{0ex}}\mathrm{K})$ the Te $p$ states form an obvious dip at Fermi level ${E}_{F}$; but at high temperatures $(833--1073\phantom{\rule{0.3em}{0ex}}\mathrm{K})$ this dip disappears, approaching the free-electron form. Our results suggest that besides the contribution from the modification of the local environment around Ge atoms with the temperature the unexpected temperature dependence of the local environment around Te atoms is also responsible for the structural change of liquid ${\mathrm{Ge}}_{15}{\mathrm{Te}}_{85}$ characterized by anomalous temperature behaviors of thermodynamic quantities.