Downloaded from orbit.dtu.dk on: dic.

18, 2023

New Catalysts for Methanol Synthesis

Jakobsen, Helene Hagemann

Publication date:
2023

Document Version
Publisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA):
Jakobsen, H. H. (2023). New Catalysts for Methanol Synthesis. Department of Physics, Technical University of
Denmark.

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
 You may not further distribute the material or use it for any profit-making activity or commercial gain
 You may freely distribute the URL identifying the publication in the public portal

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
 New Catalysts for Methanol Synthesis

Helene Hagemann Jakobsen

Ph.D. dissertation

Supervisors:
Professor Ib Chorkendorff, DTU Physics
Professor Jakob Doganli­Kibsgaard, DTU Physics
Associate Professor Christian Danvad Damsgaard, DTU Physics

Section of Surface Physics and Catalysis

Department of Physics
Technical University of Denmark
March, 2023
Preface
This thesis is submitted as a partial fulfilment of the degree of PhD in Physics at the
Technical University of Denmark (DTU). The work was carried out at the Section of Surface
Physics and Catalysis (DTU SurfCat) at the Department of Physics (DTU Physics). It has
been supervised by Professor Ib Chorkendorff and co­supervised by Professor Jakob
Doganli­Kibsgaard and Associate Professor Christian Danvad Damsgaard. The work has
been carried out between the 1st of November 2019 and the 28th of February 2023 and
was supported by a research grant (V­SUSTAIN, grant 9455) from the Villum Foundation.

ii
Abstract
In the context of a fossil­fuel free and CO2 ­neutral society, where we inevitably will still
depend on liquid fuels, methanol is considered an important chemical. Methanol is sug­
gested as the key chemical in the ‘methanol economy’ and so it has the potential to aid in
the reduction of greenhouse gas emissions. If CO2 hydrogenation is to be applied within
a sustainable infrastructure, either new catalysts or the optimization of the currently used
one is required. This thesis presents research within the development and characteriza­
tion of Cu­based catalysts for the synthesis of methanol from CO2 and H2 . Furthermore,
it describes the findings from the study of a model system that can potentially aid in the
understanding of the Cu/ZnO/Al2 O3 catalyst.
For methanol synthesis from the hydrogenation of CO2 , a series of new promoters for a
Cu­based catalyst were investigated in close collaboration with theory. The study focused
on the Cu­Ge system and investigated methods for synthesizing supported Cu(1­x) Gex
nanoparticles with very small amounts of Ge promoter. The investigations lead to a syn­
thesis that produced metallic Cu(1­x) Gex nanoparticles with variations in the Ge molar
fraction and Cu particle sizes. Activity tests of SiO2 ­supported Cu(1­x) Gex nanoparticles
showed that while catalysts of high Ge­loading performed worse than those of low Ge­
loading, the series of catalysts within the latter category performed similarly. Character­
ization of the Cu­Ge catalysts showed a large variance in Cu particle sizes across the
range of tested catalysts, which prompted the need to separate the effect of active sur­
face area from that of promotion by Ge. The Cu surface­averaged mass activities indicate
that the Cu­Ge catalyst with the lowest fraction of Ge perform slightly better. However,
when compared to an industrial grade Cu/ZnO/Al2 O3 /MgO, none of the Cu­Ge catalysts
performed better both neither in terms of activity nor selectivity. The question of the state
of Ge remains since neither analysis of PXRD, PDF analysis of X­ray total scattering data,
nor (HR­)TEM imaging proved useful in determining the presence and state of Ge in the
most relevant Cu­Ge catalysts.
A CuZn/C model for the industrial­type methanol catalyst was studied using a combination
of PXRD, X­ray total scattering, STEM­EDX, and activity measurements. By combining
a weakly interacting carbon support of high surface area with the stepwise increase of
Zn/Cu, the relevant fraction of Zn could be determined: The highest mass activities were
obtained for carbon­supported catalysts with Zn/(Zn+Cu) ratios above 0.2. This was in
line with previous findings and hypothesized to be the gradual reduction of ZnO to Zn
in the surface of Cu. Characterization of the catalysts post activity­testing illustrates the
important role of ZnO as a structural promoter: Smaller Cu nanoparticles are stabilized
with larger Zn/Cu fractions.
Work on bringing back a combined HPC­UHV chamber was done, and now a foundation
has been laid for the future characterization of the chemical speciation of the relevant
fraction of any promoter. Whether it be Zn in the CuZn/C model, or Ge in the new Cu­
Ge­based catalysts for CO2 hydrogenation, the combination of quasi in­situ XPS/ISS and
high­pressure measurements will prove very useful in the future.

iii
Resumé
I et samfund, som både er CO2 ­neutralt såvel som fri for fossilt brændstof, men hvor vi
stadig vil have brug for flydende brændstof, betragtes methanol som et vigtigt kemikalie.
Methanol ses som et essentielt kemikalie i den såkaldte ’methanoløkonomi’ med et poten­
tiale til at hjælpe med at reducere udledning af drivhusgasser. Hvis hydrogenering af CO2
skal anvendes i en bæredygtig infrastruktur, er der enten brug for nye katalysatorer eller
for at optimere den alment anvendte katalysator. Denne afhandling præsenterer forskning
inden for udvikling og karakterisering af Cu­baserede katalysatorer, med henblik på syn­
tese af methanol fra CO2 og H2 . Den beskriver desuden forskningsresultaterne fra studiet
af et modelsystem, der kan hjælpe til en bedre forståelse af Cu/ZnO/Al2 O3 ­katalysatoren.
En serie af nye promotere til Cu­baserede katalysatorer, der har til formål at blive brugt
til hydrogeneringen af CO2 , blev undersøgt i tæt samarbejde med teoretikere. Studiet
fokuserede på Cu­Ge­systemet og undersøgte metoder til at syntetisere supporterede
Cu(1­x) Gex ­nanopartikler, der indeholder meget små mængder Ge­promoter. Studiet førte
til en syntese, der frembringer metalliske Cu(1­x) Gex ­nanopartikler med variationer i molar­
fraktionen af Ge og partikelstørrelsen af Cu. Ved test af aktiviteten af de SiO2 ­supporterede
Cu(1­x) Gex ­nanopartikler for deres aktivitet, sås det, at katalysatorer med høj vægtkon­
centration af Ge præsterede ringere end dem med lav vægtkoncentration. Samtidig sås
det, at katalysatorerne i den sidstnævnte kategori præsterede nogenlunde identisk. Ved
karakterisering af hele rækken af de testede Cu­Ge­katalysatorer sås det, at der var en
stor varians i størrelsen af Cu partikler. Dette medførte et behov for at adskille effekten af
det aktive overfladeareal fra effekten af Ge­promovering. Ved at normalisere aktiviteten af
katalysatorerne med et mål for overfladearealet af Cu sås det, at Cu­Ge­katalysatorerne
med den laveste vægtkoncentration præsterer lidt bedre. Dog præsterer ingen af Cu­Ge
katalysatorerne lige så godt som den industrielle Cu/ZnO/Al2 O3 ­katalysator ­ hverken når
det kommer til aktivitet eller selektivitet. Spørgsmålet vedrørende den kemiske tilstand af
Ge står stadig tilbage, eftersom hverken analyse af PXRD, PDF­analyse af røntgentotal­
spredning, eller (HR­)TEM viste sig at være behjælpelig med at fastslå tilstedeværelsen
eller tilstanden af Ge i de mest relevante Cu­Ge­katalysatorer.
En CuZn/C­model er blevet undersøgt med en kombination af PXRD, røntgentotalspred­
ning, STEM­EDX, og aktivitetsmålinger. Kombinationen af en højoverfladeareal­support
af kul med en trinvis forøgelse af Zn/Cu­forholdet gjorde det muligt at observere den rel­
evante fraktion af Zn: De højeste aktiviteter sås for de kul­supporterede katalysatorer,
der havde et Zn/(Zn+Cu)­forhold over 0.2. Dette var på linje med tidligere fund og for­
modes at skyldes den gradvise reduktion af ZnO til Zn i overfladen af Cu. Karakterisering
af katalysatorerne, efter de var blevet testet, illustrerede den vigtige rolle ZnO har som
strukturel promoter: Mindre Cu­partikler stabiliseres ved at have et højere Zn/Cu­forhold.
Der er blevet lagt et fundament for at karakterisere den kemiske tilstand af den relevante
fraktion af hvilken som helst slags promoter vha. et kombineret HPC­UHV­setup. Hvorvidt
det er Zn i CuZn/C­modellen, eller om det er Ge i den nye Cu­Ge­katalysator, der skal
bruges til hydrogenering af CO2 , så vil kombinationen af quasi in­situ XPS/ISS og ak­
tivitetsmålinger blive særdeles brugbar i fremtiden.

iv
Acknowledgements
This thesis is a product of a little over 3 years of working as a PhD student at SurfCat, DTU.
It has been the hardest years, but also the most educational, the most transformative, and
the most rewarding. There is so much knowledge at SurfCat, just waiting for you to go
and grab it. Thank you to the holders of that knowledge: All of my friends and colleagues
at SurfCat and DTU Physics. You have made the years at SurfCat truly memorable. I
am proud of the way we have each other’s backs in science as well as in life. A special
shout out is given to Olivia (for making my dissertation look pretty), Rikke, Julius, Asger,
Krabbe, Celia and Sarah, and all the inspiring women in STEM I have met over the years.
It has been a privilege to have Professor Ib Chorkendorff as a main supervisor. I thank
him for taking in a theoretical chemist and for giving me the chance to work with exper­
imental catalysis. His ability to both see the grander scheme of things and to focus on
the nitty­gritty details have been of great value to me. Secondly, I owe gratitude to my
co­supervisors, Professor Jakob Doganli­Kibsgaard and Associate Professor Christian D.
Damsgaard, the latter being a valuable help on a daily basis, whether it be in the labora­
tories or by a blackboard.
To the researchers I have collaborated with: It’s been a true privilege to work with and
learn from you. Jette K. Mathiesen opened my eyes to the world of total scattering and
has been a great collaborator and an even greater support during our joined project. I
want to thank Mads Lützen, a brilliant and kind researcher with DTU Nanolab, for collabo­
rating with me on the CuZn project. A thank you is owed to Stefan Kei Akazawa from DTU
VISION and Filippo Romeggio from DTU SurfCat for their hard work on the CuGe project.
It has been a privilege to work with the people of Jens Nørskov’s group at CatTheory, es­
pecially Ang Cao, who has been the main contributor to the theoretical basis behind the
Cu­X study. Jakob Munkholt and Jonas Abitz Boysen from Chemical Engineering DTU
are thanked for inviting me into their lab with open arms and for always being open for
a scientific discussion. Jens Sehested from Topsøe is acknowledged for his suggestion
to explore carbon­supported CuZn catalysts and to give the introduction to the synthesis
and testing of these.
A special thanks is dedicated to the backbone of SurfCat, namely the team of floorman­
agers. Patrick Strøm­Hansen and Brian Knudsen first introduced me to the experimental
setup and took their time getting to know it alongside me. Later, Jakob Ejler joined to
be of great help when installing new stepper motors and to laugh at and with me when I
chose to save the electronics by spilling cooling water all over myself.
I owe my partner in crime and highly cherished colleague, Thomas Smitshuysen, a whole
lot. Had it not been for his endless knowledge (and opinions) about nearly everything, the
last few years would have been a very different experience. I thank him for being such an
excellent teacher (and for enjoying teaching), for always encouraging me and for being
genuinely excited about my research. I am looking forward to the future, for I not only
gained a great colleague in Thomas, I also found a great friend.
Thank you to my family for being the kind of support network every researcher should
have. It means a lot to me that you have always made an effort of understanding what
I work with so that you could be able to explain it to colleagues and friends. Thank you,
mommy, for telling me I can do more than I think.
To my queer love, Esben. You have been a better support than I could ever hope for.
Thank you for being there from the very beginning despite a pandemic, the endless chal­
lenges we’ve had to face, and the inconsistent work schedule a PhD offers. Most impor­
tantly, thank you for being my home: A place, where I can recharge my batteries and
where I can always find warmth and love. Thank you for choosing us.

