molecular-level kinetics in chemical engineering: from
TRANSCRIPT
9/9/2015
1
Molecular-level Kinetics in Chemical Engineering: From Chemical to
Mathematical ModelingKRG
University of Delaware [email protected]
Outline
• Chemical Modeling
• Elements of Chemical Modeling
Structure
Reaction Paths and Kinetics
Detailed Kinetic Modeling Approaches
• Computational Tools
• A Practical Hybrid Approach
• Conclusions
2
9/9/2015
2
Chemical Modeling Complements Real System Experiments
3
• Generally Whole-feed (e.g., Resid, Coal, Lignin) Experiments
• Pilot Plant or Lab Scale• Global feed, Global Product Measurements• Lumped Model
The direct and practical approach to obtaining chemical information:
Real Experiment
Real
System
Real
Results
Chemical Modeling
Limitations of Lumped Models
• Absence of chemical structure
• Consequence: absence of properties beyond definition of lump
• Conflict: new questions of unprecedented molecular detail
performance
environmental
reactivity
• Motivates molecular-level modeling
4
9/9/2015
3
Chemical Modeling Complements Real System Experiments
5
Accurate Chemical Modeling Requires Rigorous Mapping Steps
MS and RSbetween sdifference theaddress toTransform Modeling Chemical Inverse
domains Model and Real common to kinetics intrinsic procure todesigned Experiment System Model
MS and RSbetween sdifference theaddress toTransform"" Modeling Chemical
1
CM
ME
CM
Chemical Modeling
Elements of Chemical Modeling
6
CM• Simulation of Structure: CME• Optimization of 10K+ Molecular Composition: CME
ME• Network Analysis: DelPlots• LFER Analysis: Organization of 10K+ Rate, Adsorption and Thermodynamic Parameters• Extrinsic Effects: Kinetic Coupling; Solvent Effects; Phase Behavior, Transport• Novel Reactors
CM-1
• 10K+ Multifunctional Molecules: Implicit Equations via Monte Carlo Simulation of Kinetics• Equation Generation: NetGen, OdeGen• Kinetic Coupling: Accounting for Interactions; Optimal System Design• Transport: Molecule vs. Moiety; Size and Matrix Effects• ARM: Modeling N molecules with fewer than N equations
9/9/2015
4
Basic DelPlot Analysis (9/8/15)
7
A A, B, C, D, ...
Object is to determine the rank* of each product
y
x
i
A
xA
r > 1
non-zero intercept 1o
0 1
*Generations removed from reactant
Basic DelPlot Exposes Primary Products
8
0.60.50.40.30.20.10.00.0
0.2
0.4
0.6
0.8
X
Y/X
1 B 3 C; A 2 DA
-1k1 /min -1 = 1, k 2 /min = 2, and k 3 /min -1 = 4.
D
B
C
9/9/2015
5
Second-rank DelPlot Exposes Generations
9
k1/min -1 = 1, k1/min -1 = 2, and k3/min -1 = 4.
0.60.50.40.30.20.10.0.1
1
10
100
X
Y/XD
B
C
2
1 B 3 C; A 2 DA
Extended DelPlot Analysis
10
•
R = r
R < r
R > r
yi
XA
r
XA
9/9/2015
6
Elements of Chemical Modeling
11
CM• Simulation of Structure: CME• Optimization of 10K+ Molecular Composition: CME
ME• Network Analysis: DelPlots• LFER Analysis: Organization of 10K+ Rate, Adsorption and Thermodynamic Parameters• Extrinsic Effects: Kinetic Coupling; Solvent Effects; Phase Behavior, Transport• Novel Reactors
CM-1
• 10K+ Multifunctional Molecules: Implicit Equations via Monte Carlo Simulation of Kinetics• Equation Generation: NetGen, OdeGen• Kinetic Coupling: Accounting for Interactions; Optimal System Design• Transport: Molecule vs. Moiety; Size and Matrix Effects• ARM: Modeling N molecules with fewer than N equations
