Re: [Open Babel] Mods to allow OpenBabel to handle radicals

[Peter Murray-R. <pm...@ca...>]

At 17:39 27/09/2003 -0500, Geoff Hutchison wrote:

>On Saturday, September 27, 2003, at 08:49  AM, Peter Murray-Rust wrote:
>
>><molecule formula="K 3 Fe 1 C 6 N 6"/>
>>would have a spinMultiplicity of ?? (help I have to go back to 
>>undergraduate inorganic...) high-spin or low-spin? temperature dependent? etc.
>
>There are 77 valence electrons, so there's definitely at least one 
>unpaired spin. (Or you can look at it and realize it's Fe(3+) and so it's 
>d5.) This is low-spin since cyano ligands force a large d-d splitting. So 
>it's likely a triplet. It's not likely temperature-dependent. Those are 
>usually Fe(2+) ions in an octahedral nitrogen environment--at least for 
>iron compounds.

Yes :-) this makes my point. These are the things someone has to think 
about before assigning a multiplicity. If it defaults to 0 it's wrong.


>>  Also if we have:
>><atom elementType="Fe">...
>>then we have to give a spinMultiplicity every time we have a transition 
>>metal.
>
>Here, no. If it's a lone atom, it must have charge 0, and you can quickly 
>figure out the spin. IIRC, lone iron atoms are quintets in the gas phase.
>
>>If we don't then it is *automatically* set to 0, regardless of the actual 
>>value. XML does not transmit default attribute values so it is impossible 
>>for a later user to determine whether the "0" was intended or merely 
>>added  automatically.
>
>Total spins take the values 2S+1, so a 0 "spin" is not a physical value. 
>It *must* have been the default value. A singlet is a spin of 1. So for 
>example, if I want to specify a neutral, singlet molecule to Q-Chem, I 
>supply the line "0,1" before the coordinates.
>
>>However we probably should allow the spin to take values of zero.
>
>If you want to allow the spin to take values of zero, it must be the 
>default value

A default must not be overloaded to mean "either it is this value or I 
don't care". In this case it is clear that no default is possible. A 
missing multiplicity for either an atom or a molecules mean "I don't know"

>and imply "I don't know the spin, or I do not care about spin." I do not 
>think that code should allow *setting* the spin to zero.

What I meant is that 0 should be an allowable value. [In CML V2.01 the spin 
multiplicity *if given* was required to be a positive integer. This is 
simply a bug and 0 should be allowed, but not negative integers.]

>IMO, if the spin is set by the user, the code should check to make sure 
>the spin is at least reasonable, e.g. calculating the number of valence 
>electrons and doing a simple aufbau summation.

That is up to the OB community. It seems a useful thing to do, but I don't 
know how many exceptions there are.

To give some more examples of defaults taken from CML. If no field is given 
the interpretation is

atoms:
elementType =>unknown elementType
hydrogenCount => unknown hydrogenCount (NOT 0)
formalCharge    for atom  => 0 This is based on the assumption that "most" 
atoms are not charged and that most authors know the importance of formal 
Charge
occupancy => 1.0 based on the assumptions that most authors either don't 
know about occupancy and assume that if an atom is given then there is 
exactly one
x2, x3, xFract => unknown
isotope => "natural abundance" (but this is not defined)
atomParity => unknown

molecules:
formula => the sum of the explicit atoms present, including explicit 
hydrogenCounts if given. This is likely to be wrong in many cases, e.g. 
from MOL files without explicit hydrogens
formalCharge => sum of explicit atom formal charges
spinMultiplicity => unknown, even if atom values are given, as their 
combination may be unknown

bonds
order => unknown (NOT 1)
stereo => unknown

Peter