Investigating reaction oscillations during CO oxidation on Pd(100) using optical microscopy

(1)

Investigating reaction oscillations

during CO oxidation on Pd

(

₁₀₀

)

using optical microscopy

THESIS

submitted in partial fulfillment of the requirements for the degree of

BACHELOR OF SCIENCE

in PHYSICS

Author : G. Westra

Student ID : 1298275

Supervisor : Dr. I.M.N. Groot

2ndcorrector : Prof.dr.ir. T.H. Oosterkamp

(2)

(3)

Investigating reaction oscillations

during CO oxidation on Pd

(

₁₀₀

)

using optical microscopy

G. Westra

Huygens-Kamerlingh Onnes Laboratory, Leiden University P.O. Box 9500, 2300 RA Leiden, The Netherlands

July 6, 2017

Abstract

Roughening and oxidation of the crystal catalyst Pd(100)surface, seen by the reflectivity of the surface, is due to the CO oxidation reaction. The

surface exhibits self-sustained reaction oscillations, which are investigated with the use of optical microscopy. Changing the temperature, pressure and reactant mixture conditions have a great effect

on the intensity and periodicity of these oscillations. The goal for this research is to relate the reactivity and surface structure of the sample and also to create reproducible experiments, for which the right gas mixture is

needed to cause the surface to exhibit these predictable reaction oscillations. Furthermore plotting the reflectivity data will reveal the texture of the sample in different moments of time. Combined with the

mass spectrometer data we investigate these reaction oscillations. Keywords: Heterogeneous catalysis, CO oxidation, reaction oscillations, optical

(4)

(5)

Chapter

1

Introduction

This thesis describes the research in the Operando research in heterogeneous catal-ysis group from Leiden University. This group is led by Dr. I.M.N. Groot. The research was done between February and June 2017.

Catalysts increase the reaction rate of a chemical reaction. They are sepa-rated into two types: heterogeneous and homogeneous catalysts. Study-ing the surface structure of sStudy-ingle-crystal model catalysts involves hetero-geneous catalysis, because of the difference in phase between catalysts and reactants, which means there is an interface between the two phases. This is in contract to homogeneous catalysis where there is no phase difference between the catalysts and reactants.

Heterogeneous catalysis is the basis of many areas in the chemical indus-try. The reaction mechanisms of catalysts, the understanding of how to adjust the reactivity and the elementary steps in surface catalysis play an important role in the research of heterogeneous catalysis. In recent times the investigation and development of catalysts was done by empir-ical research. Catalysts and reactants were studied by modelling and by surface-science techniques. Nowadays the geometrical surface structure determines the reaction rate. One must change the working conditions in the laboratory to more realistic high-pressure and high-temperature condi-tions, to bridge the gap between laboratory and industrial conditions. [1– 3]

The palladium sample we use in our experiments is a perfect crystal with an (100) crystallographic surface structure. We use this sample to describe the carbon monoxide oxidation reaction described by equation 1.1. This metal is an important component of the three-way catalytic converter. Ex-periments have shown that during CO oxidation at atmospheric pressure

(8)

lic surface to an oxide surface with high activity. [4] They predict with their model and considerations that the period of reaction oscillation is a func-tion of the temperature. However, there are differences between results of several experiments of different research groups of these spontaneous reaction oscillations. [4–6] This diversity in results is likely because of the differences in the surface structures. If we can look at the activity of the surface and the structure of the surface at the same time, we can relate those two.

This is exactly what we want to investigate. For our project we will use an optical microscope to investigate CO oxidation. The set-up can also be used to examine other oxidation reactions. We make use of mass spec-trometry, that is a technique for detecting chemical species like molecules. First an ion source ionizes the species, thereafter a mass analyzer sorts the ions by their mass and charge and finally a detector detects the amount of each species present. Further we make use of a gas supply system where we can mix reactant gases. This set-up can be heated and used to mea-sure the phenomenon of spontaneous reaction oscillations and also the formation of surface oxides in situ under high-pressure, high-temperature conditions. This can be done by looking at the reflectance of the sample. In this way we can find a relation between the activity and the structure. This analysis can give us more insight in the heterogeneous catalysis of CO oxidation. In the second chapter we give more information about CO ox-idation and the catalytic increase in the reaction on the palladium sample described by equation 1.1. Furthermore the set-up is described in the third chapter and in the fourth chapter we characterize the reaction oscillations and interpret the findings of the measurements.

2 CO+O2−−→ 2 CO2 (1.1)

The goal of this research is to answer the following research questions: What can we say about the reaction rate of reaction oscillations under dif-ferent pressures and temperatures for the same gas ratio of carbon monox-ide and oxygen? How do different parts of the surface of the sample spond to these reaction oscillations? Under what conditions can we re-produce our measurements and gain more understanding of these

(9)

spon-3

taneous oscillations?

The research is done in partial collaboration with Vladimir Calvi Chaushchak (Leiden University). The execution of the measurements is done together, but the design of the experiments, interpretation of the results and pro-cessing of the data is done individually.

(10)

(11)

Chapter

2

Theory

2.1 Palladium (100) crystal

Simple lattices help us to describe the crystal structure of a clean single crystal palladium. The palladium atoms form a single, homogeneous crys-tal lattice that is perpetual to the edges of the sample for a clean cryscrys-tal. To fully understand the geometry of these lattice, we first take a look at the terminology of these structures.

There are multiple definitions of the unit cell. When looking at a crys-tal lattice we break it down into small regions of space containing lattice point(s), which can be used to construct the full structure by repeatedly stacking them together, called unit cells. Of course it is possible to have more configurations for this cell. The primitive unit cell has the least vol-ume, i.e. only one lattice point, and a Wigner-Seitz cell has the same sym-metry as the lattice. The best way to describe the unit cell with formulas is to use the conventional unit cell, which has orthogonal axes.

We are only interested in cubic lattices, because they describe most of the crystals. To describe objects inside this cube, we define a basis. In this description we give the position of atoms with respect to a reference lat-tice point. With a basis and a latlat-tice we can describe the whole crystal structure. One writes a lattice vector

R_[_rst_] =ra1+sa2+ta3 (2.1)

where r, s and t are integers and a1, a2 and a3 are the basis vectors. The

crystal structure of palladium is a Face-Centered Cubic (fcc) lattice, also called a Cubic Close-Packed lattice (ccp). Furthermore we know that the length of all sides are the same: a = 389.07 pm and the axes are all per-pendicular. [7] The fcc lattice has an extra lattice point on every face of the

(12)

Figure 2.1: Face-centered cubic lattice sketch taken from Simon ”The Oxford Solid State Basics”.[8] On the left you see the 3D cubic cell and on the right the top view with a label on some of the lattice points. The distance between the lattice points is given by a. All lattice points describe palladium atoms.

cubic unit cell as you can see in figure 2.1. The next step is to define a point

Ginside a reciprocal lattice of the direct lattice with points R. This lattice is constructed in the reciprocal space. This means that

a_i·b_j =2πδij (2.2)

needs to be satisfied. With δij the Kronecker delta. The points G of the

reciprocal space vectors bj are given by:

G(h,k,l) =hb1+kb2+lb3 (2.3)

Where the integers h, k and l are called Miller indices. For our sample those are (hkl) = (100), meaning the lattice plane described by these indices is on one of the faces of the cubes and all the planes described are parallel to this plane.

