This study used automated transition state theory (AutoTST) calculations to determine rate constants for hydrogen abstraction reactions by hydroperoxyl radicals (HOO), and compared the results to 72 published kinetic models. The researchers identified 331 relevant H abstraction reactions from the models, generated transition state geometries for 316 of them using distance geometry, and successfully calculated rate constants for 228 reactions using AutoTST. Analysis found the AutoTST approach matched existing models for 68.9% of reactions, with ongoing work seeking to improve accuracy and apply the method to new fuels.
Pests of castor_Binomics_Identification_Dr.UPR.pdf
Automated transition state theory calculations of abstraction reactions by hydroperoxyl, compared to literature model values
1. Automated transition state theory
calculations of abstraction
reactions by hydroperoxyl,
compared to literature model values
Nathan Harms, Belinda Slakman, Jason Cain, Richard West
Department of Chemical Engineering
Northeastern University, Boston, MA
10th U.S. National Combustion Meeting
April 24, 2017
1
.edu/comocheng
10. Some methods rely on rate rules to
determine reaction kinetics…
5
Advantages
• Fast
• Many rules exist
Disadvantages
• Chemistry doesn’t always
follow the rule
• Rules are misapplied
CH + RH → Products
D. L. Baulch et al, Evaluated Kinetics Data for Combustion
Modeling, J. Phys. Chem. 2005
CH4
C2H6C4H10
C3H8
11. … others rely on transition state theory
calculations to determine kinetics.
V
X‡
Reaction coordinate
Energy
Reactants
Products
6
Advantages
• More accurate
than rate rules
Disadvantages
• High computational time
• Requires initial guess of
transition state geometry
12. We would like to estimate fewer
of our important reaction rates
7
Measured
Calculated
Estimated
Confidence level
13. We would like to estimate fewer
of our important reaction rates
7
Measured
Calculated
Estimated
Confidence level
14. Our group has automated
the entire TST process
8
⇌RMG
Estimate
Distances
Create TS
Geometry
TS Search
and
Validation
Determine
Kinetics
Y
X
H
CanTherm
P. Bhoorasingh et al, FigShare (2016).
15. dXH: 1.332
dHY: 0.831
dXY: 2.160
Use group additivity and a hierarchical
tree database to predict distances
9
R
O
H
C
O
H
O
R
C
HO
R
X
H
dXH: 1.242
dHY: 0.874
dXY: 2.111
dXH: 1.336
dHY: 0.998
dXY: 2.334
dXH: 1.301
dHY: 0.906
dXY: 2.205
C
HC
R
Functional Group X-H H-Y X-Y
1.235 0.875 2.104
1.242 0.873 2.109
1.241 0.873 2.109
1.243 0.873 2.112
1.248 0.874 2.119
1.329 0.999 2.328
1.334 0.998 2.332
1.337 0.997 2.334
1.338 0.998 2.336
1.341 0.999 2.340
1.338 0.829 2.163
1.332 0.830 2.159
1.332 0.831 2.161
1.325 0.835 2.158
R
O
H
C
HO
R
C
O
H
O
R
Training Set Group Tree
Bhoorasingh & West Phys. Chem. Chem. Phys. 17, 32173–32182 (2015).
16. Generate the 3D transition state
estimate using distance geometry
10
⇌RMG
Molecule
Connectivity
3D
Structure
RDKit: Open-source cheminformatics. http://www.rdkit.org/
17. Generate the 3D transition state
estimate using distance geometry
10
⇌RMG
Molecule
Connectivity
Atoms List
AtomsList
Upper limits
Lower limits
Generate
distance
matrix
Embed
in 3D
RDKit: Open-source cheminformatics. http://www.rdkit.org/
18. Generate the 3D transition state
estimate using distance geometry
10
⇌RMG
Molecule
Connectivity
Atoms List
AtomsList
Upper limits
Lower limits
Generate
distance
matrix
Atoms List
AtomsList
Embed in 3D
Edit
distance matrix
RDKit: Open-source cheminformatics. http://www.rdkit.org/
19. New workflow is slightly more robust,
using a series of partial optimizations.
