Computational study of the catalytic dehydrogenation of ...lib.ugent.be/fulltxt/RUG01/002/224/444/RUG01-002224444_2015_0001_AC.pdf · Computational study of the catalytic dehydrogenation

Timothy Huygens

of propane on Pt and Pt3Ga catalystsComputational study of the catalytic dehydrogenation

Academic year 2014-2015Faculty of Engineering and ArchitectureChairman: Prof. dr. ir. Guy MarinDepartment of Chemical Engineering and Technical Chemistry

Master of Science in Chemical EngineeringMaster's dissertation submitted in order to obtain the academic degree of

Counsellor: Stephanie SaerensSupervisors: Prof. dr. Marie-Françoise Reyniers, Dr. ir. Maarten Sabbe

Timothy Huygens

of propane on Pt and Pt3Ga catalystsComputational study of the catalytic dehydrogenation

Academic year 2014-2015Faculty of Engineering and ArchitectureChairman: Prof. dr. ir. Guy MarinDepartment of Chemical Engineering and Technical Chemistry

Master of Science in Chemical EngineeringMaster's dissertation submitted in order to obtain the academic degree of

Counsellor: Stephanie SaerensSupervisors: Prof. dr. Marie-Françoise Reyniers, Dr. ir. Maarten Sabbe

FACULTY OF ENGINEERING AND ARCHITECTURE

Department of Chemical Engineering and Technical Chemistry Laboratory for Chemical Technology

Director: Prof. Dr. Ir. Guy B. Marin

Laboratory for Chemical Technology

Declaration concerning the accessibility of the master thesis Undersigned, Timothy Huygens Graduated from Ghent University, academic year 2014-2015 and is author of the master thesis with title: Computational study of the catalytic dehydrogenation of propane on Pt and Pt3Ga catalysts The author(s) gives (give) permission to make this master dissertation available for consultation and to copy parts of this master dissertation for personal use. In the case of any other use, the copyright terms have to be respected, in particular with regard to the obligation to state expressly the source when quoting results from this master dissertation.

Laboratory for Chemical Technology • Technologiepark 914, B-9052 Gent • www.lct.ugent.be Secretariat : T +32 (0)9 33 11 756 • F +32 (0)9 33 11 759 • [email protected]

Acknowledgements The last five years as chemical engineering student have rushed past and now graduation is

already coming up. However, no graduation is possible without the final act: writing your

master thesis. During that last year, you put some much work and effort into this work so you

can be proud on the work that you summit at the end of May.

First of all, I want to thank prof. dr. ir. Guy B. Marin and prof. dr. Marie-Françoise Reyniers

for the opportunity to cooperate on the topic of “Design of Pt-based bimetallic catalysts for the

catalytic dehydrogenation of propane” and finish my Chemical Engineering studies at the

Laboratory of Chemical Technology, in the University of Ghent.

Special thanks goes to my coach ir. Stephanie Saerens for her yearlong dedication and support.

She was available to answer all my questions, even in the weekends, and without her, this work

should not be what it is today. Further, I would like to thank dr. ir. Maarten Sabbe for providing

me insights in the complicated matter, during our meetings and discussions.

Verder ben ik dankbaar voor de korte, maar leuke ontspanningsmomenten in ons thesislokaal.

Ook wil ik graag mijn familie bedanken voor de steun tijdens al mijn examens en ook mijn

thesis. Ik weet nog altijd niet of al die kaarsen geholpen hebben, maar ze kunnen zeker en vast

geen kwaad. Tot slot, wil ik mijn vriendin Eva bedanken. Samen hebben we vaak recht

tegenover elkaar gewerkt aan onze thesis en het is ons gelukt om deze mooi af te ronden. Nu

beginnen we aan een nieuw hoofdstuk in ons leven.

Timothy Huygens

May 2015

Computational study of the catalytic dehydrogenation of propane on Pt and Pt3Ga catalysts Timothy Huygens Supervisors: prof. dr. M.-F. Reyniers

dr. ir Maarten Sabbe Counsellor: ir. Stephanie Saerens Master’s dissertation submitted to obtain the academic degree of Master of Science in Chemical Engineering Department of Chemical Engineering and Technical Chemistry Chairman: prof. dr. ir. Guy Marin Faculty of Engineering and Architecture Academic year 2014-2015

Summary

Catalytic dehydrogenation of propane towards propylene provides an on purpose alternative for propylene production via steam cracking. Pt-based catalysts for light alkane dehydrogenation have been extensively studied but bimetallic catalysts exhibit more promising activity and selectivity towards propylene for this reaction. For a novel Pt-Ga/Mg(Ga)(Al)Ox catalyst developed earlier, with excellent catalytic properties, the most probable surface composition is the Pt3Ga composition. In this work, ab initio calculations are conducted with an optPBE vdW-DF functional to unravel the propane dehydrogenation characteristics. Primarily, the role of Ga in Pt-Ga catalysts is investigated on the Pt3Ga(111) catalyst model. The calculations on the Pt3Ga(111) model show a potential higher catalytic activity, increased selectivity towards propylene and a greater deactivation resistance with respect to Pt(111), confirming the experimental results on Pt-Ga/Mg(Ga)(Al)Ox. Furthermore, a cokes deactivated Pt catalyst model is proposed to determine the influence of the cokes on propane dehydrogenation on nearby empty Pt atoms. When the cokes enclose the intermediates, the catalytic activity drops significantly and propane dehydrogenation is unlikely to occur. However, the steric effects due the cokes are short-ranged and the catalytic activity is partially retained further away from the cokes. Near the graphene ribbon, a higher selectivity to propylene is found. Finally, a model for the stepped Pt sites is investigated to determine their role in the propane dehydrogenation characteristics. The step sites show an upward shift in catalytic activity with respect to the Pt terrace surface.

Keywords Catalytic propane dehydrogenation, propylene, ab initio calculations, Pt-Ga catalyst, cokes

deactivated Pt, undercoordinated Pt

Computational study of the catalytic dehydrogenation of propane on Pt and Pt3Ga catalysts

Timothy Huygens Supervisors: prof. dr. M.-F. Reyniers, dr. ir. Maarten Sabbe, ir. Stephanie Saerens

Abstract Catalytic dehydrogenation of propane

towards propylene provides an on purpose alternative for propylene production via steam cracking. Pt-based catalysts for light alkane dehydrogenation have been extensively studied but bimetallic catalysts exhibit more promising activity and selectivity towards propylene for this reaction. For a novel Pt-Ga/Mg(Ga)(Al)Ox catalyst developed earlier [1, 2], with excellent catalytic properties, the most probable surface composition is the Pt3Ga composition. [3] In this work, ab initio calculations are conducted with an optPBE vdW-DF functional to unravel the propane dehydrogenation characteristics. Primarily, the role of Ga in Pt-Ga catalysts is investigated on the Pt3Ga(111) catalyst model. The calculations on the Pt3Ga(111) model show a potential higher catalytic activity, increased selectivity towards propylene and a greater deactivation resistance with respect to Pt(111), confirming the experimental results on Pt-Ga/Mg(Ga)(Al)Ox. Furthermore, a cokes deactivated Pt catalyst model is proposed to determine the influence of the cokes on propane dehydrogenation on nearby empty Pt atoms. When the cokes enclose the intermediates, the catalytic activity drops significantly and propane dehydrogenation is unlikely to occur. However, the steric effects due the cokes are short-ranged and the catalytic activity is partially retained further away from the cokes. Near the graphene ribbon, a higher selectivity to propylene is found. Finally, a model for the stepped Pt sites is investigated to determine their role in the propane dehydrogenation characteristics. The step sites show an upward shift in catalytic activity with respect to the Pt terrace surface.

Keywords Catalytic propane dehydrogenation, propylene, ab initio calculations, Pt-Ga catalyst, cokes deactivated Pt, undercoordinated Pt

I. INTRODUCTION

Recently, the catalytic dehydrogenation of light

paraffins (C2-C3) has emerged as promising technology for on purpose production of light olefins, independent of the olefin production via steam cracking. The focus of this thesis is on the catalytic dehydrogenation of propane towards propylene. Propane dehydrogenation is an endothermic reaction that is limited by chemical equilibrium and the equilibrium constant is rather low for smaller carbon chains such as ethane and propane, so high temperatures in presence of a catalyst are required. Platinum metal catalysts are an adequate choice as

Timothy Huygens is with the Chemical Engineering Department, University (UGent), Ghent, Belgium. ([email protected])

catalyst for light alkane dehydrogenation as they exhibit a high activity for the dehydrogenation steps. However, those catalysts lack a high selectivity towards olefins, as many side reactions occur on the surface. Eventually these side reactions initiate coke formation, leading to unwanted deactivation of the catalyst. [4] To improve the catalytic properties of platinum-based catalysts, the metal phase is alloyed with an additional metal such as Ga, In, Sn.

II. COMPUTATIONAL METHODOLOGY

Periodic DFT calculations are carried out with the Vienna Ab Initio Simulation Package (VASP). A non-local van der Waals-density functional (optPBE vdW-DF) is employed in all calculations. The description of the atoms is conducted with Projected Augmented Wavefunction (PAW) pseudopotentials and a plane wave energy cutoff of 400 eV is employed.

A four-layers slab with 4×2 unit cell is employed to represent the Pt(111), Pt3Ga(111) and Gr/Pt(111) catalyst model. A Monkhorst-Pack grid of 3×5×1 is employed for the Brillion zone sampling. For the representation of Pt(211), a four-layers slab with 3×2 unit cell is employed. The Brillouin zone is sampled with a k-mesh of 5×5×1. The ground state geometries are obtained under strict convergence criteria of 10-8 eV/atom and 0.015 eV/Å.

To determine the transition states, the dimer method is employed in conjugation with the Nudged Elastic Band (NEB) method, to locate the saddle point on the potential energy surface (PES) along the reaction coordinate. The geometries of the intermediates and transition states are validated using vibrational analyses.

III. RESULTS To evaluate the propane dehydrogenation

characteristics on various catalyst models, reaction energies and barriers are calculated. Based on these parameters, energy profiles relative to gaseous propane are constructed. In order to satisfy the conservation of mass, the total energy of a dehydrogenated adsorbate is the sum of the electronic energy of the adsorbate and the dehydrogenated hydrogen atom(s).

A. Propane dehydrogenation kinetics on Pt(111) catalyst model

To determine the propane dehydrogenation characteristics on pure Pt catalysts as a reference for more advanced models, a 4×2 catalyst model of the most abundant Pt(111) phase is constructed. The considered elementary steps are illustrated in Figure 1.

Figure 1. Elementary steps involved in propane dehydrogenation to propylene and deep dehydrogenation.

The proposed reaction network is divided in three main sections: propane dehydrogenation to propylene, deep dehydrogenation of C3H6-species and hydrogenolysis of C3

intermediates. The elementary steps in the first section are employed to quantify the catalyst activity for propane dehydrogenation towards propylene.

In the second section, the deep dehydrogenation of the surface intermediates towards coke precursors is investigated. The selectivity towards these species is an indication for the catalyst deactivation kinetics.

In the third section, the hydrogenolysis reactions of the C3Hx-species (x=6-8) are considered. These reactions form C1- and C2-adsorbates on the surface. These species can eventually lead to gaseous side products such as methane, ethane and ethylene, which affect the selectivity towards gaseous propylene.

For all three sections, the thermodynamics and kinetics of propane dehydrogenation of the reactions involved are evaluated. The propane dehydrogenation reaction towards propylene can occur via 1-propyl and via 2-propyl but the reaction path via 2-propyl is preferred as it has a lower reaction barrier (12 kJ/mol) with respect to 1-propyl. However, as the Gibbs free energies at 873 K of these species and the reaction barriers of the elementary step are similar, it is expected that both reaction paths contribute to the formation of propylene. Both 1-propyl and 2-propyl preferentially dehydrogenate further to propylene, as it is the thermodynamically most stable C3H6-species. The resulted electronic activation energy is similar for both reactions.

Deep dehydrogenation leads to the thermodynamically most stable species, i.e. 1-propylidyne on the Pt(111) surface. Multiple reaction paths towards 1-propylidyne are possible if isomerization reactions are included. However, it is shown that the isomerization reactions are highly activated and unlikely to occur. The most probable path to coke precursors is via propane →1-propyl →1-propylidene →1-propylidyne, as the final step is kinetically favored. However, this is solely based on the electronic energy (enthalpic contributions).

Figure 2. Gibbs free energy diagram at 900 K for the propane dehydrogenation to propylene (···) and propane deep dehydrogenation to 1-propylidyne ( ̶ ̶ ) on Pt(111).

The entropic contributions are included by determining the Gibbs free energy of the intermediates and transition states. The entropic contributions are larger for gaseous species than for adsorbates, so based on the Gibbs free diagram at 900 K, gaseous propylene is 11 kJ/mol more stable than 1-propylidyne, see Figure 2.

Hydrogenolysis competes with dehydrogenation reactions, but C-C cleavage leads to the formation of less stable species with respect to the dehydrogenated products expect for the cracking of 2-propylidene. However, all hydrogenolysis reactions are higher activated than the competing dehydrogenation reactions. Therefore, it can be concluded that these reactions are to occur less for the considered species on the Pt(111) surface.

Microkinetic simulation of the reaction network shows a high catalytic activity for propylene formation at short (0.1 s) and intermediate (10 s) time scale. The most abundant surface species is initially 1-propylidyne, but at longer simulation time this species is converted to CH3C≡Pt and HC≡Pt as these species are thermodynamically favored.

B. Propane dehydrogenation kinetics on Pt3Ga(111) as a model for Pt-Ga catalysts

The most interesting bimetallic catalyst are Ga-promoted Pt catalysts. The recent experimental studies on Pt-Ga catalyst (on hydrotalcite support) during propane dehydrogenation have already shown an increase in catalytic activity and selectivity towards propylene with respect to an unmodified Pt catalyst on the same support. [2] In previous work, the PtxGay alloy at the catalytic surface of the bimetallic surface is determined by comparing frequencies of CO adsorption experiments with the results of Pt-Ga catalyst models with various Pt/Ga ratios. The best candidate of the different studied catalyst models is the Pt3Ga bulk alloy, with the lowest Ga content. [3]

Hence, to assess the role of Ga on propane dehydrogenation thermodynamics and kinetics on Pt-Ga catalysts, a 4×2 Pt3Ga(111) unit cell is constructed. Propane dehydrogenation towards propylene consists of two elementary steps and these can occur via two reaction pathways (via 1-propyl and via 2-propyl).

2-propyl

1-propylidene

10

CH2=CH-CH3,phys

CH2=CH-CH3(g)

13

_CH-CH-CH3 + H

PtPt2

_

Pt

= CH2-C-CH3 + H

Pt

_

Pt

_

Pt2

=

2-propenyl1-propenyl

C-CH2-CH3 + H

Pt3

_

Pt

≡

14 1517 20 21

4 5

= _CH-CH2-CH3 + H

Pt2 Pt2-propylidene

CH3-C-CH3 + H

Pt2

=

Pt

_

7 8

CH3-CH-CH3 + H

Pt Pt

_ _CH2-CH2-CH3+ H

Pt Pt

_ _

1-propyl

1 2

CH3-CH2-CH3,phys

CH3-CH2-CH3(g)Gaseous propane

0

Physisorbed propane

CH2-CH-CH3 + H

Pt

__

Pt

_

PtPropylene

11 12

1-propylidyne

Physisorbed propylene

Gaseous propylene

-75

-25

25

75

125

175

Gib

bs f

ree

ener

gy (k

J/m

ol)

Propane (g)

Propane, phys

1-propyl + H*

Propylene,phys + 2H*

Propylene +2H*

Propylene (g)+ 2H*

TS1

TS5

1-propylidene + 2H*

TS17TS4

1-propylidyne + 3H*

Reactant Product

1-propylidyne + 3/2 H2 (g)

Propylene (g) + H2 (g)

Figure 3. Relative electronic energy profile for the propane dehydrogenation towards propylene on Pt(111) (black) and Pt3Ga(111) (green).

On Pt3Ga(111), both 1-propyl are 2-propyl are more stabilized than on Pt(111). As 2-propyl is thermodynamically more stable than 1-propyl, the reaction pathway via 2-propyl selected to define the catalytic activity on Pt3Ga(111).

Furthermore, adsorbed propylene bonds stronger on the Pt3Ga(111) surface and the calculated reaction barriers are lower than on Pt(111), while the propylene desorption barrier remains constant, see Figure 3. It can be concluded that catalytic activity is increased on Pt3Ga(111). In contrast, Sn alloying lowers the catalytic activity. Sn is frequently used as Pt catalyst modifier for alkane dehydrogenation. [5]

In addition, the deep dehydrogenation of C3H6-species is investigated on Pt3Ga(111). The formation of these species is an indication for the deactivation of the catalyst. Consistent with Pt(111), 1-propylidyne is thermodynamically the most favored species on the Pt3Ga(111) surface. To quantify the selectivity to the coke precursors, the reaction energies and barriers of reaction path via 1-propyl and 1-propylidene towards 1-propylidyne are determined, see Figure 4.

This reaction path has higher reaction barriers on Pt3Ga(111) than on Pt(111) and consequently 1-propylidyne is not as kinetically favored as on Pt(111). Furthermore, the electronic energy difference between propylene and 1-propylidyne on Pt3Ga(111) is much smaller than on Pt(111). As it can be assumed that the entropic contributions for adsorbed species such as 1-propylidyne are similar on Pt3Ga(111) as on Pt(111), the relative Gibbs free energy of 1-propylidyne on Pt3Ga(111) with respect to Pt(111) is determined by the difference in electronic energy on Pt(111) and Pt3Ga(111). Based on this reasoning, it is expected that 1-propylidyne is less likely to be formed on Pt3Ga(111) than on Pt(111). This indicates that fewer coke precursors will be formed and it can be concluded that Ga reduces the formation of cokes on the surface and increases the stability of the catalyst with respect to Pt(111).

Finally, the hydrogenolysis of propane is studied on Pt3Ga(111) as this reaction forms C1- and C2-adsorbates, which are critical for the formation of gaseous side products such as ethane, ethylene and methane.

Figure 4. Relative electronic energy profile for the propane dehydrogenation to propylene (···) and propane deep dehydrogenation to 1-propylidyne ( ̶ ̶ ) on Pt3Ga(111).

However, this reaction is highly activated (191 kJ/mol), 4 kJ/mol more than on Pt(111) while the electronic activation energies for the competing dehydrogenation reactions are lower on Pt3Ga(111).

Hence, the formation of gaseous side products via this way is unlikely. This result is supported by the experimental results that show that the propylene selectivity is above 98% on Pt-Ga/ Mg(Ga)(Al)Ox for the total time on stream. [2]

C. Propane dehydrogenation kinetics on graphene ribbon covered Pt(111) as a model for cokes deactivated Pt catalysts

The coke formation on the surface of the Pt catalysts leads inherently to the deactivation of the catalyst surface. Furthermore, the effect of cokes deactivated Pt on nearby Pt atoms is investigated in terms of catalytic activity and further deactivation of the catalyst. [6] Graphene is selected as representation of the coke on the cokes deactivated Pt catalyst based on TEM observations. [7] The catalyst model Gr/Pt(111) is primarily constructed in a 4×2 unit cell, in which half of the supercell is covered with a continuous 1-D graphene ribbon. The catalytic activity on this catalyst model is poor as steric effects of the graphene ribbon strongly destabilize the adsorbates, see Figure 5. It is postulated that dehydrogenation reactions are improbable to occur on this catalyst model. Apart from the poor activity, an improved selectivity towards propylene with respect to deactivation reactions is observed.

Thermodynamically, 1-propylidyne is favored on the catalyst surface and an exothermic reaction energy is observed for its formation, but all preceding intermediates are more unstable than on clean Pt(111). The electronic energy difference between gaseous propylene and 1-propylidyne is smaller. Additionally, the approximation is made that the entropic contributions are similar on Pt(111) and Gr/Pt(111). This leads to the conclusion that the Gibbs free energy of 1-propylidyne is higher on Gr/Pt(111) than on Pt(111), indicating that nearby the graphene ribbon, deactivation is reduced.

-100

-75

-50

-25

0

25

50

75

100

125

150

Rel

ativ

een

ergy

(kJ/

mol

)

Propane (g)

Propane, phys2-propyl + H*


Propylene +2H*


TS2 TS7

Reactant Product

Propylene (g)+ 2H*

-125

-75

-25

25

75

125

Rea

ltive

ene

rgy

(kJ/

mol

)

Reactant Product

Propane (g)



Propylene + 2H*


TS1 TS2 TS7Propylene (g) + 2H*

1-propyl + H*

TS4

1-propylidene+ 2H*

1-propylidyne + 3H*


TS17

Figure 5. Comparison of reaction path towards propylene via 2-propyl on clean Pt(111) (black), 4×2 Gr/Pt(111) (blue) and 5×2 Gr/Pt(111) (purple), based on relative electronic energy.

Furthermore, the obtained reaction energy of propane hydrogenolysis is extremely high (265 kJ/mol) and the hydrogenolysis reaction barrier must be even higher, so this reaction is improbable.

It is clear that the steric effects in the 4×2 unit cell are too large due the graphene ribbon inclusion, so a 5×2 unit cell is constructed on which intermediates can adsorb further away from the graphene ribbon. This model restores most of its activity compared with the smaller unit cell, see Figure 5. The strength of the steric interactions is reduced as they have a small range, however repulsion by the graphene ribbon still affects the intermediates, but on a smaller scale. For the considered reaction path, the catalytic activity remains lower than on the clean Pt(111) catalyst model.

D. Propane dehydrogenation kinetics on Pt(211) as a model for undercoordinated sites of Pt catalysts

The initial coke formation and origin of gaseous side products can be assigned to the reactivity of undercoordinated sites such as steps and edges of the Pt catalyst. [7] As model for the undercoordinated sites on Pt catalysts, a stepped Pt(211) catalyst model is constructed. For this model, a 3×2 unit cell is employed to obtain a coverage of 0.17 ML, which is higher than the reference coverage on Pt(111) of 0.13 ML. Furthermore, the hydrocarbon intermediates are optimized in adsorption sites located near the step and the reaction path towards propylene via 2-propyl is investigated and compared with literature.

Both thermodynamically and kinetically, the steps are preferred as adsorption sites due to their higher reactivity than the terrace plane (Pt(111)), see Figure 6. However, it should be noted that by strengthening the propylene bonds with the surface, a higher reaction barrier is observed for propylene desorption, indicating that propylene is less likely to desorb on Pt(211) than on Pt(111). Furthermore, Yang et al. propose that step sites are more sensitive for deep dehydrogenation and hydrogenolysis reactions. [8] Deep dehydrogenation steps are thermodynamically and kinetically favored until formation of propyne. Consecutively, these deeply dehydrogenated species have low reaction barriers for hydrogenolysis, leading to C1 and C2-species on the catalyst surface.

Figure 6. Relatieve electronic energy profile for propylene reaction path via 2-propyl on (black) Pt(111) and (red) Pt(211). (respectively coverage of 0.13 ML and 0.17 ML).

IV. CONCLUSIONS Ga alloying of Pt catalysts shows superior catalytic

properties over alloying with Sn. With respect to the reference, the Pt3Ga catalyst model shows an increased catalytic activity for propane dehydrogenation. Furthermore, it is shown that the formation of coke precursors is reduced, indicating a higher deactivation resistance. The hydrogenolysis of propane is unlikely and higher selectivity towards propylene is obtained on this Pt-Ga catalyst.

On the cokes deactivated Pt catalyst, the propane dehydrogenation does not occur near the graphene ribbon due to steric effects. Beyond the short-ranged steric effects, the catalytic activity is retained, but it is lower compared to clean Pt(111) due to the repulsion by the graphene ribbon. Near the graphene ribbon, the selectivity towards coke precursors and hydrogenolysis is reduced.

Undercoordinated sites on the Pt catalysts are thermodynamically and kinetically preferred as sites for propane dehydrogenation towards propylene with respect to the terrace Pt surface.

REFERENCES 1. Sun, P.P., et al., Journal of Catalysis, 2010.

274(2): p. 192-199. 2. Siddiqi, G., et al., Journal of Catalysis, 2010.

274(2): p. 200-206. 3. Saerens, S., in Department of Chemical

Engineering and Technical Chemistry. 2014, University of Ghent: Ghent.

4. Vora, B.V., Topics in Catalysis, 2012. 55(19-20): p. 1297-1308.

5. Yang, M.L., et al., Acs Catalysis, 2012. 2(6): p. 1247-1258.

6. Vu, B., et al., Korean Journal of Chemical Engineering, 2011. 28(2): p. 383-387.

7. Peng, Z., et al., Journal of Catalysis, 2012. 286: p. 22-29.

8. Yang, M.L., et al., Physical Chemistry Chemical Physics, 2011. 13(8): p. 3257-3267.

-75

-25

25

75

125

175

225

Rel

ativ

e en

ergy

(kJ/

mol

)

ProductReactant

Propane (g)

Propane, phys 2-propyl + H*

Propylene,phys + 2H

Propylene + 2H*

TS2

TS7

Propylene,phys + H2(g)

Propylene (g) + H2(g)

-100

-50

0

50

100

150

Rel

ativ

e en

ergy

(kJ/

mol

)

Reactant Product

Propane (g)

Propane, phys

2-propyl + H*


Propylene +2H*

Propylene (g) + 2H*

TS2


TS7

Table of Contents i

Table of Contents Table of Contents ........................................................................................................................ i

List of symbols .......................................................................................................................... ix

Acronyms .............................................................................................................................. ix

Roman symbols ...................................................................................................................... x

Greek symbols ....................................................................................................................... xi

Chapter 1 Introduction ................................................................................................................ 1

1.1 Catalytic propane dehydrogenation ............................................................................. 1

1.2 Recent developments in Pt-based catalysts ................................................................. 3

1.2.1 The novel Pt-Ga/Mg(Ga)(Al)Ox catalyst .............................................................. 4

1.2.2 Cokes formation and deactivation on Pt-based catalysts ..................................... 5

1.3 Goal of this work ......................................................................................................... 5

1.4 Structure of this work .................................................................................................. 6

1.5 References .................................................................................................................... 6

Chapter 2 Literature review ........................................................................................................ 8

2.1 Catalysts for propane dehydrogenation reaction ......................................................... 8

2.1.1 Catalysis in general ............................................................................................... 8

2.1.2 Pure platinum catalysts ....................................................................................... 10

2.1.3 Bimetallic Pt catalysts ........................................................................................ 11

2.1.4 Bimetallic Sn-Pt catalysts ................................................................................... 13

ii Table of Contents

2.1.4.1 Light alkane dehydrogenation .................................................................... 15

2.1.5 Bimetallic Ga-Pt catalysts .................................................................................. 16

2.1.5.1 Light alkane dehydrogenation .................................................................... 17

2.1.5.2 Determination of active phase of Pt-Ga catalysts ....................................... 19

2.2 Coke formation during propane dehydrogenation..................................................... 20

2.2.1 Coke formation on pure Pt catalysts .................................................................. 20

2.2.2 Coke formation on bimetallic Sn-Pt catalysts .................................................... 23

2.2.3 Coke formation on bimetallic Ga-Pt catalysts ................................................... 24

2.3 Interaction between active metal phase and support ................................................. 25

2.3.1 Active metal phase ............................................................................................. 26

2.3.2 Catalyst support.................................................................................................. 29

2.3.3 Active metal phase and support interaction ....................................................... 31

2.4 Conclusions ............................................................................................................... 32

2.5 References ................................................................................................................. 35

Chapter 3 Computational methodology ................................................................................... 39

3.1 Catalyst models ......................................................................................................... 39

3.1.1 Adaptations of the ideal catalyst model ............................................................. 41

3.1.2 Catalyst model used in this work ....................................................................... 43

3.2 Periodic ab initio calculations ................................................................................... 44

3.3 Computational framework used in this work ............................................................ 47

3.3.1 Vienna Ab initio Simulation Package ................................................................ 47

3.3.2 Geometry optimizations ..................................................................................... 49

3.3.3 Transition state calculation techniques .............................................................. 50

3.3.3.1 Nudged Elastic Band (NEB) method .......................................................... 51

3.3.3.2 Dimer method ............................................................................................. 52

3.3.4 Frequency calculations ....................................................................................... 53

Table of Contents iii

3.3.4.1 Dimmins.pl script ........................................................................................ 54

3.4 References .................................................................................................................. 54

Chapter 4 Propane dehydrogenation kinetics on Pt(111) catalyst model ................................. 57

4.1 Catalyst model ........................................................................................................... 57

4.1.1 Adsorption site nomenclature ............................................................................. 59

4.1.2 Determination of the degree of coverage ........................................................... 60

4.1.3 Selection of the DFT functional ......................................................................... 60

4.2 Adsorption ................................................................................................................. 61

4.2.1 Alkanes ............................................................................................................... 61

4.2.1.1 Propane ........................................................................................................ 62

4.2.1.2 Ethane .......................................................................................................... 63

4.2.1.3 Methane ....................................................................................................... 65

4.2.2 Alkenes ............................................................................................................... 66

4.2.2.1 Propylene ..................................................................................................... 66

4.2.2.2 Ethylene ....................................................................................................... 69

4.3 Thermodynamics ....................................................................................................... 71

4.3.1 Propane dehydrogenation to propylene .............................................................. 75

4.3.2 Deep dehydrogenation of propylene and other C3H6 species ............................. 79

4.3.3 Hydrogenolysis of C3 intermediates ................................................................... 81

4.3.3.1 Formation of gaseous side products ............................................................ 83

4.3.4 Overall thermodynamics..................................................................................... 84

4.4 Kinetics ...................................................................................................................... 85

4.4.1 Propane dehydrogenation to propylene .............................................................. 86

4.4.2 Deep dehydrogenation of propylene and other C3H6-species ............................ 90

4.4.3 Hydrogenolysis of C3 intermediates ................................................................... 93

4.4.3.1 Formation of gaseous side products ............................................................ 94

4.5 Microkinetic modelling .............................................................................................. 96

iv Table of Contents

4.5.1 Determination of the thermodynamic and kinetic parameters ........................... 97

4.5.2 Construction of the microkinetic and reactor model .......................................... 97

4.5.3 Results of the simulation .................................................................................... 98

4.5.3.1 Effect of varying the simulated time .......................................................... 99

4.5.3.2 Effect of varying temperature ................................................................... 102

4.5.3.3 Effect of varying propylene pressure ........................................................ 105

4.5.3.4 Effect of varying side product pressure .................................................... 107

4.6 Conclusions ............................................................................................................. 109

4.7 References ............................................................................................................... 111

Chapter 5 Propane dehydrogenation kinetics on Pt3Ga(111) as a model for Pt-Ga catalysts 114

5.1 Catalyst model ......................................................................................................... 115

5.1.1 Adsorption site nomenclature .......................................................................... 116

5.1.2 Determination of the degree of coverage ......................................................... 117

5.2 Adsorption ............................................................................................................... 117

5.2.1 Propane............................................................................................................. 117

5.2.2 Propylene ......................................................................................................... 118

5.3 Thermodynamics ..................................................................................................... 119

5.3.1 Propane dehydrogenation to propylene............................................................ 121

5.3.2 Deep dehydrogenation of propylene ................................................................ 124

5.3.3 Hydrogenolysis of C3 intermediates ................................................................ 126

5.4 Kinetics .................................................................................................................... 127


5.4.2 Deep dehydrogenation of propylene ................................................................ 129

5.4.3 Hydrogenolysis of C3 intermediates ................................................................. 131

5.5 Conclusions ............................................................................................................. 132

5.6 References ............................................................................................................... 133

Table of Contents v

Chapter 6 Propane dehydrogenation kinetics on graphene ribbon covered Pt(111) as a model

for cokes deactivated Pt catalysts ........................................................................................... 123

6.1 Catalyst model ......................................................................................................... 124

6.2 Adsorption ............................................................................................................... 127

6.2.1 Propane ............................................................................................................. 128

6.2.2 Propylene .......................................................................................................... 128

6.3 Thermodynamics ..................................................................................................... 129

6.3.1 Propane dehydrogenation to propylene ............................................................ 132

6.3.2 Deep dehydrogenation of propylene................................................................. 135

6.3.3 Hydrogenolysis of C3-intermediates ................................................................. 137

6.4 Kinetics .................................................................................................................... 137

6.4.1 Propane dehydrogenation to propylene ............................................................ 138

6.4.2 Deep dehydrogenation of propylene................................................................. 140

6.5 The range and strength of cokes influence on propane dehydrogenation ................ 142

6.5.1 Thermodynamics .............................................................................................. 142

6.5.2 Kinetics ............................................................................................................. 144

6.6 Conclusions .............................................................................................................. 146

6.7 References ................................................................................................................ 146

Chapter 7 Propane dehydrogenation kinetics on Pt(211) catalyst model ............................... 159

7.1 Catalyst model ......................................................................................................... 159

7.1.1 Adsorption site nomenclature ........................................................................... 160

7.1.2 Determination of the degree of coverage ......................................................... 161

7.2 Adsorption ............................................................................................................... 161

7.2.1 Propane ............................................................................................................. 161

7.2.2 Propylene .......................................................................................................... 162

7.3 Thermodynamics ..................................................................................................... 163

vi Table of Contents


7.4 Kinetics .................................................................................................................... 167

7.5 Conclusions ............................................................................................................. 169

7.6 References ............................................................................................................... 170

Chapter 8 Conclusions and prospects .................................................................................... 171

Future work ........................................................................................................................ 176

References .......................................................................................................................... 176

Appendix A Examples of INCAR files .................................................................................. 178

A.1 INCAR file for strict geometry optimization .......................................................... 178

A.2 INCAR files for transition state optimization ......................................................... 179

A.2.1 NEB optimization ............................................................................................ 179

A.2.2 Dimer optimization .......................................................................................... 180

A.3 INCAR file for frequency calculations ................................................................... 181

Appendix B Optimized geometries on the Pt(111) catalyst model ........................................ 182

B.1 Reaction network ..................................................................................................... 183

B.2 Optimized geometries.............................................................................................. 184

B.2.1 Gasphase species .............................................................................................. 184

B.2.2 Hydrogen .......................................................................................................... 186

B.2.3 (Deeply) dehydrogenated intermediates .......................................................... 187

B.2.4 Dissociation products ....................................................................................... 189

B.2.5 C1 and C2 hydrocarbon intermediates .............................................................. 190

B.2.6 Transition states ............................................................................................... 191

Appendix C Thermodynamic and kinetic parameters of the elementary steps for the Pt(111)

catalyst model ........................................................................................................................ 195

C.1 Thermodynamic parameters .................................................................................... 195

C.2 Kinetic parameters ................................................................................................... 196

Table of Contents vii

Appendix D Optimized geometries on the Pt3Ga(111) catalyst model .................................. 198

D.1 Reaction network ..................................................................................................... 199

D.2 Optimized geometries .............................................................................................. 200

D.2.1 Gasphase species .............................................................................................. 200

D.2.2 Hydrogen .......................................................................................................... 200

D.2.3 (Deeply) dehydrogenated intermediates ........................................................... 200

D.2.4 Dissociation products ....................................................................................... 202

D.2.5 Transition states ................................................................................................ 202

Appendix E Optimized geometries on the graphene ribbon covered Pt(111) catalyst model 204

E.1 Reaction network ..................................................................................................... 205

E.2 Optimized geometries in the 4×2 unit cell ............................................................... 206

E.2.1 Gasphase species .............................................................................................. 206

E.2.2 Hydrogen .......................................................................................................... 207

E.2.3 (Deeply) dehydrogenated intermediates ........................................................... 207

E.2.4 Dissociation products ....................................................................................... 209

E.2.5 Transition states ................................................................................................ 209

E.3 Optimized geometries in the 5×2 unit cell ............................................................... 210

E.3.1 Gasphase species .............................................................................................. 210

E.3.2 Hydrogen .......................................................................................................... 211

E.3.3 Dehydrogenated species ................................................................................... 211

E.3.4 Transition states ................................................................................................ 212

Appendix F Optimized geometries on the Pt(211) catalyst model ......................................... 213

F.1 Reaction network ..................................................................................................... 214

F.2 Optimized geometries .............................................................................................. 214

F.2.1 Gasphase species .............................................................................................. 214

F.2.2 Hydrogen .......................................................................................................... 215

viii Table of Contents

F.2.3 Dehydrogenated intermediates ......................................................................... 215

F.2.4 Transition states ............................................................................................... 216

List of symbols ix

List of symbols

Acronyms

ALISS Alkali ion scattering spectroscopy BEEF Bayesian Error Estimation Functional BET Brunauer–Emmett–Teller DFT Density functional theory DP Deposition-precipitation EDX Energy-dispersive X-ray spectroscopy EXAFS Extended X-ray adsorption fine structure fcc Face-centered cubic crystal structure or hollow adsorption site FFT Fast Fourier transformation FTIR Fourier transform infrared spectroscopy GGA Generalized gradient approximation hcp Hollow adsorption site HREELS High-resolution electron energy loss spectroscopy HRTEM High-resolution transmission electron microscopy IMP Incipient wetness impregnation IRAS Infrared adsorption spectroscopy IUPAC International Union of Pure and Applied Chemistry LDA Local density approximation LEED Low energy electron diffraction MA Mesoporous alumina MAS NMR Magic angle spinning nuclear magnetic resonance MD Molecular dynamics MEP Minimum energy path MO Molecular orbital ML Monolayer NEB Nudged Elastic Band PAW Projector-Augmented Wave PBE Perdew-Burke-Ernzerhof functional PES Potential energy surface

x List of symbols

PSSA Pseudo stationary state approximation PW91 Perdew-Wang 91 functional RAIRS Reflection-absorption infrared spectroscopy SMSI Strong metal-support interactions SOMC Surface organometallic chemistry STEM Scanning transmission electron microscopy TAP Temporal Analysis of Products TEM Transmission electron microscopy TEOM Tapered element oscillating microbalance TGA Thermogravimetric analysis TOF Turn-over frequency TPD Temperature programmed desorption TPO Temperature programmed oxidation TPR Temperature programmed reduction TST Transition state theory VASP Vienna Ab initio Simulation Package vdW van der Waals vdW-DF van der Waals density functional XANES X-ray absorption near edge structure XPS X-ray photoelectron spectroscopy XPD X-ray photoelectron diffraction XRD X-ray diffraction ZPE Zero point energy

Roman symbols

A Pre-exponential factor (for 1st order reaction) s-1 as Asymmetrical vibration mode - Ct Concentration of active sites mol kgcat

-1

Ea Activation energy J mol -1 Ead,ads Electronic energy of the adsorbed species J mol -1

Ead,gas Electronic energy of the adsorbate species in the gas phase J mol -1 𝐸𝐸C1𝐻𝐻𝑧𝑧/surface Electronic energy of the C1Hz-species on the surface J mol -1 𝐸𝐸C2𝐻𝐻6(g) Electronic energy of gaseous ethane J mol -1 𝐸𝐸C2𝐻𝐻𝑦𝑦/surface Electronic energy of the C2Hy-species on the surface J mol -1 EC2𝐻𝐻𝑦𝑦 Relative electronic energy of the C2Hy-species J mol -1 𝐸𝐸C3𝐻𝐻8(g) Electronic energy of gaseous propane J mol -1 EC3𝐻𝐻𝑥𝑥 Relative electronic energy of the C3Hx-species J mol -1 𝐸𝐸C3𝐻𝐻𝑥𝑥/surface Electronic energy of the C3Hx-species on the surface J mol -1 𝐸𝐸C3𝐻𝐻𝑥𝑥−1/surface Electronic energy of the C3Hx-1-species on the surface J mol -1 𝐸𝐸C3Hy+z/surface Electronic energy of the C3Hy+z-species on the surface J mol -1

List of symbols xi

𝐸𝐸H/surface Electronic energy of the adsorbed hydrogen J mol -1 Eslab or Esurface Electronic energy of the clean catalyst surface J mol -1 Ereact,ads Electronic energy of the adsorbed reactant(s) J mol -1 Eprod,ads Electronic energy of the adsorbed reaction product(s) J mol -1 ETS,ads Electronic energy of the transition J mol -1 F Molecular flux s-1 Fi Molar flow of component i mol s-1

𝐹𝐹𝑖𝑖NEB Artificial nudged elastic band force on NEB image i N 𝐹𝐹𝑖𝑖⊥ Perpendicular force of NEB image i N 𝐹𝐹𝑖𝑖

S∥ String force of NEB image i N k Rate coefficient s-1 kB Boltzmann constant J K-1 nt Number of active sites per m² m-2

s Symmetrical vibration mode - s0 Sticking coefficient - W Catalyst mass kg

Greek symbols

Г Special gamma point to sample the Brillouin zone - γ Vibrational scissoring mode - ∆ǂ𝐸𝐸 Electronic reaction barrier J mol -1 ∆adsE(ad) Electronic adsorbate adsorption energy J mol -1 ∆Er Electronic reaction energy J mol -1 δ Vibrational deformation mode - ν Vibrational stretching mode - τ Vibrational twisting mode - τ�𝑖𝑖 Tangent between NEB image i and image i+1 - ω Vibrational wagging mode -

Chapter 1: Introduction 1

Chapter 1 Introduction

1.1 Catalytic propane dehydrogenation

Recently, the catalytic dehydrogenation of light paraffins (C2-C3) has emerged as promising

technology for on-purpose production of light olefins. This work focuses on the production of

propylene (C3) by catalytic propane dehydrogenation. The global propylene demand for

polymer grade propylene is expected to double in 2030 with respect to 2005. [1] Nowadays, the

main production route to ethylene and propylene is steam cracking of ethane or naphtha.

However, the main product of this route is ethylene and the yield of co-product propylene is

limited to 13-17 wt%, independent of the feedstock. [2] Innovation has led to new on-purpose

production routes towards propylene without interference with steam cracker production, such

as methanol-to-olefins (MTO) [3] and propane dehydrogenation (PDH) [4]. Additionally, the

latter production route is applicable for exploitation of shale gas (which consists of a large

proportion of propane). [5]

Dehydrogenation processes of paraffins are chemical reactions in which alkanes are converted

into the corresponding alkenes by double bond formation. The resulting release of hydrogen

gas is considered as valuable side product. Dehydrogenation of propane (C3H8) leads to the

formation of propylene (C3H6) and hydrogen (H2).

C3H8 → C3H6 + H2 (1)

Paraffin dehydrogenation chemistry is strongly endothermic and its conversion is limited by

chemical equilibrium. According to Le Châtelier’s principle, higher conversion is obtainable at

either higher temperatures or lower pressures. [6] To obtain 40% conversion at 1 atmospheric

2 Chapter 1: Introduction

pressure absolute, the dehydrogenation of propane requires a temperature of at least about 853

K. [2] At these temperatures, side reactions such as hydrogenolysis and coke formation are

unavoidable. The use of a catalyst will increase the overall conversion of the selective

dehydrogenation of propane at similar reaction temperatures. [7] Furthermore, a catalyst can

also improve selectivity by attenuating the side reactions.

Since the late 1930’s, paraffin catalytic dehydrogenation for the production of olefins has been

commercialized. During World War II, chromia-alumina catalysts have been used for butane

dehydrogenation towards butenes, which were then converted to yield high-octane fuel. [7]

However, chromia-alumina catalysts are less frequently used as the regenerated catalyst

contains Cr(IV), which poses a significant health and environmental risk as it is a well-known

carcinogenic compound. [8]

In the 1940’s, Haensel proved that noble Pt catalysts have a high activity for the

dehydrogenation of paraffins to the corresponding olefins. [9] Bloch developed Pt-based

catalysts in the 1960’s that were used successfully in the production of biodegradable detergents

based on dehydrogenation of heavy paraffins. These Pt-based catalysts excelled at heavy

paraffin dehydrogenation as they exhibited high activity and stability with a minimum of

cracking. [10] Based on the promising results of heavy paraffin dehydrogenation on Pt-based

catalysts, it was expected that the extrapolation to light paraffins would be straightforward.

However, the required temperature for dehydrogenation of light paraffins is much higher than

for the dehydrogenation of heavy paraffins. As mentioned above, paraffin dehydrogenation is

equilibrium limited and smaller equilibrium coefficients are obtained for paraffin

dehydrogenation of smaller saturated hydrocarbons. Therefore, there is a larger driving force

for the dehydrogenation of heavy paraffins than of light paraffins. To obtain similar conversion

for the dehydrogenation of light paraffins as heavy paraffins, a higher temperature is required,

as illustrated in Figure 1-1.


Figure 1-1. Temperatures required to achieve 10 and 40% equilibrium conversion of C2–C15 n-paraffins at 1 atm. [7] The equilibrium conversions are based on the equilibrium constant of the corresponding paraffin dehydrogenation.

The requirement of higher temperatures during light paraffin dehydrogenation causes

acceleration of side reactions and catalyst deactivation. Suppression of these reactions can be

achieved by a careful catalyst, selection, e.g. a well-chosen Pt-based catalyst. Therefore,

catalyst development for the dehydrogenation of light paraffins has led to new, promising

catalysts.

1.2 Recent developments in Pt-based catalysts

Pure platinum is a well-known active metal phase for dehydrogenation reactions of light

alkanes. Theoretical studies, performed by Yang et al. [11, 12] and Valcarcel et al. [13]

proposed a reaction network for the propane dehydrogenation reaction, see Figure 1-2.

However, platinum suffers from low olefin selectivity and fast deactivation due to coke

accumulation. By promoting Pt with modifiers such as Sn or Ga, coke formation on the active

phase can be reduced, resulting in a higher olefin selectivity of the catalyst. [14, 15] The

catalytic properties of Pt-Sn catalysts are intensively studied both experimentally and

theoretically. However, recent studies have highlighted the promising role of Ga on the catalytic

properties of Pt-based catalysts during light paraffin dehydrogenation. Furthermore, a new Pt-

Ga catalyst has been developed for paraffin dehydrogenation reactions at the Laboratory for

Chemical Technology. [16]


Figure 1-2: Elementary dehydrogenation steps involved in propane dehydrogenation on Pt(111). Each row consists of C3Hx-species with one less hydrogen until C3H3. The dehydrogenated hydrogens are not shown. Employed color code: C (grey), H (white) and Pt (blue). [12]

1.2.1 The novel Pt-Ga/Mg(Ga)(Al)Ox catalyst

A new hydrotalcite-supported Pt-Ga catalyst has been developed for alkane dehydrogenation

reactions. The novel Pt-Ga/Mg(Ga)(Al)Ox catalyst is active for light paraffin dehydrogenation

and remains highly selective for the formation of the corresponding olefin. [14] Furthermore,

experiments have shown that the catalyst is less prone to surface carbon formation and exhibits

higher stability during regeneration cycles than Pt/HT. Ab initio techniques aid to determine

the surface composition of the Pt-Ga catalyst since modern spectroscopic techniques have been

applied to the catalyst and they do not provide a decisive answer. Based on a comparison of

calculated C-O stretch frequencies of adsorbed CO with experimental FTIR frequencies, Pt3Ga

is proposed as most probable surface composition of the Pt-Ga/Mg(Ga)(Al)Ox catalyst. [17]

The novel Pt-Ga/Mg(Ga)(Al)Ox catalyst shows superior qualities for the propane

dehydrogenation reaction compared to the pure platinum catalyst. However, the origin of the

increased selectivity and activity of the promoted catalyst is unclear.


1.2.2 Cokes formation and deactivation on Pt-based catalysts

Pt-based catalysts are prone to deep dehydrogenation reactions, which can eventually lead to

coke deposits on the surface. However, the coke formation cannot be directly coupled to

deactivation of Pt-based catalysts as only a small part of the formed coke deposits on the active

metal phase, while the majority is formed on the support or spills over onto the support from

the active phase. [18, 19] Furthermore, studies have shown that undercoordinated sites such as

edges and steps are preferred for initiating the coke formation on Pt-based catalysts. [20]

Characterization of the coke deposits could lead to further understanding of the coke formation

mechanism and enabling the minimization of coke formation. Recent TEM studies have shown

that cokes formed during propane dehydrogenation has a graphitic nature. [21] However, the

exact effect of cokes formation on the reaction thermodynamics and kinetics remains elusive.

1.3 Goal of this work

The goal of this master thesis is to obtain fundamental insight into the propane dehydrogenation

mechanism on Pt-based catalysts using optPBE vdw-DF DFT calculations. The focus is on the

determination of the activity (propylene formation rate), selectivity towards gaseous side

products (methane, ethane and ethylene) and towards coke precursors (tendency to form

carbonaceous species) of the catalyst.

First, a reaction network on Pt(111) is constructed based on the articles of Yang et al. [12] and

Valcarcel et al. [13], including all elementary steps to describe the propane dehydrogenation

reaction, the side product formation and coke formation. This Pt(111) catalyst model is

considered as a reference model and its results can be validated with literature.

Thereafter, three other catalyst models are investigated: a Pt-Ga catalyst model representing the

novel Pt-Ga/Mg(Ga)(Al)Ox catalyst, a cokes deactivated Pt catalyst model and a step edge Pt

catalyst model, representing the undercoordinated Pt atoms where coke formation could

preferably occur. With these models it is possible to study the effect on the reaction

thermodynamics and kinetics of, respectively, Ga-alloying, catalyst deactivation by coke

formation, and structure sensitivity. The focus is on the quantification of the activity and

selectivity towards propylene.


1.4 Structure of this work

In Chapter 2, a detailed literature study is performed on Pt-based catalysts for light alkane

dehydrogenation. The main topics covered are pure, Sn-modified, Ga-modified Pt catalysts,

and their catalytic properties. Furthermore, the coke formation and deactivation of Pt-based

catalysts are discussed. A section is dedicated to the interaction between the active metal phase

and the support. Chapter 3 consists of an overview of the computational methodology used for

the DFT study.

In Chapter 4, the considered intermediates and transition states are calculated in a DFT

framework on a pure Pt catalyst model (the Pt(111) model). The results are validated with

literature and are used to determine propane dehydrogenation thermodynamics and kinetics.

Furthermore, a kinetic study using a microkinetic model is conducted to obtain insight into the

surface coverages and the rate-determining step.

In Chapter 5, a Pt-Ga catalyst model (the Pt3Ga(111) model) is used to determine the role of Ga

alloying on the thermodynamics and kinetics of propane dehydrogenation. Chapter 6 makes use

of a cokes deactivated Pt catalyst model to quantify the influence of cokes on propane

dehydrogenation characteristics. Chapter 7 discusses the last catalyst model, the

undercoordinated sites of the pure Pt catalyst. In previous three chapters, the catalytic activity

and selectivity towards propylene are determined with respect to the reference catalyst model

of Chapter 4.

Finally, in Chapter 8, the overall conclusions of this work are stated and an outlook on the future

is given.

1.5 References

1. Mackenzie, W., Global propylene long-term outlook 2H 2014. September 2014. 2. Vora, B.V., Development of Dehydrogenation Catalysts and Processes. Topics in

Catalysis, 2012. 55(19-20): p. 1297-1308. 3. Ren, T., M.K. Patel, and K. Blok, Steam cracking and methane to olefins: energy use,

CO 2 emissions and production costs. Energy, 2008. 33(5): p. 817-833. 4. Rahimi, N. and R. Karimzadeh, Catalytic cracking of hydrocarbons over modified ZSM-

5 zeolites to produce light olefins: A review. Applied Catalysis A: General, 2011. 398(1–2): p. 1-17.

