Molecule masses from symbol issue

[dgasmith]

Describe the bug
Nucleon symbols and values are in the periodictable information, but do not propagate their values into a Molecules mass. For example symbol D or He3 do not correctly review masses of 2 or 3, respectively.

To Reproduce

>>> qcel.models.Molecule(symbols=["He3"], geometry=[0,0,0]).masses
array([4.00260325413])

Expected behavior
The mass of the Helium3 to be 3 and change.

Additional context
The issues comes from reconcile_nucleus where everything is validated off of atomic number rather than the input symbol. This looks like a non trivial fix to that logic.

[loriab]

did you try 3He? that should recognize the isotope. He3 is read as a user-convenience tag of 3 on regular helium.

[dgasmith]

qcelemental.exceptions.NotAnElementError: 3He

There are unambiguous issues as well with known isotopes:

>>> qcel.models.Molecule(symbols=["D"], geometry=[0,0,0]).masses
array([1.00782503])

[loriab]

ok. it's probably in the handing over of field symbol to molparse -- may need a different field. it shouldn't be deep, though, since this is literally the problem reconcile_nucleus was built to handle. https://github.com/MolSSI/QCElemental/blob/master/qcelemental/molparse/nucleus.py#L381

[loriab]

the below works (though it's not set up to recognize D, T). understandably you'd like to not have so many guts showing to make a simple deuterium.

>>> qcel.models.Molecule(**qcel.molparse.to_schema(qcel.molparse.from_arrays(elbl=["2H"], geom=[0,0,0]), dtype=2)).masses
array([2.01410177812])

The problem comes about pretty much because https://github.com/MolSSI/QCElemental/blob/master/qcelemental/molparse/from_schema.py#L82 . https://github.com/MolSSI/QCElemental/blob/master/qcelemental/molparse/from_arrays.py#L191-L195 I hesitate to flip it without really studying the consequences.

[dgasmith]

Yea, I don't think this is either a priority or a quick fix. More of a "I was writing a test was and was somewhat surprised". Something to keep in mind I think.

[loriab]

I don't know if it's so much a fix as just the current setup enforces schema field definitions. Below are a couple schema-approved ways to do your original problem. Redefining symbols where as input it's a spec label and as output it's a string is a new kettle of fish. But if that is wanted someday, the underlying tech is ready.

>>> qcel.models.Molecule(symbols=["He"], mass_numbers=[3], geometry=[0,0,0]).masses
array([3.01602932])
>>> qcel.models.Molecule.from_data("""3He 0 0 0""").masses
array([3.01602932])

[dgasmith]

In this case perhaps symbols=["He3"] should throw an error that information in the symbols is overloaded?

[loriab]

Sounds sensible at first. You have hit on an inconsistency, though. speclabel takes prefix mass number so that post can be label or mass, so 3He. meanwhile, the periodictable takes suffix mass number: He3.

So symbols is being broadly interpreted (6 lines below map to the same Mol) but only to extract the info that is properly contained in symbols. Probably something like an option on to_Z(..., strict=True) that restricts interpretation to element only.

# all same
qcel.models.Molecule(symbols=['He'], geometry=[0,0,0])
qcel.models.Molecule(symbols=['He3'], geometry=[0,0,0])
qcel.models.Molecule(symbols=['He4'], geometry=[0,0,0])
qcel.models.Molecule(symbols=['2'], geometry=[0,0,0])
qcel.models.Molecule(symbols=[2], geometry=[0,0,0])
qcel.models.Molecule(symbols=['helium'], geometry=[0,0,0])
# replies "you fool"
qcel.models.Molecule(symbols=['He300'], geometry=[0,0,0])