High Temperature Water Gas Shift Reaction over Iron Oxide CatalystsRainee VanNatter, Carl RF Lund

Introduction

Kinetic Modeling

Computational Chemistry

?

Experimental Data

Reaction Mechanism

Model Catalyst Surface

Estimated Adsorption

& Activation Energies

Estimated Adsorption

& Activation Energies

CO + H2O ↔ H2 + CO2

Uses: Adjust synthesis gas to a H2:CO ratio needed for another process (such as ammonia synthesis); Hydrogen production; Reduction of carbon monoxide.

Commercial Process:Water gas shift reaction is reversible and exothermic. • Maximum conversion favored by low temperature • High reaction rate favored by high temperature.

Thus, reaction typically performed in 2 stages:• High temperature stage with an iron oxide catalyst containing 5-10% chromia; followed by• Low temperature stage with Copper/ZnO/alumina catalyst.

Quantum chemical calculations and microkinetic modeling are being carried out to study the water gas shift reaction over iron oxide catalysts. Cluster models corresponding to potential active sites on (100), (110), and (111) surfaces of an Fe3O4 crystal have been generated. The mechanistic thermochemistry of water-gas shift is being studied by computational chemistry using these model surfaces. At the same time, microkinetic modelling is being used to study various proposed redox and formate water gas shift mechanisms. These 2 sets of information can then be compared to discriminate between potential reaction mechanisms.

Selected Mechanisms being Studied

****

******

)5()4()3()2()1(

22

2

2

2

22

COCOCOOCO

OHHHOHHOOH

OHOH

Forming the Model Catalyst Surfaces: 2-Unit (100) Example

The Kinetic Model:2-Step Redox ‘Mechanism’ Example

Magnetite (Fe3O4) Unit Cell

Magnetite is ferrimagnetic. The spins of the octahedral and tetrahedral iron cations have opposite ‘directions’.

Electron ‘hopping’ between octahedral iron cations results in an effective charge of +2.5

Final Cluster Models

Selected Cluster Models with an Adsorbed Oxygen Adatom

The Computational Chemistry

Cleaving of (100) Surface from Magnetite unit cell

Selection of Cluster Atoms

Selected Results of the Kinetic Modeling & Computational Chemistry: A Comparison

2rxn1rxnO*ads

exo‡2rxn

exo‡1rxn

O*ads

SSS

HHH

Adjustable

Fixed

Parameters

(100)(110)

(111)

Half the Octahedral iron cations were assigned a charge of +2. The other half were assigned a charge of +3.

Tetrahedral iron cations assigned a spin of -5/2.

Magnetite Electronic Properties

Settings for the Initial Wavefunction Guess

Geometry Optimizations

Jaguar SettingsUnrestricted DFT, tzv** basis set, ultrafine DFT grids

Transition State Searches & Geometry Scans

exo‡jrxnHi*

adsH

Assumptions• Steady-State, Isothermal, Isobaric Plug-Flow Reactor• Ideal Gas Behavior• Uniform Catalyst Sites• Heats of Adsorption independent of Surface Coverage & Temperature within Experimental Range

• Adsorption Entropy ≈ -ideal gas translational entropy of the gas phase species at experimental Tavg.

• Simple Transition State Theory rate expressions

T, P, mcat 0 rates flow gasInlet

in expn %Conversio COy

0 rates flow gasInlet in simCOy Conversion

SimulatedSimulation Routine (Single

Data Point)GuessesParameter

Calculate Rates

vacCOOCO

OHvacOH

PKkPkrate

PKkPkrate

2

22

2

222

1

111

Material Balance Equations(ODE’s)

1

2

1

22

2

2

2 0

ratedmnd

ratedmnd

ratedmnd

ratedmnd

dmnd

H

CO

OH

CO

N

Surface Coverage Equations(NLAE’s)

1

0 21

Ovac

O rateratedm

d

Load Initial Parameter Guesses

Load Experiment

cati mnPT 0

Calculate Rate Constants

RTH

RSk

RTH

RSk

‡2

‡2

2

‡1

‡1

1

expexp

expexp

RTH

RSK

RTH

RSK

222

111

expexp

expexp

adsO

fO

fCO

fCO

adsO

fO

fH

fOH

HTHTHTHTH

HTHTHTHTH

)()()()(

)()()()(

2

22

2

1

adsO

fO

fCO

fCO

adsO

fO

fH

fOH

STSTSTSTS

STSTSTSTS

)()()()(

)()()()(

2

22

2

1

Repeat for entire data set of 189 Experiments

Calculate simulated conversion

0

,0

,CO

finalsimCOCO

n

nnsimkCOy

Integrate through the simulated reactor

Calculate Objective Function

k

kCOsim

kCO s,Experiment All

2exp,, yy

Adjust Parameters to minimize Objective

Function

(Nelder Mead Simplex Method)

Continue .

until Objective Function

Converges to a minimum value

Results of the 2-step Redox ‘Mechanism’ Fit

Θ(O)

Θ(vac)

Simulated Catalyst Coverage

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.900.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

R² = 0.909498965529318

CO Conversion

Measured (Experiment)

Sim

ulat

ed (S

hort

Red

ox 'M

echa

nism

')

-5% -4% -3% -2% -1% 0% 1% 2% 3% 4% 5%0%

500%

1000%

1500%

2000%

2500%

3000%

3500%

4000%Sensitivity of Fit to Parameters

ΔH_ads(O)ΔH_act(1)ΔH_act(2)ΔS_ads(O)ΔS_act(1)ΔS_act(2)

%Change in Parameter

%Ch

ange

in O

bjec

tive

Func

tion

• The 2-step redox model provided a good fit to the experimental results, with an R2 value of 0.91

• The model predicted a catalyst surface fully covered with adsorbed oxygen

• The model was very sensitive to the oxygen atom adsorption energy (predicted to be -605 kJ/mol), and insensitive to the other parameters.

