Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
We generally run 2 versions of VASP - 5.4.4 and 6. VASP 6 is used primarily on nodes operating on the Knights Landing (KNL) CPU architecture, while VASP 5.4.4 is generally used on all other CPUs.
NERSC maintains compiled, and optimized versions of the vasp binaries. To get access to these binaries, you must be added to the group VASP license. If/when you determine that your research requires the use of VASP, please contact Kristin and request to be added to the VASP License. We currently have two sets of licenses i) UC Berkeley Affiliated (Purchased through VASP Software GmbH) and ii) LBNL Affiliated (Purchased through Materials Design). Once you have been added to one of the licenses, fill out the VASP License Confirmation Request to get access to the binaries on NERSC machines.
More information can be found in the NERSC documentation
The potcar files can be found in the following directory.
*Note*: Users with only jcesr access will need to request access to matgen to access these POTCARs. Contact the NERSC Resource Manager and request to be added to the 'matgen' UNIX group.
When running vasp on savio, we have access to our own compilations. To access the vasp binaries on savio, add the following line to your .bashrc file.
POTCARs are located at:
For issues with these compilations, please contact Eric Sivonxay (esivonxay@lbl.gov)
When running vasp on Lawrencium, we have access to our own compilations. To access the vasp binaries on Lawrencium, add the following line to your .bashrc file.
Pre-release versions of the VASP modules are located at export /clusterfs/mp/temp_modfiles
POTCARs are located at:
For issues with these compilations, please contact Eric Sivonxay (esivonxay@lbl.gov)
Special thanks to the original authors of this page: Eric Sivonxay
Gaussian is available on Savio. Anyone wanting to use Gaussian on Savio must sign Exhibit A. Print out two copies of the form and sign both (i.e. they should not be signed electronically). Then provide both physical copies to Alice. If you are remote, an e-mailed copy will suffice for the time being.
Talk to Rishabh, Jingyang, or Ryan if you have LAMMPS questions!
More Information to come
Create the molecule in Avogadro or wherever you want to
Then use antechamber to get the partial charges. Antechamber is part of AmberTools. Installation instructions are available in their manual
The relevant command:-
antechamber -fi pdb -fo prepi -i <infile.pdb> -o output.prepi -c bcc -j 4 -at gaff -nc -1
The -nc
flag is the net charge in the molecule. It will error out if not invoked for ions.
Sometimes it might be advisable to take partial charges and LJ pair coefficients directly from papers. This is especially important if you are trying to benchmark your results against previous publications. For instance, in case of TFSI-, antechamber
partial charges are significantly different than partial charges available in previous publications - https://pubs.acs.org/doi/full/10.1021/jp0686893
In short, antechamber
is a good starting point, but it is always advisable to tread carefully with partial charges in MD. Also remember, that these charges are not scaled. Why is charge scaling important? Go here β βKirby, B. J.; Jungwirth, P. Charge Scaling Manifesto: A Way of Reconciling the Inherently Macroscopic and Microscopic Natures of Molecular Simulations. J. Phys. Chem. Lett. 2019, 10 (23), 7531β7536. https://doi.org/10.1021/acs.jpclett.9b02652.β
Open the .pdb file in VMD. Use Extensions -> TK console
to open the TCL scripting menu
Then use this sequence of commands to get the bonds, angles, dihedrals and most importantly, the box size
topo retype bonds
topo guessangles
topo guessdihedrals
set sel [atomselect top all]
set center [measure center $sel weight none]
$sel moveby [vecscale -1.0 $center]
set minmax [measure minmax $sel -withradii]
set box [vecscale 1.1 [vecsub [lindex $minmax 1] [lindex $minmax 0]]]
pbc set $box
Once all these TCL commands execute without any errors, we can get the lammps data file using: -
topo writelammpsdata <lammps.data> full
[Note: we can change the atom type to something else like atomic depending on the system, but mostly for molecular systems, full will be the way to go]
This is just a skeletal data file with the relevant bonds, angles, and dihedrals. No parameters are currently in this datafile. Sadly, the grunt work starts here. On paper, we can use the moltemplate package to get all the parameters, but it has never worked reliably (at least for me) . The work around is to visit the relevant force field parameter page on the moltemplate GitHub. For instance, for GAFF go to, https://github.com/jewettaij/moltemplate/blob/master/moltemplate/force_fields/gaff.lt and extract all the parameters (LJ, bond, angle, dihedral) from there. Remember to benchmark your LJ parameters.
Here is what a finished file will look like
module load ms/lammps/3Mar20-mpi
More Information to come
Talk to Orion if you have OpenMM questions!
Our group has a robust infrastructure for setting up and running molecular MD simulations. The core of this infrastructure is in the Pymatgen OpenMM IO add-on. Once you have installed that package and its dependencies, you should be able to use the following code. For a quick install of the dependencies, execute these commands:
First, you should import all of the functions and objects that we'll use below.
A pymatgen InputSet
contains all of the files needed to set up and run a simulation. Writing those files by hand would be incredibly tedious, so we use a Generator
to create the InputSet
. The Generator
is instantiated with simulation properties as input, like temperature and force field. Once it is created, it can generated many InputSets
with different sets of molecules.
An InputSet
is created by calling generator.get_input_set()
and providing the molecules (as smiles) and the density of the solution. When this happens, a ton of code is run behind the scenes that sets up the simulation.
We might not want to run our simulation immediately! For example, we might want to upload it to a supercomputer. We can easily do this by saving all of the files with input_set.write_input
.
It is equally easy to load an InputSet
from a directory.
An InputSet
can create a dynamic OpenMM Simulation
object. This simulation object can be run and modified using the OpenMM software package.
Once we have a simulation, we will first need to minimize the energy, otherwise the simulation box will explode. Once that's done, we can start the simulation!
Congratulations! You ran your first OpenMM simulation!
By default, an OpenMM Simulation
will not output any information. If we want data, we will need to add OpenMM Reporters
to write out the trajectory and simulation state. In the next cell, I attach the most useful reporters, StateData
, which writes state information, DCDReporter
, which saves the trajectory, and PDBReporter
, which saves the topology.
200 is an extremely short time interval, in reality, you would want to use something like 1,000 for the StateDataReporter and 10,000 for the DCD reporter. Next we can run the simulation and see that it ran. We are using a short time interval here because this is just a demonstration.
A real molecular dynamics simulation is slightly more complex than the toy examples we've covered here. We first need to equilibrate the pressure of the simulation, which will change the volume until the simulation contents are at atmospheric pressure. Then we need to anneal the simulation by heating it to 400K to make sure we find a true equilibrium. Fortunately, the Pymatgen code has very convenient functions for doing both of these things. A real simulation would look something like below (though in reality we would run it for many more timesteps).