Abstract

Local Hyperdynamics

Arthur Voter
Los Alamos National Laboratory
Theoretical Division

I will discuss the characteristics and implementation of local hyperdynamics,
which differs from standard hyperdynamics in that the biasing is
performed locally, making it suitable for large systems. In standard
hyperdynamics, the requirement that the bias potential be zero everywhere
on the dividing surface bounding the current state has the consequence
that for large systems the boost factor decays to unity, regardless
of the form of the bias potential. In local hyperdynamics, the bias
force on each atom is obtained by differentiating a local bias energy
that depends only on the coordinates of atoms within a finite range
D of this atom. This bias force is thus independent of the bias
force in distant parts of the system, providing a method that gives
a constant boost factor, independent of the system size. Although
the resulting dynamics are no longer conservative, the method is
surprisingly accurate. We have shown that local hyperdynamics should
give correct results for a homogeneous system (all atoms equivalent),
but it has been more difficult to explain why it remains extremely
accurate for inhomogeneous systems as well. In this talk, I will
present our best understanding of why the method works so well, and
discuss our recent progress in implementing it in the LAMMPS code,
allowing million-atom simulations on the millisecond time scale.