2.2 Reaction mechanisms

To start a heterogeneous catalytic process there must be at least one of the reactants attached for a significant period of time to the surface of the catalyst, otherwise this cannot be called a reaction on a surface. In the previous paragraph we were speaking about a clean surface, but in prac-tice there will always be an amount of unfamiliar atoms attached to the surface. This atoms can only be attached by adsorption, i.e. by condensa-tion or by segregacondensa-tion from the bulk of the sample and are called adatoms. Movement of these atoms or molecules on the solid sample is called sur-face diffusion. The sample is the substrate and the reagent on the sursur-face

(13)

2.2 Reaction mechanisms 7

is called the adsorbate. Adding these chemical species together will result in a monomolecular reaction. [9]

We are interested in bimolecular reactions, which means that two molecules plus the surface are involved. When experimenting at the right tempera-tures, this results in three distinct reaction mechanisms taking place on the surface. Figure 2.2 shows these different mechanisms. The Langmuir-Hinshelwood reaction mechanism is where both species are adsorbed on the surface and form products by reactions with each other. After this cre-ation the product leaves the surface by desorption. It is the most common mechanism and the three dynamics of the Langmuir-Hinshelwood mech-anism are adsorption, surface diffusion and desorption. Another mecha-nism is called the Eley-Rideal mechamecha-nism where only one species is bound and another species reacts directly with it from the gas phase. The L-H kinetics is the one which is observed in CO oxidation on platinum and palladium at low pressures. [3, 10–12]

At higher oxygen pressures CO oxidation will follow Mars-Van Kreve-len reaction mechanism: A reactant bonds with the surface, creating metal oxides, and the other reactant reacts directly with the O atom from the ox-ide, so the active part is the surface itself in this reaction. This will result in the formation of a thin layer of palladium oxide and the formation of vacancies in the surface that will be filled with the reactant again. The L-H and M-VK mechanisms describe the CO oxidation on the Pd sample: [5, 13]

Langmuir-Hinshelwood mechanism (Low pressure)

O2+2∗ −−→2 Oads (2.4)

CO+ ∗ COads (2.5)

COads+Oads −−→ CO2+2∗ (2.6)

Mars-Van Krevelen mechanism (High pressure)

O2+2 Pd−−→ 2 PdO (2.7)

CO+PdOCO2+Pd (2.8)

(14)

Figure 2.2: Different reaction mechanisms for catalysis from gas and surface reactions. Part A describes the Langmuir-Hinshelwood mechanism, part B the Eley-Rideal mechanism and part C the Mars-Van Krevelen mechanism. Reac-tion A1 (= equaReac-tion 2.5), reacReac-tion A2 (= equaReac-tion 2.4) and reacReac-tion B1 are ad-sorption reactions, reaction A3 (= equation 2.6), reaction B2 and reaction C1a (= equation 2.8) are desorption reactions. Reaction C1a (= equation 2.7) is where the reactant forms a chemical bond with the surface. Figure taken from Herbschleb. (2011) [13]

2.3 Stages of oxidation

Pd(100)oxidises at high pressure through four stages. [12] First is the cre-ation of an oxygen overlayer by adsorption, second the ejection of pal-ladium atoms on terraces creating islands, after that the surface

recon-structs to a (5 x 5) structure or for temperatures above 400 K a (√5 x

√

5)R27◦structure of PdO(101). Finally the surface roughens due to the appearance of three-dimensional clusters with a different crystallographic structure of PdO and are reactive to reduction with CO. [12] Zheng and Altman determined that the following oxygen states can exist on the

palla-dium surface, ranked by the temperature needed for CO2production: [14]

• Production of CO2below 335 K:

– c(2 x 2) oxygen overlayer

• Production of CO2between 375 K and 405 K:

– (5 x 5) surface

– Bulk PdO

– p(2 x 2) overlayer

– (2 x 2) high density state

• Production of CO2for temperatures above 400 K:

(15)

2.4 Reaction oscillations 9

The process of catalysis is non-equilibrium one, because reactant mole-cules are continually provided and the products are removed. So each reaction results in a structure that is different than in equilibrium. In order to investigate the surface you have to do research during actual reactions (in situ). The rates of individual processes and the structure are coupled and could lead to reaction oscillations. Only during the reaction this struc-ture could be described as stable. [12] The roughening and smoothening of Pd(100) causes a change in the reflectivity of the sample that can be seen with an optical microscope during actual heterogeneous catalysis. The observed changes in intensity of reflection of the sample is due to the ap-pearance and disapap-pearance of the oxide layer on this sample. [15] When the reflecting surface is smooth, it means that the irregularities on the sur-face are smaller than the wavelength of an incident light beam. However when the surface is rough, the incident light rays are going to diffract in every way resulting in diffuse reflection. The reflecting behaviour of our surface lies in between. [16]

2.4 Reaction oscillations

The oscillatory behaviour of CO oxidation on the palladium(100) sam-ple is previously observed in scanning probe microscopy (SPM) experi-ments. [17] Hendriksen and colleagues made a model in which these os-cillations go through 4 stages. They have also done their experiments

with Synchotron X-Ray Diffraction (SXRD) experiments on Pd(001) and

made clear that this model is also suitable for other palladium crystal sur-faces. [4] These stages are shown in figure 2.3.

When the CO oxidation starting point is a rough oxidized palladium sample this means that at high pressures the sample will follow the Mars-Van Krevelen reaction mechanism. This mechanism will make the sur-face more rough and at a certain point there will be an abrupt switch to the Langmuir-Hinshelwood reaction mechanism because of a decrease in PdO, which results in a phase transition on the surface. This is

accom-panied by an increase in CO pressure, a decrease in CO2pressure and an

increase in brightness. The Langmuir-Hinshelwood mechanism makes the surface less rough, as you can see in figure 2.3, which leads to a next phase transition, where again there will be an abrupt switch, but this time back to the Mars-Van Krevelen mechanism. Due to this, the surface gets rough again, which means the cycle is complete.