Reaction Families Organize Pathways and Kinetics*
12
• Reaction Families:– Allow for computer-generated reaction paths
– Allow for massive parameter reduction via LFER organization
• Reaction Family: a set of reactionshaving similar transition states
index reactivity is where,ba = ln
H G
or 0
RIRIk
HSS
RI
lnk
ln k = a + b RI
*9/22/15
9/9/2015
7
Geometric Rationalization of LFER
13
• Physical picture of hypothetical reaction:Na + RCl NaCl + R•
• Variation of R gives Reaction Family
E*
q
BDE+Ip-Ae
†1
†2
Na+, Cl-, R•
Na, Cl•, R•
Na + RClR• + NaCl
BDE (R-Cl)
E*=q = ' RI
Elements of Chemical Modeling
14
CM• Simulation of Structure: CME• Optimization of 10K+ Molecular Composition: CME
ME• Network Analysis: DelPlots• LFER Analysis: Organization of 10K+ Rate, Adsorption and Thermodynamic Parameters• Extrinsic Effects: Kinetic Coupling; Solvent Effects; Phase Behavior, Transport• Novel Reactors
CM-1
• 10K+ Multifunctional Molecules: Implicit Equations via Monte Carlo Simulation of Kinetics• Equation Generation: NetGen, OdeGen• Kinetic Coupling: Accounting for Interactions; Optimal System Design• Transport: Molecule vs. Moiety; Size and Matrix Effects• ARM: Modeling N molecules with fewer than N equations
9/9/2015
8
DBE Pyrolysis Mechanisms*
15
Free-Radical Mechanism Concerted Mechanism
k1 +
+k2 +
+k2’ +
k3 +
+k4 products
O
HH
O
H
RDS
+
+
‡
‡
b
c
RDS
*11/17, 19
The kinetics of hydrogen abstraction provide the probe of the mechanism
16
The options for hydrogen abstraction:
+ A H + '
X'+ A XH +
Consequences:
+ XD D + X'
D favored as XD/A D/(D + H) = DI 1.0 as XD/A Kinetics analysis shows XH = D
Hydrogen Abstraction Kinetics
9/9/2015
9
Free-Radical Pyrolysis of PDD
171/R (Moles PDD/Mole tetralin - d )12
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.2 0.4 0.6 0.8 1.0
DI
Experimental Example of Kinetic Coupling
18
DBE enhanced with PPE PPE impeded with DBE
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 20 40 60 80 100time / min
Mo
lar
Yie
ld D
BE DBE Neat
5:1 DBE:PPE
1:1 DBE:PPE
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 20 40 60 80 100
time / min
Mo
lar
Yie
ld D
BE
PPE Neat
1:1 PPE:DBE
9/9/2015
10
Two-Component Rice-Herzfeld Pyrolysis
19
A1 2 1 A2 2
1 + A1 P1 1 2 + A2 2 + P2
1 1 + Q1 2 2 + Q2
1 + A2 2 + P1
2 + A1 1 + P2
1 + 1 i + j 2 + 2
1 + 1 i + j 2 + 2
1 + 1 i + j 2 + 2
k11
k1
1
''11
11
'11
k22
k2
2
''22
22
'22
k12
k21
ij
' ij''ij
Rate Laws for Coupled Binary Systems
20
12
12
1 1111 1 1 2 2 2
111 1 2
11 1
211 1 51 22 21 1 22 1 1 2 3 2 2 2 12 3
2
22 1
12
1 122 2 2
11
2 1 2
1 1
,1
1 1 1 11
1
,
Ak A k S S
r A AN
Nk S
M k S k S M k SN
Nk S
Ak A k
r A A
1
1 222 11 2 2 2 1 2
1 21 2 2 1
11 1
21 1 1 51 22 21 1 22 1 1 2 3 2 2 2 1
2 32
22 1
1 1 1
1
1 1 1 11
N NS S k S N N
k S k SN
Nk S
M k S k S M k SN
Nk S
9/9/2015
11
Chain Transfer Limiting Case
21
One Component Pyrolysis
A1 21
1 + A 1k11 1H +1 1 + A2
k12 1H + 2
1 k1 1 + Q1 2 + A1
k'21 A2 + 1
2 1t
Products1 + 2
t
1 + 2t Products
Additional Reactions with Solvent
Volcano Curve for Pyrolysis with Chain Transfer Solvent
22
120110100908070
0
1
2
Bond Dissociation Energy µ2-H [kcal/mol]
E
Fixing species 1 [d°(1-H) and d°(1-H)]
