ASE Manual Release 3.6.1.2825 CAMd - CampOS Wiki

header: Print a header line defining the columns.
stress: Print the six components of the stress tensor.
peratom: Print energy per atom instead of total energy.
mode: If ‘a’, append to existing file, if ‘w’ overwrite existing file.
ASE Manual, Release 3.6.1.2828
Despite appearances, attaching a logger like this does not create a cyclic reference to the dynamics.
Note: If building your own logging class, be sure not to attach the dynamics object directly to the logging object.
Instead, create a weak reference using the proxy method of the weakref package. See the ase.md.MDLogger
source code for an example. (If this is not done, a cyclic reference may be created which can cause certain
calculators, such as Jacapo, to not terminate correctly.)
7.21.4 Constant NVE simulations (the microcanonical ensemble)
Newton’s second law preserves the total energy of the system, and a straightforward integration of Newton’s
second law therefore leads to simulations preserving the total energy of the system (E), the number of atoms (N)
and the volume of the system (V). The most appropriate algorithm for doing this is velocity Verlet dynamics, since
it gives very good long-term stability of the total energy even with quite large time steps. Fancier algorithms such
as Runge-Kutta may give very good short-term energy preservation, but at the price of a slow drift in energy over
longer timescales, causing trouble for long simulations.
In a typical NVE simulation, the temperature will remain approximately constant, but if significant structural
changes occurs they may result in temperature changes. If external work is done on the system, the temperature is
likely to rise significantly.
Velocity Verlet dynamics
class md.verlet.VelocityVerlet(atoms, timestep)
VelocityVerlet is the only dynamics implementing the NVE ensemble. It requires two arguments, the atoms
and the time step. Choosing a too large time step will immediately be obvious, as the energy will increase with
time, often very rapidly.
Example: See the tutorial Molecular dynamics.
7.21.5 Constant NVT simulations (the canonical ensemble)
Since Newton’s second law conserves energy and not temperature, simulations at constant temperature will somehow
involve coupling the system to a heat bath. This cannot help being somewhat artificial. Two different approaches
are possible within ASE. In Langevin dynamics, each atom is coupled to a heat bath through a fluctuating
force and a friction term. In Nosé-Hoover dynamics, a term representing the heat bath through a single degree of
freedom is introduced into the Hamiltonian.
Langevin dynamics
class md.langevin.Langevin(atoms, timestep, temperature, friction)
The Langevin class implements Langevin dynamics, where a (small) friction term and a fluctuating force are added
to Newton’s second law which is then integrated numerically. The temperature of the heat bath and magnitude of
the friction is specified by the user, the amplitude of the fluctuating force is then calculated to give that temperature.
This procedure has some physical justification: in a real metal the atoms are (weakly) coupled to the electron gas,
and the electron gas therefore acts like a heat bath for the atoms. If heat is produced locally, the atoms locally
get a temperature that is higher than the temperature of the electrons, heat is transferred to the electrons and then
rapidly transported away by them. A Langevin equation is probably a reasonable model for this process.
7.21. Molecular dynamics 153