Heat current
Using the force expression, one can derive the following expression for the heat current for the whole system (\(E_i\) is the total energy of atom \(i\)) [Fan2015]:
where
and
Exchanging the summation indices gives the following equivalent expression for the total potential heat current [Fan2015]:
These expressions are equivalent after summing over all atoms, but their individual atomic contributions generally differ. GPUMD uses the following per-atom partition, which gives the same total heat current:
With this partition, the per-atom heat current can be expressed in terms of the per-atom virial \(\mathbf{W}_i\) defined above [Gabourie2021]:
where the per-atom virial tensor cannot be assumed to be symmetric and the full tensor with 9 components should be used [Gabourie2021]. This result has actually been clear from the derivations in [Fan2015] but it was wrongly stated there that the potential part of the heat current cannot be expressed in terms of the per-atom virial.
One can also derive the following expression for the heat current from a subsystem \(A\) to a subsystem \(B\) [Fan2017]: