No abstract

We study in J +1 dimensions a new deposition and evaporation model of a J-dimensional surface which bears a Potts-spin representation. For the pure deposition case, our simulations on systems up to 11 520^ sites in d'^2 and 2x192^ sites in t/^S yield roughness exponents which violate recent conjectures. Including evaporation, we observe a nonequilibrium surface-roughening transition in d'^3, but only a smooth crossover behavior in t/ -2. A logarithmic anomalous scaling form for surface width at the transition is conjectured. PACS numbers: 61.50.Cj, 05.40.+j, 05.70.Ln, 81.10.Bk Surface roughening due to thermal fluctuations has been studied extensively over the years. ^ There now exist well-established theories''^ as well as exactly solvable models ^' which come to fair agreement with observed surface roughening in copper^ and other experimental systems. In certain class of growth processes, such as Eden growth"* and vapor deposition,^ the moving surface of a compact cluster may also become rough under a stochastic growth rule. The scaling properties of the surface height fluctuation and characteristics of possible nonequilibrium roughening transitions have been the topic of a number of recent numerical and analytical investigations.^"^^ Kardar, Parisi, and Zhang proposed a nonlinear Langevin equation dh/dt-^vS/^h-hiX/Dmy-^Tjixj),

The roughening behavior of a moving surface under a deposition and evaporation dynamics is explored within the hypercube-stacking model. One limiting case of the model is an equilibrium surface, which exhibits thermal roughening for surface dimension d & 2. Another limiting case is nonequilibrium irreversible growth, where the model is shown to map exactly to zero-temperature directed polymers on a hypercubic lattice with a random energy distribution.Results of exact calculations for d = 1 and of large-scale Monte Carlo simulations [N = 2, 11520, and 2 x 192 surface sites for d = 1, 2, and 3, respectively] are presented that establish the Kardar-Parisi-Zhang equation as the correct continuum description of the growth process. For pure deposition (i.e., irreversible growth), careful analysis of surface width data, yields the exponents P(2) = 0.240 6 0.001 and P(3) = 0.180 6 0.005, which violate a number of recent conjectures. By allowing for evaporation, we observe a less rapid increase of the surface roughness as a function of time. This plienomenon is consistently explained by a crossover scenario for d = 1 and 2 but a nonequilibrium roughening transition for d = 3, as predicted by a perturbative renormalization-group analysis of the Kardar-Parisi-Zhang equation. Detailed predictions on crossover scaling from the renormalizationgroup analysis are also confirmed by simulation data. In the d = 1 case, some of the continuum parameters characterizing the renormalization-group Bow are obtained explicitly in terms of the lattice parameters via the exact calculation of steady-state properties of the model.

An effective-potential approach is presented for improving sampling efficiency in simulations of atomistically detailed models of dense long-chain liquids. The motion of atoms on short-time scales is described as rapid fluctuations about the slowly moving mean conformations of the chain molecules. The distribution of these fluctuations is approximated by that of isotropic elastic motion. The interactions between nonbonded pairs of atoms are preaveraged over this distribution and a much softer, effective interaction is obtained, allowing a correspondingly faster exploration of configuration space. The approximate sampling scheme is then tested on two model systems of united-atom liquid hydrocarbons—a melt of twenty C24 chains and one of ten C71 chains. In the C24 melt, where a comparison with fully equilibrated samples from rigorous algorithms is possible, the preaveraging is shown to produce an improvement in sampling efficiency of up to an order of magnitude at the expense of only a moderate loss in accuracy. The method is then applied to the C71 melt, where currently available, rigorous sampling algorithms fail to produce thorough equilibration. Also in this case, the preaveraging method proves to significantly enhance the sampling speed while producing satisfactory distributions of observables.