Fluctuating hydrodynamics for multiscale modeling and simulation: Energy and heat transfer in molecular fluids

This work illustrates that fluctuating hydrodynamics (FHD) simulations can be used to capture the thermodynamic and hydrodynamic responses of molecular fluids at the nanoscale, including those associated with energy and heat transfer. Using all-atom molecular dynamics (MD) trajectories as the reference data, the atomistic coordinates of each snapshot are mapped onto mass, momentum, and energy density fields on Eulerian grids to generate a corresponding field trajectory. The molecular length-scale associated with finite molecule size is explicitly imposed during this coarse-graining by requiring that the variances of density fields scale inversely with the grid volume. From the fluctuations of field variables, the response functions and transport coefficients encoded in the all-atom MD trajectory are computed. By using the extracted fluid properties in FHD simulations, we show that the fluctuations and relaxation of hydrodynamic fields quantitatively match with those observed in the reference all-atom MD trajectory, hence establishing compatibility between the atomistic and field representations. We also show that inclusion of energy transfer in the FHD equations can more accurately capture the thermodynamic and hydrodynamic responses of molecular fluids. The results indicate that the proposed MD-to-FHD mapping with explicit consideration of finite molecule size provides a robust framework for coarse-graining the solution phase of complex molecular systems.     

A multiscale coarse-graining method for biomolecular systems

A new approach is presented for obtaining coarse-grained (CG) force fields from fully atomistic molecular dynamics (MD) trajectories. The method is demonstrated by applying it to derive a CG model for the dimyristoylphosphatidylcholine (DMPC) lipid bilayer. The coarse-graining of the interparticle force field is accomplished by an application of a force-matching procedure to the force data obtained from an explicit atomistic MD simulation of the biomolecular system of interest. Hence, the method is termed a "multiscale" CG (MS-CG) approach in which explicit atomistic-level forces are propagated upward in scale to the coarse-grained level. The CG sites in the lipid bilayer application were associated with the centers-of-mass of atomic groups because of the simplicity in the evaluation of the forces acting on them from the atomistic data. The resulting CG lipid bilayer model is shown to accurately reproduce the structural properties of the phospholipid bilayer.     

Multiscale coarse-graining of monosaccharides

A systematic multiscale coarse-graining (MS-CG) algorithm is applied to build coarse-grained models for monosaccharides in aqueous solution. The methodology is demonstrated for the example of alpha-D-glucopyranose. The nonbonded interactions are directly derived from the force-matching approach, whereas the bonded interactions are obtained through Boltzmann statistical analyses of the underlying atomistic trajectory. The MS-CG model is shown to reproduce many structural and thermodynamic properties in the constant NPT ensemble. Although the model is derived at a single temperature, pressure, and concentration, it is shown to be reasonably transferable to other thermodynamic states. In this model, long-range interactions are effectively mapped into short-range forces with a moderate cutoff and are evaluated by table look-up. As a result, molecular dynamics employing the MS-CG model is approximately 3 orders of magnitude more efficient than its atomistic counterpart. Consequently, the model is particularly suitable for simulating carbohydrate systems at large length and long time scales. Results for an alpha-(1-->4)-d-glucan with 14 glucose units are also presented, demonstrating that the MS-CG algorithm is also applicable to the coarse-graining of other saccharide systems.     

The multiscale coarse-graining method. VII. Free energy decomposition of coarse-grained effective potentials

The potential of mean force (PMF) with respect to coarse-grained (CG) coordinates is often calculated in order to study the molecular interactions in atomistic molecular dynamics (MD) simulations. The multiscale coarse-graining (MS-CG) approach enables the computation of the many-body PMF of an atomistic system in terms of the CG coordinates, which can be used to parameterize CG models based on all-atom configurations. We demonstrate here that the MS-CG method can also be used to analyze the CG interactions from atomistic MD trajectories via PMF calculations. In addition, MS-CG calculations at different temperatures are performed to decompose the PMF values into energetic and entropic contributions as a function of the CG coordinates, which provides more thermodynamic information regarding the atomistic system. Two numerical examples, liquid methanol and a dimyristoylphosphatidylcholine lipid bilayer, are presented. The results show that MS-CG can be used as an analysis tool, comparable to various free energy computation methods. The differences between the MS-CG approach and other PMF calculation methods, as well as the characteristics and advantages of MS-CG, are also discussed.     

