In this article, we describe the Desmos supercomputer that consists of 32 hybrid nodes connected by a low-latency high-bandwidth Angara interconnect with torus topology. This supercomputer is aimed at cost-effective classical molecular dynamics calculations. Desmos serves as a test bed for the Angara interconnect that supports 3-D and 4-D torus network topologies and verifies its ability to unite massively parallel programming systems speeding-up effectively message-passing interface (MPI)-based applications. We describe the Angara interconnect presenting typical MPI benchmarks. Desmos benchmarks results for GROMACS, LAMMPS, VASP and CP2K are compared with the data for other high-performance computing (HPC) systems. Also, we consider the job scheduling statistics for several months of Desmos deployment.

Diffusion is one of the key subjects of molecular modeling and simulation studies. However, there is an unresolved lack of consistency between Einstein-Smoluchowski (E-S) and Green-Kubo (G-K) methods for diffusion coefficient calculations in systems of complex molecules. In this paper, we analyze this problem for the case of liquid n-triacontane. The non-conventional long-time tails of the velocity autocorrelation function (VACF) are found for this system. Temperature dependence of the VACF tail decay exponent is defined. The proper inclusion of the long-time tail contributions to the diffusion coefficient calculation results in the consistency between G-K and E-S methods. Having considered the major factors influencing the precision of the diffusion rate calculations in comparison with experimental data (system size effects and force field parameters), we point to hydrogen nuclear quantum effects as, presumably, the last obstacle to fully consistent n-alkane description.

In this paper, the predictive power of molecular dynamics methods is demonstrated for the cases of model paraffinic and aromatic lubricant liquids at pressures up to 400 MPa. The shear viscosity and self-diffusion coefficients are calculated for 2,2,4-trimethylpentane (C₈H₁₈) at 298 K and 1,1-diphenylethane (C₁₄H₁₄) at 333 K. Three force fields with different levels of accuracy are compared by the ability to predict the experimental data. The Stokes–Einstein correlation between viscosity and self-diffusion is demonstrated for both compounds.

Computational predictions of the high-pressure viscosity of hydrocarbon mixtures could help to accelerate the development of fuels and lubricants with improved performance. In this study, we use molecular dynamics simulations to study the viscosity and density of methylcyclohexane, 1-methylnaphthalene, and their binary mixtures at 323 K and pressures of up to 500 MPa. The simulation results are in excellent agreement with previous experiments available up to 100 MPa for both pure compounds (200 MPa for 1-methylnaphthalene) and the binary mixtures. For 1-methylnaphthalene, the viscosity initially increases slower-than-exponential with pressure before it reaches an inflection point and then increases faster-than-exponential. The inflection point (300 MPa) occurs at a pressure slightly below the one at which 1-methylnaphthalene is expected to enter the supercooled phase (400 MPa). For methylcyclohexane, the increase in viscosity with pressure is slower-than-exponential over the entire pressure range studied. The binary mixtures show intermediate pressure–viscosity responses between the two pure cases. The applicability of equations commonly used to describe the pressure dependence of viscosity, as well as the viscosity of binary mixtures, is evaluated against the computational predictions.

Decades of molecular simulation history proved that the Green-Kubo method for shear viscosity converges without any problems in atomic and simple molecular liquids, unlike liquids with high values of viscosity. In the case of highly viscous liquids, the time decomposition method was developed in 2015 by Maginn and co-authors [Y. Zhang, A. Otani, and E. J. Maginn, J. Chem. Theory Comput. 11, 3537–3546 (2015)] which allows us to improve the convergence of the Green-Kubo integral. In this paper, the contributions of intramolecular and intermolecular force field parts to the viscosity integral are discovered to gain the understanding of the Green-Kubo method. The n-alkanes from n-ethane to n-pentane at 330 K in the optimized potentials for liquid simulations-all atom force field are used as reference models. The dependencies of these contributions and decay times of the corresponding correlation functions on the chain length are observed. The nonequilibrium simulations are carried out to verify the Green-Kubo results. The obtained values of viscosity are compared with experimental data.