5. Bruijnincx, P.C.A. and B.M. Weckhuysen, Shale Gas Revolution: An Opportunity for the Production of Biobased Chemicals? Angewandte Chemie International Edition, 2013. 52(46): p. 11980-11987.


6. Bond, G.C., Metal-Catalysed Reactions of Hydrocarbons. Fundamental and applied Catalysis. Vol. XXI. 2005: Springer. 666.

7. Bhasin, M.M., et al., Dehydrogenation and oxydehydrogenation of paraffins to olefins. Applied Catalysis a-General, 2001. 221(1-2): p. 397-419.

8. Davis, R.J., R.H. Griffith, and J.D.F. Marsh, The physical properties of chromia-alumina catalysts. Advances in Catalysis, 1957. 9: p. 155-168.

9. Haensel, V., Hydrocarbon Conversion Process with Subsequent Reforming of Selected Hydrocarbon Fractions. US Patent, 1959.

10. Bloch, H.S., UOP discloses new way to make linear alkylbenzene. Oil Gas J 79–81, 1967(US Patent 3,448,165,).

11. Yang, M.L., et al., Density functional study of the chemisorption of C-1, C-2 and C-3 intermediates in propane dissociation on Pt(111). Journal of Molecular Catalysis a-Chemical, 2010. 321(1-2): p. 42-49.

12. Yang, M.L., et al., DFT study of propane dehydrogenation on Pt catalyst: effects of step sites. Physical Chemistry Chemical Physics, 2011. 13(8): p. 3257-3267.

13. Valcarcel, A., et al., Theoretical study of dehydrogenation and isomerisation reactions of propylene on Pt(111). Journal of Catalysis, 2006. 241(1): p. 115-122.

14. Siddiqi, G., et al., Catalyst performance of novel Pt/Mg(Ga)(Al)O catalysts for alkane dehydrogenation. Journal of Catalysis, 2010. 274(2): p. 200-206.

15. Galvita, V., et al., Ethane dehydrogenation on Pt/Mg(Al)O and PtSn/Mg(Al)O catalysts. Journal of Catalysis, 2010. 271(2): p. 209-219.

16. Sun, P.P., et al., Synthesis and characterization of a new catalyst Pt/Mg(Ga)(Al)O for alkane dehydrogenation. Journal of Catalysis, 2010. 274(2): p. 192-199.

17. Saerens, S., Determination of the active phase of a Pt/Mg(Ga)AlOx catalyst of the catalytic dehydrogenation of propane, in Department of Chemical Engineering and Technical Chemistry. 2014, University of Ghent: Ghent.

18. Vu, B., et al., Electronic density enrichment of Pt catalysts by coke in the propane dehydrogenation. Korean Journal of Chemical Engineering, 2011. 28(2): p. 383-387.

19. Vu, B.K., et al., Location and structure of coke generated over Pt–Sn/Al2O3 in propane dehydrogenation. Journal of Industrial and Engineering Chemistry, 2011. 17(1): p. 71-76.

20. Peng, Z., et al., High-resolution in situ and ex situ TEM studies on graphene formation and growth on Pt nanoparticles. Journal of Catalysis, 2012. 286: p. 22-29.

21. Preto, I., Graphene formation and oxidation on Pt/Mg(Ga)AlOx dehydrogenation catalysts, in Department of Chemical Engineering and Technical Chemistry. 2015, University of Ghent: Ghent.

Chapter 2: Literature review 8

Chapter 2 Literature review

2.1 Catalysts for propane dehydrogenation reaction

2.1.1 Catalysis in general

The International Union of Pure and Applied Chemistry (IUPAC) has defined a catalyst as “a

substance that increases the rate of a reaction without modifying the overall standard Gibbs

energy change in the reaction”. [1] The chemical process of increasing the reaction rate is called

catalysis, and the catalyst is both a reactant and a product of the reaction, i.e., the catalyst is

restored after each catalytic act. Furthermore, the catalyst does not influence the

thermodynamical equilibrium composition after the cessation of the reaction. [2] Catalysts only

enhance the kinetics of the occurring reactions, not the thermodynamics and the chemical

equilibrium since it influences both the forward and reverse reaction rate. While catalysts are

not consumed during the reaction, several side reactions such as inhibition and deactivation can

influence the catalyst chemical and mechanical properties.

By means of catalysis, other reaction mechanisms with different transition states are

energetically more favorable, but often have a higher level of complexity. Nonetheless,

catalysis is widely employed in both research and industrial applications as it improves two

important characteristics of the overall reaction mechanism: activity and selectivity. The most

straightforward principle is the activity of the catalyst as the catalyst increases the reaction rate

with multiple orders of magnitude. Often the highest activity can be correlated to the interaction

between substrate and catalyst, which should neither be too strong, so the substrate remains

available for reaction, nor too weak so the catalyst enhances the reaction rate. The catalyst can

9 Chapter 2: Literature review

also improve the selectivity towards a desired product by increasing exclusively its production

rate. Activity and selectivity can play a major role to select a catalyst.

While several types of catalysis are available, the focus in this work will be on heterogeneous

catalysis. Here, the catalyst consists of a different phase, mostly solid, that differs from the

reactants phase. Its main advantage is that no chemical separation is needed to separate the gas

or liquid reaction products from the solid-state catalyst. The reactions occur on the catalyst

surface and typically consist of multiple steps: adsorption, reaction and desorption, often

accompanied with diffusion towards, outwards and on the catalyst surface and dissociation of

the adsorbate. The first chemical reaction step is adsorption where a bond is formed between

the solid catalyst, the adsorbent, and reactants, the adsorbates. Two main types of adsorption

are possible: physisorption, which is generally a weak bond formation between adsorbent and

adsorbate based on long-range van der Waals forces and chemisorption, where stronger bonds

are formed due to the overlap of orbitals between the adsorbent and adsorbate. The latter is

necessary to enhance the further catalytic reactions on the surface. The adsorbed molecules can

diffuse on the surface, react to products and eventually desorb from the surface. [3-5]

In this work, the focus lays on the dehydrogenation of paraffins and more specific that of

propane. Platinum, an archetype catalyst for this reaction, has a high activity but suffers from

low olefin selectivity and coke formation, which leads to catalyst deactivation. Paraffin

dehydrogenation is an endothermic reaction that is limited by chemical equilibrium and as

illustrated in Figure 2-1, the equilibrium constant is rather low for smaller carbon chains.

Consequently, to obtain higher conversion either higher temperatures or lower pressures are

necessary, according to Le Chatelier’s principle. [6, 7]

Promoting Pt catalysts with metals such as As, Bi, Pb, Ga, Ge and Sn will weaken the platinum-

olefin interaction without influencing the platinum-paraffin interaction. Hence, the paraffin

dehydrogenation rate will remain constant while the dehydrogenation rate of the desired olefins

decreases and the olefin desorption barrier lowers. This can result in an overall increase of

olefin production rate. The above-mentioned modifiers will also restrict the formation of coke

precursors, this way decreasing the coke deposit on the catalyst active sites. [8, 9]


Figure 2-1: Equilibrium constants for n-paraffin dehydrogenation at 500 °C [6]

In the following sections, platinum and its modified bimetallic catalysts employed in

dehydrogenation reactions are discussed. Apart from the pure platinum catalyst, the Sn-

promoted and Ga-promoted Pt catalysts are discussed in particular. Pt-Sn catalysts have already

been researched extensively and the effects of Sn-promoting are well known. This could be

useful to link to Ga-promoting, since recent studies have shown that Ga-promoted catalysts

have improved catalytic properties. [10, 11] The catalyst characteristics either found

experimentally and/or computationally are evaluated.

2.1.2 Pure platinum catalysts

Platinum is a noble metal and its bulk phase has a face-centered cubic (fcc) structure with a

coordination number of twelve, a packing factor of 0.74 and a lattice distance of 3.92 Å. [12]

The most stable surface cut of the fcc unit cell is the (111) plane. [13] Each atom on the (111)

surface has a coordination number of nine, so there are three broken bonds per surface atom.

Three broken bonds is the minimal amount possible, making the (111) plane the most abundant

surface plane in Pt particles. However, the surface of an actual solid-state catalyst does not

solely consist of a single crystal plane. Multiple irregularities can be encountered such as edges,

steps and kinks. These irregularities will have a higher activity due to the geometrical and

electronic effect of the higher undercoordination.

Pure platinum catalyst has frequently been utilized to determine experimentally the activity,

selectivity towards products and deactivation during paraffin dehydrogenation. [14-19] The

most employed experimental methods to investigate the surface chemistry on Pt catalysts are

temperature programmed desorption (TPD), infrared adsorption spectroscopy (IRAS) and


transmission electron microscopy (TEM). Since many data are available for pure Pt catalysts,

the Pt catalysts can be used as a reference case in this study.

The platinum metal catalyst is very active for paraffin dehydrogenation; however, it lacks a

high selectivity towards olefins, as many side reactions occur on the surface such as coke

formation, leading to unwanted deactivation. It is known that the formation of coke on the

catalyst surface results from both consecutive dehydrogenation and hydrogenolysis of adsorbed

carbon species. Zaera et al. conducted TPD experiments with propylene on Pt(111) to unravel

the thermal chemistry and hence, propose a set of consecutive reactions of propylene that lead

to coke formation. When adsorbed alone, propylene thermal activation leads mainly to

molecular desorption and dehydrogenation reactions. The latter reaction leads to the formation

of propylidyne (Pt3≡C-CH2-CH3), which is the trigger for surface coke formation as consecutive

dehydrogenation steps form surface carbon. In addition, a small amount of propane is formed

through self-hydrogenation with hydrogen, formed in the dehydrogenations steps. Zaera et al.

propose that coke formation reactions demand larger ensembles of active metal atoms to

accommodate the formed hydrogens in consecutive dehydrogenation reactions and to be able

to cleave the bond in the adsorbed species during hydrogenolysis. [17, 18]

Coke formation can be reduced by alloying the platinum catalysts with other elements as it

distorts the larger ensembles of active metal atoms and hereby increases the olefin selectivity

by reducing the coke formation. Furthermore, promoting metals can increase the catalyst

activity. Two major steps in the dehydrogenation reaction are the dissociative adsorption of

saturated hydrocarbons and the desorption of the olefin product. The bimetallic catalyst can

attenuate the paraffin adsorption and the olefin desorption, increasing the rate of olefin

formation. [10, 15]

Catalyst models based on pure platinum catalysts are often employed in a computational

approach or in a combined experiment-and-theory study to correlate the above-mentioned

experimental properties. However, as the modified platinum catalyst excels in multiple

properties in comparison with pure platinum catalyst, the latter are used to obtain fundamental

insights or as reference with respect to the modified Pt catalysts. [14, 20-25]

2.1.3 Bimetallic Pt catalysts

The drawbacks of pure platinum catalysts are the poor olefin selectivity and the high

deactivation due to the coke formation on the surface. The main role of platinum modifiers is

to circumvent these drawbacks in the light of paraffin dehydrogenation. [8] Bimetallic surfaces


exhibit catalytic properties different from their pure counterparts. The bimetallics local

electronic and geometric structure influences their catalytic properties and this way enhances

the desired characteristics of the dehydrogenation reaction.

It is rather difficult to pinpoint the surface structure and properties after modification with

another metal, as spectroscopic techniques such as X-ray diffraction (XRD), X-ray

photoelectron spectroscopy (XPS) and TEM give limited results in the sub-nanometer range.

Three major effects can possibly occur when altering the catalyst metal composition. First, the

electronic structure of the surface is changed due to the bond formation with the modifier metal

as it can act as a ligand and induce an electronic effect. The alloying metal can directly influence

the geometric structure through different metal-metal bonding distance, as so inducing

additional strain in the catalyst. Moreover, it is possible that the surface atoms rearrange

themselves, but this depends heavily on the composition of the catalyst and reaction conditions.

Therefore, the configuration of the bimetallic alloy has to be determined: bulk alloys, where the

heteroatoms are present in the bulk and surface layers and surface alloys, where the heteroatoms

are only present in the surface layers, as illustrated in Figure 2-2.

Figure 2-2: (left) four layered model of a Pt3M/Pt(111) surface alloy. (right) four layered model of a Pt3M bulk alloy. Pt atoms are indicated in blue while the hetero atoms M are red (left) and green (right). [26]

This is not a static given, but rather dynamical depending on the reaction conditions. The

bimetallic alloy always strives towards the lowest Gibbs free energy and in this case, this

corresponds to the lowest surface energy of the exposed surface. Two rearrangement

phenomena can occur: segregation and antisegregation. In segregation, the heteroatoms

segregate to the surface while Pt atoms migrate to the bulk, while in antisegregation the reverse

will happen. The rearrangement of the surface atoms can alter the surface reactivity in several

ways as illustrated in Figure 2-3. [27, 28]


Figure 2-3: Geometric factors determining the adsorption and dissociation of molecules on bimetallic surfaces. [27]

The site blocking refers to the blocking of adsorption sites by the heteroatoms as these atoms

can only weakly interact with adsorbate molecules. The ensemble effect occurs when an atom

is removed or another heteroatom is incorporated in the ensemble or adsorption site and this

hereby alters the adsorption properties. Furthermore, the template effect induces a shape change

of the adsorption site. At last, there is the coordination effect, which inhibits reactions that need

a minimum number of adjacent adsorption sites. These rearrangement effects induce an

additional electronic effect, so it is often impossible to discriminate between the different

contributions of geometric and electronic effects.

Platinum can be alloyed with both transition metals e.g., Ag, Co, Fe and non-transition metals

e.g. Ga, In, Sn. The main effect that is observed for alloys with non-transition metals is site

blockage, while this does not occur with metals alloyed with transition metals. [27] In the

following sections, bimetallic Pt catalysts are outlined. The focus of this work is on Ga-

promoted Pt catalysts, but since Sn-promoted Pt catalyst are already extensively researched,

these could offer a useful guideline for the study of the relative new Ga-promoted Pt catalysts.

2.1.4 Bimetallic Sn-Pt catalysts

Sn-promoted Pt catalysts exhibit excellent characteristics for light paraffin dehydrogenation

since both high reaction rates and high selectivity towards olefins are observed. Surface science


studies of Pt-Sn chemistry have provided insight in the catalyst properties. However, the lack

of detailed knowledge on the surface composition and structure has made it difficult to correlate

the surface structure with the superior catalytic properties. [29]

The role of Sn is to improve activity, selectivity and stability of catalyst particles. It is expected

that the addition of Sn neutralizes acidity of supports, interacts electronically with Pt and

reduces the ensemble effect that favors coke formation. Several models are proposed for Pt-Sn

interaction, as illustrated in Figure 2-4. When increasing the Pt+Sn load, Sn/Pt ratio or reduction

temperature, the system shifts towards the formation of Pt-Sn alloys compared to formation of

Sn clusters. In this work, it is assumed that Pt-Sn alloys have formed on the catalyst surface.

[30]

Figure 2-4: Models of Pt–Sn interaction. Pt-Sn alloys are formed under higher Pt+Sn loading, higher Sn/PT ratio and increased reduction temperature [30]

High-resolution transmission electron microscopy (HRTEM), alkali ion scattering

spectroscopy (ALISS) and X-ray photoelectron diffraction (XPD) are employed to shed a light

on the formation of the Pt-Sn phase. The exact mechanisms contributing to the formation of Pt-

Sn alloys are still unknown. [15, 29]

In the case of promoting platinum metal particles with Sn, several ordered surface alloys can

be distinguished as illustrated in Figure 2-5. At low Sn content, Pt3Sn or PtSn are the most

probable ordered alloys at dehydrogenation reaction conditions (~800 K), but other alloys such

as Pt2Sn3 and PtSn2 can be formed under certain reaction conditions as well. These compositions

of Pt-Sn alloys can be produced accurately, using controlled deposition methods. These alloys

are further utilized to determine the most stable and the most catalytic active alloy under certain

reaction conditions.

It is challenging to determine the composition of the active phase of the Pt-Sn alloy since the

reaction conditions can alter the surface composition. The role of Sn on the catalytic properties

of Pt-Sn catalyst is correlated to the oxidation state of tin, however, this state is difficult to

determine without sophisticated analytical techniques. Possibilities are extended X-ray

Pt+Sn loading ↑ Sn/Pt ↑ Reduction T ↑

Pt+Sn loading ↑↑ Sn/Pt ↑↑ Reduction T ↑↑


adsorption fine structure (EXAFS) and XPS, but even these techniques have limits

characterizing the catalyst surface certainly if operated under reaction (in situ).

Figure 2-5: Binary alloy phase diagram of Pt-Sn [31]

2.1.4.1 Light alkane dehydrogenation

Galvita et al. investigated the effect of Sn-modified Pt catalysts during ethane dehydrogenation.

In most studies an alumina support (Al2O3) is used, however Galvita et al. made use of a

calcined hydrotalcite (Mg(Al)Ox) support during the experiments. The introduction of Sn on

the dispersed Pt-particles by impregnation leads to geometrical and electronic effects,

enhancing the catalyst activity and selectivity. The addition of Sn reduces the size of Pt

ensembles on the surface (geometrical effect), hereby reducing the hydrogenolysis reaction rate

as it requires sufficient large Pt clusters to cleave the carbon-carbon bond in the adsorbed

species. This leads also to a higher selectivity towards ethylene. The electronic effects of Sn-

promoting enhance the dissociative adsorption of ethane on the catalyst surface, improving the

catalyst activity and selectivity towards ethylene. [15]

Analogous conclusions were made for propane dehydrogenation on Pt-Sn catalysts, supported

on SBA-15, alumina [14] and ZSM-5 [10]. The geometrical effect of Sn inclusion in the Pt

particles decreases the size of the active Pt particles, hence reducing coke formation and other

side reactions. Furthermore, sticking coefficient measurements, low energy electron diffraction

(LEED) and TPD have proven that these Pt-Sn surface alloys induce a weaker chemisorption

bond between propylene and Pt (electronic effect). [32]


As limited theoretical work is performed to determine a detailed reaction mechanism of propane

dehydrogenation on supported Pt-Sn catalysts, the fundamental understanding of the role of Sn

is limited. Yang et al. conducted ab initio density functional theory (DFT) calculations of

propane dehydrogenation on pure Pt and Pt-Sn catalysts. [24, 33] In the latter, they used five

models with different Sn to Pt ratios to represent the Pt-Sn catalysts: Pt3Sn/Pt(111), Pt3Sn(111),

Pt2Sn/Pt(111), Pt2Sn(111) and PtSn2(111). The first and third are surface alloys where only the

top layer is alloyed with Sn. The Pt2Sn alloys are theoretical alloys and cannot be found in the

Pt-Sn phase diagram (see Figure 2-5). [31] The descriptors for the catalyst activity are defined

by the activation energies of the dehydrogenation steps from propane to propylene. This

sequence consists out of two major elementary steps: dissociative propane adsorption to propyl

and dehydrogenation of propyl to propylene. The selectivity descriptor is determined as the

difference in activation energy between dehydrogenation and desorption of propylene.

Alloying Pt with Sn will significantly lower the reaction rate of propane dehydrogenation and

the variation in the activation energies depends strongly on the change in the binding strength

of H in the geometry of the transition state. Analyzing the competition between

dehydrogenation and C-C scission leads to the conclusion that the cracking is kinetically

hindered because of the much higher energy barriers. [33] The deposition of Sn makes it

difficult to achieve a large ensemble, which is essential to activate cracking reactions because

sufficient large ensembles are required to accommodate detached fragments. The introduction

of Sn lowers the desorption barrier for propylene to the gas phase and simultaneously increases

the energy barrier for propylene dehydrogenation. As the Sn content increases, the selectivity

toward propylene desorption is significantly improved. Considering the trade-off between the

catalytic activity and the selectivity, the Pt3Sn bulk alloy is proposed as the best candidate of

those five models for propane dehydrogenation. [21, 33]

2.1.5 Bimetallic Ga-Pt catalysts

Recent studies have shown that alloying Pt catalysts with Ga leads to promising catalyst

properties such as reduced coke formation and deactivation of the catalyst during light alkane

dehydrogenation. [34] Furthermore, a novel synthesis method is proposed by Sun et al. for

incorporating Ga in the Pt catalyst particles, which facilitates the use of Ga as modifier

element. [35]

Sun et al. describe a new approach for preparing Ga-promoted Pt particles in the context of light

alkane dehydrogenation. For the novel Pt-Ga/Mg(Ga)(Al)Ox catalyst, the modifying element,


Ga, was introduced by transference from the support, a calcined Mg(Ga)(Al)Ox hydrotalcite.

First, Pt nanoparticles were dispersed onto the calcined Mg(Ga)(Al)Ox starting from an

organometallic precursor and secondly the catalyst precursor is reduced. The formation of Pt-

Ga alloy particles is dependent on reduction temperature. Reduction at 723 K or lower produces

mainly metallic Pt particles, while at reduction temperatures of 773–873 K, PtxGay alloys were

observed by EXAFS and STEM-EDX results. It is proposed that at high reduction temperatures,

hydrogen formed on the surface of the metal particles spills over onto the support where they

reduce Ga3+ cations to Ga, which then interact with the supported Pt to form PtxGay alloys with

different compositions, as illustrated in Figure 2-6. [35, 36]

Unfortunately, the exact composition of these alloys remains elusive as the average particle size

is in the order of a few nanometers and even sophisticated spectrometry methods such as TEM

and STEM fall short to determine the exact stoichiometry. To discover the unknown

composition of the Pt-Ga alloys, other methods can be employed and these will be discussed

further on in this chapter.

Figure 2-6: Proposed mechanism for migration of Pt alloy formation during synthesis and reaction of Pt/Mg(Ga)(Al)Ox catalysts. [35]

2.1.5.1 Light alkane dehydrogenation

In Figure 2-7, the results of the catalyst activity (propylene formation rate) and the propylene

selectivity as function of time on stream are illustrated. The activity of the Pt-Ga/Mg(Ga)(Al)Ox

catalyst is initially more than two times higher than the pure Pt/Mg(Al)Ox catalyst. While both

suffer from activity loss as function of time on stream, the former catalyst is more stable as it

loses only 30% of its activity after two hours on stream with respect to 40% activity loss on

Pt/Mg(Al)Ox (see Figure 2-7a). Furthermore, the experiments show that the propylene

selectivity is enhanced for the Pt-Ga/Mg(Ga)(Al)Ox catalyst as it remains 99% the entire time


on stream. In contrast, the Pt/Mg(Al)Ox catalyst has initially a low propylene selectivity of 78%,

but increases to 87% after one hour on stream (see Figure 2-7b).

Figure 2-7: (a) Comparison of activity of Pt/Mg(Ga)(Al)O (Ga/Pt = 2.86) and Pt/Mg(Al)O and (b) selectivity for propane dehydrogenation. Reaction temperature of 873 K, 20 vol.% C3H8 in feed, H2/C3H8 = 1.25, with balance He for total flowrate of 60 ml/min. [11]

Siddiqi et al. determined the optimal ratio of Ga to Pt in a Pt-Ga/Mg(Ga)(Al)Ox catalyst by

performing experiments with Pt-Ga catalysts with various Ga/Pt ratios. As illustrated in

Figure 2-8, a trade-off has to be made between catalyst activity and propylene selectivity. The

rate of propylene formation has a distinct maximum of 64 µmol/s/gcat at a Ga/Pt ratio of two

and the rate sharply decreases for different Ga/Pt ratios (see Figure 2-8a). The propylene

selectivity initially increases sharply until a maximum of 99% at Ga/Pt of 5.4 and further

decreases monotonically (see Figure 2-8b). The approximated optimal Ga/Pt ratio is 5.4, when

the propylene selectivity is maximal and the rate of formation is 50 µmol/s/gcat.

Figure 2-8: Effect of Ga/Pt ratio on Pt/Mg(Ga)(Al)O catalysts for C3H8

dehydrogenation (a) activity and (b) selectivity, feed composition of H2/C3H8 = 1.25 and all data points are after 120 min time on stream. Reaction temperature of 873 K, 20 vol.% C3H8 in feed, H2/C3H8 = 1.25, with balance He for total flowrate of 60 ml/min. [11]


2.1.5.2 Determination of active phase of Pt-Ga catalysts

Almost no theoretical work is performed on propane dehydrogenation on supported Pt-Ga

catalysts, and consequently the fundamental understanding of the role of Ga as a promotor is

slight. In previous work, the active phase of the Pt-Ga/Mg(Ga)(Al)Ox catalyst is investigated

by conducting ab initio DFT calculations on various Pt-Ga alloys to determine an appropriate

catalyst model that describes the surface composition of the novel catalyst. [37]

As illustrated in Figure 2-9, the following Pt-Ga compositions are taken into account in

ascending order: Pt3Ga, Pt2Ga, Pt5Ga3, PtGa and Pt2Ga3. These are all the alloys up to 0.6 mole

fraction of Ga and stable above reduction temperatures of 773 K. On the Pt-Ga/Mg(Ga)(Al)Ox

catalyst, CO adsorption experiments are performed in unison with frequency calculations of

adsorbed CO on each catalyst model, accounting for different coverages and segregation. The

experimental results indicate a small shift in CO frequencies (~5 cm-1) for increasing Ga

content. Compared to performed DFT calculations on the candidate structures, this small shift

indicates the formation of the Pt-Ga alloy with the lowest Ga content, which is Pt3Ga in this

study. Other catalyst models with low Ga-content, which can be correlated with the small shift,

are e.g. a surface alloy Pt3Ga/Pt. [37]

Figure 2-9: The phase diagram of the Pt–Ga binary system as reassessed with first-principles calculations by Wang et al. [38]

The role of gallium on increased activity and selectivity during propane dehydrogenation

remains elusive. The focus on this work is to provide a fundamental insight in the reaction


mechanism on this novel Pt-Ga/Mg(Ga)(Al)Ox catalyst. Based on previous studies, the Pt3Ga

alloy will be chosen as a model for the Pt-Ga/Mg(Ga)(Al)Ox catalyst. [37]

2.2 Coke formation during propane dehydrogenation

During propane dehydrogenation, side reactions such as cracking to lighter hydrocarbons,

skeletal isomerization and aromatization lead to formation of coke deposit on the surface of a

platinum catalyst, causing rapid deactivation and lower yields. [39] Characterization of the coke

deposits could lead to further understanding of the coke formation mechanism and enabling the

minimization of coke formation. This section focuses on the characterization of the coke and

its formation rate, which is essential for understanding the mechanism of coke formation as

well as the mechanism of catalyst deactivation.

2.2.1 Coke formation on pure Pt catalysts

Coke formation on supported Pt catalysts depends on the selected reaction temperature. At

lower temperature regimes, coke formation can be contributed to condensation and

rearrangement steps. However, at high temperature, also hydrogen transfer and

dehydrogenation steps are observed. The higher temperature regime is a more realistic reaction

condition in light alkane dehydrogenation reactions since higher conversions are obtained at

these conditions. [40]

The coke mobility has been studied with temperature programmed oxidation (TPO) techniques

and results show that coke can be formed on the Pt metal, Pt-support boundary or on the support.

These studies have shown that reaction conditions influence the rate of coke formation and the

nature of the coke deposited on different sites are essential for the understanding of the

mechanism of coke formation as well as the mechanism of catalyst deactivation. [41] TPO

studies show that different type of cokes can be found on the supported catalyst. The coke

formed on those different sites could have other effects on the catalyst deactivation. Larsson et

al. proposed a model where only a small part of the formed coke was responsible for catalyst

deactivation, while the major part of the coke was formed regardless of the gas composition but

did not contribute to the deactivation. [41] It is also observed that the coke formation rate

consists out of two time regimes; initially the coking rate is high, but eventually it decreases to

a lower constant rate. These regimes can be contributed to the formation of different types of

coke deposits. Essentially, there is constant coke formation during both regimes, which can be


contributed to coke deposits on the support. However, the initial high coke formation rate is

due to the formation of coke precursors on the active phase. The coke precursors migrate to the

boundary between the platinum and the support and when most of these sites are occupied, the

coke formation rate is greatly reduced. These coke deposits are responsible for the deactivation

of the active sites on the metal as it hinders the transport of coke precursors to the support,

which thus increases the coke accumulation on the platinum. [42]

Vu et al. investigated the nature of coke generated in propane dehydrogenation. Two supported

Pt catalysts were synthesized: each with 3% Pt and the first on mesoporous alumina and the

other on SBA-15 (mesoporous silica). In this study, Vu et al. characterized the coke structure

of both spent catalysts with both XRD and the sophisticated magic angle spinning nuclear

magnetic resonance (13C-CP/MAS NMR) technique. The XRD results show that the nature of

the coke exhibits a pregraphite-like carbon structure. Further characterization with NMR leads

to the determination of the pregraphite-like carbon structure as a distinct aromatic structure

while no detectable presence of aliphatic coke was found, independently of the different acidity

of the support. Further investigation was conducted with the Fourier transform infrared

spectroscopy (FTIR) and the XPS techniques, which confirmed the polyaromatic structure of

the deposited coke. It was also observed that the electron-rich state of the polyaromatic ring in

the coke donates electrons to the Pt active metal, inducing a higher electron density of Pt metals

as illustrated in Figure 2-10. This causes a shift in binding energy towards lower values. [43]

Figure 2-10: A schematic diagram for coke-Pt interaction and electron transfer [43]

The most favorable adsorption site for coke formation and possible coke spillover are not yet

discussed. Peng et al. performed in situ and ex situ HRTEM studies on pure platinum catalysts,

supported on MgO. Various average metal particle diameters (1.4 nm, 3.5 nm and 6.1 nm) were


investigated. The employed carbon source for the coke deposits was a mixture of ethylene and

hydrogen under light alkane dehydrogenation conditions. The TEM observations confirmed

that the formed coke species are a type of graphene and their formation initiates at highly

preferred low-coordination sites such as step sites on the surface of the metal particles. The

growth of the graphene layers depends on the size and shape of the catalyst particles. In these

studies, particles with particle diameter larger than 6 nm are covered with multiple layers of

piled graphene, the number of layers depending on the hydrocarbon exposure time. In the region

with particle diameters of 2-6 nm, the formation of graphene nanotubes are observed and at

even smaller particle diameters (< 2 nm), a graphene sheet is formed, which migrates towards

the support. The size dependency of graphene growth is attributed to the accommodation of

strain energy generated in the graphene layers and the minimization of overall free energy in

the growth process. The graphene growth models are illustrated in Figure 2-11. However, this

classification based on the average particle diameter is ambiguous, as other factors clearly

influence the graphene growth such as the shape of the particle. [44]

Figure 2-11: Schematic illustration of graphene layer growth on Pt particles of increasing size: (left) envelopment of Pt particles by graphene for particles greater than ∼6 nm in diameter; (middle) formation of graphene nanotubes on Pt particles of 2–6 nm; (right) formation of graphene sheets and their migration to the support for Pt particles less than 2 nm in diameter. [44]

Additionally, Larsson et al. reported the effect of hydrogen on suppressing coke formation.

Hydrogen is active in preventing coke formation but does not remove any significant amount

of coke already formed. On the one hand, the coke formed in dedicated coking experiments,

which means higher temperature and longer time on stream, probably has been more deeply

dehydrogenated and strongly bonded to the catalyst. During the reaction, on the other hand, the

formation of coke precursors is suppressed by hydrogen. This explains why hydrogen helps

maintain the catalytic activity. [41]


2.2.2 Coke formation on bimetallic Sn-Pt catalysts

Li et al. and Vu et al. both determined the influence of alloying the Pt catalyst with Sn on the

coke characteristics and its formation rate. Both employ alumina supported Pt-Sn catalysts in

their research but Vu et al. focused more on the location and structure of coke generated during

the propane dehydrogenation reaction, while Li et al. determine different coke formation

mechanisms. [41, 42, 45]

Vu et al. prepared different Pt-Sn catalysts with a constant 1% Pt content, while varying the Sn

content between 0 and 1.67%. Based on a thermogravimetric analysis (TGA), the best coke

tolerance is found with the 1% Pt-1.67% Sn catalyst, which has the slowest deactivation rate

and lowest coke content. Furthermore, Vu et al. determined, based on XRD and XPS analyses,

that the coke on the Pt-Sn/Al2O3 catalyst has the same pregraphite-like carbon structure as on

the Pt/Al2O3 catalyst. Consequently, the addition of Sn has no influence on coke structure. The

TPO technique allows determining the coke location on the supported bimetallic catalyst. Two

types of coke are identified on the spent catalysts and can be assigned to coke on the metal

(lower TPO temperature) and the support (higher TPO temperature). For increasing Sn contents,

the TPO profiles shift to higher temperatures, indicating that Sn alters the electronic properties

of the Pt surface and weakens the propylene-platinum bond. Therefore, the Sn addition

accelerates the coke spillover from the catalytic phase to the support, decreasing the coke

fraction on the metal and hence ensuring a longer stability and activity of the catalyst. [45]

Li et al. confirmed the results of the formation of two types of coke on the alumina supported

Pt-Sn catalyst and assigned them to coke on the metal and the support. Furthermore, Li et al.

varied the reaction conditions, especially the gas composition (propane and/or hydrogen), to

determine its influence on the coke formation. Based on results with various H/C ratios, Li et

al. conclude that the coke formed on the metal has aliphatic hydrocarbon characteristics,

containing more hydrogen than that formed on the support, which has an aromatic

characteristic. The proposed model of coke formation is divided in two parts, one for each type

of coke formed.

The rate of coke formation on the metal is weakly dependent on the propylene and hydrogen

pressures but increases simultaneously with the propane pressure, while the rate of coke

formation on the support increases concurrently with the propane and propylene pressures and

decreases with the hydrogen pressure. In the following paragraph, the focus will be on the coke

formation on the active metal phase only.


Li et al. propose the following reaction mechanism for coke formation on active metal phase of

supported Pt-Sn catalysts based on a kinetic analysis:

Reaction mechanism 1: Coke formation on active metal phase of supported Pt-Sn catalyst [42]

𝐶𝐶3𝐻𝐻8(𝑔𝑔) + 2 ∗ → 𝐶𝐶3𝐻𝐻7 ∗ +𝐻𝐻 ∗ (1)

𝐶𝐶3𝐻𝐻7 ∗ + ∗ → 𝐶𝐶3𝐻𝐻6 ∗ +𝐻𝐻 ∗ (2)

𝐶𝐶3𝐻𝐻6 ∗ → 𝐶𝐶3𝐻𝐻6(𝑔𝑔) + ∗ (3)

2𝐶𝐶3𝐻𝐻6 ∗ → 𝐶𝐶6𝐻𝐻12 ∗ + ∗ (4)

2𝐻𝐻 ∗ ⇌ 𝐻𝐻2(𝑔𝑔) + 2 ∗ (5)

The reaction between the two strong adsorbed C3H6* molecules, which are formed by

dehydrogenation of propane, is identified as the kinetic relevant step (4) for the coke formation

on the Pt surfaces. It is noted here that in this model the hydrogen pressure has little effect on

the rate of coking on the metal, but it changes significantly the hydrogen content in the coke.

At higher hydrogen partial pressure, the coke is less dense and compact due to a higher H/C

ratio in the coke itself. [46] However, this contradicts the observations of Larsson et al. which

report a stronger dependence on hydrogen pressure, namely the reduction of coke precursor

formation, together with less dense coke due the presence of hydrogen in the feed.

A large portion of the precursors migrates to the acid sites of the alumina and is involved in the

coke formation on the support. In addition, the propylene in the gas phase can also adsorb on

the alumina support and form coke through complex reaction mechanisms. [42]

2.2.3 Coke formation on bimetallic Ga-Pt catalysts

The characterization of coke is less researched for supported Pt catalysts alloyed with Ga.

Siddiqi et al. studied the catalyst performance of Pt-Ga catalysts, supported on calcined

hydrotalcite for both ethane and propane dehydrogenation. Here the coke formation and catalyst

deactivation were also reported. The presence of Ga compared to Sn reduces the deactivation

and the amount of coke deposited during both ethane and propane dehydrogenation. The ratio

of deposited carbon to surface Pt atoms are much greater than unity in all cases, but suddenly

drops from 47 to 15 at Ga/Pt ratio of 5.4, as illustrated in Figure 2-12. The large carbon to

surface Pt atoms ratio suggests that most of the accumulated coke is present on the support and

that only a small fraction remains on the surface of the metal particles, in agreement with the

model for Pt-Sn catalysts. It is hypothesized that Ga affects the surface, similar to that of Sn,

by reducing the desorption barrier of the desired alkene product. [11, 42]


Figure 2-12: Carbon formation after 120 min of propane dehydrogenation at different Ga/Pt ratios, Reaction temperature of 873 K, 20 vol.% C3H8 in feed, H2/C3H8 = 1.25, with balance He for total flowrate of 60 ml/min. [11]

Additionally, Siddiqi et al. determined the effect of the extra methyl moiety of propane with

respect to ethane on the coke formation and the deactivation. The overall amount of coke

generated during ethane dehydrogenation is greater than during propane dehydrogenation,

while the deactivation was larger in the latter. Therefore, this confirms additionally that the total

coke amount cannot be directly linked to deactivation. However, it is expected that the mobility

of carbonaceous species is lower for such species originating from propane than those

originating from ethane, resulting in larger coke buildup on the active phase. [11]

Peng et al. suggested that the suppression of coke formation on bimetallic Pt catalysts, modified

with Ga, Sn, In, etc. may be assigned to the presence of these elements at the undercoordinated

sites. [44]

2.3 Interaction between active metal phase and support

Platinum is a highly active catalytic element and is not required in large quantities to catalyze

the reaction when it is dispersed on a high surface-area support. The high dispersion is also

necessary to achieve high selectivity towards dehydrogenation relative to undesirable side

reactions. [8] In the following section, the active metal phase properties of the pure and

modified platinum catalyst are discussed with focus on their dispersion and particle structure.

Furthermore, several high surface-area supports with high Pt dispersion inducement are

evaluated. Finally, the link is made between the two sections and their interaction will be

elaborated on.


2.3.1 Active metal phase

The degree of dispersion and thus the particle size depends on various factors such as the

amount of platinum precursors, the type of the support and the utilized impregnation method.

As platinum particles exhibit a high catalytic activity, small particles are favored since they

have a higher area-to-volume ratio with respect to larger particles and less bulk platinum atoms

are present. There are different particle structures and shapes for transition metal particles in

relation to their length scale regimes, as illustrated in Figure 2-13. Each structure and shape

contributes differently to the chemical reactivity of the metal particle.

Figure 2-13: Schematic overview of structure-stability regimes of transition metal particles [47]

It is not possible to distinguish between interior and exterior surface atoms in metal particles

with a size less than 1 nm. Such small clusters often have a low energy barrier for

reconstructing, especially when in contact with adsorbing molecules and atoms. They react

typically as a sole molecule. Their reactivity can be directly related to their orbital structure,

which varies strongly with number of atoms. These small metal clusters can be distributed on

a microporous support and are highly reactive. Between one and three nm, the number of

surface atoms increases over the number of interior atoms. The shape of these particles is often

that of an ideal Platonic or Archimedian structure, composed of similar regular polyhedral, as

illustrated in Figure 2-14. [47] At the upper boundary of this particle size regime, the ratio of

surface-to-bulk atoms decreases beneath unity. The surface of larger particles is determined by

the termination of the most stable bulk structure. However, in the size regime 3-10 nm, the

interior atoms in the bulk may not yet have a structure similar to the corresponding most stable

bulk structure. Apart from the different length regimes, the formation of step sites can be


relevant to explain the structure dependence of the catalytic activity. These are observed in the

size domain of 2-20 nm. [47]

It is expected that alloying with a transition metal such as Sn and Ga influences the average

particle size of Pt catalysts. Galvita et al synthesized and characterized a Sn-promoted Pt

supported on calcined hydrotalcite (Mg(Al)Ox) and compared it with a Pt/Mg(Al)Ox catalyst as

reference. The Pt/Mg(Al)Ox catalyst has a calcined support with a BET (Brunauer–Emmett–

Teller) surface of 200 m²/g and H2-chemisorption experiment resulted in a dispersion of 84%,

which means an average Pt particle diameter of 1.35 nm. This is confirmed by TEM analysis

results which lead to a particle diameter of 1.0-1.5 nm. [15, 48] The alloying of Pt with Sn did

not lead to a loss in the surface area of the calcined support, but since the majority of the

observed particles were so small (~1.5 nm) and the Sn-to-Pt ratio was 0.23, diffraction patterns

were inconclusive to determine the effect of Sn on the dispersion. However, bimetallic Pt-Sn

particles were observed based on HRTEM and fast Fourier transformation (FFT) analysis,

indicating that for small Sn-to-Pt ratios, Sn has no distinct effect on the dispersion. [15]

Other researchers such as Zhu et al. [49] and Zhang et al. [10] prepared Pt-Sn catalysts with

various Sn-to-Pt ratios and determined a more profound effect of Sn on the average particle

size. Zhu et al. developed a new one pot, surfactant-free, synthetic route based on the surface

organometallic chemistry (SOMC) concept for the synthesis of Sn surface-enriched Pt-Sn

nanoparticles, with Sn-to-Pt ratios between 0 and 2. XRD and TEM analyses were performed

to determine the average particle size and both conclude that tin alloying leads to smaller

particle sizes e.g. for a Sn-to-Pt ratio of 1/12, the average particle size is 3.4 nm, while for a

ratio of 1, the particle size is decreased to 2.0 nm. [49] Zhang et al. prepared Pt-Sn/ZSM-5

catalysts with higher Sn-to-Pt ratios and confirmed the trend observed by Zhu et al. However,

when the concentration of Sn is excessive, Sn0 species are formed, of which the formation is

proposed as the reason for the reduced metal dispersion.

Sun et al. prepared and characterized Pt-Ga catalysts, supported on calcined hydrotalcite with

various Ga-to-Pt ratios. The dispersion of the Pt particles on calcined hydrotalcite-like supports

is determined by H2-chemisorption. Pure Pt supported on Mg(Al)Ox (support without Ga) has

a dispersion of 84%, corresponding to an average particle size of 1.35 nm. The dispersion of Pt

on Mg(Ga)(Al)Ox (support with Ga) tends to decrease with increasing Ga/Pt ratio from about

78% to 52%, corresponding to an increase in average Pt particle size from 1.5 to 2.2 nm.

However, the observed decrease in dispersion with increasing Ga-to-Pt ratio is not contributed


to increase of Ga atoms, but is likely due to the decrease in the concentration of Al cations at

the surface of the support, which have been observed to facilitate the dispersion of Pt. [35]

As was earlier illustrated in Figure 2-13, modified Pt catalysts can be classified based on their

average particle size in either ‘Platonic and Archimedian structures’ (1-3 nm, see also

Figure 2-14) or ‘surface structures, that can deviate from their most stable bulk configuration’

(3-10 nm).

Figure 2-14: (a) Archimedian cuboctahedron, (b) Archimedian decahedron and (c) Platonic icosahedron. [47]

Kumar et al. performed experiments on pure Pt catalysts, supported on SBA-15, to determine

the effect of the particle size on propane dehydrogenation characteristics. In this study, two

catalysts are synthesized, using incipient wetness impregnation (IMP) and deposition-

precipitation (DP) techniques. This results respectively in a catalyst with an average particle

size of 3 nm (small) and an average particle size of 21 nm (large). The results from tapered

element oscillating microbalance (TEOM) experiments show that small Pt particles are very

active for C–C bond activation, namely cracking reactions, while relatively large Pt particles

are more selective for C–H bond activation, namely dehydrogenation reactions. Consequently,

selectivity to propylene depends on the activity ratio between C–H and C–C bond activation,

confirming that propane dehydrogenation is structure sensitive. [50, 51]

Furthermore, Yang et al. performed combined experiments and DFT calculations on larger

particles with different surface geometries, as the surface of large particles cannot be described

as an extension of the most stable bulk geometry. Two Pt catalysts with different particle shapes

and close particle sizes, namely, octahedron (12.0 nm) and cube (11.5 nm), have been used to

test the catalytic performances. The octahedral particles are dominated by Pt (111) facets and

the cubic particles are surrounded by Pt (100) facets. As indicated by the measured turnover


frequency (TOF) for propylene formation and selectivity toward propylene, the cubic particles

exhibit a higher catalytic activity for propane dehydrogenation but a lower selectivity for

propylene than the octahedral ones. In order to explain the experimental observations, the

differences in activation energies for the dehydrogenation process from propane to propylene

between Pt (111) and Pt (100) are compared. Based on DFT calculations, the activation energies

for dehydrogenation and cracking on Pt (100) are lower than those on Pt (111), which indicates

that cubic particles exhibits a higher catalytic activity than the octahedral particles. However,

the desired product, propylene, may undergo further dehydrogenation readily on Pt (100),

leading to a low selectivity toward propylene. [25]

2.3.2 Catalyst support

Properties that are desired for support materials are high platinum dispersion inducement, as

small particles are the most active, a high surface area and thermal stability. Another important

support aspect is a low pore diffusional resistance. Without the low diffusional resistance, the

catalytic reaction rate is limited by the intraparticle mass transfer rate, so both activity and

selectivity are lowered. [8] Among many high surface area materials, alumina is the support of

choice. Alumina (Al2O3) has excellent thermal stability and mechanical strength under

processing, transport and catalyst regeneration conditions. However, the most important reason

why alumina is chosen as support material is its superior capability of maintaining a high degree

of platinum dispersion. However, also other high surface support materials are proposed and

employed in various studies such as SBA-15, ZSM-5, SAPO-34, ZnAl2O4, mesoporous alumina

(MA) and calcined hydrotalcite Mg(Al)Ox. [15, 51-54]

Apart from the selected support, the employed impregnation method plays a key role in the

characteristics of the active metal phase. Therefore, if a comparison of different supports is

conducted, other preparation conditions are kept constant such as the impregnation method.

Zhang et al. prepared a set of different bimetallic Pt-Sn catalysts on four supports: γ-Al2O3,

MA, ZSM-5, SBA-15 (in order of increasing BET surface), all prepared with the co-

impregnation method with constant loading of Pt (0.5 wt.%) and Sn (1.0 wt.%). In the case of

first three supports, the metallic Pt-Sn particles are distributed homogeneously with an average

particle sizes of 16.7 (γ-Al2O3), 15.5 (MA) and 11.6 nm (ZSM-5), indicating that the dispersion

increases with the BET surface of the support. However, the SBA-15 sample has the highest

specific surface, but here an average particle size of 26.4 nm is observed. This can be explained

based on the weak interaction between active phase and support. These weak interactions can


facilitate migration of metallic particles during the catalyst preparation, leading to

conglomeration of particles. [54]

Zhang et al. propose also that the support influences the stability of the promoting metal tin, as

it can be present in different oxidation states with different catalytic properties. When this

promoter exists in a reduced state (Sn0), it is expected to poison the catalyst; while it exists in

an oxidized state (Sn4+ or Sn2+), it acts as a promoter. Therefore, it is deduced that the SBA-15

support cannot effectively stabilize the oxidative state of tin species, explaining the decrease in

catalytic activity. The acidity of the support is another important parameter, as it is proposed

that it aids in the high dispersion of the metal particles. However high support acidity is

disastrous as side reactions are mainly catalyzed by strong acid centers, which is expected in

ZSM-5 supported catalysts. [10, 54]

Furthermore, Zhang et al. determined the performance of the catalysts during propane

dehydrogenation, as illustrated in Figure 2-15. The poorest catalytic activity and activity loss is

described by the Pt-Sn/SBA-15 catalyst, as it exhibits weak interaction between support and

the metal particles, leading to agglomeration of metal particles and unstable Sn promotors.

However, the selectivity towards propylene is high since almost no side reactions occur. The

PtSn/ZSM-5 catalyst shows the highest catalyst activity, but its high acidity facilitates side

reactions such as C-C scission, reducing its propylene selectivity. A trade-off has to be made

between activity and selectivity; the most promising supports with the highest propylene yield

are γ-Al2O3 and MA.

Figure 2-15: (a) Conversion and (b) selectivity as function of time for the different catalysts: (1) Pt-Sn/ZSM-5; (2) Pt-Sn/γ-Al2O3; (3) Pt-Sn/MA; (4) Pt-Sn/SBA-15. Reaction conditions: 590 °C, H2/C3 = 0.25 (molar ratio), m(cat) = 1.0 g, WHSV = 3.0 h−1. [54]


Recent studies have shown that calcined hydrotalcite is a promising support for light alkane

dehydrogenation as it results in high dispersion of the modified platinum particles and maintains

a high thermal stability under light alkane dehydrogenation reaction conditions. [15, 35, 55] As

hydrotalcite support is a double-layered Mg-Al mixed oxide, the presence and strength of basic

or acid sites depends on the Mg-to-Al ratio. [56, 57] Furthermore, hydrotalcite-derived supports

can be utilized to promote platinum catalysts with other elements (e.g., Zn, Ga, In), which

replace the Al in the hydrotalcite support. These promoters are able to form an alloy with Pt

under certain conditions, hereby enhancing its catalytic properties.

2.3.3 Active metal phase and support interaction

Metal-support interactions are widespread phenomena that differ in strength. This is influenced

by three key properties of the interacting metal-support: energetics of metal particles, geometric

properties and electronic properties.

The interactions are often divided based on their strength, varying from weak to strong metal-

support interactions. Strong metal-support interactions (SMSI) are the most promising kind as

they have the most influence on the active metal phase. The general examples for the SMSI

phenomenon are Group VIII metals such as Pt, Ir, Os and Pd, dispersed on titania support

(TiO2). This strong effect can alter both the morphology of the active phase and the support. In

the case of Pt clusters on titania, the metal particles take on a more pillbox-like morphology as

opposed to the rounded globular morphology in the absence of SMSI. Moreover, the titania

support is transformed to Ti4O7 in the presence of Pt clusters. The appearance of SMSI strongly

depends on the reduction temperature, as at low reduction temperature these morphology

alterations are not observed.

In the case of a bimetallic active phase, the interaction with support becomes more important,

especially as the support strength differs between the two metals. This difference in support

interactions could lead to phase separation of the two metals. When this reconstruction of the

surface occurs, the catalytic activity can change dramatically and it is important to consider this

phenomenon. [58, 59]

Ren-Yuan et al. found evidence of strong metal-support interaction in alumina-supported

platinum catalysts. Based on temperature programmed reduction (TPR) experiments, it is

concluded that a Pt-Al alloy is formed after the reduction of the alumina supported Pt catalysts.

