[gpaw-users] constrained optimization in ASE

[Sampsa Riikonen]

Dear List,

I am in the need of doing some structural relaxations involving fixed 
bond lengths.

The correct way of doing such relaxations is, of course, by the aid of 
Lagrance multipliers.

However, in ASE, the constraints are implemented by "force filtering"

https://wiki.fysik.dtu.dk/ase/ase/constraints.html

In the example given, atom is constrained to move on a certain path by 
simply modifying its force to be
parallel to that path.

How correct is this "force filtering" .. ?  Any profound literature on 
the subject? .. At least, at a first glance I can't
see that the forces induced by lagrance multipliers would remove exactly 
the forces perpendicular to the path..?
Is this the case (i.e. force filtering is fundamentally the same thing 
as Lagrance multipliers) ?

Regards,

Sampsa