alongd
commented
6 years ago Looking more into this, the above examples do not fall under the same family recipe.
For the sulfur case we have [::SH][::SH] <=> [:SH2]=[::S]
while the nitrogen case could either be described as [:NH]=[:NH] <=> [:NH2][::N], or more correctly [:NH]=[:NH] <=> [NH2+]=[::N-] (where the products here are resonance structures of each other, the second being the representative one according to the octet rule)
(the above was written using augmented-SMILES which haven't been invented yet :smiley:
: represents a lone pair, . would represent a radical site)
This is problematic because in the sulfur case the bond order increases, while in the nitrogen case it either decreases or remains the same for the same reaction direction. (the charge separation isn't an issue.)
I don't think we should create families with just one reaction in each. The alternative would be to teach RMG to use high-P limit isomerization\association\dissociation library reactions if available.
To illustrate this problem for one case: RMG finds the network HSS + H <=> HSSH, but it won't further expand it to HSSH <=> H2SS <=> H2S + S
We should teach RMG about transitions like
HSSH <=> H2SSandHNNH <=> H2NNIn both cases H migrates internally, but there's no radical site to "abstract it", so this is different than theintra_H_migrationfamily. The products are closed shell species: H2SS and H2NN have two lone pairs on the S and N not connected to the hydrogens, respectively.If this is only added as library reactions, we're missing it out in PDep networks.