v
Abbreviations
BET: Brunauer­Emmett­Teller, in relation to surface area.
CTY: Copper time­yield.
CZA: The industrial­type catalyst used as a reference in this study, Cu/ZnO/Al2 O3 /MgO.
in­situ: ’while it happens’.
ex­situ: Used for characterisation performed on a sample before or after it has been ex­
posed to an environment, also meaning it has been exposed to air.
EDX: Energy dispersive X­ray (spectroscopy), also referred to as EDS.
FCC: Face­centered cubic crystal structure.
FFT: Fast­Fourier transform.
FID: Flame ionization detector.
GC: Gas chromatography.
GHSV: Gas hourly space velocity. The speed of the gas passing over a catalyst normal­
ized by the volume of the catalyst bed. Usually in the units of h−1 .
GUI: Graphical user interface.
HAADF: High­angle annular dark field.
HCP: Hexagonal close­packed crystal structure.
HPC: High­pressure cell.
HSA: Hemispherical energy analyzer.
H2 ­TPD: Temperature­programmed desorption of H2 .
ICP­OES: Inductively coupled plasma optical emission spectroscopy.
ISS: Ion scattering spectroscopy. Also referred to as low­energy ion spectroscopy (LEIS).
IWI: Incipient wetness impregnation. Referred to as either conventional (1­step) or se­
quential (2­step).
MeOH: Methanol.
MFC: Mass­flow controller.
N2 O­RFC: Reactive frontal chromatography by nitrous oxide.
NP: Nanoparticle. Referring to a particle that is between 1 −100 nm in diameter.
OQMD: Open quantum materials database.
QMS: Quadropole mass spectrometry.
RA­PDF: Rapid­acquisition PDF.
rWGS: Reverse water­gas­shift reaction.
SEM: Scanning electron microscopy.
SMSI: Strong metal–support interaction (SMSI) effect.
(S)TEM: (Scanning) transmission electron microscopy.
Syngas: H2 ­containing gas composition produced from steam reforming of natural gas.
The name is used for synthesis gas in different industries, but in the methanol industry
and this thesis it refers to a gaseous mixture of CO2 , CO and H2 .
TrEG: Triethylene glycol.
UHV: Ultra­high vacuum, the regime characterized by a pressure lower than 1.0 × 10−9 mbar.
WHSV: Weight hourly space velocity. The speed of the gas passing over a catalyst nor­
malized by the weight of the catalyst or the weight of what is assumed to be the active
material. Usually in the units of N·cm3 ·hour−1 ·g−1 .
XAS: X­ray absorption spectroscopy.
XPS: X­ray photoemission spectroscopy.
X­ray TS PDF: X­ray total scattering (TS) with pair­distribution function (PDF) analysis.
(P)XRD: (Powder) X­ray diffraction.

vi
 Contents

Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Resumé . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

1 Introduction 1
1.1 The role of methanol in a CO2 ­neutral future . . . . . . . . . . . . . . . . . . 2
1.2 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Applied Heterogeneous Catalysis 6

2.1 Thermal Methanol Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 The Industrial­type Cu/ZnO Catalyst . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 The use of model systems to elucidate the active site . . . . . . . . 11

3 Experimental Methods 14
3.1 Catalyst Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.1 Incipient wetness impregnation . . . . . . . . . . . . . . . . . . . . . 14
3.1.2 Mixed suspension approach . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Catalyst Pre­treatment and Activity Measurements . . . . . . . . . . . . . . 16
3.2.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3 X­ray Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3.1 PXRD data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3.2 PXRD apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 X­ray Total Scattering with PDF Analysis . . . . . . . . . . . . . . . . . . . . 23
3.5 Microscopy: (S)TEM and SEM . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5.1 Scanning electron microscopy . . . . . . . . . . . . . . . . . . . . . 26
3.5.2 Transmission electron microscopy . . . . . . . . . . . . . . . . . . . 27
3.6 X­ray Photoelectron Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . 28
3.6.1 Setup and data analysis . . . . . . . . . . . . . . . . . . . . . . . . . 30

4 New Catalysts for CO2 Hydrogenation 31

4.1 Proposing a New Cu­X system . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Getting to Know the Cu­Ge System . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.1 Optimization of synthesis route for an active Cu­Ge catalyst . . . . . 36
4.2.2 Activity of a CuO/GeO2 /SiO2 catalyst . . . . . . . . . . . . . . . . . . 41
4.2.3 In­situ XRD study of CuO/GeO2 /SiO2 . . . . . . . . . . . . . . . . . 42
4.3 Study of Ge Promotion in Cu­Ge/SiO2 . . . . . . . . . . . . . . . . . . . . . 45
4.3.1 Synthesis of Cu(1­x) Gex catalysts . . . . . . . . . . . . . . . . . . . . 47
4.3.2 CO2 hydrogenation over Cu­based catalysts: Setting the stage . . . 52
4.3.3 Effect of size and composition in the Cu­Ge System . . . . . . . . . 54
4.3.4 Discussion on the single­atom surface alloy Cu­Ge system . . . . . 66
4.4 Preliminary Studies of Possible Promoters for Cu­X/C . . . . . . . . . . . . 72
4.5 Conclusion on New Catalysts for CO2 Hydrogenation . . . . . . . . . . . . 73

5 Carbon­supported Cu/Zn as a Model System 75

5.1 Effect of Zn Content on Syngas Conversion . . . . . . . . . . . . . . . . . . 78

vii
Contents

5.2 A Question of Surface Area or Zn Promotion . . . . . . . . . . . . . . . . . 84

5.2.1 Bulk­averaged information: PXRD analysis . . . . . . . . . . . . . . 84
5.2.2 Local information: HAADF­STEM coupled with elemental analysis . 86
5.2.3 Bulk­averaged, but sensitive to local structure: X­ray TS/PDF . . . . 89
5.2.4 Separating the effects of surface area and Zn promotion . . . . . . . 93
5.3 Modifications to the Synthesis and Testing of Cu(Zn)/C . . . . . . . . . . . . 94
5.4 Outlook and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.5 Conclusion on Carbon­supported Cu/Zn as a Model System . . . . . . . . . 99

6 Optimizing a UHV Chamber for Surface­Sensitive Studies 101

6.1 Setting the stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.2 Back to basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.3 Outlook: Future work and possibilities . . . . . . . . . . . . . . . . . . . . . 108

7 Conclusion 110
7.1 Carbon­supported Cu(Zn) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.2 New Methanol Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Appendix A Supporting data for X­ray methods 113

A.1 PXRD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

Appendix B Supporting data for Cu and CuGe catalysts 114

B.1 Supporting data for in­situ XRD study of CuO/GeO2 /SiO2 . . . . . . . . . . 114
B.2 Early synthesis studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
B.3 XRD spectra of unsupported Cu and CuGe nanoparticles . . . . . . . . . . 119
B.4 PXRD of Cu1−x Gex /SiO2 ­ before and after activity testing . . . . . . . . . . 122
B.5 XPS of selected Cu1−x Gex /SiO2 . . . . . . . . . . . . . . . . . . . . . . . . 124
B.6 Supporting data for activity of Cu and CuGe­based catalysts . . . . . . . . 126
B.7 PXRD of Cu­based catalysts on different supports . . . . . . . . . . . . . . 131
B.8 PDF refinement of selected Cu(1­x) Gex /SiO2 catalysts . . . . . . . . . . . . 132
B.9 SEM­EDX and (HR­)TEM of (1)­Cu0.85 Ge0.15 /SiO2 . . . . . . . . . . . . . . 134
B.10 SEM images for early synthesis studies . . . . . . . . . . . . . . . . . . . . 136

Appendix C Supporting data for the Cu/ZnO/Al2 O3 /MgO catalyst 137

Appendix D Supporting data for CuX catalysts 139

Appendix E Supporting information for experimental setup 141

E.1 Data treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
E.2 Setup optimizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

Appendix F Supporting data for Cu(Zn)/C catalysts 148

F.1 Activity data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
F.2 Representative microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
F.3 In­situ XRD study of Cu/C . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
F.4 Example of total scattering data: CuZn­0.05/C . . . . . . . . . . . . . . . . 155
F.5 PDF refinement of selected Cu(Zn)/C catalysts . . . . . . . . . . . . . . . . 156
F.6 Rietveld refinement of selected Cu(Zn)/C catalysts . . . . . . . . . . . . . . 159
F.7 PXRD of post­mortem Cu(Zn)/C catalysts . . . . . . . . . . . . . . . . . . . 160

List of Figures 161

viii
 Contents

List of Tables 171

References 172

ix
Chapter 1

Introduction
Our way of life as we know it in the Anthropocene is coming to an end. We cannot sustain
the energy consumption of the ever­growing human population the way we have been
used to since the technological revolution. Not only are we running out of resources, and
faster each year as we reach the ’Earth Overshoot Day’ earlier these days than we did
in the 1970’s, but we are also running out of viable places to live due to climate change.
Changes to the global climate causes changes in weather extremes, changes in precip­
itation patterns, and agricultural and ecological droughts, and scientists are observing
changes in the Earth’s climate in every region and across the whole climate system (1).
In their sixth assessment report of 2021, the Intergovernmental Panel on Climate Change
(IPCC) conclude that “it is unequivocal that human influence has warmed the atmosphere,
ocean and land”, as illustrated in the graphic of Figure 1.1 that show the changes in global
surface temperatures relative to 1850­1900 (1, 2). The report shows that emissions of
greenhouse gases (GHGs, abbrev.) from human activities are responsible for 1.1 ◦ C of
warming since 1850­1900, and that the global temperature will reach or exceed 1.5 ◦ C of
warming over the next 20 years.

Figure 1.1: Changes in global surface temperature relative to 1850–1900: History and
causes of warming. Reprinted with permission. Copyright IPCC (2).