11
20. Comparing literature kinetics
against automatic TST calculations
for H abstraction by OOH radical
Calculate the
reaction rate
using AutoTST
Filter for
reaction of
interest
Gather models
published
Identify species
involved
Compare the
results
12
21. We have collected 72 different models
so far…
PCI2013/353-Malewicki
PCI2017/037-Sakai
PCI2017/032-Cheng
PCI2017/082-Hemken
PCI2017/111-Jin
PCI2017/052-Stagni
PCI2017/022-Thion
PCI2017/052-Li
PCI2017/058-Sun
PCI2017/024-Bohon
PCI2017/145-Sun
PCI2017/051-Pelucchi
PCI2017/012-Felsmann
PCI2017/047-Rodriguez
PCI2017/038-Labbe-Zhao(30Torr-10Atm)
PCI2017/087-Tran
PCI2017/036-Rashidi
PCI2017/025-Sudholt EL24115
CombFlame2013/2712-Sarathy
CombFlame2013/1958-Zhao
CombFlame2013/487-Schenk
CombFlame2013/1541-Zhang
CombFlame2013/17-Malewicki
CombFlame2013/1315-Chang
CombFlame2013/1939-Cai
CombFlame2013/2291-Somers
CombFlame2013/1609-Veloo
CombFlame2013/2680-Vranck
MB-Dooley
Narayanaswamy
H2-3
Shamel-Propane
Reduced-DRG-GRI-mech
PCI2015/0409-Zhang
PCI2015/0325-Nawdiyal
PCI2015/0153-Marshall
AramcoMech_1.3
USC_Mech_ii
n-Heptane
CombFlame2015/3755-Konnov
Chernov
MB-Farooq
GRI-mech-3.0
CombFlame2014/405-Cai
CombFlame2014/65-Darcy
CombFlame2014/885-Xiong
CombFlame2014/1135-Dames
CombFlame2014/84-Wang
CombFlame2014/657-Jin
CombFlame2014/798-Cai
CombFlame2014/818-Zhang
AramcoMech_2.0
Biomass
H2
MB-Fisher
MatheuCH4
Gasoline_2
GRI-17-species-mech
Gasoline_Surrogate
IJCK2013/638-Metcalfe
PCI2013/269-Matsugi
PCI2013/401-Liu
PCI2013/335-Wang
PCI2013/325-Husson
PCI2013/297-Herbinet
PCI2013/361-Malewicki
PCI2013/599-Veloo
PCI2013/259-Labbe
PCI2013/225-Somers
PCI2013/527-Sheen
PCI2013/411-Darcy
PCI2013/289-Dagaut
13
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
22. … and each one has slightly different
naming conventions for each molecule
14
MVOX
IIC4H7Q2-T
C3H5-A C3H5-SC6H101OOH5-4
TC4H8O2H-I
C4H8O1-3
C3KET21
CH3COCH2O2H
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
23. CH4 + HO2• → CH3• + H2O2
CH2O + HO2• → HCO + H2O2
C2H6 + CH3• → C2H5• + CH4
PCI2013/353-Malewicki
C2H6 + HO2• → C2H5• + H2O2
•••
PCI2013/353-Malewicki
C2H6 + HO2• → C2H5• + H2O2
PCI2013/353-Malewicki
CH4 + HO2• → CH3• + H2O2
Using python scripts, the existing models
are filtered for OOH Abstraction reactions
15
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
24. CH4 + HO2• → CH3• + H2O2
CH2O + HO2• → HCO + H2O2
C2H6 + CH3• → C2H5• + CH4
PCI2013/353-Malewicki
C2H6 + HO2• → C2H5• + H2O2
•••
PCI2013/353-Malewicki
C2H6 + HO2• → C2H5• + H2O2
PCI2013/353-Malewicki
CH4 + HO2• → CH3• + H2O2
Using python scripts, the existing models
are filtered for OOH Abstraction reactions
15
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
25. The filtered reactions were given to
AutoTST to calculate reaction rates
16
PCI2013/353-Malewicki
CH4 + HO2• → CH3• + H2O2
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
26. The filtered reactions were given to
AutoTST to calculate reaction rates
16
PCI2013/353-Malewicki
CH4 + HO2• → CH3• + H2O2
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
27. Rates from existing models then compared
to AutoTST rates and other models
17
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
🆚
28. Rates from existing models then compared
to AutoTST rates and other models
17
Gather models
published
Identify
species
involved
Filter for
reaction of
interest
Calculate the
reaction rate
using AutoTST
Compare the
results
PCI2013/353-Malewicki
CombFlame2014/405-Cai
CH4 + HO2• → CH3• + H2O2
🆚
PCI2017/082-Hemken
CH4 + HO2• → CH3• + H2O2
CombFlame2014/405-Cai
CH4 + HO2• → CH3• + H2O2
33. Finding our bugs
and measuring our progress
21
Step in the Methods
Number of
Reactions
Percent of
Reactions
Plausible Reactions
Identified from Models 407 N/A
Reaction Attempted 407 N/A
Matched an H-Abstraction
Reaction in RMG 331 100%
Started Making TS
Geometry 316 95.5%
Started Canonical TST
Calculations 246 74.3%
Overall Successes 228 68.9%
34. Finding our bugs
and measuring our progress
21
Step in the Methods
Number of
Reactions
Percent of
Reactions
Plausible Reactions
Identified from Models 407 N/A
Reaction Attempted 407 N/A
Matched an H-Abstraction
Reaction in RMG 331 100%
Started Making TS
Geometry 316 95.5%
Started Canonical TST
Calculations 246 74.3%
Overall Successes 228 68.9%
35. Finding our bugs
and measuring our progress
21
Step in the Methods
Number of
Reactions
Percent of
Reactions
Plausible Reactions
Identified from Models 407 N/A
Reaction Attempted 407 N/A
Matched an H-Abstraction
Reaction in RMG 331 100%
Started Making TS
Geometry 316 95.5%
Started Canonical TST
Calculations 246 74.3%
Overall Successes 228 68.9%
36. Finding our bugs
and measuring our progress
21
Step in the Methods
Number of
Reactions
Percent of
Reactions
Plausible Reactions
Identified from Models 407 N/A
Reaction Attempted 407 N/A
Matched an H-Abstraction
Reaction in RMG 331 100%
Started Making TS
Geometry 316 95.5%
Started Canonical TST
Calculations 246 74.3%
Overall Successes 228 68.9%
37. Ongoing and future efforts
22
• Complete Imported
OOH Kinetics
• Improve the
robustness and
accuracy of AutoTST
• Look at OOH
Abstraction of new
fuels
All Combustion Reactions
Hydrogen
Abstraction
Reactions
H Abstraction
by OOH
Published Kinetics
38. Acknowledgements
• Nathan Harms, Belinda Slakman,
Jason Cain, Dr. Pierre Bhoorashingh
• ACS Petroleum Research Fund
• National Science Foundation
Grant No.1605568
• Northeastern University,
Department of Chemical Engineering
23
39. Contributions
• Identified 331 “RH + ·OOH ⇌ ·R + H2O2” reactions
from published kinetic models.
• Found and validated (with IRC) transition states
for 228 of them and calculated k(T)
• Developed tools to do the above automatically.
24
r.west@neu.edu richardhwest rwest
Richard H. West
24 April 2017