Pyrolysis with Chain Transfer Solvent
9/9/2015
12
Coal Liquefaction Efficiency
23
0
10
20
30
40
50
60
70
80
90
100
70 80 90 100
An-H2
Py-H2 Phen-H2
AceN
TetInd
Me-N
Bond Dissociation Energy / kcal mol
Coa
l Con
vers
ion
toT
HF
Sol
uble
s
UD Computational Tools*
• Develop usable approaches and software tools for molecule-based kinetic modeling of complex feeds
• Maintain fundamental chemistry as the primary focus
• Link quantum and engineering levels
• Produce models that are compatible across a process (e.g., refinery)
• Deliver in a user-friendly format
24*10/13, 15, 20 and 11/22
9/9/2015
13
UD Software Tools
Total Molecular Footprint
Feedstock Mole Fractions
Feedstock MoleculesMolecule Reactivity
Experimental Data on Feedstock
Interactive Network
Generator• Automatically builds reaction
network based on feedstock molecules, reaction chemistries, and reaction rules
Kinetic Model Editor• Builds and solves kinetic model
• Optimizes kinetic parameters using reactor experimental data
Composition Model Editor• Optimizes mole fractions
using experimental data
Experimental Data on Reactor
25
Interactive Network Generator
Kinetic Model EditorComposition Model Editor
UD Software Tools
26
9/9/2015
14
Composition Model
Editor
Interactive Network
Generator
Kinetic Model Editor
UD Supporting Apps and Tools
CoreGen• Creates core structures
HOUGen• Creates molecular identities of general hydrocarbon mixtures
PropGen• Estimates physical properties of molecules
Universal Sampling Builder• Sampling protocol between structural attribute pdfs and molecular compositions
Composition Model Visualizer• Bulk
Properties
Molecular Weight
CME‐Plastics• Light‐weight CME
applied to linear polymers
27
Interactive Network Generator
Kinetic Model EditorComposition Model
Editor
UD Supporting Apps and Tools
INGen Network Merge• Merges reaction networks, removes duplicates
Reaction NetworkVisualizer
File Converters• Molecule file conversion between BE matrix, adjlists, SMILES code, CML, MOPAC, CT, and images
28
9/9/2015
15
Kinetic Model Editor
Interactive Network
Generator
Composition Model
Editor
UD Supporting Apps and Tools
KME Results Analyzer• Visualize details of kinetic model solution
Reactions producing propane
TGA Simulator• Kinetic model for ThermogravimetricAnalysis
CodeGen• Converts reaction list to differential equations
KME Flowsheet• multiple reactors• simple separation
Milano Script Gen• Generates Milano‐style equations for a model
29
UD Supporting Apps and Tools
INGen Network Merge• Merges reaction networks, removes duplicatesReaction
NetworkVisualizer
KME Results Analyzer• Visualize details of kinetic model solution
Reactions producing propane
TGA Simulator• Kinetic model for ThermogravimetricAnalysis
CodeGen• Converts reaction list to differential equations
CoreGen• Creates core structures
KME Flowsheet• multiple reactors• simple separation
HOUGen• Creates molecular identities of general hydrocarbon mixtures
PropGen• Estimates physical properties of molecules
File Converters• Molecule file conversion between BE matrix, adjlists, SMILES code, CML, MOPAC, CT, and images
Universal Sampling Builder• Sampling protocol between structural attribute pdfs and molecular compositions
