simtk.org

Hi,

I am trying to use ForceBalance to optimize the Lennard-Jones parameters of a molecular liquid. My input file follows. I've based my input file on the dms example and it is executing and optimizing the parameters effectively. I'm not clear on what non-bonded options are used by default and how they can be changed. I would like to use PME with a 12 vdw cutoff but I can't find a way to specify this in the npt.py script or forcebalance itself. The npt2.py script seems to be incompatible with forcebalance right now. ForceBalance seems to automatically link to the system npt.py script rather than using the one in the target directories.

Thanks,
Chris

$options
forcefield ff.xml
jobtype newton
convergence_step 0.01
convergence_objective 0.1
$end


$target
type Liquid_OpenMM
name molliquid
liquid_md_steps 2000000
liquid_eq_steps 500000
liquid_timestep 2.0
liquid_interval 0.1
gas_equ_steps 10000
gas_prod_steps 100000
gas_timestep 2
gas_interval 0.1
minimize_energy 0
$end

Hi Chris,

Thanks for getting in touch. I think that with OpenMM, there are some defaults for the nonbonded settings which are being set in the source code, and this option isn't exposed in the input file. However, it's very reasonable that this should be an option that you can specify in running the calculation.

I'm not sure which version you're running, but in the latest development version the code for specifying the cutoff is here: https://github.com/leeping/forcebalance ... io.py#L596 . The npt.py script creates an Engine object to run the simulation, and the Engine objects are defined in the various xxio.py modules.

As you can see, the cutoff is 8.5 Angstrom with an isotropic vdW correction in the case of PBC, and infinite in the case of no PBC. For the time being, you may change the source code to get the cutoff you want, although I should probably add this as a feature (let me know if you're interested).

Thanks,

- Lee-Ping

PS: I think npt2.py should probably not be a part of the software distribution. ForceBalance automatically links to the installed version of npt.py, because I decided it was unsustainable to have a separate npt.py script for each job.

Some suggestions:

1) It's best if you use the most recent ForceBalance code that is downloaded straight from the GitHub repository, if you're not already doing this.

2) If the optimization is already working effectively for your problem, you may keep the liquid simulation settings -- but if you're still setting things up I suggest you set the liquid_xx_steps to 10% of their current values.