However, at higher reduction temperature, the alumina support reduces further and dilutes the

Pt-Al alloy with Al atoms, indicating that selecting the appropriate reduction temperature is key


in the formation of desired alloys. It is important to note that interactions are reversible and can

be destroyed by reoxidation, which could be the key for understanding the regeneration process

of deactivated catalysts. [60]

As cokes formation on the active phase remains unavoidable, regeneration under oxidizing

atmosphere is often necessary in industrial applications. However, two deactivation modes of

the metal particle can occur under reaction conditions, depending on the active metal-support

interactions. Pt catalyst supported on oxides such as Al2O3, Si2O3, Ti2O3 etc. show sintering

which results in loss of small catalyst particles and a larger average particle diameter. It is also

possible that reactions with the support itself leads to the formation of MPtOx (M=Mg, Ce),

which reduces the amount of catalytic exposed active Pt particles and consequently a loss in

catalytic activity. To reactivate these metal particles often expensive H2-reduction and/or oxy-

chlorination are required [61]

Li et al. proposed a well-defined cuboctahedral MgAl2O4 spinel support, which is able to

stabilize platinum particles in the range of 1-3 nm and does not show deactivation modes under

oxidizing conditions. This support material has two dominant {100} and {111} facets and in

the cuboctahedral shape, six {100} facets are isolated by eight {111} facets. Both spectrometric

analysis and DFT calculations are employed to unravel the stabilization mechanism of Pt

particles on this spinel support. The HRTEM analysis shows that fresh prepared catalyst has

small Pt particles on both above-mentioned facets, however during aging, the small Pt particles

on the {100} surfaces are sintered and unstable, while the high dispersion is maintained on the

{111} surfaces. Due to the small {111} surfaces and their isolation in the cuboctahedral

structure, the supported Pt particles are limited in size. DFT calculations are conducted on Pt

depositions on Mg/O-terminated {100} facet and O/(Mg-Al-Mg) terminated {111} facet

dynamically for 10 ps at 800 °C. These experiments revealed that Pt(111) is attracted by O/(Mg-

Al-Mg) terminated {111} facet and grows epitaxial in a lattice-matching way on the support,

while Pt(100) is repelled by Mg/O-terminated {100} facet, confirming the spectrometric results.

2.4 Conclusions

Light alkane dehydrogenation is an endothermic reaction that is limited by chemical

equilibrium and the equilibrium constant is rather low for smaller carbon chains such as ethane

and propane. Consequently, to obtain higher conversion either higher temperatures or lower

pressures are necessary. Platinum metal catalysts are an adequate choice as catalyst for light


alkane dehydrogenation as they exhibit a high activity for the dehydrogenation steps. However,

those catalysts lack a high selectivity towards olefins, as many side reactions occur on the

surface. Eventually these side reactions initiate coke formation, leading to unwanted

deactivation of the catalyst. To improve the catalytic properties of platinum-based catalysts, the

metal phase is alloyed with an additional metal such as Ga, In, Sn. Primarily, the alloying leads

to an increase in catalyst activity. The rate of olefin formation is higher for bimetallic catalyst

as the alloying metal attenuates the paraffin adsorption and the olefin desorption, two major

steps in the dehydrogenation reaction mechanism. Secondly, the coke formation is reduced by

alloying the platinum catalysts with other elements as it distorts the larger ensembles of active

metal atoms and hereby increases the olefin selectivity by reducing the coke formation.

The two most interesting bimetallic catalysts are Ga-promoted Pt and Sn-promoted Pt catalysts.

The first is employed as catalyst model throughout this thesis, but little research is conducted

on this catalyst type, while the properties of the second catalyst are already heavily investigated

and can be employed as a guide. The introduction of Sn lowers the desorption barrier for

propylene to the gas phase and simultaneously increases the energy barrier for propylene

dehydrogenation. The deposition of Sn makes it difficult to achieve a large ensemble of Pt

atoms, which is essential to activate cracking reactions because sufficient large ensembles are

required to accommodate detached fragments. As the Sn content increases with respect to the

Pt content, the selectivity toward propylene desorption is significantly improved. Considering

the trade-off between the catalytic activity and the selectivity, the Pt3Sn bulk alloy is proposed

as the best candidate for propane dehydrogenation.

The recent experimental studies on Ga-promoted Pt catalyst during light alkane

dehydrogenation have already shown an increase in catalytic activity and propylene selectivity

with respect to an unmodified Pt catalyst on the same support. The PtxGay alloy at the catalytic

surface of the bimetallic surface is determined by comparing frequencies of CO adsorption

experiments with the results of Pt-Ga catalyst models with various Pt/Ga ratios. The best

candidate of the different studied catalyst models is the Pt3Ga bulk alloy, with the lowest Ga

content. This model will be further used to correlate experimental observed catalytic properties

with those based on DFT calculations.

As coke formation on the catalyst surface is an unavoidable, it is rudimental to find the origin,

nature and location of the coke and the mechanism that leads to the deactivation of the catalyst.

At high temperature, TPO results shown three possible locations where cokes can be formed:

the Pt metal itself, Pt-support boundary or on the support. A different type of cokes can be


formed on each location and models are proposed that only a small part of the coke is

responsible for the deactivation of the catalyst while the larger part is formed on the support

directly or spills over from the catalyst surface. The nature of the coke is a pregraphite-like

carbon structure such as graphene and the coke formation is mainly initiated at low-

coordination sites such as step sites. However, the final structure of the coke and its spillover

characteristics is highly dependent on the particle size of the catalyst.

The effect of alloying of the Pt with Sn accelerates the coke spillover from the catalytic phase

to the support, decreasing the coke fraction on the metal and hence ensuring a longer stability

and activity of the catalyst. Several models are also proposed to predict the coke formation on

Pt-Sn catalysts and the dependency of the hydrogen effect on the formation rate. Few

observations are made concerning the coke formation on Pt-Ga catalysts. However, it is

proposed that the presence of Ga acts similar as Sn and reduces the deactivation and the amount

of coke deposited during light alkane dehydrogenation.

Besides alloying the Pt metal particles with other metals, the active metal phase structure and

the support type can contribute to the overall activity and selectivity of the catalyst as the

dehydrogenation reactions are expected to be structure sensitive and support-metal phase

interaction can enhance the catalytic properties. Platinum is a highly active catalytic and

expensive element, so high dispersion of the metal particles is desired property. Various factors

such as the amount of platinum precursors, the type of the support and the utilized impregnation

method can be altered to achieve a higher degree of dispersion. Various researchers have

established that the alloying of the Pt with either Sn or Ga has a small negative to no influence

on the average particle size.

The selection of an appropriate support is based on the desired properties such as high platinum

dispersion inducement, as small particles are the most active, a high surface area and thermal

stability. Among many high surface area materials, alumina is the support of choice in industrial

applications. However multiple other high surface materials such as SBA-15, ZSM-5, SAPO-

34, ZnAl2O4, (MA) and calcined hydrotalcite Mg(Al)Ox are proposed to stabilize the active

metal phase and replace alumina as support.

Furthermore, the metal phase-support interactions and the acidity of the support also play a

major role. When the metal phase-support interaction is too weak, the Pt cannot be stabilized

and conglomerate to larger particles. If the acidity of the support is high, a higher overall

catalytic activity is observed, however poor selectivity towards propylene is obtained due to


side reactions occurring on the support surface. Another interesting support is calcined

hydrotalcite Mg(Al)Ox as it induces a high degree of dispersion and can impregnate other metals

e.g., Zn, Ga, In by replacing the Al in the hydrotalcite layers.

Finally, deactivated catalysts have to remain stable under oxidizing regeneration conditions as

often the morphology of the active metal phase is altered during this process since sintering or

the formation of unreactive MPtOx can occur.

2.5 References

1. McNaught, A.D. and A.D. McNaught, Compendium of chemical terminology. Vol. 1669. 1997: Blackwell Science Oxford.

2. Fechete, I., Y. Wang, and J.C. Vedrine, The past, present and future of heterogeneous catalysis. Catalysis Today, 2012. 189(1): p. 2-27.

3. Ross, J.R.H., Chapter 1 - Heterogeneous Catalysis – Chemistry in Two Dimensions, in Heterogeneous Catalysis, J.R.H. Ross, Editor. 2012, Elsevier: Amsterdam. p. 1-15.

4. Ross, J.R.H., Chapter 2 - Surfaces and Adsorption, in Heterogeneous Catalysis, J.R.H. Ross, Editor. 2012, Elsevier: Amsterdam. p. 17-45.

5. Ross, J.R.H., Chapter 6 - The Kinetics and Mechanisms of Catalytic Reactions, in Heterogeneous Catalysis, J.R.H. Ross, Editor. 2012, Elsevier: Amsterdam. p. 123-142.

6. Bhasin, M.M., et al., Dehydrogenation and oxydehydrogenation of paraffins to olefins. Applied Catalysis a-General, 2001. 221(1-2): p. 397-419.

7. Sanfilippo, D., Dehydrogenation of paraffins: synergies between catalyst design and reactor engineering. Catalysis today, 2006. 111(1-2): p. 133-139.

8. Vora, B.V., Development of Dehydrogenation Catalysts and Processes. Topics in Catalysis, 2012. 55(19-20): p. 1297-1308.

9. Bricker, J.C., Advanced Catalytic Dehydrogenation Technologies for Production of Olefins. Topics in Catalysis, 2012. 55(19-20): p. 1309-1314.

10. Zhang, Y., et al., Propane dehydrogenation on PtSn/ZSM-5 catalyst: Effect of tin as a promoter. Catalysis Communications, 2006. 7(11): p. 860-866.


12. Tilley, R.J., Crystals and crystal structures. 2006: John Wiley & Sons. 13. Downs, R.T. and M. Hall-Wallace, The American Mineralogist crystal structure

database. American Mineralogist, 2003. 88(1): p. 247-250. 14. Feng, J., M. Zhang, and Y. Yang, Dehydrogenation of Propane on Pt or PtSn Catalysts

with Al2O3 or SBA-15 Support. Chinese Journal of Chemical Engineering, 2014. 22(11–12): p. 1232-1236.

15. Galvita, V., et al., Ethane dehydrogenation on Pt/Mg(Al)O and PtSn/Mg(Al)O catalysts. Journal of Catalysis, 2010. 271(2): p. 209-219.

16. Yu, C.L., et al., Propane dehydrogenation to propylene over Pt-based catalysts. Catalysis Letters, 2006. 112(3-4): p. 197-201.

17. Zaera, F. and D. Chrysostomou, Propylene on Pt(111) I. Characterization of surface species by infra-red spectroscopy. Surface Science, 2000. 457(1-2): p. 71-88.

18. Zaera, F. and D. Chrysostomou, Propylene on Pt(111) II. Hydrogenation, dehydrogenation, and H-D exchange. Surface Science, 2000. 457(1-2): p. 89-108.


19. Zaera, F., T.V.W. Janssens, and H. Öfner, Reflection absorption infrared spectroscopy and kinetic studies of the reactivity of ethylene on Pt(111) surfaces. Surface Science, 1996. 368(1–3): p. 371-376.

20. Nykanen, L. and K. Honkala, Selectivity in Propene Dehydrogenation on Pt and Pt3Sn Surfaces from First Principles. Acs Catalysis, 2013. 3(12): p. 3026-3030.

21. Nykänen, L. and K. Honkala, Density Functional Theory Study on Propane and Propene Adsorption on Pt(111) and PtSn Alloy Surfaces. The Journal of Physical Chemistry C, 2011. 115(19): p. 9578-9586.


23. Valcárcel, A., et al., Theoretical study of the structure of propene adsorbed on Pt(111). Surface Science, 2002. 519(3): p. 250-258.


25. Yang, M.-L., et al., Tuning selectivity and stability in propane dehydrogenation by shaping Pt particles: A combined experimental and DFT study. Journal of Molecular Catalysis A: Chemical, 2014. 395(0): p. 329-336.

26. Sabbe, M.K., et al., Benzene adsorption on binary Pt3M alloys and surface alloys: a DFT study. Physical Chemistry Chemical Physics, 2013. 15(29): p. 12197-12214.

27. Held, G., Adsorption and dissociation of benzene on bimetallic surfaces—the influence of surface geometry and electronic structure. Journal of Physics: Condensed Matter, 2003. 15(34): p. R1501.

28. Müller, S., Bulk and surface ordering phenomena in binary metal alloys. Journal of Physics: Condensed Matter, 2003. 15(34): p. R1429.

29. Koel, B.E., Structure, Characterization and Reactivity of Pt–Sn Surface Alloys, in Model Systems in Catalysis. 2010, Springer. p. 29-50.

30. Sanfilippo, D. and I. Miracca, Dehydrogenation of paraffins: synergies between catalyst design and reactor engineering. Catalysis Today, 2006. 111(1-2): p. 133-139.

31. Massalski, T.B., et al., Binary alloy phase diagrams. Vol. 1. 1986: American Society for Metals Metals Park, OH.

32. Tsai, Y.L., C. Xu, and B.E. Koel, Chemisorption of ethylene, propylene and isobutylene on ordered Sn/Pt(111) surface alloys. Surface Science, 1997. 385(1): p. 37-59.

33. Yang, M.L., et al., First-Principles Calculations of Propane Dehydrogenation over PtSn Catalysts. Acs Catalysis, 2012. 2(6): p. 1247-1258.

34. Jablonski, E.L., et al., Effect of Ga addition to Pt/Al2O3 on the activity, selectivity and deactivation in the propane dehydrogenation. Applied Catalysis a-General, 1999. 183(1): p. 189-198.


36. Redekop, E.A., et al., Delivering a Modifying Element to Metal Nanoparticles via Support: Pt–Ga Alloying during the Reduction of Pt/Mg(Al,Ga)Ox Catalysts and Its Effects on Propane Dehydrogenation. ACS Catalysis, 2014. 4(6): p. 1812-1824.


38. Wang, J.S., et al., First-principles calculations assisted thermodynamic assessment of the Pt–Ga–Ge ternary system. Calphad, 2009. 33(3): p. 561-569.

39. Bond, G.C., Metal-Catalysed Reactions of Hydrocarbons. Fundamental and applied Catalysis. Vol. XXI. 2005: Springer. 666.


40. Magnoux, P., H.S. Cerqueira, and M. Guisnet, Evolution of coke composition during ageing under nitrogen. Applied Catalysis a-General, 2002. 235(1-2): p. 93-99.

41. Larsson, M., et al., The effect of reaction conditions and time on stream on the coke formed during propane dehydrogenation. Journal of Catalysis, 1996. 164(1): p. 44-53.

42. Li, Q., et al., Coke Formation on Pt–Sn/Al2O3 Catalyst in Propane Dehydrogenation: Coke Characterization and Kinetic Study. Topics in Catalysis, 2011. 54(13-15): p. 888-896.



45. Vu, B.K., et al., Location and structure of coke generated over Pt–Sn/Al2O3 in propane dehydrogenation. Journal of Industrial and Engineering Chemistry, 2011. 17(1): p. 71-76.

46. Liu, K., et al., Hydrogasification of coke in heptane reforming over Pt-Re/Al2O3. Industrial & Engineering Chemistry Research, 2003. 42(8): p. 1543-1550.

47. Van Santen, R.A., Complementary Structure Sensitive and Insensitive Catalytic Relationships. Accounts of Chemical Research, 2008. 42(1): p. 57-66.

48. Pérez, O.-L., D. Romeu, and M.J. Yacamán, The relation between dispersion and particle size on supported catalysts. Journal of Catalysis, 1983. 79(1): p. 240-241.

49. Zhu, H., et al., Sn surface-enriched Pt–Sn bimetallic nanoparticles as a selective and stable catalyst for propane dehydrogenation. Journal of Catalysis, 2014. 320(0): p. 52-62.

50. Santhosh Kumar, M., et al., Dehydrogenation of propane over Pt-SBA-15 and Pt-Sn-SBA-15: Effect of Sn on the dispersion of Pt and catalytic behavior. Catalysis Today, 2009. 142(1–2): p. 17-23.

51. Santhosh Kumar, M., et al., Dehydrogenation of propane over Pt-SBA-15: Effect of Pt particle size. Catalysis Communications, 2008. 9(5): p. 747-750.

52. Vu, B.K., et al., Pt–Sn alloy phases and coke mobility over Pt–Sn/Al2O3 and Pt–Sn/ZnAl2O4 catalysts for propane dehydrogenation. Applied Catalysis A: General, 2011. 400(1–2): p. 25-33.

53. Nawaz, Z. and F. Wei, Pt–Sn-based catalyst's intensification using Al2O3–SAPO-34 as a support for propane dehydrogenation to propylene. Journal of Industrial and Engineering Chemistry, 2011. 17(3): p. 389-393.

54. Zhang, Y., et al., Comparative study of bimetallic Pt-Sn catalysts supported on different supports for propane dehydrogenation. Journal of Molecular Catalysis A: Chemical, 2014. 381(0): p. 138-147.

55. Sun, P.P., et al., Novel Pt/Mg(In)(Al)O catalysts for ethane and propane dehydrogenation. Journal of Catalysis, 2011. 282(1): p. 165-174.

56. McKenzie, A.L., C.T. Fishel, and R.J. Davis, Investigation of the surface-structure and basic properties of calcined hydrotalcites. Journal of Catalysis, 1992. 138(2): p. 547-561.

57. Di Cosimo, J.I., et al., Structure and surface and catalytic properties of Mg-Al basic oxides. Journal of Catalysis, 1998. 178(2): p. 499-510.

58. Tauster, S., Strong metal-support interactions. Accounts of Chemical Research, 1987. 20(11): p. 389-394.

59. Spencer, M.S., Models of strong metal-support interaction (SMSI) in Pt on TiO2 catalysts. Journal of Catalysis, 1985. 93(2): p. 216-223.


60. Ren-Yuam, T., W. Rong-An, and L. Li-Wu, Evidence of strong metal-support interaction in alumina-supported platinum catalysts. Applied catalysis, 1984. 10(2): p. 163-172.

61. Li, W.Z., et al., Stable platinum nanoparticles on specific MgAl2O4 spinel facets at high temperatures in oxidizing atmospheres. Nature Communications, 2013. 4: p. 8.

Chapter 3: Computational methodology 39

Chapter 3 Computational methodology In the last three decades, quantum chemistry has become a valid tool in the study of atoms and

molecules and, increasingly, in modeling complex systems such as in biology, chemistry and

materials science. The current ab initio techniques are generally based on the computational

solution of the electronic Schrodinger equation. For given positions of a collection of atomic

nuclei and the total number of electrons in the system the electronic energy and electron density

are calculated. The ability to obtain appropriate solutions to the electronic Schrodinger equation

within the desired convergence for systems containing tens, or even hundreds, of atoms has

revolutionized the ability of theoretical chemistry to address important problems in a wide range

of disciplines. [1]

Nowadays, computational chemistry is able to provide adequate results complementing

experimentally obtained data, which have the disadvantage of being labor intensive and time

consuming to obtain. Computational methods also add another advantage with respect to

experimental methods since they can describe molecular behavior on a smaller scale than the

available experimental tools.

This chapter firstly introduces the available computational methods to represent catalyst

models. Secondly, the density functional theory techniques and the applied framework of this

work are described.

3.1 Catalyst models

Industrial dehydrogenation processes are generally based on a heterogeneous noble metal

catalyst on a support. The active sites for catalysis are located on the surface of the metal

40 Chapter 3: Computational methodology

particles. An appropriate catalyst model is necessary to obtain relevant results from ab initio

calculations. Due to the computational limitations, it is not possible to describe every atom in

the metal particles, so three main approaches are proposed to model the catalyst on nanoscale:

cluster approaches, embedded cluster approaches and periodic models, as visualized in Figure

3-1. Each method has their own merits and flaws, and the choice depends mainly on the

industrial context of the catalyst, but constraints on computational resources often result in the

selection of the computationally less demanding method. [2]

The cluster approach (Figure 3-1, left) represents the surface by a finite number of atoms. This

type of model is the most basic and is computationally the fastest. Its accuracy however is

variable, depending on the absolute size of the catalyst and the chosen bond saturation at the

cluster periphery. In the case of a metal catalyst with a diameter of few nm this approach is

inadequate due to the delocalized nature of metal bonding. [3]

Embedding techniques (Figure 3-1, middle) improve upon the cluster approach by introducing

electrostatic, steric and possibly elastic constraints on the cluster imposed by its environment.

As a result, this will reduce the periphery effects. The degree of detail in the description of the

environment and cluster interactions determines the level of accuracy and performance of

different embedding methods. For example, an accurate reproduction of the electrostatic field

in the region of interest is crucial for the study of the majority of solid metal oxide catalysts.

[4]

Periodic methods (Figure 3-1, right) place the catalyst geometries in periodic boundary

conditions and usually employ a full quantum chemical treatment for the total system. These

approaches have been widely used in studies of catalysis on metal and semiconductor surfaces

where the delocalized electron states are of importance and edge effects become negligible. As

periodic images are employed, the size of the unit cell is constrained; on the one side, a well-

chosen size is required to model correctly adsorbate coverage or less periodic effects e.g. point

defects. On the other side, larger unit cells will heavily increase the computational cost. [2, 4]


Figure 3-1: Different catalyst models for a fcc(111) surface: (left) cluster model, (middle) embedded cluster approach, and (right) periodic slab model [2]

The approaches mentioned above are easily implemented for an ideal description of the catalyst,

but the representation deviates from the actual catalyst as these models exclude surface defects,

multifaceted surfaces to account for structural sensitivity and support interaction influencing

the catalytic phase and enacting possible spillover from the metal catalyst to the support.

3.1.1 Adaptations of the ideal catalyst model

Non-idealities such as surface defects, structural sensitivity and support interactions can be

implemented into a more sophisticated and representable catalyst model, however including

one or multiple of these complexities will give rise to higher computational effort. For this

reason, only the most relevant one is incorporated in the catalyst model.

Figure 3-2: Modeling of more complex surfaces: facet edges, steps and three-phase boundary, shown for an fcc metal: (left) zig-zag slab structure; (middle) high-index surface, here the (211) surface; (right) modeling of the particle-support interface: infinite metal rod on support. [2]


In the case of structural sensitivity, all the involved multifaceted surfaces have to be calculated.

Multiple calculations for differently indexed surfaces have to be performed, hereby increasing

steeply the computational cost. Furthermore, it is also necessary to model the undercoordinated

sites e.g. edges, steps, kinks and terraces in a representable vicinity of the catalyst particle as

they often have an essential influence on the catalyst properties, e.g. B5-sites of Ru catalysts.

[5] In both cluster and embedded cluster approaches, the structural sensitivity can be addressed

straightforwardly by devising models that represent the undercoordinated sites of the catalyst

particle. However, the construction of finite clusters, whether or not embedded, has to be done

with care, as studies on cluster size convergence are necessary. [6]

In the periodic slab model, it is harder to incorporate structural sensitivity since the surface unit

cell is periodically extended and a zig-zag structure, where the edge, kink and/or step are

exposed, is used to model the structure (i.e. facet edge, Figure 3-2, left). A flaw of these zig-

zag structures is that some undercoordinated faces are in an unnatural local environment.

Another option to model these undercoordinated surfaces is through implementation of high

index surfaces, which are arrays of the desired step or kink, separated by terraces with specific

width (i.e. step edge, Figure 3-2, middle). It is important to remark that properties of the system

calculated by these two models could deviate due to interference of the periodicity of the

approach, as certain edge effects are artificially induced on each other. This flaw could be

avoided by increasing the size of the unit cell, but this is disastrous for the needed computational

effort.

Interaction between the catalyst particle (the active phase) and the support phase (Figure 3-2,

right) can strongly alter the catalytic properties of the active phase as the support can induce

charge transfer or strain effects on the catalyst itself. When modeling the catalyst-support

interaction, it is important to note that its properties can depend on the molecular coverage, as

this influences the catalyst geometry relaxation and metal-support strength. [7] Another

phenomenon that will occur is spillover i.e. adsorbate diffusion between the support and the

particle. In the vast majority of computational studies, these support effects are neglected, while

they can contribute to specifics of the catalyst activity. [2, 8]

To incorporate support effects, it is essential to model simultaneously the catalyst and the

support. Due to the high computational cost, the support effects are often neglected when

modeling a supported catalyst, but several methods are available that integrate the support

effect. Two models are highlighted: the first is based on the periodic film model, where periodic

layers of metal catalyst are deposited on the periodically modeled support (Figure 3-3, left).


The second is the cluster-on-periodic-support approach, in which finite metal clusters are

deposited on a periodic modeled cluster (Figure 3-3, right). As the first model utilizes the

smallest unit cell, it is computationally more attractive, but the results can be influenced by

incorrect charge transfer between periodic images or false induced strain between the catalyst

and support. The other model has the merit that relaxation of the catalyst is not hindered.

Another merit of cluster-on-periodic-support model is that also adsorption on the support phase

or catalyst-support interface can be accounted for. [9, 10]

Figure 3-3: (left) Periodic film model: V2O5−TiO2 slabs. Side view of a (001) V2O5 monolayer supported on (001) TiO2 anatase (weak interaction with support).[9] (right) Cluster-on-periodic-support model: The structures of anatase TiO2 (001)-supported Pt with bond distances are in Å. [10]

The main difficulty with incorporating structure sensitive and metal-support effects is that you

have to do it fully; there is no middle ground where abstraction can be made of certain effects.

Thus, it is essential to have a clear understanding of the catalyst structure and its composition

under reaction conditions. In situ methods, such as in situ X-ray absorption near edge structure

(XANES), extended X-ray absorption fine structure (EXAFS) and X-ray diffraction techniques

(XRD) provide valuable information of catalysts when particles have the size of a few

nanometers. While the development of in situ characterization techniques of nanocatalysts

continues, it is impossible to obtain useful results on small particles (size <2 nm) and

consequently a well-educated guess of the catalyst model has to be proposed. [11, 12]

3.1.2 Catalyst model used in this work

In this work, the periodic slab approach will be used in first instance to model the catalyst

surface and determine the reaction network. Later on, the model can be extended or altered to

incorporate the support and multifaceted effects. The configurations of each model will be

discussed further in this work.


3.2 Periodic ab initio calculations

Ab initio quantum calculations are based on so-called “first principles” as no input from

experimental or other sources are employed to calculate molecular geometries and properties.

At its basis stand the postulates of quantum mechanics and other laws of physics and chemistry

to define the framework of these calculations. Centrally the Schrödinger equation, which

represents the most stable electronic structure of the nuclei-electron system, is solved.

In its exact form, the electronic Schrödinger equation is a many-body problem of which the

computational complexity grows exponentially with the number of electrons, and hence, a brute

force solution is time-consuming and inefficient. Several approximating equations can be

applied to the electronic Schrödinger equation to reduce calculation time. Two major

approaches are utilized to solve this equation: Molecular orbital (MO) method and Density

functional theory (DFT) method. The first method is generally based on the Hartree-Fock

theory: a wave function method with a mean field approach, neglecting the correlation between

electrons. This method produces reasonable results for many properties, but is incapable of

providing a robust description of the electronic energy and the electron correlation that is

essential to describe systems with delocalized electrons. Post Hartree-Fock techniques improve

the robustness of the electronic energy calculation, but are lagging for molecular systems with

high amount of electrons as this sharply increases the computational cost of ab initio

calculations. [1, 13]

Nowadays, first-principles calculations on heterogeneous metal catalysts are entirely based on

the Kohn-Sham DFT methods. This method employs the optimization of three-dimensional

electron density to solve the electronic Schrödinger equation in contrast to Hartree-Fock’s high

dimensional wave function. The electron density uniquely determines the Hamiltonian operator

and thus the properties of the system as stated by the first Hohenberg-Kohn theorem. Hence,

the ground state energy can be described with a unique functional as function of the electron

density. This universal functional, independent of the system at hand, is the key of the DFT

methods. The explicit form cannot be determined entirely because description of the electron-

electron correlation cannot be extracted directly and hence DFT methods employ the

description of an exchange-correlation functional to approximate this. The Jacob’s ladder, as

illustrated in Figure 3-4, summarizes different types of functional with increasing complexity

and computational cost. [2, 14, 15]


Figure 3-4: Jacob's ladder of density-functional approximations (after Perdew). [2]

Basic DFT functionals employ semi-local exchange correlations such as local-density (LDA)

or generalized gradient approximations (GGA). [16, 17] Both have a relative low computational

cost and still yield decent results of covalent bonds and geometric structure. Two major flaws

can be attributed to these types of functionals: artificial electron delocalization and

underestimation of long-range interactions e.g. van der Waals (vdW) interactions. The first can

be attributed to an incomplete cancellation of repulsive Coulomb self-interactions by the

approximate exchange energy calculated by the selected functional. The second arises from the

nature of these more basic functionals as they only consider local electron density and neglect

the non-local effects like dispersion interactions.

Initially, the Perdew-Burke-Ernzerhof (PBE) functional, belonging to the GGA functional

category, is employed for the ab initio calculations in this work. [17] The corresponding results

are expected to be mostly inaccurate due to mentioned shortcomings, especially when applied

in heterogeneous catalysis. The results obtained by PBE are used as initial geometry in a next

calculation using a more advanced functional and hence reducing the computational cost of this

progressive functional. Promising DFT functionals are the hybrid and advanced non-local

functionals. In this work, it was opted to select a vdw-DF functional, the optPBE-vdW

functional. [18, 19] Generally, vdw-DF functionals introduce long-range dispersion in

approximate exchange-correlation functionals. The exact form of this non-local correlation

energy is the key parameter that determines the excellent properties of this type of functional.

In the optPBE-vdW functional, the exchange functional is optimized with respect to the


correlation part of PBE-types and the long-range dispersion part of vdW-DF functionals to

describe accurately the bonding energies in the S22 dataset. [20, 21] However, the vdW-DF

functional has some flaws e.g. it gives falsified results if hydrogen bonds are present in the

molecular system as it overestimates the bonding strength of hydrogen.

The accuracy of RPBE, a PBE type of functional is in range of 20-30 kJ/mol for DFT

calculations. [22] No direct accuracy results are available for covalent bindings for DFT

calculations with the vdw-DF functional. Still, it is reported that especially for vdW interactions

the accuracy is lower than 15 kJ/mol. [21] Literature states that vdW-DF functionals can give

overbinding of chemisorbed species on a metal surface. However, a new Bayesian error

estimation (BEEF) vdw-DF functional is proposed that better fits various datasets and correctly

describes the binding of chemisorbed species. [23] However, none of these tested benchmarks

consisted of chemisorption on transition metal catalyst models. As verification of this

statement, the calculated adsorption energies of propane (physisorbed) and propylene on

Pt(111) (4×2 unit cell) using optPBE and BEEF vdW-DF will be compared to experiment. [24]

It is reported that the BEEF functional predicts less strong overbinding compared to optPBE

vdw-DF. As seen in Table 3-1, the BEEF vdw-DF functional indeed gives better predictions

for adsorption energies. However, it should be noted that these are solely preliminary results as

different coverage are employed in the DFT study than experimentally.

Table 3-1. Comparison of propylene/propane adsorption strength for two different functionals and experiment.

ΔEads

(kJ/mol) Experimental data

(0.2 ML) [24] OptPBE vdW-DF

(0.13 ML) BEEF vdW-DF

(0.13 ML) Propylene chemisorption -68 -134 -105

Propane physisorption - -43 -33

Recent literature reports that the selected optPBE vdW-DF is adequate to predict the adsorption

strength of benzene on transition metals, indicating that this functional indeed satisfy for

propane dehydrogenation on Pt-based catalysts. [25] However, to reliable asses the

performance of optPBE vdW-DF functional for propane dehydrogenation further

benchmarking is needed.


3.3 Computational framework used in this work

3.3.1 Vienna Ab initio Simulation Package

The calculations from this work are executed with the Vienna Ab initio Simulation Package

(VASP 5.3.3). [26-29] VASP is a complex package for performing ab initio quantum

mechanical and molecular dynamics (MD) simulations using pseudopotentials and a plane-

wave basis set. In this work, it is opted for the projector-augmented wave (PAW)

pseudopotentials method with a plane-wave cutoff of 400 eV. [30, 31] To start up a calculation

in this framework four input files have to be configured: INCAR file, KPOINTS file, POSCAR

file and POTCAR file. Their meaning and configurations are discussed below. [32]

• INCAR file: The main input file of VASP. It determines the calculation type and the

corresponding settings, and contains a relatively large number of parameters. Several

parameters can be kept on their standard value, but some need to be tuned to obtain the

desired result. The first is the setting for the selection of DFT functional (the ‘GGA’ tag).

The PBE functional corresponds to the GGA=PE, while the optPBE-vdW functional

corresponds to GGA=OR.

The determination of the minimal energy of the potential energy surface (PES) can be

achieved by two main algorithms: the quasi-Newton RMM-DIIS algorithm and the

conjugate gradient algorithm. The IBRION tag determines how atomic position are updated

and moved in each iteration. In the case of relaxation problems close to the minimum, the

quasi-Newton RMM-DIIS algorithm (IBRION=1) is recommended and when the minimum

is still far away, the conjugate gradient algorithm (IBRION=2) should be selected. Other

IBRION values are available, but these are for other optimization problems e.g. transition

state and frequency analysis. The working of these algorithms can be tuned by a scaling

constant for the forces (POTIM tag in the INCAR file). Standard value is 0.5 for relaxation

problems, decreasing this value will lead to smaller position changes each iteration.

Another parameter is the EDIFF tag in VASP, which represents the electronic energy

convergence criterion. This parameter is set to 10-7 for loose geometry optimizations and

10-8 for strict geometry optimizations. It is obliged to obtain strict geometry optimizations,

as they are required for frequency calculations. Otherwise, imaginary frequencies cannot be

excluded.

For the forces, there is also a convergence criterion (EDIFFG tag in the INCAR file) with a

threshold of 0.05 eV/Å for loose geometry optimizations and 0.015 eV/Å for strict geometry


optimizations. The relaxation will be converged if both the energy and forces criteria are

met. Examples of INCAR files for geometry, frequency and transition state calculations can

be found in Appendix A.

• KPOINTS file: The file KPOINTS must contain the k-point coordinates and weights or the

mesh size for creating the k-point grid. The formulation of a so-called Monkhorst-Pack grid

is obligatory as integrals in real space over the periodic system are replaced by integrals

over the first Brillouin zone in reciprocal space, by virtue of Bloch's theorem to reduce

computational time. [33]

• POSCAR file: This file can be split into two sections: the lattice parameters and the initial

atom positions. The lattice parameters contain the unit cell shape, size and the total number

of each element present. A vacuum layer is introduced in the z-direction (c-coordinate)

where gas phase molecules can adsorb. It is attempted to keep the height of this vacuum

layer constant on 12 Å. In the z-direction between subsequent unit cells, an intermediate

artificial dipole layer is used to correct for the dipole interaction between the slabs. In the

lower part of the POSCAR file, the initial atomic positions are noted. Each line stands for

one atom and contains three numbers and three letters. The numbers represent the atomic

position in the unit cell of that specific atom, noted in Cartesian coordinates or fractional

with respect to the lattice parameters. The letters (T or F) denote on the fact that atoms need

relaxation (T) or are kept fixed (F) in the x-, y- or z-direction. In this work, slab

configurations utilize four atomic layers where the bottom two are kept fixed and the upper

two can relax. The bottom layers represent the bulk layers of the catalyst.

• POTCAR file: This file is a collection of information about each atomic species present in

the calculation, merged with the PAW pseudopotentials needed for the calculation.

When VASP calculations are completed, each time several output files are generated. The most

relevant ones are the CONTCAR file, the OUTCAR file and the standard output (stdout) file:

• CONTCAR file: This file has the same format as the initial POSCAR file and contains the

atomic positions of each atom after the last ionic step is converged in VASP. As this file

has the same format as POSCAR files, they can easily be converted and utilized as an initial

position for more accurate calculations.


• OUTCAR file: Detailed file that summarizes extensively all information obtained during

the calculations.

• Stdout file: During the calculations, data are collected in this file of every iteration step.

More information can be retrieved about convergence speed and details on each sequence.

At the bottom of this file, either errors or the desired quantities of the converged result can

be retrieved.

By combining several output and input files, it is possible to determine all kinds of desired

properties of a system. The desired properties are the most stable geometries, the reaction

energies and reaction barriers. These properties can be attained by geometry optimizations of

intermediates on the surface and transition state calculations.

3.3.2 Geometry optimizations

In this section, the practical execution of the VASP geometry optimizations are described. All

computational settings are implemented according to the strict convergence criteria as discussed

in previous paragraphs. In this work, the reaction network of propane dehydrogenation will be

examined and discussed in detail. Here an overview of the practical approach will be described,

while in later chapters the results are discussed. The most stable slab geometries of the desired

PtxGay catalyst were calculated in previous work. [34] The following four catalyst models are

employed in this work i.e. Pt(111), Pt3Ga(111), Gr/Pt(111) and Pt(211). On these slabs,

adsorbate molecules or intermediates in the propane dehydrogenation network are modelled.

Initial geometries and most probable adsorption sites are based on the article of Yang et al. [35]

As first iteration, the calculations are conducted with the PBE functional and the resulted

geometries are utilized as initial condition for the consecutive iteration with the vdW-DF

functional.

In dehydrogenation reactions, adsorbed hydrogen is formed on the catalyst surface. Due the low

hydrogen coverage on the surface and repulsion between the adsorbate and hydrogen, it is

assumed that hydrogen diffuses away from the adsorbate. This is implemented in DFT

calculations by optimizing the adsorbate and the hydrogen in a separate unit cell. For this

reason, the most stable geometry of hydrogen needs to be found. The hydrogen atom is placed

on several possible adsorption sites in the empty unit cell and the most stable hydrogen

geometry is employed in further calculations. Simultaneously, the dehydrogenated adsorbate is


optimized on different adsorption sites and the several optimized possibilities are compared

based on the adsorbate adsorption energy ΔadsE(ad) conform following formula:

∆𝑎𝑎𝑎𝑎𝑎𝑎𝐸𝐸(𝑎𝑎𝑎𝑎) = 𝐸𝐸𝑎𝑎𝑎𝑎,𝑎𝑎𝑎𝑎𝑎𝑎 − 𝐸𝐸𝑎𝑎𝑎𝑎,𝑔𝑔𝑎𝑎𝑎𝑎 − 𝐸𝐸𝑎𝑎𝑠𝑠𝑎𝑎𝑠𝑠 (1)

where Ead,ads is the electronic energy of the dehydrogenated species inside the catalyst unit cell,

Ead,gas is the electronic energy of the adsorbate species in the gas phase and Eslab is the energy

of the clean catalyst surface. The electronic energy of the adsorbate species in the gas phase are

calculated in a 20×20×20 Å³ unit cell and a special gamma point Г is used to sample the

Brillouin zone. These most stable geometries will be used in further calculations.

When the most stable geometries of several multiple adsorbed species are calculated, reaction

energies can be determined. When adsorbed species are used to determine the reaction energy,

it is essential that both reactant and product have the same atoms present in the unit cell. The

reaction energy ΔEr is determined based on the following formula:

∆𝐸𝐸𝑟𝑟 = 𝐸𝐸prod,ads − 𝐸𝐸react,ads (2)

Where Eprod,ads is the electronic energy of the adsorbed reaction product(s) and Ereact,ads is the

electronic energy of the adsorbed reactant(s). The type of reaction such as dehydrogenation,

isomerization or dissociation depends on the reacting molecules.

For the calculation of the reaction barrier, it is essential to model the transition state between

the reactant and the product. The employed computational techniques are discussed in the

following section.

3.3.3 Transition state calculation techniques

To determine the transition state of a reaction, it is necessary to find a saddle point along the

reaction coordinate on the potential energy surface between the reactant and reaction product

with only one imaginary frequency. As finding transition states in a PES is computationally

demanding, the computational strategy will be twofold. First, the Nudged Elastic Band (NEB)

method will be employed to optimize several intermediate images along the reaction path. The

most promising intermediate image will be used in the dimer method to search the nearby saddle

point and to converge towards the desired transition state. In the following sections, these

methods and their configurations will be discussed.


3.3.3.1 Nudged Elastic Band (NEB) method

The NEB is a method to find a minimum energy path (MEP) between a pair of stable states. In

the context of reaction rates, this reactant-product pair has an initial (reactant) and a final state

(product), both of which are local minima on the potential energy surface. The MEP has the

property that any point on the path is at an energy minimum in all directions perpendicular to

the path. This path passes through at least one first-order saddle point. The NEB method utilizes

a string of images (geometric configurations of the system) to describe a reaction pathway.

These configurations are initially created linearly between the initial and final state.

Additionally, the neighboring images are connected by spring forces to ensure equal spacing

along the reaction path, during the calculation. The artificial nudged elastic band force 𝐹𝐹𝑖𝑖NEB is

equal to the spring force 𝐹𝐹𝑖𝑖S∥, along the tangent τ�𝑖𝑖, and the perpendicular force due to the

potential 𝐹𝐹𝑖𝑖⊥, as illustrated in Figure 3-5. Upon convergence of the NEB method, the images

describe the reaction mechanism along the MEP, with a resolution determined by the number

of images. [36-38]

Figure 3-5: Two components make up the nudged elastic band force FNEB: the spring force 𝑭𝑭𝒊𝒊

𝐒𝐒∥, along the tangent 𝛕𝛕�𝒊𝒊, and the perpendicular force due to the potential 𝑭𝑭𝒊𝒊⊥. The unprojected force due to the potential Fi is also shown for completeness. [36]

This technique results in a MEP diagram between the initial and final state and while infinitely

increasing the number of images will eventually lead to the accurate pinpointing of the

transition state, this is a computational impossibility. So this technique is rather used in


combination with the dimer method to optimize the transition state. From the optimized images

in the NEB method, the energetic most unstable image (highest electronic energy) will be used

as an initial guess in the dimer method.

In VASP, the NEB method is implemented as follows: first, a set of images are generated

between two optimized states by using their converged CONTCAR and stdout files. In general,

a number of ten images will be generated, including the initial and final state. The corresponding

VASP function will generate a number of POSCAR files, with initial geometries between the

converged CONTCAR geometries. Furthermore, a special INCAR file has to be loaded (see

Appendix A). Additional parameters have to be configured such as the energy of the initial state

and final state. Those have to be filled respectively in the EFIRST- and ELAST-tag. To obtain

faster convergence, loose convergence criteria are used. The KPOINTS and POTCAR file can

be used from either initial or final state. The electronic energy results of the MEP can be

obtained by applying the corresponding script (TS_Extract.sh) which gives the energies of the

images in the MEP.

3.3.3.2 Dimer method

The dimer method uses a minimum-mode following algorithm to climb to a saddle point,

starting from any initial configuration. The method is designed for finding saddle points without

knowledge of the final configuration of the transition state. The method only makes use of first

derivatives of the potential energy, which makes it computationally more interesting and is

therefore applicable in situations where second derivatives are too costly or too tedious to

evaluate, e.g. in this work where plane wave based density functional theory calculations are

utilized. The method makes use of two replicas -dimer- of the system, spaced apart by a finite

small distance. The dimer alters the force in such a way that the original system converges to a

saddle point rather than a minimum. [39, 40] In this work, the NEB method is used prior to the

dimer calculations to obtain an initial geometry, closer to the transition state geometry and so

further reducing the computational cost.

To obtain the transition state in VASP, the dimer method is employed as follows: primarily two

intermediate images from the converged NEB method are transferred: the electronic most

unstable image (highest electronic energy) and the previous image. The former’s CONTCAR

file is converted to a new POSCAR as the initial transition state geometry. Both images are

used to generate a MODECAR file, which gives an initial direction along the dimer to state the

coordinates that are likely involved in the reaction. If this file is not supplied, the dimer method


will random generate a direction, costing extra computational power. The KPOINTS and

POTCAR file from either the initial or the final state can be used, while a special version of the

INCAR with adaptations is used (see Appendix A). The convergence criteria are set on strict

for this method.

Apart from the general CONTCAR, OUTCAR and stdout output files, a converged dimer

calculation also generates another important file named DIMCAR. This file is also updated

each iteration step. Six values are monitored but force, torques and especially curvature are the

most important parameters. To correctly obtain convergence, the force has to decrease and the

curvature should be negative each iteration.

If the dimer method converges strictly, the transition state geometry and its electronic energy

are available and it will be possible to calculate the electronic reaction barrier based on the

following formula:

∆ǂ𝐸𝐸 = 𝐸𝐸TS,ads − 𝐸𝐸react,ads (3)

where ETS,ads is the electronic energy of the transition state as obtained from the dimer

calculation and Ereact,ads is the electronic energy of the adsorbed reactant(s).

3.3.4 Frequency calculations After the geometry of an adsorbate is optimized under strict convergence criteria, consecutively

a frequency calculation can be conducted. Resulted frequencies of the normal modes should all

be real in the case of adsorbates and all except one in the case of transition states. For transition

states, the normal mode corresponding to the imaginary frequency should coincide with the

direction of the reaction. The strict geometry optimizations are essential to avoid additional

imaginary frequencies, which are indications that the most stable geometry is not reached. Even

if additional imaginary frequencies are found, these can be solved by applying the dimmins.pl

script, which will be discussed in section 3.3.4.1.

In agreement with the strict geometry optimizations, largely the same computational settings

are employed for frequency calculations, unless noted differently. For the consecutive

frequency calculation, the same types of input files are used with often some alternations. The

KPOINTS and POTCAR file can be transferred from the strict geometry calculations. As

POSCAR file, the resulted CONTCAR of the strict geometry optimization has to be used as

initial starting point in the frequency calculations. The electronic settings of the INCAR are

kept the same as for the strict geometry optimization. However, the IBRION-tag is set on a


value of 5, which corresponds to frequency calculations. Additionally, the POTIM-tag is

decreased to 0.015 to ensure that only the vibrations are captured during the calculations, which

can be seen as small deviations from the most stable geometry.

The results from the frequency calculations can be used as an indication of the quality of the

found most stable geometry because the occurrence of an imaginary frequency -or multiple in

the case of transition state- points out that there is a more stable geometry in direction of the

imaginary normal mode. Moreover, the calculated normal modes can be linked with

experiments based on vibrational techniques such as reflection-absorption infrared

spectroscopy (RAIRS).

3.3.4.1 Dimmins.pl script In the case that frequency calculations resulted in an additional imaginary frequency, the

dimmins.pl script should be applied. The following procedure in cooperation with this script

will lead to the most stable geometry with all real frequencies in the case of adsorbates. First,

the movements of the imaginary frequency are copied from the OUTCAR to a new MODECAR

file. Next, the dimmins.pl script is applied together with the POSCAR-file from the frequency

calculations and the newly generated MODECAR file. Additionally, a certain distance can set

in the dimmins.pl script, which corresponds to which extent the atoms are moved along the

direction of the imaginary frequency. As a default, this distance is equal to 0.1 Å. This script

constructs directories: ‘min1’ and ‘min2’, where the adsorbate is moved over the inserted

distance according to the imaginary mode, but in opposite directions. Then, the geometries of

the ‘min1’ and ‘min2’ directories are optimized in accordance with the strict convergence

criteria as the most stable geometry of the two is employed for an additional frequency

calculation. If an imaginary frequency remains, the procedure is repeated, eventually with an

enlarged distance until the most stable geometry is effectively reached.

3.4 References

1. Friesner, R.A., Ab initio quantum chemistry: methodology and applications. Proc Natl Acad Sci U S A, 2005. 102(19): p. 6648-53.

2. Sabbe, M.K., M.-F. Reyniers, and K. Reuter, First-principles kinetic modeling in heterogeneous catalysis: an industrial perspective on best-practice, gaps and needs. Catalysis Science & Technology, 2012. 2(10): p. 2010-2024.

3. Murzin, D. and D.Y. Murzin, Nanocatalysis 2006. 2006: Research Signpost. 4. Catlow, C.R.A., et al., Computational approaches to the determination of active site

structures and reaction mechanisms in heterogeneous catalysts. Philosophical


Transactions of the Royal Society a-Mathematical Physical and Engineering Sciences, 2005. 363(1829): p. 913-936.

5. Van Santen, R.A., Complementary Structure Sensitive and Insensitive Catalytic Relationships. Accounts of Chemical Research, 2008. 42(1): p. 57-66.

6. Deutschmann, O., Modeling and simulation of heterogeneous catalytic reactions: from the molecular process to the technical system. 2013: John Wiley & Sons.

7. Valero, M.C., P. Raybaud, and P. Sautet, Interplay between molecular adsorption and metal-support interaction for small supported metal clusters: CO and C2H4 adsorption on Pd-4/gamma-Al2O3. Journal of Catalysis, 2007. 247(2): p. 339-355.

8. Marshall, S.T. and J.W. Medlin, Surface-level mechanistic studies of adsorbate–adsorbate interactions in heterogeneous catalysis by metals. Surface Science Reports, 2011. 66(5): p. 173-184.

9. Alexopoulos, K., et al., Theoretical Study of the Effect of (001) TiO2 Anatase Support on V2O5. The Journal of Physical Chemistry C, 2010. 114(7): p. 3115-3130.

10. Wanbayor, R. and V. Ruangpornvisuti, A periodic DFT study on binding of Pd, Pt and Au on the anatase TiO2 (001) surface and adsorption of CO on the TiO2 surface-supported Pd, Pt and Au. Applied Surface Science, 2012. 258(7): p. 3298-3301.

11. Stierle, A. and A.M. Molenbroek, Novel in situ probes for nanocatalysis. Mrs Bulletin, 2007. 32(12): p. 1001-1005.

12. Zhang, S., et al., In-Situ Studies of Nanocatalysis. Accounts of Chemical Research, 2013. 46(8): p. 1731-1739.

13. Amusia, M.Y., A.Z. Msezane, and V.R. Shaginyan, Density functional theory versus the Hartree-Fock method: Comparative assessment. Physica Scripta, 2003. 68(6): p. C133-C140.

14. Perdew, J.P. and K. Schmidt. Jacob's ladder of density functional approximations for the exchange-correlation energy. in AIP Conference Proceedings. 2001. Iop Institute Of Physics Publishing Ltd.

15. Perdew, J.P., et al., Prescription for the design and selection of density functional approximations: More constraint satisfaction with fewer fits. The Journal of chemical physics, 2005. 123(6): p. 062201.

16. Kohn, W. and L.J. Sham, Self-Consistent Equations Including Exchange and Correlation Effects. Physical Review, 1965. 140(4A): p. A1133-A1138.

17. Perdew, J.P., K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple. Physical Review Letters, 1996. 77(18): p. 3865-3868.

18. Dion, M., et al., Van der Waals density functional for general geometries. Physical Review Letters, 2004. 92(24): p. 4.

19. Klimes, J., D.R. Bowler, and A. Michaelides, Van der Waals density functionals applied to solids. Physical Review B, 2011. 83(19): p. 13.

20. Klimes, J., D.R. Bowler, and A. Michaelides, Chemical accuracy for the van der Waals density functional. Journal of Physics-Condensed Matter, 2010. 22(2).

21. Lee, K., et al., Higher-accuracy van der Waals density functional. Physical Review B, 2010. 82(8): p. 081101.

22. Hammer, B., L.B. Hansen, and J.K. Norskov, Improved adsorption energetics within density-functional theory using revised Perdew-Burke-Ernzerhof functionals. Physical Review B, 1999. 59(11): p. 7413-7421.

23. Wellendorff, J., et al., Density functionals for surface science: Exchange-correlation model development with Bayesian error estimation. Physical Review B, 2012. 85(23): p. 235149.


24. Tsai, Y.-L., C. Xu, and B.E. Koel, Chemisorption of ethylene, propylene and isobutylene on ordered Sn/Pt (111) surface alloys. Surface science, 1997. 385(1): p. 37-59.

25. Matos, J., H. Yildirim, and A. Kara, Insight into the Effect of Long Range Interactions for the Adsorption of Benzene on Transition Metal (110) Surfaces. The Journal of Physical Chemistry C, 2015. 119(4): p. 1886-1897.

26. Kresse, G. and J. Hafner, Ab initio molecular dynamics for liquid metals. Physical Review B, 1993. 47(1): p. 558.

27. Kresse, G. and J. Hafner, Ab initio molecular-dynamics simulation of the liquid-metal–amorphous-semiconductor transition in germanium. Physical Review B, 1994. 49(20): p. 14251.

28. Kresse, G. and J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Computational Materials Science, 1996. 6(1): p. 15-50.

29. Kresse, G. and J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical Review B, 1996. 54(16): p. 11169.

30. Sun, G.Y., et al., Performance of the Vienna ab initio simulation package (VASP) in chemical applications. Journal of Molecular Structure-Theochem, 2003. 624: p. 37-45.

31. Blöchl, P.E., Projector augmented-wave method. Physical Review B, 1994. 50(24): p. 17953.

32. Kresse G., M.M., Furthmüller J, VASP the GUIDE. 2014. 33. Monkhorst, H.J. and J.D. Pack, Special points for brillouin-zone integrations. Physical

Review B, 1976. 13(12): p. 5188-5192. 34. Saerens, S., Determination of the active phase of a Pt/Mg(Ga)AlOx catalyst of the

catalytic dehydrogenation of propane, in Department of Chemical Engineering and Technical Chemistry. 2014, University of Ghent: Ghent.


36. Sheppard, D., R. Terrell, and G. Henkelman, Optimization methods for finding minimum energy paths. Journal of Chemical Physics, 2008. 128(13): p. 10.

37. Jónsson, H., G. Mills, and K.W. Jacobsen, Nudged elastic band method for finding minimum energy paths of transitions. 1998.

38. Henkelman, G., G. Jóhannesson, and H. Jónsson, Methods for finding saddle points and minimum energy paths, in Theoretical Methods in Condensed Phase Chemistry. 2002, Springer. p. 269-302.

39. Pedersen, A., S.F. Hafstein, and H. Jónsson, Efficient Sampling of Saddle Points with the Minimum-Mode Following Method. SIAM Journal on Scientific Computing, 2011. 33(2): p. 633-652.

40. Henkelman, G. and H. Jónsson, A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. The Journal of Chemical Physics, 1999. 111(15): p. 7010-7022.

Chapter 4: Propane dehydrogenation kinetics on Pt(111) catalyst model 57

Chapter 4 Propane dehydrogenation kinetics on Pt(111) catalyst model In this chapter, ab initio calculations are employed to unravel the reaction mechanism of

propane dehydrogenation on a Pt(111) catalyst model. The most stable geometries of the

adsorbed intermediates are calculated to determine the thermodynamics. Furthermore,

transition states of the selected reaction paths are optimized to obtain information on the

kinetics. Eventually, these data are used to perform a kinetic study to obtain coverages of

adsorbed species and dominant paths. The Pt(111) catalyst model is interesting to consider as a

reference for more advanced catalyst models, which are studied in further chapters. This model

can be used as benchmark for the applied method since it can be compared with available

literature on light alkane dehydrogenation on pure Pt catalysts.

4.1 Catalyst model

The bulk phase of platinum has a face-centered cubic (fcc) structure with a coordination number

of twelve, a packing factor of 0.74 and a lattice distance of 3.92 Å. [1] The most stable surface

cut of the fcc unit cell is the (111) plane. Each atom on the (111) surface has a coordination

number of nine, so there are three broken bonds per surface atom. [2] Three broken bonds is

the minimal amount possible, making the (111) plane the most abundant surface plane in Pt

particles and the most validate candidate for this highly coordinated catalyst model.

The periodic slab approach is employed as method to model the metal surface geometries. As

discussed in section 3.1, this approach is the most suitable for metal catalysts as it can describe

the delocalized nature of the metallic bonding. At the basis of this method stands the selection

58 Chapter 4: Propane dehydrogenation kinetics on Pt(111) catalyst model

of the unit cell, of which the periodic extension describes the infinite metal catalyst surface.

The determination of the optimal size of the unit cell is essential as larger cells demand high

computational effort, while smaller unit cells can incorrectly describe the coverages of the

adsorbates.

Figure 4-1: Isometric representation of the 4×2 unit cell of the Pt(111) catalyst model

In this work, a 4×2 unit cell is used to represent the Pt(111) catalyst model. The choice of this

4×2 unit cell will be explained further in section 4.1.2. In this 4×2 unit cell, four Pt atoms are

taken in the a-direction, two Pt atoms in the b-direction and four Pt atoms in the c-direction.

The a-, b- and c-directions are the lattice vectors of the unit cell, which can be correlated with

the x-, y- and z-direction of the Cartesian coordinate system. In the c-direction, the bottom two

layers are kept fixed as they represent the bulk atoms of the catalyst model, while the upper two

layers represent the surface of the catalyst model and are allowed to relax. A vacuum layer of

12 Å and an artificial dipole layer are constructed to avoid periodic interactions between unit

cells in the c-direction. The dimensions of the unit cell are 11.29×5.64×18.91 Å³. For geometry

optimization of the unit cell, strict convergence settings are employed conform Chapter 3.3.1.

The non-local vdW-DF functional is utilized during the strict optimizations. An isometric view

is illustrated in Figure 4-1.

b

a

c


From the optimized geometry of the unit cell, a lattice parameter of 3.99 Å is determined, which

is an acceptable overestimation from the reported value of 3.92 Å. [1] In literature, catalyst

models have mostly four layers in the c-direction for similar DFT studies. [3-8] As verification

of the four-layer model, the adsorption energy of propylene, an essential intermediate during

propane dehydrogenation, is calculated with a four-layer model (bottom two are fixed) and a

six-layer model (bottom four are fixed). An energetic difference of -9 kJ/mol is observed with

respect to the four-layer model (propylene binds 9 kJ/mol stronger on the six-layered model).

This deviation (6.5%) is too great to be within a precision of 5%. However, the deviation is

acceptable for this work, as including two additional bulk layers greatly increase the

computational cost.

4.1.1 Adsorption site nomenclature

On this catalyst model, four types of adsorption sites are possible: top, fcc, hcp and bridge sites.

The top sites are defined as adsorption on a single surface atom of the catalyst. On top of these

single atom sites, the adsorbate species is placed. Bridge sites involve two adjacent surface

atoms. Between those atoms, the adsorbate species is placed. Both the fcc and hcp sites use

three adjacent atoms in the surface layer. Between these three atoms, a hollow site is present

where the adsorbate species is placed. The difference between the fcc and hcp sites is made

based on the second layer underneath the surface layer. When there are another three atoms

beneath the three surface atoms, this site is defined as an fcc site. Otherwise, when there is only

one atom beneath the three surface atoms, this site is defined as an hcp site. This various

adsorption site on the catalyst model are illustrated in Figure 4-2.

Figure 4-2: Adsorption sites on Pt(111) catalyst model: top ●, bridge █ , three-folded hcp ▲ and fcc ▼.

b

a


In more advanced catalyst models, this nomenclature for adsorption sites is too common, as

multiple sites of the same type are possible. Each site has other characteristics because the atoms

in the layers underneath differ between various sites. The identification of each site is simplified

by incorporating the atoms in the layers underneath in the site naming, until each site can be

distinguished from each other. When a more sophisticated nomenclature is needed, this will be

pointed out.

4.1.2 Determination of the degree of coverage

When catalyst models are employed, it is essential to define the coverage of the adsorbate

species on the surface. To describe the degree of coverage, equation (1) is employed. In the

case of the Pt(111) catalyst model, eight surface atoms are present in the unit cell, so when one

adsorbate species is on the surface, a coverage of 0.13 ML is obtained for that species.

θx =

# of adsorbate molecule X# of surface atoms in the unit cell

(1)

By selecting a Pt(111) 4×2 unit cell, a large flexibility in degrees of coverage is obtained.

However, it is essential to validate this selection with respect to literature. The unit cell selection

for Pt(111) is based on two main reasons: the catalyst model extension towards Pt3Ga and the

monolayer coverage of propylene, determined experimentally. In the next chapter, this catalyst

model will be extended to Pt3Ga to study the effect of gallium alloying. To do this, every fourth

Pt atom has to be replaced with a Ga atom and as periodic slab approach model is employed,

only multiples of four are possible as upper surface layer. A 2×2 or a 4×2 unit cell are the best

candidates, as the computational cost is too great with larger unit cells e.g. 6×2 unit cell. During

our calculation, it is preferable to remain under the monolayer coverage of propylene and this

influences the final decision. TPD analysis has shown that the monolayer of chemisorbed

propylene is 0.20 ML, so based on the equation above, this means one propylene molecule for

five surface atoms. [9, 10] Consequently, the 4×2 unit cell remains valid for this work.

4.1.3 Selection of the DFT functional

In literature, PBE is a widely used functional for DFT calculations. However, in this work, this

functional is solely employed to optimize the initial geometries, as it does not address non-local

interactions such as van der Waals forces. Hence, it was opted to select a vdw-DF functional,

the optPBE-vdW functional, to optimize the final geometries. This non-local vdW-DF


functional introduces long-range dispersion in the approximate exchange-correlation

functionals.

4.2 Adsorption

In this section, the adsorption of reactants and products that appear in the gasphase during

propane dehydrogenation are discussed. Logically, the main products are propane, propylene

and hydrogen. Additionally, ethane, ethylene and methane are also included as they can be

formed through cracking of adsorbed intermediates. Essentially, it is tried to correlate the results

of this work with experimental aspects of the adsorption of hydrocarbons on the surface of

platinum, preferably on Pt(111). Throughout this section, the following equation is employed

to describe the adsorption energy of each compound.

∆ads𝐸𝐸(ad) = 𝐸𝐸ad,ads − 𝐸𝐸ad,gas − 𝐸𝐸surface (2)

Vibrational spectrometry such as RAIRS is a useful tool to describe the adsorbate based on its

vibrational modes. However, it is essential to compose a nomenclature of the possible

vibrational modes. The abbreviation of a vibrational mode consists out of three parts: the type

of vibrational mode denoted by a Greek letter, the direction of the movement (asymmetrical or

symmetrical, denotes as subindex) and the moiety on which the vibrational mode acts, denoted

between brackets. The types of vibration modes are deformation (δ), rocking (ρ), scissoring (γ),

stretching (ν), twisting (τ) and wagging (ω). The subindices indicate that the vibration mode is

symmetrical (s) or asymmetrical (as).

4.2.1 Alkanes

Alkanes are chemically saturated hydrocarbons that show a low bond breaking activity even on

transition metal surfaces. In the absence of bond breaking, these molecules prefer to adsorb

with their carbon chain parallel to the surface and their heat of adsorption increases by ~9

kJ/mol per -CH2-group. Thus at low temperatures and on relatively chemically inactive low

Miller-index metal surfaces the molecules adsorb intact and in a flat-lying configuration. [11]

This phenomenon is called physisorption as the alkane keeps drifting over the Pt surface, as has

only a weak bond with surface. Molecular beams techniques are employed at low temperatures,

under the alkane desorption temperature, to obtain information on the physisorbed geometries

and adsorption energies.


4.2.1.1 Propane

The chemisorption of propane on Pt surface has been a subject of various studies, both

experimentally and theoretically. As propane is a saturated hydrocarbon with no unpaired

electrons, it will not dissociatively adsorb on the Pt surface by binding with the surface atoms.

Consequently, a low adsorption energy is expected. McMaster et al. measured the molecular

adsorption dynamics of propane on Pt(110) and Pt(111) at 95 K using molecular beams. They

conducted their studies on clean and propane covered surfaces to describe the coverage effect

on the adsorption characteristics of propane. They reported that the propane oriented itself

parallel to the metal surface in the case of a monolayer. The molecular beam experiments show

that adsorbed propane facilitates trapping of physisorbed propane on the surface as the

adsorption probability increases linearly with propane coverage. Experimentally, the saturated

monolayer coverage is 3·1014 molecules/cm² and a monolayer of Pt atoms consists of 1.5·1015

molecules/cm²; hence the adsorption energies determined at a coverage of 0.2 ML. [12, 13]

Furthermore, both Nykänen et al. [14] and Yang et al. [3] have conducted a DFT study that

included the physisorption of propane on Pt(111). The resulted adsorption energies can be found

in Table 4-1.

Table 4-1. Adsorption energies of physisorbed propane on Pt(111) surface, determined both experimentally as theoretically. Nykänen et al. employed a vdW-DF functional, while Yang et al. used of a PBE functional. This work uses an optPBE vdW-DF functional.

ΔEads

(kJ/mol) Experimental data (0.2 ML) [13, 15]

Nykänen et al. (0.13 ML) [14]


Yang et al. (0.11 ML)[3]

This work

(0.13 ML)

Pt(111) -41 ~ -44 -33 -36 -6 -43

Due the use of a GGA functional, Yang et al. underestimate the energy of propane physisorption

and evaluate it as nearly thermoneutral. Their calculations did not include the non-local vdW

interactions, which are essential to describe the physisorbed propane on the Pt(111) surface.

The results of this work and of Nykänen et al. are comparable with the experimental results.

The discrepancy between the results of Nykänen et al. and this work for the same coverage can

be contributed to slightly different functionals are employed (vdW-DF by Nykänen and optPBE

vdW-DF in this work). The geometry of the adsorbed propane is identical to the optimized

structure of gasphase propane, which suggests that the interaction between propane and surface

is non-covalent, as expected for physisorption. From the results of Nykänen, a weaker

adsorption is observed at low coverages, indicating that there is an attractive interaction

between the physisorbed propane molecules. These results confirm the molecular beams

experiments of McMaster et al., which show a higher adsorption probability at higher coverage.


Chesters et al. studied the RAIRS spectra of monolayer and multilayer propane adsorbed on

Pt(111) at 95 K. The (sub-)monolayer coverage is achieved by dosing 1 Langmuir on a clean

Pt(111) surface. [16] The frequencies they obtained of physisorbed propane are compared with

the calculated frequencies of physisorbed propane in this work. As a low propane coverage is

employed in the DFT calculations, only the monolayer frequencies of propane are evaluated,

see Table 4-2.

Table 4-2. Comparison of vibrational modes of physisorbed propane between experiments and this work. Experiments made use of monolayer coverage of propane, while in this work a coverage of 0.13 ML propane is used.

Vibrational modes νas (CH3) νas (CH3) νas (CH2) δas (CH3)

Experimental (cm-1) [16] 2949 2937 2915 1454

This work (cm-1) 2997 2985 2930 1455

Chesters et al. observes four vibrational modes: three asymmetrical stretchings and one

asymmetrical deformation of the methyl groups. While compared with the experiments the

asymmetrical stretching of both methyl and methylene group are shifted respectively with 48

cm-1 and 15 cm-1, while for the deformation of the methyl group, no shift is observed. Three

imaginary frequencies are reported for the physisorbed propane. In general, physisorbed species

have a higher degree of freedom than chemisorbed species. These imaginary frequencies are

assigned to translations and external rotation of the molecule above the surface. It is not possible

to alter these vibrational modes to become real. When kinetic parameters are determined, these

frequencies can be modified accordingly. This modification is discussed in section 4.5.1. [17]

However, at higher temperatures, dissociative chemisorption of propane into 1-propyl or 2-

propyl and hydrogen (in the case of C-H bond breaking) or methyl and ethyl (in the case of C-

C bond breaking) occurs. These reactions are respectively dehydrogenation reactions and

dissociation reaction and play a major role in propane dehydrogenation. Their description falls

under the thermodynamics section of this chapter, see section 4.3.

4.2.1.2 Ethane

Ethane is a saturated C2 hydrocarbon, which can be formed during propane dehydrogenation as

a product of C-C bond breaking. As it is similar to propane, it is expected that ethane will

physisorb on the Pt surface, conform the physisorption of propane. Only a weak bond is formed

with the surface at a higher distance from the surface than chemisorbed species. Arumainaygam

et al. investigated the dynamics of the ethane adsorption with molecular beams on clean and


ethane-covered Pt(111) at 95 K. The geometry of ethane in a monolayer, adsorbed on Pt(111),

is parallel to the metal surface. Conform to the molecular beam results on propane adsorption,

the ethane sticking probability increases with a higher ethane coverage. The adsorption energy

is experimentally determined at coverage range of 0.1 – 0.4 ML. [15, 18, 19] Various DFT

studies have been conducted on ethane dehydrogenation and hydrogenolysis, but did not

consider ethane physisorption as a reaction step prior further reactions. [20, 21] However, from

the comparison of this work with the experimental data, it can be concluded that the molecular

adsorption energy of ethane can be predicted adequately.

Table 4-3. Adsorption energies of physisorbed ethane on Pt(111) surface, determined both experimentally as theoretically. This work uses an optPBE vdW-DF functional.

ΔEads (kJ/mol) Experimental data (0.1 – 0.4 ML) [15, 18, 19]

This work

(0.13 ML) Propane physisorption -32 ~ -33 -31

Chesters et al. studied the RAIRS spectra of monolayer and multilayer ethane adsorbed on

Pt(111) at 95 K. [16] The (sub-)monolayer coverage is achieved by dosing 1 Langmuir on a

clean Pt(111) surface. The frequencies they obtained of physisorbed ethane are compared with

the calculated frequencies of physisorbed ethane in this work. As a low ethane coverage is

achieved during DFT calculations, only the monolayer frequencies of ethane are evaluated, see

Table 4-4.

Table 4-4. Comparison of vibrational modes of physisorbed ethane between experiments and this work. Experiments made use of monolayer coverage of propane, while in this work a coverage of 0.13 ML propane is used.

Vibrational modes νas (CH3) νs (CH3) δas (CH3)

Experimental (cm-1) [16] 2958 2853 1457

This work (cm-1) 2969 2921 1468

Chesters et al. reported three vibration modes: a symmetrical and asymmetrical stretching of

the methyl groups and an asymmetrical deformation of the methyl groups. A shift compared to

experiment is observed of 11 cm-1 in the case for the asymmetrical modes, while a larger shift

is found for the symmetrical stretching of 68 cm-1. Again, three imaginary frequencies are

observed during the vibrational analysis. However, the allocation of these frequencies is

discussed previously for propane.


At higher temperatures, thermal dissociation of ethane occurs, providing enough energy to

dehydrogenate to adsorbed ethyl or dissociate to two adsorbed methyl groups. The first reaction

is accounted for in the thermodynamics section of this chapter, see section 4.3.

4.2.1.3 Methane

Methane is a saturated compound that can be formed from C-C bond breaking of intermediates

during propane dehydrogenation. As it has no unpaired electrons, it is expected that methane

will physisorb on the Pt surface, conform the previous discussed physisorptions. Apart from

ethane adsorption, Arumainaygam et al. also investigated the dynamics of the methane

adsorption with molecular beams on clean Pt(111) at 100 K. From these experiments, the

adsorption energy of physisorbed methane could be determined at low coverages. [22] Further,

adsorption also determined by Cushing et al. at intermediates coverages in the range of 0.1 –

0.4 ML with respect to the surface atoms of clean Pt(111). DFT studies have been conducted

by Qi et al. [23] and Zhang et al. [24] on the adsorption and dissociation of methane on Pt

surfaces, which included the physisorption of methane, prior to the C-H bond breaking.

Table 4-5. Adsorption energies of physisorbed methane on Pt(111) surface, determined both experimentally as theoretically. Qi et al. employed a PBE functional, while Zhang et al. used a PW91 functional. This work uses an optPBE vdW-DF functional.

ΔEads (kJ/mol) Experimental data (0.1 – 0.4 ML) [15]

Qi et al. (0.25 ML) [23]

Zhang et al. (0.25 ML) [24]

This work

(0.13 ML) Pt(111) -16 -3 -4 -20

As both DFT studies do not account for van der Waals interactions, the physisorption energy

of methane is underestimated. In contrast, the result of this work has the same order of

magnitude as the experimental results, but still overestimates the methane adsorption energy up

to 5 kJ/mol. Physisorbed methane has three imaginary frequencies. However, the allocation of

these frequencies is discussed previously for propane and ethane.

Fuhrman et al. have shown that methane dissociates thermally at 120 K, when molecular

beaming methane at kinetic energies between 30 and 80 kJ/mol. Hence, methane adsorbs on

the Pt(111) surface as a methyl and a hydrogen. At temperatures of 300 K, the adsorption of

methane is observed to lead directly to deeply dehydrogenated methylidyne (CH) as surface

species. [25, 26] The first dehydrogenation step of methane will be discussed further in the

thermodynamics chapter, see section 4.3.


In general, it is proposed that the heat of adsorption of alkanes increases by ~9 kJ/mol per -CH2-

group. [11] The alkane adsorption energies, calculated based of DFT theory, increase by 11

kJ/mol, which is in accordance with the experimental observation.

4.2.2 Alkenes

Because of the availability of π-electrons that can readily be donated to the metal, the bonding

of alkenes to the metal surface is much stronger than the bonding of alkanes. Furthermore, the

absorbed alkene undergoes molecular rearrangements depending on temperature and the metal

surface structure. Additionally, the metal substrate may restructure under the influence of

alkene adsorption to optimize bonding to the adsorbed species. For propane dehydrogenation,

propylene and ethylene are considered as alkene compounds on the surface.

4.2.2.1 Propylene

Adsorbed propylene plays a major role in propane dehydrogenation. When adsorbed on the

surface, several reaction ways are possible: hydrogenation to propane, desorption to gaseous

propylene as desired product and dehydrogenation to deep dehydrogenated species such as

propenyl. To promote propane dehydrogenation, the desorption of propylene should be favored

compared to the side reactions, and hence the binding of propylene with the surface should be

weakened.

Zaera et al. characterized the adsorption of propylene on Pt(111) single-crystal surfaces by

infrared spectroscopy (RAIRS). The uptake of propylene was studied at 90 K to avoid further

thermal decomposition. As function of the coverage, four adsorption modes of propylene are

observed.

Figure 4-3: Adsorption modes of propylene on Pt(111): propylene di-σ V-shape (a), propylene di-σ (b) and propylene π (c) White, H, grey, C and light blue, Pt atoms.

At the lowest propylene dosages, a di-σ bonded adsorbate with a V-shape is seen, as the central

carbon atom is preferentially bonded with the platinum surface. Above half saturation of the


monolayer, propylene rearranges itself toward a more horizontal double bond and vertical

methyl group. This orientation leads to better packing of the propylene molecules on the

surface. Above the monolayer saturation, a second layer of weakly π-bonded propylene is

observed. In this layer, propylene orients itself parallel to the surface and interacts weakly with

the surface via π-bonding. At high propylene (> 2.0 L exposure), a solid film with randomly

oriented propylene is formed. [27, 28] In this work, di-σ adsorbed propylene is studied.

Valcarcel et al. performed DFT calculations on propylene dehydrogenation and isomerization.

A part of the study was to conduct a frequency analysis on di-σ adsorbed propylene and deep

dehydrogenated species to verify the origin of certain surface species during propylene

adsorption. They performed their analysis with PW91 as functional. [7, 8] To validate the

geometry of di-σ adsorbed propylene, the calculated frequencies are compared with the RAIRS

spectra, conducted at 90 K by Zaera et al. and the frequency analysis of Valcarcel et al. in

Table 4-6 .

Table 4-6. Comparison of propylene vibration modes between RAIRS and DFT based frequency analysis of the di-σ adsorption mode. Valcarcel et al. used a PW91 for the di-σ propylene optimization, while this work uses an optPBE vdW-DF functional.

Vibrational mode (cm-1)

Zaera et al. [27] Valcarcel et al. [7] This work

0.4 – 0.8 L exposure 0.25 ML 0.13 ML

ρ(CH3) 1015 1007 1017

τ(CH2) 1037 1030 1036

ν(C-CH3) 1088 1092 1081

ω(CH2) 1260 1161 1162

δ(CH) 1309 1296 1305

δs(CH3) 1375 1337 1358

γ(CH2) 1437 1400 1408

νs(CH3) - - 2902

νs(CH2) 2830 2982 2951

2δas(CH3),

νs(CH3) 2860 2933 -

ν(CH) 2883 2986 2972

νas(CH2) 2915 3013 3022

The di-σ adsorption mode of propylene appears on the platinum surface at low exposures, so

the observed RAIRS spectrum at low exposure is compared with the theoretical calculated

frequency analysis of the di-σ adsorption mode of Valcarcel et al. and this work. The vibrational

modes of both theoretical models are similar and each can be linked to identical modes. In the


theoretical frequency analysis, all possible vibrational modes are calculated in contrast to the

RAIRS spectra where only the strongest and most IR-active frequencies are observed. Still,

almost all vibrational modes can be coupled with those based on DFT calculations. Some

discrepancy arises in the 2800-3100 cm-1 range as the C-H stretching modes are strongly

interfered by anharmonicity. Moreover, the experimental coverage and periodicity has a strong

influence on the orientation of the methyl group and its environment. These factors cause the

change in the frequencies. [27] In the di-σ configuration, the double C=C is stretched to 1.5 Å

with respect to the C=C bond length of 1.34 Å of gasphase propylene. This elongated bonding

is similar to the bonding length of gasphase propane, indicating the rehybridisation of these two

C atoms. Another indicator for this phenomenon is the loss of planarity; the methyl group

orientates itself away from the surface and the angle between the planar double bond plane and

the methyl group is ~21°. The rehybridisation of double bonds C atoms points out the formation

of covalent bonds with the platinum surface.

Various DFT studies have been conducted on the adsorption of propylene, especially on the di-

σ mode. All calculated adsorption energies are tabulated in Table 4-7 and when compared to

the experimental values, a clear overestimation is observed, especially for this work. The

explanation is two-fold. Firstly, the selected vdW-DF functional has a tendency of overbinding

adsorbed species on the surface, overestimating their adsorption energies. Compared to other

functionals, e.g. PW91 and PBE, vdw-DF overbinds propylene. Secondly, Zaera et al. proposed

that the adsorption energy based on TPD experimental data may be influenced by thermal

decomposition of propylene before desorption at 230-250 K. The hydrogen, which is released

due the thermal decomposition, remains on the surface and weakens the propylene binding with

the surface, leading to the observation of a lower adsorption energy. [28] Furthermore, also a

difference is noted between the work of Nykänen et al. and this work, even though both use a

vdW-DF functional. However, Nykänen et al. used other computational settings for his

functional. Nykänen et al. observed that the adsorption energy increases at higher coverage,

indicating the occurrence of small repulsion between adsorbed propylene. [14]

Table 4-7. Adsorption energy of di-σ adsorbed propylene, determined both experimentally and theoretically. Valcarcel et al., Yang et al. and Nykänen et al. used respectively a PW91, a PBE and a vdW-DF functional. This work uses an optPBE vdW-DF functional.

ΔEads

(kJ/mol) Experimental data (0.2 ML) [29, 30]

Valcarcel et al. (0.13 ML) [8]

Yang et al. (0.25 ML)[3]


This work (0.13 ML)

Pt(111) -51 ~ -68 −87 -90 -80 -134


Furthermore, the physisorption of propylene is considered on the Pt(111) surface. This

adsorption mode can act as a precursor state prior to the di-σ adsorption mode. In contrast with

di-σ propylene, the double bond is not lost during this mode. The observed C=C length is 1.34

Å, which is identical to that of gasphase propylene. The geometry of this mode is differently as

the planarity remains and the double bond length is the same as in gasphase propylene.

However, the geometry does not orientate itself parallel, but with an angle of 42° between the

surface and propylene due the hindrance of the methyl group. In this work, the optimized

geometry of physisorbed has a remaining root mean square (RMS) of 0.026 with respect to the

imposed strict ionic convergence criterion of 0.015. However, the resulted RMS is sufficient

low for this work. Furthermore, three imaginary frequencies are obtained for the physisorbed

propylene. Still, these can be treated accordingly for the determination of kinetic parameters.

4.2.2.2 Ethylene

As C-C cleavage reactions occur on the catalyst surface, ethylene is one of the most likely side

product to be formed. Furthermore, adsorbed ethylene is a central component to describe the

hydrogenation of ethylene and dehydrogenation of ethane and various DFT and experimental

studies have dedicated to it. [20, 29-33] Two distinct adsorption modes of ethylene are

experimentally observed: di-σ and π-adsorption mode. The majority of ethylene adsorbs in the

di-σ mode, as it is more stable than the π mode. High-resolution electron energy loss

spectroscopy (HREELS) analyses have shown that the di-σ mode solely occurs at low coverage,

while the minor adsorption mode π-ethylene generally is observed at higher coverages.

TPD experiments around 100 K have shown that for low coverage the adsorption energy of di-

σ ethylene is around -70 kJ/mol, however all of the DFT studies overestimate this value, even

beneath ethylene monolayer coverage of 0.25. Discrepancies can occur when comparing these

values as energies deduced from TPD correspond to a kinetically controlled out-of-equilibrium

process, which does not necessarily yield a correct adsorption energy, whereas DFT reports

adsorption energies based on electronic energies at zero K. It is proposed in literature that at

100 K ethylene already dehydrogenates to more stable species, clouding the detected adsorption

energy. [32]

In this work, di-σ adsorbed ethylene is studied, as it is generally more stable on the Pt (111)

surface. In this mode, the C=C bond of ethylene is elongated, conform propylene. The observed

C=C bond length is 1.5 Å, which is more similar to the length of single bonded C atoms (1.53


Å) than double bonded C atoms (1.34 Å). Furthermore, rehybridisation of the C atoms occurs

as covalent bonds are formed with the surface.

Table 4-8. Adsorption energy of di-σ adsorbed ethylene determined both experimentally and theoretically. Essen et al. employed a PW91 functional, while Watwe et al. used a RPBE functional. This work employs an optPBE vdW-DF functional.

ΔEads


(0.2- 0.25 ML)[9, 32, 33] Watwe et al.

(0.25 ML) [21] Essen et al.

(0.11 ML) [32] This work

(0.13 ML) Pt(111) -71 ~ -73 -117 -100 -131

Apart of the di-σ adsorbed ethylene, the physisorbed ethylene is studied. During physisorption,

ethylene retains its double bond as no elongation with respect to gaseous ethylene is observed.

However, it should be noted that physisorbed ethylene is not converged under the strict ionic

criterion. The remaining RMS is 0.026 and as this mode is metastable, four imaginary vibration

modes are obtained during the frequency analysis. The lowest frequency (7 cm-1) is attributed

to lattice vibration, while the upper three are allocated to translation and external rotation of

ethylene. These frequencies are treated prior the kinetic simulation.


4.3 Thermodynamics

In this work, the following intermediates are included to describe the reaction network of

propane dehydrogenation on the Pt(111) surface: C3Hx (x=5-8), C2Hy (y=3-6), CHz (z=1-4) and

atomic hydrogen. For each component, the most stable geometry on the Pt(111) catalyst model

is calculated and a frequency analysis is conducted to evaluate its stability. The considered

species are listed in Table 4-9 and a streamlined version of the proposed reaction network is

shown in Figure 4-4. The optimal geometries can be found in Appendix B, together with a

visualized reaction network based on those optimal geometries.

Table 4-9. Adsorbed hydrocarbon species on the Pt(111) surface. * indicates with how many C-Pt bonds the species is adsorbed to the surface.

C3Hx C2Hy CHz

CH3-CH2-CH3 Propane,phys CH3-CH3 Ethane,phys CH4 Methane,phys

CH3-CH2-CH2* 1-Propyl CH3-CH2

* Ethyl CH3* Methyl

CH3-CH*-CH3 2-Propyl CH2*-CH2

* Ethylene CH2* Methylidene

CH3-CH*-CH2* Propylene CH2=CH2 Ethylene,phys CH2

* Methylidyne

CH3-CH=CH2 Propylene,phys CH3-CH** Ethylidene

CH3-CH2-CH** 1-Propylidene CH3-C*** Ethylidyne

CH3-CH**-CH2 2-Propylidene

CH3-CH2-C*** 1-Propylidyne

CH3-CH*-CH** 1-Propenyl

CH3-C**-CH2* 2-Propenyl


Figure 4-4: Reaction netw

ork for propane dehydrogenation on Pt(111). T

he consecutive reactions of the methyl and ethyl species are given in Figure 4-6

2-propyl

1-propylidene

10

CH

2 =CH

-CH

3,phys

CH

2 =CH

-CH

3 (g)

13

CH

2 + CH

-CH

3

=Pt2

=Pt2_

CH

-CH

-CH

3+ HPt

Pt2

_Pt

=

Methyleneand ethylidene

CH

2 -C-C

H3

+ HPt _

Pt _

Pt2

=2-propenyl

CH

3 -C + C

H3

Pt3

≡

Pt _

CH

+ CH

2 -CH

3

Pt3Pt

≡

_

1-propenyl

C-C

H2 -C

H3

+ H

Pt3

_

Pt

≡

1415

1617

1819

2021

Methyl and ethyl

CH

2 + CH

2 -CH

3

Pt2Pt

=

_

45

6

=

_

CH

-CH

2 -CH

3+ H

Pt2Pt

CH

3 -CH

+ CH

3

Pt2Pt

=

_

2-propylidene

CH

3 -C-C

H3

+ H

Pt2

=

Pt _

78

9

CH

3 + CH

2 CH

3

PtPt

_

_C

H3 -C

H-C

H3

+ H

PtPt

_

_

CH

2 -CH

2 -CH

3 + H

PtPt

_

_

1-propyl

12

3

CH

3 -CH

2 -CH

3,phys

CH

3 -CH

2 -CH

3 (g)Gaseous

propane

0

Physisorbed propane

Ethylideneand m

ethyl

CH

2 -CH

-CH

3+ HPt _

_Pt

_PtPropylene

1112

Methylidene

and ethyl

Methylidene

and ethyl1-propylidyne


Gaseouspropylene

Ethylidyneand m

ethyl


The reaction network will be divided in three sections: propane dehydrogenation to propylene

(4.3.1), deep dehydrogenation of propylene (4.3.2) and hydrogenolysis of C3 intermediates

(4.3.3). In each section, reaction energies are determined and the stability of each intermediate

is evaluated. The first section can be interpreted to quantify the activation of the catalyst, while

section two and three quantify the selectivity towards propylene with respect to deep

dehydrogenation reactions and cracking reactions that can lead to cokes formation. In the

second section, the relative energy of adsorbed propylene and deep dehydrogenated species are

compared, as the latter species are coke precursors and indicate the catalyst sensitivity to

deactivation. In the third section, hydrogenolysis or dissociation of the C-C bond are discussed.

These reactions lead to smaller hydrocarbons on the surface and eventually gaseous side

products such as methane, ethane and ethylene. Their formation indicates a decrease of

propylene selectivity as propane is converted to other gaseous species.

Three main reaction types can be distinguished: dehydrogenation, isomerization and

hydrogenolysis. In the first type, the breaking of C-H plays a major role, while in the last, the

breaking of a C-C bond is considered. The location of the fragment(s) is essential to describe

their reaction energy. In the case of dehydrogenation, this fragment is always an adsorbed

hydrogen. In this work, it is assumed that the fragmented hydrogen diffuses away from the

adsorbate due to repulsion between hydrogen and adsorbate. This assumption is valid because

the calculations are done at low H coverages, and at propane dehydrogenation temperatures of

873 K, hydrogen is mobile on the surface. [34] Furthermore, this validation can be extended to

dissociation reactions, as there is also repulsion observed between the C1 and C2 fragments. In

addition, the product C-fragments will be calculated in a separate unit cell. In this work, this

assumption is implemented as follows. The product adsorbate and the atomic hydrogen are both

optimized in separate unit cell and equation (3) is used to calculate the reaction energy for

dehydrogenation reactions, while for the hydrogenolysis reactions equation (4) is employed and

both hydrocarbon fragments are optimized in a separate unit cell. For other reactions such as

isomerization and physisorption, equation (5) is used as the reaction is treated in one unit cell.

ΔE𝑟𝑟 = 𝐸𝐸C3𝐻𝐻𝑥𝑥−1/surface + 𝐸𝐸H/surface − 𝐸𝐸C3𝐻𝐻𝑥𝑥/surface − 𝐸𝐸surface (3)

ΔE𝑟𝑟 = 𝐸𝐸C2𝐻𝐻𝑦𝑦/surface + 𝐸𝐸C1𝐻𝐻𝑧𝑧/surface − 𝐸𝐸C3𝐻𝐻𝑦𝑦+𝑧𝑧/surface − 𝐸𝐸surface (4)

ΔE𝑟𝑟 = 𝐸𝐸𝑝𝑝𝑟𝑟𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝/surface − 𝐸𝐸reactants/surface (5)

As for each dehydrogenation step, the hydrogen is optimized in separate unit cell, it is essential

to find its most stable adsorption site on the surface. As hydrogen adsorption is dissociative


adsorption, a small adaptation is made to equation (2) to ensure that the conservation laws are

respected. Equation (6) is applied as follows and can be extended to other dissociative

adsorption reactions.

∆ads𝐸𝐸(H2) = 2𝐸𝐸H,ads − 𝐸𝐸ad,gas − 2𝐸𝐸surface (6)

Various researchers have studied the adsorption of hydrogen gas on Pt(111) surfaces with either

molecular beams or TPD analysis. They report energies all in the same order of magnitude;

however, the adsorption energy of hydrogen depends strongly on the hydrogen coverage on the

surface. The hydrogen adsorption strength decreases for increasing hydrogen coverage on

Pt(111). [35-37] The experimental allocation of the most stable adsorption site of hydrogen is

difficult as different sources contradict each other. This discrepancy can be explained based on

the employment of the harmonic oscillator to allocate the vibration modes during vibration

analysis and to whether or not the zero point energy (ZPE) effects are included. Bădescu et al.

performed a combined experiment and theory study on hydrogen adsorption at both low and

high hydrogen coverage. At low coverage (θ < 0.7), the three-folded fcc site is most stable

according to their theoretical calculations, while the top site is 2.5 kJ/mol less stable. [38] At

monolayer coverage, top site becomes more stable than the fcc site, but only if the zero point

energy correction is excluded. [39] Over the complete coverage range, the hcp site is less stable

than the other adsorption sites.

Yang et al. determined hydrogen adsorption energies on a 2×2 unit cell with PBE as functional.

However, the adsorption energy was calculated with respect to atomic hydrogen in the

gasphase. [3] In Table 4-10, the adsorption energies are listed; the results of this work are

modified based on the formula used by Yang et al. to compare with their results (see right

column Table 10). The adsorption energy of each site falls into the experimental range of

values. However, the most stable adsorption site is top site at low coverage (θ =0.13), while

Yang et al. and Bădescu et al. both determined the hollow fcc site as most stable under these

conditions. When comparing both DFT studies, it can be observed that the vdW-DF functional

stabilizes the three-folded adsorption sites with ~89 kJ/mol and the top site with ~97 kJ/mol,

which can be explained based on van der Waals interactions. For further calculations, the

adsorption energy of hydrogen on the top site is employed to determine the reaction energies

of dehydrogenation reactions, as formulated in equation (3).


Table 4-10. Hydrogen adsorption energies on Pt(111), determined experimentally and theoretically. Yang et al. used a PBE functional while this work uses an optPBE vdW-DF functional.

ΔEads


(0 – 0.55 ML) [35-37] This worka

(0.13 ML) Yang et al.b

(0.25 ML) [3] This workb

(0.13 ML)

On top -73 ~ -80 -77 -257 -354

Fcc -73 ~ -80 -71 -263 -351

Hcp -59 -256 -346 (a) Energies calculated according to equation (6). (b) Adsorption energy is determined with respect to atomic hydrogen in the

gasphase.

4.3.1 Propane dehydrogenation to propylene

The dehydrogenation of propane to propylene consists at least of two elementary steps on the

platinum surface. Additionally, the physisorption of propane is included prior to the propane

activation on the surface. This initial activation of propane is the first elementary step and

dehydrogenates at either the methyl or methylene group, respectively generating 1-propyl or 2-

propyl. The most stable adsorption of these species is located on top site of a platinum atom.

Consequently, these species form propylene via β-dehydrogenation. Propylene orientates itself

most stable over a bridge site. Each step generates also a detached hydrogen. Its location plays

a major to determine the kinetics as the activation energy is strongly influenced by the (de-

)stabilization due to this hydrogen. However, for the thermodynamics, it is assumed that

hydrogen is highly mobile on the surface and diffuses away from the product hydrocarbon.

Both 1-propyl and 2-propyl can also form respectively 1-propylidene and 2-propylidene via α-

dehydrogenation. An additional hydrogen is detached from the carbon atom that is already

bonded with the surface. The bridge adsorption site is the most stable for both these species.

Apart from dehydrogenation reactions, isomerization reactions are also included between 1-

propyl and 2-propyl, 1-propylidene and propylene and 2-propylidene and propylene. The

reaction energies determined in this work are compared with the DFT studies of Yang et al. [4]

and Valcarcel et al. [7] Yang et al. report all reaction energies studied in this work, while

Valcarcel investigated all reactions originating from propylene. Here, the formation of 1-propyl

and 2-propyl are treated as hydrogenation reactions with respect to propylene, but by altering

the sign of the reaction energy, they can be treated as the reaction energies of the respective

dehydrogenation reactions.


Table 4-11. Comparison between reaction energies for propane dehydrogenation towards propylene of different DFT studies. Yang et al. [3] used a PBE functional while Valcarcel et al. [7] employed a PW91 functional. This work uses an optPBE vdW-DF functional.

ΔEr (kJ/mol) Yang et al.

(0.25 ML)

Valcarcel et

al. (0.25 ML)

This work

(0.13 ML) # Surface reaction

0 Propane(g) → Propane,phys Physisorption -2 - -43

1 Propane,phys → 1-propyl + H Adsorption -7 - -7

2 Propane,phys → 2-propyl + H Adsorption -6 - -14

4 1-propyl → 1-propylidene + H Dehydrogenation 3 1 7

5 1-propyl → propylene + H Dehydrogenation -22 -18 -22

7 2-propyl → propylene + H Dehydrogenation -23 -31 -15

8 2-propyl → 2-propylidene + H Dehydrogenation 7 - 15

10 1-propyl → 2-propyl Isomerization 1 13 -7

11 1-propylidene → propylene Isomerization -25 -19 -29

12 2-propylidene → propylene Isomerization -30 - -30

13a Propylene→ propylene,phys Desorption 90† 87† 92

13b Propylene,phys →Propylene (g) Desorption - - 43 † Reaction energy is determined with respect to propylene in the gasphase, as physisorbed propylene is not studied.

Aside from the adsorption of propane and the desorption of propylene, it can be observed that

the order of magnitude of the results of the DFT studies are similar. However, because van der

Waals interactions are incorporated in our calculations, some discrepancies are noticed with the

literature. The discussion of the difference in propane adsorption and propylene desorption can

be found in 4.2.2.1 and 4.2.1.1 respectively.

In the first elementary step, Yang et al. observe 1-propyl as more stable intermediate compared

to 2-propyl with an energetic difference of one kJ/mol. However, in this work, 2-propyl is more

stabilized than 1-propyl and a larger energy difference of 7 kJ/mol beneficial for 2-propyl is

reported. In the case of 1-propyl, Yang et al. report that the ethyl moiety orientates itself

perpendicular with respect to the surface, while d(Pt-C) is 2.1 Å and the d(C-C) is 1.52 Å and

1.54 Å. [3] Compared with this work, this geometry of 1-propyl is identical. However, an

additional geometry with the ethyl moiety parallel to the surface is calculated, see Figure 4-5.

The resulted electronic energy is a fraction lower (~2 kJ/mol). However, the electronic energy

only takes into account enthalpic contributions, so to allocate the most stable geometry of 1-

propyl, it is essential to evaluate the entropic factors.


Figure 4-5: Isometric representation of two possible ethyl moiety orientation of 1-propyl: ethyl moiety perpendicular to the Pt surface (left) and ethyl moiety parallel to the surface (right). Following color code us used: H (white), C (gray) and Pt (blue).

Hence, the Gibbs free energy is determined based on the vibrational analysis of both geometries

at 873 K. This calculation is based on the partition functions. Additionally, the imaginary

frequencies and frequencies assigned to lattice movement are neglected to determine the Gibbs

free energy. Based on those results, the geometry with ethyl moiety perpendicular to the surface

(Figure 4-5, left) is ~2 kJ/mol more stable than with ethyl moiety parallel to the surface

(Figure 4-5, right). Hence, the geometry with ethyl moiety perpendicular to the surface is

employed in further calculations. This geometry is stabilized due to entropic contributions, as

the ethyl moiety is farther away for the surface. Still, as the Gibbs free energy difference is so

small, both species are expected to from on the surface, certainly at low coverages. At higher

coverage, the 1-propyl with the upwards ethyl group is more favorable as it blocks fewer

adsorption sites.

The geometries of 2-propyl in this work and in the work of Yang are almost identical, only the

distance between Pt and the bonded C is slightly shorter in the case of this work (1.52 Å) with

respect to Yang et al. (1.54 Å). The small geometric differences can be explained based on van

der Waals interactions. An energetic difference of -8 kJ/mol in favor of this work was found.

This stabilization can be explained based on the lower coverage (0.13 ML) employed in this

work, which leads to less interaction between adsorbates. It is expected that this species will be

hindered at higher coverage as the two methyl groups of 2-propyl partially block two nearby

atoms.

In this work, 2-propyl is more stable than 1-propyl based on their electronic energy. However,

this energy calculation neglects zero point energy (ZPE) and entropic contributions, so for the

most stable geometries of 1-propyl and 2-propyl, a frequency analysis is conducted to determine


the Gibbs free energy at 873 K. From this calculation, it can be concluded that 1-propyl is more

stable than 2-propyl with a difference of -9 kJ/mol. The entropic effects of the ethyl moiety of

1-propyl cause additional stabilization with respect to the two methyl groups of 2-propyl.

However, as the difference in Gibbs free energy is still quite low, both pathways will be

important at propane dehydrogenation temperatures.

For the second elementary step, the β-dehydrogenation is most favorable, but has to compete

with α-dehydrogenation, that occurs simultaneously. Both literature and this work reports that

α-dehydrogenation generate less stable species than β-dehydrogenation. Propylene is the most

stable C3H6-species on the surface, with 1-propylidene as second and 2-propylidene as third.

When comparing the stable geometries in this work with those of Yang et al, it can be concluded

that propylene is identical, while 1-propylidene and 2-propylidene slightly differ at the upper

C-C bonds. For 1-propylidene, the upper C-C bond of the ethyl moiety is elongated from 1.53

Å to 1.54 Å, while both C-C bonds of 2-propylidene are stretched from 1.52 Å to 1.53 Å. This

small elongation is contributed to the van der Waals interactions between the intermediates and

the surface. As with 1-propyl, the location of the ethyl moiety of 1-propylidene is important to

determine the global minimum of this species. Hence, two configurations are optimized: with

the ethyl moiety perpendicular to the surface (same as Yang et al.) and parallel to the surface.

The resulted electronic energy differs slightly in favor of the first geometry. However, a

conclusion is made based on the Gibbs free energy calculation at 873 K. The most stable

geometry remains the first with less than 1 kJ/mol difference. Therefore, it is expected that both

species will be present on the surface at low coverage, where the first geometry is preferential

at higher coverage due to the direction of the ethyl moiety.

The propane dehydrogenation to propylene can be treated as a catalytic cycle via two pathways:

via 1-propyl (reactions: 0 – 1 – 5 – 13a – 13b) and 2-propyl (reactions: 0 – 2 – 7 – 13a – 13b).

Additionally, hydrogen desorption is included to complete these loops. Both reaction paths are

compared with the gasphase reaction based on DFT study and microcalorimetry experiments,

see Table 4-12.

The global reaction energy of each path is identical with a value of 139 kJ/mol. To validate this

energy, it has to be compared with experimental microcalorimetry data. In literature, no

information is found on the microcalorimetry of propane dehydrogenation. However, the

reaction energy can be interpolated from microcalorimetry data, obtained for ethane and

isobutene dehydrogenation. The reaction energies are respectively 157 kJ/mol [40, 41] and 110


kJ/mol [42], which means an interpolated value of 134 kJ/mol is obtained for propane

dehydrogenation. This estimate is in accordance with the calculated values of reaction energies

in this work. Based on the thermodynamical data, both reaction pathways will occur on the

surface.

Table 4-12. Comparison of DFT calculated reaction energies (kJ/mol) to form propylene via two reaction paths on the Pt(111) surface and the gasphase reaction.

Elementary step Via 1-propyl Via 2-propyl Gasphase

Propane(g) → Propane,phys -43 -43 -

Propane,phys → 1-propyl + H -7 - -

Propane,phys → 2-propyl + H - -14 -

H → ½ H2(g) 38 38 -

1-propyl → propylene + H -22 - -

2-propyl → propylene + H - -15 -

H → ½ H2(g) 38 38 -

Propylene→ propylene,phys 92 92 -

Propylene,phys → Propylene (g) 43 43 -

Propane(g) → Propylene (g) + H2(g) 139 139 139

4.3.2 Deep dehydrogenation of propylene and other C3H6 species

The deep dehydrogenation, especially the dehydrogenation of propylene, is essential to describe

the selectivity towards coke precursors. In this work, three deeply dehydrogenated species are

investigated: 1-propylidyne, 1-propenyl and 2-propenyl. 1-propylidyne is formed via α-

dehydrogenation of 1-propylidene, while β-dehydrogenation of 1-propylidene leads to the

formation of 1-propenyl. Propylene forms 1-propenyl and 2-propenyl, respectively via

dehydrogenation of the methylene and methylidyne groups. 2-propenyl is also formed via β-

dehydrogenation of 2-propylidene. 1-propylidyne prefers adsorption on a three-folded fcc site

and orientates its ethyl moiety perpendicular to the surface. The adsorption orientation of 1-

propenyl and 2-propenyl are similar, the carbon atoms that misses two hydrogen adsorbs on a

bridge site, while the carbon with a single-detached hydrogen prefers the top position. It can be

observed that all adsorbed hydrocarbons on Pt(111) are sp3-bound.

In Table 4-13, the reaction energies determined in this work are compared with those of DFT

studies performed by Yang et al. [4] and Valcarcel et al. [7]. While the same trend appears for

all three DFT studies, it can be observed that the reaction energies of this work shift with ~10


kJ/mol with respect to Yang et al. Van der Waals interactions cause the deeply dehydrogenated

species to be less stable on the surface. This work proposes 1-propylidyne as the

thermodynamic most stable species on the surface, which is confirmed by literature.

Table 4-13. Comparison between reaction energies for deep dehydrogenation of propylene on Pt(111) of different DFT studies. All considered reactions are dehydrogenation reactions. The employed functionals are the same as in Table 4-11.


(0.25 ML) [4]

Valcarcel et al.