Composition Model Visualizer• Bulk
Properties
Molecular Weight
CME Express• Fast local optimization for similar compositions
Milano Script Gen• Generates Milano‐style equations for a model
CME‐Plastics• Light‐weight CME
applied to linear polymers
30
9/9/2015
16
Molecular-level Modeling Requirements
1. Composition Models: CME– Measurements (GS-MS, NMR, etc.)
– Modeling: Properties -> Molecular Structure
2. Reaction Network Modeling: INGen– N < NC: “Conventional” deterministic methods
– N > NC: Hybrid methods, e. g., The Attribute Reaction Model
3. Reaction Modeling: KME– Model parser and solver in various engineering modes
– Order 10 [O(10)] LFER’s for every process chemistry
4. Property Estimation: CME– Molecular Structure -> Properties
– End-use and internal-use properties
CCCCc1c2ccc(Sc3cc4cc(C)c5cc(CC)cc6cc(Sc7ccc(CCC)c(c7)c7cccc(CCCCCC)c7)c(c3)c4c56)cc2cc2sc3cc(C)ccc3c12
RI
lnk
ln k = a + b RI
31
GasolineNaphtha
Kerosene LGO HGO VGO
MW(g/mol) 83.2 130.5 168.5 211.5
345.1
469.0
H/C ratio 1.97 1.46 1.62 1.64 1.77 1.63
Distillation
IBP(K) 297 376 476 545 594 696
10BP(K) 302 437 479 546 633 705
30BP(K) 310 460 498 546 653 723
50BP(K) 318 460 517 559 669 743
70BP(K) 326 461 517 572 677 791
90BP(K) 334 461 532 584 690 805
EBP(K) 358 461 544 588 693 810
CME Converts Measurements to Molecules
CME
AttributePDFs
SimulatedCompositions
AvailableMeasurements
Molecules
SimulatedProperties
Adjust
32
9/9/2015
17
XH
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
Attribute Probability Distribution to Generate Molecular Species
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
Pro
babi
lity
XX
XXX
X
XO
X
XS
SX
X XO
H
X
O
OH
X
2. Optimization of Probability Distributions:• Initially uniform probability• Minimization of objective function
Cores Inter-unit Links Side Chains
33
Molecular-level Modeling Requirements
1. Composition Models: CME– Measurements (GS-MS, NMR, etc.)
– Modeling: Properties -> Molecular Structure
2. Reaction Network Modeling: INGen– N < NC: “Conventional” deterministic methods
– N > NC: Hybrid methods, e. g., The Attribute Reaction Model
3. Reaction Modeling: KME– Model parser and solver in various engineering modes
– Order 10 [O(10)] LFER’s for every process chemistry
4. Property Estimation: CME– Molecular Structure -> Properties
– End-use and internal-use properties
34
9/9/2015
18
Equivalent Computational Representations
Bond Electron Matrix Adjacency List
C(C(C(H3)H2)C(C(H3)H2)H2)String Code
CCCCCSMILES Code
35
Chemical Reaction as a Matrix Operation
+
C1 C2 H1 H2
C1 0 1 0 0
C2 1 0 0 0
H1 0 0 0 1
H2 0 0 1 0
C1 C2 H1 H2
C1 0 0 1 0
C2 0 0 0 1
H1 1 0 0 0
H2 0 1 0 0
C1 C2 H1 H2
C1 0 ‐1 1 0
C2 ‐1 0 0 1
H1 1 0 0 ‐1
H2 0 1 ‐1 0
+ =
Reduced Reactant Matrix
Reaction MatrixReduced
Product Matrix
+
MA MR MB
36
9/9/2015
19
Structure/Reactivityparameters
Rate law parameters
InputGraph of reactant molecule
Determine species type
Determine connected components
Determine products uniqueness (isomorphism)if unique, add to unreacted species list
Add reaction expression to the list
Obtain species from unreacted list
Determine reaction sites and apply reaction matrices
Determine Reactivity Index(CQC, GA, Data Base, etc.)
Network Building Algorithm
37
INGen Output: Species Analysis
38
9/9/2015
20
INGen Output: Network
KME Compatible
Network
Reaction Family
39
Co-Processing of BPO & HGO
Pyrolysis Unit
BiomassBPO
(Biomass Pyrolysis
Oil)
HGO (Heavy
Gas Oil)
HydrotreatingUnit
Commodity Chemicals
& Fuels1-α
α
Hydrogen
Water Water
9/9/2015
21
Lignin Monomers
41
Freudenberg Lignin Model as a String
Substituents: 39 Attributes
Cores : 19 Attributes
63 Connections:
CML Input from ChemDraw
O=CC=Cc1cc(OC)c(c(c1)c1cc(cc(c1O)OC)C1OCC2C1COC2c1ccc(c(c1)OC)OC(C(c1cc(OC)c(c(c1)Oc1ccc(cc1OC)C(C(Oc1ccc(cc1)C(C(Oc1ccc(cc1OC)C(C(Oc1ccc(cc1)C(C(CO)C)C)CO)Oc1ccc(cc1OC)C(c1cc(ccc1O)CC(=O)CO)C(CO)C)CO)O)CO)O)O)C)CO)OC(C(c1ccc(c(c1)OC)O)Oc1c(OC)cc(cc1c1cc(cc(c1O)OC)C1C2C(=O)OCC2C(c2c1cc(OC)c(c2)OC)O)C(C(Oc1ccc(cc1OC)C(c1cc(O)c(cc1C1Oc2c(C1CO)cc(cc2OC)CC(=O)CO)OC)C(Oc1c(OC)cc(cc1OC)C(C(CO)C)O)CO)CO)O)CO
The computer “knows” Lignin as its SMILES code
42
9/9/2015
22
Pyrolysis Reaction Chemistry
β-O-4 Ether Cleavage Reactions β-O-4 Ether Dehydration α -O-4 Ether Cleavage Vinyl Degradation Pathways
Virk, Michael T. Klein and Preetinder S. "Model Pathways in Lignin Thermolysis. 1. Phenethyl Phenyl Ether." Industrial Engineering Chemistry Fundamentals, 1983: 35-45.John B. McDermott, Cristian Libanati, Concetta LaMarca, and Michael T. Klein. Ind. Eng. Chem. Res., 1990: 22-29.