(16)

Figure 2.3: Reaction oscillations cycle as described by Hendriksen 2010.[4] We distinguish 4 stages and 4 transitions. The dark areas on the palladium surface from the figure are palladiumoxides. Transition A and transition C are phase tran-sitions. Variations in CO pressure, CO2pressure and surface brightness happen

because of these transitions. The CO2production increases when the surface is in

the oxide phase. The brightness increases because the surface of the oxygen over-layer is less reflective. Transition B is the smoothening of the surface due to the Langmuir-Hinshelwood reaction mechanism and transition D is the roughening of the surface due to the Mars-Van Krevelen reaction mechanism.

(17)

Chapter

3

Set-up

A schematic representation of the set-up we used for the measurements is shown in figure 3.1. Every component can be fully controlled by the computer, except for the opening of the gas bottles inside the gas supply system and the intensity of the LED for the optical microscope. This micro-scope is used to take pictures of the zoomed in sample and the brightness of these pictures can be calculated. The sample is inside a small flow reac-tor of 16 ml that is build by the Leiden Probe Microscopy B.V. (LPM). [18] A gas supply system (LPM) can be used to set the flow rate of the

reac-tant gasses with a maximum of 10 mln/min. The gases coming out of the

reactor chamber are measured with a mass spectrometer (LPM) and the residual gases are ventilated by the fume hood. The LED light source is used to illuminate the sample and a camera is used to record the out com-ing signal. The sample is heated with a heatcom-ing element and can be heated

Figure 3.1: Block diagram of the set-up used for the measurements. The gas supply system, optical microscope and mass spectrometer are controlled by the computer. The gases supplied enter the optical microscope and after that the mass spectrometer, before the gases come back and leave via the fume hood.

(18)

Shown in figure 3.2 is a schematic design of the complete optical micro-scope. The sample is inside the reactor which is not indicated in this fig-ure. This reactor can be seen in figure 3.3. The gases coming from the gas supply system enter the reactor and leave the reactor towards the mass spectrometer. The window on top is transparent and the reactor is closed in vacuum, but not in Ultra-High Vacuum (UHV). The pre-sample part of the microscope consists of a 625 nm mounted LED light, which has a maximum intensity of 700 mW/min, but we cannot measure the intensity while doing experiments. It is the M625L3 LED with a collimator both from Thorlabs. [21] The parallel emitted light first passes a lens and then a diffuser to filter unwanted spatial patterns. Immediately after passing a pinhole, another lens and a diaphragm, the light passes through to a beam splitter, where half of its intensity is reflected and half is transmitted. The reflected light then reflects on the surface of the heated sample in the flow reactor and is transmitted through: the beam splitter, a lens, another pin-hole and one more lens before it reaches the CCD camera. The resolution of this camera is 2592 x 1944 pixels and it is an acA2500 - 14gm Basler camera. Data is processed with the Pylon API 4 software. Output format: Tiff. [19]

3.2 Thermocouple and heating

We measure the temperature of the sample with a thermocouple wire type-C, which consists of tungsten and rhenium alloys, because they can be used to measure at quite high temperatures in an oxygen atmosphere. These wires are laser beam welded on the surface. Since the National In-struments temperature input device we use, the NI USB TC01 [22], can-not measure the temperature for this type, hence we need to convert the voltage, that is already ADC-converted by the box, with a LabVIEW pro-gram [22] to a type-C temperature. We only need to know the Cold-Junction Compensation (CJC), because the device does not perform any

(19)

3.2 Thermocouple and heating 13

Figure 3.2: Optical microscope from the set-up based upon figures 2 and 3 from Onderwaater 2017.[19] The LED light has a wavelength of 625 nm. The com-ponents are: an aspherical collimator: focus of 25 mm, lens L2: 100 mm focal length, diffuser: scotch tape, pinhole, lens L3: 60 mm focal length, diaphragm, beam splitter, sample, lens L4: 100 mm focal length, pinhole, lens L5: 60 mm focal length and the detector: CCD camera.

(20)

Figure 3.3: Graphic view of the reactor, not in scale. A is the complete reactor and B is the sample labeled 11 on the sample holder with heater label 14. The water cooling, label 5 and label 6 are not used. Label 10 indicates the thermo-couple wires and the heater feed troughs are connected to the heating element of label 14. Label 9 marks the gas inlet connected with label 16 and the gas outlet is marked by label 17. The rest of the components are described in and this figure is based upon figure 5 from Bremmer 2013. [20]

0.02 mV at room temperature. With this CJC and the measured voltage, we can use equation 3.1 and 3.2 to calculate the temperature. The coeffi-cients biare given by NI. [22]

VType−C =VMeasured+VCJC (3.1) T(VType−C) = n

∑

i=0 bi(VType−C)i (3.2)

The heating is done with a power supply which has a voltage range of 0 to 30 volt and a current range of 0 to 5 ampere. To keep a constant temperature we apply feedback between the readout of the temperature and the heating power.

3.3 Gas supply system and mass spectrometer

The gas supply system we use for the experiments is shown in figure 3.4. Before experimenting we have to open the gas bottles by hand. The CO gas bottle is contaminated with metal carbonyl content, so we make use of the LPM Carbonyltrap 2.1. [18] The mass flow controllers, the pump and mass spectrometer can be controlled by the computer. We can mix

(21)

3.3 Gas supply system and mass spectrometer 15

Figure 3.4: Gas supply system from the set-up. This figure is based upon the layout of the gas supply system GC4 designed by LPM. [18] The three gases can be mixed after opening the valves with multiple mass flow controllers (MFC). The quadrupole mass analyzer (QMS) analyzes the gases and the back pressure of the reactor is regulated (BPC) before gases reach the pump.

gases in different ratios and save data with a python program. After the residual gases leave the reactor they enter to a connected LPM T100 gas analyzer [18], which contains a quadrupole mass spectrometer that can measure molar masses up to 100 amu. The gases pass a leak valve we can manually change to control the gas inlet. The software we use for this spectrometer is RGA software for Windows. We measure the pressure of the gases in Torr versus the time in hours and minutes. With a python

program we can convert pressure to mbar. 1 mbar≈0.750 Torr.

Timing is important for conducting experiments. We have to moni-tor pictures, measure the gases coming out of the reacmoni-tor, the gas inlet from the gas supply system and the temperature of the sample at the same time. The refresh time of the reactor has been previously calculated and measured by PhD. candidate G.M. Bremmer and is approximately 230 sec-onds. We need to take this into account when doing experiments. Because this synchronization in time is important we make use of Spacetime [23] open source software. It is written in Python and developed particularly for datasets of heterogeneous catalysis. Gas flow data, pictures, mass spec-trometer data, reflectance data and also temperature data can be presented together. The graphs in the third chapter are made with this program.