2-step Redox ‘Mechanism’

****

2

22

COOCOOHOH

(100) (110) (111)

Estimation of Parameters: Examples

molkJE estimateadsO /8.933,

H2O*

H2O-Fe Distance

Results of Geometry Scan

0,‡n/desorptioadsorption 2 estimateexo

OHE

ads*CO

adsO*

adsHHO*

adsO*H

exo‡5rxn

exo‡4rxn

exo‡3rxn

exo‡2rxn

exo‡1rxn

exo‡5rxn

exo‡4rxn

exo‡3rxn

exo‡2rxn

exo‡1rxn

ads*CO

adsO*

adsHHO*

adsO*H

22

22

SSSS

SSSSS

HHHHH

HHHH

Adjustable

Fixed

Parameters

****

****

**

22

22

2

22

COCOCOHHHCOO

HHCOOHCHOOHCHOOOHCO

OHOHFormate Mechanism 1

OCOOCOOCOOOCO

OOHOHHOOHHOOOHOOHOOH

****

******

22

2

2

2

22

Redox Mechanism 2

0% 10% 20% 30% 40% 50% 60% 70% 80% 90%0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

R² = 0.920095537288159

CO Conversion, Measured

CO C

onve

rsio

n, S

imul

ated

Θ(H2O)Θ(HO*H)Θ(O)

Θ(CO2)

Θ(vac)

Average Simulated Site Coverage

-5% -4% -3% -2% -1% 0% 1% 2% 3% 4% 5%

Sensitivity to Parametershloc_H2O

hloc_HO*H

hloc_O

hloc_CO2

hact(1)

hact(2)

hact(3)

hact(4)

hact(5)

%Change in Parameter

• The redox model provided a good fit to the experimental results, with an R2 value of 0.92

• The model predicted a catalyst surface largely covered with adsorbed oxygen

• The model was most sensitive to the adsorption energies of dissociated water (-565 kJ/mol) and oxygen (-558 kJ/mol), and the activation energy of the hydrogen formation step (123.0 kJ/mol).

Formate Mechanism 2

****

******

22

22

2

22

COCOCOHHHCOO

HHCOOHHOCOHHOOH

OHOH

adsO*H

K.M.: 2-step Redox

K.M.: Redox1, 9-Param

K.M.: Redox1, 14-Param

(111) pre-capped, O + * → O*

(111) pre-capped, O + *O → O*O

(111), O + * → O*

(111), O + *O → O*O

(110) 3x, O + * → O*

(100) 2x, O + * → O*

(100) 2x, O + *O → O*O

(100) 4x pre-capped, O + * → O*

(100) 4x, O + * → O*

K.M.: Redox1, 9-Param

K.M.: Redox1, 14-Param

(111) pre-capped, H + OH + *O → HO*OH †

(110) 3x, H + OH + * → HO*H

(100) 2x, H + OH + * → HO*H

K.M.: Redox1, 9-Param ‡

K.M.: Redox1, 14-Param ‡

(111) 4x, H2O + * → H2O*

(111) 4x, Pre-capped, H2O + *O → H2O*O

(110) 3x, H2O + * → H2O*

(100) 2x, H2O + * → H2O*

K.M.: Redox1, 9-Param ‡

K.M.: Redox1, 14-Param ‡

(111) 4x, CO2 + * → CO2*

(111) 4x, Pre-capped, CO2 + *O → CO2*O

(110) 3x, CO2 + * → CO2*

(100) 2x, CO2 + * → CO2*

-605.0 kJ/mol

-557.5 kJ/mol

-581.6 kJ/mol

-849.9 kJ/mol

-668.8 kJ/mol

-680.9 kJ/mol

-701.5 kJ/mol

-404.5 kJ/mol

-163.0 kJ/mol

-22.4 kJ/mol

-193.2 kJ/mol

-130.7 kJ/mol

-565.0 kJ/mol

-604.5 kJ/mol

-810.4 kJ/mol

-446.5 kJ/mol

-505.6 kJ/mol

-108.7 kJ/mol

-135.8 kJ/mol

-62.2 kJ/mol

-31.0 kJ/mol

-158.6 kJ/mol

-54.9 kJ/mol

-123.0 kJ/mol

-148.1 kJ/mol

-103.1 kJ/mol

-37.5 kJ/mol

-69.1 kJ/mol

-51.2 kJ/mol

adsHHO*H

adsO*H2

H

ads*CO2

H

† HO*H not possible on this surface. ‡ This kinetic model is fairly insensitive to this parameter

Preliminary Results

• The fitted 2-step redox model’s sensitive parameter (hloc(O)) best agrees with the computational chemistry results for the (111) surfaces. The computational chemistry results for the (100) and (110) surfaces predict too weak of a bond with the surface for any reaction to happen.

• The adsorption energy of O estimated using the cluster models varies largely with coverage. The Kinetic modeling results are thus best compared with surfaces corresponding to the coverage (largely O-covered) predicted by the kinetic modeling results.

Future Work

• Larger clusters are being studied to examine the effect of cluster size on the results, and whether our current clusters are too small to obtain accurate estimates.

• Kinetic Modeling needs to be carried out for other mechanisms. This includes a redox mechanism utilizing neighboring Fe-sites (for the (111) surfaces) and various formate and carbonate mechanisms.

• Quantum chemistry computations need to be carried out for cluster models having occupied neighboring sites as predicted by the kinetic modeling, where not already performed.

• Transition state searches and geometry scans need to be performed for the majority of clusters to estimate activation energies.

Kinetic Modeling Results of Redox Mechanism 1Redox Mechanism 1