43
Model Results
• INGen Results: beta and alpha linkage reactions on adapted molecule• 1851 reactions • 295 species
• Kinetic Model Solution Time• C#=30, 2‐3 minutes• C#=60, 1 hour
44
9/9/2015
23
Attribute Reaction Modeling
45
Reducing the Freudenberg Lignin into its Attributes
O
O OH
O HO
O
OHO
OH
O
OH
OHO
HO
O
O
OH
OH
O
HO
O
OH
O
O
O
OOH
O
O
O
O
OH
O
O
O
OH
HOO
OH
OH
OO
O
O
OH
OH
O
O
HO
OH
O
O
OH
OOH
OH
O
O
O
Core Identity IL Identity SC Identity
O
O
O H
O
O H
CoresInter-core Linkages (IL)Side Chains (SC)
46
9/9/2015
24
Attribute Reaction Model Equations
47
Juxtaposition Tracks Molecules
XH
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
Pro
babi
lity
XX
XXX
X
XO
X
XS
SX
X XO
H
X
O
OH
X
Cores Inter-unit Links Side Chains
9/9/2015
25
Model Results
• Parity plot comparison with literature data
• Once-through solution time: 5 seconds
0
0.05
0.1
0.15
0.2
0.25
0.3
0 0.05 0.1 0.15 0.2 0.25
Predicted values
Observed Values
Data Set 4
Data Set 11
Data Set 2
Data Set 3
Data Set 12
Data Set 13
Data Set 14
Data Set 15
y=x
49
Computational Issues
• Conventional Model
– More than 2000 species (equations) and 10,000 reactions
– Solution time: minutes to hours depending on complexity
• Attribute Reaction Model
– Less than 100 equations and hundreds of reactions
– 5 second solution time
50
9/9/2015
26
Co-Processing of BPO & HGO
Pyrolysis Unit
BiomassBPO
(Biomass Pyrolysis
Oil)
HGO (Heavy
Gas Oil)
HydrotreatingUnit
Commodity Chemicals
& Fuels1-α
α
Hydrogen
Water Water
511
HGO Composition
• Aromatic species
– Cores
– Substituents
• Substituents are further divided into side chains and inter-core linkages
• Paraffins
– C5-C25 n-Paraffins
– C4-C15 2-methyl and 3-methyl Paraffins
52
9/9/2015
27
Hydrotreating Reaction Families*
Reaction Family log A E0*
C‐O Hydrogenolysis 10 21 H2 + →
Dealkylation 10 25 H2 + → +
Decarbonylation 12 6 → + CO
Decarboxylation 14 31 → + CO2
Denitrogenation 10 25 2H2 + → + NH3
Desulfurization 20 36 2H2 + → + H2S
Hydrogenation 10 20 H2 + →
Saturation 2H 10 25 H2 + →
Saturation 4H 10 25 2H2 + →
Saturation 6H 10 18 3H2 + →
Water Gas Shift 10 11 H2O + CO → H2 + CO2
S
53*10/22/15
Hydrotreating Reaction NetworkReaction Family BPO HGO Sum of Models Combined Model
COHydrogenolysis‐Oxygen 1,052 0 1,052 1,052
Dealkylation‐Aromatic 12 39 51 45
Dealkylation‐MultiRing 0 119 119 119
Dealkylation‐Nap6 12 39 51 45
Dealkylation‐Oxygen 269 0 269 269
Decarbonylation‐Oxygen 107 0 107 107
Decarboxylation‐Oxygen 21 0 21 21
HDN‐Nitrogen 1 4 5 4
HDS_4H‐Sulfur 0 4 4 4
Hydrogenation‐Aromatic 5 7 12 10
Hydrogenation‐IsoOlefin 0 1 1 1
Hydrogenation‐MultiRing 0 8 8 8
Hydrogenation‐Nap6 5 7 12 10
Hydrogenation‐NOlefin 2 3 5 3
Hydrogenation‐Oxygen 64 0 64 64
Saturation2H‐MultiRing 1 3 4 3
Saturation4H‐MultiRing 6 60 66 60
Saturation4H‐Nitrogen 0 2 2 2
Saturation4H‐Oxygen 8 0 8 8
Saturation6H‐Aromatic 18 45 63 51
