Opened 8 years ago

Closed 6 years ago

Last modified 5 years ago

#118 closed defect - convergence (fixed)

wild oscillations in ionization and temperature on second iteration, does not abbort as dr -> 0

Reported by: Gary J. Ferland Owned by: nobody
Priority: blocker Milestone: c13 release
Component: ionization convergence Version: trunk
Keywords: Cc:


The HII region model reported by Valentina has major problems with the trunk, although it runs fine with c08. There are two separate problems.

The major problem is a very large osciallation in the ionization and resulting temperature on the second iteration. This sets in by zone 3 or 4, and the H0 fraction oscillates between ~1e-7 and ~1e-13 in adjacent zones.

In response to this oscillation the code makes the zones thinner. It very quickly reaches the smallest possible zone thickness where the code used to abort. This abort feature appears to have been removed from radius_next by Ryan in R3621 - as a result the simulation continues with the zone thickness of 1e-36 cm until the limit of 1400 zones is hit. This takes more than an hour on this mac. Had the abort been left in the sim would have failed in ten minutes rather than more than an hours.

We need to get the abort feature back in. Why was this removed? is it a simple matter of reverting R3621?

Next problem - the oscillation. Have not debugged it yet but this has all the hallmarks of the classic Lya oscillation. There were dampers in the code to prevent this, and I know that some were removed.

Attachments (1) (38.1 KB) - added by Gary J. Ferland 8 years ago.
valentinas model with stop zone to prevent 90 minute execution time

Download all attachments as: .zip

Change History (4)

Changed 8 years ago by Gary J. Ferland

Attachment: added

valentinas model with stop zone to prevent 90 minute execution time

comment:1 Changed 6 years ago by rjrw

Runs through without problems @ r5801 both on trunk and newmole.

comment:2 Changed 6 years ago by Ryan Porter

Resolution: fixed
Status: newclosed

Still runs through with no problems at r6328. Closing ticket.

comment:3 Changed 5 years ago by Ryan Porter

Milestone: C10 branchc13 release

Milestone C10 branch deleted

Note: See TracTickets for help on using tickets.