1
Chapter 1. Introduction

As of 2022, we were at a 1.06 ◦ C warming (3), and the 1.5 ◦ C­limit will be reached un­
less there are immediate, rapid and large­scale reductions in the emissions of GHGs.
The IPCC assessment report finds clear evidence that CO2 is the main driver of cli­
mate change. The GHG that is most important for the natural greenhouse effect, which
makes the Earth habitable, is now found in atmospheric concentrations higher than ever
recorded. As of writing this thesis, the monthly average of January 2023 was 419.47 ppm,
more than 1 ppm than that of January 2022, and part of a seemingly ever­increasing av­
erage since Charles D. Keeling began his recordings in 1958 (4). The concentration of
atmospheric CO2 is increasing mostly due to the burning of fossil fuels. Returning the
carbon that plants pulled out of the atmosphere many million years ago without returning
it consequently pushes the carbon pool in the wrong direction ­ today, plants and oceans
can only take up a part of the CO2 we are releasing, and therefore much of it is bound
to stay in the atmosphere until the carbon pool balances out. Since the middle of the
20th century, annual emissions of CO2 from burning fossil fuels have increased from an
average of 11 billion tons/year in the 1960s to 35 billion tons/year in the 2010s (5). In the
future, we cannot depend on fossil fuels. Both due to the damage the release of CO2 is
doing to the global climate, but also due to the limited supply. But, as Nobel Laureate
Richard Smalley asks, “what will we do when there is no longer enough oil and gas?” (6).
The answer is to find a source of energy, just as abundant as oil, but more green, that
can sustain the growing population without compromising our way of living.

1.1 The role of methanol in a CO2 ­neutral future

The biggest challenge is to find an energy source that is just as reliable and versatile as
oil. Alternative, renewable energy sources such as geothermal energy, hydropower, so­
lar energy, ocean energy, and wind power, have the greatest potential but cannot always
consistently produce energy at all hours of the day. Being able to store excess energy
produced during peak cycles is key to using the full potential of the renewable energy
sources. What first comes to mind is naturally batteries. But comparing the volumetric
and gravimetric energy density of batteries to a group of materials and technologies, as
done in Figure 1.2, they are not nearly as effective a storage option as for example hy­
drogen. H2 can be produced by the electrolysis of water, a process that in itself does not
produce CO2 , but it is difficult to handle and store. Although H2 in both gas and liquid form
has a favourable gravimetric energy density, the high pressure used for storage, the use
of special materials to minimize the diffusion and leakage, and the extensive safety mea­
surements that must be taken, are all drawbacks. An energy carrier that is comparable
to conventional diesel and gasoline is methanol. As argued by Nobel Laureate George
Olah, methanol provides an efficient means to store energy and can be used as a fuel
or be blended with gasoline (7). He suggested the idea of a methanol economy, where
CO2 is recycled into methanol using green hydrogen (7–10). In contrast to H2 , methanol
is easier to handle and store since it is a liquid with a boiling point of approximately 65 ◦ C.

2
 1.1. The role of methanol in a CO2 ­neutral future

Figure 1.2: Volumetric (in MJ/L) and gravimetric (in MJ/kg) energy density of selected
energy carriers. Data points are taken with permission from (11). Copyright 2014 Elsevier.

Methanol does not only have the potential to be used as a fuel alone, it is also an impor­
tant raw material in the chemical industry. Since it is the simplest alcohol, it is a chemical
building block for hundreds of everyday products, such as plastics and paints. As illus­
trated in Figure 1.3, the demand for primary chemicals (methanol, ammonia and high­
value chemicals) has increased in recent years, with the growth in methanol production
increasing the most of all. The main end uses for methanol are formaldehyde, fuel ap­
plications and other high­value chemicals (12), and the demand and therefore growth is
set to increase further into a ’net­zero scenario’. According to the International Energy
Agency, the chemical sector is not on track for a 2050 scenario of net zero emissions and
in the agency’s view “both the private and public sectors will need to achieve technological
innovation, efficiency gains and higher recycling rates” (13). Since the chemical sector is
the largest industrial energy consumer and the third largest industry subsector in terms of
direct CO2 emissions, there is a great incentive to optimize. Recent innovations related
to the production of methanol include small­ and large scale emission­to­liquids plants in
Iceland (the ”George Olah Carbon Dioxide to Renewable Methanol Plant”), Norway (next
to another chemical plant in Finnfjord), and in China (next to a coke oven gas production).

For quite a long time, methanol has been produced at large, industrial­scale plants using
mainly gas and coal as feedstock. The high pressures and temperatures required for a
high enough conversion are reason for a large energy input, even though the methanol
synthesis is carried out using a highly active and selective catalyst. There is no doubt that
the conventional way of producing methanol is not sustainable in a CO2 ­neutral future.
Future efforts in renewable methanol production could go two ways: We could use the
current infrastructure, or we could fully embrace the methanol economy as suggested by
George Olah.

3
Chapter 1. Introduction

Figure 1.3: Growth in primary chemical production in the ’Net Zero Scenario’ over
the year 2000­2030 as reported by the International Energy Agency (IEA). Coloured
dashed lines past the year 2021 represent the projected growth. Data set is from (13),
https://www.iea.org/reports/chemicals. License: CC BY 4.0.

The first way requires a change to unconventional feedstocks, such as agricultural waste
or forestry residues, which would subsequently have to be treated and turned into syn­
thesis gas to be used in the current infrastructure. However, a change of feedstock would
likely mean a change in impurities. And a change in impurities means either having to
change the purification stage of the production line, or to optimize the catalyst to han­
dle new poisons. Besides the handling of new poisons, the general optimization of the
catalyst in terms of long­term stability and selectivity is also high on the list. Any optimiza­
tions of the currently used industrial­type methanol catalyst requires understanding of the
catalyst material system to the fullest.

The second route to a renewable methanol production requires catalyst systems that are
optimized for the specific reaction that is direct hydrogenation of CO2 . To lower the envi­
ronmental footprint of the methanol industry, the combination of CO2 from waste streams
with H2 from electrolysis of water is preferred, and even better is to move the production
from large­scale, centralized plants to small, de­centralized plants that operate at milder
conditions (lower temperatures and pressures). In the end, this would mean optimizing
the current methanol catalyst to operate under the conditions than what it is normally used
for, or to formulate a completely new catalytic system.

4
 1.2. Thesis outline

1.2 Thesis outline

The work presented in this thesis is the final written product of my PhD project. The
thesis is part focused on the development of new catalysts for CO2 hydrogenation, and
part focused on the understanding of the Cu/ZnO/Al2 O3 catalyst using a model system.
In the beginning of Chapters 4 and 5, the reader will find a more in­depth introduction to
the subject. Some of the subsections include research conducted by someone else, but
in close collaboration with the author of this thesis. It is always noted when research or
data treatment has been carried out by collaborators.

Chapter 1. An introductory chapter that serves as a reminder of why we are doing this.

Chapter 2. A chapter that introduces the relevant concepts of catalysis, with thermal
methanol catalysis as a case, and describes the literature relevant for the subsequent
chapters.

Chapter 3. A chapter to present the methods and apparatus used in this PhD project.
Procedures written in this chapter are general for the subsequent chapters, and any de­
viations from these procedures are presented when relevant.

Chapter 4. Embracing the methanol economy: New catalysts for CO2 hydrogenation.

Chapter 5. Understanding and optimizing the industrial­type catalyst: A model system

for the Cu/ZnO/Al2 O3 methanol catalyst.

Chapter 6. An outlook­style chapter describing the efforts done to optimize a combined

HPC­UHV setup that in the future can be used for surface­sensitive studies of the catalysts
discussed in the preceding chapters.

5
Chapter 2. Applied Heterogeneous Catalysis

Chapter 2

Applied Heterogeneous Catalysis

Before introducing thermal methanol catalysis, the basics of catalysis are shortly intro­
duced. In the case of an elementary reaction like the one shown in Figure 2.1 (i.e. a
reaction that occurs in a single step), the rate of reaction is defined as the disappearance
of the reactants (A and B) or appearance of products (C). Since we don’t have to consider
any intermediates in an elementary reaction, the rate of reaction is the difference between
the forward rate of reaction, r+ , and that of the backwards reaction, r− . The individual rate
expression include a temperature­dependant reaction rate constant, k+ and k− :

( )
− − −Ea
r = r − r = k [A] [B] − k [C]
+ + a b c
, k(T ) = ν · exp (2.1)
RT

Figure 2.1: A simple potential energy diagram of a heterogeneous catalytic reaction.

By IUPAC definition, “catalyst is a compound that can increase the speed at which reac­
tions occur, and lower the energy required for the reaction to take place” (14, 15). What a
catalyst effectively does is to lower the energy needed to overcome the potential energy
barrier Ea of Equation (2.1). It does so by introducing an alternative path for the reac­
tion. From Figure 2.1 it is obvious that the alternative path is not easier, as it consists of

6
 2.1. Thermal Methanol Synthesis

multiple steps. The energy barrier of the catalytic reaction is, however, lower than that
of the reaction without a catalyst. The role of the catalyst is to offer a surface on which
the reacting species can adsorb and the intermediate species be stabilized. According
to the Sabatier Principle, the interaction between the adsorbing species and the surface
must not be too strong, but nor too weak. If too weak an interaction, the catalyst has little
effect, and in the case of too strong an interaction the adsorbates will not desorb from the
catalyst surface.
What is important to mention is that a catalyst does not affect the equilibrium constant of
the overall reaction ­ that is inherent to the specific reaction. What it does is to change
the kinetics but not the thermodynamics of a reaction, which is why it is important to test
catalysts in the kinetic regime. Without going into too much detail with thermodynamics,
the temperature­dependant equilibrium constant for a reaction such as that shown in Fig­
ure 2.1 can both be written in terms of Gibbs free energy or with respect to equilibrium
concentrations. At equilibrium, the rates of the forward and the backwards reactions are
equal, and so the equilibrium constant can also be written in terms of those:

( )
k+ [C]ceq −∆G◦
K(T ) = − = = exp , ∆G◦ = ∆H ◦ − T · ∆S ◦ (2.2)
k [A]aeq · [B]beq kB T

The temperature­dependant equilibrium constant is important to introduce as it explains

how a certain reaction behaves according to temperature and pressure. It is also a helpful
tool in comparing the (theoretical) highest achievable output of a product of a certain
reaction when a catalyst is not involved.

