Modeling the Ammonia Decomposition Reaction for Hydrogen ProductionDanielle A. Hansgen, Jingguang G. Chen, and Dionisios G. Vlachos

Center for Catalytic Science and Technology (CCST)Department of Chemical Engineering, University of Delaware, Newark, DE 19716 -2570

Motivation Interaction ParametersThe Vienna Ab-initio Simulation Package (VASP) was used to calculate heats of chemisorption as a function of adsorbate coverages from 1/9 to 1 monolayer (ML)

• The slope of the heat of chemisorption as a function of coverage is called the interaction parameter (Ixy)

Ammonia as a hydrogen carrier in a “hydrogen economy”• Ammonia is widely produced (100 million metric tonnes per year)• Liquid at low pressures (9.2 atmospheres at 298 K) • Compatible with the current infrastructure• Can be catalytically decomposed on-board to create CO free H2 for a fuel cell

1/9 ML 2/9 ML

1/3ML 2/3ML

Model Results

10

100

onve

rsio

n (%

)

Rh,

Co

Cu

FeIr Mo

Ru,

Ni

Pd

Pt Rh,

Co

Ru,

Ni

W

Plot of conversions at the end of the reactor at 950 K•When N-N and NHx-N interactions are included, the peak of the volcano curve shifts to metals with higher heats of chemisorption

Microkinetic Modeling

p ( xy)

Rational Catalyst Design• Microkinetic models are an inexpensive and efficient way to screen catalysts• Insights gained from models will help develop better catalysts

Interaction parameters were calculated for different adsorbates on ruthenium

• Strength of interactions are a function of atomic radius

1.0

Cu

Pt

Ir Pd Rh,

Co

Rh,

Co

Ru,

Ni

Ru,

Ni

Fe W Mo

1

Co

150140130120

Heat of nitrogen chemisorption (kcal/mol)

No Interactions N-N, NHx-N interactions

1.0

Cu

Pt Ir Pd Rh,

Co

Rh,

Co

Ru,

Ni

Ru,

Ni

Fe W Mo

2N* -> N2 +2* 950K 2N* -> N2 +2* 673KNH * > NH* +H* 950K

higher heats of chemisorption•Looking at surface coverages and sensitivity coefficients show why this shift occurs

Surface Coverages and Sensitivity Coefficients (SC) without Interactions

Qx(Өy) =Qxo-IxyӨy

The ammonia decomposition reaction consists of 12 elementary reaction steps

1: * *R NH NH+ → 7 * * * *R NH N H

Microkinetic Analysis-Modeling overall reaction in terms of elementary reactions

-No assumptions concerning the kinetically significant elementary steps, or most abundant reaction intermediate

3 2 21 32 2

N H N H→ +

a function of atomic radius• N-N interactions are much stronger than H-H interactions

N-N and H-H interaction parameters were calculated for a

f

MetalInteraction Parameters

INN IHH

Co -34 - 1.6

0.8

0.6

0.4

0.2

0.0

Cov

erag

e (M

L)

150140130120QN (kcal/mol)

N 950K N 673K H 950K H 673K NH3 950K NH3 673K

0.8

0.6

0.4

0.2

0.0

Sen

sitiv

ity C

oeffi

cien

t

150140130120

QN (kcal/mol)

NH2 -> NH +H 950K NH2* -> NH* +H* 673K NH3* -> NH2* +H* 950K NH3* -> NH2* +H* 673K

•For metals with heats of chemisorption higher than 125 kcal/mol, the surface is poisoned by nitrogen•For these metals the rate determining step is the removal of nitrogen based on the3 3

3 3

3 2

2 3

2

2

1: * *2 : * *3: * * * *4 : * * * *5 : * * * *6 : * * * *

R NH NHR NH NHR NH NH HR NH H NHR NH NH HR NH H NH

+ →→ ++ → ++ → ++ → ++ → +

2

2

2

2

7 : * * * *8 : * * * *9 : 2 * 2*10 : 2* 2 *11: 2 * 2*12 : 2* 2 *

R NH N HR N H NHR N NR N NR H HR H H

+ → ++ → +→ ++ →→ +

+ →

12 activation energies 12 pre-exponentialsOperating conditions

ConversionsSurface coverages

Reaction rates

PFRDifferential - algebraic

equation solver

number of metals• The averages and standard deviations are are shown in the figure to the right• Similar interaction parameters for all metals studied

IrMoNiPdPtRhRu

- 27- 43- 42- 32- 39- 37

- 2.7

+ 0.1- 1.6- 2.3- 0.5- 1.4

Average -36 -1.2

The binding energies of the NHx species were calculated

1.0

0.8

0.6

04vera

ge (M

L)

