Ticket #65 (new defect - convergence)
chemistry solution is greatly disturbed for no apparent reason
| Reported by: | gary | Owned by: | nobody |
|---|---|---|---|
| Priority: | critical | Milestone: | newmole onto trunk |
| Component: | chemical network | Version: | |
| Keywords: | chemistry solver bonkers | Cc: |
Description
the attached bug.in falls over. with trace convergence print turned on we see that the solver has happily found that a solution is somewhere around 9200K. the chemistry falls apart when the solver changes modes.
bug.out shows what happened. Through line 130 of the output things are on track. The print at line 132 occurs at line 570 of conv_init_solution.cpp.
the trouble starts during the call to ConvPresTempEdenIoniz? at line 616 to conv_init_solution.cpp. This happens at a temperature of 9216K at printed at line 135 of the output - the previous temperature at line 130 was identical to all printed figures so T may not have changed at all. A series of chemistry failures occur which are fatal.
something really kicked the chemistry network at this point even though the external conditions appears to have not changed.
is it possible to see what rates going into the chemistry changed by large amounts? this does not appear to be something subtle - the solver totally jumps the rail.
this does not occur if the large H2 is turned off - could there be comething from the H2 chemistry done by the large H2 model that has not been evaluated but suddenly is?