Saturation6H‐MultiRing 6 130 136 130
Saturation6H‐Nitrogen 0 4 4 4
Saturation6H‐Oxygen 157 0 157 157
WaterGasShift‐Oxygen 1 0 1 1
Total 1,747 475 2,222 2,178
541
9/9/2015
28
Profile Plot of Major Species
0
10
20
30
40
50
60
70
80
90
100
0.E+00 2.E‐06 4.E‐06 6.E‐06 8.E‐06 1.E‐05
Molar Flow Rate (mol/s)
Reactor Length (dm)
0
2
4
6
8
10
12
14
0.E+00 2.E‐06 4.E‐06 6.E‐06 8.E‐06 1.E‐05
Molar Flow Rate (mol/s)
Reactor Length (dm)
water
methane
carbon dioxide
carbon monoxide
cyclohexane
methylcyclohexane
furan
1,3‐dimethylcyclohexane
oxepane
hexane
o‐cresol
2‐methoxyphenol
4‐hydroxy‐3‐methoxybenzaldehydephenol
6,8‐dioxabicyclo[3.2.1]octane‐2,3,4‐triol2‐methoxy‐4‐methylphenol
2,4‐dimethylphenol
formic acid
furan‐2‐carbaldehyde
2‐hydroxyacetaldehyde
acetic acid
hydrogen
PFR, 1,000 atm pressure, 800 K
551
Preliminary Status
• Quantitative organization of science and engineering knowledge base
• Identified some key technology gaps, science needs• Petroleum was inherently profitable, which allowed
for 100+ years of technology development• Biomass not inherently competitive, issues need to
be addressed• H2 consumption: expensive• H2 consumption: CO2 footprint• H2O release
• O rejection as H20 or CO2?• Energy density
56
9/9/2015
29
Conclusions
• Irreducible complexity, as measured by model parameters:
– Structure: O(10) 5-10 pdf’s 2-3 parameters each
– Reactivity: O(10) 10 LFER’s 2 parameters each
– Reaction Families:O(10) Reaction Matrices
• Rigorous kinetics can model interactions and competitions
• Net reduction of complexity: O(105 105) parameters reduced to
0(30)
• Combinatorial detail easily handled by
– INGen
– KME
– CME
57
CME Converts Measurements to Molecules
CME
AttributePDFs
SimulatedCompositions
AvailableMeasurements
Molecules
SimulatedProperties
Adjust
58
9/9/2015
30
XH
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
Attribute Probability Distribution to Generate Molecular Species
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
Pro
babi
lity
XX
XXX
X
XO
X
XS
SX
X XO
H
X
O
OH
X
2. Optimization of Probability Distributions:• Initially uniform probability• Minimization of objective function
Cores Inter-unit Links Side Chains
59
Molecular-level Modeling Requirements
1. Composition Models: CME– Measurements (GS-MS, NMR, etc.)
– Modeling: Properties -> Molecular Structure
2. Reaction Network Modeling: INGen– N < NC: “Conventional” deterministic methods
– N > NC: Hybrid methods, e. g., The Attribute Reaction Model
3. Reaction Modeling: KME– Model parser and solver in various engineering modes
– Order 10 [O(10)] LFER’s for every process chemistry
4. Property Estimation: CME– Molecular Structure -> Properties
– End-use and internal-use properties
60
9/9/2015
31
Lignocellulosic Biomass
[1] Yang H, Yan R, et al. Energy & Fuels. 2006. 20 (1), 388-393.[2] Freudenberg K, Neish AC. Science. 1969. 165 (3895), 784.
