The Lennard-Jones 12-6 potential (LJ) is arguably the most widely used pair potential in Molecular Simulations. In fact, it is so popular that the question is rarely asked whether it is fit for purpose. In this paper, we argue that, whilst the LJ potential was designed for noble gases such as argon, it is often used for systems where it is not expected to be particularly realistic. Under those circumstances, the disadvantages of the LJ potential become relevant: most important among these is that in simulations the LJ potential is always modified such that it has a finite range. More seriously, there is by now a whole family of different potentials that are all called Lennard-Jones 12-6, and that are all different -and that may have very different macroscopic properties. In this paper, we consider alternatives to the LJ 12-6 potential that could be employed under conditions where the LJ potential is only used as a typical short-ranged potential with attraction. We construct a class of potentials that are, in many respects LJ-like but that are by construction finite ranged, vanishing quadratically at the cut-off distance, and that are designed to be computationally cheap. Below, we present this potential and report numerical data for its thermodynamic and transport properties, for the most important cases (cut-off distance r c =2σ ("LJ-like") and r c =1.2σ (a typical "colloidal" potential).subsequently, the first Molecular Dynamics (MD) simulations by Rahman 7 , there was an unexpectedly good agreement between the simulation results and the experimental data for liquid argon. The reason, as was argued in 8 was due to a fortuitous cancellation of errors. However, towards the end of the 20-th century, the LJ 12-6 potential had already achieved an almost unassailable status in classical simulations: it was (and is) used to describe atoms, molecules, coarse-grained models of proteins, and in some cases even larger particles such as nano-colloids. However, for these systems, there is no evidence at all that the LJ 12-6 potential is better than other possible choices. Yet, whenever new simulation techniques are tested, the LJ 12-6 potential is the first model to try.However, even if the true Lennard-Jones 12-6 potential would have been a satisfactory all-purpose potential, there is a practical problem: the Lennard-Jones potential has an infinite range, which would make it less suited for efficient numerical simulations (note that the infinite range was an advantage for (Lennard-)Jones's analytical calculations). This problem is normally resolved in practice by truncating the potential at a finite distance r c , e.g. r c =2.5σ . Unfortunately, not all authors use the same truncation procedure, and in recent years this confusion has become worse, as the cost of using a larger (but still finite) cut-off distance has become less prohibitive. In addition, in MD simulations, one should truncate the force, rather than the potential. So actually, the potential is truncated and shifted. Yet even such a 1 arXiv:1910.05746v2 [cond...

We report a numerical study of the diffusiophoresis of short polymers using non-equilibrium molecular dynamics simulations. More precisely, we consider polymer chains in a fluid containing a solute that has a concentration gradient and examine the variation of the induced diffusiophoretic velocity of the polymer chains as the interaction between the monomer and the solute is varied. We find that there is a non-monotonic relation between the diffusiophoretic mobility and the strength of the monomer–solute interaction. In addition, we find a weak dependence of the mobility on the length of the polymer chain, which shows clear difference from the diffusiophoresis of a solid particle. Interestingly, the hydrodynamic flow through the polymer is much less screened than for pressure driven flows.

Gradients in temperature, concentration or electrostatic potential cannot exert forces on a bulk fluid; they can, however, exert forces on a fluid in a microscopic boundary layer surrounding a (nano)colloidal solute, resulting in so-called phoretic flow. Here we present a simulation study of phoretic flow around a spherical colloid held fixed in a concentration gradient. We show that the resulting flow velocity depends non-monotonically on the strength of the colloid-fluid interaction. The reason for this non-monotonic dependence is that solute particles are effectively trapped in a shell around the colloid and cannot contribute to diffusio-phoresis. We also observe that the flow depends sensitively on the anisotropy of solute-colloid interaction.

The methodology to simulate transport phenomena in bulk systems is well-established. In contrast, there is no clear consensus about the choice of techniques to model cross-transport phenomena and phoretic transport, mainly because some of the hydrodynamic descriptions are incomplete from a thermodynamic point of view. In the present paper, we use a unified framework to describe diffusio-osmosis(phoresis), and we report non-equilibrium molecular dynamics (NEMD) on such systems. We explore different simulation methods to highlight some of the technical problems that arise in the calculations. For diffusiophoresis, we use two NEMD methods: boundary-driven and field-driven. Although the two methods should be equivalent in the limit of very weak gradients, we find that finite Peclet-number effects are much stronger in boundary-driven flows than in the case where we apply fictitious color forces. Graphic abstract

The existing kinetic theory of gases is based on an analytical approach that becomes intractable for all but the simplest molecules. Here we propose a simple numerical scheme to compute the transport properties of molecular gases in the limit of infinite dilution. The approach that we propose is approximate, but our results for the diffusivity D, the viscosity η, and the thermal conductivity λ of hard spheres, Lennard-Jones particles, and rough hard spheres agree well with the standard (lowest order) Chapman−Enskog results. We also present results for a Lennard-Jones-dimer model for nitrogen, for which no analytical results are available. In the case of polyatomic molecules (we consider n-octane), our method remains simple and gives good predictions for the diffusivity and the viscosity. Computing the thermal conductivity of polyatomic molecules requires an approximate treatment of their quantized internal modes. We show that a wellknown approximation that relates λ to D and η yields good results. We note that our approach should yield a lower limit to the exact value of D, η, and λ. Interestingly, the most sophisticated (higher-order) Chapman− Enskog results for rough hard spheres seem to violate this bound.