Systematic coarse-graining of nanoparticle interactions in molecular dynamics simulation

A recently developed multiscale coarse-graining procedure [Izvekov, S.; Voth, G. A. J. Phys. Chem. B 2005, 109, 2469] is extended to derive coarse-grained models for nanoparticles. The methodology is applied to C(60) and to carbonaceous nanoparticles produced in combustion environments. The coarse-graining of the interparticle force field is accomplished applying a force-matching procedure to data obtained from trajectories and forces from all-atom MD simulations. The CG models are shown to reproduce accurately the structural properties of the nanoparticle systems studied, while allowing for MD simulations of much larger self-assembled nanoparticle systems.     

Molecular latent space simulators

Small integration time steps limit molecular dynamics (MD) simulations to millisecond time scales. Markov state models (MSMs) and equation-free approaches learn low-dimensional kinetic models from MD simulation data by performing configurational or dynamical coarse-graining of the state space. The learned kinetic models enable the efficient generation of dynamical trajectories over vastly longer time scales than are accessible by MD, but the discretization of configurational space and/or absence of a means to reconstruct molecular configurations precludes the generation of continuous atomistic molecular trajectories. We propose latent space simulators (LSS) to learn kinetic models for continuous atomistic simulation trajectories by training three deep learning networks to (i) learn the slow collective variables of the molecular system, (ii) propagate the system dynamics within this slow latent space, and (iii) generatively reconstruct molecular configurations. We demonstrate the approach in an application to Trp-cage miniprotein to produce novel ultra-long synthetic folding trajectories that accurately reproduce atomistic molecular structure, thermodynamics, and kinetics at six orders of magnitude lower cost than MD. The dramatically lower cost of trajectory generation enables greatly improved sampling and greatly reduced statistical uncertainties in estimated thermodynamic averages and kinetic rates.

Latent space simulators learn kinetic models for atomistic simulations and generate novel trajectories at six orders of magnitude lower cost.     

Toward a density functional description of liquid pH2

A finite-temperature density functional approach to describe the properties of parahydrogen in the liquid-vapor coexistence region is presented. The first proposed functional is zero-range, where the density-gradient term is adjusted so as to reproduce the surface tension of the liquid-vapor interface at low temperature. The second functional is finite-range and, while it is fitted to reproduce bulk pH(2) properties only, it is shown to yield surface properties in good agreement with experiments. These functionals are used to study the surface thickness of the liquid-vapor interface, the wetting transition of parahydrogen on a planar Rb model surface, and homogeneous cavitation in bulk liquid pH(2).     

Structure, dynamics, and the free energy of solute adsorption at liquid-vapor interfaces of simple dipolar systems: molecular dynamics results for pure and mixed Stockmayer fluids

We have performed molecular dynamics simulation studies of the structural, thermodynamic, and dynamical properties of liquid-vapor interfaces of pure and binary Stockmayer fluids of different polarity. The density profiles, the width of the liquid-vapor interface, and the orientational structure of the interfaces are calculated to characterize the structural aspects of the interfaces. Among the thermodynamic properties, we have computed the surface tension and also the free energy of transfer of a charged solute across the liquid-vapor interface for both pure and mixed fluids. Among the dynamical properties of the interfaces, we have calculated the time dependence of the velocity and angular velocity autocorrelation functions, continuous and intermittent survival probabilities, mean square displacements, diffusion coefficients, and also the dipole correlation functions and orientational relaxation times of interfacial solvent molecules. It is found that the width of the interfaces decreases with increase of concentration of the more polar component. The dipole vectors of the interfacial molecules tend to align parallel to the surfaces and this alignment is enhanced with increasing dipole moment of the fluid molecules. Also, the surface tension shows an increasing trend with increase of dipole moment of the molecules. The dynamical properties of the interfaces are found to be different from those of the corresponding bulk liquid phases. In general, the molecules at the interfaces are found to rotate and translate in the parallel direction at a somewhat faster rate than the bulk molecules. Also, on increase of concentration of the more polar component, the diffusion and orientational relaxation of interfacial molecules are found to show a weaker slowing down than those of the bulk molecules, which can be attributed to the preferential presence of the more polar component in the bulk liquid regions. The temporal behavior of the interfacial survival probabilities reveals a decrease of the survival times with increasing polarity, which can be attributed to a corresponding decrease in the interfacial thickness. Results are presented for both continuous and intermittent survival times and the origins of their differences are discussed. The free energy calculations reveal no minimum at the interfaces for adsorption of a charged solute, which shows that the ions would prefer to stay in the interior of the liquid phases, rather than at interfaces, for these model dipolar systems.     

