Minimum image problem, atomic

The Finite Range Problem. The minimum image convention requires (1) the use of interaction energy functions that are of finite range, i.e. that are non-zero only for distances below a certain limit, R. As a consequence, only a fraction of all minimum image pairs actually interact with non-zero energy this fraction must be less than ir/2d, i.e., in two dimensions maximally 0.79, in three dimensions maximally 0.52 (and actually often less than 0.3). It is desirable to efficiently eliminate from consideration all noninteracting atom pairs. 

In the work of Edwards et al [36], which included interactions only to the third neighbor shell (approximately 4.4 A), the problem of energy instability for a 1000 K simulation did not arise the timestep was lO s at all temperatures. Since no analysis of the effects of timestep and cutoff was presented, it is difficult to determine the reasons for the differences in the behavior observed by Edwards et al and in the present study. Some possibilities are (a) They may have employed a minimum image approach for the periodic boundary conditions, whereas we have calculated both primaryprimary and primary-image interactions for each atom, (b) Their system was much larger than ours (2048 vs 256 atoms), (c) Edwards et al used the NVE ensemble for the production simulations, in contrast to our choice of the NPT ensemble. 

---