Kinetic Monte Carlo simulation accuracy

Kinetic Monte Carlo Simulations. The approximations of the previous sections generally make it rather easy to interpret the results of a kinetic model. Their drawback is that it is very difficult to assess their accuracy. KMC simulations do not have this drawback. For a given reaction model the results of a kMC simulation are exact. 

For the equihbrium properties and for the kinetics under quasi-equilibrium conditions for the adsorbate, the transfer matrix technique is a convenient and accurate method to obtain not only the chemical potentials, as a function of coverage and temperature, but all other thermodynamic information, e.g., multiparticle correlators. We emphasize the economy of the computational effort required for the application of the technique. In particular, because it is based on an analytic method it does not suffer from the limitations of time and accuracy inherent in statistical methods such as Monte Carlo simulations. The task of variation of Hamiltonian parameters in the process of fitting a set of experimental data (thermodynamic and... 

By contrast, few such calculations have as yet been made for diffusional problems. Much more significantly, the experimental observables of rate coefficient or survival (recombination) probability can be measured very much less accurately than can energy levels. A detailed comparison of experimental observations and theoretical predictions must be restricted by the experimental accuracy attainable. This very limitation probably explains why no unambiguous experimental assignment of a many-body effect has yet been made in the field of reaction kinetics in solution, even over picosecond timescale. Necessarily, there are good reasons to anticipate their occurrence. At this stage, all that can be done is to estimate the importance of such effects and include them in an analysis of experimental results. Perhaps a comparison of theoretical calculations and Monte Carlo or molecular dynamics simulations would be the best that could be hoped for at this moment (rather like, though less satisfactory than, the current position in the development of statistical mechanical theories of liquids). Nevertheless, there remains a clear need for careful experiments, which may reveal such effects as discussed in the remainder of much of this volume. 

This selection process is then iterated, beginning from an initial state of the system, as defined by species populations, to simulate a chemical evolution. A statistical ensemble is generated by repeated simulation of the chemical evolution using different sequences of random numbers in the Monte Carlo selection process. Within limits imposed by computer time restrictions, ensemble population averages and relevant statistical information can be evaluated to any desired degree of accuracy. In particular, reliable values for the first several moments of the distribution can be obtained both inexpensively and efficiently via a computer algorithm which is incredibly easy to implement (21, 22), especially in comparison to now-standard techniques foF soTving the stiff ordinary differential equations (48, 49) which may arise in the deterministic description of chemical kinetics (53). Now consider briefly the essential features of a simple chemical model which illustrates well the attributes of stochastic chemical simulations. 

---