Descripancy in energy of non-interacting atoms with springs

Dear lammps users,

I tried to simulate single atom attached to its lattice site by a harmonic
spring in NVT ensemble using bussi thermostat in lammps. The spring
constant is 1326.4 kcal/molA2. The average harmonic energy in the
simulation was estimated to be 0.68105 kcal/mol.

But the analytical result for the same case is 0.7606kcal/mol and using
Monte Carlo simulations for the same case is 0.76 kcal/mol.

Instead of single spring if i use 10 springs the average energy is close
to the analytical value..

Can any one comment on this why do we have the variation in single
molecule simulation results with lammps?

Even in case of multiple springs, the average energy of all springs are

please find the input file i used for this simulation in the attachment

IISc Banglaore

conf-Ih.txt (876 Bytes)

input-real1.txt (2.13 KB)

I don’t know the details of how the Bussi thermostat works

or whether it is statistically valid for thermostatting a single atom.

I don’t think Nose/Hoover would be. Langevin would be fine

(you could try that one and compare). However, computing

the temperature of a single atom, which T is used by

some thermostats (not sure about Bussi) is problematic.

You will need to make sure the LAMMPS is computing the

temperature you expect, as used by the thermostat. Generally

the T computation subtracts degrees-of-freedom (e.g. 3 for

center-of-mass), so that needs careful attention if you

are trying to thermostat a single atom.