Viabilty of atomistic potentials for thermodynamic properties of carbon dioxide at low temperatures

Investigation into volumetric and energetic properties of several atomistic models mimicking carbon dioxide geometry and quadrupole momentum covered the liquid-vapor coexistence curve. Thermodynamic integration over a polynomial and an exponential-polynomial path was used to calculate free energy. Computational results showed that model using GROMOS Lennard-Jones parameters was unsuitable for bulk CO(2) simulations. On the other hand, model with potential fitted to reproduce only correct density-pressure relationship in the supercritical region proved to yield correct enthalpy of vaporization and free energy of liquid CO(2) in the low-temperature region. Except for molar volume at the upper part of the vapor-liquid equilibrium line, the bulk properties of exp-6-1 parametrization of ab initio CO(2) potential were in a close agreement with the experimental results. Copyright 2001 John Wiley & Sons, Inc. J Comput Chem 22: 1772-1781, 2001     

Transferability of coarse-grained force fields: the polymer case

A key question for all coarse-graining methodologies is the degree of transferability of the resulting force field between various systems and thermodynamic conditions. Here we present a detailed study of the transferability over different thermodynamic states of a coarse-grained (CG) force field developed using the iterative Boltzmann inversion method. The force field is optimized against distribution functions obtained from atomistic simulations. We analyze the polymer case by investigating the bulk of polystyrene and polyamide-6,6 whose coarse-grained models differ in the chain length and in the number of atoms lumped in one bead. The effect of temperature and pressure on static, dynamic, and thermodynamic properties is tested by comparing systematically the coarse-grain results with the atomistic ones. We find that the CG model describing the polystyrene is transferable only in a narrow range of temperature and it fails in describing the change of the bulk density when temperature is 80 K lower than the optimization one. Moreover the calculation of the self-diffusion coefficient shows that the CG model is characterized by a faster dynamics than the atomistic one and that it overestimates the isothermal compressibility. On the contrary, the polyamide-6,6 CG model turns out to be fully transferable between different thermodynamic conditions. The transferability is checked by changing either the temperature or the pressure of the simulation. We find that, in this case, the CG model is able to follow all the intra- and interstructural rearrangements caused by the temperature changes. In addition, while at low temperature the difference between the CG and atomistic dynamics is remarkable due to the presence of hydrogen bonds in the atomistic systems, for high temperatures, the speedup of the CG dynamics is strongly reduced, leading to a CG diffusion coefficient only six times bigger than the atomistic one. Moreover, the isothermal compressibility calculated at different temperatures agrees very well with the experimental one. We find that the polymer chain length does not affect the transferability of the force field and we attribute such transferability mainly to the finer model used in describing the polyamide-6,6 than the polystyrene.     

Nonequilibrium multiscale computational model

A computational multiscale method is proposed to simulate coupled, nonequilibrium thermomechanical processes. This multiscale framework couples together thermomechanical equations at the coarse scale with nonequilibrium molecular dynamics at the fine scale. The novel concept of distributed coarse scale thermostats enables subsets of fine scale atoms to be attached to different coarse scale nodes which act as thermostats. The fine scale dynamics is driven by the coarse scale mean field. A coarse-grained Helmholtz free energy is used to derive macroscopic quantities. This new framework can reproduce the correct thermodynamics at the fine scale while providing an accurate coarse-grained result at the coarse scale.     