(22)

(23)

Chapter

4

Sample brightness

In this chapter we discus some of the results we found during experiments on the palladium sample surface. The pressure we discus is the total pres-sure of the mixture of gases entering the reactor. When drafting experi-ments we wanted to have a clear starting point. Looking at the possible surface structures of figure 2.3 for these reaction oscillations we choose a smooth metal as our starting point, because we expect we could create this surface texture by smoothening the surface and also the reduction of the oxygen overlayer with a CO gas flow. However creating this surface is harder than it seems as will be shown in the following chapter.

4.1 About the sample

The polished sample we used for our experiment was actually contami-nated with nickel and silicon. Probably this has no effect on the CO oxida-tion reacoxida-tion mechanism on the surface, but we do not want to explain cer-tain phenomena, other than this oxidation reaction, that happen because

of this contamination. That is why we cleaned another Pd(100) sample

after we have done some experiments. This cleaning has been done by several rounds of sputtering with argon ions and annealing up to 1100 K in an Ultra-High Vacuum (UHV) chamber. Experiment E discusses the re-sults from the clean sample and we checked the surface by Auger Electron Spectroscopy (AES) with an electron beam of 2000 eV and 0.075 mA emis-sion. We have found no nickel contamination on the surface of the second sample.

After the acquisition of pictures we can analyse the brightness of these pictures with a python program, created by another bachelor student Jeroen

(24)

Figure 4.1: Examined areas on a picture (2592 x 1944 pixels) of the sample taken with the Basler camera. On the left is indicated where the gas flow is coming from. On the right you see the areas examined after experimenting: Area A: x = 1950:2050, y = 220:320, area B: x = 1175:1275, y = 200:300, area C: x = 1150:1250, y = 970:1070, area D: x = 1950:2050, y = 1090:1190 and area E: x = 460:560, y = 1300:1400.

van Doorn, but adjusted to new areas. These new areas are chosen at ran-dom as you can see in figure 4.1. The area closest to where the reactant gases enter the reactor is area A and area E is the furthest away. This python program averages the colour intensity of the small areas and re-turns one value for each area per unit time when the pictures were taken. Furthermore, as previously mentioned, we cannot exclude the fact that due to light intensity changes the light exposure reached on the camera differs between pictures. Therefore we calculate a scientifically relevant value called the normalized reflected intensity of the sample given by the following equation:

∆R = I−I0

I0 (4.1)

where I is the measured intensity and I0is the maximum reflected

in-tensity during the experiments. This is preferably at the start of our exper-iments, but since we cannot analyse the pictures during actual measure-ments, we have no indication for the reflected intensity when starting an experiment.

(25)

4.2 Expectations 19

4.2 Expectations

We first take a look at the possibility of adsorption by one of the reactant gases on the surface or other possible disturbing chemical effects before we mix the reactants. There are three reactant gases we use in our set-up. Oxygen is responsible for the oxidation and thus the creation of an oxygen overlayer. Argon gas is an inert gas, which means this gas does not react with the other gases and the surface of the sample. We use it to flush the reactor to remove gases out of the supply system and those that are present due to the desorption from either the surface of the sample or the inside surface of the reactor. We have seen during measurements that argon gas flow does not influence the reflectivity of the sample.

Carbon monoxide is the reactant gas which becomes oxidized. Without the presence of oxygen we expect that at low pressures the CO gas will adsorb on the surface due to one of the adsorption reactions the Langmuir-Hinshelwood mechanism described in the first chapter. This results in a small increase in carbon dioxide production, but at high pressures the CO gas will react with the oxygen overlayer formed on the surface and will not adsorb on the surface because of the Mars-Van Krevelen mechanism. So for the following experiments we expect that at high temperature and high pressure conditions the CO reactant gas will reduce the surface and at low pressure conditions the surface gets smoother.

4.3 Experiment A: Smoothening and roughening

Table 4.1 and figure 4.2 display the results of experiment A. Images are sampled once every second and we try to create a smooth metallic surface, out of a surface that is oxidized for a small amount. The experiment starts

by letting a 3 mln/min CO gas flow inside the reactor for 4 hours at a

constant pressure of 0.3 bar and constant temperature of 300◦C.

Table 4.1: Experiment A.Smoothening and roughening

Time CO (mln/min) Pressure (bar) Temperature (◦C)

12:34 3 0.3 300

16:34 - 0.3 300

We observe no changes in the CO pressure inside the mass spectrome-ter and also we see no differences in temperature measured on the sample. But we do see differences in the reflectance of the surface. From the start of the measurement the surface starts with a small dip in brightness and after

(26)

Figure 4.2: Experiment A. Coloured graphs of pure CO gas flow. First graph: Gas flow data in mln/min. Second graph: Partial pressure data in 10e-7 mbar. Third

graph: Normalized reflectance data. Fourth graph: Temperature of the sample in

(27)

approximately 130 minutes of almost constant brightness the surface gets more shiny. However after 168 minutes this value decreases again and af-ter 4 hours almost all parts of the sample examined are as much reflective as at the start of the experiment.

The reason for this behaviour is that the CO molecules do not ad-sorb on the surface following step 2.5 of the Langmuir-Hinshelwood re-action mechanism, which rere-action is responsible for the smoothening of the surface, and when they do adsorb, they get desorbed after a short pe-riod of time. Generally for the Langmuir-Hinshelwood reaction mecha-nism one reaction is the lowest and this reaction determines the complete mechanism. The literature tells us that this is the Rate-Determining Step (RDS). [10] In the case of experiment A this step can be conditioned by the reactant gas flow pressures, the coverage of the surface or the tempera-ture of the sample. An experiment at a pressure of 1.0 bar with the same gas flow and temperature did not show adsorption of CO on the surface,

which is necessary for the formation of CO2, see step 2.6, and thus the

smoothening of the surface. In our set-up only oxygen and CO can cover the surface and we know that the adsorption of CO on an oxygen covered surface is possible. [10] Further experiments with pure CO gas flow at a

sample temperature of 350◦C also showed no CO adsorption.

There could be another possibility for the rate-determining step de-pending on something else then these reaction conditions, but we predict that the CO oxidation over this surface is due to the Mars-Van Krevelen mechanism as previously reported by others. [17] This means that if we want to smoothen the surface, the reaction of CO with PdO resulting in CO2 and surface sites (step 2.8) should happen faster than the creation of

PdO (step 2.7). We were not able to find the best mixture of CO and oxygen reactant gases applicable for this solution of the problem.

4.4 Experiment B: Oxidation and reduction

The goal for the next experiment is to reduce an oxidized surface. The experiment consists of oxidizing a surface followed by flushing the reactor and then adding a CO gas flow. The expectation is that the surface gets brighter. Mixing reactant gases with argon and flush with argon ensures no gaseous oxygen present. The CO gas flow is able to reduce the sample and the only present oxygen is adsorbed on the surface.

