On searching neighbors in computer simulations of macromolecular systems

Various algorithms for evaluating nonbonded interactions in molecular dynamics (MD) simulations of macromolecular systems are considered, and a combination of two techniques using a space grid for finding neighbor atoms is proposed. The application of grid search techniques to nonrectangular periodic systems is discussed. Finally, the computing time required by different algorithms is compared on a Cray‐1 vector processing computer as well as on a CDC Cyber 170/760 sequential computer. Neighbor list techniques turn out to be faster than the grid search techniques for the systems considered here. However, storage requirements may exclude the use of neighbor list techniques for large systems. Finally, MD of complex macromolecular systems turned out to be about 4–14 times faster on a Cray‐1 than on a Cyber 170/760.