Journal of Computational Chemistry
Volume 22, Issue 2; Pages: 184-195
ABSTRACT A tree code algorithm is presented for rapid evaluation of the potential energy in classical molecular systems. The algorithm treats a general nonbonded pairwise particle interaction, including the Coulomb and London dispersion potentials as special cases. The energy of the system is computed recursively as a sum of group-group interactions. The algorithm uses several adaptive techniques to reduce the execution time such as an adaptive tree structure and variable order multipole approximation. Two test cases are presented, a random set of particles representing a solvent and a set of particles lying on a B-spline curve representing a supercoiled DNA molecule. The results show that the tree code is significantly faster than direct summation for systems having a large number of particles; the crossover point and speedup depend on the accuracy requested.