Table 4.2 and figure 4.3 display the results of experiment B. The

tem-perature of 350 ◦C and the pressure of 0.2 bar are constant. Images are

(28)

mix-Time Ar (mln/min) CO (mln/min) O2(mln/min) Press. (bar) Temp ( C)

15:16 2 - 3 0.2 350

16:16 5 - - 0.2 350

16:26 2 3 - 0.2 350

17:26 - - - 0.2 350

We clearly see for the mass spectrometer data that the increased step

in CO2 pressure is due to the added CO gas to the reactor. This step is

very fast, about 5 seconds, but for the reflectivity data this is faster than a second. It has become evident that the reduction of the oxidized surface can only be investigated if we sample pictures faster than once a second.

If we now look back at figure 2.3 we see that this reduction for reac-tion oscillareac-tions is described by cycle A, the PdO overlayer decreases from the surface and the Mars-Van Krevelen reaction mechanism is replaced by the Langmuir-Hinshelwood mechanism. But this is not the reduction that happens here. The L-H mechanism assumes an adsorption of CO on the surface (step 2.5) before the productions of CO2(step 2.6) and experiment

A just showed that there was no CO adsorption at this pressure or tem-perature! For the oxidation of PdO both reaction mechanisms describe the same chemical reaction. The only conclusion is that during the oxidation and reduction of PdO the M-VK mechanism was still responsible for the formation of CO2and surface sites (step 2.7 and 2.8).

This is a very interesting result, because we have not yet been able to smoothen the surface with just one reactant gas. The M-VK mechanism roughens the surface and only the L-H mechanism is able to smoothen the surface. What we do know from the theory described in chapter 2 is that there is a possibility that the reduced surface follows the L-H mecha-nism during reaction oscillations. That is why we now want to take a look at these reaction oscillations, despite we are still unable to create a clear starting point from which we can start doing experiments. And also to gain more information regarding our research questions.

(29)

Figure 4.3: Experiment B. Coloured graphs of oxidation and reduction. First graph: Gas flow data in mln/min. Second graph: Partial pressure data in 10e-7

mbar. Third graph: Normalized reflectance data. Fourth graph: Temperature of the sample in◦C. The time is depicted on the x-axes and all graphs are synchro-nized in time.

(30)

(31)

Chapter

5

Reaction oscillations

From experiment A we suggest that smoothening the surface can be done with the right mixture of CO and oxygen, but this mixture is also respon-sible for the reaction oscillations. We can only reach a rough metal with the knowledge from experiment B. That means we know how to control the oxidation and reduction of the surface by adding reactant gases to the reactor. The decrease in reflectivity is due to both the oxidation and rough-ening of the surface and this is previously mentioned by Willem Onder-waater [15]. We want to know how much the roughness interferes with the measurements, even though we are looking at the normalized reflected light intensity difference.

Now we know that the starting conditions of the following experi-ments are always different it will be more difficult to gain more under-standing of these spontaneous oscillations. But we can try to find these oscillations and describe the conditions for which these happen and also see what happens at different parts of the sample. Also the observations can be used for future experiments.

Preferably for the start of each experiment the surface of the sample needs to be prepared to reach a smooth metallic texture. We start our ex-periments with a rough metallic structure.

5.1 Conditions

Both reaction mechanisms produce CO2, so we want to find the

temper-ature where the production of CO2 starts. The experiment we did was

ramping a temperature between 100 ◦C and 360 ◦C by increasing it 1 ◦C

(32)

oscillations, see figure 2.3, showed that at 174◦C and a gas flow ratio of

CO:O2:Ar = 25:500:675, spontaneously oscillatory behaviour was found on

a Pd(001) sample. We wanted to recreate these and therefore we started

doing multiple experiments with a gas flow ratio of CO:O2:Ar = 1:20:28

and each experiment with a duration of 120 minutes. The results are de-scribed in table 5.1.

Table 5.1: Experimental observations. Experiments at different temperatures and pressures with a gas flow ratio CO:O2:Ar = 1:20:28 for 120 minutes.

Temperature (◦C) Pressure (bar) Reaction oscillation observation

175 0.2 No

175 0.5 Yes

175 1.0 No

315 1.0 No

Oscillations with a bigger intensity are found at different gas ratios for our set-up. At a gas ratio of CO:O2= 1:5 we found no oscillations, but this

was also not expected. Experimenting with a gas ratio of CO:O2 = 1:10

gave us the results shown in table 5.2. This lead to the following optimal

conditions for our set-up: A temperature of 165 ◦C and a pressure of 1.0

bar for this ratio.

The oscillations we found at 200 ◦C had a much longer oscillation

cy-cle than experiments of 165 ◦C. We did not observe this alternations for

temperatures above 300◦C and below 155◦C. We use the following

termi-nology; introduced by Yuranov [24]: high, low and intermediate-reactivity states. The self-sustained oscillations are in an intermediate-reactivity state. Experiments have shown that a low-reactivity state is correlated with an oxidized surface and a high-reactivity state correlated with a reduced sur-face.

(33)

Table 5.2: Experimental observations. Experiments at different temperatures and pressures with a gas flow ratio CO:O2:Ar = 3:30:20 for at least 50 minutes.

Temp. (◦C) Pressure (bar) Reaction oscillation observation Reactivity state

135 1.25 No Low

155 1.0 No Low

165 0.2 Yes Intermediate

165 (exp. C) 1.0 Yes Intermediate

200 1.0 Yes (low activity) Intermediate

300 1.0 No High

5.2 Experiment C: Reaction oscillations for 2 hours

From the literature we know that the oscillations are limited to a regime of temperatures with the matching pressures. [24] This is the reason why we only find sustained oscillations at a gas ratio of CO:O2 = 1:10 for 165

±10◦C and 1.0 bar pressure as you can see in table 5.2. The expectation

is that adding a gas mixture of CO and oxygen to a rough and reduced surface, the surface will start from a high-reactivity state. We do not know if and how long it will take for the surface to reach the intermediate re-activity state downwards from the high-re-activity state. This will happen if the right temperature and pressure are applied and the surface reaches an equilibrium state equal to the intermediate state.

Table 5.3 and figure 5.1 display the results of experiment C. We distin-guish a preparation stage and an experimental stage. For the preparation stage the temperature of 300◦C and pressure of 0.3 bar are constant. First

we start with a 15-minute-long 3 mln/min CO flow for the reducing of the

sample. Second we flush the reactor with a 5 mln/min argon flow for 10

minutes which is longer than the prescribed 230 seconds from chapter 2. The experimental stage is where there is a constant temperature of 165

◦_{C, after the sample cooled down, and where there is a pressure of 1.0 bar,}

after the pressure in the reactor was increased. Images are sampled once

every second and the experiment starts by letting a mixture of 6 mln/min

argon flow and 9 mln/min CO gas flow inside the reactor for 10 minutes.

