An energy basin finding algorithm for kinetic Monte Carlo acceleration.

We present an energy basin finding algorithm for identifying the states in absorbing Markov chains used for accelerating kinetic Monte Carlo (KMC) simulations out of trapping energy basins. The algorithm saves groups of states corresponding to basic energy basins in which there is (i) a minimum energy saddle point and (ii) in moving away from the minimum the saddle point energies do not decrease between successive moves. When necessary, these groups are merged to help the system escape basins of basins. Energy basins are identified either as the system visits states, or by exploring surrounding states before the system visits them. We review exact and approximate methods for accelerating KMC simulations out of trapping energy basins and implement them within our algorithm. Its flexibility to store varying numbers of states, and ability to merge sets of saved states as the program runs, allows it to efficiently escape complicated trapping energy basins. Through simulations of vacancy-As cluster dissolution in Si, we demonstrate our algorithm can be several orders of magnitude faster than standard KMC simulations.

[1]  Pascal Bellon,et al.  Identification of novel diffusion cycles in B2 ordered phases by Monte Carlo simulation , 1997 .

[2]  Novotny.  Monte Carlo algorithms with absorbing Markov chains: Fast local algorithms for slow dynamics. , 1995, Physical review letters.

[3]  L. Sander,et al.  Multiscale kinetic Monte Carlo algorithm for simulating epitaxial growth , 2005, cond-mat/0504272.

[4]  W. Fichtner,et al.  Arsenic deactivation in Si: Electronic structure and charge states of vacancy-impurity clusters , 2003 .

[5]  Michael L. Falk,et al.  Multiscale diffusion Monte Carlo simulation of epitaxial growth , 2006, J. Comput. Phys..

[6]  D. R. Mason,et al.  Stochastic kinetic Monte Carlo algorithms for long-range Hamiltonians , 2004, Comput. Phys. Commun..

[7]  K. Saarinen,et al.  Formation of vacancy-impurity complexes by annealing elementary vacancies introduced by electron irradiation of As-, P-, and Sb-doped Si , 2004 .

[8]  J. Mullen.  Isotope effect in intermetallic diffusion , 1961 .

[9]  J. Xie,et al.  Diffusion and Clustering in Heavily Arsenic-Doped Silicon: Discrepancies and Explanation , 1999 .

[10]  First-principles calculation of intrinsic defect formation volumes in silicon , 2005, cond-mat/0508329.

[11]  D. Gillespie.  Markov Processes: An Introduction for Physical Scientists , 1991 .

[12]  D. Gillespie.  Exact Stochastic Simulation of Coupled Chemical Reactions , 1977 .

[13]  A. Voter,et al.  Classically exact overlayer dynamics: Diffusion of rhodium clusters on Rh(100). , 1986, Physical review. B, Condensed matter.

[14]  Clinton Dew Van Siclen,et al.  Stochastic method for accommodation of equilibrating basins in kinetic Monte Carlo simulations , 2007, Journal of physics. Condensed matter : an Institute of Physics journal.

[15]  M. Novotny,et al.  A Tutorial on Advanced Dynamic Monte Carlo Methods for Systems with Discrete State Spaces , 2001, cond-mat/0109182.

[16]  K. Saarinen,et al.  Formation of thermal vacancies in highly As and P doped Si. , 2004, Physical review letters.

[17]  A. B. Bortz,et al.  A new algorithm for Monte Carlo simulation of Ising spin systems , 1975 .

[18]  Dietrich Stauffer,et al.  Anual Reviews of Computational Physics VII , 1994 .

[19]  William H. Press,et al.  Numerical recipes in C. The art of scientific computing , 1987 .

[20]  W. H. Weinberg,et al.  Theoretical foundations of dynamical Monte Carlo simulations , 1991 .

[21]  Ab initio calculations of the structure and energetics of As-vacancy complexes in silicon , 1999 .

[22]  M. A. Novotny.  Low-temperature long-time simulations of Ising ferromagnets using the Monte Carlo with Absorbing Markov Chains method , 2002 .

[23]  Ed Anderson,et al.  LAPACK Users' Guide , 1995 .

[24]  David J Wales,et al.  Graph transformation method for calculating waiting times in Markov chains. , 2006, The Journal of chemical physics.

[25]  David J. Srolovitz,et al.  First passage time Markov chain analysis of rare events for kinetic Monte Carlo: double kink nucleation during dislocation glide , 2002 .

[26]  A. Satta,et al.  Ab initio structures of As{sub m}V complexes and the simulation of Rutherford backscattering channeling spectra in heavily As-doped crystalline silicon , 2005 .

[27]  A. La Magna,et al.  Accelerated Monte Carlo algorithms for defect diffusion and clustering , 2000 .

[28]  S. Dunham,et al.  Atomistic models of vacancy‐mediated diffusion in silicon , 1995 .

[29]  Cleve B. Moler,et al.  Nineteen Dubious Ways to Compute the Exponential of a Matrix, Twenty-Five Years Later , 1978, SIAM Rev..

[30]  K. Saarinen,et al.  Formation of vacancy-impurity complexes by kinetic processes in highly As-doped Si. , 2002, Physical review letters.

[31]  Jack Dongarra,et al.  LAPACK Users' Guide, 3rd ed. , 1999 .

[32]  Sean X. Sun.  Path summation formulation of the master equation. , 2006, Physical review letters.

[33]  H. Ryssel,et al.  Atomistic modeling of high-concentration effects of impurity diffusion in silicon , 1998 .

[34]  Ramamoorthy,et al.  Complex dynamical phenomena in heavily arsenic doped silicon. , 1996, Physical review letters.

[35]  M. Kalos,et al.  First-passage Monte Carlo algorithm: diffusion without all the hops. , 2006, Physical review letters.