You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
hPF-MD.jl => RobertoMD.jl due to the naming policy of Julia general package.
The roadmap below was originally intended to write a toy code for testing new algorithms based on the particle field scheme.
The goal of this package is now to become a production simulator for particle-field simulations, since the performance of Julia and the parallelization capability of MPI.jl are quite satisfactory.
The support force-fields are limited. Next tasks will be implementations of:
Angle
Dihedral
Electronics
Spectral method to solve fields [Bore, S. L.; Cascella, M. J. Chem. Phys. 2020, 153 (9), 094106.]
Improvements on the formulation of the particle-field interactions
Improvements of hPF-MD.jl:
benchmark this serial version against OCCAM or GALAMOST (MD-SCF) using the simple CG-polymer systems
implement harmonic bond interactions ✅
implement Andersen thermostat ✅
implement unwrapped trajectory writer ✅
implement the auto-differentiation for determination of the force and pressure tensor from energy function.✅
The implementation of auto-differentiation is okay-ish but the computational consumption is very high.
add functionality for modeling multicomponent systems
implement the spectral method for the calculation of density gradient with convolution of Gaussian filter ✅
general step:
construction of wave vectors using: k=fftfreq(N_grids,2pi/(dcell))
forward transform density grids: rho_=fft(rho)
convolution with gaussian function: rho_conv=rho_*exp(-|k|^2)
backward transform density grids: rho=real(ifft(rho_conv))
parallelization (may use MPI not MP) ✅
using particle decomposition algorithm with MPI.jl
add pressure function via incorporating other equation of state of liquids.
Future work of hPF-MD method:
replace the physics-based potential form (Flory-Huggins Interactions Energy) with numerical form: e.g. train a neural network using the local density grids as features to predict the potential energy and forces.
automatic determination of the interaction parameters for coarse-graining
The text was updated successfully, but these errors were encountered:
hPF-MD.jl => RobertoMD.jl due to the naming policy of Julia general package.
The roadmap below was originally intended to write a toy code for testing new algorithms based on the particle field scheme.
The goal of this package is now to become a production simulator for particle-field simulations, since the performance of Julia and the parallelization capability of MPI.jl are quite satisfactory.
The support force-fields are limited. Next tasks will be implementations of:
Improvements of hPF-MD.jl:
Future work of hPF-MD method:
The text was updated successfully, but these errors were encountered: