Thanks for reply. I also wrote a PDE to simulate the setting, it gives completely different results. I expect the middle region without fixed charged should not be empty. Also, the free ions should not stack around the fixed charges, it should form a relatively smooth decreasing curve away from the ...

Thanks for replying. I just ended my travel. Your suggestions are very helpful, I changed the cutoff distance, and used latin hypercube for sampling the initial positions, before simulation I also minimize the energy. These changes allow the system to be simulated in 0.01 ps timestep. However, the r...

Hi Peter, Thanks dor reply, I am sure I created a state before as state = simulation.context.getState(getForces=True) Now I change the sample code according to your mention. length = 1e-8 num_particles = 50 rho_perm_pos = np.random.rand(36, 3) * length # 36 fixed ions at x = 0 rho_perm_pos[:,0] = 0 ...

Sorry about being confusing, please see below, Initialize the positions import numpy as np from simtk import openmm as mm from simtk.openmm import app from simtk.unit import nanometer, picosecond, angstrom, kilojoule_per_mole, kelvin, kilocalorie_per_mole import openmm.unit as unit import time from ...

Thanks for replying. The physical system I tried to simulate is 'a cubic system (L = 1e-8m), with 2 layer of fixed charges at x = 0 and x = L/2, and certain number of free ions '. There are two approaches I implemented, one is to use openmm to calculate the PME coulombic and LJ forces and use python...

Thanks for replying. The values of positions are from 0 to 1e-8m, I transfered them to the unit of angstrom. Thats why I multiple them with 1e+10

Hi Peter, How are you, I have now looked back on this problem. And recently I found some problems about the computation. box_size = 100*angstrom # define a new system system = mm.System() system.setDefaultPeriodicBoxVectors((box_size, 0, 0), (0, box_size, 0), (0, 0, box_size)) # define the ion numbe...

Hi Peter, Thanks for your replying, now the code can run very fastly. However, the results did not match with my PDE simulation, I am not sure which part was wrong. So I am trying to implement all the Brownian dynamic simulation (including LJ force and random diffusion) via Openmm. My problem settin...

My system size is 1000 angstrom actually, which is equivalent to 100 nm. My code can be changed to that size but no difference showed. I changed my code alittle bit, and it can run very fastly, only <0.1s, but I am not sure it make sense or not. I changed the size to 100A, cutoff to 10A, tolerance t...

Hi Peter, Thanks for replying. I am timing the whole script but certainly the simulation.context.getState(getForces=True) is the only time consuming line. I use a normal PC to run this code, of which cost 2.5 second to complete a Pymatgen ewald evaluation. Openmm is written in C++, it should be much...