To run your first pyrticle simulation, change to the examples/pic/ directory. To keep things simple, we will run a 2D simulation of a Gaussian beam in a beam tube. A parameter file for this simulation comes in the above-mentioned directory pyrticle distribution, it is called simple.py.
To first give you an idea of what is going to happen, run this simulation now. Type the following command into your shell:
$ python driver.py simple.cpy
You will see pyrticle perform a number of preparatory steps, and then, once the screen starts looking like this:
step=0 | t_sim=0 | W_field=2.94401e-06 | t_step=0.932571 | t_eta=0 | n_part=2000
step=1 | t_sim=4.02369e-11 | W_field=2.94297e-06 | t_step=0.934013 | t_eta=820.506 | n_part=2000
step=2 | t_sim=8.04738e-11 | W_field=2.94239e-06 | t_step=0.569611 | t_eta=643.318 | n_part=2000
step=3 | t_sim=1.20711e-10 | W_field=2.94206e-06 | t_step=0.66167 | t_eta=611.747 | n_part=2000
step=4 | t_sim=1.60948e-10 | W_field=2.94188e-06 | t_step=0.704858 | t_eta=604.52 | n_part=2000
step=5 | t_sim=2.01185e-10 | W_field=2.94188e-06 | t_step=0.693117 | t_eta=614.649 | n_part=2000
step=6 | t_sim=2.41422e-10 | W_field=2.94205e-06 | t_step=0.860489 | t_eta=618.254 | n_part=2000
...
pyrticle is performing the actual timestepping. simple.cpy tells pyrticle to write out a visualization file every 10 timesteps, so after a hundred or so timesteps (which should take less than a minute on a modern computer), you will have a few files to look at. We suggest that you download VisIt, and play around with visualizing them.
Next, let us look at the anatomy of the simulation control file. It is broken up into a few sections. We will look at each section in turn. Please open simple.cpy in an editor.
The first section of the control file is labeled “PIC setup”, and contains just that:
# -----------------------------------------------------------------------------
# pic setup
# -----------------------------------------------------------------------------
pusher = PushMonomial()
reconstructor = RecShape()
dimensions_pos = 2
dimensions_velocity = 2
final_time = 10*units.M/units.VACUUM_LIGHT_SPEED
vis_interval = 10
It sets up the solver’s dimensionality, which deposition and push methods it should use, how long it should run, and how often it should dump the visualization output to disk.
Next up is the mesh setup:
# -----------------------------------------------------------------------------
# geometry and field discretization
# -----------------------------------------------------------------------------
element_order = 3
_tube_width = 1*units.M
_tube_length = 2*units.M
import hedge.mesh as _mesh
mesh = _mesh.make_rect_mesh(
[0, -_tube_width/2],
[_tube_length, _tube_width/2],
periodicity=(True, False),
subdivisions=(10,5),
max_area=0.02)
Here we observe the first use of user variables. pyrticle wants to interpret every setting in a control file as a simulation parameter. If it does not something you are setting, it will stop with an error. If a setting starts with an underscore (like _tube_length above), however, this setting becomes a user variable and is ignored by pyrticle. This way, you can perform your own symbolic calculations in your control files.
Apart from user variables, this section is reasonably straightforward–it sets up an x-periodic rectangular domain.
The final section in the control file is concerned with setting up a particle distribution:
# -----------------------------------------------------------------------------
# particle setup
# -----------------------------------------------------------------------------
_cloud_charge = 10e-9 * units.C
nparticles = 2000
_electrons_per_particle = abs(_cloud_charge/nparticles/units.EL_CHARGE)
_pmass = _electrons_per_particle*units.EL_MASS
_c0 = units.VACUUM_LIGHT_SPEED
_mean_v = numpy.array([_c0*0.9,0])
_sigma_v = numpy.array([_c0*0.9*1e-3, _c0*1e-5])
_mean_beta = _mean_v/units.VACUUM_LIGHT_SPEED
_gamma = units.gamma_from_v(_mean_v)
_mean_p = _gamma*_pmass*_mean_v
distribution = pyrticle.distribution.JointParticleDistribution([
pyrticle.distribution.GaussianPos(
[_tube_length*0.25,0.0],
[0.1, 0.1]),
pyrticle.distribution.GaussianMomentum(
_mean_p, _sigma_v*_gamma*_pmass,
units,
pyrticle.distribution.DeltaChargeMass(
_cloud_charge/nparticles,
_pmass))
])
As you can see, it makes extensive use of user variables and finally sets up a Gaussian distribution in position and momentum, coupled with delta distributions in particle charge and mass.
More documentation on the details of how control files are written will become available as time permits, for now we invite you to tinker and ask questions. If something is unclear, please do not hesitate to get in touch.
Thanks for trying Pyrticle!