Sanitization error for Carbon Monoxide?

[nickvandewiele]

My understanding is that we represent carbon monoxide as the triple-bonded structure with charged atoms, each atom bearing 1 lone pair.

When I run this Molecule through RDKit's Chem.Sanitize, it complains the valency is not correct.

    from rmgpy.molecule import Molecule
    from rdkit import Chem
    import rmgpy.molecule.generator as generator
    adj = """
    1 O u0 p1 c+1 {2,T}
    2 C u0 p1 c-1 {1,T}
    """

    mol = Molecule().fromAdjacencyList(adj)
    rdkitmol = generator.toRDKitMol(mol, sanitize=False)
    Chem.SanitizeMol(rdkitmol)

RMG-Py, by default, runs RDKit's sanitize functionality for compounds that do not contain nitrogen atoms.

I wonder how the error of CO should be treated. The options I see:

• change the representation of CO to comply with RDKit's Sanitization procedure?
• avoid having CO in RMG-Py altogether?
• change the default settings for sanitization calls in RMG-Py

I could not find prior raised issues on this topic, apart from #367, in which the representation of carbon monoxide in the solvation.transformLonePairs function of RMG-Py was changed into the biradical to avoid the same SanitizationError. bslakman@aa15c5c

[rwest]

Not sure how we ought best to represent CO, but a molecule search for "carbon monoxide" returns formaldehyde, and one for [C][O] returns something even stranger.

[alongd]

In Wikipedia (for what it's worth) carbon monoxide's SMILES is [C-]#[O+], as @nickvandewiele 's AdjList suggests.
In the latest test I ran in RMG-Py (my ref xa1179) I added [C-]#[O+] to the initial mixture, looking for its reactions with ethylamine (NCC). Surprisingly, the [C-]#[O+] species did not react with ethylamine (did react with H2, NH3, CH4...). Instead, RMG predicted the species [C]=O (triplet), which gave the reactions I was expecting.

See Issues: RMP-Py #184, ReactionMechanismGenerator/RMG-database#31

@connie reccomended to represent CO as [C-]#[O+] in issue ReactionMechanismGenerator/RMG-database#31. I agree we should continue to update this in the database. For example, In the 2+2_cycloaddition_CO kinetic family (in groups.py), CO is still defined as [C]=O (to the best of my understanding):

entry(
    index = 1,
    label = "CO",
    group = 
"""
1 *1 CO u0 {2,D}
2 *2 Od u0 {1,D}
""",
    kinetics = None,
)

Another problem (which brought up this issue for me in the first place) is that currently only radicals can react to abstract a proton in the H_Abstraction kinetic family (reactants=["X_H_or_Xrad_H_Xbirad_H_Xtrirad_H", "Y_rad_birad_trirad_quadrad"]), so I'm not getting the reactions I was expecting from CO (e.g.,CH3CH2NH2 + CO <=> CH2CH2NH2 + HCO) using the [C-]#[O+] species.

[soumsrani]

Hi, I am using 796adab version of RMG-Py. I am also getting both [C-]#[O+] and [C]=O in the mechanism. The mechanism generated contains CH4, C2H6 and C3H8 chemistry.
input.docx

[connie]

I believe the 2+2_cycloaddition_CO is not meant for carbon monoxide but instead of C=O double bonded forms, as the child groups of that entry are all things like methanal, etc.

I do believe that H_Abstraction only should occur with radical abstraction. I forget what we were thinking but I thought there were routes for this reaction to occur not through H_Abstraction but I'm not sure..

[alongd]

Just ran 4 simulations of:
1% [C-]#[O+], 1% X in 98% Ar at 1500K for 10 s,
where X = H2 \ CH4 \ NH3 \ NCC

this structure of CO seems to react in RMG only via the 1,2 Insertion CO family.
RMG generated reactions of this sort:

CO(1) + H2(2) = CH2O(4)
CO(1) + CH4(2) = C2H4O(7)
CO(1) + NH3(2) = CH3NO(5)
CO(1) + NCC(2) = C3H7NO(13)
CO(1) + C2H4(29) = C3H4O(37)
..........

I did not get any of the following reactions (looked in chem_edge_annotated.inp as well):

CO + H2 = HCO + H
CO + CH4 = HCO + CH3
CO + NH3 = HCO + NH2
CO + NCC = HCO + NCCrad
CO + C2H4 = HCO + C2H3
..........

These are probably important to many RMG works, and should be added to the Disproportionation family template.

[rwest]

Sounds like this needs some serious attention.

[alongd]

In second thought, maybe Disproportionation is not the right family for these reactions.

Disproportionation:
*R1 + *R3-R2-H <=> R1-H + R3=R2

While we need to form a unique triple bond ([R3+]#[R2-]):

 *R1 + *R3-R2-H <=> R1-H + R3#R2

(*R  + *HCO     <=> R-H + [C-]#[O+])

I couldn't see any other fit from our reaction families.
Should we open a new reaction family, or did I miss something?

[alongd]

Perhaps we should open a CO_disprop family.

Strengthening the Wikipedia comment from above, the following paper says that [C-]#[O+] is indeed the right non-excited form of CO:
Electronic structure of CO—An exercise in modern chemical bonding theory

[mliu49]

I think CO representation has been fixed, and the CO_disprop family has also been added already.