Conversation
…ed in input/thermo/groups/group.py
A bunch of groups were modified such that a H atoms could also be a Cl. The names were also modified, but the rate rules were not. This commit updates the rate rules to match those changes.
All other families use underscores, and this was already called that in some places!
All other families use underscores not hyphens, and this was already called that in some places!
What we need to do is figure out why the group additive estimate was bad for this! (And test it on some other things) I converted the NASA/CHEMKIN-like block into polynomials using http://rmg.mit.edu/simulate/evaluate_nasa
Not clear what group this was meant to be. The listed C_rad/H2/Cl group is in the Yrad tree, as is the Cl_rad. Neither is an XH group.
Otherwise the child nodes with chlorine in cannot be reached.
I'm not sure where these were meant to be, but we don't seem to have training data for them anyway. Group C/H2/Cs/Cl in H_Abstraction family was found to be identical to group C/H2/CsCl
Rather than chloromethanes be under C_pri (which usually means it is bonded to another carbon) or C_sec (which would have 2 neighboring ligands) they are now siblings of methane, under a 'C1' node. i.e. Cl doesn't count as a "heavy atom" when defining primary/ secondary/tertiary; chloromethane is more like methane than ethane. Not much training data for these, and there are no comments about where they came from, so I'm a bit nervous. Also, we have no data for CH2Cl2
It is not a child. If they are similar enough we could add a common parent as another level in the tree, but no evidence to do so at this time.
Remove the unreachable duplicate definition
Their parent wasn't a real parent so they couldn't be reached, and we have no data for them anyway.
These could not be reached, and have no data
L6: C/H/MustCl
L7: C/H/Cs2Cl
L7: C/H/CsCl2
This family now passes the $ make test-database unit tests
…ion DOI: 10.1039/c7cp07552a
|
Thanks! |
|
Yes, let's rename it |
|
Tree generation which will be the norm soon is much easier to do for smaller families than larger families. So if we expect the type of atom being abstracted to have the strongest impact on the rate, we should keep them separate. If we have good reasons to expect that we will want to use Cl rates for F reactions we should have one family. |
|
bond BDE(kj/mol) Good point Matt. |
|
Sounds like consensus is to leave it as Things remaining to do?
Anything else? |
The current RMG does not have Bromine atom types
Writing these adjacency list and species names by hand is pretty tedious and easy to make mistakes. Is there a semi-automated way?
|
That's weird, the test that failed. A segfault when generating trial molecules for a group somewhere. Seems familiar from when I was merging RMG-cat. |
|
The segfault should be fixed by ReactionMechanismGenerator/RMG-Py#1656, then we should run these tests again. |
|
@davidfarinajr, are there any intentions to rebase fix this PR? (or is there an updated one?) A colleague of mine would like to use RMG for halogens. Thanks! |
|
This branch is way behind. I'm waiting on my calculations to finish to re-fit group additivity values, HBIs and long-range correction groups for fluorine. I do have kinetics and thermo from NIST HFC mech on another branch (https://github.com/davidfarinajr/RMG-database/tree/fluorine_PR). But the branch is messy and I haven't gotten to opening PRs yet. I will make this higher priority for myself if others are wanting to use RMG for halogens. Which halogens are they interested in? We have the most data for fluorine, less for chlorine, even less for bromine and vey very little for mixed halogens. |
|
Thanks @davidfarinajr! |
6 participants
Kinetics:
Thermo:
-- CCSD(T) was extrapolated using a two-point method implemented in Orca: SCF part extrapolate based on the scheme proposed in J. Chem. Phys. 2008, 129, 184116 using the exponents from J. Chem. Theory Comput., 7, 33-43 (2011) and the correlation part extrapolate based on the scheme proposed in J. Chem. Phys. 1997, 106, 9639 using the exponents from J. Chem. Theory Comput., 7, 33-43 (2011))