Hands-on, part 2

Each group are assigned six simulations to carry out. The simulations will be used for post-processing in session 4 (03.02.22), so complete all six simulations before then. Use the Zoom breakout rooms assigned during the session to agree on who does which simulations. For your own learning, make sure that each member run at least one simulation (each simulation can be run by several members if you wish). The simulations are a bit time-consuming, so start early.

Below is a description of the sphere and cylinder assignments (depending on which one you’re assigned). After that follows some questions to be answered during the Zoom event, before instructions on how to run the simulations.

Sphere assignment

Neglect the effect of the magnetic field of the Earth. Instead of simulating true atomic oxygen, which has a mass of 16 amu and is the predominant species of the F-layer, we will use a reduced ion mass of 1/16 amu in order to speed up computations. A representative orbital velocity is 7000 m/s, but increase this so you preserve the Mach number when reducing the ion mass. Carry out five simulations of the sphere, where

it is biased to 1V
it is biased to 2V
it is biased to 3V
one half is biased to 1V and the other to 3V
it is left at the floating potential

Save diagnostic output every 100th timestep.

Then, create a new mesh where the sphere has a radius of 5 debye lengths and repeat the third simulation. Save diagnostic output every 10th timestep.

Cylinder assignment

Neglect the effect of the magnetic field of the Earth, and to begin with, also neglect the effect of the drift velocity by setting it to zero. Instead of simulating true atomic oxygen, which has a mass of 16 amu and is the predominant species of the F-layer, we will use a reduced ion mass of 1/16 amu in order to speed up computations. Carry out five simulations of the cylinder, where

it is biased to 1V
it is biased to 2V
it is biased to 3V
one half is biased to 1V and the other to 3V
it is left at the floating potential

Save diagnostic output every 100th timestep.

Then, repeat the third simulation, but with a drift velocity perpendicular to the cylinder. A representative orbital velocity is 7000 m/s, but increase this to preserve the Mach number when reducing the ion mass. Save diagnostic output every 10th timestep.

Questions for Zoom discussion

In the Zoom break discuss the following questions with your group, and try to answer to arrive at an answer.

How long will it take to reach steady-state, i.e., how long should the simulation run (in microseconds)? [Suggestion: 10 us]
How many simulation particles would you use? [Suggestion: 5 million of each species]

We will drop by your room and see how your doing, as well as provide you with the numbers you are actually to use in the simulations.

Setting up a new simulation

First, make sure you’re conda environment is active:

conda activate ptetra

For each simulation, you need to create a subfolder inside PTetraWorkshop, for instance:

$ mkdir Cylinder_0.2R_10L_1V_1V_0kms

(recall the naming conventions in the introduction. You will need to run PTetra from within this folder, so create a symlink to the PTetra executable inside the folder (symlinks, or symbolic links, are like shortcuts in Windows, but behave better):

$ cd Cylinder_0.2R_10L_1V_1V_0kms
$ ln -s ../MPI_V50i/mptetra

It is important that you execute ln from with the target folder when you use relative paths such as ../MPI_V50i/mptetra. Otherwise the symlink will point to the wrong relative path.

Next, PTetra will look for the mesh (converted to topo-format using msh2topo in part 1) in the file meshpic.dat. Create a symlink so PTetra can find it:

$ ln -s ../Geometry/cylinder_0.5R.topo meshpic.dat

Finally, PTetra also needs the input file pictetra.dat. You can copy the file from the example directory already in the repository:

$ cp ../Sphere_0.5R_1V_1V_112kms/pictetra.dat .

Next, you need to configure the simulation parameters in this file with the text editor of your choice in accordance to the simulation you are running. Documentation is available in the file, but you do not have to make any changes to the injection parameters, solar parameters, charge exchange parameters or reflection parameters. Pay attention to the plasma parameters.

Under simulation control parameters and numerical parameter, the number of simulation particles and the physical end time of the simulations needs to be edited according to the answers of your Zoom questions. The frequency of diagnostic output should also be edited according to your assignment’s description. Other parameters under these categories are fine.

In addition, to specify fixed potentials on the two halves, you must set sc_fixedpotentials=.true.. Then, the voltages are specified in the list sc_fixedPot. The voltages in sc_fixedPot are assigned to the physical groups in ascending order. For the floating potential simulation, you must set sc_fixedpotentials=.false. in order to let the potentail be self-consistently determined from the charging due to the plasma. In this case, sc_nnetBias and netBias is used to control the relative voltage between the two halves, which in our case, is to be zero. Since the two halve constitutes one circuit, sc_nnetBias=1. netBias is as follows:

$begin netBias
nnodes=2
1  0.0
2  0.0
$end netBias

This means that if the first physical group is zero, then so is the second group. It is only the relative voltages between the elements in netBias that matters.

You can now start PTetra:

$ ./mptetra

From the output of PTetra, you should be able to answer (to yourselves):

What is the surface area of the physical surfaces created in Gmsh?
What is the statistical weight of the simulation particles? That is to say, how many physical particles does each simulaion particle correspond to?

Starting over

If a simulation for some reason is erroneous, you should delete the output files:

$ rm *.vtk *.rdm *.topo *.hst fort.20

and then start over. If the mesh is incorrect, you need to replace meshpic.dat with a symlink to the correct mesh, and also remove alujlu.prcnd so PTetra can recreate it. alujlu.prcnd stores the preconditioning matrix, and is mesh-dependent.

Continue the simulation later

Since the simulations are a bit time-consuming, you may need to stop the simulations, and continue later. To stop PTetra gracefully, enter the simulation folder and create the file .quit:

touch .quit

PTetra will complete the time-step, and save restart files named pictetra<timestep>_<processor>.rdm.

To continue the simulations, remove the .quit file:

rm .quit

Change the entry restartfrom in pictetra.dat to point to the respective restart file(s) and start PTetra as usual.

Inspecting the simulations

While or after running simulations, it may be useful to run some diagnostics to ensure that the simulations are correct. Use the attached script plot.py to plot the current collected by the object, e.g.,

./plot.py Sphere_0.5R_1V_1V_112kms --OML
./plot.py Cylinder_0.2R_10L_1V_1V_0kms --FL

The --OML and --FL flags allow you to compare the simulations with OML theory (most relevant for the spherical assignment) or finite length theory (for cylinders), respectively.