Multiscale coupling of mesoscopic- and atomistic-level lipid bilayer simulations

A multiscale method is presented to bridge between the atomistic and mesoscopic membrane systems. The atomistic model in this case is the united atom dimyristoylphosphatidylcholine membrane system, although the method is completely general. Atomistic molecular dynamics provides the expansion modulus which is used to parametrize a mesoscopic elastic membrane model. The resulting elastic membrane model, including explicit mesoscopic solvent, shows appropriate static and dynamic undulation behaviors. Large membranes of approximately 100 nm in length can then be easily simulated using the mesoscopic membrane system. The critical feedback from the mesoscopic system back down to the atomistic-scale system is accomplished by bridging the stress (or surface tension) of a small region in the mesoscopic membrane to the corresponding atomistic membrane system. Because of long length-scale modes of membranes such as undulation and buckling, the local tension responds differently from the frame tension, when subjected to external perturbations. The effect of these membrane modes is shown for the stress response of a local membrane region and therefore the atomistic membrane system. In addition, certain equilibrium static and dynamic properties of stand-alone and multiscale coupled systems are presented for several different membrane sizes. Although static properties such as two-dimensional pair-correlation function and order parameters show no noticeable discrepancy for the different systems, lipid self-diffusion and the rotational relaxation of lipid dipoles have a strong dependence on the membrane size (or long-wavelength membrane motions), which is properly modeled by the present multiscale method.     

Multiscale coarse-graining of ionic liquids

A recently developed multiscale coarse-graining (MS-CG) approach for obtaining coarse-grained force fields from fully atomistic molecular dynamics simulation is applied to the challenging case of the EMIM+NO3- ionic liquid. The force-matching in the MS-CG methodology is accomplished with an explicit separation of bonded and nonbonded forces. While the nonbonded forces are adopted from this force-matching approach, the bonded forces are obtained from fitting the statistical configurational data from the atomistic simulations. The many-body electronic polarizability is also successfully broken into effective pair interactions. With a virial constraint fixing the system pressure, the MS-CG models rebuild satisfactory structural and thermodynamic properties for different temperatures. The MS-CG model developed from a modest atomistic simulation is therefore suitable for simulating much larger systems, because the coarse-grained models show significant time integration efficiency. This approach is expected to be general for coarse-graining other ionic liquids, as well as many other liquid-state systems. The limitations of the present coarse-graining procedure are also discussed.     

Mixed atomistic and coarse-grained molecular dynamics: simulation of a membrane-bound ion channel

The recently developed multiscale coarse-graining (MS-CG) method (Izvekov, S.; Voth, G. A. J. Phys. Chem. B 2005, 109, 2469; J. Chem. Phys. 2005, 123, 134105) is used to build a mixed all-atom and coarse-grained (AA-CG) model of the gramicidin A (gA) ion channel embedded in a dimyristoylphosphatidylcholine (DMPC) lipid bilayer and water environment. In this model, the gA peptide was described in full atomistic detail, while the lipid and water molecules were described using coarse-grained representations. The atom-CG and CG-CG interactions in the mixed AA-CG model were determined using the MS-CG method. Molecular dynamics (MD) simulations were performed using the resulting AA-CG model. The results from simulations of the AA-CG model compare very favorably to those from all-atom MD simulations of the entire system. Since the MS-CG method employs a general and systematic approach to obtain effective interactions from the underlying all-atom models, the present approach to rigorously develop mixed AA-CG models has the potential to be extended to many other systems.     

Effect of flexibility on surface tension and coexisting densities of water

Molecular dynamics simulations of pure water at the liquid-vapor interface are performed using direct simulation of interfaces in a liquid slab geometry. The effect of intramolecular flexibility on coexisting densities and surface tension is analyzed. The dipole moment profile across the liquid-vapor interface shows different values for the liquid and vapor phases. The flexible model is a polarizable model. This effect is minor for liquid densities and is large for surface tension. The liquid densities increase from 2% at 300 K to 9% at 550 K when the force field is changed from a fully rigid simple point charge extended (SPCE) model to that of a fully flexible model with the same intermolecular interaction parameters. The increases in surface tension at both temperatures are around 11% and 36%, respectively. The calculated properties of the flexible models are closer to the experimental data than those of the rigid SPCE. The effect of the maximum number of reciprocal vectors (h(z) (max)) and the surface area on the calculated properties at 300 K is also analyzed. The coexiting densities are not sensitive to those variables. The surface tension fluctuates with h(z) (max) with an amplitude larger than 10 mN m(-1). The effect of using small interfacial areas is slightly larger than the error in the simulations.     