(0.25 ML) [7]

This work


14 Propylene → 1-propenyl + H 6 -6 20

15 Propylene → 2-propenyl + H -1 -2 7

17 1-propylidene → 1-propylidyne + H -76 -77 -67

20 1-propylidene → 1-propenyl + H -18 -25 -10

21 2-propylidene → 2-propenyl + H -32 - -24

Zaera et al. performed infrared analyses on propylene adsorption at higher temperature than 95

K. At 275 K, the RAIRS spectra show a new species on the surface, which is identified as 1-

propylidyne through vibrational analysis. This thermal conversion takes place between 230 and

270 K, indicating that this species in the most stable on surface. However, propylene cannot

directly react to 1-propylidyne. The shortest path is a mechanism of two elementary steps. The

best guess is an initial isomerization (1,2 hydrogen shift) from propylene to 1-propylidene,

followed by α-dehydrogenation to 1-propylidyne. Still, no direct experimental evidence is

found of this mechanism, so other two steps or even three steps mechanism are possible. [27]

In previous DFT studies, even further dehydrogenated species are determined on the catalyst

surface: 1-propenylidene, propyne and propynyl. [4, 7] Both 1-propenylidene and propyne have

four bounds with the Pt surface. 1-propenylidene is formed via dehydrogenation of 1-propenyl

or 1-propylidyne and its outer carbon atoms adsorbs on a fcc site, while the center carbon atom

prefers the top site. Propyne has two carbon atoms that are double bonded on the Pt surface and

they both prefer the bridge site. Theoretically, it is also possible that this intermediate desorbs

as gaseous propyne. Both four-bonded species can dehydrogenate to form propynyl. This

species has five bonds with surface and orientates itself in a fcc site (three-bonded C atom) and

a bridge site (two-bonded C atom). Their reaction energies can be found in Table 4-14 and are

subsequently discussed.


Table 4-14. Additional reactions of deep dehydrogenation of propylene and their reaction energies, reported in literature. All considered reactions are dehydrogenation reactions Yang et al. used a PBE functional while Valcarcel employed a PW91 functional.


(0.25 ML)[4]

Valcarcel et al. (0.25 ML)[7] Surface reaction

1-propylidyne → 1-propenylidene + H 25 22

1-propenyl → 1-propenylidene + H -32 -30

1-propenyl → propyne + H -14 -11

2-propenyl → propyne + H -7 -25

1-propenylidene → propynyl + H 72 83

Propyne → propynyl + H 54 64

Dehydrogenation of 1-propylidyne forms a less stable species so, this adsorbate will probably

not be formed. While 1-propenyl and 2-propenyl both form more stable species, themselves are

not favored in the dehydrogenation process. Eventually those four-bonded species can form

propynyl, which is the most unstable species on the surface as both reaction energies are

strongly endothermic. These deeply dehydrogenated species are expected to be less involved in

the overall dehydrogenation reaction as 1-propylidyne is the most stable C3-species on the

surface. For this reason, these intermediates will not be added to the reaction network in this

work.

4.3.3 Hydrogenolysis of C3 intermediates

The scission reactions of C-C, C=C and C≡C bonds in the gasphase are unfavorable as the bond

energies are respectively 350, 620 and 835 kJ/mol at 298 K. [43] However, literature reports

that the metal surface weakens the C-C bond of hydrocarbon intermediates on the surface and

scission reactions occur more easily. [44, 45] Via these reactions, smaller adsorbates (C1 and

C2 hydrocarbons) are formed. They either hydrogenate and desorb from the surface as alkanes

such as methane and ethane or dehydrogenate further, leading to coke precursors and adsorbed

carbon. The inhibition of these steps can greatly improve the selectivity to propylene, as no

other gaseous hydrocarbon species can be formed. Experiments have shown that during propane

dehydrogenation on Pt catalysts, before steady state, the selectivity towards propylene is low

(< 80 %) and even at steady state the propylene selectivity is only 87 %. Experimentally, the


selectivity is determined with respect the other gaseous hydrocarbon products. During the

complete time on stream, hydrogenolysis of C3 intermediates occurs on pure Pt catalysts. [46]

In this work, the scission reaction of each intermediate up to the C3H6-species is included in the

reaction mechanism. The reaction energies are compared with those of Yang et al. [3] in

Table 4-15. Yang et al. did not report how the reaction energies of the dissociation reaction are

determined. It is assumed that the dissociated species are optimized in separate unit cells as

done in this work.

Table 4-15. Comparison of reaction energies for hydrogenolysis during propane dehydrogenation on Pt(111) between different DFT studies. Yang et al. employ a PBE functional, while this work uses an optPBE vdW-DF functional.


(0.25 ML) [3]

This work


3 Propane,phys→ methyl + ethyl Dissociation 11 -5

6 1-propyl → methylidene + ethyl Dissociation 36 22

9 2-propyl → methyl + ethylidene Dissociation 27 21

16 Propylene → methylidene + ethylidene Dissociation 62 57

18 1-propylidene → methylidyne + ethyl Dissociation -22 -21

19 2-propylidene → methyl + ethylidyne Dissociation -55 -62

- 1-propylidyne → C + ethyl Dissociation 105 -

In general, the calculated reaction energies of Yang et al. and in this work for these reactions

are quite similar. It is observed that in general the reaction energies are lower in this work with

respect to Yang et al. The dissociation of 1-propyl, 2-propyl and propylene are

thermodynamically unfavorable (positive reaction energy). In contrast, propane hydrogenolysis

leads to more stable species (methyl and ethyl), which are both important for the formation of

gaseous side products. However, the formation of 1-propyl and 2-propyl remains

thermodynamically more favorable with respect to the dissociation products. The

hydrogenolysis of 1-propylidene and 2-propylidene forms stronger adsorbed species with

respect to the reactants, but the 1-propylidene and 2-propylidene are thermodynamically less

favored than propylene. Furthermore, it is interesting to look into the dissociation of 1-

propylidyne, which as thermodynamic most favorable product could be important as

intermediate for dissociated species. However, Yang et al. report strong endothermic reaction

energy of 105 kJ/mol for this reaction. Based on this reasoning, this reaction is not likely occur


and has not been incorporated in this work. The most realistic paths toward dissociated species

are via 1-propylidene and 2-propylidene.

4.3.3.1 Formation of gaseous side products

The cleavage of the C-C bonds results in the formation of C1 and C2 hydrocarbons on the

surface. Further reaction of these species leads either to adsorbed coke precursors or to the

desorption of smaller alkanes and alkenes, which strongly influences the selectivity for

propylene. The formation of methane, ethane and ethylene are included in the reaction

mechanism and compared with DFT studies from literature, see Figure 4-5 and Table 16. [20,

32, 47] The optimal geometries of these species can be found in Appendix B.

Figure 4-6: Reaction network for formation of gaseous side products on Pt(111).

The poor description of the physisorbed state of the alkanes is due the negligence of van der

Waals interactions in the other studies. Adsorbed ethyl prefers to dehydrogenate to ethylene

instead of to hydrogenate to ethane in both studies. However it should be noted that the

desorption energy of ethylene is high (131 kJ/mol) with respect to ethane (27 kJ/mol). However,

it should be noted that solely the enthalpic factors are included in the electronic energy.

CH2CH3

Pt

_

CH3-CH3,phys

CH4(g)

CH3-CH3(g)

CH2-CH2 + H

+ H-Pt

CH2=CH2,phys CH2=CH2(g)

Pt

_ CH4,physCH3 + H

Pt

_Pt

_

Pt

_

Pt

_

Methyl Physisorbed methane Gaseous methane

Ethyl

Physisorbed ethane Gaseous ethane

Ethylene Physisorbed ethylene Gaseous ethylene

A1 A2

B1

C1

B2

C2 C3


Table 4-16. Comparison between reaction energies for side product formation on Pt(111) of different DFT studies. The employed functionals and coverages in literature are noted underneath the table. This work uses an optPBE vdW-DF functional.

ΔEr (kJ/mol) Literature

[20, 32, 47]

This work


A1 Methyl + H → Methane,physisorbed Hydrogenation -21a -15

A2 Methane,physisorbed → Methane(g) Desorption 0 a 20

B1 Ethyl + H → Ethane,physisorbed Hydrogenation 43b

-4

B2 Ethane,physisorbed → Ethane(g) Desorption 31

C1 Ethyl → Ethylene + H Dehydrogenation -30b -13

C2 Ethylene → Ethylene,physisorbed Desorption 100c 90

C3 Ethylene,physisorbed →Ethylene(g) Desorption 35c 41 (a) A coverage of 0.11 ML is achieved and RPBE is used as functional. In addition, the ZPE is added to the reaction energy.

[47] (b) A coverage of 0.25 ML is achieved and RPBE is used as functional. [20] (c) A coverage of 0.11 ML is achieved and

PW91 is used as functional. [32]

4.3.4 Overall thermodynamics

To evaluate relative electronic energies of the various C3Hx-species an energy profile is

constructed with the energy of each intermediate with gaseous propane as reference. To satisfy

the conservation laws, it is essential that the mass of every species is identical. In the case of

dehydrogenated species, the relative energy is a combination of the energy of the

dehydrogenated species and the energy of the dehydrogenated hydrogen(s), conform equation

(7) (see below). The constructed energy profile is illustrated in Figure 4-7.

EC3𝐻𝐻𝑥𝑥 = 𝐸𝐸C3𝐻𝐻𝑥𝑥/surface + (8 − 𝑥𝑥)(𝐸𝐸H/surface − 𝐸𝐸surface)

− (𝐸𝐸C3𝐻𝐻8(g) + 𝐸𝐸surface) (7)

The activation of propane leads to both 1-propyl and 2-propyl, as the Gibbs free energies of

these species are similar. These species dehydrogenate preferentially further to propylene, as it

is the thermodynamically the most stable C3H6-species. Still propylidyne is the most energetic

stable species on the Pt(111) surface, so it is expected that the most likely reaction will be this

formation reaction and if isomerization reactions are excluded and Gibbs free energy is taken

into account as discussed in section 4.3.1, the most probable path is via propane →1-propyl

→1-propylidene →1-propylidyne.


Figure 4-7: Energy profile of adsorbed C3Hx (x= 5–8) species on the Pt(111) surface. Energies are determined relative to gaseous propane. In addition, the dissociated species are shown (red).

Hydrogenolysis reactions lead to the formation of endothermic products, except for the cracking

of 1-propylidene and 2-propylidene. Cracking of 2-propylidene leads to the formation of

species more stable than propylene; through this cracking reaction, C1 and C2 hydrocarbons are

formed on the surface. However, the description of the kinetics is necessary to evaluate the

dominant reaction paths and, eventually, the distribution of the intermediates on the surface.

4.4 Kinetics

In this work, the kinetic descriptors of our reactions are determined based on the transition

states, in this case the electronic activation energies. The geometry of the transition state has to

be found along the reaction coordinate. These structures are so-called first order saddle points

on the potential energy surface (PES) of the considered species. Hence, the transition state is

energetic minimum on the PES, in all directions except the reaction coordinate. In vibrational

analysis of transition states, this corresponds to one imaginary frequency assigned along the

reaction coordinate. Furthermore, the Hammond–Leffler postulate is applied to determine the

structure of the transition state. This postulate predicts a transition state that resembles the

-125

-105

-85

-65

-45

-25

-5

15R

elat

ive

ener

gy (k

J/m

ol)

C3H8 (g) C3H8* C3H7

* + H* C3H5* + 3H*

Propane (g)

Propane, phys

2-propyl

Propylene,phys

Propylene

1-propenyl1-propyl 1-propylidene

2-propylidene

2-propenyl

1-propylidyne

Ethyl & methyl

Ethyl & methylidene

Ethyl idene & methyl

Ethylidene & methylidene

Ethyl & methylidyne

Ethylidyne & methyl

C3H6* + 2H*


products (late) for endothermic reaction and transition state that resembles the reactants (early)

for exothermic reactions. [48]

If both the optimal geometry of transition state and reactants are obtained, the electronic

activation energy Δ‡E can be determined, as shown in equation (8). This value can be used as

an indication of the activation barrier. In equation (8), the forward electronic activation energy

of an elementary step is calculated, while to determine the reverse electronic activation energy,

either the reaction energy has to be added or the electronic activation energy has to be calculated

with respect to the electronic energy of the products.

∆ǂ𝐸𝐸 = 𝐸𝐸TS,ads − 𝐸𝐸react,ads (8)

To describe the kinetics of the reaction network, the network will be divided in three sections:

propane dehydrogenation to propylene (4.4.1), deep dehydrogenation of propylene (4.4.2) and

C-C cleavage of C3 intermediates (4.4.3). In each section, the stability of transition states is

evaluated and the electronic activation energy is determined for each reaction. The first section

can be interpreted to quantify the activity of the catalyst, while section two and three quantify

the selectivity towards propylene with respect to side reactions that lead to coke precursors.


dissociation. In the first and second, the breaking of C-H plays a major role, while in the last,

the breaking of a C-C bond is considered.

As the transition state geometries are calculated with the NEB method, an initial and final state

has to be given. The final state of dehydrogenation reaction has to be optimized with the

hydrogen nearby the intermediate in the same unit cell. The location of the adsorbed hydrogen

is either on top a Pt site or in a fcc site. The obtained geometries of the transition states can be

found in Appendix B.


As discussed previously, the dehydrogenation of propane towards propylene consists of two

elementary steps: the activation of propane to either 1-propyl or 2-propyl via respectively

dehydrogenation of the methyl or methylene group and β-dehydrogenation of 1-propyl or 2-

propyl leads to propylene. In this work, before the first elementary step, the physisorption of

propane has been taken into account. However, for this type of reaction it is assumed that the

electronic activation energy is zero.


Table 4-17. Comparison of the electronic activation energies between DFT studies for propane dehydrogenation towards propylene on Pt(111). Yang et al. used a PBE functional, while this work uses an optPBE vdW-DF functional.

† Reaction 11: an additional imaginary frequency has been observed. Reaction 12: Electronic energy of transition state is

determined based on the NEB calculation.

The electronic activation energies in this work are compared with the DFT study of Yang et al.

In general, the obtained activation barriers show a rather similar trend. However, for all

dehydrogenation reactions, an increase is observed of ~13 kJ/mol, except for the reaction path

towards propylene via 2-propyl. The results of Yang et al. and the reaction path via 1-propyl of

this work have shown that the two elementary steps for propylene formation have similar

reaction barriers. In contrast, the reaction barriers of the reaction path via 2-propyl are dissimilar

as the second reaction barrier is 12 kJ/mol higher than the first.

The location of detached hydrogen is the key parameter to describe the geometry of the

transition state. The transition state geometries of both DFT studies are similar. The distance

between the carbon and hydrogen atom along the reaction coordinate differs only ~0.1 Å.

Additionally, the detached hydrogen is located either on the same or on a nearby Pt atom.

Δ‡E (kJ/mol) Yang et al.

(0.25 ML) [4]

This work


0 Propane(g) → Propane,phys Physisorption 0 0

1 Propane,phys → 1-propyl + H Adsorption 67 80


4 1-propyl → 1-propylidene + H Dehydrogenation 70 83

5 1-propyl → propylene + H Dehydrogenation 68 79



10 1-propyl → 2-propyl Isomerization - 203

11 1-propylidene → propylene Isomerization - 154†

12 2-propylidene → propylene Isomerization - 168†

13a Propylene→ propylene,phys Chemisorption 90 92

13b Propylene,phys →Propylene (g) Physisorption - 43


Figure 4-8: Relative energy profile of the propane dehydrogenation towards propylene. The energies are determined relative to gaseous propane. Dehydrogenated hydrogen(s) are optimized in separate unit cells.

Based on the energy profile, shown in Figure 4-8, the reaction path via 2-propyl to form

propylene is the least activated. However, in reality, both reaction paths will occur on the

surface, as both the reaction and activation energies are similar for the two elementary steps.

The activation energies for the second elementary step of both reaction paths are higher than in

the first, indicating this may be the rate-determining step for the formation of propylene.

However, this is solely based on enthalpic contributions. When entropic factors are included, it

is expected that the first elementary step is rate determining as the entropy decreases during

adsorption since gaseous propane loses part of its degree of freedom.

The formation of 1-propylidene and 2-propylidene is disadvantaged as the activation energy of

the α-dehydrogenation is higher than of the β-dehydrogenization. This is illustrated in a relative

energy profile, see Figure 4-9. Hence, it is observed that propylene thermodynamically and

kinetically is preferred with respect to 1-propylidene and 2-propylidene. From those two

species, the formation of 2-propylidene is the highest activated.

-100

-50

0

50

100

150

Rel

attiv

een

ergy

(kJ/

mol

)

Reactant Product

Propane (g)

Propane, phys

1-propyl + H*


Propylene +2H*

Propylene (g) + 2H*

TS1TS5

TS2

2-propyl + H*

TS7


Figure 4-9: Relative energy profile of the propane dehydrogenation towards C3H6-species. The energies are determined relative to gaseous propane. Dehydrogenated hydrogen(s) are optimized in separate unit cells.

Furthermore, it is shown that the electronic activation energies of the isomerization reactions

are difficult to obtain within in the strict convergence criteria. Only the electronic activation

energy of the isomerization of 1-propylidene to propylene is obtained. However, for the

isomerization between 2-propylidene and propylene, the electronic energy of the transition state

is determined based on the NEB calculation. Furthermore, the transition state of the

isomerization between 1-propylidene and propylene has two imaginary frequencies. The

observed reaction barriers are in the range of 150 to 200 kJ/mol, indicating that these reactions

are unlikely to occur; certainly, as other paths have much smaller reaction barriers. Chen et al.

performed a DFT analysis on ethane dehydrogenation. Here, the isomerization reactions were

included and the obtained reaction barriers were at least 200 kJ/mol or higher, confirming our

assumption that isomerization reactions are unlikely to occur on the platinum surface. [20]

It is assumed that the adsorption of propylene is not an activated process. However, for

propylene to desorb, it has to overcome the reaction energy of desorption. In this case, the

reaction barrier of desorption is 135 kJ/mol, which is much higher than the determined barrier

of 90 kJ/mol by Yang et al. This discrepancy is discussed in 4.2.2.1.This value is essential to

determine the descriptor of selectivity towards propylene.

-80

-60

-40

-20

0

20

40

60

Rel

attiv

e en

ergy

(kJ/

mol

)

Reactant Product

Propane (g)

Propane, phys

1-propyl + H*

Propylene +2H*

TS1 TS6

TS2

2-propyl + H*

TS7

TS4

2-propylidene +2H*

1-propylidene +2H*


4.4.2 Deep dehydrogenation of propylene and other C3H6-species


the selectivity towards propylene with respect to formation of coke precursors. The

dehydrogenation reactions towards 1-propenyl, 2-propenyl and 1-propylidyne are studied in

this work.

Table 4-18. Comparison of the electronic activation energies between DFT studies for propane dehydrogenation towards propylene. Yang et al. used a PBE functional, while in this work, an optPBE vdW-DF functional is employed.


(0.25 ML) [4]

This work


14 Propylene → 1-propenyl + H Dehydrogenation 73 91


17 1-propylidene → 1-propylidyne + H Dehydrogenation 22 45†

20 1-propylidene → 1-propenyl + H Dehydrogenation 60 72

21 2-propylidene → 2-propenyl + H Dehydrogenation 52 66 † The optimized transition state has one additional imaginary frequency.

For all reactions, an increase of electronic activation energy, varying between 7 and 23 kJ/mol,

is observed with respect to the results of Yang et al. However, based on previous results, an

increase of at least 10 kJ/mol is expected based on the reaction energies in 4.3.2. Therefore, the

discrepancy is the largest for reaction 15, as a smaller increase is observed. Based on the

Brønsted–Evans–Polanyi principle, it is expected that the formation of 2-propenyl would have

a lower barrier than 1-propenyl as 2-propenyl is more stable than 1-propenyl, contradicting the

results of Yang et al. that observe a similar barrier for both reactions.

Deep dehydrogenation plays a major role in the selectivity of propane dehydrogenation with

respect to deactivation reactions. As 1-propylidyne is the most stable on the surface, its

formation can be related to the surface coke formation and the deactivation of the catalyst.

Hence, a selectivity descriptor is proposed as the difference between the activation energy of

propylene desorption and propylene dehydrogenation. In literature, the activation energy of

propylene dehydrogenation with the lowest barrier is taken, mostly the dehydrogenation to 2-

propenyl. [4] However, in this work, the activation energy for deactivation is determined based

on all deeply dehydrogenated species and not only those formed out of propylene. Hence, 1-

propylidyne is the most stable species on the surface and its formation has a low reaction barrier

via 1-propylidene dehydrogenation. Instead of comparing two reactions, two reaction paths are


used to define the selectivity descriptor in this work. Each reaction path consists out of two

elementary steps: for the desired product propylene, this is first the dehydrogenation of 1-propyl

to propylene and consequently the desorption of propylene, while for deep dehydrogenation,

this is consecutive α-dehydrogenation of 1-propyl to 1-propylidene and to 1-propylidyne.

To visualize the selectivity descriptor an energy profile, relative to gaseous propane, for both

considered reaction paths is constructed, see Figure 4-9. The mass conservation laws are

satisfied, so for the determination of relative energies of the dehydrogenated species, the energy

of dehydrogenated hydrogens is added, conform equation (7). In the final step, all adsorbates

desorb. In the case of 1-propylidyne, this is solely the adsorbed hydrogen atoms.

Figure 4-10: Relative energy profile for the propane dehydrogenation towards propylene (···) and propane deep dehydrogenation towards 1-propylidyne ( ̶ ̶ ) on Pt(111). The energies are determined relative to gaseous propane.

The dehydrogenation reaction to 1-propylidyne has the lowest barrier and its formation is

thermodynamically the most preferable. Hence, it is proposed that the most likely reaction

pathway in the network is propane → 1-propyl → 1-propylidene → 1-propylidyne. However,

it is necessary to note that both 1-propyl and 1-propylidene are both less stable and have higher

activation barriers with respect to 2-propyl and propylene. Still, the activation barriers of the

first elementary steps are more similar than those for deep dehydrogenation, indicating that the

most likely reaction is indeed the dehydrogenation towards 1-propylidyne. The activation

-125

-75

-25

25

75

125

Reactant Product

Propane (g)



Propylene +2H*

Propylene (g) + 2H*

TS1

TS5


1-propylidyne + 3/2 H2(g)

1-propylidyne + 3H*

TS4

1-propylidene +2H*

TS17


energies for the first elementary step are similar, however for the second step, deep

dehydrogenation (45 kJ/mol) is less activated than desorption (135 kJ/mol).

However, it is important to note that the solely the enthalpic contributions are accounted for.

Otherwise, solely 1-propylidyne would be formed and would rapidly deactivate the catalyst. To

include the entropic contributions, a Gibbs free energy diagram for the considered reactions

path is constructed. The Gibbs free energy is determined at 900 K, and the entropy and enthalpy

for each intermediates and transition state is calculated based on partition functions and

vibrational analysis.

Figure 4-11: Gibbs free energy diagram at 900 K for the propane dehydrogenation towards propylene (···) and propane deep dehydrogenation to 1-propylidyne ( ̶ ̶ ) on Pt(111). The energies are determined relative to gaseous propane.

The entropic contribution causes the first step to have the highest activation energy, as the

entropy loss is the greatest on adsorption of gaseous species. On the surface, the intermediates

are less influenced by the entropic changes. However, in contrast with the electronic energy,

the desorption of propylene is favored with respect to dehydrogenation of 1-propylidyne as

during desorption, propylene obtains a higher degree of freedom. At the final step, all hydrogens

-75

-25

25

75

125

175

Gib

bs fr

ee e

nerg

y (k

J/m

ol)

Propane (g)

Propane, phys

1-propyl + H*


Propylene +2H*

Propylene (g)+ 2H*

TS1

TS5

1-propylidene *H2+

TS17TS4

1-propylidyne + 3H*

Reactant Product


Propylene + H2 (g)


desorb and it is observed that the difference in Gibbs free energy is 11 kJ/mol in favor of

gaseous propylene.


Hydrogenolysis reactions lead to smaller hydrocarbons, which form either surface cokes or

gaseous side products. Determining the kinetic parameters of these reactions is essential to

describe propylene selectivity with respect to other gaseous products. However, the dissociation

reactions included in this reaction network are endothermic except for the dissociation of 1-

propylidene and 2-propylidene. The breaking of C-C leads to higher reactive barriers than for

C-H bond breaking. Wang et al. propose that the reaction barrier depends not only on the

thermodynamics, but also additionally on the bond polarity. Under similar conditions, breaking

of C-H bonds occurs before C-C bonds, as the difference in bond polarity between C-C and C-

H enhance the breaking of C-H bonds. [49]

Table 4-19. Comparison of the electronic activation energies between DFT studies for hydrogenolysis of C3-intermediates. All considered reactions are dissociation reactions. Yang et al. used a PBE functional, while this work uses an optPBE vdW-DF functional.


(0.25 ML) [4]

This work

(0.13 ML) Surface reaction

3 Propane,phys→ methyl + ethyl 235 187

6 1-propyl → methylidene + ethyl 163 168†

9 2-propyl → methyl + ethylidene 175 185

16 Propylene → methylidene + ethylidene 193 214

18 1-propylidene → methylidyne + ethyl 114 119

19 2-propylidene → methyl + ethylidyne 126 138

- 1-propylidyne → C + ethyl 184 - † The optimized transition state has one additional imaginary frequency.

For all reactions, except propane dissociation, an increase of electronic activation energy is

observed. The propane dissociation differs so much because physisorbed propane is used a

reference species. This species is much more stabilized in this work, with respect to Yang et al.

because the vdw-DF functional has been applied. However, similar results are obtained if the

gaseous propane is used as reference. This means that the reaction energy for propane

physisorption, as reported in Table 4-11, needs to be subtracted. This results in an electronic

activation energy of 237 kJ/mol and 230 kJ/mol, respectively for Yang et al. and this work. The

dissociation of propylene is unlikely as both reaction energy as its barrier is the highest, while


other reactions such as dehydrogenation and desorption are more favorable. In addition, the

dissociation reaction barriers for 1-propylidene and 2-propylidene are higher than the barrier

for the dehydrogenation reactions of these species, while those have the lowest dissociation

reaction barriers. This indicates that cracking reactions are not likely to occur on flat Pt(111)

surfaces and these reactions solely occur with deeper dehydrogenated species or on other

adsorption sites, such as steps or edges.

4.4.3.1 Formation of gaseous side products

Gaseous side products such as ethylene, ethane and methane can be formed through reaction of

C1- and C2-hydrocarbons on the surface, as proposed in Figure 4-6. Various DFT studies have

reported the reaction barriers of these reactions. [20, 32, 47]

Table 4-20. Comparison between electronic activation energies for side product formation of different DFT studies. The employed functionals and coverages in literature are noted underneath the table. This work uses an optPBE vdW-DF functional.

Δ‡E (kJ/mol) Literature

[20, 32, 47]

This work


A1 Methyl + H → Methane,physisorbed Hydrogenation 69a 68

A2 Methane,physisorbed → Methane(g) Desorption 0 a 20

B1 Ethyl + H → Ethane,physisorbed Hydrogenation 98b 68

B2 Ethane,physisorbed → Ethane(g) Desorption - 31

C1 Ethyl → Ethylene + H Dehydrogenation 78b 93

C2 Ethylene → Ethylene,physisorbed Desorption 100c 90

C3 Ethylene,physisorbed →Ethylene(g) Desorption 35c 41 (a) A coverage of 0.11 ML is achieved and RPBE is used as functional. In addition, the ZPE is added to the reaction energy.

[47] (b) A coverage of 0.25 ML is achieved and RPBE is used as functional. [20] (c) A coverage of 0.11 ML is achieved and

PW91 is used as functional. [32]

The electronic activation energy for the hydrogenation of methyl is for both studies equivalent.

Furthermore, the description of the methane physisorption fails as van der Waals interactions

are excluded in literature. The electronic activation energy for methyl dehydrogenation is

smaller than the hydrogenation reaction of 1-propyl and 2-propyl. The electronic activation

energies of 1-propyl and 2-propyl hydrogenation are respectively 87 kJ/mol and 82 kJ/mol.

Furthermore, physisorbed methane is more stable than adsorbed methyl and hydrogen (-15

kJ/mol). Hence, it is expected that methyl hydrogenation occurs more frequently than methane


dehydrogenation. However, it should be noted that dehydrogenation is second order reaction

and depends on the hydrogen coverage.

The hydrogenation step of ethyl towards ethane and the consecutive desorption step are reported

together with one electronic activation energy, which represents the C-H formation as this step

is more activated than desorption. The electronic activation energy for ethyl hydrogenation is

lower than described in literature, while the ethyl dehydrogenation is less activated than

literature. However, it should be noted that in literature different functionals and coverage are

employed. To determine the kinetically favored reaction path, a reaction profile (see

Figure 4-12) of C2Hy-species on the surface is constructed based on equation (9) (see below).

EC2𝐻𝐻𝑦𝑦 = 𝐸𝐸C2𝐻𝐻𝑦𝑦/surface + (6 − 𝑥𝑥)(𝐸𝐸H/surface − 𝐸𝐸surface)

− (𝐸𝐸C2𝐻𝐻6(g) + 𝐸𝐸surface) (9)

Figure 4-12. Relative energy profile of ethane dehydrogenation towards ethylene on Pt(111). The electronic energy is calculated relative to gaseous ethane. The dehydrogenated hydrogens are optimized in separate unit cells.

Since both ethane and ethylene are not supplied as feed, the only way to enter this mechanism

is through the formation of ethyl through cracking. In this network, ethyl dehydrogenation

towards ethylene competes with ethyl hydrogenation towards ethane. On the surface, di-σ

adsorbed ethylene is the thermodynamic product as it is more stable than physisorbed ethane.

However, physisorbed ethane is the kinetically preferred product as it is kinetically favored

-60

-40

-20

0

20

40

60

80

100

Rel

ativ

e en

errg

y (k

J/m

ol)

Ethane (g)

Ethane,physEthyl + H*

Ethylene + 2H*

Ethylene + 2H*

Ethylene (g) + 2H*

TSB1

TSC1

Reactant AlkeneAlkane


with respect to adsorbed ethylene. This energy profile accounts solely for enthalpic

contributions and for the formation of gaseous species, entropy plays a major role as driving

force. To determine the most probable gaseous side product, the Gibbs free energy is

determined for gaseous ethane and gaseous ethylene plus gaseous H2, to satisfy the conservation

laws. The selected temperature is 900 K, which is a typical reaction temperature for propane

dehydrogenation [46]. At this temperature, gaseous ethane has a lower Gibbs free energy than

ethylene and gaseous H2. The Gibbs free energy difference is 10 kJ/mol. This indicates that

gaseous ethane is as the thermodynamically preferred C2 side product. However, it is noted that

the hydrogenation step also depends on hydrogen coverage, which is rather low in most

dehydrogenation processes.

To determine the Gibbs free energy driving force for the formation of small gaseous

hydrocarbons from gaseous propane, the Gibbs free energy of gaseous methane and ethane (as

preferential gaseous C2 hydrocarbon) at 900 K is compared with gaseous propane. However, to

satisfy the carbon and hydrogen balance, the Gibbs free energy of gaseous ethane and methane

are calculated together with respect to gaseous propane and gaseous H2. Gaseous methane and

ethane are together thermodynamically more stable as the resulted Gibbs free energy difference

is -17 kJ/mol. This indicates that these species will be formed despite the high reaction barrier

for propane hydrogenolysis.

As gaseous propylene is the desired product, the Gibbs free energy of this species should be

lower than those of the most probable gaseous side products ethane and methane. The Gibbs

free energy of gaseous propylene and H2 is 59 kJ/mol more stable than gaseous propane, as

illustrated in Figure 4-11. As the Gibbs free energy is lower for gaseous propylene than gaseous

methane and ethane, the formation of gaseous propylene is favored.

4.5 Microkinetic modelling

In this section, a microkinetic simulation of propane dehydrogenation will be performed. First,

all kinetic and thermodynamic parameters are calculated based on the electronic geometry and

transition state optimizations. Furthermore, the kinetic model is constructed based on the

elementary steps in the reaction network of Pt(111) as discussed previously, see section 4.3

(Figure 4-4 and Figure 4-6). Finally, the model is implemented in a computer code and

simulations with the constructed kinetic model are performed. In literature, Sui et al. have also


performed microkinetic simulations on a similar Pt catalyst surface. [50] However, they solely

focused on the dominant dehydrogenation path towards propylene while in this work all side

reactions and deep dehydrogenation reactions have also been considered.

4.5.1 Determination of the thermodynamic and kinetic parameters

Vibrational frequencies are necessary in order to calculate thermodynamic and kinetic

properties at finite temperatures. Vibrational analyses have been performed, as mentioned

before, on strictly optimized geometries to avoid spurious imaginary frequencies.

After the calculation of vibrational frequencies, vibrational partition functions, zero-point

energies and thermal corrections to the enthalpy and entropy can be derived from statistical

thermodynamic equations. [51] From these values, kinetic and thermodynamic parameters are

calculated as function of temperature.

In this work, all species are assumed to be immobile on the surface, which is mainly for

adsorbed hydrogen and the physisorbed species a strong approximation at the high propane

dehydrogenation temperature (~873 K). Consequently, for surface species only vibrational

contributions have been taken into account. Free rotation and translation have been considered

explicitly only for gas phase species. All thermodynamic and kinetic parameters (pre-

exponential A and activation energy Ea) have been determined at a temperature of 900 K. The

actual values of the calculated parameters can be found in Appendix C.

4.5.2 Construction of the microkinetic and reactor model

A microkinetic model can be constructed using adsorption and desorption steps and the

elementary reaction steps (dehydrogenation, dissociation and isomerization) on the surface,

based on the kinetics and thermodynamics calculated from DFT. Adsorption rate coefficients

are described as a product of the incident molecular flux, F, and the initial sticking coefficient

s0 (see equation 10). It is assumed that adsorption is non-activated for all species, and

consequently the sticking coefficient is approximately equal to that on a clean surface. This is

considered to be equal to one for all adsorbates. The rate coefficients for desorption are obtained

using the thermodynamic equilibrium coefficient as determined by the DFT calculations (see

equation 11).

𝑘𝑘𝑎𝑎𝑝𝑝𝑝𝑝 =𝑠𝑠0 ∙ 𝐹𝐹𝑝𝑝

=𝑠𝑠0

𝑛𝑛𝑝𝑝�2𝜋𝜋 ∙ 𝑚𝑚 ∙ 𝑘𝑘𝐵𝐵 ∙ 𝑇𝑇 (10)


𝑘𝑘𝑝𝑝𝑑𝑑𝑝𝑝 =

𝑘𝑘𝑎𝑎𝑝𝑝𝑝𝑝𝐾𝐾𝑑𝑑𝑒𝑒

(11)

In equation 10, F is the incident flux, p is the pressure in bar, nt is the number of active sites per

m2 (equal to 1.66×1019 m-2), m is the molecular mass of the adsorbate molecule in kg, kB is the

Boltzmann constant and T is the temperature.

An Arrhenius expression is used to describe the surface rate coefficients (k in s-1), using the

activation energies (Ea) and pre-exponential factors (A):

𝑘𝑘 = 𝐴𝐴 exp �-𝐸𝐸𝑎𝑎𝑅𝑅𝑇𝑇

� (12)

The microkinetic simulation will be performed using an in-house developed python code. First,

the reaction rates of all elementary reaction steps are written using basic rate laws. Production

rates of all intermediate species are also formulated. Furthermore, a site balance is constructed

to make sure that the sum of all fractional surface coverages amounts to one. Finally, the

turnover frequency for all gaseous species is calculated. The turnover frequency (TOF) of all

gaseous species is obtained as the molecules consumed/formed per active site per second (s-1).

For example, the production rate of the surface intermediate 2-propylidene (CH3)2C=Pt2 (s-1) is

given by:

𝑅𝑅(CH3)2C=Pt2 =d𝜃𝜃(CH3)2C=Pt2

d𝑡𝑡= 𝑘𝑘for,8𝜃𝜃2-propyl𝜃𝜃free − 𝑘𝑘for,12𝜃𝜃(CH3)2C=Pt2

− 𝑘𝑘for,19𝜃𝜃(CH3)2C=Pt2𝜃𝜃free − 𝑘𝑘for,21𝜃𝜃(CH3)2C=Pt2𝜃𝜃free

(13)

For reactants and products, the production rate is increased with an adsorption minus a

desorption term. The site balance comprises all surface species:

𝜃𝜃free + ��𝜃𝜃CiHj

8

𝑗𝑗=1

3

𝑖𝑖=0

= 1 (14)

The pseudo-steady-state surface concentrations are obtained transiently by solving the

differential equations of equation (13), using the LSODA integration routines from the

ODEPACK Fortran library, as implemented in the Python package SciPy.

4.5.3 Results of the simulation

In this section the simulation results of the microkinetic model will be discussed. First, the time

of the simulation is varied in order to study the different regimes of the propane


dehydrogenation reaction. The most interesting regime will be selected for further discussion

on the effect of varying temperature and pressures.

4.5.3.1 Effect of varying the simulated time

Three distinct simulated times will be discussed: one simulation over a very short integration

time interval of 0.1 seconds, one over an intermediate time of 10 seconds and one over a very

long time of 109 seconds. All simulations have been performed at a fixed reactant pressure (H2:

0.1 bar and C3H8: 0.3 bar) and fixed temperature (typical dehydrogenation temperature of 873

K) without an initial product pressure. These reactant pressures have been chosen based on

literature for typical reaction conditions for a comparable Pt catalyst. [52]

4.5.3.1.1 Simulated time of 0.1 seconds

In first instance, the shortest integration time of 0.1 seconds is considered (with intermediate

results printed out every 0.001 seconds). By selecting this short time scale it is possible to

capture the transient behavior of the reaction as can also be observed in a TAP reactor

(Temporal Analysis of Products).

The result for the turnover frequency (TOF) can be found in Figure 4-13 (blue). The TOF is the

highest for this small simulated time. This indicates that there is a high initial activity on the

Pt(111) catalyst.

Figure 4-13. Turnover frequency (TOF) in s-1 of H2 and C3H6 for different integrated times. Blue: 0.1s (left axis), red: 10s (left axis), green: 109s (right axis).

0.0E+00

2.0E-09

4.0E-09

6.0E-09

8.0E-09

1.0E-08

1.2E-08

1.4E-08

950

1000

1050

1100

1150

1200

H2 C3H6

TOF

[s-1

] for

109

sim

ulat

ion

tim

e onl

y

TOF

[s-1

]

Species

0.1 s 0.1 s

10 s

0.1 s 0.1 s

10 s

109 s

109 s


Furthermore, the coverage of the catalyst surface can also be examined, see Figure 4-14 (blue).

For all simulations it is assumed that the initial coverage on the surface equals zero for all

species. 1-propylidyne is dominantly present on the surface for the short simulated time. Only

a very small amount of hydrogen covers the surface. The whole surface is not yet covered

(approximately 80 %) and still a large amount of free sites is available on the surface.

Figure 4-14. Coverage of the surface for different simulated times. The last column gives the total occupancy of the catalytic surface. Blue: 0.1s, red: 10s, green: 109s. The coverage of all other species on the surface is equal to zero. For all simulations it is assumed that the initial coverage on the surface equals zero for all species.

The fact that there is still a large amount of free sites on the surface ensures that the activity of

the catalyst is maintained in this regime (high TOF of C3H6) . Furthermore, it is noticed that

1-propylidyne is dominant on the surface. This could already be guessed based on the kinetic

parameters (see Appendix C2). Reaction 17 (the reaction that leads to the formation of

1-propylidyne) has the smallest activation energy of all elementary steps in the reaction network

(Ea of 14.9 kJ/mol). For this reason, 1-propylidyne is kinetically favored and is formed in the

beginning of the reaction, i.e. in the short-time integration regime.

4.5.3.1.2 Simulated time of 10 seconds

For the second regime, an average integration time of 10 seconds (with intermediate results

printed out every second) has been selected. The selection of this time scale makes it possible

to look at phenomena that occur past the initial transient regime while still maintaining high

catalytic activity.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

H CH3CH-Pt2 1-propylidyne CH-Pt3 CH3C-Pt3 Occupied

Perc

enta

ge o

f the

surf

ace

Species

0.1 s10 s1000000000 s


The TOF (red, see Figure 4-13) is only smaller by 12 % than compared to the short-time

integration regime of 0.1s, but is still high enough to have sufficient conversion of propane to

propylene. The coverages (red, see Figure 4-14) show another interesting feature. The coverage

of 1-propylidyne has decreased compared to the short-time integration regime and small

coverages of CH≡Pt3 and CH3C≡Pt3 are observed. It is proposed that 1-propylidyne, although

it is the kinetically favored deeply dehydrogenated species, is converted into these

thermodynamically very stable C1 and C2 species. This is confirmed, for e.g. CH≡Pt3, by the

net reaction rates of the interesting elementary steps, see Table 4-21.

Table 4-21. Net reaction rates for reaction 17 and reaction 18 (see Figure 4-4) for different simulated times. Reaction 17: 1-propylidene → 1-propylidyne + H-Pt, reaction 18: 1-propylidene → CH≡Pt3 + CH3CH2-Pt.

Simulated time 0.1 s 10 s

net17 [s-1] -0.00619 -0.00535 net18 [s-1] 0.00472 0.00408

The net reaction rate of reaction 17 is negative, which means that 1-propylidyne is converted

back to 1-propylidene and part of this species is converted to CH≡Pt3 via reaction 18. At a

simulated time of 10s this effect is clearly visible in the coverages (red, see Figure 4-14). The

Gibbs free energies of CH≡Pt3 + CH3CH2-Pt and 1-propylidyne + H-Pt are rather similar (see

Table 4-22), but the thermodynamic driving force for CH≡Pt3 + CH3CH2-Pt is larger because

CH3CH2-Pt has the ability to react further via reaction B or reaction C (see Figure 4-4) and

eventually to desorb as respectively gaseous ethane or ethylene. Furthermore, it is also possible

that CH3CH2-Pt undergoes a C-C scission reaction to form thermodynamically very stable C1

species. However, these reactions have not been taken into account yet and will be considered

for future work.

Table 4-22. Gibbs free energy of 1-propylidyne + H-Pt and CH≡Pt3 + CH3CH2-Pt at 900 K (electronic energy + thermal contributions + entropy).

Gibbs free energy (900 K)

kJ/mol 1-propylidyne + H-Pt -23430 CH≡Pt3 + CH3CH2-Pt -23364

It is proposed that, in terms of deactivation, 1-propylidyne is the ‘fast’ coke precursor, while

CH≡Pt3 and CH3C≡Pt3 species are ‘slow’ coke precursors. This interesting result could not have


been obtained by solely looking at reaction energies and barriers and this stresses the

importance of microkinetic simulations.

4.5.3.1.3 Simulated time of 109 seconds

Finally, a very long integration time of 109 seconds (approximately 32 years, with intermediate

results printed out every 1000 seconds) has been selected. By selecting this long time scale it is

possible to further study the deactivation of the catalyst as it was pointed out in the previous

section that deactivation has some interesting features.

From Figure 4-13 (green, right axis) it is clear that the activity of the catalyst after 109 seconds

has deteriorated substantially. The TOF of C3H6 has decreased by 12 orders of magnitude

compared to the situation after 0.1 seconds integration. This is confirmed by looking at the

surface coverage (see Figure 4-14, green): no free sites are available for reaction (total

occupancy of 100 %).

The hypothesis that the initial high amount of 1-propylidyne is converted into CH≡Pt3 and

CH3C≡Pt3 species is confirmed by looking at the coverages. There is no 1-propylidyne left on

the surface and the whole surface is blocked by the C1 and C2 species. At a simulated time of

109 seconds the coverages have reached a quasi steady-state regime since all coverages have

converged except for those of CH≡Pt3 and CH3C≡Pt3 species, but these only increase slightly

since the coke formation at this point has already reached 99.99 %.

Consequently, the 109 seconds time regime is not very interesting because activity is very low

and all free sites are blocked. For further study, the equations will not be solved until (quasi)

steady-state has been reached and only the most interesting time scale regime will be regarded.

Steady state can only be reached after a very long integration time (i.e. >32 years) and it is not

very useful to go to these long time scales. In order to investigate the effect of varying

temperature and pressures, the 10 seconds simulated time will be implemented since in this

regime activity is still high, but also a substantial amount of 1-propylidyne has already been

converted to CH≡Pt3 and CH3C≡Pt3.

4.5.3.2 Effect of varying temperature

All simulations have been performed at a fixed reactant pressure (H2: 0.1 bar and C3H8: 0.3 bar).

Again, these reactant pressures have been chosen based on literature for typical reaction

conditions for a comparable Pt catalyst. [52] The temperature is varied between 400 °C and 700


°C with an interval of 25 °C (typical dehydrogenation temperature is 600 °C or 873 K). The

simulated time has been kept fixed at 10 seconds. The resulting TOFs can be found in

Figure 4-15 and Figure 4-16.

Figure 4-15. TOF of H2 and C3H6 in s-1 as function of temperature in °C for a simulated time of 10 seconds. The reactant pressures are fixed (C3H8/H2: 0.3/0.1 bar).

Figure 4-16. TOF of CH2, C2H4 and C2H6 in s-1 as function of temperature in °C for a simulated time of 10 seconds. The reactant pressures are fixed (C3H8/H2: 0.3/0.1 bar).

The TOF increases for C3H6 and H2 with increasing temperature (Figure 4-15). Propane

dehydrogenation is an endothermic reaction so this is the expected behavior. The highest TOF

is obtained at a temperature of approximately 675 °C. At higher temperatures, deep

dehydrogenation barriers can be overcome and the TOF decreases again. Also, desorption can

0

200

400

600

800

1000

1200

1400

1600

1800

2000

400 450 500 550 600 650 700

TOF

[s-1

]

T [°C]

C3H6

H2

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

400 450 500 550 600 650 700

TOF

[s-1

]

T [°C]

CH4C2H4C2H6


occur more easily at these high temperatures, leading to more free sites on the surface. Both

TOFs overlap because hydrogen and propylene are formed together in a 1:1 ratio for the

propane dehydrogenation reaction when the amount of side reactions is still limited.

However, at the temperature of 675 °C, the selectivity of the catalyst becomes worse. This can

be seen from Figure 4-16 where at this temperature the relative TOF of methane and ethylene

compared to the TOF of C3H6 becomes higher. However, the TOF of ethane remains unaltered

with increasing temperature. This is because at these conditions dehydrogenation reactions,

such as elementary reaction C, are preferred compared to hydrogenation reactions, such as

elementary reaction B. This is also clearly seen in the net reaction rates, see Table 4-23. At high

temperatures (600 and 700 °C) the net reaction rate of reaction C is higher than reaction B.

However, at low temperatures (400 °C) this is no longer the case. Furthermore, at low

temperatures, the C-C scission reactions (e.g. elementary step 3) do not occur on the surface

due to a high activation barrier.

Table 4-23. Net reaction rates of the elementary reactions (see Figure 4-4) on Pt(111) for fixed reactant pressure (H2: 0.1 bar and C3H8: 0.3 bar) and varying temperature.

Temperature [°C] net rate in s-1 400 °C 600 °C 700 °C net1 2.7485 528.8687 836.0576 net2 4.3533 501.4381 659.6029 net3 0.0000 0.0001 0.0009 net4 0.3040 69.8104 116.4396 net5 2.4444 459.0585 719.6181 net6 0.0000 0.0000 0.0000 net7 4.1487 467.7158 609.5194 net8 0.2046 33.7220 50.0828 net9 0.0000 0.0001 0.0005 net10 0.0000 -0.0001 -0.0001 net11 0.0000 0.0073 0.0335 net12 0.0000 0.0000 0.0000 net14 -0.3040 -69.8043 -116.4092 net15 -0.2046 -33.7194 -50.0703 net16 0.0000 0.0000 0.0000 net17 0.0000 -0.0053 -0.0169 net18 0.0000 0.0041 0.0138 net19 0.0000 0.0026 0.0125 net20 0.3040 69.8043 116.4092 net21 0.2046 33.7194 50.0703 netA 0.0000 0.0028 0.0140 netB 0.0000 0.0000 0.0000 netC 0.0000 0.0042 0.0147


Table 4-23 also makes clear that at high temperature (700 °C) the deep dehydrogenation

reactions become favored, e.g. reaction 18 and 19 relatively to reaction 1, which results in a

lower TOF for C3H6.

From Table 4-23 it is also observed that both reaction pathways via 1- and 2-propyl towards

propylene are followed and both are the important paths toward propylene formation. There is

no clear distinction between the two paths, which was also assumed in previous section

(see 4.4.1).

It is also interesting to study the coverage of the surface at different temperatures, see

Figure 4-17.

Figure 4-17. Coverage of the Pt(111) surface (fraction) of species at different temperatures. Blue: 400 °C, red: 600 °C and green: 700 °C.

It is again confirmed that at the very high temperature (700 °C) side reactions and deep

dehydrogenation reactions are favored and the conversion of 1-propylidyne to

thermodynamically more stable C1 and C2 species is earlier initiated. Furthermore, due to easier

desorption, more free sites become available.

4.5.3.3 Effect of varying propylene pressure

All simulations have been performed at a fixed reactant pressure (H2: 0.1 bar and C3H8: 0.3 bar)

and fixed temperature (600 °C). The simulated time has been kept fixed at 10 seconds. The

C3H6 product pressure has been varied from 0 to 1 bar in order to evaluate the effect of product

pressure on the TOF, with steps of 0.1 bar. The resulting TOFs can be found in Figure 4-18 and

Figure 4-19.

0.00.10.20.30.40.50.60.70.80.91.0

H CH3CH-Pt2 1-propylidyne CH-Pt3 CH3C-Pt3 Occupied

Frac

tion

of th

e sur

face

Species

400 °C600 °C700 °C


Figure 4-18. TOF of H2 and C3H6 in s-1 as function of C3H6 pressure in bar for a simulated time of 10 seconds. The reactant pressures and temperature are fixed (H2: 0.1 bar and C3H8: 0.3 bar, 600 °C).

Figure 4-19. TOF of CH2, C2H4 and C2H6 in s-1 as function of C3H6 pressure in bar for a simulated time of 10 seconds. The reactant pressures and temperature are fixed (H2: 0.1 bar and C3H8: 0.3 bar, 600 °C).

It can immediately be noticed that changing the pressure of C3H6 does not have a large influence

on the TOFs. Even at the highest evaluated pressure of C3H6 (1 bar), which is over three times

the pressure of propane, the TOF of C3H6 and H2 only decrease by 10 % compared to the 0 bar

C3H6 case. Also the TOFs of methane, ethylene and ethane remain unaffected. This justifies the

fact that no reactor model equations have been taken into account in the simulation, since

920

940

960

980

1000

1020

1040

1060

0.0 0.2 0.4 0.6 0.8 1.0

TOF

[s-1

]

p_C3H6 [bar]

C3H6H2

0

0.0005

0.001

0.0015

0.002

0.0025

0.003

0.0035

0.004

0.0045

0.0 0.2 0.4 0.6 0.8 1.0

TOF

[s-1

]

p_C3H6 [bar]

CH4C2H4C2H6


accumulation of product pressures in the gas phase does not significantly change the overall

turnover rate.

It is also remarkable that the inverse reaction, i.e. propylene hydrogenation, does not occur at

very high pressures of C3H6. However, hydrogenation only occurs at high pressures of both

C3H6 and H2 and at lower temperatures. It was validated that the applied kinetic model predicts

propylene hydrogenation with a simulated TOF of 0.05 s-1 for a pressure of 0.3 bar C3H8, 3 bar

H2, 3 bar C3H6 and a temperature of 300 °C.