We do this to see the change in CO2production when we switch to a

mix-ture of 6 mln/min argon flow, 0.9 mln/min CO flow and 9 mln/min oxygen

(34)

Figure 5.1: Experiment C. Coloured graphs of oscillation reactions. First graph: Gas flow data in mln/min. Second graph: Partial pressure data in 10e-7 mbar.

Third graph: Normalized reflectance data. Fourth graph: Temperature of the sample in◦C. The time is depicted on the x-axes and all graphs are synchronized in time.

(35)

Table 5.3: Experiment C.Reaction oscillations for 2 hours

Time Ar (mln/min) CO(mln/min) O2 (mln/min) Pressure (bar) Temp. (◦C)

15:00 - 3 - 0.3 300

15:15 5 - - 0.3 300

15:25 6 9 - 1.0 165

15:35 6 0.9 9 1.0 165

17:35 - - - 1.0 165

There it is, the first disclosure of measured reaction oscillations in this thesis in the experimental stage. Nonetheless we have to check if the os-cillations match with the work of Hendriksen described in figure 2.3 and

this is a positive match. The CO2 pressure decreases in anti-phase with

CO pressure. The mass spectrometer data also shows no alteration in ar-gon and oxygen pressure, as expected.

The temperature of the sample reveals the start of the oscillation and

when we zoom in we see the temperature first drops. After the CO2

pres-sure increases again the temperature also increases. We explain the drop in temperature as the fact that for the CO2production, the reformation of

chemical bonds of CO to CO2 releases energy resulting in an exothermic

reaction. When there is less CO2production the constantly heated sample

starts to cool down and the heater will start, with the use of feedback, to

heat the sample again to 165◦C. The sudden increase in temperature then

is when the production of CO2starts to increase again.

It is good to determine that we see these fluctuations also in the reflec-tivity data. All 5 parts of the surface get instantly brighter at the start of the oscillations. The reason given in figure 2.3 for this behaviour is the switch in mechanism. We see that is also a good explanation for the case of experiment C. If this oscillation is still in the M-VK reaction mechanism,

an increase of CO pressure would lead to more CO2production. Exact the

opposite is what happens. The increase of CO pressure leads to more CO

adsorption and less CO2 production, following the L-H reaction

mecha-nism.

If our statement that the L-H reaction mechanism is responsible for the increase of normalized reflectance is correct, than this would mean that the PdO layer on the surface of the sample is decreased for the duration of approximately 20 seconds in our measurements. The abrupt switch back to the M-VK reaction mechanism is described in figure 2.3 as a phase tran-sition from a smooth metal to a smooth oxide, meaning the fast growth of a PdO layer. The experimental data shows that after the oscillation the

(36)

Figure 5.2 shows the reflectivity and mass spectrometer graphs from three times the execution of experiment C with each time a different outcome for the oscillation period. Experiment C1 is the same as described in figure 5.1.

The cut out part is the experimental part with the mixture of 6 mln/min

argon flow, 0.9 mln/min CO flow and 9 mln/min oxygen flow inside the

reactor for 2 hours. Experiment C1 shows a longer period before the start of the first oscillation then experiment C2 and C3. Also there is a lot of variety in intensity of CO2and CO pressure alteration as we can see by the

mass spectrometer data.

Would this mean that 2 hours is too short to reach an equilibrium state of sample reflectivity and CO2production for which the oscillation period

between the oscillations is reproducible? Investigating this is the objective for the next experiment D. Also due to memory effects on the surface, after a lot of experiments, the surface is more rough and the oscillations are far more complex than we previously expected.

5.4 Experiment D: Reaction oscillations for

approx-imately 22 hours

The aim for the following experiment is the creation of predictable reaction oscillations. We expect that after an unknown amount of time, different for each experiment after the start, the period between each reaction oscilla-tion and the intensity and duraoscilla-tion of each oscillaoscilla-tion will be the same by prolonging the experiment.

Table 5.4 and figure 5.3 display the results of experiment D. This time we increase the experimental stage so the preparation also should be longer. For the preparation stage the temperature of 300◦C and pressure of 1.0 bar

are constant. First we start with a 30-minute-long 6 mln/min CO flow for

the reducing of the sample. Second we flush the reactor for 20 minutes

with a 5 mln/min argon flow which is longer than the prescribed 230

(37)

5.4 Experiment D: Reaction oscillations for approximately 22 hours 31

Figure 5.2: Three times experiment C.Coloured graphs of oscillation reactions. Left graphs: Normalized reflectance data. Right graphs: Partial pressure data in 10e-7 mbar. The 2 hours durations are depicted on the x-axes and all graphs are synchronized in time.

(38)

Table 5.4: Experiment D:Reaction oscillations for approximately 22 hours

Time Ar (mln/min) CO (mln/min) O2(mln/min) Pressure (bar) Temp. (◦C)

15:07 - 6 - 1.0 300

15:37 5 - - 1.0 300

15:57 6 9 - 1.0 165

16:07 6 0.9 9 1.0 165

13:33 - - - 1.0 165

Despite the difference in preparation stage of our experiment with re-spect to experiment C , the mass re-spectrometer data seems to be to some extent the same at the experimental stage. In this almost a day long ex-periment we see a lot of varieties in intensity strength and time periods between peaks of oscillations. We conclude that the memory effect is re-sponsible for these deviations. By sputtering and annealing of this surface in another set-up, this effect can be temporarily excluded from the surface. We observe for the temperature data that before the start of each oscil-lation, directly after another osciloscil-lation, there is a difference in fluctuation of the temperature. The temperature first fluctuates for about 0.05◦C and after approximately 30 minutes, just for the start of an oscillation, this is increased to a 0.15◦C fluctuation from 165◦C. Whereas the oscillations are

accompanied by an approximately 0.6◦C fluctuation. These observations

are related to the whole 22 hours experimental stage.

The reflectivity data gives another reason to assume surface memory. All parts of the surface, and especially the parts closest to the entrance of the gas flow inside the reactor, decrease in reflectivity when we disre-gard the oscillations. Part E, the part the farthest away from where the reactants enter the reactor, decreases the least. There is a temperature gra-dient, caused by the placement of the heater closer to the middle of the sample, which could explain differences in reactivity. But part A and part

(39)

5.4 Experiment D: Reaction oscillations for approximately 22 hours 33

Figure 5.3: Experiment D. Reaction oscillations for approximately 22 hours. Coloured graphs of oscillation reactions. First graph: Gas flow data in mln/min.

