cudaMPCD

Fast hydrodynamic solver using the method of multi-particle collision dynamics MPCD. Python frontend using pybind11 with backends implemented in C++ and CUDA for usage on CPU and GPUs. In the current state, this code can be used to simulate a Poiseuille flow, i.e. a flow between to parallel plates. This can be used for viscosity measurements. The code has implementations of 2 collision operators: standard stochastic rotation dynamics (SRD) and extended MPC with non-ideal equation-of-state. The current implementation was tuned for older GTX 2080Ti and Quadro RTX 6000 cards.

Usage

After installation, you can import and use the module in Python:

import pympcd

# Create simulation parameters
params = pympcd.Params()
params.n = 10
params.temperature = 1.0
params.volume_size = (10, 10, 100)
params.periodicity = (1, 1, 0)
params.drag = 0.001
params.delta_t = 0.02
params.experiment = "standart"
params.algorithm = "srd"

# Create and run simulation
sim = pympcd.Simulation(params, "cuda")
sim.step(1000)
sim.step_and_sample(100000)

# Get results
density, velocity = sim.get_mean_fields()

Use algorithm = "srd" for standard stochastic rotation dynamics, use algorithm = "extended" for extended MPC with non-ideal equation-of-state. The Poiseuille flow used here, is a geometry for which the Navier-Stokes equations can be solved analytically. Thus, we can use this geometry for testing the code. After the simulation, we may use the flow field data as follows:

import matplotlib.pyplot as plt
import math

# Calculate velocity profile
x_velocity = velocity[:,:,:,0]; 
vel_profile = x_velocity.mean(axis=(0,1))

plt.plot(vel_profile) # will show a parabola
plt.xlabel('channel cross section')
plt.ylabel('flow speed')

# Simulation parameters:
L = params.volume_size[2] - 2 #subtract walls
g = params.drag
n = params.n
dt = params.delta_t

# Viscosity measurement:
eta = L * L * n * g / (8 * vel_profile.max())

# Viscosity theoretical for SRD algorithm:
eta_theo = (1 - math.cos(120/180*math.pi)) / (6*3*dt) * (n - 1 + math.exp(-n))

print( "SRD Viscosity; theoretical: ", eta_theo, ", measured:", eta ) 

The last line prints the two values of the fluid viscosity "SRD Viscosity; theoretical: 37.50, measured: 37.27"

Installation

Prerequisites

Required:

• CMake (version 3.12 or higher)
• C++ compiler with C++14 support or higher

Optional:

• CUDA-capable GPU and NVIDIA CUDA Toolkit
• Python 3.6+ with development headers
• pybind11
• NumPy

Building and Installing

1. Clone the repository:

git clone https://github.com/azantop/cudaMPCD.git
cd cudaMPCD

2. use pip to install the package:

pip install .

3. For running simulations with standalone binary use Cmake:

mkdir build
cd build
cmake ..
cmake --build .
cmake --install .

If you find this repository helpful, please consider citing our article (doi.org/10.1063/5.0037934)

@article{zantop2021multi,
  title={Multi-particle collision dynamics with a non-ideal equation of state. I},
  author={Zantop, Arne W and Stark, Holger},
  journal=JCP,
  volume={154},
  number={2},
  pages={024105},
  year={2021},
  publisher={AIP Publishing LLC}
}