How to set specific boundary conditions

[EricAComstock]

Hello All,

In the config file, is there any way to set specific boundary conditions for the incoming ions on the sides of the simulation grid? And can an external electromagnetic field be imposed? My problem has nonuniform boundary conditions, which is why I'm asking.

Sincerely,
Eric Comstock

[JonasSuni]

Hi Eric,

Could you give some more details regarding your problem setup, so we can try to figure out if it's possible or not?

Best regards,
Jonas

[EricAComstock]

Hello Jonas,

Thank you! I'm just wondering if I can simulate a plasma shock with a quasi-neutral plasma of high-mass ions (30 amu) with gamma = 7/5, with an incoming Mach number of M_1 = 2. This is mainly a test problem that I would like to use for both my own learning and for profiling the code.

Sincerely,
Eric Comstock

[JonasSuni]

Hi Eric,

In this case you want to start with the configuration and input file in this sample directory: https://github.com/fmihpc/vlasiator/tree/master/samples/IPshock

The config options under [boundaries] specify which boundaries are periodic, if any, and which other boundary types are used in the simulation. The [maxwellian] and [proton_maxwellian] specify the inflowing plasma, and [outflow]/[proton_outflow] specify the outflow boundary. The initial upstream and downstream magnetic field, density, temperature, and velocity are defined under [IPShock]/[proton_IPShock]. You will need to calculate these yourself for your case.

Remember to rename all the proton_ configuration options if you change the name of the population when you change its properties.

Best regards,
Jonas

[EricAComstock]

Hello Jonas,

Thank you! I got the baseline working, so I'm trying to make edits to match my problem. May I ask what the units of the timestep are?

And what does the error code (-6) mean?

Sincerely,
Eric Comstock

[JonasSuni]

Hi Eric,

Everything in the config is in SI units, unless otherwise specified (for instance the proton mass and its unit). So dt and time are in seconds.

Regarding the error, could you give some more details? Did the run produce a logfile.txt? Was there any other output?

Best regards,
Jonas

[EricAComstock]

Hello Jonas,

Thank you for the information!

No logfile.txt was produced - just the error. I'm trying to simulate a problem with relatively small size and high plasma density, so it likely was a timestep issue - 1 second was too long. I'm using ionospheric conditions for my pre-shock plasma, where the density is ~9e10/m^3, and running it over a volume of only 20 m on a side, so the plasma would only take ~0.02s to cross from one side to the other.

Sincerely,
Eric Comstock

[JonasSuni]

Hi Eric,

Could you post the configuration file, inflow file, and command that you used to launch the simulation so we can try to figure out what's going on?

Best regards,
Jonas

[EricAComstock]

Hello Jonas,

Thank you! I named my edits "ex4shock," since it is example four in my paper.

The command I run (from the vlasiator/samples/ex4shock folder) was

../../vlasiator --run_config ex4shock.cfg

I attached the file below - it is supposed to model a plasma shock in a small region of the Earth's ionosphere. I had to change the extension to be able to attach it.

Sincerely,
Eric Comstock

ex4shock.txt

[alhom]

Hi Eric!

You have the top-level parameters wrong in your config.

project is used to choose from the project setups in the folder https://github.com/fmihpc/vlasiator/tree/master/projects. You should use IPShock instead (and replace ex4shock in the config elsewhere as well!). Setting project gives you access to the respective config options for that project, such as [IPShock] and [proton_IPShock].

dynamic_timestep, propagate_field, propagate_vlasov_acceleration, propagate_vlasov_translation are all Boolean values - set these to 1 for enabling the dynamic timestep (that will take into account CFL limits etc. automatically) and the solver portions you wish to use (which likely is all of them! :) )

While you can name your particle populations as you like, I would still suggest using proton for protons - that is the default population for Analysator tools, so it will save you a bit of trouble later.

Further, under [io] all the intervals are in seconds, so you will want to adjust those to a cadence that fits your simulated time.

Best wishes,
Markku

[EricAComstock]

Hello Markku,

Thank you! Since I'm in the ionosphere I'm using a single ion species with mass 30 amu, to represent the N2+, NO+, and O2+ ions that are common in that band. IPShock by itself runs correctly, but I'm trying to modify it to get a different plasma shock case.

In [io], are the terms in seconds of real time or seconds of simulated time?

Sincerely,
Eric Comstock

[alhom]

Hi

using a single ion species with mass 30 amu

Ah! Carry on, then! :)

Under [io], the times are in simulated time, except for the specified walltime interval, which is in real time (to e.g. place restarts at the end of a job slot with a given length).

-Markku

[EricAComstock]

Hello Markku,

Thank you!

Currently, I'm getting the error

(MAIN): Completed grid initialization.
[aewlgst04094d:03889] *** Process received signal ***
[aewlgst04094d:03889] Signal: Segmentation fault (11)
[aewlgst04094d:03889] Signal code: Address not mapped (1)
[aewlgst04094d:03889] Failing at address: 0x10
[aewlgst04094d:03889] [ 0] Segmentation fault (core dumped)

What does this mean?

Sincerely,
Eric Comstock

[alhom]

This looks to be because of misconfigured boundaries - you have specified +-x boundaries as inflow and outflow, but y and z boundaries are not set and not set as periodic either. Congratulations, you have found something(s!) we should add a sanity check for! :)

Setting

[boundaries]
periodic_x = no
periodic_y = yes
periodic_z = yes
boundary = Outflow
boundary = Maxwellian

lets the simulation go forward into the main loop, but it will crash there promptly for NaN/inf floating point values. This suggests double-checking that all the physical input values are compatible.

Some suggestions to this end:

• use the VDF input inspection tool https://github.com/fmihpc/vlasiator/blob/dev/tools/vdf_params.py to probe that your sparsity thresholds and velocity space extents and resolution are sane, including the Maxwellian cutoff value (maxwCutoff)
  • this is further pointed at by zero-block counts in diagnostics.txt. Check that your VDF is sufficiently resolved by your velocity space.
• in fieldsolver, check the inflow (adiabatic) electron density and temperature to be compatible
• The config-checker script https://github.com/fmihpc/vlasiator/blob/master/tools/check_vlasiator_cfg.sh is worth using regularly, although at this point it shows the config to be clean (checking for input parameters being accounted for).
• your IPShock.Width is much larger than the spatial extents. I doubt this should be the root cause of this particular error, however, but the results would definitely not be what you expect :)

[EricAComstock]

Hello Alhom,

Thank you! I managed to get it working. Is there any way to generate graphs of the results or check convergence?

Sincerely,
Eric Comstock

[alhom]

Great! See analysator: https://github.com/fmihpc/analysator and its guides http://fmihpc.github.io/analysator/ for Python-based plotting and data handling of vlsv files, or the https://github.com/fmihpc/vlsv library for its VisIt plugin. If you prefer ParaView, analysator does cobtain a plugin but using it involves installing analysator to the ParaView environment as well.