The coverages and net rates remain mainly unaffected by a change in C3H6 pressure. The

dominant reaction path (via 1- and 2-propyl) is not changed with increasing C3H6 pressure since

the net rates scale with the TOF. The net rates of the elementary reactions leading to propylene,

i.e. reactions 1, 2, 5 and 7, are lower at high pressure of C3H6 (1 bar) compared to the 0 bar

case. This is also reflected in the values of the TOFs (Figure 4-18).

4.5.3.4 Effect of varying side product pressure

All simulations have been performed at a fixed reactant pressure (H2: 0.1 bar and C3H8: 0.3 bar)

and fixed temperature (600 °C). The simulated time has been kept fixed at 10 seconds. The

CH4, C2H4 and C2H6 side product pressures have been varied from 0 to 1 bar with steps of 0.2

bar. All pressures are simultaneously increased by the same amount as a general check on the

effect of the side products pressures on the TOF. The resulting TOFs can be found in

Figure 4-20 and Figure 4-21.


Figure 4-20. TOF of H2 and C3H6 in s-1 as function of side products pressure (CH4, C2H4 and C2H6) in bar for a simulated time of 10 seconds. The reactant pressures and temperature are fixed (H2: 0.1 bar and C3H8: 0.3 bar, 600 °C).

Figure 4-21. TOF of CH2, C2H4 and C2H6 in s-1 as function of side products pressure (CH4, C2H4 and C2H6) in bar for a simulated time of 10 seconds. The reactant pressures and temperature are fixed (H2: 0.1 bar and C3H8: 0.3 bar, 600 °C).

Changing the pressure of the side product does not have a large influence on the TOFs of C3H6

and H2. They only change in a range of 0.5 %, which is practically negligible.

There are some interesting features to see for the TOFs of CH4, C2H4 and C2H6. First of all, it

is noticed that the TOF of CH4 remains unaffected, which is surprising since the CH4 pressure

is raised substantially. However, it is well-known that methane activation is difficult, as also

1040

1040.5

1041

1041.5

1042

1042.5

1043

1043.5

1044

1044.5

1045

0.0 0.2 0.4 0.6 0.8 1.0

TOF

[s-1

]

p_SideProduct [bar]

C3H6H2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.0 0.2 0.4 0.6 0.8 1.0

TOF

[s-1

]

p_SideProduct [bar]

CH4C2H4C2H6


stated in literature by Vines et al., and this is also observed here. [47] Furthermore, it is also

surprising that with increasing C2H4/C2H6 pressure, the TOF of C2H4 increases. This is due to

the conversion of ethane into ethylene (i.e. ethane dehydrogenation), which is also confirmed

by the net rates at 1 bar side product pressure. At high side product pressure of 1 bar the net

reaction rate of reaction B (ethane formation from ethyl) equals -0.1157 s-1, while the net

reaction rate of reaction C (ethylene formation from ethyl) equals 0.1193 s-1. This confirms that

ethane adsorbs and subsequently dehydrogenates to form ethylene.

4.6 Conclusions

In this chapter, the propane dehydrogenation characteristics on pure Pt catalysts are

investigated. The abundant Pt(111) phase is selected to represent the pure Pt catalyst. A 4×2

unit cell is constructed for the DFT calculations on the Pt(111) surface. This catalyst model is

used as reference for more advanced catalyst models later in this work. The proposed reaction

network is divided in three sections: propane dehydrogenation to propylene, deep

dehydrogenation of propylene and hydrogenolysis of C3 intermediates. The first section consists

of elementary steps towards propylene and other C3H6-species. Furthermore, this first section

is employed to quantify the catalyst activity of pure platinum. In the second section, the deep

dehydrogenation towards coke precursors is investigated. The selectivity towards these species

is an indication for the catalyst deactivation. In the third section, the hydrogenolysis reactions

are taken into account. These reactions form C1 and C2 adsorbates on the surface. These species

eventually lead to gaseous side products such as methane, ethane and ethylene and affect the

selectivity towards propylene.

Both the thermodynamics and kinetics of propane dehydrogenation are evaluated for the

considered reactions. The propane dehydrogenation towards propylene occurs both via 1-propyl

and via 2-propyl, as the Gibbs free energies at 873 K of these species are similar. However the

reaction path via 2-propyl is preferred as it has a lower reaction barrier with respect to 1-propyl.

Both 1-propyl and 2-propyl dehydrogenate preferentially further to propylene, as it is the

thermodynamically most stable C3H6-species. The resulted electronic activation energy is

similar for both reactions.

Thermodynamically, 1-propylidyne is the most stable species on the Pt(111) surface, so it is

expected that the most likely reaction will be its formation reaction. Multiple reaction paths


towards 1-propylidyne are possible if isomerization reactions are included. However, it is

shown that the isomerization reactions are highly activated and unlikely to occur. The most

probable path is via propane →1-propyl →1-propylidene →1-propylidyne. However, this is

solely based on the electronic energy. The entropic contributions are included by determining

the Gibbs free energy of the intermediates and transition states. The entropic contributions are

greater for gaseous species than for adsorbates, so based on the Gibbs free diagram, gaseous

propylene is 11 kJ/mol more stable than 1-propylidyne.

Hydrogenolysis reactions compete with the dehydrogenation reactions and C-C bond breaking

leads to the formation of less stable products with respect to dehydrogenated species, except

for the cracking of 2-propylidene. However, cracking of 2-propylidene and other

hydrogenolysis reactions are higher activated than the competing dehydrogenation reactions.

Therefore, it can be concluded that these reactions are unlikely to occur for the considered

species on the Pt(111) surface.

The microkinetic simulation of propane dehydrogenation on Pt(111) is initially performed at

873 K and total reactant pressure of 0.4 bar (C3H8/H2: 3/1), while varying the simulated time

(0.1 s, 10 s and 109 s). At 0.1 s, the TOF for propylene and hydrogen gas is the highest

(1191 s-1) and the surface is 80 % covered with adsorbates, dominantly 1-propylidyne. At

intermediate simulated time (10 s), the TOF for propane dehydrogenation decreases with 12 %

with respect to 0.1 s simulation time. Furthermore, the coverage of 1-propylidyne starts to

decline and ethylidyne (CH3C≡Pt) and methylidyne (HC≡Pt) are formed. This indicates that 1-

propylidyne is kinetically preferred as adsorbates (fast coke precursors), while the other species

are thermodynamically favored (slow coke precursors). At the very high simulation time (109

s), which represents the deactivated catalyst regime, no catalyst activation for propane

dehydrogenation is reported and the surface is completely covered with slow precursors.

The intermediate simulated time of 10 s is further employed as it still shows high catalytic

activity, while 1-propylidyne is substantially converted to slow coke precursors. While varying

the simulation temperature, a maximum TOF for propylene dehydrogenation is observed at 948

K. However, at this temperature, formation of gaseous side products (methane and ethylene) is

also increased, lowering the selectivity towards gaseous propylene. In addition, the effect of

increasing initial propylene pressure is investigated. It is observed that for increasing propylene

pressure, the TOF for propylene decreases linearly, but the absolute effect is small as only a

10% decrease is reported at 1 bar pressure of propylene. Additionally, the effect of increasing


the pressure of all gaseous side products on the TOF for propylene is negligible. However, the

formation of gaseous ethylene increases as ethane is converted to ethylene on the catalyst

surface.

4.7 References

1. Tilley, R.J., Crystals and crystal structures. 2006: John Wiley & Sons. 2. Downs, R.T. and M. Hall-Wallace, The American Mineralogist crystal structure

database. American Mineralogist, 2003. 88(1): p. 247-250. 3. Yang, M.L., et al., Density functional study of the chemisorption of C-1, C-2 and C-3

intermediates in propane dissociation on Pt(111). Journal of Molecular Catalysis a-Chemical, 2010. 321(1-2): p. 42-49.







10. Lee, A.F., et al., A Fast XPS study of propene decomposition over clean and sulphated Pt{111}. Catalysis Letters, 2002. 78(1-4): p. 379-382.

11. Bonzel, H., Adsorbed Layers on Surfaces. Part 5: Adsorption of molecules on metal, semiconductor and oxide surfaces. Landolt Börnstein, 2006. 1.

12. McMaster, M.C., S.L.M. Schroeder, and R.J. Madix, Molecular propane adsorption dynamics on Pt(110)-(1x2). Surface Science, 1993. 297(3): p. 253-271.

13. McMaster, M.C., C.R. Arumainayagam, and R.J. Madix, Molecular propane adsorption dynamics on Pt(111). Chemical Physics, 1993. 177(2): p. 461-472.


15. Cushing, G.W., et al., C-H Bond Activation of Light Alkanes on Pt(111): Dissociative Sticking Coefficients, Evans-Polanyi Relation, and Gas-Surface Energy Transfer (vol 114, pg 17222, 2010). Journal of Physical Chemistry C, 2010. 114(51): p. 22790-22790.

16. Chesters, M.A., P. Gardner, and E.M. McCash, The reflection-absorption infrared spectra of n-alkanes adsorbed on Pt(111). Surface Science, 1989. 209(1–2): p. 89-99.

17. De Moor, B.A., M.-F. Reyniers, and G.B. Marin, Physisorption and chemisorption of alkanes and alkenes in H-FAU: a combined ab initio-statistical thermodynamics study. Physical Chemistry Chemical Physics, 2009. 11(16): p. 2939-2958.

18. Arumainayagam, C.R., M.C. McMaster, and R.J. Madix, The dynamics of precursor adsorption: ethane on Pt(111). Surface Science, 1990. 237(1–3): p. L424-L431.


19. Arumainayagam, C.R., et al., Dynamics of molecular adsorption of ethane with Pt(111) - a supersonic molecular-beam study. Journal of Physical Chemistry, 1991. 95(3): p. 1041-1047.

20. Chen, Y. and D.G. Vlachos, Hydrogenation of Ethylene and Dehydrogenation and Hydrogenolysis of Ethane on Pt(111) and Pt(211): A Density Functional Theory Study. Journal of Physical Chemistry C, 2010. 114(11): p. 4973-4982.

21. Watwe, R.M., et al., Theoretical Studies of Stability and Reactivity of C2 Hydrocarbon Species on Pt Clusters, Pt(111), and Pt(211). The Journal of Physical Chemistry B, 2000. 104(10): p. 2299-2310.

22. Arumainayagam, C.R., et al., Dynamics of molecular CH4 adsorption on Pt(111). Surface Science, 1989. 222(1): p. 213-246.

23. Qi, Q.H., et al., Methane dissociation on Pt(111), Ir(111) and PtIr(111) surface: A density functional theory study. Applied Surface Science, 2013. 284: p. 784-791.

24. Zhang, R.G., L.Z. Song, and Y.H. Wang, Insight into the adsorption and dissociation of CH4 on Pt(h k l) surfaces: A theoretical study. Applied Surface Science, 2012. 258(18): p. 7154-7160.

25. Fuhrmann, T., et al., Activated adsorption of methane on Pt(111) - an in situ XPS study. New Journal of Physics, 2005. 7.

26. Steinrueck, H.P., et al., A detailed analysis of vibrational excitations in X-ray photoelectron spectra of adsorbed small hydrocarbons. Journal of Chemical Physics, 2006. 125(20).


28. Zaera, F. and D. Chrysostomou, Propylene on Pt(111) II. Hydrogenation, dehydrogenation, and H-D exchange. Surface Science, 2000. 457(1-2): p. 89-108.

29. Tsai, Y.-L., C. Xu, and B.E. Koel, Chemisorption of ethylene, propylene and isobutylene on ordered Sn/Pt (111) surface alloys. Surface science, 1997. 385(1): p. 37-59.

30. Salmeron, M. and G.A. Somorjal, Desorption, decomposition, and deuterium-exchange reactions of unsaturated-hydrocarbons (ethylene, acetylene, propylene, and butenes) on the Pt(111) crystal-face. Journal of Physical Chemistry, 1982. 86(3): p. 341-350.

31. McCrea, K.R. and G.A. Somorjai, SFG-surface vibrational spectroscopy studies of structure sensitivity and insensitivity in catalytic reactions: cyclohexene dehydrogenation and ethylene hydrogenation on Pt(111) and Pt(100) crystal surfaces. Journal of Molecular Catalysis a-Chemical, 2000. 163(1-2): p. 43-53.

32. Essen, J.M., et al., Adsorption of ethene on Pt(111) and ordered PtxSn/Pt(111) surface alloys: A comparative HREELS and DFT investigation. Surface Science, 2007. 601(16): p. 3472-3480.

33. Paffett, M.T., et al., Chemisorption of ethylene on ordered Sn/Pt(111) surface alloys. Surface Science, 1989. 223(3): p. 449-464.

34. Papoian, G., J.K. Nørskov, and R. Hoffmann, A comparative theoretical study of the hydrogen, methyl, and ethyl chemisorption on the Pt (111) surface. Journal of the American Chemical Society, 2000. 122(17): p. 4129-4144.

35. Godbey, D.J. and G.A. Somorjai, The adsorption and desorption of hydrogen and carbon-monoxide on bimetallic Re-Pt(111) surfaces. Surface Science, 1988. 204(3): p. 301-318.

36. Lu, K.E. and R.R. Rye, Flash desorption and equilibration of H2 and D2 on single-crystal surfaces of platinum. Surface Science, 1974. 45(2): p. 677-695.


37. Salmeron, M., R.J. Gale, and G.A. Somorjai, Modulated molecular-beam study of the mechanism of the H2-D2 exchange-reaction on Pt(111) and Pt(332) crystal-surfaces. Journal of Chemical Physics, 1979. 70(6): p. 2807-2818.

38. Bădescu, Ş.C., et al., Energetics and Vibrational States for Hydrogen on Pt(111). Physical Review Letters, 2002. 88(13): p. 136101.

39. Bădescu, Ş.C., et al., Vibrational states of a H monolayer on the Pt(111) surface. Physical Review B, 2003. 68(20): p. 205401.

40. Shen, J.Y., et al., Microcalorimetric, infrared spectroscopic, and DFT studies of ethylene adsorption on Pt/SiO(2) and Pt-Sn/SiO(2) catalysts. Journal of Physical Chemistry B, 1999. 103(19): p. 3923-3934.

41. Sattler, J.J., et al., Catalytic Dehydrogenation of Light Alkanes on Metals and Metal Oxides. Chemical reviews, 2014. 114(20): p. 10613-10653.

42. Liu, Y. and C. Wang. Thermodynamic calculation of isobutane dehydrogenation reaction. in Advanced Research and Technology in Industry Applications (WARTIA), 2014 IEEE Workshop on. 2014. IEEE.

43. Cox, J.D. and G. Pilcher, Thermochemistry of organic and organometallic compounds. 1970.

44. Song, L.J. and L.V.C. Rees, Diffusion of propane in theta-1 and silicalite-1 zeolites. Microporous and Mesoporous Materials, 2000. 41(1-3): p. 193-200.

45. Koestner, R.J., et al., Evidence for the formation of stable alkylidyne structures from C-3 and C-4 unsaturated-hydrocarbons adsorbed on the Pt(111) single-crystal surface. Surface Science, 1982. 116(1): p. 85-103.


47. Vines, F., et al., Methane Activation by Platinum: Critical Role of Edge and Corner Sites of Metal Nanoparticles. Chemistry-a European Journal, 2010. 16(22): p. 6530-6539.

48. Pechukas, P., Transition State Theory. Annual Review of Physical Chemistry, 1981. 32(1): p. 159-177.

49. Wang, H.F. and Z.P. Liu, Comprehensive mechanism and structure-sensitivity of ethanol oxidation on platinum: New transition-state searching method for resolving the complex reaction network. Journal of the American Chemical Society, 2008. 130(33): p. 10996-11004.

50. Sui, Z.-J., et al., Chapter Two - Kinetics of Catalytic Dehydrogenation of Propane over Pt-Based Catalysts, in Advances in Chemical Engineering, B.M. Guy, Editor. 2014, Academic Press. p. 61-125.

51. McQuarrie, D.A. and J.D. Simon, Physical chemistry: a molecular approach. Vol. 1. 1997: University science books Sausalito, CA.

52. Bariås, O.A., A. Holmen, and E.A. Blekkan, Propane Dehydrogenation over Supported Pt and Pt–Sn Catalysts: Catalyst Preparation, Characterization, and Activity Measurements. Journal of Catalysis, 1996. 158(1): p. 1-12.

Chapter 5: Propane dehydrogenation kinetics on Pt3Ga(111) as a model for Pt-Ga catalysts 114

Chapter 5 Propane dehydrogenation kinetics on Pt3Ga(111) as a model for Pt-Ga catalysts Pure platinum catalysts have drawbacks as they show low olefin selectivity and short lifetime due

to deactivation of the catalyst. The addition of metal modifiers to the Pt catalyst enhances the

catalytic properties. Various transition and non-transition metals are employed as alloy element

for pure platinum. In literature, PtxSny-catalysts are intensively researched, both experimentally

and theoretically, to obtain a fundamental understanding of the role of Sn as modifier on the

dehydrogenation of light alkanes and especially propane. [1-9] In this work, the role of Ga addition

on the catalytic properties during propane dehydrogenation is investigated. An innovative

synthesis method facilitates the preparation of Pt-Ga catalysts on a hydrotalcite support

(Mg(Ga)(Al)Ox). [10] Experimental results have shown that under the same reaction conditions,

an increase in activity and selectivity towards propylene are observed with respect to pure platinum

catalysts. [11] Still, Pt-Sn catalysts can be employed as guideline because gallium and tin are both

transition metal close to metal-nonmetal gap in the periodic table.

To construct the catalyst model for Pt-Ga catalysts, the active phase of the Pt-Ga/Mg(Ga)(Al)Ox

catalyst is determined based on DFT calculations of PtxGay catalysts. The selection of alloys is

based on the Ga content (max. 60 mol%) and their stability under both reaction and oxidizing

conditions. Additionally, the vibration modes of adsorbed CO are investigated to couple the

calculated shift on the various alloys with respect to the pure Pt catalyst with the observed

experimental shift of ~5 cm-1. The Pt3Ga(111) catalyst model resulted in best agreement with

experiments. This catalyst model is employed to determine the propane dehydrogenation kinetics

and evaluate the effects of Ga on the catalytic properties.

115 Chapter 5: Propane dehydrogenation kinetics on Pt3Ga(111) as a model for Pt-Ga catalysts

5.1 Catalyst model

Bulk Pt3Ga has a similar fcc-like crystal structure as Pt. Based on the similarity, the high

coordinated terrace plane Pt3Ga(111) is the most abundant plane as it is the most stable plane of

the fcc structure. [12] In this work, it was opted to select a non-segregated bulk alloy of Pt3Ga(111)

as catalyst model. Saerens et al. observed the lowest surface tension in a non-segregated model in

comparison with segregated and anti-segregated models. [13] The Pt(111) catalyst model from

Chapter 4 is employed as basis for the construction of the Pt3Ga(111) catalyst model. The same

4×2 unit cell is used to represent the Pt3Ga(111) catalyst model. For every four Pt atoms, a Ga

atom replaces one. The optimization is conducted with the same Monkhorst Pack grid (5×5×1) and

vacuum layer of 12 Å as in Chapter 4. Strict convergence criteria and the optPBE vdW-DF as

functional are selected. The optimized 4×2 unit cell is shown in Figure 5-1.

Figure 5-1: Isometric representation of the 4×2 unit cell of the Pt3Ga(111) catalyst model

The dimensions of the unit cell are 11.21×5.60×18.8 Å³, which is slightly smaller than the Pt(111)

catalyst model because the atomic radius of Ga is smaller than of Pt. The lattice parameter for this

model is 3.96 Å, which is a small overestimation with the reported value of 3.90 Å in literature.

[12] However, the same discrepancy is observed with the Pt(111) catalyst model. It is expected

that the non-local vdW-DF functional slightly overestimates the lattice parameter of the catalyst

models.

c

a

b



Basically, the same four distinct adsorption sites are still possible (top, bridge, fcc, hcp) with

respect to the nomenclature of 4.1.1. However, the incorporation of Ga leads to multiple sites of

the same type. The characteristics of each site depend on the nearby atoms, so in the site naming,

the locations of gallium and platinum in the top (and if necessary second) layer are incorporated

in the name of site, similar to previous work. [13] The extended nomenclature is reported in Table

5-1, corresponding with the allocated adsorption sites, illustrated in Figure 5-2.

Figure 5-2: Adsorption sites on Pt3Ga(111) catalyst model: top ●, bridge █ , three-folded hcp ▲ and fcc ▼. The numbering of the sites corresponds to the site naming in Table 5-1.

Table 5-1. Adsorption site nomenclature for Pt3Ga(111) catalyst model.

Number Site Number Site Number Site

1 Pt-top 4 Pt2-bridge-Pt2Ga 7 Pt2Ga-fcc

2 Ga-top 5 Pt2-bridge-Pt3 8 Pt2Ga-hcp

3 PtGa-bridge-Pt3 6 Pt3-fcc 9 Pt3-hcp

Previous work has shown that hydrocarbon intermediates do not directly bond with gallium. This

results in a reduction of preferable adsorption sites per unit cell due the incorporation of Ga.

Furthermore, sites that exist of multiple atoms such as the hollow and bridge sites are more affected

by this Ga addition with respect to the single atom sites. In the context of propane dehydrogenation,

the geometrical effect of Ga will lead to less deeply dehydrogenated species as they bond on

multiple-atoms sites.

1 2

3 4

5 6 7

8 9 b

a



The degree of coverage as defined in 4.1.2, is used in this chapter as well. No distinction is made

between the Pt and Ga atoms in the surface layer. Therefore, the coverage is 0.13 ML in the case

of one adsorbate molecule.

5.2 Adsorption

In this section, the adsorption of the main gaseous reactants and products is discussed. These

compounds are propane, propylene and H2. Beside them, gaseous products such as ethane, ethylene

and methane are formed via cracking reaction of C3-species. However, these species are excluded.

The optimized geometries of the adsorbates can be found in Appendix D.

Most of the ab initio studies on modified Pt-based catalysts are conducted on Pt-Sn catalyst models

and none on Pt-Ga catalyst models. The Pt-Sn catalyst models for propane dehydrogenation are

used as guideline for Pt-Ga catalysts. The Pt3Sn catalyst model is the most interesting for this work

as it has the same alloy composition as the Pt3Ga catalyst model.

The adsorption energies are determined based on the equation formulated in 4.2.

5.2.1 Propane

The adsorption of propane is the first step of the reaction mechanism for propane dehydrogenation.

As a saturated compound, propane can adsorbed either via physisorption or dissociative

chemisorption. The first is the topic in this section as no internal bond breaking occurs in this

adsorption mode. Physisorbed propane acts as a precursor prior the dissociative chemisorption

step.

Conform the propane physisorption in Chapter 4, propane retains its gaseous structure. The C-C

bond lengths are 1.53 Å, which are identical to the bond lengths of gaseous propane. It is observed

that the physisorption height (shortest distance Pt-C) is 4.0 Å, which is 0.2 Å shorter with respect

to the observed height on Pt(111).

Nykänen et al. and Yang et al. observe that alloying with Sn increases the adsorption energies with

respect to pure Pt. A decrease of ~4 kJ/mol is calculated for both DFT studies, so Sn destabilizes

slightly the physisorbed species. Yang et al. also observe that the adsorption height increases due

the Sn impregnation. In contrast, the reverse is observed for Ga as modifier element. The

physisorption on Pt3Ga(111) is stabilized with 8 kJ/mol and, as previously mentioned, a shorter


adsorption height is found. These effects show that Ga addition stabilizes the physisorbed species

better than their Pt-Sn counterpart.

Table 5-2. Adsorption energies of propane on Pt(111), Pt3Sn (111) and Pt3Ga(111). Yang et al. and Nykänen et al. used a vdW-DF functional. This work uses an optPBE vdW-DF functional.

ΔEads

(kJ/mol) Nykänen et al. (0.13 ML) [4]


Yang et al. (0.08 ML) [7]

This work

(0.13 ML) Pt(111) -33 -36 -41 -43

Pt3Sn(111) -27 -33 -37 -

Pt3Ga(111) - - - -51

In this work, vibrational analysis of physisorbed propane resulted in two imaginary frequencies.

These are expected as physisorbed species have a higher degree of freedom with respect to

adsorbates. The vibration modes are assigned to translation and external rotation.

5.2.2 Propylene

As propylene is the main product, the description of propylene adsorption and desorption is

essential. At low propylene coverage (θ < 0.25 ML), di-σ propylene is the most stable adsorption

mode. Other adsorption modes are formed at higher propylene coverages. Π-adsorption mode is

an example of such a mode and is weakly bonded to the catalyst surface via its π-bond. [14]

In this work, solely the di-σ adsorption mode is investigated as this mode is more stable than π-

propylene. Conform 4.2.2.1, di-σ propylene loses its planar geometry upon adsorption and its C=C

bond is elongated from 1.34 Å in the gasphase to 1.49 Å adsorbed. This length is more similar to

C-C bond length of 1.53 Å. These effects indicate a covalent interaction between propylene and

the catalyst surface.

The adsorption energies of di-σ propylene are compared with the DFT studies of Nykänen et al.

and Yang et al, see Table 3. As discussed in Chapter 4, the discrepancy of propylene adsorption

energy on Pt(111) between this work and other studies can be based on the expected overbinding

of the employed optPBE vdW-DF functional. On Pt3Sn(111), propylene is less stabilized. The

resulted adsorption energies are increased with ~30 kJ/mol with respect to Pt(111). A deviation in

the energy difference is observed for the work of Yang et al. as both the Sn addition and different

coverages play a role. On Pt3Ga(111), propylene is slightly more stabilized. The absolute effect of

Ga (-4 kJ/mol) is smaller than the effect of Sn (30 kJ/mol)


Table 5-3. Adsorption energies of di-σ propylene on Pt(111), Pt3Sn (111) and Pt3Ga(111). Yang et al. and Nykänen et al. used respectively a PBE and vdW-DF functional. This work employs an optPBE vdW-DF functional.

ΔEads

(kJ/mol) Nykänen et al. (0.13 ML) [4]

Nykänen et al. (0.25 ML) [4] Yang et al. [7, 15] This work

(0.13 ML) Pt(111) -72 -80 -94 (0.25 ML) -131

Pt3Sn(111) -40 -52 -48 (0.08 ML) -

Pt3Ga(111) - - - -135

Beside di-σ adsorption mode, the physisorption of propylene above the catalyst surface is

investigated. This species is weakly bonded with the surface and is predominantly stabilized by

van der Waals interactions. The geometry of physisorbed propylene is similar on Pt(111) and

Pt3Ga(111), as it retains the structure of its gaseous form. However, a shorter adsorption height is

observed with respect to Pt(111), similar to propane physisorption. On Pt(111), the distance is 3.8

Å, while on Pt3Ga(111) the height is reduced to 3.3 Å.

The smaller adsorption height on Pt3Ga(111) indicates that propylene physisorbs stronger on the

Pt3Ga(111) model. This is confirmed by the calculated adsorption energies; on Pt3Ga(111) the

adsorption energy is -63 kJ/mol, while on Pt(111) a value of -43 is found for ΔadsE of physisorbed

propylene.

5.3 Thermodynamics

To describe propane dehydrogenation on Pt3Ga(111), the same reaction network as proposed in

4.3 is employed. However, some reactions are excluded as mainly the catalytic activity and

selectivity towards propylene with respect to deactivation reactions are focused on. Hence, the

main topics are the reaction pathways towards propylene and deep dehydrogenation of propylene

and other C3H6-species. Dissociation reactions (C-C scission reactions) are evaluated, initially

solely of propane as on Pt(111) these reactions are unfavorable. The reactions that lead to gaseous

side products such as ethane, ethylene and ethane are not considered

The considered species are listed in Table 5-4 and a streamlined version of the calculated reaction

network is shown in Figure 5-3. For each component, the most stable geometry on the Pt3Ga(111)

catalyst model is calculated and a frequency analysis is conducted to evaluate its stability.

Furthermore, the optimal geometries can be found in Appendix D, together with a visualized

reaction network based on those optimal geometries.


Table 5-4. Adsorbed hydrocarbon species on the Pt3Ga(111) surface in descending order of molecular weight. * indicates with how many C-Pt bonds the species is adsorbed to the surface.

CH3-CH2-CH3 Propane,phys CH3-CH2-CH2* 1-Propyl

CH3-CH*-CH3 2-Propyl CH3-CH*-CH2* Propylene

CH3-CH=CH2 Propylene,phys CH3-CH2-CH** 1-Propylidene

CH3-CH**-CH2 2-Propylidene CH3-CH2-C*** 1-Propylidyne

CH3-CH*-CH** 1-Propenyl CH3-C**-CH2* 2-Propenyl

CH3-CH2* Ethyl CH3* Methyl

Figure 5-3: Reaction network for propane dehydrogenation on Pt3Ga(111).


(5.3.1), deep dehydrogenation of propylene (5.3.2) and hydrogenolysis of C3 intermediates 5.3.3).

In each section, reaction energies are determined and the stability of each intermediate is evaluated,

2-propyl

1-propylidene

CH2=CH-CH3,phys

CH2=CH-CH3(g)

13

_CH-CH-CH3 + H

PtPt2

_

Pt

= CH2-C-CH3 + H

Pt

_

Pt

_

Pt2

=


C-CH2-CH3 + H

Pt3

_

Pt

≡

14 1517 20 21

Methyl and ethyl

4 5

= _CH-CH2-CH3 + H

Pt2 Pt2-propylidene

CH3-C-CH3 + H

Pt2

=

Pt

_

7 8

CH3 + CH2CH3

Pt Pt

_ _CH3-CH-CH3 + H

Pt Pt

_ _CH2-CH2-CH3+ H

Pt Pt_ _

1-propyl

1 2 3

CH3-CH2-CH3,phys


0

Physisorbed propane

CH2-CH-CH3 + H

Pt

__

Pt

_

PtPropylene

11 12

1-propylidyne


Gaseous propylene


conform Chapter 4. The activity of the catalyst is investigated in section 5.3.1, while the selectivity

towards propylene with respect to the most stable coke precursors is discussed in section 5.3.2, as

propylene desorption is compared with deep dehydrogenation of propylene. Section 5.3.3

determines the selectivity to other side products, however on Pt(111) it was observed that these

reactions are kinetically unfavorable, so solely the dissociation of propane is selected as model

reaction for this section and will be discussed.

Three main reaction types are distinguished: dehydrogenation, isomerization and dissociation

reactions. The focus will be mainly on the dehydrogenation reactions as the C-H bond breaking

has the lowest reaction barriers on Pt(111) with respect to the isomerization and dissociation. Due

the low hydrogen coverage and the repulsion between the hydrocarbon intermediates and

fragmented hydrogen, it is expected the adsorbed hydrogen diffuses away from the hydrocarbon

intermediates after the dehydrogenation reactions. For the description of the thermodynamics, the

products of dehydrogenation reactions are optimized in different unit cells. The equations

proposed in 4.3 are employed to describe the reaction energies of the considered reactions.

The hydrogen in the separate unit cell orientates itself most favorable in the Pt3-hcp site with Ga

in the sublayer. The most stable adsorption site i.e. Pt3-hcp-Ga is determined with respect to other

considered adsorption sites. It is observed that hydrogen does not directly adsorb on a Ga atom so

it is expected that neighboring and sublayer Ga atoms induce an electronic effect on the Pt atoms,

thereby increasing the stabilization of hydrogen on the Pt3Ga(111) surface. Due these electronic

effects, the most stable adsorption site for hydrogen is different on Pt3Ga(111) than on the Pt(111)

catalyst model. The calculated adsorption energy for dissociative adsorption of H2 is -102 kJ/mol,

which is 25 kJ/mol more stable than on Pt(111).


Two main reaction pathways are possible towards propylene: via 1-propyl and via 2-propyl. Each

path consists of two elementary steps (three if the physisorption of propane is included). The first

elementary step dehydrogenates a hydrogen at either the methyl or methylene group, respectively

generating 1-propyl or 2-propyl. The most stable adsorption of these species is located on top site

of a platinum atom. Consequently, these species form propylene via β-dehydrogenation in the

second elementary step. Propylene orientates itself most stable over a bridge site, which consists

of two Pt atoms. Aside from the β-dehydrogenation towards propylene, 1-propyl and 2-propyl can

also react further via α-dehydrogenation towards respectively 1-propylidene and 2-propylidene.

Both species adsorb both on a Pt2-bridge-Pt2Ga site.


Each step generates also a detached hydrogen. Its location plays a major role to determine the

kinetics as the activation energy is strongly influenced by the (de-)stabilization due to this

hydrogen. However, for the thermodynamics, it is assumed that hydrogen is highly mobile on the

surface.

The optimal geometries on Pt3Ga(111) are similar to those on Pt(111) so Ga induces no direct

geometrical changes in the intermediates, similar findings are reported for the Pt3Sn(111) model.

[7] It is also observed that the hydrocarbons cannot be stabilized by binding with Ga, as no Ga-C

bonds are formed. Vibrational analysis of the intermediates have shown that all intermediates,

except for the physisorbed species, have zero imaginary frequencies.

In literature, DFT studies report that the addition of Sn to Pt catalysts reduces the stability of

intermediates such as 1-propyl and 2-propyl. The same trend was already observed for adsorbed

propylene and propane. [7] However, the addition of Ga shows a different effect and stronger

binding with the surface is reported (see Table 5). For both reaction pathways, a similar reaction

energy is observed for the first elementary step on Pt(111) and Pt3Ga(111). With respect to Pt(111),

1-propyl and 2-propyl have a similar electronic energy on Pt3Ga(111), as the difference is only 2

kJ/mol in favor of 2-propyl, while on Pt(111) this is 7 kJ/mol. It is then also expected that both

reaction paths occur on the surface. However, the electronic effect due to Ga alloying is observed

in the second dehydrogenation step. On both catalyst models, propylene is thermodynamically

preferred, as it is the strongest adsorbate bonded with the surface. On Pt3Ga(111), the reaction

energies of the second elementary step are more exothermic with respect to Pt(111).


Table 5-5. Reaction energies of propane dehydrogenation towards propylene on Pt(111) and Pt3Ga(111). The energies are determined with an optPBE vdW-DF functional.

ΔEr (kJ/mol) This work (0.13 ML)

# Surface reaction Pt(111) Pt3Ga(111)

0 Propane(g) → Propane,phys Physisorption -43 -51

1 Propane,phys → 1-propyl + H Adsorption -7 -7

2 Propane,phys → 2-propyl + H Adsorption -14 -9

4 1-propyl → 1-propylidene + H Dehydrogenation 7 -4

5 1-propyl → propylene + H Dehydrogenation -22 -40

7 2-propyl → propylene + H Dehydrogenation -15 -38


10 1-propyl → 2-propyl Isomerization -7 -2

11 1-propylidene → propylene Isomerization -29 -36

12 2-propylidene → propylene Isomerization -32 -44

13a Propylene→ propylene,phys Desorption 85 71

13b Propylene,phys →Propylene (g) Desorption 49 63

Based on these intermediates, propylene is thermodynamically favored as can be seen in the

relative energy profile (see Figure 5-4). While the electronic energy difference is small, 2-propyl

is slightly favored as reaction intermediate towards propylene. With respect to Pt(111), all C3H6-

species are stronger bonded to the surface. Propylene is the most stable species and is 26 kJ/mol

stronger bonded to the Pt3Ga(111) surface than on Pt(111). Ga alloying affects 1-propylidene and

2-propylidene less in this catalyst model, as a decrease of respectively 19 kJ/mol and 12 kJ/mol in

stability is reported.


Figure 5-4: Energy profile of adsorbed C3Hx (x= 6–8) species on the Pt3Ga(111) surface at 0.13 ML. Energies are determined relative to gaseous propane. Dehydrogenated hydrogen is optimized in separate unit cell.

Conform Chapter 4, to validate the conducted calculations on propane dehydrogenation towards

propylene, the catalytic cycle of this reaction mechanism is evaluated. Hence, the electronic energy

difference for gaseous propane dehydrogenation has to be identical to total electronic energy

difference for the surface reactions on the catalyst. In the gasphase, the reaction energy is 139

kJ/mol. The catalytic reaction paths via 1-propyl or 2-propyl have both a total reaction energy of

138 kJ/mol. The 1 kJ/mol difference can be neglected due to rounding of the numbers and the

reaction pathways have the same total reaction energies.

5.3.2 Deep dehydrogenation of propylene


the selectivity towards propylene with respect to deactivation of the catalyst during propane

dehydrogenation. Hence, the desorption reaction of propylene competes with deep

dehydrogenation reactions that form coke precursors and eventually deactivate the catalyst. Three

deeply dehydrogenated species are investigated: 1-propylidyne, 1-propenyl and 2-propenyl.

Formation of 1-propylidyne occurs via α-dehydrogenation of 1-propylidene, while 1-propenyl is

either formed via β-dehydrogenation of 1-propylidene or via dehydrogenation of the methylene

group of propylene. 2-propenyl is also formed via two paths: dehydrogenation of the methylidyne

-120

-100

-80

-60

-40

-20

0

20

40

60R

elat

ive

ener

gy (k

J/m

ol)

C3H8 (g) C3H8, phys C3H7* + H* C3H6

* + 2H* C3H6 (g) + 2H*

Propane (g)

Propane, phys

2-propyl

Propylene,phys

Propylene

Propylene (g)

1-propyl

1-propylidene

2-propylidene

C3H6, phys + 2H*


of propylene and via β-dehydrogenation of 2-propylidene. Reaction energies of these reaction can

be found in Table 6.

Table 5-6. Reaction energies of deep dehydrogenation reactions of C3H6-species on Pt(111) and Pt3Ga(111). The energies are determined with an optPBE vdW-DF functional.

ΔEr (kJ/mol) This work (0.13 ML)




17 1-propylidene → 1-propylidyne + H Dehydrogenation -67 -42

20 1-propylidene → 1-propenyl + H Dehydrogenation -10 -27


1-propylidyne, 1-propenyl and 2-propenyl all three adsorb on a hollow three-folded site. In the

case of 1-propylidyne, one carbon atom directly bond with three Pt atoms, while for 1-propenyl

and 2-propenyl, one carbon atom single bonds with one Pt atom and a second carbon atom bonds

with two Pt atoms. On Pt(111), the most stable adsorption site is in a fcc site. However, on

Pt3Ga(111), the most stable adsorption site is shifted to a hcp site with underneath a Ga atom. The

electronic effect induced by the sublayer Ga atom stabilizes the deeply dehydrogenated species.

With respect to each other, 1-propylidyne is the most stable C3H5-species, conform Chapter 4.

However, Ga alloying stabilizes 1-propylidyne less and an increase in reaction energy of 25 kJ/mol

is observed compared to the pure Pt(111) catalyst. In contrast, for the other two C3H5-species, a

decrease (of ~16 kJ/mol) in reaction energies is observed, consistent with the previous species (see

5.3.1). Thermodynamically, 1-propylidyne is still the most favorable species, however, the

electronic energy difference between 1-propylidyne and propylene is smaller on Pt3Ga(111) (-6

kJ/mol) with respect to Pt(111) (-38 kJ/mol). This smaller energy difference reduces the

thermodynamic driving force towards 1-propylidyne and indicates that the selectivity towards

propylene is improved with respect to the pure Pt catalyst model.


Figure 5-5: Energy profile of adsorbed C3Hx (x= 5–8) species on the Pt3Ga(111) surface at 0.13 ML. Energies are determined relative to gaseous propane. Dehydrogenated hydrogen is optimized in separate unit cell.


Cracking of C3-species leads to C1 and C2 species on the surface. Further surface reactions of these

species either lead to gaseous side products such as ethane, ethane, and ethylene or to coke

precursors and eventually cokes. These reactions mainly affect the selectivity towards propylene,

as other undesirable gaseous products are formed. On Pt(111), it was shown that hydrogenolysis

is solely favorable for deeply dehydrogenated species, as in general higher reaction barriers are

obtained for C-C cleavage than C-H bond breaking. On Pt3Ga(111), the dissociation of propane is

primarily evaluated as a model reaction for C-C bond breaking during propane dehydrogenation.

In this reaction, propane dissociates into ethyl and methyl. Both adsorb on top of a Pt atom. The

reaction products are optimized in the same unit cell. The optimized geometry on Pt3Ga(111) is

similar on Pt(111). The bond lengths are slightly elongated on this model, indicating weaker

adsorption on the surface. The calculated reaction energy is 14 kJ/mol, which is 19 kJ/mol larger

than on Pt(111), confirming the weaker adsorption. However, methyl and ethyl are optimized in

separate unit cells on Pt(111), eliminating repulsion between the dissociated products. Therefore,

no clear conclusions can be drawn for the difference between Pt(111) and Pt3Ga(111) in terms of

hydrogenolysis reaction energies.

-110

-90

-70

-50

-30

-10

10R

elat

ive

ener

gy (k

J/m

ol)

C3H8 (g) C3H8* C3H7

* + H* C3H6* + 2H* C3H5

* + 3H*

Propane (g)

Propane, phys

2-propyl

Propylene,phys

Propylene

1-propenyl

1-propyl

1-propylidene

2-propylidene

2-propenyl

1-propylidyne


5.4 Kinetics

To determine the kinetic descriptors for the considered reactions, the optimized geometry and

electronic energy of transition state are calculated. Each determined transition state is validated by

vibrational analysis, as every transition state should have one imaginary frequency of which the

mode is along the reaction coordinate. The resulted geometries can be found in Appendix D.

Yang et al. investigate the role of Sn on the propane dehydrogenation kinetics. The determination

of reaction barriers is focused on the elementary steps towards propylene, the dehydrogenation of

propylene and hydrogenolysis of dehydrogenated species. Pt(111) and various Pt-Sn catalyst

models are employed in the calculations, however the results on Pt(111) and Pt3Sn(111) are the

most interesting for this work. The resulted energy barriers on Pt3Sn(111) are all larger than on

Pt(111). Furthermore, the largest increase is reported for the deep dehydrogenation reactions

towards 1- and 2-propenyl. Yang et al. conclude that the addition of Sn lowers the catalytic activity

for propane dehydrogenation, while a higher propylene selectivity with respect to deeply

dehydrogenated species is observed. [7]

In the following sections, the quantification of the role of Ga on the catalytic activity (5.4.1) and

the selectivity towards propylene (5.4.2 and 5.4.3) is the point of focus. The selectivity towards

propylene is evaluated with respect to deeply dehydrogenated species (5.4.2) and gaseous side

products, formed via hydrogenolysis (5.4.3).


As mentioned above, the focus lays on the role of Ga on the catalytic activity during propane

dehydrogenation towards propylene. Conform Chapter 4, it is assumed that the

adsorption/desorption steps are not activated, however during the desorption step, the desorption

energy must be overcome, so this value is set equal to the electronic activation energy for

desorption steps. For the first elementary step, the reaction barriers for 1- and 2-propyl formation

are calculated. As the electronic activation energies are identical, both reaction paths are likely to

occur. However, for the further determination of the activity, the reaction path via 2-propyl is

selected as on both Pt(111) and Pt3Ga(111) 2-propyl is electronically more stable than 1-propyl.

For the second elementary step, solely the β-dehydrogenation of 2-propyl is considered.

The calculated transition state energies (as shown in Table 5-7) are converged under strict criteria

and have one unique imaginary frequency with a vibrational mode along the reaction coordinate.


Table 5-7. Electronic activation energies for propane dehydrogenation towards propylene on Pt(111) and Pt3Ga(111). The energies are determined with an optPBE vdW-DF functional.

Δ‡E (kJ/mol) This work (0.13 ML)








To validate the role of Ga on the catalytic activity, the reaction path via 2-propyl towards propylene

is compared on Pt(111) and Pt3Ga(111). This is illustrated in Figure 5-6. For the energy profile,

all energies are determined relative to gaseous propane. To satisfy the conservation laws, the

energy of adsorbed hydrogen(s) is added in the case of dehydrogenated species. In the last steps,

all adsorbates (propylene and H2) desorb and the final energy should be the same for both catalyst

models, as the energies of those products are independent of the catalyst model.

All considered intermediates are stronger bonded on Pt3Ga(111) with respect to Pt(111).

Furthermore, lower reaction barriers are calculated on Pt3Ga(111), indicating that propane

dehydrogenation is less activated on Pt3Ga(111) than on Pt(111). The electronic energies of the

transition states downshift with at least 11 kJ/mol for these steps. The stabilizing effect of Ga is

the largest for di-σ propylene, as it is 27 kJ/mol more stable on Pt3Ga(111) than on Pt(111).

However, it is expected that the desorption energy of propylene is increased with the same amount,

but in contrast, the desorption energy for propylene to the gasphase is identical on both catalyst

models. It is the desorption energy of hydrogen that has become more endothermic on the

Pt3Ga(111) catalyst model compared to Pt(111).

Hence, the addition of Ga decreases the activation barriers of the elementary steps for propane

dehydrogenation towards propylene, without stronger propylene adsorption with respect to its

gaseous form. It can be concluded that Ga enhances the catalytic activity, as was proposed by

experiments by Siddiqi et al. [11]. In contrast, it was shown that alloying Pt with Sn lowered the

catalytic activity. [7] In terms of activity, alloying with Ga is a better choice in contrast to alloying

with Sn.


Figure 5-6 : Relative energy profile for the propane dehydrogenation towards propylene on Pt(111) (black) and Pt3Ga(111) (green). The energies are determined relative to gaseous propane.


Deeply dehydrogenated species are considered coke precursors, as these species are strongly

bonded to the surface and further reactions lead to coke formation and eventually to deactivation

of the catalyst. To quantify the selectivity of propylene with respect to these coke precursors, two

reaction paths are compared. The first path consists of all elementary steps towards gaseous

propylene as discussed in previous section. 2-propyl is selected as intermediate for this reaction

path, as it is slightly more stable than 1-propyl on Pt3Ga(111). The second path consists of the

elementary steps towards the most stable deeply dehydrogenated species i.e. 1-propylidyne. The

reaction barrier is calculated for the following elementary steps: α-dehydrogenation of 1-propyl to

1-propylidene and the consecutive dehydrogenation of 1-proylidene to 1-propylidyne. The

activation energies for this path can be found in Table 8.

-100

-75

-50

-25

0

25

50

75

100

125

150

Rel

ativ

een

ergy

(kJ/

mol

)

Propane (g)



Propylene +2H*


TS2 TS7

Reactant Product

Propylene (g)+ 2H*


Table 5-8. Electronic activation energies for deep dehydrogenation reaction pathway on Pt(111) and Pt3Ga(111). The energies are determined with an optPBE vdW-DF functional.

Δ‡E (kJ/mol) This work (0.13 ML)


4 1-propyl → 1-propylidene + H Dehydrogenation 83 94†

17 1-propylidene → 1-propylidene +H Dehydrogenation 45† 69† † On Pt(111), vibrational analysis resulted in an additional imaginary frequencies. On Pt3Ga(111), the electronic energies of the

transition states are based on the NEB results.

The resulted electronic activation energies are considerably larger on Pt3Ga(111) than on Pt(111),

indicating that this reaction path is less favored on Pt3Ga(111). However, to evaluate both reaction

paths, an energy profile is constructed, relative to gaseous propane (see Figure 5-7). To satisfy

conservation laws, the energies of adsorbed hydrogen(s) are added to the dehydrogenated species.

In the final steps, all adsorbates are desorbed, however as 1-propylidyne cannot desorb, solely its

dehydrogenated hydrogens are desorbed.

Figure 5-7: Relative energy profile for the propane dehydrogenation towards propylene (···) and propane deep dehydrogenation towards 1-propylidyne ( ̶ ̶ ) on Pt3Ga(111). The energies are determined relative to gaseous propane.

The competition between these reaction paths is centered at the desorption step and

dehydrogenation to 1-propylidyne. For the first elementary step, both intermediates are equally

-125

-75

-25

25

75

125

Rea

ltive

ene

rgy

(kJ/

mol

)

Reactant Product

Propane (g)

Propane, phys

2-propyl + H*


Propylene + 2H*


TS1 TS2 TS7 Propylene (g) + 2H*

1-propyl + H*

TS4

1-propylidene+ 2H*

1-propylidyne + 3H*


TS17


likely to be formed. In the second step, propylene is more likely to be formed as both its reaction

energy and barrier is lower than 1-propylidene. However, since 1-propylidyne is 6 kJ/ more stable

on the surface than propylene, this species is preferentially formed.

However, in this electronic energy profile (Figure 5-7), solely enthalpic contributions are taken

into account. As discussed in 4.3.2, the Gibbs free energy accounts for both the enthalpic and

entropic contributions. On Pt(111), the Gibbs free energy diagram at 900 K showed that gaseous

propylene was preferred with respect to 1-propylidyne because its entropic contributions in the

gaseous state are much larger than adsorbates such as 1-propylidyne due the higher degree of

freedom. Furthermore, the description of the Gibbs free energy on Pt(111) can be extended to

Pt3Ga(111). The Gibbs free energy of gaseous propylene is independent of the catalyst model,

however this is not the case for adsorbates such as 1-propylidyne. Since similar geometries were

obtained for the adsorbates on Pt3Ga(111) and Pt(111), it can be assumed that the entropic

contributions are also similar. Based on this reasoning, the Gibbs free energy of 1-propylidyne on

Pt3Ga(111) can be estimated based on the Gibbs free energy diagram on Pt(111). As the entropic

contributions are similar on both catalyst models, the electronic energy difference of 1-propylidyne

between Pt(111) and Pt3Ga(111) determines the Gibbs free energy of 1-propylidyne on

Pt3Ga(111). The electronic energy of 1-propylidyne is higher on Pt3Ga(111) than on Pt(111), so it

is expected that Gibbs free energy of 1-propylidyne is also higher. As the Gibbs free energy

difference between 1-propylidyne and gaseous propylene is higher on Pt3Ga(111), it can be

concluded that Pt3Ga(111) is more deactivation resistant than Pt(111) as it is less likely that coke

precursors are formed on Pt3Ga(111).


The cracking of the C-C bond of C3 intermediates leads to formation of smaller gaseous

hydrocarbons such as methane, ethane and ethylene. The description on the C-C cleavage is thus

essential for the propylene selectivity with respect to other gaseous side products. Siddiqi et al.

have reported that the selectivity towards propylene is above 98% on Pt-Ga/Mg(Ga)(Al)Ox for

total time on stream, while for Pt/Mg(Al)Ox the selectivity towards increases during the reaction,

but never pass a propylene selectivity of 87 %. [11] This indicates that pure Pt catalyst are more

prone to hydrogenolysis than Pt-Ga catalysts. Via the hydrogenolysis of propane, the formation of

these gaseous side products is evaluated. The resulted electronic activation energy is 191 kJ/mol,

which is 136 kJ/mol more than the dehydrogenation reactions of propane. Conform Pt(111), it is

expected that C-C cleavage is unfavorable with respect to the breaking of C-H bonds. The


considered dissociation reaction is even higher activated on Pt3Ga(111) than on Pt(111). This

confirms that on Pt3Ga(111) dissociation of propane is unlikely to occur and inherently inhibits

the formation of gaseous side products.

5.5 Conclusions

In this chapter, the characteristics of propane dehydrogenation on Pt-Ga catalysts is evaluated. In

a previous combined experiment-and-theory study, the Pt3Ga(111) surface cut is assigned as the

most probable surface composition of the novel Pt/Mg(Ga)(Al)Ox catalyst. Hence, a 4×2 unit cell

of is employed for the Pt3Ga(111) catalyst model. To determine the role of Ga on the activity and

propylene selectivity during propane dehydrogenation, our results are compared with literature on

pure Pt catalysts and literature on Pt3Sn catalysts.