2.1 Thermal Methanol Synthesis

The technology for methanol production is well­established and mature. The industrial
process of methanol synthesis was first described in the 1920’s by Patart and Lormand
(16, 17). It involves heating and pressurizing synthesis gas (syngas, abbrev.) consisting
of CO2 , CO and H2 , and the use of a catalyst, to produce methanol and water. Formation
of methanol via syngas proceeds via the reactions in Equations (2.3) to (2.5) (18):

CO + 2H2 ⇀
↽ CH3 OH ∆H ◦ = −89 kJ · mol−1 (2.3)

CO2 + 3H2 ⇀
↽ CH3 OH + H2 O ∆H ◦ = −48 kJ · mol−1 (2.4)

CO2 + H2 ⇀
↽ CO + H2 O ∆H ◦ = 41 kJ · mol−1 (2.5)

The reactions producing methanol are both exothermic, hence a low temperature is pre­
ferred to reach a higher theoretical equilibrium amount of methanol (given the definition of
Gibbs free energy in Equation (2.2)). When analyzing the expression for the equilibrium

7
Chapter 2. Applied Heterogeneous Catalysis

constants of Equations (2.3) and (2.4) we also find that the methanol concentration is pro­
portional to the pressure, meaning an elevated pressure will provide a higher equilibrium
concentration of methanol. Both of these trends are illustrated in Figure 2.2b, which il­
lustrates the theoretical concentration of methanol at equilibrium at 1 bar and 10 bar as a
function of reaction temperature.
While producing methanol from just CO/H2 or CO2 /H2 seems straight forward, formation of
methanol from the latter comes at the cost of simultaneously producing CO and more H2 O
via the so­called reverse water­gas­shift (rWGS) reaction (Equation (2.5)). To counter­act
the rWGS reaction, CO is added to the feed gas, thereby recycling water into H2 that will in
turn, according to the Le Chatelier’s principle, increase the equilibrium amount of CH3 OH
in Equation (2.4).

In Figure 2.2a, the impact of the rWGS reaction is investigated at 1 bar for a gas mixture
of 75 % H2 and 25 % CO2 by allowing or prohibiting the rWGS reaction (the ’H2 /CO2 incl.
rWGS’ and ’H2 /CO2 excl. rWGS’, respectively). The thermodynamic calculations have
been executed using HSC 7 (19).
At low temperatures, where the CO2 hydrogenation is favored, the methanol equilibrium
concentration of the two cases is close to each other. As the temperature increases, the
difference quickly becomes apparent. The low equilibrium concentration of the ’H2 /CO2
incl. rWGS’ profile perfectly illustrates how CO inhibits the equilibrium. Part of the syn­
thesis gas is allowed to participate in the endothermic rWGS reaction, producing CO and
H2 O. The produced water will additionally push the equilibrium of Equation (2.4) to the
reactant­side, further decreasing the equilibrium concentration of methanol. Comparing
the H2 /CO2 feeds with that of the synthesis gas (H2 /CO/CO2 ) feed graphically shows why
CO is used as an added reactant gas in the industrial application. Including CO in the feed
gas will dramatically improve the methanol equilibrium by suppressing the rWGS reaction.
At 1 bar total pressure, including CO as a reactant will increase the allowed amount of
methanol from below 1 % (’H2 /CO2 incl. rWGS’ line) to almost 3 % (’H2 /CO/CO2 ’ line).

8
 2.1. Thermal Methanol Synthesis

(a) (b)

Figure 2.2: Equilibrium methanol profiles. (a) Equilibrium concentration with a feed of
25/75 CO/H2 , 10/10/80 CO/CO2 /H2 and 25/75 CO2 /H2 . For the latter, the rWGS is in­
cluded or excluded by either considering CO as a product (’incl. rWGS’) or excluding it
(’excl. rWGS’). Sub­figure (b) shows the equilibrium concentration in a CO/CO2 /H2 feed
as a function of temperature and pressure. Simulated using HSC 7 (19).

As evident from Figure 2.2a, the thermodynamics favor methanol production from CO op­
posed to CO2 . The reason for the industry to use a mixture of CO and CO2 comes down to
the catalyst that is used for the industrial methanol production. The first industrially used
catalyst was that of a Zn/Cr2 O3 (20–22). This catalyst required fairly high temperatures
and pressures (300 bar and 300 −400 ◦ C) to obtain high conversion (23). In the 1960’s,
it was replaced by the more active Cu/ZnO/Al2 O3 catalyst which can be operated under
much milder conditions. This is still in use today, with the production of methanol oper­
ating at 200 −300 ◦ C, 50 −100 bar and using a feed gas of H2 /CO2 /CO (90:5:5) (18, 23).
Although conversion of CO/H2 is possible over the Cu/ZnO­based catalyst, the methanol
mass activity is higher in a CO2 ­rich feed. The data to support this is discussed in Ap­
pendix C. The reason for using both CO2 and CO is the outcome of a series of studies that
investigate the mechanism for methanol synthesis over Cu­based catalysts (24–32). Ex­
periments using isotopically marked CO and CO2 show that CO2 is the preferred carbon
source for methanol formation over the ternary Cu/ZnO/Al2 O3 catalyst at low conversion
(24, 25, 29). It seems that the governing consensus is that methanol is formed from CO2
over a formate­pathway, while CO reacts into CO2 via the rWGS reaction, thereby adding
more reactant to the carbon source pool (33–35).

9
Chapter 2. Applied Heterogeneous Catalysis

2.2 The Industrial­type Cu/ZnO Catalyst

The recent review­type study by Schwiderowski et al. (36) and the catalyst­overview by
reference (37) show that although there is progress in the development of alternative
catalysts for the CO2 hydrogenation reaction, none of the reported catalysts can compete
with the well­established Cu/ZnO/Al2 O3 catalyst. The superiority of the Cu/ZnO/Al2 O3
catalyst is owed to the years of extensive research and optimizations of this chemical
system. Despite the vast amount of literature made on the methanol synthesis catalyst,
it is still a dynamic research area with many different directions and contradicting views.
The knowledge gained from over the years is briefly outlined in this section. To begin, a
description of the three components of the Cu/ZnO/Al2 O3 catalyst and how they are each
found to contribute to the catalytic methanol synthesis:

• Metallic Cu is regarded as the catalytically active component of the catalyst.

Early studies of Cu single crystals show that Cu synthesizes methanol from CO2 and
H2 (27, 38–42), and more recent studies attribute the catalytic activity to defective
Cu step sites and Cu lattice strain (28).

• Al2 O3 acts as a structural promoter by facilitating a high Cu dispersion (43).

• ZnO works as both a structural and electronic promoter as found by references

(27, 29, 32, 41, 44–54). The early studies by Nakamura and Fujitani report an im­
provement in methanol activity when adding Zn to Cu single crystals and polycrys­
talline Cu (27, 41, 45, 47, 55), and this improvement has been reported in multiple
studies since (32, 48, 52, 53). Similarly to the non­reducible oxide Al2 O3 , ZnO ad­
ditionally acts as a structural promoter by increasing Cu dispersion.

The last point of the promotional role that Zn has on the methanol activity, referred to as the
Cu­ZnO synergy effect, is the subject that has been most intensely studied (and debated)
over the years. And with good reason. If we wish to improve the stability and selectivity
or increase the activity of the industrial methanol catalyst, we need to understand it to its
fullest. The Cu­ZnO synergy is also a classical example of metal­promoter structure, why
understanding it fully will benefit similar systems and aid in the discovery of new catalytic
systems. Models for the Cu­ZnO synergy are at this point many, but they can be grouped
into the following categories: Changes to the morphology of Cu as a function of the gas
composition (56–60); partly reduced single Zn atoms in the surface of Cu (a Cu­Zn surface
alloy) (27–29, 45, 47, 50, 51, 54, 55, 61–64); the interface between Cu and ZnO (32, 53,
65–69). It is argued that some or all of these phenomena might coexist (depending on
the chosen reaction conditions and time on stream) (70, 71), but the critical issue is to
determine the relative contribution that each of the phenomena have and whether they
are interlinked or not (18).

10
 2.2. The Industrial­type Cu/ZnO Catalyst

2.2.1 The use of model systems to elucidate the active site

The following subsection will give an oversight of the observations and conclusions drawn
in literature about the Cu­ZnO synergy. It is focused on the use of model systems used
to elucidate the active site of the Cu/ZnO catalyst, specifically focusing on the studies of
model systems that eventually inspired the study of carbon­supported CuZn in the coming
Chapter 5.

The early models of the industrial­type Cu/ZnO­based catalyst include that of Nakamura
and Fujitani. They found that by increasing the Zn coverage on Cu(111) and polycrys­
talline Cu the turnover frequency of methanol increased significantly. The increase in
activity until a Zn coverage of 0.2 monolayers on Cu(111) was attributed to metallic Zn
substituting into the Cu surface, and it was shown that a change in the surface chemistry
occurred just before this fraction, namely at 0.15 monolayers where Zn was more readily
oxidized to ZnO (45, 47). The formation of a surface CuZn alloy at the optimal Zn cover­
age on Cu(111) was later indicated by the same group using a combination of UHV­STM
and LEES­AES (63). Although the conclusion on the state of Zn might differ from the
studies of the 90’s, recent studies reach the same optimal Zn coverage on Cu(111) and
Cu(100) single crystals (32, 68).
Moving on from single crystals, Holse et al. studied metallic CuZn (Cu:Zn ≈ 75:25)
nanoparticles prepared via cluster synthesis in a combined UHV­HPC setup (72). By
combining in­situ TEM with XPS they revealed important trends in the Cu/ZnO system:
Firstly, that the surface behaviour of the Cu/ZnO system is dynamic in reducing and oxidiz­
ing gas environments, with CuO reversibly encapsulating/decapsulating ZnO, supporting
the earlier findings Grunwaldt et al. found for a co­precipitated Cu/ZnO system (58); sec­
ondly, that metallic Zn is only observed when metallic Cu is present, indicating a spillover
mechanism of dissociated hydrogen from Cu to Zn ­ a mechanism previously suggested
to aid in the reduction of ZnO (60, 73). In their own words, Holse et al. demonstrated “a
new model approach that should be generally applicable to address metal−support inter­
actions in coprecipitated catalysts and multicomponent nanomaterials” (72).
While very useful to reveal the dynamics of the Cu/ZnO system without having to worry
about (strong) metal­support interactions (SMSI, abbrev.) interfering with the results,
studies of these kind of model systems inevitably create a material gap. The industrial
catalyst is made up of not only Cu and a small amount of Zn, but also an excess of
ZnO, Al2 O3 , and additives and promoters such as MgO, ZrO2 , etc. In an effort to bridge
the material gap, many studies focus on an industrial­type Cu/ZnO/Al2 O3 catalyst, the
preparation of which differs depending on research group. With the use of a combined
HPC­UHV setup (the exact same one as presented in the following Chapter 6), Kuld et al.
followed the chemical state of Zn for a Cu/ZnO/Al2 O3 catalyst prepared by co­precipitation
(50), as well as a ZnO reference, under pre­treatments with variations in the H2 pressure.
XPS analysis of the ZnO reference found that metallic Zn was not observed, even after a
pre­treatment in 1 bar H2 , illustrating the importance of Cu for the reduction of ZnO in the
Cu/ZnO/Al2 O3 . The combination of (1) a decrease in the Cu surface area (from H2 ­TPD
and H2 ­TA), (2) an increasing amount of Zn in Cu/ZnO/Al2 O3 (from analysis of the Zn

11
Chapter 2. Applied Heterogeneous Catalysis

Auger spectrum), and (3) an increase in the combined Cu/Zn area (from N2 O­RFC) lead
the authors to suggest the formation of a Cu­Zn surface alloy rather than a ZnOx overlayer
(50). As a continuation of this study, the same group provided a quantitative description
of the Zn coverage as a function of pre­treatment conditions and related it to the methanol
activity. In their 2016 study, Kuld et al. found that increased reduction potentials (going
as high as 40 bar H2 at 280 ◦ C) produced more Cu­Zn surface alloy (51), and that the
methanol activity could be correlated to the Zn coverage. Their results demonstrated two
ways of optimizing the performance of a Cu­ZnO­based methanol catalysts: Either as a
function of H2 pre­treatment, or by increasing the reduction potential of the synthesis gas.
At the end, the authors suggest that specifically for a Cu particle with dCu =85 Å, a CO/CO2
ratio of 10 combines a high Zn coverage of approximately 0.25 with the CO2 partial pres­
sure needed for an adequate formate coverage, ind the end resulting in a high methanol
activity (51). From this study, it seems as though vital parameters for the methanol activity
are not only the Zn coverage, but also the size of the ZnO and Cu particles. It was shown
that smaller ZnO particles have a positive effect on the Cu promotion, where the opposite
applied to the size of the Cu particle. Both of the trends agree with independent studies
of the effect of ZnO doping and the effect of Cu particle size on the methanol TOF (52,
67, 74).