Simulations of solvation free energies and solubilities in supercritical solvents

Computer simulations are used to study solvation free energies and solubilities in supercritical solvents. Solvation free energies are calculated using the particle insertion method. The equilibrium solvent configurations required for these calculations are based on molecular dynamics simulations employing model solvent potentials previously tuned to reproduce liquid-vapor coexistence properties of the fluids Xe, C(2)H(6), CO(2), and CHF(3). Solutes are represented by all-atom potentials based on ab initio calculations and the OPLS-AA parameter set. Without any tuning of the intermolecular potentials, such calculations are found to reproduce the solvation free energies of a variety of typical solid solutes with an average accuracy of +/-2 kJmol. Further calculations on simple model solutes are also used to explore general aspects of solvation free energies in supercritical solvents. Comparisons of solutes in Lennard-Jones and hard-sphere representations of Xe show that solvation free energies and thus solubilities are not significantly influenced by solvent density fluctuations near the critical point. The solvation enthalpy and entropy do couple to these fluctuations and diverge similarly to solute partial molar volumes. Solvation free energies are also found to be little affected by the local density augmentation characteristic of the compressible regime. In contrast to solute-solvent interaction energies, which often provide a direct measure of local solvent densities, solvation free energies are remarkably insensitive to the presence of local density augmentation.     

Effect of Neck Formation on the Measurement of Dynamic Interfacial Tension in a Drop Volume Tensiometer

Dynamic interfacial tension values obtained by drop volume tensiometry will be affected under certain experimental conditions by the formation of a neck between the drop and the capillary tip. This phenomenon must be accounted for to obtain accurate values of interfacial tension. In this work, neck formation for a water–mineral oil system is studied under conditions where hydrodynamic effects can be neglected. A model originally developed for the determination of the surface tension of a suspended drop is modified for application to dynamic interfacial tensions of surfactant-containing liquids. The model relates apparent values of interfacial tension calculated from drops possessing necks to actual values. Experiments with Span 80 (sorbitan monooleate) and sodium dodecyl sulfate (SDS) surfactants in a mineral oil–water system are used to test the validity of the developed model. For the small tip diameter used, good agreement is obtained for Span 80 up to the critical micelle concentration, and for low concentrations of SDS, when the surfactant adsorption is diffusion-limited. In both cases, the neck diameter of the growing drop can be considered constant over the range of dynamic interfacial tensions tested.     

Molecular dynamics study of the liquid-vapor interface of acetonitrile: equilibrium and dynamical properties

The equilibrium and dynamical properties of the liquid-vapor interface of pure acetonitrile are studied by means of molecular dynamics simulations. Both nonpolarizable and polarizable models are employed in the present study. For the nonpolarizable model, the simulations are carried out for two different system sizes and at two different temperatures whereas the simulation with the polarizable model is done for a single system. The inhomogeneous density, anisotropic orientational profile, the width of the interface, and also the surface tension are calculated at room temperature and also at a lower temperature of 273 K. The dynamical aspects of the interface are investigated in terms of the single-particle dynamical properties such as the relaxation of velocity autocorrelation and the translational diffusion coefficients along the perpendicular and parallel directions and the dipole orientational relaxation of the interfacial acetonitrile molecules. The results of the interfacial dynamics are compared with those of the corresponding bulk phases at both temperatures. The convergence of the calculated results with respect to the length of simulation runs and the system size are also discussed.     

On Coarse-Graining by the Inverse Monte Carlo Method: Dissipative Particle Dynamics Simulations Made to a Precise Tool in Soft Matter Modeling

We present a promising coarse-graining strategy for linking micro- and mesoscales of soft matter systems. The approach is based on effective pairwise interaction potentials obtained from detailed atomistic molecular dynamics (MD) simulations, which are then used in coarse-grained dissipative particle dynamics (DPD) simulations. Here, the effective potentials were obtained by applying the inverse Monte Carlo method [Lyubartsev and Laaksonen, Phys. Rev. E. 52, 3730 (1995)] on a chosen subset of degrees of freedom described in terms of radial distribution functions. In our first application of the method, the effective potentials were used in DPD simulations of aqueous NaCl solutions. With the same computational effort we were able to simulate systems of one order of magnitude larger than the MD simulations. The results from the MD and DPD simulations are in excellent agreement.