Propane dehydrogenation towards propylene consists of two elementary steps and occurs via two

reaction pathways (via 1-propyl and via 2-propyl). The reaction barriers of the two elementary

steps towards propylene define the catalytic activity. Literature reports that on Pt3Sn(111), the

transition states are destabilized with respect to Pt(111) and hence higher reaction barriers are

observed. The addition of Sn lowers the catalytic activity for propane dehydrogenation. In contrast

on Pt3Ga(111), both 1-propyl are 2-propyl are more stabilized than on Pt(111) and as 2-propyl is

thermodynamically more stable than 1-propyl, the reaction pathway via 2 propyl is selected to

define the catalytic activity on Pt3Ga(111). Furthermore, adsorbed propylene bonds stronger on

the Pt3Ga(111) surface and the calculated reaction barriers for this reaction path are lower than on

Pt(111), while the propylene desorption energy remains constant (see Figure 5-6). It can be

concluded that catalytic activity is increased on Pt3Ga(111).

Furthermore, the deep dehydrogenation of C3H6-species is investigated. The formation of these

species are an indication for the deactivation of the catalysts and can be used as a descriptor for

the selectivity of the reaction. Conform Pt(111), 1-propylidyne is thermodynamically the most

favored species on the Pt3Ga(111) surface. To quantify the selectivity to the coke precursors, the

reaction path via 1-propyl and 1-propylidene towards 1-propylidyne is calculated. This reaction

path has higher reaction barriers on Pt3Ga(111) than on Pt(111), see Table 5-8 and 1-propylidyne

is not kinetically favored as on Pt(111) (see Figure 5-7). Furthermore, the electronic energy

difference between propylene and 1-propylidyne on Pt3Ga(111) is much smaller than on Pt(111).

It can be assumed that the entropic contributions for adsorbed species such as 1-propylidyne are

similar on Pt3Ga(111) as on Pt(111). Based on the Gibbs free energy, it is expected that 1-


propylidyne is less likely to be formed on Pt3Ga(111) than on Pt(111). This indicates that fewer

coke precursors will be formed and that Pt3Ga(111) has a higher stability than Pt(111).

Finally, the hydrogenolysis of propane is studied as this reaction forms C1 and C2 adsorbates, which

are critical for the formation of gaseous side products such as ethane, ethylene and methane.

However, this reaction is highly activated (191 kJ/mol) and the formation of gaseous side products

via this way is unlikely. This result is supported by the experimental results that show that the

propylene selectivity is above 98% for the total time on stream.

5.6 References

1. Feng, J., M. Zhang, and Y. Yang, Dehydrogenation of Propane on Pt or PtSn Catalysts with Al2O3 or SBA-15 Support. Chinese Journal of Chemical Engineering, 2014. 22(11–12): p. 1232-1236.

2. Koel, B.E., Structure, Characterization and Reactivity of Pt–Sn Surface Alloys, in Model Systems in Catalysis. 2010, Springer. p. 29-50.

3. Li, Q., et al., Kinetics of propane dehydrogenation over Pt–Sn/Al2O3 catalyst. Applied Catalysis A: General, 2011. 398(1–2): p. 18-26.





8. Zhang, Y., et al., Propane dehydrogenation on PtSn/ZSM-5 catalyst: Effect of tin as a promoter. Catalysis Communications, 2006. 7(11): p. 860-866.

9. Zhu, H., et al., Sn surface-enriched Pt–Sn bimetallic nanoparticles as a selective and stable catalyst for propane dehydrogenation. Journal of Catalysis, 2014. 320(0): p. 52-62.



12. Schweitzer, A., et al., Pt3Ga: Thermodynamics and nonstoichiometry. Zeitschrift für Naturforschung B, 2004. 59(9): p. 999-1005.




Chapter 6: Propane dehydrogenation kinetics on graphene ribbon covered Pt(111) as a model for cokes deactivated Pt catalysts 134

Chapter 6 Propane dehydrogenation kinetics on graphene ribbon covered Pt(111) as a model for cokes deactivated Pt catalysts The coke formation on platinum catalyst plays a major role in the alternation of catalytic

properties of Pt catalysts during propane dehydrogenation. Initial coke formation starts rapidly

at highly undercoordinated sites such as steps and edges as those are prone to C-C cleavage.

When those sites are deactivated, the coke formation reaches it lower steady state rate as coke

formation continues on Pt(111) and the support. In literature, it is proposed that near cokes

deactivated Pt atoms, gaseous side products are less probable to be formed than on a clean Pt

catalyst. [1] Furthermore, the influence of cokes deactivated Pt atoms on propane

dehydrogenation characteristics in terms of catalytic activity and selectivity towards coke

precursors remains elusive.

To construct an appropriate catalyst model for the cokes deactivated Pt catalyst model, the

location and structure of the coke in the model has to represent the nature of the cokes,

determined experimentally. Literature has reported that only a small part of coke deposits is

formed on the active metal particles. The majority either is formed on the support or spills over

from the active metal phase to the support. The characterization of the coke deposits on the

platinum surface shows the aromatic nature of the cokes. Based on TEM experiments, an

interlayer space of the carbon deposits of 3.66 Å is found. This distance corresponds to the

interlayer distance of graphene. [2-5]

On the graphene covered Pt(111) catalyst model, the thermodynamics and kinetics of propane

dehydrogenation are calculated, with focus on the activity and selectivity during propane

Chapter 6: Propane dehydrogenation kinetics on graphene ribbon covered Pt(111) as a model 135 for cokes deactivated Pt catalysts

dehydrogenation. For the activity, the reaction barriers and energies of the most dominant path

towards the desired product propylene will be studied. For selectivity, the difference in reaction

barriers and energies between the most dominant path toward propylene and the most important

side reaction towards coke precursors will be investigated.

6.1 Catalyst model

Graphene is used as model compound to represent coke on the cokes deactivated Pt(111)

catalyst model. Furthermore, it was opted to start from the 4×2 Pt(111) catalyst model (see

Chapter 4.1). A graphene ribbon is constructed on a 2×2 island of Pt atoms. In the unit cell, the

graphene ribbon itself consist of eight carbon atoms and at its periphery, four hydrogen atoms

are added. The hydrogen atoms are added because during propane dehydrogenation hydrogen

is co-fed. Consequently, in these hydrogen-rich conditions, the graphene ribbon will most likely

be hydrogen-terminated. In the optimization, the strict convergence criteria and the non-local

optPBE vdW-DF functional is employed. The optimized unit cell is shown in Figure 6-1. The

catalyst model of the cokes deactivated Pt(111) catalyst model will be abbreviated as

Gr/Pt(111).

Figure 6-1: Isometric representation of the 4×2 unit cell of the cokes deactivated Pt(111) catalyst model

c

a

b


Hence, the dimensions of the unit cell are 11.29×5.64×18.91 Å³, including a vacuum layer in

c-direction. This is the same as for the Pt (111) catalyst model without cokes (see Chapter 4).

The constructed graphene ribbon can be constructed inside the dimensions of the Pt(111)

catalyst model as the cokes atoms does not extend the unit cell in a- and b-direction. In the c-

direction, the vacuum layer of 12 Å and an artificial dipole layer suffice to avoid periodic

interactions in this direction.

The carbon atoms of the graphene bind by two on a Pt atom. This leads to Pt-C distances

between 2.15 Å and 2.45 Å. Still, the graphene layer keeps its planar geometry, so three

alternate C-C bonds are observed in an aromatic ring. The distances are 1.47 Å, 1.54 Å and 1.59

Å and are denoted in Angstrom below (see Figure 6-2). An extension of the unit cell in the a-

and b-direction leads to the visualization of a graphene ribbon, as illustrated in Figure 6-2. The

graphene ribbon alters the structure of the two top Pt layers of the unit cell. In the top layer, an

elongation of Pt-Pt bonds is observed of the Pt atoms beneath the graphene ribbon in a-

direction. This elongation leads to a bond length of 3.07 Å, while the bulk Pt-Pt length is 2.82

Å. As consequence, Pt-Pt bonds of the empty Pt atoms are contracted in the same direction from

2.82 to 2.67 Å. In the second Pt layer, the reverse is observed. The Pt atoms in the second layer

beneath the graphene ribbon contract with 0.4 Å, while the bonds of Pt atoms underneath the

empty Pt atoms are elongated with 0.4 Å. The graphene ribbon also affects the unit cell in the

c-direction. The Pt atoms on which the graphene is bonded, are 0.2 Å elevated with respect to

the empty Pt atoms.

Figure 6-2: Adsorption sites on 4×2 Gr/Pt(111) catalyst model: top ●, bridge █ , three-folded hcp ▲ and fcc ▼. Additionally, the three alternate C-C distances in an aromatic ring are shown in Angstrom.

1.47

1.59

1.53

b

a


The 4×2 unit cell is one of the smallest supercells to investigate the effects on cokes on the

catalytic properties. Two rows of non-deactivated Pt atoms remain as empty sites. Again four

adsorption sites are distinguishable conform the nomenclature determined in 4.1.1. The cokes

equally affect the top sites of both rows, as they are equidistant to the coke. However, as it is

expected that nearby coke destabilizes the adsorbates, the bridge and hollow sites located

between these two rows are preferable. Those adsorption sites are the least affected by the

destabilization of the cokes.

Additionally, a 5×2 unit cell is constructed to determine the range and strength of the cokes

influence on propane dehydrogenation. Here, the 4×2 unit cell is extended with one row in the

a-direction and optimized under strict criteria and with vdW-DF as functional. The optimized

unit cell is shown in Figure 6-3. The dimensions of the enlarged unit cell are 14.11×5.64×18.91

Å³. The structure of the optimized graphene ribbon is identical with respect to the 4×2 unit cell

as the same planarity and alternate C-C bond distances are observed.

Figure 6-3: Isometric representation of the 5×2 unit cell of the cokes deactivated Pt(111) catalyst model

Three rows of non-deactivated Pt atoms remain available as adsorption site, see Figure 6-4.

Adsorption occurs on four distinct adsorption sites conform the nomenclature determined in

4.1.1. Of the empty three Pt rows, the middle row top sites are more preferable than the outer

c

b

a


rows. The most preferable bridge and hollow sites are located on both sides of the middle empty

row. Here, the adsorption modes are the least affected by the repulsion of the adsorbed coke.

While the effect of cokes on the adsorbate is less in this extended unit cell, the required

computational cost is larger. Hence, it is opted to focus on a certain reaction path to evaluate

the properties observed on this unit cell.

Figure 6-4: Adsorption sites on 5×2 Gr/Pt(111) catalyst model: top ●, bridge █ , three-folded hcp ▲ and fcc ▼. Additionally, the three alternate C-C distances in an aromatic ring are shown in Angstrom.

The description of the degree of coverage is essential and should be unambiguous. Hence, it is

opted to maintain the definition for degree of coverage, as formulated in 4.1.2 (see equation

(1)). This means that the coverage is determined based on the number of Pt atoms in the top

layer, and not with respect to the free Pt atoms in the top layer. Hence, a coverage of 0.13 ML

is obtained for the 4×2 unit cell and a coverage of 0.10 ML for the 5×2 unit cell. However, it

should be remembered that in fact four of the Pt atoms in the top layer of the unit cell are

unattainable for reaction since they are covered with cokes.

In the following sections, the 4×2 unit cell is always employed, while the last section both unit

cells, 4×2 and 5×2 Gr/Pt(111), are employed to determine the range and strength of cokes

influence.

6.2 Adsorption

The focus in this section lays on the adsorption of the main reactants and products of propane

dehydrogenation. The considered gaseous species are propane, propylene and H2. However, the

b

a

1.48 1.54

1.59


adsorption of propane, propylene and H2 on cokes deactivated Pt catalysts has not been

described before in literature. Instead, the adsorption of the compounds is compared to clean Pt

to assess the influence of cokes. The adsorption energies are calculated conform equation (2)

in 4.2. The optimized geometries are visualized in Appendix E.

6.2.1 Propane

The adsorption of propane plays a major role during propane dehydrogenation. Two modes of

propane adsorption can be observed: physisorption and dissociative chemisorption. The first is

discussed in this section, while the latter is discussed in the Thermodynamics section as it can

be treated as dehydrogenation step (see 6.3). As propane is a saturated compound, it has no

unpaired electron to directly interact with the surface, so solely physisorption occurs.

Physisorbed propane is stabilized by van der Waals interactions and this metastable species acts

as a precursor for further elementary steps in the propane dehydrogenation reaction mechanism.

Physisorbed propane orientates itself above the middle of the island of free Pt atoms. The

bottom C-C bond is parallel to the surface and the perpendicular distance to the Pt surface is

4.2 Å, which is similar to the 4.3 Å distance on Pt(111). Furthermore, the two C-C bonds have

a length of 1.53 Å, identical to the C-C bond length of gaseous propane.

The adsorption energy of physisorbed propane on Gr/Pt(111) is 10 kJ/mol lower than on

Pt(111). Rather the reverse was expected as repulsion by the graphene ribbon would increase

the adsorption energy. However, as the distance between the physisorbed species and the

graphene is more than 4 Å, it possible that van der Waals interactions stabilize the physisorbed

propane, while repulsion occurs closer to the graphene ribbon.

Table 6-1. Comparison between adsorption energies of propane on Pt(111) and Gr/Pt(111) (4×2 unit cell), calculated in this work at a coverage of 0.13 ML with optPBE vdW-DF as functional.

ΔEads (kJ/mol) Pt(111) Gr/Pt(111) Propane physisorption -43 -55

Vibrational analysis of physisorbed propane on Gr/Pt(111) has resulted in one imaginary

frequency. However, this can be attributed to the translation of the metastable physisorbed

species.


6.2.2 Propylene

Propylene has an unsaturated double bond so it can directly form covalent bonds with the Pt

surface during adsorption. Literature makes a distinction between four different adsorption

modes in function of the employed propylene coverage (see also 4.2.2.1). The two most stable

adsorption modes are di-σ propylene and π propylene. Literature indicates that the di-σ

adsorption mode is more stable than the π adsorption mode as di-σ propylene directly bonds

with Pt surface while π propylene is weakly bonded to surface. [6, 7] So solely the di-σ

adsorption mode is discussed out if these two. Physisorbed propylene is also a probable

adsorption mode of propylene. Both the di-σ and physisorbed adsorption modes are investigated

in this work as physisorbed propylene can act as a precursor for the di-σ adsorption mode.

The di-σ adsorption mode of propylene adsorbs on the bridge site, located on one row of the

empty Pt atoms. The methyl moiety orientates over the other empty row of Pt atoms, to reduce

the interaction with the graphene ribbon. Clearly, the elongation of the C=C bond of propylene

is observed as it is stretched from 1.36 Å (C=C bond) gaseous propylene to 1.51 Å, which is

similar to 1.53 Å of the C-C bond of gaseous propane. The physisorption mode of propylene

orientates itself above three empty Pt atoms, parallel to the surface. The shortest distance

between propylene and platinum surface is 4.5 Å. The C=C bond of propylene is not elongated

in this physisorption mode. The optimized geometries on Gr/Pt(111) and Pt(111) are similar as

the C-C and Pt-C length differ less than 5%.

Table 6-2. Comparison between adsorption energies of propylene on Pt(111) and Gr/Pt(111) (4×2 unit cell), calculated in this work at a coverage of 0.13 ML with optPBE vdW-DF as functional.

ΔEads (kJ/mol) Pt(111) Gr/Pt(111) Di-σ adsorption -135 -14

Physisorption -43 -47

While the adsorption energy for di-σ propylene is strongly negative on Pt(111), this is not the

case on Gr/Pt(111). Di-σ propylene is 120 kJ/mol less stable on Gr/Pt(111) due to steric effects

of the inclusion between the graphene ribbons. The physisorption of propylene is slightly more

stabilized on Gr/Pt(111) as the same was observed with physisorbed propane. The nearness of

graphene destabilizes the chemisorption of propylene, while it stabilizes the physisorption of

propylene due to long-range van der Waals interactions.


6.3 Thermodynamics

The same reaction network as in Chapter 4 is employed to describe the propane

dehydrogenation on cokes deactivated Pt(111). However, as the focus lays on the catalyst

activity and selectivity towards propylene, certain intermediates are excluded in the calculations

such as the dissociation products and the cracked C1 and C2 hydrocarbons. The remaining

intermediates that are considered on Gr/Pt(111) are mainly C3Hx (x=5-8) species, ethyl, methyl

and atomic hydrogen. For each component, the most stable geometry on the Gr/Pt(111) catalyst

model is calculated and a frequency analysis is conducted to evaluate its stability. The

considered species are listed in Table 6-3 and a streamlined version of the calculated reaction

network is shown in Figure 6-5. The optimal geometries can be found in Appendix E, together

with a visualized reaction network based on those optimal geometries.

Table 6-3. Adsorbed hydrocarbon species on the Gr/Pt(111) (4×2 unit cell) surface in descending order of molecular weight. * indicates with how many C-Pt bonds the species is adsorbed to the surface.

CH3-CH2-CH3 Propane,phys CH3-CH2-CH2* 1-Propyl

CH3-CH*-CH3 2-Propyl CH3-CH*-CH2

* Propylene

CH3-CH=CH2 Propylene,phys CH3-CH2-CH** 1-Propylidene

CH3-CH**-CH2 2-Propylidene CH3-CH2-C*** 1-Propylidyne

CH3-CH*-CH** 1-Propenyl CH3-C**-CH2* 2-Propenyl

CH3-CH2* Ethyl CH3

* Methyl


Figure 6-5: Reaction network for propane dehydrogenation on Gr/Pt(111) (4×2 unit cell).


(6.3.1), deep dehydrogenation of propylene (6.3.2) and hydrogenolysis of C3 intermediates

(6.3.3). In each section, reaction energies are determined and the stability of each intermediate

is evaluated. The first section can be interpreted to quantify the activity of the catalyst, while

section two mainly quantifies the selectivity towards propylene with respect to deactivation of

the catalyst by cokes precursor formation. Section three determines the selectivity to other side

products, however on Pt (111) it was concluded that, these reactions are kinetically unfavorable,

so solely the dissociation of propane (C-C scission) is selected as model reaction for this section

and will be discussed.


dissociation reactions. The focus will be mainly on the dehydrogenation reactions as the C-H

2-propyl

1-propylidene

CH2=CH-CH3,phys

CH2=CH-CH3(g)

13

_CH-CH-CH3 + H

PtPt2

_

Pt

= CH2-C-CH3 + H

Pt

_

Pt

_

Pt2

=


C-CH2-CH3 + H

Pt3

_

Pt

≡

14 1517 20 21

Methyl and ethyl

4 5

= _CH-CH2-CH3 + H

Pt2 Pt2-propylidene

CH3-C-CH3 + H

Pt2

=

Pt

_

7 8

CH3 + CH2CH3

Pt Pt

_ _CH3-CH-CH3 + H

Pt Pt_ _CH2-CH2-CH3+ H

Pt Pt

_ _

1-propyl

1 2 3

CH3-CH2-CH3,phys


0

Physisorbed propane

CH2-CH-CH3 + H

Pt

__

Pt

_

PtPropylene

11 12

1-propylidyne


Gaseous propylene


bond breaking is the least activated on Pt(111), with respect to the isomerization and

dissociation. For the description of the thermodynamics, the products of dehydrogenation

reactions are optimized in different unit cells. It is assumed that the low hydrogen coverage and

the repulsion between the hydrocarbon intermediates and fragmented hydrogen lead to

diffusion of the hydrogen on the catalyst surface, away from the hydrocarbon intermediates.

The equations proposed in section 4.3 are employed to describe the reaction energies of the

considered reactions.

For the dehydrogenation reactions, the fragmented hydrogen is optimized in a separate unit cell.

It is essential to locate the most stable adsorption site for hydrogen on the cokes deactivated

Pt(111) catalyst. Four adsorption sites for hydrogen sites are studied: on top, bridged, hollow

fcc and hcp. These are compared with their respective adsorption energies on clean Pt(111). A

large upward shift in adsorption energies is observed for all adsorption modes compared to

clean Pt(111), while maintaining the same order of stability (top > bridge > fcc > hcp) (see

Table 4). Hydrogen in the bridge site is second most stable adsorption site, indicating that

hydrogen is stabilized if it is further away from the surface. The graphene ribbon alters the

geometry of on top hydrogen. While hydrogen on a Pt(111) top site has a perpendicular

orientation with respect to the surface, it bends away from the cokes on Gr/Pt(111). The optimal

geometries of hydrogen in fcc and hcp are similar to those of on clean Pt(111). For further

calculations, the adsorption energy of hydrogen on the top site is employed to calculate the

reaction energies of dehydrogenation reactions.

Table 6-4. Comparison of hydrogen adsorption energies on Pt(111) and Gr/Pt(111) (4×2 unit cell) with a coverage of 0.13 ML. This work uses an optPBE vdW-DF functional.

ΔEads (kJ/mol) On top Bridge Fcc Hcp Pt(111) -77 - -71 -59

Gr/Pt(111) -42 -37 -30 -23


The dehydrogenation of propane to propylene consists at least of two elementary steps on the

platinum surface. Prior to those two elementary steps, the physisorption is included in our

reaction network. However, this step is already discussed in previous section (see 6.2.1). The

further reaction of physisorbed propane is the first elementary step, which leads to formation

of 1-propyl or 2-propyl by respectively dehydrogenation of the methyl or methylene group. The


second elementary step is the β-dehydrogenation of the C3H7-species into propylene. However,

the step competes with α-dehydrogenation of 1-propyl and 2-propyl, respectively leading to 1-

propylidene and 2-propylidene, which are undesirable intermediates on the surface. Each step

generates also a detached hydrogen. Its location plays a major role to determine the kinetics as

the activation energy is strongly influenced by the (de-)stabilization due to this hydrogen.

However, for the thermodynamics, it is assumed that hydrogen is highly mobile on the surface.

Table 6-5. Comparison between reaction energies for propane dehydrogenation towards propylene on clean Pt(111) and cokes deactivated Pt(111) (4×2 unit cell). A coverage of 0.13 ML is achieved. This work uses an optPBE vdW-DF functional.

ΔEr (kJ/mol)

Pt(111) Gr/Pt(111) # Surface reaction

0 Propane(g) → Propane,phys Physisorption -43 -55

1 Propane,phys → 1-propyl + H Adsorption -7 67

2 Propane,phys → 2-propyl + H Adsorption -14 120


5 1-propyl → propylene + H Dehydrogenation -22 71

7 2-propyl → propylene + H Dehydrogenation -15 20


13a Propylene→ propylene,phys Desorption 92 -33


In the Gr/Pt(111) catalyst model, both 1-propyl and 2-propyl adsorb on top of one of the four

empty Pt atoms. The optimized geometries of 1-propyl and 2-propyl are similar to that on

Pt(111) as the same bond lengths are observed on both catalyst models. However, it should be

noted that, while the hydrocarbon moiety remains the same, the Pt atom on which adsorption

occurs protrudes above the other Pt atoms of the top layer. The distance between that Pt atom

and the nearby Pt atoms in the underlying layer is 3.07 Å for 1-propyl and even 3.17 Å for 2-

propyl, confirming that the surface layers are further distorted by adsorption of hydrocarbon

species. 1-propyl bends away from the cokes and over the island of the four clean Pt atoms. The

ethyl moiety of 1-propyl is orientated perpendicular with the surface, to reduce the repulsion of

the nearby cokes. The methyl moieties of 2-propyl are orientated above two other empty Pt

atoms, however one methyl group is always near the graphene ribbon, see also Appendix E.


The second elementary step produces propylene, 1-proylidene and 2-propylidene. Propylene

orientates itself on the bridge site, parallel to the graphene ribbon. The bonds in propylene are

identical to those on clean Pt(111), however the two Pt atoms on which propylene is bonded

protrude above the other Pt atoms of the top layer. The methyl moiety of propylene is orientated

away from the graphene ribbon. 1-propylidene adsorbs on the bridge site, parallel with the

graphene ribbon. The ethyl moiety is orientated perpendicular with the surface and above the

remaining empty Pt atoms. 2-propylidene prefers to adsorb on the bridge site between the two

rows of clean Pt atoms. The methyl groups are orientated above the nearby fcc and hcp site.

The Pt atoms where these species are adsorbed on, are protruded above the top layer of the

catalyst.

To obtain useful insight on the stability of the intermediates with respect of each other, an

energy profile is constructed for the considered reactions. The reaction energies of the first

elementary steps are strongly endothermic on Gr/Pt(111) as it is unfavorable to adsorb propane

dissociatively on the surface. 1-propyl is the least unstable with respect to 2-propyl, as the

perpendicular orientation of its ethyl moiety reduces the repulsion by the cokes, while both

methyl groups of 2-propyl are on the same height as the graphene ribbon. Here are the steric

effects with the graphene the largest.

Figure 6-6: Energy profile of adsorbed C3Hx (x= 6–8) species on the Gr/Pt(111) surface (4×2 unit cell). Electronic energies are determined relative to gaseous propane. Dehydrogenated hydrogen is optimized in a separate unit cell.

-70

-50

-30

-10

10

30

50

70

90

110

Rel

ativ

e en

ergy

(kJ/

mol

)

C3H8 (g) C3H8* C3H7

* + H* C3H6* + 2H* C3H6 (g) + 2H*

Propane (g)

Propane, phys

1-propyl

2-propyl1

Propylene,phys

Propylene

2-propylidene

1-propylidene Propylene (g)


Furthermore, the reaction energies for the second elementary steps are also endothermic, so

further dehydrogenation leads to even more unstable species. However, the reaction energies

are the most endothermic for 1-propyl as the dehydrogenated products are less stabilized, while

the reaction energies for 2-propyl are at least one magnitude lower as the reactant is less

stabilized. The most stable C3H6-species is 2-propylidene as it adsorbs in between the rows of

clean Pt atoms. Second most stable species is propylene, which is 2 kJ/mol less stable than 2-

propylidene, while 1-propylidene is 20 kJ/mol less stable than propylene (see Figure 6-5).

Thermodynamically, it is expected that propylene is still the most favorable species on the

surface (if deeply dehydrogenated species are neglected, but these are discussed in the next

section). The difference in relative energy between propylene and 2-propylidene is small, so

both species are expected on the surface. However, propylene can be formed by

dehydrogenation of 1-propyl and 2-propyl, while 2-propylidene solely via 2-propyl. As 1-

propyl is more stable than 2-propyl, the reaction paths via 1-propyl are favored. Additionally,

propylene can desorb to a more stable physisorbed state, inducing extra driving force for this

reaction.

The considered isomerization reactions are excluded in this reaction network as it was shown

in 4.4.1, that they generally have a large reaction barrier on Pt(111). This makes those reactions

improbable to occur.

It is important to conclude that steric effects between the cokes and the adsorbates lead to

unstable species on the surface. Thermodynamically speaking less or even no reaction will

occur on this model of cokes deactivated Pt(111) surface compared to the clean Pt(111) surface.


The considered deeply dehydrogenated species on Gr/Pt(111) are 1-propenyl, 2-propenyl and

1-propylidyne. 1-propylidyne is formed via α-dehydrogenation of 1-propylidene and adsorbs in

a hollow fcc site. Steric interactions cause the 1-propylidyne to bend slightly away from the

graphene ribbon so that its ethyl moiety is orientated over the neighboring hcp site. 1-propenyl

is formed either via β-dehydrogenation of 1-propylidene or dehydrogenation of the outer

bonded C atom of propylene. 2-propenyl is formed either via β-dehydrogenation of 2-

propylidene or dehydrogenation of the middle C atom of propylene. These species orientate

themselves on a bridge and a top site, respectively with the double bonded C atom and single

bonded C atom.


Table 6-6. Comparison between reaction energies for deep dehydrogenation on clean Pt(111) and cokes deactivated Pt(111) (4×2 unit cell). A coverage of 0.13 ML is achieved on these catalyst models. This work uses an optPBE functional.

ΔEr (kJ/mol)

# Surface reaction Pt(111) Gr/Pt(111)



17 1-propylidene → 1-propylidyne + H Dehydrogenation -67 -74


21 2-propylidene → 2-propenyl + H Dehydrogenation -24 5

Based on the reaction energies (see Table 6-6) and energy profile (see Figure 6-7), 1-

propylidyne is the most thermodynamically favored species on Gr/Pt(111). The same is

observed on Pt(111). Furthermore, 1-propenyl is less stable than 2-propenyl, which is also

observed on Pt(111). The reaction energies are similar on clean Pt(111) and Gr/Pt(111). This

indicates that no additional repulsion occurs from the graphene ribbon in the reactions towards

C3H5-species.

Figure 6-7: Energy profile of adsorbed C3Hx (x= 5–8) species on the Gr/Pt(111) surface (4×2 unit cell). Electronic energies are determined relative to gaseous propane. Dehydrogenated hydrogen is optimized in a separate unit cell.

-70

-50

-30

-10

10

30

50

70

90

110

Rel

ativ

e en

ergy

(kJ/

mol

)

C3H8 (g) C3H8* C3H7

* + H* C3H6* + 2H* C3H5

* + 3H*

Propane (g)

Propane, phys

1-propyl

2-propyl

Propylene,phys

Propylene

2-propylidene

1-propylidene

1-propylidyne

2-propenyl

1-propenyl


To quantify the selectivity towards propylene with respect to deactivation by deep

dehydrogenation reactions, the same reaction paths (desorption and dehydrogenation paths) as

on Pt(111) are considered. The propylene desorption pathway consists of 1-propyl

dehydrogenation to propylene and propylene desorption via physisorbed propylene, while the

dehydrogenation pathway consists of consecutive dehydrogenations of 1-propyl to 1-

propylidyne via 1-propylidene. Based on thermodynamical aspects and stability of the

intermediates, it is shown that propylene desorption is less endothermic near the graphene

ribbon, while the stabilization of 1-propylidyne remains constant with respect to clean Pt(111).

On the Gr/Pt(111) surface, propylene desorption will occur easier and side reactions to 1-

propylidyne will occur less, thereby decreasing the deactivation of the catalyst.

6.3.3 Hydrogenolysis of C3-intermediates

Cracking of C3-species leads to C1 and C2 species on the surface. They either hydrogenate and

desorb from the surface as alkanes such as methane and ethane or dehydrogenate further,

leading to coke precursors and adsorbed carbon. On Pt(111), it was shown that cracking is

solely favorable for deeply dehydrogenated species, as higher reaction barriers are obtained for

C-C cleavage than C-H bond breaking. On Gr/Pt(111), the dissociation of propane is primarily

evaluated to determine if this and other cracking reactions are probable to occur near the

graphene ribbon. The optimized geometry of the dissociated ethyl and methyl shows clear the

effect of repulsion by the graphene ribbon, see Appendix E. Both the methyl and ethyl adsorb

on top of a Pt atom. The Pt-C bond of ethyl is elongated with respect to Pt(111); also, the Pt

atom on which the ethyl is adsorbed is protruded, as the distance between the two top layers is

3.2 Å while the bulk distance 2.82 Å. This distortion is due the repulsion of the two hydrocarbon

species and the graphene ribbon. It is expected that the steric effects result in a strong

endothermic reaction energy for the dissociation reaction. The resulted reaction energy is 265

kJ/mol. This is strongly endothermic, indicating that the reaction is highly improbable to occur

on the Gr/Pt(111) surface. Hence, the electronic reaction energy for this reaction is not

calculated, as it has to be higher than 265 kJ/mol.


6.4 Kinetics

The determination of the kinetic descriptors is based on the transition states, in this case the

electronic activation energies. The electronic energy of the optimized transition states with an

imaginary frequency along the reaction coordinate is used to determine the electronic activation

energy of a reaction, a measure for the reaction barrier. The electronic activation energy Δ‡E is

determined with respect to the electronic energy of the reactant, conform Chapter 4.

The description of the kinetics is divided into two sections: propane dehydrogenation to

propylene (6.4.1) and deep dehydrogenation of propylene (6.4.2). The kinetics of the

dissociation reactions are not considered as these products are thermodynamically unfavorable.

The first section quantifies the activity of the cokes deactivated Pt(111), while the second

describes the selectivity towards propylene with respect to deactivation reactions that lead to

cokes precursor formation.

As the transition state geometries are calculated with the NEB/dimer method, an initial and final

state has to be provided. In contrast to the assumption of low hydrogen coverage for the

adsorbate optimization, the final state of dehydrogenation reactions has to be optimized with

the hydrogen nearby the intermediate in the same unit cell. For all optimized transition state

geometries, one imaginary frequency corresponding to the reaction coordinate is calculated.


Electronic activation energies of all considered reactions in the thermodynamics section are

determined. It is assumed that the adsorption and desorption of the gaseous C3-species (propane

and propylene) are not activated. However, the reaction barrier for desorption is considered to

be equal to the desorption energy. On Gr/Pt(111), the reaction energy for desorption of

propylene is negative, so it is assumed that the reaction barrier is equal to zero. The other

elementary steps are indeed activated and the reaction barriers of these reactions are much

higher with respect to the same reactions on Pt (111). The steric effects of the graphene ribbon

destabilize the transition states in a similar way as the intermediates, leading to higher electronic

activation energies.


Table 6-7. Comparison between electronic activation energies for propane dehydrogenation towards propylene on Pt(111) and Gr/Pt(111) (4×2 unit cell). A coverage of 0.13 ML is achieved. This work uses an optPBE vdW-DF functional.

Δ‡E (kJ/mol)

Pt(111) Gr/Pt(111) # Surface reaction










The repulsion between the graphene ribbon and the intermediate prevents chemisorption on the

surface. Additionally, all reactions on the surface are so highly activated that they probably do

not occur. In Figure 6-8, the reaction towards propylene via 1-propyl is highlighted, as this is

least activated pathway on Gr/Pt(111). However, with respect to the same path on Pt(111), all

intermediates and transition states are more unstable due to repulsive interaction with the

graphene ribbon, except for physisorbed propane. The transition state of the first elementary

step is 81 kJ/mol more unstable on Gr/Pt(111) than on Pt(111), for the second the difference is

even 133 kJ/mol. On both the clean as the cokes deactivated Pt catalyst model, the first

elementary step has the highest electronic activation energy. The relative energy of propylene

desorption is 34 kJ/mol lower on Gr/Pt(111) than on Pt(111). This discrepancy is equal to the

difference in adsorption energies of two atomic hydrogen on the respective models. Cokes

deactivation leads not only to blockage of adsorption sites, but also lowers the activity of the Pt

catalyst. However, it should be noted that the observed magnitude of graphene ribbon repulsion

is due to the selection of the 4×2 unit cell. If this unit cell is periodically extended, the adsorbed

intermediates are surrounded on two sides by the graphene ribbons.


Figure 6-8: Comparison of reaction path towards propylene via 1-propyl on clean Pt(111) (black) and Gr/Pt(111) (blue) (4×2 unit cell). The energy is calculated relative to gaseous propane.


The kinetics of the deep dehydrogenation reactions are necessary to describe the selectivity

towards gaseous propylene with competes with formation of coke precursors via deep

dehydrogenation. Thermodynamically, 1-propylidyne is the most stable species on the surface

on the cokes deactivated Pt catalyst, so this species is selected a model compound for the coke

precursors. To define the selectivity descriptor between propylene desorption and deep

dehydrogenation towards 1-propylidyne, their respective reaction paths are compared, see

Figure 6-8. For the deep dehydrogenation reaction path, the transition state of dehydrogenation

of 1-propylidene to 1-propylidyne is calculated. The electronic activation energy of this reaction

is 18 kJ/mol, which is 27 kJ/mol lower than on Pt(111).

Both reaction paths start at 1-propyl, which can either dehydrogenate to propylene (desorption

reaction path) or 1-propylidene (coke precursor reaction path). However, for the first step, the

reaction barrier towards propylene is lower than for 1-propylidene and subsequently the latter

leads to a more unstable intermediate. The second step is the competition of 1-propylidene

dehydrogenation and propylene desorption via a physisorbed species. As last step in the

-75

-25

25

75

125

175

Rel

ativ

een

ergy

(kJ/

mol

)

Propane (g)

Propane, phys

1-propyl + H*

Propylene,phys + 2H

Propylene + 2H*

TS1

TS5

Reactant Product




formation 1-propylidyne has the lowest reaction barrier, this step is kinetically favored.

However, it should be noted that solely enthalpic factors are included in the energy profile.

Figure 6-9: Relative energy profile for propylene desorption (···) and dehydrogenation ( ̶ ̶ ) reaction paths as descriptor for the propylene selectivity on Gr/Pt(111) (4×2 unit cell). The energy is calculated relative to gaseous propane.

Referring to the Gibbs free energy diagram in 4.4.2, the inclusion of entropic contributions,

which are neglected in the relative energy profile, results into that gaseous propylene (+ H2

(g)) has a lower Gibbs free energy than 1-propylidyne (+ 3/2 H2 (g)) at 900K. As gaseous

species are independent of the catalyst model, the Gibbs free energy remains constant.

Furthermore, it is assumed that the entropic contributions of adsorbates are of the same

magnitude on different catalyst models. However, it should be noted that this assumption is less

applicable because the intermediates are weaker bonded to the catalyst surface. This leads to

different vibration modes and thus entropic contributions. So the relative stability of gaseous

propylene with respect to 1-propylidyne is mostly determined by the electronic energy

difference. On clean Pt(111), the energetic difference between gaseous propylene and 1-

propylidene, (accounting for the conservations laws, by adding extra gaseous hydrogen), is 132

kJ/mol. However, a general upward shift in electronic energies for adsorbate species is observed

on Gr/Pt(111). Hence, this leads to a smaller energetic difference on Gr/Pt(111) of 47 kJ/mol.

-75

-25

25

75

125

175

Rel

ativ

e en

ergy

(kJ/

mol

)

Propane (g)

Propane, phys

1-propyl +H*


Propylene + 2H*

Propylene (g) + 2H*

TS1

TS5

1-propylidene+ 2H*

TS17

TS4

1-propylidyne + 3H*

Reactant Product




The decrease points out that the formation of 1-propylidyne is less favored on Gr/Pt(111) with

respect to Pt(111).

6.5 The range and strength of cokes influence on propane dehydrogenation

In the previous sections, a 4×2 unit cell was used to describe the cokes deactivated Pt catalyst.

It can be concluded that the inclusion between graphene ribbons induces destabilization of the

adsorbates. The steric effects are so great that catalytic activity near these ribbons is small and

dehydrogenation is improbable to occur. However, due to the weakening of the adsorbate bonds

with the surface, the selectivity towards propylene is improved as the physisorbed species are

more stabilized than the chemisorbed species. It is expected that adsorbates less enclosed by

the graphene ribbons are more stable and the catalytic activity is retained. Hence, the 5×2 unit

cell is constructed (see 6.1) to determine the range and strength of the graphene ribbon on the

adsorbate stability and catalytic activity. Thermodynamically, the intermediates are expected to

form further away from the graphene ribbon. The reaction path for propane dehydrogenation

towards propylene via 2-propyl is considered to evaluate this effect. This reaction path is

compared with the 4×2 cokes deactivated Pt(111) unit cell and the clean Pt(111) 4×2 unit cell.

The optimized geometries can be found in Appendix E, together with a graphical representation

of the considered reaction path based on those geometries. Further, this section is divided into

two parts: thermodynamics and kinetics.

6.5.1 Thermodynamics

For the considered dehydrogenation reactions, the most stable adsorption site of hydrogen is

determined in a separate unit cell, as it is assumed that the hydrogen coverage is low. The most

stable adsorption site of hydrogen is on top on the 5×2 Gr/Pt(111) unit cell. The energy for

dissociative hydrogen adsorption on the 5×2 unit cell is -57 kJ/mol. As expected, the value is

situated between the ΔEads (H2) of -76 kJ/mol on clean Pt(111) and of -42 kJ/mol on cokes

deactivated Pt(111) (4×2 unit cell).

The optimized geometry of all intermediates can be found in Appendix E. Under strict

convergence criteria, the physisorbed propane and propylene have a residual RMS forces of

respectively 0.017 and 0.021 eV/Å, which is acceptable for these metastable species.


Furthermore, vibrational analyses are conducted to evaluate their stability. No imaginary

frequencies are found, except for physisorbed propane and propylene. These species have

respectively three and four imaginary frequencies, which can be assigned to translation and

external rotation above the surface.

Table 6-8. Comparison between reaction energies for propane dehydrogenation towards propylene via 2-propyl on clean Pt(111) and on two catalyst models for cokes deactivated Pt(111). This work uses an optPBE vdW-DF functional.

ΔEr (kJ/mol) 0.13 ML 0.1 ML Surface reaction Pt(111)

Gr/Pt(111)

(4×2) Gr/Pt(111)

(5×2) Propane(g) → Propane,phys Physisorption -43 -55 -49

Propane,phys → 2-propyl + H Adsorption -14 120 22

2-propyl → propylene + H Dehydrogenation -15 20 9

Propylene→ propylene,phys Desorption 92 -33 60

Propylene,phys → Propylene (g) Desorption 43 47 40

The calculated reaction energies on the larger unit cell show that the steric effects caused by

the graphene inclusion of the intermediates are reduced because the intermediates can adsorb

further from the graphene ribbon. It is logical that the largest decrease in reaction energy (-98

kJ/mol) is observed for the dehydrogenation of propane to form 2-propyl, as during this step

steric effects are too large when enclosed by cokes. The reverse is observed for the desorption

step (adsorbed propylene to physisorbed propylene), as now the chemisorbed species are

favorable with respect to the physisorbed.


Figure 6-10: Relative energy profile for propane dehydrogenation towards propylene via 2-propyl on clean Pt(111) (black), 4×2 Gr/ Pt(111) (blue) and 5×2 Gr/Pt(111) (purple). Energies are determined relative to gaseous propane. Dehydrogenated hydrogen is optimized in a separate unit cell.

However, the graphene still induces repulsive interaction with the adsorbates, as can be seen in

the endothermic nature of the reaction energies. In the constructed energy profile, relative

energies of intermediates on the 5×2 Gr/Pt(111) model situate themselves between

intermediates of the two other models. The relative energies of the clean Pt(111) intermediates

are the most stabilized, while those of the 4×2 Gr/Pt(111) model are strongly endothermic with

respect to gaseous propane. It can be concluded that the 4×2 Gr/Pt(111) catalyst model was too

small to accurately describe the effect of cokes on the reaction mechanism since this catalyst

would not be active anymore.

6.5.2 Kinetics

The determination of the electronic activation energy is identical as in 6.4. It is assumed that

adsorption and desorption are not activated. In the case of desorption, the electronic activation

energy is equal to the desorption energy. For the considered activated reactions, it is observed

that on the larger unit cell of Gr/Pt(111), the reaction barriers are lower with respect to the

smaller unit cell, see Table 9 and Figure 6-10. The decrease of steric effects helps to stabilize

the transition states in a similar way as the intermediates.

-75

-50

-25

0

25

50

75

100

Rel

ativ

e en

ergy

(kJ/

mol

)


* + 2H* C3H6 (g) + 2H*

Propane (g)

Propane, phys 2-propyl

Propylene,phys

Propylene

Propylene (g)

C3H6, phys + 2H*


Table 6-9. Comparison between electronic activation energies for propane dehydrogenation towards propylene via 2-propyl on clean Pt(111) and on two catalyst models for cokes deactivated Pt(111). This work uses an optPBE vdW-DF functional.

Δ‡E (kJ/mol) 0.13 ML 0.1 ML Surface reaction Pt(111)

Gr/Pt(111)

(4×2) Gr/Pt(111)

(5×2) Propane(g) → Propane,phys Physisorption 0 0 0

Propane,phys → 2-propyl + H Adsorption 68 193 108†

2-propyl → propylene + H Dehydrogenation 84 143 115

Propylene→ propylene,phys Desorption 92 0 60

Propylene,phys →Propylene (g) Desorption 43 47 40 † The frequency analysis of the transition state for reaction 2 reported two imaginary frequencies.

The relative energy profile (Figure 6-11) shows that the activity is retained on 5×2 Gr/Pt(111)

in comparison with 4×2 Gr/Pt(111), both the energies of the intermediates and transition states

are shifted downwards. Kinetically, it is expected that the reaction rates are the highest on clean

Pt(111) and the proximity of graphene will lower the reaction rate until all reaction are

improbable to occur. In the final steps, all adsorbate are desorbs and the total reaction energy is

identical for all catalyst models, as electronic energies of the gaseous species do not depend on

the catalyst model.

Figure 6-11: Comparison of reaction path towards propylene via 2-propyl on clean Pt(111) (black), 4×2 Gr/Pt(111) (blue) and 5×2 Gr/Pt(111) (purple)

-75

-25

25

75

125

175

225

Rel

ativ

e en

ergy

(kJ/

mol

)

ProductReactant

Propane (g)


Propylene,phys + 2H

Propylene + 2H*

TS2

TS7




6.6 Conclusions

In this chapter, a model is constructed to quantify the effect of cokes deactivated Pt on the

activity towards propylene and deactivation of the catalyst during propane dehydrogenation.

Characterization of coke deposits on pure Pt catalyst have led to the selection of graphene as

model compound to represent the cokes on the Pt surface. Initially, a 4×2 unit cell is employed

for the calculations on Gr/Pt(111), directly extended from the Pt(111) catalyst model of Chapter

4. However, the catalytic activity on this catalyst model is poor as steric effects of the graphene

ribbon strongly destabilize the adsorbates. It is postulated that dehydrogenation reactions are

improbable to occur on this catalyst surface. Aside the poor activity, an improved selectivity

towards propylene with respect to deactivation reactions is observed. Thermodynamically, 1-

propylidyne is favored on the catalyst surface and an exothermic reaction energy is observed

for its formation, but all preceding intermediates are more unstable than on clean Pt(111). This

leads to a selectivity descriptor that favors propylene desorption with respect to the same

descriptor on Pt(111). So a distinct trade-off is made between activity and propylene selectivity.

It is clear that the steric effects in the 4×2 unit cell are too large, so a 5×2 unit cell is constructed

on which intermediates can adsorb further away from the graphene ribbon. This model restores

most of its activity compared with the smaller unit cell. The strength of the steric interactions

is reduced as they have a small range, however repulsion by the graphene ribbon still affects

the intermediates, but on a smaller scale. For the considered reaction path, the catalytic activity

is lower than on the clean Pt(111) catalyst model.

6.7 References


2. Preto, I., Graphene formation and oxidation on Pt/Mg(Ga)AlOx dehydrogenation catalysts, in Department of Chemical Engineering and Technical Chemistry. 2015, University of Ghent: Ghent.

3. Larsson, M., et al., The effect of reaction conditions and time on stream on the coke formed during propane dehydrogenation. Journal of Catalysis, 1996. 164(1): p. 44-53.







Chapter 7 Propane dehydrogenation kinetics on Pt(211) catalyst model The solid state Pt catalyst consists of several surface planes such as the (111) and (110) planes

of which the Pt(111) is the most abundant. However, the transition of one plane to another is

coupled to the formation of undercoordinated edges. Furthermore, irregularities on the surface

leads to undercoordinated sites such as steps and kinks. Literature has shown that those

undercoordinated sites have a high catalytic activity and are preferable to the terrace Pt(111)

sites at the initial stage of propane dehydrogenation. Apart from high activity, a low

propylene selectivity and large coke accumulation on these sites are observed at this stage,

indicating that coke deactivates these sites rather fast. [1-3]

It is opted to construct the catalyst model based on the undercoordinated step site. Pt(211) is

the smallest fcc surface that consists of step sites. Furthermore, Yang et al. conducted a DFT

study on the effect of step sites and also employed a Pt(211) unit cell. However, a coverage of

0.25 ML is employed, which is higher than the propylene monolayer coverage. Furthermore,

Yang et al. use the PBE functional that does not account for van der Waals interactions. To

obtain qualitative insight, the reaction path via 2-propyl to form propylene on Pt(211) is

evaluated based on its comparison with Yang et al. [2]

7.1 Catalyst model

The Pt(211) catalyst model is represented by a 3×2 unit cell, see Figure 7-1. The a-direction

consists of two Pt atoms and the b-direction of three atoms. In the latter direction, the step is


located. In the c-direction, four layers of Pt atoms are used. The upper two layers can relax as

they represent the surface atoms, while the bottom two layers are kept fixed as they represent

the bulk structure of the model. In this direction, a vacuum layer of 12 Å and an artificial

dipole layer are constructed to avoid periodic interactions. A Monkhorst-Pack grid of 5×5×1

is utilized for the Brillouin zone integration. The optimization is completed under strict

convergence criteria with optPBE vdW-DF as functional. The unit cell dimensions, including

vacuum layer, are 5.64×6.91×17.46 Å³

Figure 7-1: Isometric representation of the 4×2 unit cell of the Pt(211) catalyst model


Four types of adsorption sites conform the nomenclature of Chapter 4 are available in this

model. While most of adsorption sites are identical with respect of those in Chapter 4, the

focus will be on those nearby the step. The considered adsorption sites are illustrated in

Figure 7-2. Hence, to avoid confusion with the Pt(111) adsorption sites, the suffix “near-

edge” is added during comparison with the Pt(111) model.

Step

b

c

a


Figure 7-2: Top view of Pt(211) unit cell. Adsorption sites near the edge on the Pt(211) catalyst model are shown: top ●, bridge █ , three-folded hcp ▲ and fcc ▼.


The degree of coverage as defined in 4.1.2, is used in this chapter as well. Therefore, for one

adsorbate per six surface atoms, the coverage equals 1/6 ML (0.17 ML). This is lower than the

monolayer coverage of propylene (0.20 ML). [4]

7.2 Adsorption

The main compounds of propane dehydrogenation that occur in the gasphase are propane,

propylene and H2. Cracking reactions can lead to eventually to formation of smaller gaseous

hydrocarbons such as methane, ethane and ethylene. However, their adsorption will not be

considered on Pt(211). The optimized geometries can be found in Appendix F.

7.2.1 Propane

Propane is a saturated C3 hydrocarbon that weakly adsorbs on the Pt surface. During

physisorption, no binding is formed with the Pt surface and propane is solely stabilized by van

Step

Step

b

a


der Waals interactions. Wang et al. studied the molecular propane adsorption on a Pt(655)

surface. This surface is formed by separating Pt(111) terraces by the Pt(100) surface of single

atom height and it can be considered as a step surface. The favorable adsorption zone for

propane was near the step edge on the Pt surface. [5]

The initial position of physisorbed propane on Pt(211) is located in the favorable adsorption

zone proposed by Wang et al. The resulted geometry is located above the step, parallel with

surface. The distance between the Pt and closest C atom of propane is 3.3 Å, which is similar

to the distance calculated on Pt(111). The C-C bonds of physisorbed propane are identical to

C-C bonds of gaseous propane, so no strong interaction with the Pt surface is observed.

Additionally, frequency analyses have shown only one imaginary frequency. This imaginary

frequency corresponds to external rotation of the physisorbed propane species.

Table 7-1. Comparison of physisorption energies of propane on Pt(111) and Pt(211) surface. Yang et al. used a PBE functional, while in this an optPBE vdW-DF functional is employed.

