Soft/Stiff Spring

User discussions
1

GROMACS version: 2024
GROMACS modification: Yes

Hi,
what is the difference between a stiff spring and a soft spring in steered molecular dynamics. What are the contexts where one should be chosen over the other.

Thanks

2

Hi,

Soft and stiff are not really defined, but they depend on the system and what you want to do. Generally, with soft one means using a weak spring constant in the steered phase (e.g. a few 10s of kJ/mol/nm^2) while with stiff the spring constant values are much higher (e.g. ~1000s of kJ/mol/nm^2).

The main difference is how closely does the dragged group follow the spring. In pulling, you attach a spring to a group and, by pulling the spring, the groups follows. If the restraint is weak, then the spring can deform a lot before the force applied by it on the group is sufficient to move it, while if the restraint is strong (stiff spring) then the group will follow more closely the spring as small spring deformations will lead very quickly to strong recalling forces. You can have a better idea by taking a look at the manual.

Intuitively, think about a rock on the ground (the group you want to pull), and you (the pulling force) attaching a rope (the spring) to the rock. If the spring is made of very strong un-elastic material (stiff spring approximation) the moment you start pulling on the rope the rock will follow you, supposing you are applying enough force to move it. Whatever movement you do, the rock will follow you for the same amount of space. If now we change the rope and we put an elastic one (soft spring approx), then, when you start pulling, the rope deforms and you are able to take a few steps without the rock moving. Once you are far enough, the tension built-up in the rope will be sufficient to have a stiff enough rope to actually start moving the rock.

Where/when using one of the two approaches depends on the problem. Generally, if you need the pulled group to really follow the spring step by step, then you need a stiff potential. Similarly if you are using theories/approximations that are based on stiff spring potential approximations (like some Jarzynski implementations for free energy calculations). However, stiff potential means large forces, and large forces can easily break the simulation engine.