The previous studies that present good support of a Cu­Zn surface alloy are supported
by experiments done either in­situ or quasi in­situ, or after relevant conditions have been
applied. Examples of some of the newer studies that utilise operando techniques are
those of Pandit et al. (75), Divins et al. (76), Dalebout et al. (64), and Amann et al. (71).
Going back to a system that is more simple than oxide­supported Cu(Zn) particles, Amann
et al. investigated the nature of Zn during CO2 /CO hydrogenation over Zn/ZnO/Cu(211):
The results from XPS studies at 180 to 500 mbar show that the active state of the copper–
zinc catalyst during CO2 /CO hydrogenation involves surface metallic Zn–Cu alloy sites
(71). On the other side of the material gap, studying Cu0.7 Zn0.3 on oxide supports with
operando XAS, synchrotron­based in­situ NAP­XPS, and quasi in­situ XPS in combination
with high­pressure measurements, Divins et al. argues that a strong interaction of Cu
and ZnO, which is affected by the choice of support, is necessary for a selective and
active catalyst (76). At 20 −40 bar and in a CO2 ­enriched syngas feed, they ascribed the
ZnOx speciation to a distorted ZnOx phase with a maximum content of 9 at% Zn0 and the
rest being present as metal oxides. Interestingly, they also argue that operating with the
highest possible surface area (i.e. having small nanoparticles below 5 nm) is not always
beneficial for the catalytic process (76). The latter is in line with what is modelled by Kuld
et al. and is the opposite of what is usually thought (51): Namely, that the catalytic activity
increases as the particle size decreases, thereby increasing the accessible surface area.

12
 2.2. The Industrial­type Cu/ZnO Catalyst

Clearly, catalytic methanol synthesis and the Cu­ZnO synergy are dynamic research ar­
eas. While the use of model systems definitely helps to simplify the picture of the latter,
they may not truly represent the changes that occur in real catalysts. Supports can play
an important role in the catalytic process by providing new active sites and may strongly
affect both the physical and chemical properties of metal nanoparticles. Just as studies
are dedicated to bridging the ’material gap’, many studies are dedicated to bridging the
’pressure gap’, as heterogeneous catalysts are dynamic and known to adapt their mor­
phology and electronic structure to the surrounding gaseous atmosphere (77, 78). Thus,
for methanol synthesis catalysts, a combination of the studies that bridge the ’material
gap’ and studies that include operando studies in pressure regimes that are industrially
relevant (20–100 bar) (56, 58, 70, 76, 77, 79–84), are key to elucidating the active site of
the Cu/ZnO/Al2 O3 catalyst under industrial working conditions.

The Cu/ZnO material system is generally thought of as a prototype for studying complex
promotional interactions (and specifically the SMSI effect) in catalysis. The knowledge
accumulated for the Cu/ZnO catalyst can therefore not only help with the optimization
of the system for future purposes, it can also aid in the discovery and optimization of
new catalytic systems for the hydrogenation of CO2 , or for any other system that slightly
resembles that of the Cu/ZnO one.

13
Chapter 3. Experimental Methods

Chapter 3

Experimental Methods
Modifications and optimizations to procedures are an inevitable part of any PhD process.
Any changes to methods or procedures will be presented in the sections describing each
of the projects, which is why the following chapter will be a general description of the
methods (spanning from synthesis, to activity testing, and lastly to characterization) used
throughout the PhD process.

3.1 Catalyst Synthesis

Catalysts in this project have either been synthesised using an incipient wetness impreg­
nation (IWI) approach, or by the mixing of two separate solutions of nanoparticles and
support in their respective solutes.

3.1.1 Incipient wetness impregnation

The general procedure for synthesising catalysts via this method is as follows:

1. The solvent­capacities of the utilized support materials are 2.2 mL/g and 1.8 mL/g
for carbon black powder (Cabot Corporation, BET area of 1612 ± 38 m2 /g) and
SiO2 (44740, Alfa Aesar, BET area of 244.2 m2 /g), respectively. The carbon black
support, being nanoparticles, are always handled within a closed container until
impregnation with metal particles. SiO2 pellets are grinded to 105­300 µm sized
grains before impregnation.

2. A solution containing the metal precursors is prepared with a concentration cor­

responding to a) the wanted metal­to­metal ratio and b) the final metallic weight­
loading of the final catalyst (i.e. after reduction, with metals regarded as metallic).
The metal salts are dissolved in either water, dilute nitric acid (HNO3 ), ethanol or
acetone, depending on the solubility of the metal salts.

3. The impregnated powder is dried on a hot­plate set to <70◦ C and subsequently

calcined in static air in a muffle furnace at 220◦ C. During this treatment, the metal
precursors and solvents are thought to decompose into CO2 , N2 , O2 and NOx .

4. The dried impregnated powder is stored for further use, as described in the following
sections and chapters.

3.1.2 Mixed suspension approach

The following method of synthesising catalysts consists of two steps:
Step 1: Synthesizing metallic, crystalline nanoparticles via a hot­injection wet­chemical
synthesis.
Step 2: Supporting the newly synthesised nanoparticles on a support of our liking.

14
 3.1. Catalyst Synthesis

This approach is a collaboration with postdoc Jette K. Mathiesen. Her expertise lies in
synthesising nanoparticles with varying amounts of Ge and Cu (step 1), a process carried
out at the Department of Chemistry at the University of Copenhagen. The nanoparticles
are subsequently supported on amorphous SiO2 (step 2) in­house at SurfCat, DTU.

Step 1: In forming the nanoparticles, a hot­injection wet­chemical synthesis method is

applied. In short, the synthesis is initiated by the reduction of metal precursors, followed by
the nucleation and growth (if wanted) of the crystalline particles. Common for all synthesis
are the following steps:

• An organic solvent is added to a 100 mL round­bottom flask (RBF, abbrev.), while

stirring at 700 rotations per minute (RPM, abbrev.) under a N2 atmosphere. The
solvent is heated to a certain temperature for a certain amount of time.

• A precursor solution consisting of varying amounts of GeBr2 and Cu(O2 C5 H7 )2 pre­

cursors, also dissolved in the same solvent as originally in the RBF, is added to the
RBF while continuously stirring.

• A freshly prepared solution of the reducing agent, also dissolved in the chosen sol­
vent, is added drop­wise to the RBF while stirring. The concentration of the reducing
agent is always 200 mM.

• The resulting solution is incubated and cooled to room temperature. Once cooled,
the solution is added to a centrifuge tube with added hexane, ethanol or a mixture
of the two, then spun at 10 000 RPM at 10 minutes to precipitate the nanoparticles.

• The particles are thoroughly washed with hexane, ethanol or a mixture of the two.
Subsequently, the nanoparticles are suspended in 5 mL of either hexane or ethanol
for future use. This suspension will be referred to as a NP (nanoparticle, red.) sus­
pension.

Step 2: For the purpose of testing the nanoparticles formed during Step 1, they need to
be supported to properly disperse the nanoparticles, as well as to scale up the amount of
catalyst one has to work with. As the first step of the mixed­suspension route is a small
scale synthesis producing below 100 mg of nanoparticles (assuming full conversion), a
total metal loading of below 5 wt% was chosen for the final catalyst. SiO2 (44740, Alfa
Aesar) grinded to 105­300 µm grains was used as a support for the nanoparticles.
The amount of SiO2 is measured out according to the amount of nanoparticles available
from Step 1. A suspension of SiO2 grains and ethanol in a glass container is sonicated
for 30 minutes at room temperature to ensure a uniform mixture of the grains. To this sup­
port suspension, the NP suspension is added, and the resulting mixture of suspensions
is sonicated for another 30 minutes before being left to stir at a rotation speed of <500
RPM at room temperature for over 12 hours. The glass container is covered by parafilm
with holes in to allow the organic solvents to evaporate slowly. After stirring, the mixture is
sonicated to release any nanoparticles or support left on the sides of the glass container
during stirring, after which the solvents are evaporated in a pre­heated oven at <90◦ C.

15
Chapter 3. Experimental Methods

The resulting powder was crushed and characterized by powder X­ray diffraction to iden­
tify the crystalline phases.

3.2 Catalyst Pre­treatment and Activity Measurements

After synthesis, the dried powder catalyst has typically been characterized by X­ray diffrac­
tion or by electron microscopy, depending on the questions sought answered. Most of the
times, a sample of the catalyst, usually between 50 and 200 mg, is loaded in a reactor
made of glass, steel or glass­lined steel. If the sample amount is too small to adhere
to the required bed­length of 50 times the particle diameter, or if the sample packs too
densely in the reactor so that a pressure gradient builds up in the reactor once a gas is
flowed over the bed, the sample is diluted with inert SiC (Sigma Aldrich, A14470). The
catalyst bed is held in place by pieces of quartz­wool on both sides of the bed, and held
in place in the reactor by a thermocouple or an indent in the reactor.

Prior to testing the activity of a catalyst, i.e. measure how much synthesis gas a catalyst
can convert at a given temperature, pressure, gas flow and gas composition, one should
consider whether the catalyst has to be activated via a pre­treatment. Unless otherwise
stated, all catalysts discussed in the following chapters have undergone a pre­treatment in
the form of reduction in a gas mixture of H2 and Ar. The specifications of the pre­reduction
will always be stated before presenting any activity measurements, as the treatment will
vary from catalyst to catalyst.

3.2.1 Experimental setup

How a catalyst performs in terms of converting a reactant gas to a product is evaluated in
a reactor setup. The setup is a classical fixed­bed reactor setup comprised of four main
parts: A gas­wall for mixing reactant gases; A reactor bed and a furnace for controlling
the temperature during activity measurements and pre­treatments; A pressure­controller
down­stream of the reactor for controlling the pressure in the reactor; And lastly, an anal­
ysis part based on gas chromatography (GC) and quadrupole mass spectroscopy (QMS).

