C-C Beta scissions with biradicals lead to new biradicals instead of alkenes

[nickvandewiele]

Ring opening reactions by a unimolecular C-C scission lead to structures having two unpaired electrons, biradicals. If these are 1,4 biradicals, then these species quickly undergo a beta scission of the C-C bond in beta position of each radical.

E.g. 1,4-butadiyl leads to ethene + ethene.

since these types of species have an unpaired electron with a sigma bond in beta position, they are also eligible to undergo a "normal" beta scission.

So RMG would perform: 1,4-butadiyl = ethene + 1,2-ethadiyl which makes perfectly sense to RMG.

Both reactions are actually one and the same reaction, namely the beta scission reaction family.

@gmagoon circumvented this inconvenience by creating a new reaction family 1,2-Birad_to_alkene which would turn 1,2-ethadiyl into ethene via a "genuine" elementary reaction. This would have as a side effect that all possible double bonds in every compound would suddenly be eligible to undergo this reaction family. With desastrous results in terms of CPU time, but also quality of the generated models, since some of the newly created biradicals via this reaction would actually be added to the core due to inaccurate kinetics or thermochemistry.

The real issue was of course that the recipe for the "normal" beta-scission would create a unpaired electron for the atom whose sigma bond was being broken. In case of 1,4-biradicals, this recipe would not create the alkene structures but a new 1,2-biradical instead.

One solution would be to check whether the resulting product of a reaction is a 1,2-biradical and transform it into a alkene. But then I am maybe forgetting possible reactions where 1,2-biradicals in triplet are stable enough to exist.

Is this a typical example of where separating the data ("a new reaction family") from the model ("how do we transform reactants into products?") becomes difficult and problem-specific?

[nickvandewiele]

NB: this a probably a good case study for RMG-Py too.

[rwest]

Interesting. I think it probably is a good example of why separating the chemical rules from the computer code is tricky. (The unachieved vision of letting chemists easily update the database without needing to understand the java).

But what do we do? Let each reaction family be its own class with arbitrary methods, and let the chemists learn some Python? We're thinking this type of idea may be inevitable for the transition state location project, where each family needs a totally different approach. But that's probably a big-picture question for discussion on RMG-Py.

Here on RMG-Java (and in the short-term on RMG-Py):

Forbid biradicals from reacting in the normal beta-scission family and make a new family specifically for them, that goes straight to two alkenes?

[nickvandewiele]

"and let the chemists learn some Python" well I feel that's going to be a large barrier for a lot of them in the field.

[nickvandewiele]

My approach for dealing with the biradical beta scission was to treat 1,2-biradical species as an artefact, and checking for these kinds of structures right after they have been created.

Of course, after this was done, I found a complication.

Consider the C-C scission of 1,5-pentadiyl = propadiyl + ethene

The reverse reaction: propadiyl + ethene = 1,5-pentadiyl

is actually a radical addition to a double bond, rather than a radical recombination. So the reverse reaction family for the reaction family Bond_Dissociation is the radical addition, and not the radical recombination!

This means that an automatic perception of 1,2-biradicals and immediate conversion into alkenes actually mixes up the strict separation of reaction families, and the straightforward creation of the reverse reaction based on the recipe of the forward reaction family.

Still, I feel that 1,2-biradicals are artificial, in the sense that they, most of the times, are not representative for the structure or stability (thermochemistry) of a species. They should therefore actually never be observed by users and treated internally.

[rwest]

If the C-C scission of 1,5-pentadiyl to give propadiyl and ethene is correctly predicted by the "radical addition to a double bond" family, then the route via 1,2-ethadiyl should just be forbidden?

Could/should we: 1) accurately predict the energy of 1,2-biradicals to be very high, so they don't show up in the core and don't cause problems. 2) forbid them entirely, so we don't waste effort thinking about them on the edge and including them in PDep reaction networks.

Perhaps this all became messy when we decided O₂ is a 1,2-biradical.

[nickvandewiele]

1) you' re right about the thermochemistry for 1,2-biradicals: they will have high energy due to the double addition of a HBI, but this completely ignores the fact that both unpaired electrons are next to each other. Anyways, in my JP10 models, I do see 1,2-biradicals in the core.

2) maybe forbidding 1,2-hydrocarbon biradicals completely is the best approach if the equivalent pathways via alkenes are added anyways via another reaction family. However, currently I am not sure whether this is always the case.

3) my artefact removal patch (cf earlier post) works fine in the sense that RMG does not crash and now generates alkenes instead of 1,2-biradicals. However, I now see a lot of "ERROR: Couldn't find the rate constant for reaction ..." which probably reflects the fact that the reverse reaction now suddenly belongs to another reaction family. BTW, this type of error silently continues RMG, and eventually does not add these reactions to the edge. So, this might not be the best approach neither...

[jwallen]

I think the first thing to do is to get the thermo of the 1,2-biradicals correct. Perhaps we need another set of groups that contain corrections for biradicals (maybe just 1,2-biradicals to start, perhaps expanding to 1,3-, etc. in the future if needed). This way they are less likely to get into the core, but not expressly forbidden.

In the longer term I think we need to think about pruning them based on thermo rather than flux, especially for pdep jobs, where it costs significantly more time/accuracy to deal with very large energy ranges.

[gmagoon]

My recollection from my earlier review of the literature is that it had been found that 1,2-biradical thermochemistry is adequately taken into account by group additivity, though I haven't investigated extensively. You might want to consult the references I listed at https://github.com/GreenGroup/RMG-Java/blob/master/databases/RMG_database/kinetics_groups/1%2C2-Birad_to_alkene/rateLibrary.txt (See in particular p. 463 of ref. 4: http://dx.doi.org/10.1021/ja00184a008 )