Invalid chemkin thermo block generated for an adsorbate with 5 elements

[rwest]

Bug Description

When running a surface chemistry job, this ended up in my chem_annotated-surface.inp chemkin file

! Gas phase thermo from Thermo library: thermo_DFT_CCSDTF12_BAC. Adsorption correction: + Thermo group additivity estimation: adsorptionPt((NRO)*)
! Binding energy corrected by LSR.
SX(8582)                                    G   100.000  5000.000  948.83      1&
X   1H   3C   1O   2N   1
 8.75901373E+00 7.66223340E-03-2.06362528E-06 4.02513189E-10-3.39045287E-14    2
-2.23373310E+04-3.74620293E+01 3.45624020E+00 3.20307213E-04 5.64909893E-05    3
-7.37256390E-08 2.81889462E-11-1.99942728E+04-5.11341806E+00                   4

which then obviously crashed the cantera conversion, leading to the error message

Warning: 
ERROR: Unable to parse '/Users/rwest/Dropbox (CoMoChEng)/Northeastern/Papers/2019-02 RMG-Cat Pt database franklin/rmgcat-pt-supporting-info/RMG_CH4catcombust_Pt_coupled/chemkin/chem-surface.inp' near line 195:

Error: Could not generate Cantera files for some reason.
Traceback (most recent call last):
  File "/Users/rwest/Code/Cat/RMG-Py/rmgpy/rmg/main.py", line 923, in execute
    surfaceFile=(os.path.join(self.outputDirectory, 'chemkin', 'chem-surface.inp')))
  File "/Users/rwest/Code/Cat/RMG-Py/rmgpy/rmg/main.py", line 1407, in generateCanteraFiles
    parser.convertMech(chemkinFile, transportFile=transportFile, outName=outName, quiet=True, permissive=True, **kwargs)
  File "/Users/rwest/anaconda/envs/rmg230/lib/python2.7/site-packages/cantera/ck2cti.py", line 2183, in convertMech
    self.loadChemkinFile(surfaceFile, surface=True)
  File "/Users/rwest/anaconda/envs/rmg230/lib/python2.7/site-packages/cantera/ck2cti.py", line 1736, in loadChemkinFile
    label, thermo, comp, note = self.readThermoEntry(thermo, TintDefault)
  File "/Users/rwest/anaconda/envs/rmg230/lib/python2.7/site-packages/cantera/ck2cti.py", line 1039, in readThermoEntry
    composition[comp[i].capitalize()] = int(comp[i+1])
ValueError: invalid literal for int() with base 10: '1H'

To (hopefully) reproduce it without running the full simulation (which took ~6 hours) the species in question is

SX(8582)
1 O u0 p3 c-1 {3,S}
2 O u0 p2 c0 {3,D}
3 N u0 p0 c+1 {1,S} {2,D} {4,S}
4 C u0 p0 c0 {3,S} {5,S} {6,S} {7,S}
5 H u0 p0 c0 {4,S}
6 H u0 p0 c0 {4,S}
7 H u0 p0 c0 {4,S}
8 X u0 p0 c0

My guess is that this is perhaps the first time we've run across a species with five different elements (X, H, C, O, N) and our chemin-writing code isn't set up to do. (The fifth element count block should go in that gap at the end of line 1, if I recall correctly).

How To Reproduce

I'm guessing here, but try writing a chemkin thermo block for the species mentioned above.

Expected Behavior

It'd make a valid chemkin file.

Installation Information

Describe your installation method and system information.

• Mac OS X
• built from source, in an anaconda environment, using the 2.4.0 tagged commits from github

The current git HEAD for RMG-Py is:
	caebd4895dfaee296b6562b28f89e6c350422cca
	Fri Jun 14 18:30:27 2019 -0400
The current git HEAD for RMG-database is:
	08b8214e8b4c869411f498cb50acb76b012520b4
	Fri Jun 14 18:15:05 2019 -0400

[mliu49]

I think RMG's behavior may be somewhat correct. We have explicit handling for molecules with more than 4 elements, which (according to the comments) is supported by Chemkin 4 and later.

I confirmed the syntax with the input manual for our current version of Chemkin Pro. The & indicates line continuation and the elements are specified on the following line(s).

I think you are correct that Chemkin II supports a 5th element in the gap between the last temperature and the 1. Perhaps we could update our Chemkin writer to prefer doing that for molecules with 5 elements.

Additionally, new Chemkin also supports additional elements directly after the 1 (i.e. no & or line break). You could test to see if ck2cti can parse that properly.

I do feel like it the ck2cti converter should be able to handle the line continuation form as well though.

[rwest]

OK, thanks Max. I was not familiar with the new Chemkin 4 syntax, and it appears that the version of ck2cti that it was using is not either (I guess this is one that was distributed with RMG, but perhaps I'm in a conda environment made for RMG 2.3.0).

I suggest

• for 5 elements, use the Chemkin2 syntax (as long as it is also supported by Chemkin 4+ and Cantera)
• for 6 or more elements, use the current (Chemkin4+) syntax.
• ensure that the ck2cti version required by or distributed with RMG can read the Chemkin4+ syntax
• ensure that Cantera's ck2cti converter can read the Chemkin4+ syntax (send a pull request if needed, including a unit test)

?

Anyway, at the end of the day, RMG shouldn't crash converting its own files.

[mliu49]

I looked into it some more, and it turns out that we are actually writing it wrong. 😓

The new line syntax is not column based and should be whitespace delimited instead:

X 1 H 3 C 1 O 2 N 1

instead of

X   1H   3C   1O   2N   1

So we should definitely fix that.

While I was digging around, I found out that the ct2cti script originated as a port of RMG's Chemkin module. 😄

[mliu49]

Fixed by #1636.