We are putting a warning that an optimization with 'fixatoms' is not yet implemented. Put 0 in the forces (and/or displacement) works form some algorithm but not for others. The proper way to proceed is to reduce the dimensionality of the optimization problem.
How about instead we make it work properly? Does not seem (too) tricky, you just add a mask to the function/gradient wrappers so they map a subset of the atomic coordinates to the optimized vector.
We are putting a warning that an optimization with 'fixatoms' is not yet implemented. Put 0 in the forces (and/or displacement) works form some algorithm but not for others. The proper way to proceed is to reduce the dimensionality of the optimization problem.