The setup, as it was originally build, is described in the PhD thesis by Irek Sharafutdinov
(85), but it has undergone several changes throughout the last couple of years ­ all of
which are equally important to the current PhD project and are originally described in the
PhD thesis by Thomas Smitshuysen (37). As the setup has been the primary work­horse
of the current PhD project, a considerable amount of space will be spent on describing
the specifics of the setup and the changes made to it. Specifics on data collection and
analysis is described in depth and demonstrated in Appendix E.1.

16
 3.2. Catalyst Pre­treatment and Activity Measurements

Figure 3.1: Schematic of the experimental setup. Modified with permission from (37).

Figure 3.1 shows a schematic of the setup as of November 2022. The flow of reactant
gases are controlled via analogue Brooks mass­flow controller (MFCs). These are cali­
brated using a Definer 220 volumetric flow meter, situated down­stream of the reactor and
after the pressure controller. It is important to note that the calibration is done at standard
conditions of 273.15 K and 1 bar. All gas lines are made of stainless steel, except for the
CO­line which is made entirely of brass tubing on the high­pressure side of the MFC. Be­
sides brass tubing, the CO­line is equipped with a carbonyl filter, both additions that are
made to minimize the formation of carbonyls.
As the gases are mixed on the low­pressure side of the MFCs, they can either bypass or
flow through the reactor. For high­pressure experiments at 10 bar, only reactors made
of stainless steel or glass­lines stainless steel are used, as reactors made of glass will
not withstand that high a pressure. The reactor is situated inside a furnace capable of
heating up to 1200◦ C. The temperature in the reactor is monitored with a K­type ther­
mocouple situated in the reactor, just below the catalyst bed. The same thermocouple is
also what controls the power input of a CPX400DP Dual 420 watt DC power supply and,
via a PI loop, controls the temperature of the furnace and hence the reactor and catalyst
bed. In short, the PI loop calculates the error between a temperature setpoint and the
temperature measured in the reactor and applies a correction based on proportional (P)
and integral (I) terms. This way, it is possible to control the output of the power supply
in such a manner that the measured temperature is restored to the desired temperature
with minimal overshooting.
Down­stream of the reactor, all of the gas lines after the furnace are heated to avoid
precipitation of water and any other gaseous products. Also down­stream, a pressure­

17
Chapter 3. Experimental Methods

controller is situated. The pressure in the reactor is calculated as the average between the
pressure output of the pressure­controller and the pressure measured by gauge 1 (P1).
Besides the pressure­controller that also works as a pressure transducer, there are four
other ways of measuring pressure on the setup, all used in combination with each other
to calibrate and/or monitor pressure: A digital pressure transducer (P1) on the gas­mixing
side, on the low­pressure side of the MFCs; An analogue manometer (P2); A digital pirani
(P4, VPM­5 Smartpirani, Sens4) before the GC, monitoring the inlet pressure; And lastly
another digital pirani P5, monitoring the outlet pressure of the GC.
Besides controlling and monitoring the pressure, there are relieve valves that can be used
to relieve the flow of gases going to the GC. This is found useful during experiments with
a combination of high pressures and high flows of gas.
Analysis methods: Gas chromatograpy and mass spectrometry
As briefly mentioned, the analysis part consists of a GC and a QMS, the latter being a new
addition build by students Mathias Ribergaard Vinther and Kasper Larsen Schou. Each
of the methods have their strengths and weaknesses, with the real work­horse in terms
of analysis being the on­line gas chromatograph (7890A, Agilent) with flame­ionization
and thermal conductivity detectors, shortened FID and TCD, respectively. In short, a gas
chromatograph works by taking a sample of a gas, leading it through a column that is
coated with a stationary phase, with which the different components in a gas interacts
differently. The analysis then relies on how different components of the flowing mobile
phase, the gas sample, are separated according to how each constituent interacts with
the stationary phase. As the end of the column, a suitable detector is found.
Specific for the GC used in this PhD project is that a sample of the gas to be analyzed
is taken by a rotary valve, which then empties the gas into a sample tube. While the gas
sample is acquired, nothing is changed in the experiment. The sample is subsequently
emptied into columns, of which three are in series leading to the TCD detector, and a
fourth column leads to the FID detector. The TCD detector measures anything but the
carrier gas, which, for this system, is He. The FID measures any flammable hydrocar­
bons present in the sample gas. The schematic of the gas chromatograph is given in the
Appendix in Figure E.6.
The quadrupole mass spectrometer needs a low enough vacuum that the chances of
ions colliding with other molecules in the mass analyzer is minimized. In order to achieve
vacuum low enough, a sniffer is used to leak gas from the reactor setup into to mass
spectrometer situated in a vacuum chamber. With this technique, the QMS does not
experience the 1 bar of gas coming from the reactor setup. The base pressure of the
vacuum system is in the 1.0 × 10−8 mbar range, achieved by pumping on the system
with a turbo pump (Pfeiffer Vacuum) and an dry scroll pump (Edwards, nXDS10i). Once
ions have been produced in he ionization area of the instrument, they are focused and
passed along the middle of four rods of the quadrupole. The motion of the ions will depend
on the electric field so that only ions of a particular mass­to­charge ratio (m/z, abbrev.)
will pass through the detector. By varying the RF potentials between the four rods, ions
of different m/z are brought to the detector and thus a mass spectrum is created.

18
 3.3. X­ray Diffraction

Commenting on the strengths and weaknesses of gas chromatography as used on this

setup, it comes down to the following: A strength in terms of detection of molecular
species, with each molecular species having their own unique position in the GC spec­
trum; A main advantage is that the GC is also easily calibrated by the user; A disadvantage
is of course the low time­resolution as a sample of the analysis gas is taken every 15 min­
utes, so any changes happening within this period of time is non­detectable; A weakness
is also the fact that neither the TCD nor the FID detector is capable of detecting water
vapor.
The QMS excels in its better time­resolution, creating a mass spectrum every 2 minutes.
Furthermore, it is useful for monitoring water, an ability especially important for when one
wants to monitor the rWGS reaction or the reduction of a catalyst. However, there are
also disadvantages to this technique, which is why it is best paired up with a sensitive
method such as gas chromatography. The down­side of quadrupole mass spectrometry
is that cracking patterns of different components of a gas sample can look very similar, at
times making it difficult to distinguish between molecules of the same masses if you have
a mixture of all of them.

3.3 X­ray Diffraction

X­ray diffraction is one the most used ’hard’ X­ray methods for characterization of mate­
rials. It is based on the constructive interference of between monochromatic X­rays and
a crystalline sample. The interaction only produces constructive interference, and thus a
Bragg peak, given Bragg’s law is fulfilled:

2 · dhkl · sin(θhkl ) = n · λ n = 1, 2, ... (3.1)

Here, dhkl is the separation between crystal lattice planes with Miller indices hkl, θhkl is
the diffraction angle, n is the order, and λ is the X­ray wavelength. By scanning a sample
over a range of diffraction angles, all possible diffraction directions of the crystal lattice
should be obtained as the powdered material, and thus the crystals, are randomly ori­
ented. The diffraction of an incident beam with a wavelength λ on crystal lattices with a
given d­spacing is illustrated in Figure 3.2.
X­rays interact with each of the electrons in a sample, whether it’s crystalline or amor­
phous. As the waves scattered from each of the electrons meet, they interfere, and a
scattering pattern arises. This pattern is characteristic of how the atoms, and thus the
electron clouds, are arranged in a sample. With every crystal having a unique scattering
pattern, it is possible to compare to standard reference patterns or construct structural
models. In crystalline materials, the electrons are arranged in simple, periodic patterns,
and these periodic patterns give rise to well­defined Bragg peaks. Amorphous materials,
such as amorphous SiO2 , do not have sharp Bragg peaks, as they do not have long­
range well­defined crystalline order. But just as crystalline materials, any other atomic
arrangement found in the amorphous material will also give rise to interference effect,
called diffuse scattering.

19
Chapter 3. Experimental Methods

This thesis will not go into mathematical detail about how ones goes from scattering from
a point all the way to scattering from a crystal structure, but it is necessary to introduce a
few concepts to lay the ground for the coming sections.

Figure 3.2: Illustration of the Bragg law.

We need to introduce the scattering vector, Q, which is widely used in scattering theory as
the difference between the incoming and outgoing beam. Why is Q important? Because
assuming the scattering event is elastic, Q can be calculated from the scattering angle
and the radiation wavelength ­ in other words, it represents the momentum transfer of the
scattering event. A scattering event from a point is illustrated in Figure 3.3.

Figure 3.3: A scattering event illustrating the relation between incident and reflected
beams and their corresponding wave vectors, ki and kf , and the scattering vector Q.

From geometry, one can derive the relation between the magnitude of the vector (Q) and
the scattering angle for elastic scattering:

4 · π · sin(θ)
Q = 2 · k · sin(θ) = (3.2)
λ

Moving on to describing how a wave is scattered from one point, one can then extend to
consider how an X­ray is scattered from an atom. To do that, one must firstly consider the
scattering amplitude from a single electron and move on to integrate over the full electron
density, ρ, of an atom to end up with the ’form factor’, f (Q), of an atom:

∫
f (Q) = ρ(r) · exp(iQ · r)dr (3.3)

20
 3.3. X­ray Diffraction

If we then consider a collection of N atoms, with an atom j positioned at r, the scattering

amplitude, ψ(Q), is a sum over all the atoms in the sample:

∑
N
ψ(Q) = fj (Q) · exp(i(Q · rj )) (3.4)
j=1

The scattering amplitude of Equation (3.4) is important, but cannot be measured directly.
Only the intensity of the scattered wave can be measured. Luckily, it is related to the
scattering amplitude, so that we in theory can examine the scattered intensity to obtain
information about the structure of a material:

I(Q) ∝| ψ(Q) |2 (3.5)

How does all of this fit into powder X­ray diffraction? As previously mentioned, crystals are
randomly oriented in a powdered material, meaning there will always be crystals whose
orientation satisfy any given Bragg condition. When doing an XRD experiment, the 3D
nature of the scattering amplitude of Equation (3.4) is condensed into 1D. This happens
when the Debye­Scherrer rings, which are cones of diffracted rays coming from lattice
planes that are spaced d1 , d2 and so on apart (see Figure 3.4 for reference), are integrated
into a 1D pattern during data analysis.

3.3.1 PXRD data analysis

Broadening of the Bragg peaks in the 1D pattern not only comes from beam divergence
and wavelength distribution. It also comes from limitations of the crystallite domains and
the lattice strain. That can be derived from Equation (3.4), but is actually explained more
easily by the Scherrer equation:

K ·λ
βhkl = (3.6)
Dhkl · cos(θhkl )

Here, β is the peak profile width, K is the crystallite shape form factor (for spherical crystal­
lites, it is 0.9 (86)), θ the diffraction angle, and λ the X­ray wavelength. D is the parameter
expressing the crystallite domain size. Consequently, narrow Bragg peaks indicate large
crystallite domains, and broad Bragg peaks indicate smaller crystals. Because it is easily
applicable, the Scherrer equation is widely used to estimate the size of crystals. Broad­
ening of Bragg peaks is not only a results of the crystallite not being infinitely large and
it not being perfectly ordered (i.e. it having defects, non­uniform interplanar spacing, and
disorder, collectively referred to as microstrain). It is also influenced by a diffractome­
ter function that contributes to the peak width and profile. The diffractometer function is
determined by measuring on a standard sample with the same optics as one would use
for PXRD measurements. The intrinsic line broadening and peak profile is subsequently
separated from the peak profiles of samples of interest.

