RMG Database incomplete for Liquid Phase Kinetics

[DetlevCM]

Dear All,

tracking things down, like the expected hydroperoxide decomposition, I found that apparently the liquid_oxidation kinetics groups are not considered when RMG runs as it only searches through RMG_Database.

Is there a specific reason why liquid phase reactions were removed?
Looking through prior versions, it appears that liquid phase reactions were removed from the main RMG database from version 3.2 - while they are present prior in RMG 3.0 and 3.1.

[Alborzi]

Dear all,
I have seen this issue too.
Any comment much appreciated,
Ehsan

[rwest]

IIRC the liquid_oxidation root level database was an aborted attempt (by me) to reproduce in entirety some other rule-based system for estimating liquid hydrocarbon autoxidation. I'm afraid I forget whose - it was many years ago that I did it. It was an experiment that I think was never finished, never tested, and never used, and so eventually deleted to reduce confusion. It was before we added solvation corrections to RMG and all the work by @ajalan on liquid oxidation.

Have people been finding a use for the liquid_oxidation database?

[rwest]

Poking around the source folder, maybe I got it further than I remember.
Is it (almost) complete and functioning?
It was based on Pfaendtner & Broadbelt 2008 http://dx.doi.org/10.1021/ie0714807

I don't think we ever tested (and certainly haven't published) a model built using it though.

[DetlevCM]

Dear Dr. West,

I remember from my early fiddlings with RMG that I obtained sensible looking liquid oxidation schemes that covered hydroperoxide decomposition, forming predominantly alcohols, aldehydes and ketones.
These then vanished in later revisions - looking through it now, it seems I started using RMG in that 3.0-3.1 branch where these reactions were present.

I would say that back then it definitely worked.
As to being complete: Based on the work of Sicard (IASH conference, 2013), carboxylic acids are also formed in liquid phase oxidation (besides alcohols aldehydes, ketones), so it covers some of the chemistry, but not all.

[DetlevCM]

OK, following this up some more, I just end up being more and more perplex....
From the dates, the earliest RMG version I would have used was 3.3. - or development code from GitHub.
I have an RMG Run on a version compiled in december 2012 - which shows nice decomposition of the hydroperoxides. The same input file fails with the current RMG version.

The condition file and RMG dictionary from the historic run are - with a changed file ending (!!!! txt -> png) attached below. (Change the file ending to txt and it should show.)

[rwest]

@DetlevCM do you have any of the other output files from this run? If any of them include the git hash of the version that was used to run it in the header, eg something like

!  This chemkin file was generated by RMG - Reaction Mechanism Generator (http://rmg.mit.edu)
!  The java code was compiled by ant at:
!    2010-10-27T11:10:41
!  The git repository was on the branch:
!    master
!  And at the commit with the hash:
!    5ad4962048c4b62389c777a00a4892638d46b067

then please share that with us too.

One thing to note is that input file uses the normal RMG_database not the liquid_oxidation one discussed at the top of this thread.

@ajalan (and other RMG-Java devs) don't suppose anything springs to mind as to why hydroperoxides no longer decompose?

[DetlevCM]

Unfortunately it doesn't have a hash - none of the RMG builds I created from zipped downloads have one, sorry.

This is the start of the logfile:
In fact, from the RMG repository I know I ran it on our cluster - so it was definitely a "self compile".
A code that was available between December 2012 and the release of RMG 3.3, though I think 3.3 also had some decomposition.

On the liquid_phase library:
I have tried with copying it to the kinetics groups which works. However it still doesn't include the original reactions that I had in my older schemes, in retrospect after more time with it yesterday I think these are other additional reactions. I have tacked on the chemkin scheme I got in the past from that run, again a changed file ending.

The beginning of the log is below:

RMG execution initiated at 2012-12-21 12:05:29
!                                                                  #
!                 RMG - Reaction Mechanism Generator                 #
!                                                                    #
!                    http://rmg.sourceforge.net/                     #
!                                                                    #
!  This java code was compiled by ant at:                            #
!    2012-12-21T11:54:08                                             #
!  The git repository was on the branch:                             #
!    * Not in git repository                                         #
!  And at the commit with the hash:                                  #
!    * Not in git repository                                         #
!                                                                    #
!  For details visit:                                                #
!    http://github.com/GreenGroup/RMG-Java/                          #
!                                                                    #
!  Copyright (c) 2002-2011                                           #
!  Prof. William H. Green and the RMG Team:                          #
!    Joshua W. Allen, Dr. Robert W. Ashcraft, Dr. Gregory J. Beran,  #
!    Caleb A. Class, Connie Gao, Dr. C. Franklin Goldsmith,          #
!    Michael R. Harper, Amrit Jalan, Gregory R. Magoon,              #
!    Dr. David M. Matheu, Shamel S. Merchant, Jeffrey D. Mo,         #
!    Sarah Petway, Sumathy Raman, Dr. Sandeep Sharma,                #
!    Prof. Kevin M. Van Geem, Dr. Jing Song, Dr. John Wen,           #
!    Dr. Richard H. West, Andrew Wong, Dr. Hsi-Wu Wong,              #
!    Dr. Paul E. Yelvington, Dr. Joanna Yu                           #
!                                                                    #
!  The RMGVE graphical user interface to the RMG database            #
!  was written by John Robotham.                                     #
!                                                                    #
!  This software package incorporates parts of the following         #
!  software packages:                                                #
!    DASSL    - Written by Prof. Linda Petzold et al                 #
!      http://www.cs.ucsb.edu/~cse/software.html                     #
!    CDK      - Written by Prof. Cristoph Steinbeck et al            #
!      http://cdk.sourceforge.net/                                   #
!    InChI    - Available from IUPAC                                 #
!      http://www.iupac.org/inchi/                                   #
!    cclib                                                           #
!      http://cclib.sourceforge.net                                  #
!                                                                    #
!  For more information, including how to properly cite this         #
!  program, see http://rmg.sourceforge.net/.                         #
!                                                                    #
!