Cu

Pt

Ir Pd Rh,

Co

Rh,

Co

Ru,

Ni

Ru,

Ni

Fe W Mo

N 950K N 673K H 950K H 673K NH3 950K NH3 673K

1.0

0.8

0.6

0 4vity

Coe

ffici

ent

Cu

Pt Ir Pd

Rh,

Co

Rh,

Co

Ru,

Ni

Ru,

Ni

Fe W Mo

2N* -> N2 +2* 950K 2N* -> N2 +2* 673K NH2* -> NH* +H* 950K NH2* -> NH* +H* 673K NH3* -> NH2* +H* 950K NH3* -> NH2* +H* 673K

•For these metals, the rate determining step is the removal of nitrogen based on the sensitivity analysis

Surface Coverages and SCs with N-N and NHx-N Interactions

Model Inputs

3 differential equations6 algebraic equations

Activation Energies (Ea) - Bond Order Conservation (BOC)

Atomic heats of chemisorption (QH, QN)+

Gas phase molecular bond energies Coverage effects:density

Low coverage values: temperature programmed desorption (TPD)

g g pusing VASP at a coverage of 1/9 ML while varying nitrogen coadsorbate coverages (from 0 to 2/3 ML)

NHx-N interactions were also calculated through a DFT-BOC hybrid method where the DFT N-N interaction parameter was included in the heat of chemisorption input

The DFT-BOC hybrid method does well at approximating the NHx-N interactions and is computationally inexpensive

0.4

0.2

0.0

Cov

150140130120

QN (kcal/mol)

0.4

0.2

0.0

Sen

sitiv

150140130120

QN (kcal/mol)

• N-N interaction parameters were calculated through DFT• NHx-N interactions were calculated through the DFT-BOC hybrid method.• The amount of nitrogen on the surface is significantly reduced for most metals and the amount of hydrogen increased• When interactions are included the rate determining step is the removal of the

Future WorkPre-exponentials Fit to Ruthenium Experimental Dataa

(Assumed valid for All Metals)

Gas phase molecular bond energies

Activation energies for all 12 elementary steps

Coverage effects:density functional theory (DFT)

Conclusions

• Complete microkinetic model development• Calculate nitrogen – hydrogen adsorbate interactions using VASP

• Compare microkinetic models to experimental data

• A library of thermodynamically consistent microkinetic models has been developed for the first time to describe the chemistry on various single metals

• Nitrogen-nitrogen adsorbate interactions were estimated using VASP for metals with high nitrogen coverages and these interactions were included in the microkinetic models

• A linear correlation was used to estimate the heat of nitrogen chemisorption as

ReactionPre-exponential factor (s-1) or sticking coefficient (unitless)

NH3 + * NH3 * 1.5 x 10-4

NH3* NH3 + * 8.1 x 1011

NH3* + * NH2* +H* 2.0 x 1012

NH2* +H* NH3* +* 3.4 x 109

Experimental Operating Conditions/Parameters

When interactions are included, the rate determining step is the removal of the second hydrogen for most metals

Acknowledgments• US Department of Energy, Grant FG02-03ER15468• The Chen and Vlachos research groups

• Predict activity for more complex catalytic surfaces (bimetallics, surface defects, etc.) in order to develop higher activity catalysts using microkinetic modeling

a function of nitrogen coverage• Interactions reduced the amount of nitrogen adsorbed on the surface• Increased the amount of surface hydrogen• Increased the catalytic activity

• VASP calculations were used to confirm the validity of a DFT-BOC hybrid method of approximating NHx-nitrogen interactions

• This method was used to calculate NHx-nitrogen interactions for metals in the microkinetic library

NH2 H NH3 3.4 x 10NH2* +* NH* + H* 2.0 x 1012

NH* + H* NH2* +* 1.4 x 1010

NH* + * N* + H* 1.9 x 1012

N* + H* NH* + * 7.6 x 109

2N* N2 + 2* 1.7 x 1012

N2 + 2* 2N* 2.0 x 10-1

2H* H2 + 2* 1.1 x 1011

H2 + 2* 2H* 8.7 x 10-1

aA.B. Mhadeshwar, J.R. Kitchen, M.A. Barteau, and D.G. Vlachos, Catalysis Letters 2004, 96, 13-22.bJ.C. Slater, Journal of Chemical Physics 1964, 41, 3199-3204.

Diameter = 0.32 cmLength = 1.0 cm

Pressure =1.0 atmFlow rate = 70 sccm

Surface density = 13,100 cm2/cm3