Dynamic Scaling of Surface Growth in Simple Lattice Models Croucher ASI on Frontiers in Computational Methods and Their Applications in Physical Sciences Dec. 6 - 13, 2005 The Chinese University of Hong Kong Dynamic Scaling of Surface Growth in Simple Lattice Models D. P. L. S. Pal K. Binder Background Models and Simulation Method Results Surface properties Temporal correlations Summary and Conclusions
Simulation NATURE Experiment Theory
NATURE Simulation (Monte Carlo) Experiment Theory (MBE, LEED, RHEED) (Growth eqns.)
Why MBE? - the promise of designer materials e.g. multilayers ________________ quantum wires (vicinal surfaces) Theoretical questions: binding energies - Quant Mech large scale structures - Stat Mech
Theoretical Background non-equilibrium (equilibrium roughening transition) define: height above L L substrate mean height surface width structure factor local order parameter
Comprehensive growth equation: random noise h = deviation of surface height from the mean
Edwards-Wilkinson growth equation: (sedimentation) random noise h = deviation of surface height from the mean
Question: What happens when t ? simple model studies: EW sedimentation model (Edwards & Wilkinson, 1982) KPZ equation (Kardar, Parisi & Zhang, 1986) random deposition (Family, 1986) restricted SOS model (Kim & Kosterlitz, 1989) growth-diffusion model (Wolf & Villain, 1990) MBE models (Pal and Landau, 1993) and many more . . .
Surface Width Dynamic Finite Size Scaling Define: z =/ = dynamic exponent
Computational Study of Film Growth Surface Science Statistical Mechanics Multiple processes: deposition & diffusion Methods: “Ab initio” Molecular Dynamics Classical Molecular Dynamics (phenomenological potentials) O O O OOO OO OO surface Discrete stochastic SOS models surface
Atomistic Edwards-Wilkinson Model L L square lattice substrate (p. b. c.) Growing film held at constant temperature T Particles fall randomly on the surface, then diffuse to the neighboring site with the greatest depth constant flux
Atomistic Edwards-Wilkinson Model BUT, what if more than one neighboring site has the same depth? constant flux
Atomistic Edwards-Wilkinson Model BUT, what if more than one neighboring site has the same depth? Generate a random number to decide! constant flux
Monte Carlo is Serious Science!
Simulations of MBE: Monte Carlo (MC) versus Kinetic Monte Carlo (KMC) Deposition + diffusion O OOOOOOOOOO MC KMC In KMC we must consider more than just the final particle state!
MBE Model Growth (KMC) UDP Model RHEED intensity-growth of GaAs (Neave et al, 1985)
MBE Model Growth (KMC) What happens at “long times”? Dynamic finite size scaling shows z=1.65 But the Edwards- Wilkinson eqn. yields z=2.0
Atomistic 2+1 dim EW Model Interfacial width: What happens at “long times”? (Note: For large systems >1010 random numbers are needed per run)
Atomistic 2+1 dim EW Model Interfacial width: Dynamic Finite Size Scaling …for the EW equation = 0 , so
Atomistic 2+1 dim EW Model Interfacial width: Dynamic Finite Size Scaling
Atomistic 2+1 dim EW Model Interfacial width: Dynamic Finite Size Scaling
Atomistic 2+1 dim EW Model Structure Factor: Dynamic Finite Size Scaling
Atomistic 2+1 dim EW Model Structure Factor: Dynamic Finite Size Scaling
Atomistic 2+1 dim EW Model Structure Factor: Dynamic Finite Size Scaling Data do NOT scale for z=2.0 !
Atomistic 2+1 dim REW Model Restricted Edwards-Wilkinson Model: When two or more neighboring sites have equal depth, the particle does not diffuse!
Atomistic 2+1 dim REW Model Interfacial width: What happens at “long times”?
Atomistic 2+1 dim REW Model Interfacial width: Dynamic finite size scaling Data scale (for long times) with z=2.0 !
Atomistic 2+1 dim REW Model Structure factor: Dynamic Finite Size Scaling
Atomistic 2+1 dim REW Model Structure factor: Dynamic Finite Size Scaling
Growth equation: h = deviation of surface height from the mean surface stiffness Measure surface properties numerically: Average quantities over b b blocks of sites
Growth equation: h = deviation of surface height from the mean “generalized noise” Measure surface properties numerically: Average quantities over b b blocks of sites
Atomistic 2+1 dim EW Model Surface stiffness Stiffness decays to a constant value
Atomistic 2+1 dim EW Model Non-equilibrium contribution to the interface velocity* * i.e. “generalized noise”
Atomistic 2+1 dim EW Model Non-equilibrium contribution to the interface velocity
Atomistic 2+1 dim EW and REW Models Non-equilibrium contribution to the interface velocity For random noise, U(b,t) should decay to 0 !
Time-displaced Correlation Function To study temporal correlations in U(b,t), define blocking factor
Time-displaced Correlation Function EW model (finite size effects) Note: C(b,) is independent of L
Time-displaced Correlation Function EW model (time dependence) C(b,) decays non-exponentially !
Time-displaced Correlation Function REW model Correlations decay exponentially fast to 0 !
Time-displaced Correlation Function Dynamic scaling: where
Time-displaced Correlation Function EW model z
Time-displaced Correlation Function EW model (define: P(b,)=C(b,)/ (b)-1 ) For b > 15, get scaling with z = 1.65
Time-displaced Correlation Function EW model No scaling for z = 2.0 !
Summary and Conclusions Simple lattice models for surface growth have behavior that depends on the “noise”: The atomistic EW model does not have the same behavior as the EW equation! … but the REW model the EW equation. Time-displaced correlations are non-exponential for the atomistic EW model the use of a random number to choose one of the degenerate neighbor sites creates a 2nd source of (correlated) noise. Challenge for the future: - Study other models by simulation to extract the noise Universality classes?