21
Chapter 3. Experimental Methods

The instrumental broadening of the diffractometers used in this work are given in Ta­
bles A.1 and A.2. Broadening of Bragg peaks due to microstrain have not been included
in the analysis of the peak profiles. All data analysis of PXRD data obtained at the in­house
diffractometers was handled in the Malvern Panalytical software package HighScore Plus
(87) and the OriginPro software package. In most cases, a light smoothing of the spec­
trum has been executed to decrease the level of noise, however this is never done for
spectra subjected to analysis.
PXRD data of Cu and CuGe nanocrystals synthesized as described in Section 3.1.2 have
been measured by Jette K. Mathiesen at the University of Copenhagen. The PXRD data
have been plotted and analyzed in the OriginPro software package, using a Lorentzian
distribution for the purpose of estimating crystallite sizes. Unless otherwise mentioned,
the diffractograms presented in this thesis are smoothed, but only for visualization pur­
poses. Reference diffraction patterns are collected from ICSD (88).

3.3.2 PXRD apparatus

Two different pieces of equipment have been used throughout the PhD project: For ex­situ
measurements, an Empyrean (Malvern Panalytical) diffractometer; for in­situ measure­
ments, an X’pert Pro (Malvern Panalytical) diffractometer. All diffraction patterns were
measured at 40 kV and 45 mA.
Both diffractometers are equipped with a Cu anode (λ=1.5418 Å). As Cu radiation con­
sists of both Kα and Kβ , the Cu radiation is filtered using a Ni­filter to suppress the CuKβ
radiation. If not done, every Bragg peak will appear two times, and the W­impurities of the
radiation source will also appear. Kα radiation consists, in part, of Kα1 and Kα2 . Hence,
each Bragg peaks will consist of two different reflections. Because Kα1 has a slightly
shorter wavelength and twice the intensity of Kα2 , the latter will appear as a small shoul­
der ­ at lower values of θ, the two reflections almost overlap, and the larger the value of
θ, the larger the separation between the two reflections.
Both diffractometers are equipped with a graphite­monochromator on the diffracted beam­
side to minimize fluorescence.

For ex­situ PXRD measurement, the sample (whether it be as­synthesized, pre­treated

or post­mortem), is loaded between two polyimide (Kapton, red.) foils or on a zero­
background plate. The latter is preferred, but kapton foil has been used whenever the
sample tends to easily become static or if the sample amount was too small to load on a
zero­background plate. Kapton foil does have a broad background, but no sharp Bragg
peaks that can interfere with any Bragg peaks arising from a crystalline material. The
zero­background plate, as hinted by its name, is made of silicon cut at a special orienta­
tion to give no background noise from diffraction angles 20 to 120 (2θ).
Besides a Ni­filter, the following optical elements are used for all ex­situ PXRD measure­
ments: On the incident beam­side, the combination of a 4◦ divergence slit and a 10 mm
beam mask is used, to control the amount of sample that is irradiated by the X­ray beam
and therefore to avoid any signal coming from the sample holder; on both the incident and
the diffracted beam­sides, 0.02 RAD soller slits are used to reduce peak asymmetry, and

22
 3.4. X­ray Total Scattering with PDF Analysis

to improve peak shape and resolution; a diffracted anti­scatter slit (15.8 mm) is placed on
the diffracted beam­side to reduce background signal.

For in­situ PXRD measurements, a sample is kept in the sample holder while being sub­
jected to heat and gas. It is, however, also possible to use the diffractometer for ex­situ
PXRD measurements. A sample of catalyst is loaded into a ceramic macor (a mixture of
oxides, red.) holder. In the sample holder, the sample is ensured to have a flat surface
to ensure the correct sample height according to the placement of the furnace and the
X­ray path in the diffractometer. The furnace surrounding the sample holder is an XRK
900 (Anton Paar) model with transparent windows made of Be. The sample holder can be
pumped out via the combined pumping/gas­system connected to the diffractometer, and
gas at a maximum of 1100 mbar can be introduced to the sample and sample holder. The
gas pressure in the sample holder is controlled via a Bronkhorst pressure controller and
the continuous pumping of gases using a dry scroll pump (Edwards, nXDS6i). Bronkhorst
MFCs control the inlet of gases, and the gases are mixed before being let in either around
or through the sample. The flow of gasses are analyzed using a Pfeiffer QMS. To achieve
low enough pressures for the QMS, a heated glass­capillary sniffer is used.
On the incident beam side, the following optical elements are used: A Ni­filter, a 0.04 RAD
soller slit, and a 4◦ divergence slit. On the diffracted beam side, a diffracted anti­scatter
slit (10.4 mm) is used.

3.4 X­ray Total Scattering with PDF Analysis

Although the Scherrer equation is easily applicable for determining crystallite sizes, a
problem arises when the PXRD pattern is broadened to such a degree that it is impossible
to distinguish between the reflections from the crystallite domains and those coming from
the amorphous support. To extract structural information from samples that are either
amorphous or have no long­range atomic order, we need to treat both the Bragg peaks
and the diffuse scattering. In theory, we can also do that. When measuring signals from
isotropic powders we can simplify Equation (3.4) to only consider the length of Q instead
of its direction. That simplification comes out as the Debye scattering equation:

∑∑ sin(Q · rij )
I(Q) = fi (Q)fj (Q) · (3.7)
Q · rij
i j

This equation can, in theory, be used to calculate the atomic positions of all particles in
a sample. In practice, it is a computationally very heavy task. A method that have been
developed for structural analysis using the Debye scattering equation is X­ray total scat­
tering (TS) coupled with Pair Distribution Function (PDF) analysis. The PDF is essentially
a histogram of interatomic distances in a sample. No peak will appear in the PDF above
the longest interatomic distance, and it is therefore a useful and intuitive way of directly
monitoring nanoparticle sizes, where regular PXRD will maybe fail. It is important to note
that, as with XRD, the size estimates from the PDF are not directly the nanoparticle sizes,
but rather the coherent scattering domain. In cases where no defects are present, the

23
Chapter 3. Experimental Methods

size from the PDF equals the size of the actual nanoparticle.

Obtaining the PDF from X­ray scattering data includes going through a series of steps.
The first step is to correct the experimentally measured scattering data so that only the
data coming from elastic, coherent scattering events are left. The corrections can include
removing contributions from Compton scattering (i.e. inelastic and incoherent scattering
events), fluorescence, absorption of the beam by the sample, and signal from the sample
holder, the support, a dilutant, or air in the X­ray beam path. After corrections are done,
the total scattering structure function, S(Q) is introduced. The normalization by the atomic
form factor is done at this step to ensure that features at high scattering angles, i.e. the
local structural information, are amplified. From S(Q), the reduced total scattering struc­
ture function, F(Q), is obtained. Where F(Q) and S(Q) are in reciprocal space and can be
be used to analyse the long­range structure, the reduced atomic pair distribution function,
G(r), is in real space. This is the function that is normally referred to as the PDF and is
obtained by taking the sine Fourier transform of F(Q). It is given by:

∫ Qmax
2
G(r) = F (Q)sin(Qr)dQ (3.8)
π Qmin

Structural analysis can be performed in real space using G(r) to understand the scattering
data in terms of atomic pair correlations. While the PDF is referred to as a histogram of
interatomic distances, the expression in Equation (3.8) does not intuitively show why.
This is where the radial distribution function, R(r), is introduced:

1 ∑ ∑ fv · fu
R(r) = · δ(r − rvu ) (3.9)
N v u ⟨f 2 ⟩

As realized from Equation (3.9), the R(r) provides a histogram of interatomic distances:
A pair of atoms consisting of atoms v and u gives rise to a delta function at rvu , which is
the distance between the two atoms. The intensity of the peak is related to the scattering
power of the two atoms through their form factors f. The reduced pair distribution function,
G(r), is is closely related to R(r) through:

R(r)
G(r) = − 4πρ0 (3.10)
r

Here, ρ0 is the number density, i.e. the number of atoms in a volume of the sample. G(r)
fluctuates around 0 because of the second term of Equation (3.10), and if a certain in­
teratomic distance is present more than average, a positive peak is seen in G(r), while a
interatomic distance present less than average will give negative values for G(r). While
R(r) might seem more intuitive to use, the G(r) function is often used in PDF analysis as
this is the function that is directly obtained when Fourier transforming the experimental to­
tal scattering data (89), as shown in Equation (3.8). Using R(r) would also mean to make
a series of assumptions about the structure, while using G(r) means that assumptions are

24
 3.5. Microscopy: (S)TEM and SEM

made with the model used in the refinement. The idea of PDF analysis is illustrated visu­
ally with plots of I(Q), F(Q) and G(r) of a carbon­supported CuZn sample in Figure F.11.

The integration in Equation (3.8) should in principle run over infinite Q, but the experimen­
tal configuration dictates the range, why in practice the integral runs over a finite Q­range
given by Qmax and Qmin . A large Q­range can be achieved by a combination of high X­
ray energy and having the detector close to the sample as realized from Figure 3.3 and
Figure 3.4, and Equation (3.2). This is why all X­ray total scattering data presented in
this work has been done with synchrotron radiation using rapid­acquisition PDF (RA­PDF,
abbrev.) (90). Choosing the optimal Q­range is a compromise between including noise in
the data (large Q­range) and broadening of the PDF peaks and oscillations in the Fourier
transform (small Q­range). This matter is not something that have been investigated in
this work, but it is worth noting for anyone interested in using X­ray TS with PDF analysis
for their scientific problem.

Figure 3.4: An X­ray scattering experiment. The principle of powder diffraction is illus­
trated: More than one crystal is at the correct orientation (angled θ to the incident beam)
to satisfy Bragg’s law, resulting in more than one cone of diffracted rays.

3.5 Microscopy: (S)TEM and SEM

Moving on from X­rays to using electrons, another world opens up. Instead of a method
that provides a structural average, we now have a method that reveal the atomic structure
and distribution of individual nanoparticles.

The main difference between scanning electron microscopy (SEM) and transmission elec­
tron microscopy (TEM) is that for SEM, the contrast is due to the topology and composition
of a surface. For TEM, where the electron beam passes through a thin sample (as a rule
of thumb, no more than 100 nm (91)), we get the same information as we do in SEM,
along with signals originating from transmission.

25
Chapter 3. Experimental Methods

Figure 3.5: Events happening as a results of the interaction between a primary electron
beam and a sample in an electron microscope.