[rwest]

Do any of the files (log, or chemkin) list the reaction family that generated the reactions that now seem to be missing?

[DetlevCM]

Unfortunately, I don't think I can see any reaction family that is mising.
I would think that is is a finer change.

We initially did some products tracing - and found a difference between iso and normal alkanes in the ratio between aldehydes and ketones. A newer build then stopped producing the ketones - just aldehydes and the discrepancy disappeared. Since it again changed....
Alcohols were always present to some extent.

I would think it is an interaction somewhere in the datbase - but where and how I unfortunately do not know.

[DetlevCM]

Another comment - I'm running a scheme at 500K with an error tolerance of 0.05 andnow have the odd ketone and some alcohols, BUT with species numbers of 3500 and hihger they formed very late.
-> They formed earlier and from published work (IASH conference, Sicard 2011, temperature 423K) we know that a good number of alcohols and aldehydes form given sufficient oxygen.
We also know from Sicard that the alkanes break down into smaller molecules - this does not seem to happen.

Something that is more present in today's modern database is the formation of larger molecules from radicals combining.

[DetlevCM]

OK, I just had a stupid little idea...tried something... not quite getting there but this may be related:
I've been rerunning an old setup or rather a slightly modified old setup - decane, 500K, 24.48 atm, oxygen.

A low error tolerance model became huge - with 300 species the RMG wesite also fails to visualise the scheme... (no output file produced), however on the smaller ones, some alcohols form and a lot of C20 polymers form. (Some would be expected.)
In fact, producing polymers was a weak spot of earlier RMG versions.

I then started to limit RMG - maximum of 15 carbon atoms. No more C20 polymerisations.
I now obtains a number of alcohols. (but no species breakdown).

Next I limted RMG to 15C and 2O atoms per molecule - I get some molecuels with 3 oxygen atoms.... (See #308 )
Also some cyclics, some recombination.... it looks more like the old runs, BUT I still don't get the decomposition, the decay of molecules into smaller species as would be expected from Sicard.

Then I looked at the manual - for RMG 3.3 and 4.0.0 - the species limits. Well, there is a minor mistake in the 3.3 manual... the example says maximum carbon atoms is 100 while the text says it is 30. I wonder if the bug could be there?
The other setting that must have changed is bond breaking - because that is what no longer takes place...
(See screenshots)

I have also provided a slide from the 2013 IASH prensentation by Mikael SIcard, which shows a breakdown of the products formed from liquid phase decomposition.
The PetroOxy, the device that was used in this test pressurised the test sample with oxygen and thus offers a reservoir, thus effects would be amplified, however it should be possible to reproduce the trend in RMG.
(Screenshot again)

[DetlevCM]

OK, I did some more digging on the liquid phase aspect and I may have found an answer? Or maybe not...

I ran RMG 3.3. - and obtained the desired breakdown products. Visualizing the scheme and looking through it I found the following:
RO. + .OH <=> ROOH ! Radical Recombination

This looks very much like a hydroperoxide decomposition - and in my experience RMG often writes reactions backwards... (backwards relative to forward being determined by the dominant reaction rate)

I also obtain H2O2 <=> 2 HO. which I believe I no longer see.

This possibly directly links back to Ehsan's question here: #306

[DetlevCM]

Ok, I did some investigating:

I took my version from December 2014, removed the radical recombination reactions and dropped in the radical recombination reactions from version 3.3. instead.

Results:

	RMG 3.3	RMG 4.0.1+ with 3.3 R_recomb	RMG 4.0.1+
Species	179	96	96
Reactions	2890	3105	3105

A hydroperoxide decomposition does occur - but not significant enough. (And in the case of our models at 423K it doesn't happen.)
I might need to try chasing this through a lower temperature (not 500K - 24.48atm, but 423K and 7.14atm - which are the conditions for the experimental work referenced above and also for our case).

I'm a tad perplexed... but the modern database is definitely missing smaller hydrocarbons - initial input is C10, RMG3.3 produces <C10 hydrocarbons, RMG4.0.1+ does not...

[DetlevCM]

OK, I did a bit fo a followup recently and tried my damnest to get RMG 4.0.1 (or later) to produce some hydroperoxide decomposition.

I raised the temperature, increased the pressure... no go...
It will decompose some, produce some alcohols, but nowhere near what would be expected.

I'm truly perplexed as to why RMG no longer predicts the production of species which we know to form as a result of autoxidation...
On a side note in case I didn't mention it before: The hydroperoxide decompositon features as a radical recombination and is printed in revese, i.e. RO. + .OH -> ROOH by RMG.