Cloudy Torsten

10 Nov 2004

Torsten is still working on the temperature spike. He has been investigating variations on the basic model:

• changing the magnetic field,
• changing the density at the illuminated face
• changing the velocity

to see how that changes the temperature spike. The spike seems to be reasonably robust to changes in the parameters.

Reducing the number of zones

Since the models take a long time to run, we are interested in cutting down the domain that we have to calculate. To that end, we are trying to use the transmitted continuum just past the ionization front in a static model as the input continuum for the dynamic models.

We have had problems with matching up the temperature of the two models at the boundary. Initially this was due to the Lyman lines being optically thin close to the illuminated face of the continuation model. We have solved that by using

case B no photoionization 

However, there is still a difference between the T of the 2 models. Torsten is now trying to rerun the constant-pressure model to make sure it is consistent with our continuation models, but he has had problems with drad becoming too small. Setting drmin 13 caused the model to crash, so he is now trying again with drmin 12.

Refining the zoning logic

Problem

Currently, the code does not take account of zoning decisions that were made during the previous iterations when deciding how big the zones should be. As a result, sometimes we have a nice-looking solution at one iteration but the next iteration decides to use much larger zones and ends up "stepping over" an important feature (like the temperature spike).

Proposed solution

Since we know what the previous iteration's structure is, we can use it to limit the size of the zone in the current iteration. Say, allowing no more than 1.5 times the size of the previous drad at the same position.

Save/restore mechanism

Gary is supposed to be working on this.

11 Nov 2004

Reducing the number of zones

We have finally got this to more-or-less work. We had a lot of problems with the fact that there is an undocumented factor of 4π difference between the values that are written by punch transmitted continuum and those that need to be specified when giving the incident spectrum intensity in the init file.

So, now we can run models of just the PDR part with about 300 cells that run in about 1.5 mins per iteration. Strangely, the first one of these to run does not show the temperature spike, but rather a gentle bump. However, the starting temperature is still slightly different (higher) from that of the full model, so Torsten is now experimenting with small variations in the incident flux to see if he can get one that starts at the same temperature as the models that showed the spike.