Second graph: Partial pressure data in 10e-7 mbar. Third graph: Normalized reflectance data. Fourth graph: Temperature of the sample in ◦C. The time is depicted on the x-axes and all graphs are synchronized in time.

(40)

colder with respect to part C they decrease faster in reflectivity.

We see that when we also take in account the oscillations, the behaviour at the first two hours correspond with experiment C. However 220 min-utes after the start there is a little bump in reflectivity data. When repeat-ing this experiment we saw that the reflectivity data showed more of these random occurrences. Since the temperature and mass spectrometer data show no sign of direct relation with these events we have to conclude that this is due to the surface roughness. Although we do not observe these

events having an effect on the temperature and CO2 production, we

ex-pect that this will ensure irregularities in oscillation period and intensity. This demonstrates the importance of a smooth surface at the start of the experiments.

5.5 Grid analysis

To check if these reflectivity changes are only locally, on an atomically scale, or from a bigger area on the surface, we analyse the brightness of these pictures again, but with a new python program. This time the areas are not randomly chosen, but on a grid. The location of this areas and the results of the examination of experiment D are shown in figure 5.4.

We observe that all the 5 areas in column E are less reflective after the experiment. Column C and row 3 are the most reflective at the end. This indicates that the temperature gradient is responsible for the decreasing reflectivity on the edges and the middle of the sample. The little bump is seen on all 25 areas so this is not a local effect. We made sure that the optical microscope is fully covered so the light from outside was blocked during the experiment. Part E from the first set of parts examined in fig-ure 5.3 of experiment D showed a less decrease in reflectivity, even though this part is laying on the edge of the sample. In our new analysis we took an area closer to the edge: area A5, and we saw that this area also de-creases faster. The parts A1, E1, A5, and E5 are the corner points from our grid and they all decrease faster in reflectivity than the middle parts. This results in the conclusion that indeed the temperature gradient is indirectly

(41)

5.5 Grid analysis 35

Figure 5.4: Examined grid areas on a picture (2592 x 1944 pixels) of the sample taken with the Basler camera. On the upper left are the areas indicated that are investigated after experimenting. All the areas are 50 by 50 pixels. The location for the upper left corners from each row are: row 1: 200, row 2: 600, row 3: 1000, row 4: 1400, row 5: 1800. The location for the upper left corners from each column are: column A: 500, column B: 900, column C: 1300, column D: 1700 and column E: 2100. The duration of the experimental stage, as depicted on the x-axes, is 1346 minutes.

(42)

This experiment is done with a clean sample. We do not know whether the determined optimal conditions for our first sample at a gas ratio of CO:O2= 1:10 also apply to the new sample. We expect reaction oscillations

happening and this will confirm the expectation that the contamination, previously mentioned at the start of chapter 3, is not responsible for these oscillations.

Table 5.5 and figure 5.5 display the results of experiment E. We tried to recreate experiment C, but this time without the preparation stage, we

only flush the reactor with a 5 mln/min argon flow for 10 minutes. The

expectation is that the surface is smooth and reduced, due to the cleaning, so the sample is in the high-reactivity state and therefore the first thing that will happen is the creation of a PdO layer before the surface starts to oscillate.

Table 5.5: Experiment E.Reaction oscillations for approximately 2 hours

Time Ar (mln/min) CO(mln/min) O2(mln/min) Press. (bar) Temp. (◦C)

14:55 5 - - 1.0 165

15:05 6 9 - 1.0 165

15:15 6 0.9 9 1.0 165

17:15 - - - 1.0 165

The experimental stage is where there is a constant temperature of 165

◦_{C and where there is a constant pressure of 1.0 bar inside the reactor.}

The experiment starts by letting a mixture of 6 mln/min argon flow and

9 mln/min CO gas flow inside the reactor for 10 minutes. After this we

switch to a mixture of 6 mln/min argon flow, 0.9 mln/min CO flow and 9

mln/min oxygen flow inside the reactor for 2 hours. The result of this two

hours is shown in figure 5.5.

Only the mass spectrometer data is visualized, because we are inter-ested in the oscillation part of the partial pressures. For the reflectance

(43)

5.6 Experiment E: New Pd(100)sample 37

Figure 5.5: Experiment E: Coloured graph of reaction oscillations for approxi-mately 2 hours. Partial pressure data is in 10e-7 mbar The time is depicted on the x-axis.

data we need to choose new areas on the sample and this goes beyond the scope of our research, because for this we need to repeat the experiment several times and this could be done for future experiments. The argon and oxygen pressure are constant after the start, just as we saw in exper-iment C. The oscillations start to happen after a sudden increased step in

CO pressure and less CO2production.

Our expectation that first there should be a creation of PdO on the sam-ple seems to be confirmed by the experiment. The surface of the samsam-ple is first highly reactive to the production of CO2because of this created layer

in the high reactivity state, but after about 30 minutes this production de-creases, probably due to reaching the intermediate reactivity state. The

CO2production decreases when the CO pressure increases just as we saw

in experiment C. The oscillations are still not stable, but we have the con-firmation that the contamination is not responsible for the oscillations, as the oscillation from experiment C and experiment E behave the same. The only differences between the two is the intensity and oscillation period, but this is most likely due to the memory effect on the surface.

(44)

(45)

Chapter

6

Conclusions

Reaction oscillations The reaction oscillations during CO oxidation on Pd(100)happen in an intermediate reactivity state and the switch from the Mars-Van Krevelen reaction mechanism to the Langmuir-Hinshelwood

re-action mechanism is responsible for the decrease in CO2 production and

increase in CO production.

Reflectivity of the sampleThe surface of the Pd(100)sample exhibits sur-face memory and the roughening of the sample, due to the Mars-Van Krev-elen reaction mechanism, happens faster at the edges of the sample. The parts of the sample with a higher temperature decrease less in reflectiv-ity and are more reactive. This difference occurs for the following reasons: First because the edge is where reactant gases enter the reactor and second because there is a temperature gradient from the middle of the sample to the edges of the sample due to the placement of the heater.

Reaction rate of reaction oscillations The CO oxidation reaction on the surface takes place above a certain threshold temperature. It has been shown that the reaction oscillations on the surface for a certain gas ratio and temperature happen at different pressures for the same gas ratio of CO and O2, but the reaction rate is different.

Reproducibility When someone finds a way to smoothen the surface to a known level before the start of each experiment, this will help to re-produce the measurements of the reaction oscillations. The reflectivity of the sample is changed to the combination of two different processes, that is roughening and oxidation, therefore we cannot reproduce our results unless we stabilize both processes. Until then we are not able to find a

(46)

tween the oscillation were different. This is due to the surface memory of the sample.

