Tagging particles for time-history?

[moisesae]

What would be the most straightforward way to tag N particles and have the time-history of their entire trajectories? I would like to avoid having to print out the data for every particle in my simulations. In the documentation, I see the <diag_name>.<species_name>.random_fraction and <diag_name>.<species_name>.uniform_stride but I'm not sure if this is quite what I need. Multiple_particles is nice but I have to write out their initial conditions in the input file which is tedious for large N (Unless there is a way to load in a text file that can be read).

[roelof-groenewald]

Hi @moisesae. The easiest way would be to create a second species that has the same properties as your actual species (same initial distribution, etc.), but with only a small number of particles and statistical weight = 0. You can then set the flag do_not_deposit = True for that species so that the particles do not impact the fields. You can then output all of the particles in that species at every step (or close to every step) to get the evolution of those "trace particles". Does that make sense?

[roelof-groenewald]

Btw, we recommend using GoupBased output when outputting frequent data for best IO performance.

[moisesae]

If the boundary conditions are periodic and I don't have a particle source, the particle velocity array at index i should correspond to the same particle throughout the entire simulation correct (vx_e[i])?

where vx_e = data[('electrons', 'particle_momentum_x')].to_ndarray() /m_e

[roelof-groenewald]

Generally no, since the particles are stored in "tiles" that correspond to the domain decomposition. So when a particle moves from one tile to another it's position in the array will change. However, each particle has a unique ID, which you can use to track a given particle through different diagnostic outputs. This is what the openPMD functionality to get trajectories uses. See https://openpmd-viewer.readthedocs.io/en/latest/tutorials/4_Particle_selection.html#Tracking-particles

[moisesae]

Thank you for the suggestion @roelof-groenewald. I tried making a new diagnostic output that looked like this

but I keep getting the following errors when trying to run the simulation:

Do you know why I would see such an error? The executable was compiled with cmake so the openpmd feature is automatically turned on.

[roelof-groenewald]

You need to specify the openpmd backend to use.
For hdf5 output:

diag_full.openpmd_backend = h5

or using ADIOS2:

diag_full.openpmd_backend = bp