GPU-accelerated N-Body particle simulator with visualizer

cuda, cuda-support, nbody, nbody-gravity, nbody-gravity-simulation, nbody-sim, nbody-simulation, nbody-simulations, particle-system, particles, particles-animations, simulations, sphere
pip install nbody==0.0.4



A GPU-accelerated N-body particle simulator and animator

Create complex particle simulations the easy way: a high-level package for designing and simulating large-scale particle interactions. Let nBody do the hard work for you!


Easy to use – and fast – nBody can simulate:

  • Gravitational acceleration
  • Coulomb interactions
  • Particle collisions

nBody is highly optimized:

  • GPU acceleration available via cupy
  • CPU multiprocessing with numpy
  • Energy conservation via the velocity-verlet algorithm

Animated matplotlib visualizations included for 2-D simulations. 3-D animations are also supported through the use of vpython.


Using numpy arrays, you will need:

  • An initial position array x with shape (N,p)
    • N is the number of particles
    • p is the number of dimensions
  • An initial velocity array v with shape (N,p)
  • An array of masses mwith shape (N,)
  • An array of charges q with shape (N,)
  • An array of radii r with shape (N,)

A possible configuration is as follows:

import numpy as np
x = np.random.normal(0,   10, (N,p)) # Positions
v = np.random.normal(0,    2, (N,p)) # Velocities
m = np.random.normal(8,    1, (N, )) # Masses
q = np.random.normal(0, 1E-6, (N, )) # Charges
r = np.random.normal(1,  0.1, (N, )) # Radii

m[m < 0] = np.abs(m[m < 0])
m[m == 0] = 1E-3

Next, pass these arrays in the given order to the spheres function, so as to create a new instance S of class System with the above conditions.

from nbody import *
S = spheres(x, v, m, q, r)

After selecting a simulation runtime T and (optional) time-step dt, use the solve method to calculate the particles' trajectories.

S.solve(T, dt)

If the system is 2-D such that p == 2, an animation can be created and saved to file; here, the filename quick_start is chosen, and will produce a file animations/quick_start.mp4.

animate(S, "quick_start")

If the system is 3-D such that p == 3, animations can be created but not saved to file – simply omit the string argument shown above, and no warnings will be raised.

Once the solve method has been called, it is also possible to save the System instance to file; in this case, the data will be saved to a directory saved/quick_start.

save(S, "quick_start")