RMG-database
RMG-database copied to clipboard
ReactionMechanismGenerator
Replace an over-estimated rate rule in R_Recombination
Don't necessarily merge this PR yet - there may be a better way to address this. We just wanted to see what would happen.
To address issue https://github.com/ReactionMechanismGenerator/RMG-database/issues/353 in which it is suggested this rule was auto-generated using poor thermochemistry, and ended up much too fast, we have replaced it with the previous value for this reaction, https://rmg.mit.edu/database/kinetics/families/R_Recombination/training/40/ Apparently taken from entry: C9H7_19 + H_15 <=> indene_25 from kinetics library: kislovB This is pretty much the collision limit. (10^13 cm3/mol/s)
@mjohnson541 Is this the recommended way to address these kinds of issues with auto-generated families?
I mean improved thermochemistry libraries and estimation is the real solution, what I thought this was when I glanced at it was adding a new training reaction, which I think is the next best solution. Currently modifying the rate rules isn't preferable because they'll simply get overridden and disappear whenever I retrain the tree.
If this isn't the preferred approach, how should we fix this? It would be best to not ship 3.0 with a known bad rate rule.
I agree it's silly to leave the known problem rate in the next release when this patch is simple.
I also agree that we should fix this properly, but perhaps open a new issue to do that. That issue could suggest things like
- removing the "bad" training reaction
- adding one or more new training reactions
- adding some automated tests that rates generated from reversing training reactions end up "reasonable"
- ...
@rwest the rate that this PR fixed has since been re-autogenerated elsewhere. Could you take a look and see if this fix is still required?