Pt(111) Pt(211)

ΔEads (kJ/mol) Yang et al. (0.11 ML) [6]

This work (0.13 ML)


This work

(0.17 ML) Propane physisorption -6 -43 -4 -53

The large discrepancy for adsorption energies of propane between the DFT studies can be

explained due the difference of the utilized DFT functionals. In this work, van der Waals

interaction are included, which lead to (over-)stabilization of the physisorbed species. In the

work of Yang et al., no distinct effect of the step sites on the propane physisorption can be

observed as energetic difference between Pt(111) and Pt(211) lays within the error margin of

the employed PBE functional. However, in this work, an additional stabilization is found

nearby Pt steps of ~10 kJ/mol.

7.2.2 Propylene

As propylene has an unsaturated bond, it can directly decompose and adsorb on the Pt surface.

Zaera et al. propose four different adsorption modes in function of the employed propylene

coverage of which is the most important, followed by π propylene. Literature indicates that

the di-σ adsorption mode is more stable than the π adsorption mode. Conform 4.2.2.1; it is

opted to solely study di-σ adsorption mode of propylene. [7, 8]

On the step sites, di-σ propylene prefers to adsorb on the bridge site near the step edge, as

indicated in Figure 7-2. This is the same site as on the flat Pt(111) surface. The C=C bond of


adsorbed propylene is weakened as its bond length is 1.49 Å, which is more similar to the C-

C bond length of gaseous propane (1.54 Å) than gaseous propylene (C=C length of 1.36 Å).

Furthermore, the same geometry is observed as on the flat Pt(111) surface, with the methyl

moiety perpendicular to the surface. Aside from the di-σ adsorption mode of propylene, a

physisorption mode is investigated. In this adsorption mode, covalent bonds are formed with

the surface and physisorbed propylene retains its gasphase geometry. The most stable

adsorption site is over the edge of the step site. However, this mode did not converge under

strict criteria as an RMS force of 0.027 eV/Å remains. Still, this is adequate for a metastable

adsorption mode. Additionally, three imaginary frequencies are observed and can be assigned

to modes of translation and external rotation. The physisorbed propylene is slightly tilted with

respect to the surface, to stabilize the methyl moiety of propylene. The C=C is not stretched

and bond length of 1.34 is similar to that of gaseous propylene (1.36 Å).

Table 7-2. Comparison of adsorption energies of di-σ propylene on Pt(111) and Pt(211) surface. Yang et al. used a PBE functional, while in this a vdW-DF functional is employed.

Pt(111) Pt(211)

ΔEads (kJ/mol) Yang et al. (0.11 ML) [6]

This work (0.13 ML)


This work

(0.17 ML)

Propylene adsorption -90 -135 -138 -158

In both studies, the adsorption energies of Pt(211) are much lower than on Pt(111), indicating

that di-σ propylene prefers to adsorb on undercoordinated step sites. This additional

adsorption strength on Pt(211) will make desorption more difficult on these sites as for

desorption the reaction barrier will be more endothermic. Yang et al. did not report the

adsorption energy of physisorbed propylene. However, this work indicates that this adsorption

mode is more stabilized on the step sites as an adsorption energy of -105 kJ/mol is observed,

with respect to -43 kJ/mol on the Pt(111) terrace surface.

7.3 Thermodynamics

In this section, the reaction path towards propylene via 2-propyl is investigated as part of the

reaction network proposed in Chapter 4 to obtain qualitative insight on the effects of step sites

on the stability of the adsorbed intermediates. The following intermediates are considered for

this reaction path: physisorbed propane, 2-propyl, di-σ propylene and physisorbed propylene.

For each component, the most stable geometry on the Pt(211) catalyst model is calculated and


a frequency analysis is conducted to evaluate its stability. A condensed version of this

reaction network is shown in Figure 7-3. The optimal geometries can be found in Appendix F,

together with a visualized reaction path based on those optimal geometries. It should be noted

that this network path is a part of the proposed reaction network proposed in Chapter 4 as the

same numbering for the reactions is used.

Figure 7-3: Reaction path towards propylene via 2-propyl on Pt(211).

This reaction path was selected as it is important to quantify the catalytic activity towards

propylene. The considered reaction path consists solely of adsorption/desorption (reaction 0

and 13) and dehydrogenation steps (reaction 2 and 7). Complementary to previous chapters,

the hydrogen formed during dehydrogenation reaction diffuses away from the hydrocarbon

adsorbate. This is implemented in our calculations by optimizing the adsorbate and hydrogen

in separate unit cells. To determine the reaction energies, equation (3), (4) and (5) of section

4.3 are used. It is essential to locate the most stable adsorption site for atomic hydrogen.

Hydrogen prefers to adsorb either in a hollow fcc site or on top. However, previous pure Pt

catalyst model (see Chapter 4 and 6) have shown that the top site is the most stable adsorption

for hydrogen with the selected optPBE vdW-DF functional. So hydrogen was optimized on

various top positions: on the flat surface and near the edge. The energetic difference is 4

kJ/mol in favor for the hydrogen adsorbed nearby the edge. This most stable geometry of

adsorbed hydrogen will be used in reaction energy calculations of the dehydrogenation

reactions. With respect to Pt(111), atomic hydrogen adsorbs dissociatively 8 kJ/mol stronger

on the Pt(211) surface.

2-propyl

CH2=CH-CH3,phys

CH2=CH-CH3(g)

7CH3-CH-CH3 + H

Pt Pt

_ _2CH3-CH2-CH3,phys


0

Physisorbed propane

CH2-CH-CH3 + H

Pt

__

Pt

_

PtPropylene


Gaseous propylene

13



In the considered reaction path, two elementary steps are considered: dehydrogenation of the

methylene group of propane (reaction 2) and β-dehydrogenation of 2-propyl towards

propylene. Prior to these reaction steps, the physisorption of propane is considered. However,

this reaction is discussed in the adsorption section. Physisorbed propane is a metastable

species and acts as a precursor to C3H7-species.

In the first elementary step, 2-propyl adsorbs on top of a Pt atom, conform the adsorption on

Pt(111), but now near the edge. However, the most stable geometries differ between this work

and Yang et al., especially the orientation of the outer two methyl groups. In the optimized

geometry of Yang et al, the two methyl groups remain parallel with the step and neither of

them hangs over the step more than the other. In this work, various geometries, obtained from

rotating the propyl group around the Pt-C bond, are proposed as initial guess and the minimal

electronic energy is found where one methyl group is orientated over the edge and the other

above the terrace. For the second elementary step, propylene adsorbs on the bridge site, near

the edge. However, the orientation of methyl moiety of propylene differs between the DFT

studies. In this work, the methyl moiety is perpendicular to the surface, while Yang et al.

reported a geometry with the methyl group parallel to the surface orientated over the edge.

The different optimized geometry can be explained based on the long-range stabilization of

van der Waals interaction by the selected functional in this work.

The reported reaction energies by Yang et al. are considerably lower than in this work, except

for the physisorption of propane as the employed PBE functional cannot describe this state of

propane very well. Furthermore, reaction energies of the dehydrogenation step are 35-40

kJ/mol lower than in this work, indicating that either the used PBE functional or the higher

coverage strongly stabilize the intermediates. The general trend is that higher coverage

destabilizes the intermediates on the surface, so most probably the use of a different

functional is the reason for the discrepancy.


Table 7-3. Comparison between reaction energies for the propane dehydrogenation reaction path via 2-propyl on Pt(111) and Pt(211). Yang et al. used a PBE functional, while in this an optPBE vdW-DF functional is employed.

ΔEr (kJ/mol) Pt(111) Pt(211)

Reaction # Yang et al. (0.11 ML) [6]

This work (0.13 ML)


This work

(0.17 ML)

0 -2 -43 -4 -53

2 -6 -14 -51 -22

7 -23 -15 -59 -24

13a 90 92 138 53

13b - 43 - 105

The step sites are more active, leading to more stable intermediates and lower reaction

energies in both DFT studies. However, no general trend is observed in the literature. No

distinct energetic difference is found for the physisorption of propane, while the reaction

energies of the first and second elementary step are considerably lower, respectively 45

kJ/mol and 36 kJ/mol, see Figure 7-4. In this work, a decrease of ~9 kJ/mol is observed for

the reaction energies with respect to Pt(111) as can be seen in Figure 7-4.

Thermodynamically, the adsorption sites near the step are preferred with respect to the

Pt(111) adsorption sites.

Figure 7-4: Energy profile of adsorbed C3Hx (x= 3–6) species on the (black) Pt(111) and (red) Pt(211) surface (respectively coverage of 0.13 ML and 0.17 ML). Energies are determined relative to gaseous propane, while the detached hydrogen is optimized in separate unit cell for dehydrogenation reactions.

-120

-100

-80

-60

-40

-20

0

20

40

60

80

Rel

ativ

e en

ergy

(kJ/

mol

)


* + 2H* C3H6,phys + 2H*

Propane (g)

Propane, phys2-propyl

Propylene,phys

Propylene

Propylene (g)

C3H6 (g)+ 2H*


The reaction pathway towards propylene via 2-propyl can be treated as a catalytic cycle and,

if correct, the total reaction energy should correspond to the gasphase reaction of propane to

propylene and H2. If the reaction energies of the considered elementary steps are summated, a

total reaction energy of 138 kJ/mol is found, which corresponds with the gas phase reaction

energy.

7.4 Kinetics

To determine the reaction barriers of the proposed reactions, the corresponding transition

states of these reactions have to be found. Hence, by combining the electronic energy of the

transition state and the reactant, the electronic activation energy Δ‡E is calculated conform the

equation in Chapter 4. This energy is an estimation of the reaction barrier of that reaction.

In the considered reaction pathways, it is assumed that the physisorption of propane is not

activated and hence, has a reaction barrier equal to zero. The two considered elementary steps

are activated, but the transition states on the steps are more stable and lower barriers are found

with respect to Pt(111) for both DFT studies, see Table 4. It is concluded that propane

dehydrogenation on step sites is more active. The optimized geometries are similar to those of

Yang et al., especially the bond length between the detached hydrogen and carbon atoms are

identical. These geometries can be found in Appendix F.

Table 7-4. Comparison of the electronic activation energies on Pt(111) and Pt(211). Yang et al. used a PBE functional, while in this an optPBE vdW-DF functional is employed.

Δ‡E (kJ/mol) Pt(111) Pt(211)

Reaction # Yang et al. (0.11 ML) [6]

This work (0.13 ML)


This work

(0.17 ML)

0 0 0 0 0

2 68 68 27 49†

7 61 84 32 22

13a 90 92 138 53

13b - 43 - 105 † This transition state has an additional imaginary frequency; however, it can be assigned to lattice vibration.

However, some discrepancy is observed between the energies of the DFT studies. The

electronic energies of Yang et al. are similar for the two elementary steps on Pt(211), while in

this work, the first elementary step is higher activated (49 kJ/mol) and the second is lower

activated (22 kJ/mol), indicating that the activation of propane is the rate determining step.


Figure 7-5: Relatieve energy profile for propylene reaction path via 2-propyl on (black) Pt(111) and (red) Pt(211). (respectively coverage of 0.13 ML and 0.17 ML).The energy is calculated relative to gaseous propane while dehydrogenated hydrogen is optimized in a separate unit cell.

Based on the relative energy profile (see Figure 7-5), it is observed that adsorption sites near

the step of Pt(111) have a higher activity as lower reaction barriers are observed for the

propane dehydrogenations reactions towards propylene on Pt(211). However, the reaction

barrier for propylene desorption from adsorbed species to gasphase is 158 kJ/mol on Pt(211),

which 25 kJ/mol more than on Pt(111), indicating that propylene desorption is less likely on

Pt(211) than Pt(111).

It is also expected that the selectivity towards propylene is lower on the Pt(211) surface than

on the Pt(111) because further dehydrogenation of the adsorbed propylene can be favored.

Yang et al. reports results concerning deeply dehydrogenated species, confirming the low

selectivity towards propylene on step sites, as deep dehydrogenation steps are both

thermodynamically and kinetically favored until formation of propyne. For these types of

species, cracking reactions have low reaction barriers, leading to the formation of cokes. This

confirms the results of Peng et al. that undercoordinated sites such as steps are the initiating

site for coke formation on Pt catalysts [2, 3]

Furthermore, Yang et al. observed the hydrogenolysis of propane, which is identified as an

important step towards gaseous side product as its dissociation forms ethyl and methyl on the

-100

-50

0

50

100

150

Rel

ativ

e en

ergy

(kJ/

mol

)

Reactant Product

Propane (g)

Propane, phys

2-propyl + H*


Propylene +2H*

Propylene (g) + 2H*

TS2


TS7


surface. The reported reaction barriers are 235 kJ/mol on Pt(111) and 157 kJ/mol on Pt(211).

These values are both higher than the respective reaction barriers for dehydrogenation.

However, relative speaking, the reaction barrier for propane hydrogenolysis decrease more

than the barriers for propane dehydrogenation, indicating that steps sites are more prone to

propane dehydrogenation, and subsequently the formation of gaseous side products.

Preliminary calculations on the propane hydrogenolysis on Pt(211) have shown that the

transition state does not convergence under strict criteria.

It should be noted that apart of the effect of the steps, the coverage is also different for both

catalyst models. At lower coverages and under the saturation coverage, it is expected that

intermediates are weaker adsorbed on the Pt surface. [9] However, this effect is not observed,

as the coverage effect is subjected to the effect of the step site.

7.5 Conclusions

To obtain qualitative insight in the effect of steps on the Pt surface on propane

dehydrogenation characteristics, the reaction path towards propylene via 2-propyl is

investigated and compared with literature. As catalyst model for Pt(211), a 3×2 unit cell is

employed to obtain a coverage of 0.17 ML. The hydrocarbon intermediates are optimized in

adsorption sites located near the step.

The resulted thermodynamics and kinetics on Pt(211) are shown as a reaction profile relative

to gaseous propane and are compared with the reaction profile on Pt(111) in Figure 7-5. Both

thermodynamically and kinetically, the steps are preferred as adsorption sites due their higher

reactivity. However, it should be noted that by strengthening the propylene bonds with the

surface, a higher reaction barrier is observed for propylene desorption, indicating that

propylene is less likely to desorb on Pt(211) than on Pt(111).

Further, Yang et al. proposed that steps sites are more sensitive for deep dehydrogenation and

hydrogenolysis. Deep dehydrogenation steps are both thermodynamically and kinetically

favored until formation of propyne. Consecutively, these species have low barriers for

hydrogenolysis, leading to the formation of cokes.


7.6 References





5. Wang, J.-C., Effects of surface step on molecular propane adsorption. Surface Science, 2003. 540(2–3): p. 326-336.





Chapter 8: Conclusions and prospects 171

Chapter 8 Conclusions and prospects Recently, the catalytic dehydrogenation of light paraffins (C2-C3) has emerged as promising

technology for on purpose production of light olefins, independent of the olefin production via

steam cracking. The focus of this thesis is on the catalytic dehydrogenation of propane towards

propylene.

Ab initio calculations have been employed to obtain fundamental insight in propane

dehydrogenation kinetics. In the defined computational framework, a non-local optPBE vdW-

DF functional is used to describe the geometries of the intermediates and transition states in the

proposed reaction network on the considered catalyst models.

Platinum metal catalysts are an adequate choice as catalyst for light alkane dehydrogenation as

they exhibit a high activity for the dehydrogenation steps. However, those catalysts lack a high

selectivity towards olefins, as many side reactions occur on the surface. Eventually, these side

reactions initiate coke formation, leading to unwanted deactivation of the catalyst. [1] To

determine the propane dehydrogenation characteristics on pure Pt catalysts, a catalyst model of

the most abundant Pt(111) phase is constructed. The DFT calculations are conducted on a 4×2

Pt(111) unit cell, corresponding to a molecular coverage of 0.13 ML. The proposed reaction

network is divided in three main sections: propane dehydrogenation to propylene, deep

dehydrogenation of C3H6-species and hydrogenolysis of C3 intermediates. This first section is

employed to quantify the catalyst activity of pure platinum as reaction barriers and energies of

the elementary dehydrogenation steps towards propylene are important. In the second section,

the deep dehydrogenation towards coke precursors is investigated. The selectivity towards these

species is an indication for the catalyst deactivation. In the third section, the hydrogenolysis

reactions are taken into account. These reactions form C1 and C2 adsorbates on the surface.

172 Chapter 8: Conclusions and prospects

These species eventually lead to gaseous side products such as methane, ethane and ethylene

and affect the selectivity towards propylene.

Both the thermodynamics and kinetics of propane dehydrogenation are evaluated for the

considered reactions. The propane dehydrogenation reaction towards propylene can occur via

1-propyl and via 2-propyl but, the reaction path via 2-propyl is preferred as it has a lower

reaction barrier with respect to 1-propyl. In contrast to literature [2], adsorbed 2-propyl is more

stable than 1-propyl in this work. This catalyst model serves as reference for more advanced

models. However, as the Gibbs free energies at 873 K of these species and the reaction barriers

of the elementary step are similar, it is expected that both reaction paths contribute to the

formation of propylene. Both 1-propyl and 2-propyl preferentially dehydrogenate further to

propylene, as it is the thermodynamically most stable C3H6-species. The resulting electronic

activation energy is similar for both reactions.

Deep dehydrogenation leads to the thermodynamically most stable species on the surface, i.e.

1-propylidyne on the Pt(111) surface. Multiple reaction paths towards 1-propylidyne are

possible if isomerization reactions are included. However, it is shown that the isomerization

reactions are highly activated and unlikely to occur. Furthermore, the α-dehydrogenation of 1-

propylidene towards 1-propylidyne is kinetically favored. The most probable path is via

propane →1-propyl →1-propylidene →1-propylidyne. However, this is solely based on the

electronic energy (enthalpic contributions). The entropic contributions are included from

harmonic oscillator calculations and the Gibbs free energy of the intermediates and transition

states is determined. The entropic contributions are greater for gaseous species than for

adsorbates, so based on the Gibbs free diagram at 900 K, gaseous propylene is 11 kJ/mol more

stable than 1-propylidyne.

Hydrogenolysis reactions compete with dehydrogenation reactions and C-C bond breaking

leads to the formation of less stable products with respect to the dehydrogenated species, except

for the cracking of 2-propylidene into ethylidyne and methyl. However, C-C cleavage of 2-

propylidene and other hydrogenolysis reactions are averagely 95 kJ/mol more activated than

the competing dehydrogenation reactions. Therefore, it can be concluded that these reactions

are less favored for the considered species on the Pt(111) surface.

To validate the calculated thermodynamics and kinetics, a microkinetic simulation of the

propane dehydrogenation on Pt(111) is conducted, initially at 873 K and total reactant pressure

of 0.4 bar (C3H8/H2: 3/1). Three time regimes are investigated: short (0.1 s), intermediate (10

s) and long (109 s). The highest TOF for propylene and hydrogen gas (1191 s-1) is observed at

short simulated time. Initially, the surface is for 80% covered with 1-propylidyne, consistent


with the ab initio calculations. At intermediate simulated time (10 s), the coverage of 1-

propylidyne starts to decline and ethylidyne (CH3C≡Pt) and methylidyne (HC≡Pt) are formed.

This indicates that 1-propylidyne is kinetically preferred as adsorbate (fast coke precursors),

while the other species are thermodynamically favored (slow coke precursors). At a long

simulated time (109 s), no catalyst activity for propane dehydrogenation is reported and the

surface is completely covered with slow coke precursors. The intermediate simulated time is

further employed as it still shows high catalytic activity, while 1-propylidyne is converted

substantially to more stable coke precursors. While varying the simulation temperature, a

maximum TOF for propylene dehydrogenation is observed at 948 K. However, at this

temperature, formation of gaseous side products, especially methane and ethylene is also

increased, lowering the selectivity towards gaseous propylene. Increasing the pressure of the

products, i.e. propylene, methane, ethylene and ethane, does not have a significant influence on

the catalytic activity and selectivity.

The recent experimental studies on Ga-promoted Pt catalysts during light alkane

dehydrogenation have already shown an increase in catalytic activity and selectivity towards

propylene with respect to an unmodified Pt catalyst on the same support [3]. Saerens [4] already

identified the most likely PtxGay alloy at the catalytic surface by comparing frequencies of CO

adsorption experiments with the results of theoretical CO vibrational frequencies on Pt-Ga

catalyst models with various Pt/Ga ratios. The most likely candidate among the different studied

catalyst models is the Pt3Ga bulk alloy, the studied model catalyst with the lowest Ga content.

To assess the role of Ga on propane dehydrogenation kinetics on Pt-Ga catalysts, a 4×2

Pt3Ga(111) unit cell is constructed. The same molecular coverage is achieved as on Pt(111).

Propane dehydrogenation towards propylene consists of two elementary steps and occurs via

two reaction pathways (via 1-propyl and via 2-propyl). The reaction barriers of the two

elementary steps towards propylene define the catalytic activity. On Pt3Ga(111), both 1-propyl

and 2-propyl are more stabilized than on Pt(111). In contrast to Pt(111), is 2-propyl

thermodynamically more stable than 1-propyl and hence the reaction pathway via 2-propyl is

selected to describe the catalytic activity on Pt3Ga(111). Furthermore, adsorbed propylene

bonds stronger on the Pt3Ga(111) surface and the calculated reaction barriers are lower than on

Pt(111), while the propylene desorption barrier remains constant. It can be concluded that the

catalytic activity is potentially larger on Pt3Ga(111) than on Pt(111).

In addition, the deep dehydrogenation of C3H6-species is investigated on Pt3Ga(111). The

formation of these species are an indication for the deactivation of the catalyst. Similar to


Pt(111), 1-propylidyne is thermodynamically the most favored species on the Pt3Ga(111)

surface. To quantify the selectivity to the coke precursors, the reaction energies and barrier for

the reaction path via 1-propyl and 1-propylidene towards 1-propylidyne are determined. This

reaction path has higher reaction barriers on Pt3Ga(111) than on Pt(111) and consequently 1-

propylidyne is not as kinetically favored as on Pt(111). Furthermore, the electronic energy

difference between propylene and 1-propylidyne on Pt3Ga(111) is much smaller than on

Pt(111). As it can be assumed that the entropic contributions for adsorbed species such as 1-

propylidyne are similar on Pt3Ga(111) as on Pt(111), the relative Gibbs free energy of 1-

propylidyne on Pt3Ga(111) with respect to Pt(111) is determined by the difference in electronic

energy on Pt(111) and Pt3Ga(111). Based on this reasoning, it is expected that 1-propylidyne is

less likely to be formed on Pt3Ga(111) than on Pt(111), both from a thermodynamic and a

kinetic point of view. This indicates that fewer coke precursors will be formed and that Ga

reduces the formation of cokes on the surface, which is confirmed based on the reduced carbon

formation on Pt-Ga/Mg(Ga)(Al)Ox.

The hydrogenolysis of propane is studied on Pt3Ga(111) as this reaction forms C1 and C2

adsorbates, which are critical for the formation of gaseous side products such as ethane,

ethylene and methane. However, this reaction is highly activated (191 kJ/mol), 4 kJ/mol more

than on Pt(111) while the reaction barriers for the competing dehydrogenation reactions are

lower on Pt3Ga(111). Hence, the formation of gaseous side products via this way is unlikely.

This result is supported by the experimental results that show that the propylene selectivity is

above 98% on Pt-Ga/Mg(Ga)(Al)Ox for the total time on stream. [3]

The coke formation on the surface of the Pt catalysts leads inherently to the deactivation of the

catalyst surface. Furthermore, the effect of cokes deactivated Pt on nearby Pt atoms is

investigated in terms of catalytic activity and further deactivation of the catalyst during propane

dehydrogenation. [5] Graphene is selected as representation of the coke on the cokes

deactivated Pt catalyst based on TEM observations. [6] The catalyst model Gr/Pt(111) is

primarily constructed in a 4×2 unit cell, in which half of the supercell is covered with a

continuous 1-D graphene ribbon. The catalytic activity on this catalyst model is poor as steric

effects of the graphene ribbon strongly destabilize the adsorbates. It is postulated that

dehydrogenation reactions are improbable to occur on this catalyst surface based on the resulted

high reaction barriers and energies. In contrast the poor activity, an improved selectivity

towards propylene with respect to deactivation reactions is observed. Thermodynamically, 1-

propylidyne is favored on the catalyst surface and an exothermic reaction energy is observed

for its formation, but all preceding intermediates are more unstable than on clean Pt(111). This


leads to a selectivity descriptor that favors propylene desorption with respect to the same

descriptor on Pt(111). So a distinct trade-off is made between activity and propylene selectivity.

It is clear that the steric effects in the 4×2 unit cell are too large due to the graphene ribbon

inclusion, so a 5×2 unit cell is constructed on which intermediates can adsorb further away from

the graphene ribbon. This model restores most of its activity compared with the smaller unit

cell. The strength of the steric interactions is reduced as they have a small range, however

repulsion by the graphene ribbon still affects the intermediates, but on a smaller scale. For the

considered reaction path, the catalytic activity remains lower than on the clean Pt(111) catalyst

model.

Literature reports that the origin of gaseous side products and initial coke formation can be

assigned to the reactivity of undercoordinated sites such as steps and edges of the Pt catalyst.

[6] To obtain qualitative insight on the effects of undercoordinated sites on propane

dehydrogenation characteristics, a Pt(211) catalyst model is constructed as a representation of

an undercoordinated step on the Pt surface. For this model, a 3×2 unit cell is employed to obtain

a coverage of 0.17 ML, which is higher than the reference coverage of 0.13 ML. Furthermore,

the hydrocarbon intermediates are optimized in adsorption sites located near the step and the

reaction path towards propylene via 2-propyl is investigated.

Both thermodynamically and kinetically, the steps are preferred as adsorption sites due to their

higher reactivity than the terrace plane (Pt(111)). However, it should be noted that by

strengthening the propylene bonds with the surface, propylene adsorbs stronger, indicating that

propylene is less likely to desorb on Pt(211) than on Pt(111). The higher coverage of this model

(0.17 ML) compared to the previously considered catalyst models (0.13 ML) induces a minor

destabilization of the surface species mutually. However, as the reverse energetic effect

(stabilization) is reported, it is concluded that coverage affects the adsorbates weaker than the

step sites. Furthermore, Yang et al. proposed that steps sites are more sensitive for deep

dehydrogenation and hydrogenolysis reactions. [7] Deep dehydrogenation steps considered by

Yang et al., are both thermodynamically and kinetically favored until formation of propyne.

Consecutively, these species have low barriers for hydrogenolysis, leading to C1 and C2-species

on the catalyst surface.

Summarizing, Ga alloying of Pt catalysts show superior catalytic properties than alloying with

Sn. With respect to the reference Pt(111), the Pt3Ga catalyst model shows an increased catalytic

activity for propane dehydrogenation. Furthermore, it is shown that the formation of coke

precursors are reduced, indicating a higher deactivation resistance. The hydrogenolysis of


propane is improbable and higher selectivity towards to propylene are obtained on this Pt-Ga

catalyst. On the cokes deactivated Pt catalyst, the propane dehydrogenation does not occur near

the graphene ribbon due steric effects with the adsorbates. Beyond the short-ranged steric

effects, the catalytic activity is retained, however it is lower compared to clean Pt(111) due to

the repulsion by the graphene ribbon. Undercoordinated sites on the Pt catalysts are

thermodynamically and kinetically preferred as sites for propane dehydrogenation to propylene

with respect to the terrace Pt surface. In addition, deep dehydrogenation and hydrolysis

reactions are also less activated, indicating that these sites initiate unwanted side reactions.

Future work

For the continuation of this work, the following suggestions are made. Apart from the Pt(111)

reference catalyst model, the main focus of this work was on the quantification of the catalytic

activity and the catalyst deactivation by the formation of deeply dehydrogenated coke

precursors. However, the proposed reaction network on the more advanced catalyst models

should include the hydrogenolysis reactions and consecutive formation of gaseous side

products. The description of these products are essential to correlate the selectivity towards

propylene with experimental data. Obtaining a better understanding of the Pt-Ga particle by

investigating the Ga distribution over the catalyst particle and the interaction with the calcined

hydrotalcite support can improve the catalyst model used in the ab initio calculations.

Furthermore, the condense reaction network of the advanced catalyst models, such as the

Pt3Ga(111) catalyst model, should be extended conform the mechanism on Pt(111) and the

resulted data should be implemented in the microkinetic model for propane dehydrogenation.

Next, the proposed reaction network on the Pt(111) catalyst model can be refined by including

additional deeply dehydrogenated species and the interconversion of C1- and C2-species.

References

1. Vora, B.V., Development of Dehydrogenation Catalysts and Processes. Topics in Catalysis, 2012. 55(19-20): p. 1297-1308.








Appendix A

178

Appendix A Examples of INCAR files

A.1 INCAR file for strict geometry optimization

Electronic minimization PREC = NORMAL GGA = OR #For optPBE vdw-DF functional. Change to “PE” for PBE functional LUSE_VDW = .TRUE. #Only necessary in case of vdw-DF functional AGGAC = 0.0000 #Only necessary in case of vdw-DF functional VOSKOWN = 1 LREAL = auto #Use FALSE for small systems (bulk unit cell) ALGO = FAST ENCUT = 400 #Plane wave cutoff in eV EDIFF = 1e-8 #Strict electronic energy convergence criterion ISYM = 0 ISPIN = 1 NELM = 400 LWAVE = .FALSE. LCHARG = .FALSE. LVTOT = .FALSE. Ionic relaxation IDIPOL = 3 LDIPOL = .TRUE. ISIF = 0 ISTART = 0 ; ICHARG=2 LORBIT = 11 EDIFFG = -0.015 #Strict ionic convergence criterion NSW = 200 IBRION = 1 #1=Quasi Newton RMM-DIIS, 2=Conjugate Gradient POTIM = 0.25 DOS related values ISMEAR = 1

179 Appendix A

SIGMA = 0.2 Parallelization NSIM = 4 NPAR = 1 LPLANE = .TRUE. LSCALU = .FALSE.

A.2 INCAR files for transition state optimization

A.2.1 NEB optimization

Electronic minimization PREC = NORMAL GGA = OR #NEB calculations are only conducted with the optPBE vdw-DF functional. LUSE_VDW = .TRUE. AGGAC = 0.0000 VOSKOWN = 1 LREAL = auto ALGO = FAST ENCUT = 400 EDIFF = 1e-4 #Loose electronic energy convergence criterion ISYM = 0 ISPIN = 2 NELM = 400 LWAVE = .FALSE. LCHARG = .FALSE. LVTOT = .FALSE. Ionic relaxation IDIPOL = 3 LDIPOL = .TRUE. ISIF = 0 ISTART = 0 ; ICHARG=2 LORBIT = 11 EDIFFG = -0.20 #Loose ionic convergence criterion NSW = 200 IBRION = 1 #1=Quasi Newton POTIM = 0.30 DOS related values ISMEAR = 1 SIGMA = 0.2 Parallelization NSIM = 2 NPAR = 2 LPLANE = .TRUE. LSCALU = .FALSE. TRANSITION STATE IMAGES = 8 #Number of images between initial and final state ICHAIN = 0

Appendix A

180

SPRING = -5 SPRING2 = -5 ISPRING = 1 SPOWER = 1 EFIRST = -219.09992 #Energy of initial state ELAST = -218.94877 #Energy of final state LCLIMB = .FALSE. # If TRUE, the climbing NEB method is used LTANGENT = .TRUE.

A.2.2 Dimer optimization

Electronic minimization PREC = NORMAL GGA = OR #Dimer calculations are only conducted with the optPBE vdw-DF functional. LUSE_VDW = .TRUE. VOSKOWN = 1 AGGAC = 0.0000 LREAL = auto ALGO = FAST ENCUT = 400 EDIFF = 1e-8 #Strict electronic energy convergence criterion ISYM = 0 ISPIN = 1 NELM = 400 LWAVE = .FALSE. LCHARG = .FALSE. LVTOT = .FALSE. Ionic relaxation IDIPOL = 3 LDIPOL = .TRUE. ISIF = 0 ISTART = 0 ; ICHARG=2 LORBIT = 11 EDIFFG = -0.015 #Strict ionic energy convergence criterion NSW = 550 IBRION = 3 POTIM = 0 DOS related values ISMEAR = 1 SIGMA = 0.2 Parallelization NSIM = 4 NPAR = 2 LPLANE = .TRUE. LSCALU = .FALSE. TRANSITION STATE IOPT = 2 #Type of optimizer employed ICHAIN = 2 #Enables the dimer method

181 Appendix A

DdR = 5E-3 DRotMax = 2 DFNMin = 0.01 DFNMax = 1.0

A.3 INCAR file for frequency calculations

Electronic minimization PREC = NORMAL GGA = OR #Frequency calculations are conducted with the optPBE vdw-DF functional. AGGAC = 0.0000 VOSKOWN = 1 LREAL = auto IALGO = 48 ENCUT = 400 EDIFF = 1e-8 ISYM = 0 ISPIN = 1 NELM = 400 LWAVE = .FALSE. LCHARG = .FALSE. Ionic relaxation IDIPOL = 3 LDIPOL = .TRUE. ISIF = 0 ISTART = 0 ; ICHARG=2 LORBIT = 11 NSW = 400 IBRION = 5 #5=Vibration POTIM = 0.015 #Step size for the numerical Hessian matrix calculation DOS related values ISMEAR = 1 SIGMA = 0.2 Parallelization NSIM = 4 NPAR = 2 LPLANE = .TRUE. LSCALU = .FALSE.

Appendix B 182

Appendix B Optimized geometries on the Pt(111) catalyst model In this appendix, first the proposed reaction models are illustrated with the optimized

geometries of the intermediates. Next, the optimized geometries are enlarged and given in the

following order: gasphase, hydrogen, propane and propylene dehydrogenation, dissociation

products, C1 and C2 hydrocarbons and transition states. The following color code is employed

for the elements: carbon (gray), hydrogen (white) and platinum (blue). As platinum atoms are

rather large, solely the top layer of the catalyst model is shown. Additionally, the radius of

platinum is reduced from 1.40 Å to 1.00 Å. For the optimized geometries, interatomic lengths

are given between specific atoms. For intermediates, the lengths of the C-C bonds, one C-H

bonds and Pt-C bonds (and eventual Pt-H bond) are determined. For transition states, the bond

length between atoms that participate in the reaction are reported. All distances are determined

in Å.

183 Appendix B

B.1 Reaction network

Figure B-1: Reaction network of propane dehydrogenation on Pt(111) catalyst model. In the dehydrogenation reactions, hydrogen is formed. However, the formed hydrogen is not shown in this figure.

Appendix B 184

Figure B-2: Reaction network for formation of gaseous side products

B.2 Optimized geometries

B.2.1 Gasphase species

Figure B-3: Gaseous propane

Figure B-4: Gaseous propylene

Gaseous methanePhysisorbed methaneMethyl

A1 A2

Ethyl

Physisorbed ethane

Physisorbed ethylene

Gaseous ethane

Gaseous ethylene

B1

B2

C1C3C2

Di-sigma ethylene

+ H-Pt

+ H-Pt

- H-Pt

185 Appendix B

Figure B-5: Gaseous ethane

Figure B-6: Gaseous ethylene

Figure B-7: Gaseous methane

Figure B-8: Gaseous H2

Appendix B 186

B.2.2 Hydrogen

Figure B-9: Hydrogen on top site

Figure B-10: Hydrogen in fcc site

Figure B-11: Hydrogen in hcp site

187 Appendix B

B.2.3 (Deeply) dehydrogenated intermediates

Figure B-12: Physisorbed propane

Figure B-13: On top 1-propyl

Figure B-14: On top 2-propyl

Figure B-15: Bridged 1-propylidene

Figure B-16: Bridged 2-propylidene

Figure B-17: Di-sigma propylene

Appendix B 188

Figure B-18: Physisorbed propylene

Figure B-19: 1-propylidyne in fcc site

Figure B-20: 1-propenyl on top and in bridge site

Figure B-21: 2-propenyl on top and in bridge site

189 Appendix B

B.2.4 Dissociation products

Figure B-22: On top methyl and ethyl

Figure B-23: Bridged methylene and on top ethyl

Figure B-24: On top methyl and bridged ethylidene

Figure B-25: Bridged ethylidene and bridged methylene

Figure B-26: On top ethyl and methylidene in fcc site

Figure B-27: On top methyl and ethylidyne in fcc site

Appendix B 190

B.2.5 C1 and C2 hydrocarbon intermediates

Figure B-28: Physisorbed methane

Figure B-29: On top methyl

Figure B-30: Physisorbed ethane

Figure B-31: On top ethyl

Figure B-32: Physisorbed ethylene

Figure B-33: Di-sigma ethylene

191 Appendix B

B.2.6 Transition states

Figure B-34: TS1 Propane – 1-propyl

Figure B-35: TS2 Propane – 2-propyl

Figure B-36: TS3 Propane dissociation

Figure B-37: TS4 1-propyl – 1-propylidene

Figure B-38: TS5 1-propyl – propylene

Figure B-39: TS6 1-propyl dissociation

Appendix B 192

Figure B-40: TS7 2-propyl – propylene

Figure B-41: TS8 2-propyl – 2-propylidene

Figure B-42: TS9 2-propyl dissociation

Figure B-43: TS10 1-propyl – 2-propyl

Figure B-44: TS11 1-propylidene – propylene

Figure B-45: TS12 2-propylidene – propylene

193 Appendix B

Figure B-46: TS14 Propylene – 1-propenyl

Figure B-47: TS15 Propylene – 2-propenyl

Figure B-48: TS16 Propylene dissociation

Figure B-49: TS17 1-propylidene – 1-propylidyne

Figure B-50: TS18 1-propylidene dissociation

Figure B-51: TS19 2-propylidene dissociation

Appendix B 194

Figure B-52: TS20 1-propylidene – 1-propenyl

Figure B-53: TS21 2-propylidene – 2-propenyl

Figure B-54: TSA1 Methane – methyl

Figure B-55: TSB1 Ethane – ethyl

Figure B-56: TSC1 Ethyl – ethylene

Appendix C 195

Appendix C Thermodynamic and kinetic parameters of the elementary steps for the Pt(111) catalyst model In this appendix, the thermodynamic and kinetic parameters of the reaction network on Pt(111)

will be provided. The reaction network is the same as illustrated in Appendix B. All kinetic and

thermodynamic parameters have been determined at 900 K. All adsorbates were considered to

be immobile on the surface (including hydrogen) and only for the gaseous species free rotation

and translation have been taken into account. Spurious imaginary frequencies have been dealt

with in an adequate manner, as discussed in Chapter 4.

C.1 Thermodynamic parameters

For the thermodynamic parameters, only the reaction enthalpy, entropy, Gibbs free energy and

thermodynamic equilibrium coefficient are shown (see Table C-1).

196 Appendix C

Table C-1. Thermodynamic parameters for the elementary propane dehydrogenation reaction steps on Pt(111). All values are determined at 900 K.

∆rH ∆rS ∆rG K kJ/mol J/molK kJ/mol Total reaction 129.0 208.6 -58.7 2.56E+03 Propane physisorption -56.1 -184.8 110.3 3.98E-07 Reaction 1 8.0 55.7 -42.1 2.78E+02 Reaction 2 0.4 45.7 -40.8 2.32E+02 Reaction 3 -1.3 14.8 -14.6 7.00E+00 Reaction 4 -1.1 -11.9 9.6 2.77E-01 Reaction 5 -30.1 -28.9 -4.1 1.72E+00 Reaction 6 10.7 -36.7 43.8 2.87E-03 Reaction 7 -22.5 -18.9 -5.4 2.06E+00 Reaction 8 7.2 -0.8 7.9 3.47E-01 Reaction 9 10.2 -9.7 19.0 7.88E-02 Reaction 10 -7.7 -10.0 1.3 8.36E-01 Reaction 11 -29.0 -17.0 -13.7 6.23E+00 Reaction 12 -29.6 -18.1 -13.3 5.93E+00 Reaction 14 11.2 9.0 3.2 6.54E-01 Reaction 15 -1.5 13.5 -13.6 6.18E+00 Reaction 16 52.7 13.4 40.7 4.36E-03 Reaction 17 -72.7 5.0 -77.2 3.03E+04 Reaction 18 -31.3 -22.1 -11.4 4.56E+00 Reaction 19 -70.0 -5.4 -65.1 6.01E+03 Reaction 20 -17.7 -8.0 -10.5 4.07E+00 Reaction 21 -31.1 -4.6 -26.9 3.66E+01 Reaction A1 -4.6 4.2 -8.4 3.09E+00 Reaction B1 5.9 -9.9 14.8 1.38E-01 Reaction C1 -13.8 15.0 -27.2 3.81E+01 Propylene adsorption -137.2 -253.1 90.6 5.53E-06 Ethane physisorption -42.2 -166.9 108.0 5.38E-07 Ethylene adsorption -119.5 -159.3 23.9 4.11E-02 Methane physisorption -83.7 -141.4 43.5 2.97E-03

C.2 Kinetic parameters

For the kinetic parameters, only the forward reaction coefficient, pre-exponential factor and

activation energy are shown (see Table C-2). For adsorption, molecular flux expressions have

been used (see Chapter 4). Desorption and reverse reaction parameters have been determined

using thermodynamic consistency (see Chapter 4).

Appendix C 197

Table C-2. Kinetic parameters for the elementary propane dehydrogenation reaction steps on Pt(111). All values are determined at 900 K and units are given for first order reactions.

kforward

s-1 Aforward

s-1 Ea,forward

kJ/mol Reaction 1 1.42E+11 1.22E+16 85.0 Reaction 2 1.34E+11 1.89E+15 71.5 Reaction 3 5.07E+04 8.08E+15 193.0 Reaction 4 1.66E+08 1.13E+12 66.0 Reaction 5 1.04E+09 4.28E+12 62.3 Reaction 6 1.00E+01 1.90E+09 142.6 Reaction 7 1.07E+09 8.00E+12 66.8 Reaction 8 8.06E+07 2.05E+12 75.9 Reaction 9 3.29E+02 3.49E+12 172.7 Reaction 10 6.09E+02 3.55E+13 185.5 Reaction 11 1.62E+04 2.69E+11 124.4 Reaction 12 1.00E+01 1.90E+09 142.6 Reaction 14 3.09E+08 5.44E+12 73.2 Reaction 15 9.80E+08 5.38E+12 64.4 Reaction 16 4.26E+01 1.81E+13 200.4 Reaction 17 3.68E+08 2.69E+09 14.9 Reaction 18 4.32E+04 3.53E+10 101.9 Reaction 19 1.61E+05 5.19E+12 129.4 Reaction 20 8.80E+08 1.43E+12 55.3 Reaction 21 1.97E+09 1.76E+12 50.8 Reaction A1 1.54E+08 3.18E+12 74.4 Reaction B1 1.01E+06 7.24E+09 66.4 Reaction C1 2.46E+08 6.76E+12 76.5

Appendix D 198

Appendix D Optimized geometries on the Pt3Ga(111) catalyst model In this appendix, first the proposed reaction models are illustrated with the optimized


following order: gasphase, hydrogen, (deeply) dehydrogenated species, dissociation products

and transition states. The following color code is employed for the elements: carbon (gray),

hydrogen (white), platinum (blue) and gallium (green). As platinum and gallium atoms are

rather large, solely the top layer of the catalyst model is shown. Additionally, the radius of the

surface atoms is reduced to 1.00 Å. For the optimized geometries, interatomic lengths are given

between specific atoms. For intermediates, the lengths of the C-C bonds, one C-H bonds and

Pt-C bonds (and eventual Ga-X bonds) are determined. For transition states, the bond length

between atoms that participate in the reaction are reported. All distances are determined in Å.

199 Appendix D

D.1 Reaction network

Figure D-1: Reaction network on the Pt3Ga(111) catalyst model. In the dehydrogenation reactions, hydrogen is formed. However, the formed hydrogen is not shown in this figure.

2-propyl

1-propylidene

10

13


14 1517 20 21

Methyl and ethyl

4 5

2-propylidene

7 8

1-propyl

1 2 3

Gaseous propane0

Physisorbed propane

Propylene

11

1-propylidyne Physisorbed propylene

Gaseous propylene

12

Appendix D 200

D.2 Optimized geometries

D.2.1 Gasphase species

The optimized geometries of gaseous species can be found in B.2.1 as they are identical.

D.2.2 Hydrogen

Figure D-2: Hydrogen adsorbed on Pt3-fcc-Pt2Ga site

Figure D-3: Hydrogen adsorbed on Pt3-hcp-Ga site

Hydrogen will preferentially adsorb in hcp or fcc sites on Pt3Ga. Other sites, such as bridge and

top, are less favored and therefore not shown.

D.2.3 (Deeply) dehydrogenated intermediates

Figure D-4: Physisorbed propane

Figure D-5: On top 1-propyl

201 Appendix D

Figure D-6: On top 2-propyl

Figure D-7: 1-propylidene on Pt2-bridge-Pt2Ga site

Figure D-8: 2-propylidene on Pt2-bridge-Pt2Ga site

Figure D-9: Di-sigma propylene on Pt2-bridge-Pt2Ga site

Figure D-10: Physisorbed propylene

Figure D-11: 1-propylidyne in Pt3- hcp site

Appendix D 202

Figure D-12: 1-propenyl in Pt3- hcp site

Figure D-13: 2-propenyl in Pt3- hcp site

D.2.4 Dissociation products

Figure D-14: On top methyl and ethyl

D.2.5 Transition states

Figure D-15: TS1 Propane – 1-propyl

Figure D-16: TS2 Propane – 2-propyl

203 Appendix D

Figure D-17: TS3 Dissociation propane

Figure D-18: TS4 1-propyl – 1-propylidene

Figure D-19: TS7 2-propyl – propylene

Figure D-20: TS 17 1-propylidene – 1-propylidyne

Appendix E 204

Appendix E Optimized geometries on the graphene ribbon covered Pt(111) catalyst model In this appendix, first the proposed reaction models are illustrated with the optimized


following order: gasphase, hydrogen, (deeply) dehydrogenated species, dissociation products

and transition states. First, all geometries on the 4×2 unit cell are reported, followed by those

on the 5×2 unit cell. The following color code is employed for the elements: carbon (gray),

hydrogen (white) and platinum (blue). As platinum atoms are rather large, solely the top layer

of the catalyst model is shown. Additionally, the radius of platinum is reduced from 1.40 Å to

1.00 Å. For the optimized geometries, interatomic lengths are given between specific atoms.

For intermediates, the lengths of the C-C bonds, one C-H bond and Pt-C bonds (and eventual

Pt-H bond) are determined. For transition states, the bond length between atoms that participate

in the reaction are reported. All distances are determined in Å. In some of the shown figures,

isolated hydrogen atoms are found. This is due to the periodic nature of the unit cell and the

fact that the graphene ribbon is quite large in the unit cell: the hydrogens that terminate the

ribbon are found in the opposite position of the unit cell due to periodicity reasons.

205 Appendix E

E.1 Reaction network

Figure E-1: Reaction network on the cokes deactivated Pt(111) catalyst model (4×2 unit cell). In the dehydrogenation reactions, hydrogen is formed. However, the formed hydrogen is not shown in this figure.

2-propyl

1-propylidene

10

13


14 1517 20 21

Methyl and ethyl

4 5

2-propylidene

7 8

1-propyl

1 2 3

Gaseous propane0

Physisorbed propane

Propylene

11

1-propylidyne Physisorbed propylene

Gaseous propylene

12

Appendix E 206

Figure E-2: Reaction path via 2-propyl towards propylene on 5×2 Gr/Pt(111) catalyst model. In the dehydrogenation reactions, hydrogen is formed. However, the formed hydrogen is not shown in this figure.

E.2 Optimized geometries in the 4×2 unit cell

E.2.1 Gasphase species


2-propyl

72

Gaseous propane

0

Physisorbed propane Propylene


Gaseous propylene

13

207 Appendix E

E.2.2 Hydrogen

Figure E-3: Hydrogen on top site

Figure E-4: Hydrogen in fcc site

Figure E-5: Hydrogen on bridge site

E.2.3 (Deeply) dehydrogenated intermediates

Figure E-6: Physisorbed propane

Figure E-7: On top 1-propyl

Appendix E 208


Figure E-9: Bridged 1-propylidene

Figure E-10: Bridged 2-propylidene

Figure E-11: Di-sigma propylene

Figure E-12: Physisorbed propylene

Figure E-13: 1-propylidyne in fcc site

209 Appendix E

Figure E-14: 1-propenyl on top and in bridge site

Figure E-15: 2-propenyl on top and

in bridge site

E.2.4 Dissociation products

Figure E-16: On top methyl and ethyl

E.2.5 Transition states

Figure E-17: TS1 Propane – 1-propyl


Appendix E 210

Figure E-19: TS4 1-propyl – 1-propylidene

Figure E-20: TS5 1-propyl – propylene


Figure E-22: TS8 2-propyl – 2-propylidene

Figure E-23: 1-propylidene – 1-propylidyne

E.3 Optimized geometries in the 5×2 unit cell

E.3.1 Gasphase species


211 Appendix E

E.3.2 Hydrogen

Figure E-24: Hydrogen on top site

Figure E-25: Bridged hydrogen

Figure E-26: Hydrogen in fcc site

E.3.3 Dehydrogenated species

Figure E-27: Physisorbed propane


Figure E-29: Di-sigma propylene

Figure E-30: Physisorbed propylene

Appendix E 212

E.3.4 Transition states



Appendix F 213

Appendix F Optimized geometries on the Pt(211) catalyst model In this appendix, first the proposed reaction models are illustrated with the optimized


following order: gasphase, hydrogen, propane dehydrogenation products and transition states.

The following color code is employed for the elements: carbon (gray), hydrogen (white) and

platinum (blue). As platinum atoms are rather large, solely the two top layers of the catalyst

model is shown. Additionally, the radius of platinum is reduced from 1.40 Å to 1.00 Å. For the

optimized geometries, interatomic lengths are given between specific atoms. For intermediates,

the lengths of the C-C bonds, one C-H bonds and Pt-C bonds (and eventual Pt-H bond) are

determined. For transition states, the bond length between atoms that participate in the reaction

are reported. All distances are determined in Å.

214 Appendix F

F.1 Reaction network

Figure F-1: Reaction path via 2-propyl towards propylene on Pt(211) catalyst model. In the dehydrogenation reactions, hydrogen is formed. However, the formed hydrogen is not shown in this figure.

F.2 Optimized geometries

F.2.1 Gasphase species


2-propyl

7

Pt

_2

Gaseous propane

0

Physisorbed propane Propylene


Gaseous propylene

13

Appendix F 215

F.2.2 Hydrogen

Figure F-2: Hydrogen on top site near the edge

Figure F-3: Hydrogen on top site not near the edge

F.2.3 Dehydrogenated intermediates

Figure F-4: Physisorbed propane

Figure F-5: On top 2-propyl

216 Appendix F

Figure F-6: Di-sigma propylene

Figure F-7: Physisorbed propylene

F.2.4 Transition states

Figure F-8: TS2 Propane – 2-propyl

Figure F-9: TS7 2-propyl – propylene

Documents

Computational study of the catalytic dehydrogenation of ...lib.ugent.be/fulltxt/RUG01/002/224/444/RUG01-002224444_2015_0001_AC.pdf · Computational study of the catalytic dehydrogenation