0%
10%
20%
30%
40%
50%
Composition, m
ol%
C
H
O
CelluloseC6H10O5
HemicelluloseC5H8O4
Lignin2
C10H12O3
61
Adapted Freudenberg Structure
Formed from three monolignols Phenolic copolymer
linkages• α-O-4 linkages• β-O-4 linkages
There is a propensity toward the β-O-4 linkage forming with it comprising upwards of 50-70% of linkages of the lignin [1]
[1] Lewis, Laurence B. Davin and Norman G. "Dirigent Proteins and Dirigent Sites Explain the Mystery of Specificity of Radical Precursor Coupling in Lignan and Lignin Biosynthesis." Plant Physiology, 2000: 453 462
62
9/9/2015
32
Exhaustive Method Reaction Network
Seven degradation pathways and many reactive sites on the lignin macromolecule create a combinatorial problem
– Within INGen the amount of species reaches a computational limit due to the amount of memory required to store species information
Two separate reaction networks were created for linkage and vinyl cleavages respectively and merged to create an overall network.
63
Probability Distributions to Mole Fractions
Pro
babi
lity
Side Chains
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
Cores Inter-unit Links
XH
XO
X
XS
SXX
X
XXX
X
XX XO
H
X
O
OH
XH
XH
XX
64
9/9/2015
33
• Inter-core Linkages cleave to two side chains
• Side chains react to form smaller side chains and irreducible molecules
• Cores can break down to form smaller cores and irreducible molecules
• Char formation
Reactions between Attributes
65
BPO Composition: Major Species
# Major Species Name ChemDraw NameNetwork Species #
Mole Fraction
1 water water species2 0.566
2 acetic acid acetic acid species110 0.047
3 glycolaldehyde 2‐hydroxyacetaldehyde species42 0.047
4 furfural furan‐2‐carbaldehyde species129 0.030
5 formic acid formic acid species127 0.025
6 2,4‐dimethylphenol 2,4‐dimethylphenol species28 0.023
7 4‐methylguaiacol 2‐methoxy‐4‐methylphenol species48 0.022
8 levoglucosan 6,8‐dioxabicyclo[3.2.1]octane‐2,3,4‐triol species107 0.017
9 phenol phenol species154 0.015
10 vanillin 4‐hydroxy‐3‐methoxybenzaldehyde species94 0.015
11 2‐methylphenol o‐cresol species145 0.013
12 guaiacol 2‐methoxyphenol species52 0.013
Total 0.834
661
9/9/2015
34
Effect of Catalyst on BPO Composition
Fast Pyrolysis
# Species Name wt%
1 dimethylphenol 7.0
2 levoglucosan 5.0
3 acetic acid 5.0
4 glycolaldehyde 5.0
5 furfural 5.0
6 methoxytoluene 4.5
7 methylguaiacol 4.2
8 vanillin 4.0
9 methylphenol 3.5
10 guaiacol 2.8
~~~ ~~~~~~~~~~~~~~~~~ ~~~ ~~~~~~~
129 4‐ethyl‐2,6‐dimethylphenol
0.05
Catalyzed Pyrolysis
# Species Name wt%
1 toluene 7.1
2 ethylbenzene 3.4
3 2‐methylnaphthalene 3.3
4 naphthalene 2.8
5 dimethylnaphthalene 2.5
6 methylphenanthrene 2.3
7 xylene 2.1
8 dimethylphenol 1.6
9 benzene 1.6
10 trimethylbenzene 1.5
~~~ ~~~~~~~~~~~~~~~~~ ~~~ ~~~~~~~
70 phenanthrol 0.01
67
HGO Composition: Cores
Aromatic Cores Sulfur-containing Cores
Nitrogen-containing Cores
681
9/9/2015
35
HGO Composition: Substituents
X
Inter-Core LinkagesSide Chains
Attribute Juxtaposition DetailsMonomers & dimersCoordination number: 4
691