6.1 Outlook

For future experiments we need to clean the surface of the sample before experimenting and we need to check if the setup does not contaminate the surface with other chemicals. We also need to sample images faster, that means faster than once a second, to analyze the reaction oscillations further in terms of amplitude and examine the surface at atomic scale be-fore and after the experiments. Finally we conclude that if you want to explain the behaviour of reaction oscillations one should definitely try to use lights with different wavelengths and try to design experiments that are reproducible. Which means that the search for the relationship be-tween structure and catalytic activity of the palladium (100) sample is not ended, we are just at the beginning.

6.2 Acknowledgements

I would like to express my special thanks to my supervisor. This research could not have been done without the supervision by and fruitful dis-cussion with dr. Irene Groot. The help of Mirthe and Emiel from the fine mechanical department, the help from Mahesh and Sabine for the cleaning, sputtering and analysis of the sample and the help from Gert-jan from Leiden Probe Microscopy B.V. was deeply appreciated. Likewise I would like to thank Vladimir for his contribution to the research from the chemical and engineering point of view, and also for his assistance during measurements. The help from people from the research group wherein this project was taking place and also the contribution from the people who were present during my research presentation among which was Prof.dr.ir. Tjerk Oosterkamp was very helpful. Last I would like to thank my parents who encouraged me to finalize my thesis.

(47)

Bibliography

[1] M. Boronat and P. Concepcion, Combined theoretical and spectroscopic mechanistic studies for improving activity and selectivity in heterogeneous catalysis., Catalysis Today 285, Ahead of Print (2016).

[2] B. L. M. Hendriksen, S. C. Bobaru, and J. W. M. Frenken, Looking at heterogeneous catalysis at atmospheric pressure using tunnel vision, Topics in Catalysis 36, 43 (2005).

[3] J. W. M. Frenken and I. M. N. Groot, Live Observations of Catalysts Using High-Pressure Scanning Probe Microscopy, in Operando Research in Heterogeneous Catalysis, pages 1–30, Springer, 2017.

[4] B. L. M. Hendriksen, M. D. Ackermann, R. van Rijn, D. Stoltz, I. Popa, O. Balmes, A. Resta, D. Wermeille, R. Felici, S. Ferrer, and J. W. M. Frenken, The role of steps in surface catalysis and reaction oscillations., Nature chemistry 2, 730 (2010).

[5] Bzovska and Mryglod, Chemical oscillations in catalytic CO oxidation reaction, Condensed Matter Physics 13, 34801 (2010).

[6] M. Slinko, Oscilating reactions in heterogeneous catalysis: What new in-formation can be obtained about reaction mechanisms?, Catalysis Today

154, 33 (2010).

[7] M. W. https : //www.webelements.com/palladium/crystal structure.html, Webelements, Palladium: crystal structure, 2017.

[8] S. Simon, The Oxford Solid State Basics, Oxford University Press, 2013. [9] K. Oura, V. Lifshits, A. Saranin, A. Zotov, and M. Katayama, Surface

(48)

[12] B. L. M. Hendriksen, S. C. Bobaru, and J. W. M. Frenken, Oscillatory CO oxidation on Pd(1 0 0) studied with in situ scanning tunneling mi-croscopy, Surface Science 552, 229 (2004).

[13] C. T. Herbschleb, ReactorSTM: Imaging Catalysts under Realistic Condi-tions, PhD thesis, Leiden University, 2011.

[14] G. Zheng and E. I. Altman, The reactivity of surface oxygen phases on Pd(100) toward reduction by CO, Journal of Physical Chemistry B 106, 1048 (2002).

[15] W. G. Onderwaater, A. Taranovskyy, G. C. V. Baarle, W. M. Frenken, and I. M. N. Groot, In Situ Optical Reflectance Difference Observations of CO Oxidation over Pd(100), The Journal of Physical Chemistry C 121, 11407 (2017).

[16] E. Hecht, Optics, Fourth Edition, Pearson Education, 2002.

[17] S. C. Bobaru, High-Pressure STM Studies of Oxidation Catalysis, PhD thesis, Leiden University, 2006.

[18] LPM, Leiden Probe Microscopy B.V. http :

//www.leidenprobemicroscopy.com/, 2017.

[19] W. G. Onderwaater, A. Taranovskyy, G. M. Bremmer, G. C. van Baarle, J. W. M. Frenken, and I. M. N. Groot, From dull to shiny: A novel setup for reflectance difference analysis under catalytic conditions, Review of Scientific Instruments 88, 023704 (2017).

[20] G. Bremmer, In-situ optical investigation of Pd single crystal surfaces dur-ing catalytic CO oxidation reactions, Master’s thesis, Leiden University, the Netherlands, 2013.

[21] Thorlabs, Inc. http : //www.thorlabs.com/, 2017.

(49)

BIBLIOGRAPHY 43

[23] S. Roobol, W. G. Onderwaater, M. van Spronsen, F. Carla, O. Balmes, V. Navarro, S. Vendelbo, P. Kooyman, C. Elkjaer, S. Helveg, R. Felici, W. M. Frenken, and I. M. N. Groot, Supporting Information accompany-ing the manuscript: In situ studies of NO reduction by H2 over Pt usaccompany-ing surface X-ray diffraction and transmission electron microscopy, The Jour-nal of Physical Chemistry C 19, 1 (2017).

[24] I. Yuranov, M. Slin, E. Kurkina, E. D. Tolstunova, and A. Renken, Os-cillatory behavior during CO oxidation over Pd supported on glass fibers : experimental study and mathematical modeling, 55 (2000).

Investigating reaction oscillations during CO oxidation on Pd(100) using optical microscopy

Investigating reaction oscillations

during CO oxidation on Pd

(

100

)

using optical microscopy

Investigating reaction oscillations

during CO oxidation on Pd

(

100

)

using optical microscopy

G. Westra

Abstract

Contents

Chapter

1

Introduction

Chapter

2

Theory

2.1

Palladium (100) crystal

2.2

Reaction mechanisms

2.3

Stages of oxidation

2.4

Reaction oscillations

Chapter

3

Set-up

3.2

Thermocouple and heating

∑

3.3

Gas supply system and mass spectrometer

Chapter

4

Sample brightness

4.1

About the sample

4.2

Expectations

4.3

Experiment A: Smoothening and roughening

4.4

Experiment B: Oxidation and reduction

Chapter

5

Reaction oscillations

5.1

Conditions

5.2

Experiment C: Reaction oscillations for 2 hours

5.4

Experiment D: Reaction oscillations for

approx-imately 22 hours

5.5

Grid analysis

Chapter

6

Conclusions

6.1

Outlook

6.2

Acknowledgements

Bibliography

₁₀₀

₁₀₀