If you are indeed trying to compute the free energy of your system using an
Einstein crystal reference state, you'll have to do two things:
1. remove existing EAM potential as a fn. of lambda.
2. introduce a spring tether to equilibrium sites.
2. can be accomplished using fix spring/self where you specify the spring
constant as lambda*K_spring, and dU/dlambda can be calculated as
U_spring/lambda while post-processing.
You can do 1. using fix adapt + compute ti where you progressively scale the
interatomic interactions down to zero. However, currently, lammps cannot
scale manybody interactions using fix adapt. A temporary, inelegant solution
to this is to tweak pair eam/fs to accept a scalar input and scale energies
and forces by this, and then computing du/dlambda = U_manybody/lambda after
the fact.
Sai.