3.5.1 Scanning electron microscopy

The scanning electron microscope consists of four main components: The electron source,
from where an electron beam (referred to as the primary electron beam) with a selectable
energy is emitted; the lens system that focuses the electron beam which exits the lens
system to hit the sample; a scan unit, which makes sure to move the beam in a raster pat­
tern over the specimen area; and lastly the detection unit that, hinted by its name, picks
up and converts the signals coming from a series of events resulting from the interaction
between the electron beam and the sample. The events resulting from the interaction is
are illustrated in Figure 3.5.
As the primary beam interacts with the sample surface, we have:

• Back­scattered electrons (BSE) resulting from the elastic back­scattering when the
beam electrons collide with core electrons. These electrons have energies compa­
rable to those in the primary beam (1 −30 keV) and therefore have larger mean free
paths. The signal coming from BSEs is less sensitive to surface topography, and
the intensity depends on the composition of the sample.(91)

• The beam electrons can excite the sample, causing the sample to emit what is re­
ferred to as secondary electrons (SE). Contrary to BSEs, these low­energy electrons
(0 −100 eV) have small mean free paths, meaning the signal has to come from the
surface region of the sample.(91)

• Whenever an atom is excited, it will emit X­rays when it relaxes after emitting an
SE. The amount of energy released depends in which inner shell it is transferring
from and to. As this is unique for energy element, the detection of characteristic
X­rays allows for the determination of the chemical composition of a sample. This
technique is referred to as energy­dispersive X­ray spectroscopy (EDX or EDS).

• The three phenomena above are not the only events happening. Secondary elec­
trons, Auger electrons or X­rays created whenever an X­ray or a secondary electron
are re­absorbed into the sample can also be detected.

26
 3.5. Microscopy: (S)TEM and SEM

To simplify, SEM is all about rastering an electron beam over a surface and detecting the
yield of either secondary or backscattered electrons as a function of the position of the
primary electron beam. In this project, SEM have mostly been used in connection with
EDX detectors that detect the characteristic X­rays of a sample.

3.5.2 Transmission electron microscopy

As hinted, transmission electron microscopy makes use of a primary electron beam that
generates the same signals as for scanning electron microscopy. In addition, signals
originating from elastically scattered (diffracted) electrons, transmitted electrons with no
interactions, and inelastically scattered electrons, are generated.

Figure 3.6: Schematic diagram of image formation in a transmission electron microscope.

Modified version, reprinted with permission from (91). Copyright John Wiley and Sons,
2008.

In theory, TEM works just like an ordinary optical microscope, except that an electron
beam is used instead of light. The process of image formation is illustrated in Figure 3.6,
where the process is illustrated both in terms of aperture (original print) and with modified
text (box to the left). The image of the sample is formed by the diffracted electrons, which
are combined by the means of an electromagnetic objective lens. This lens produces a
2D diffraction pattern of the sample in what is referred to as its focal plane. The diffracted
beams recombine to form an image in what is referred to as the image plane. Optimizing
a series of lenses that are positioned under the objective lens one can obtain a diffraction
pattern or a magnified image of the sample. The information to gain from TEM rely on
the analysis of the contrasts in the image. By inserting apertures in either the diffraction
(indicated as ’contrast aperture’ in Figure 3.6) or the objective planes, we get to one of ei­
ther two regimes: Conventional TEM or high­resolution TEM (HRTEM). For conventional
TEM, the contrast intensity relates to areas that fulfill the Bragg condition or to areas that
are thicker or contain heavier elements, thereby increasing the scattering power.

27
Chapter 3. Experimental Methods

As with other techniques, higher energy means higher resolution. This, however, comes
at the expense of increased beam damage. Acquiring an image is therefore a careful
study of the highest resolution obtained with the least amount of beam damage.

A technique that is very useful in combination with conventional TEM is the procedure of
focusing the electron beam to a convergent spot and raster scanning the beam across the
sample, similar to what is done for SEM, while detecting both scattered and transmitted
electrons. This method is, surprisingly, named scanning transmission electron microscopy
(STEM). For this method, the image contrast is given by the atomic number, Z. Many
techniques are possible in STEM. In this study, the technique of high­angle annular dark
field (HAADF, abbrev.) imaging is used in combination with STEM. With the configuration
used in STEM­HAADF, or more specifically the annular detector, it is possible to collect the
electrons that are scattered at high angles while also analyzing the transmitted beam. The
first will form the image, the latter will provide elemental analysis. Since heavier atoms
will scatter more electrons at higher angles due to the greater electrostatic interactions
between the nucleus and electron beam, atoms with higher Z­number will appear brighter,
which at the end gives this method its chemical sensitivity.

3.6 X­ray Photoelectron Spectroscopy

X­ray photoelectron spectroscopy (XPS) is a widely used ’soft’ surface analytical method.
The method is based on the photoelectric effect:

Ek = hν − Eb − ϕ (3.11)

The ejection of an electron from a core level happens when an X­ray photon with the
energy hν is absorbed. The binding energy (with respect to the Fermi level of the sample)
of the core electron is given by Eb , and Ek is the kinetic energy of the photoelectron that
is emitted (relative to vacuum level). The latter is the experimental quantity measured
by the spectrometer and is dependent on the photon energy of the incident X­rays. The
last quantity of Equation (3.11), ϕ, is the work function of the spectrometer (23). It is a
constant that is compensated internally after the calibration of the spectrometer, usually
against a well­characterized standard such as pure metals (eg. Ag, Au or Cu).

A schematic of the photoemission and Augers processes is pictured in Figure 3.7. In

the case of a 1s photoelectron being ejected from the K shell of an atom, measuring the
kinetic energy of that photoelectron allows one to calculate the binding energy of it. The
atom will return to its ground state after being ionized by either emitting an X­ray or by
filling of the electron hole in the core level by an electron from a higher level. In case of the
latter process, named the Auger process, another electron is emitted to conform with the
principle of conservation of energy. The Auger electron has an element specific kinetic
energy, which is independent of the energy of the incident X­ray beam.

28
 3.6. X­ray Photoelectron Spectroscopy

Figure 3.7: The photoemission and Auger processes resulting from absorption of an X­ray
photon.

Figure 3.8: The mean free path (nm) of electrons as a function of their kinetic energy (eV).
Reprinted with permission from (23). Copyright John Wiley and Sons, 2015.

The strength of XPS as a method comes from two things: Its surface­sensitivity, and its
chemical specificity. Therefore, it is a method widely used for determining for example
which active species are on the surface of a material and what chemical state the species
are in. The depth of analysis of XPS varies with the kinetic energy of the electrons under
consideration. With the laboratory X­ray sources applied in this PhD project (AlKα and
MgKα), the kinetic energies of the photoelectrons are in the range 0­1 keV (23). According
to Figure 3.8, the inelastic mean free path (IMFP) of electrons is just below 2 nm. The
IMFP is the measure of how far an electron can travel before being scattered and can
effectively be manipulated either via angle­resolved XPS or by using synchrotron radiation
to gain optimal surface­sensitivity.

29
Chapter 3. Experimental Methods

3.6.1 Setup and data analysis

Most of the XPS measurements presented in this work have been carried out in a com­
bined HPC­UHV chamber with base pressure of approximately 1 × 10−9 mbar. The setup
and data analysis will be discussed in Chapter 6. When not using the HPC­UHV chamber
for XPS measurements, these were carried out using a Theta Probe system from Thermo
Fisher Scientific, with a base pressure of less than 5 × 10−10 mbar. It was operated with a
monochromatic Al anode using the Kα (1486.7 eV) line and a spot size of 400 µm. All data
treatment was done in either the CasaXPS or Avantage processing software packages,
depending on the experimental setup.

30
Chapter 4

New Catalysts for CO2

Hydrogenation
Methanol synthesis catalysts represent a very mature technology, but it is still a research
area with many new directions and suggestions. Since its first industrial application in
1966, the ternary Cu/ZnO/Al2 O3 (CZA, abbrev.) have dominated both in industry, but also
in research (92). But with the need to change to alternative feedstocks also comes the
need for alternative catalysts that can both live up to the task of producing methanol from
CO2 and H2 as well as overcome the challenges where the currently used CZA catalyst
comes short. A large part of the literature related to catalytic methanol synthesis review
the science, new as old, and recent collections of activity data for CO2 hydrogenation have
been presented in such a way that the reader can easily realize why the CZA catalyst is
so difficult to compete with (12, 18, 36, 37, 92). The Cu/ZnO/Al2 O3 catalyst is however
not without flaws: It is highly selective to CO via the rWGS reaction (Equation (2.5)) and
the production of methanol is further hindered by the production of water, also from the
rWGS (93, 94). A second flaw is that the lifetime of methanol catalysts in industry is 2­4
years (18), and even within this time span the activity drops continuously (49, 67, 70, 93,
95). To put it simply: The shortcomings of the Cu/ZnO/Al2 O3 catalyst are what any new
catalyst should be optimized to overcome.

The requirements for a new catalyst for methanol synthesis are diverse. There are the
usual academic and industrial ones we should always strive to fulfill, and then there are
the more specialized ones related to using CO2 ­rich feed gases at low pressures if we
want to produce methanol in small­scale plants near the end­user.
From an academic point of view, a new catalyst should not only be highly active in terms
of turnover frequency per active site, it should also be selective when it comes to prod­
uct formation (23). That way, the amount of active material used for the catalyst can
be minimized, which is especially useful if one chooses to work with some of the more
expensive metals. Furthermore, it should have a high thermal stability against sintering
and structural change to minimize the loss of active surface area. Ideally, it should also
have a good resistance towards poisons such as S, Cl and carbonyls, especially as the
impurities in alternative feedstocks might differ from those seen today. From an indus­
trial point of view, but without going in depth with the science of up­scaling, the catalyst
should basically be optimized in terms of pellet size and shape. This is to avoid a large
pressure drop over a bed of catalyst pellets and for the reaction not to be restricted by
diffusion. The pellets should have a high mechanical strength and thermal stability, and
naturally be optimized in terms of loading of any precious metals. Ideally, if the catalyst

31
Chapter 4. New Catalysts for CO2 Hydrogenation

is prone to deactivation over time, it should be possible to regenerate it and obtain, if not
100 %, then at least a significant amount of activity to limit the need for replacement. The
up­scaling of new catalysts is not studied in this work, nor is the availability and scarcity of
precious metals taken into account when proposing potential candidates for active mate­
rials, simply because the study is still at its early stages. In the future, one should confer
with studies such as that published by Laursen et al. to evaluate whether the cost or the
availability of a material could be a limiting factor in a potential up­scaling (96).
Lastly comes the more specialized requirements. If we are to move away from fossil­
fuel based methanol production at large, centralized units, and move on to small­scale,
’green’ alternatives, the challenges to consider are related to feedstock, temperature and
pressure.

• Alternative feedstocks: In recent years, there has been a big effort in utilizing
CO2 for the direct production of ’green’ methanol with no new net CO2 . Combining
H2 produced from renewable energy with CO2 trapped from the environment, or
captured before it even reaches it, is a beautiful scenario (7). This would require
a catalyst that selectively produces methanol, while also neither producing CO or
H2 O from the rWGS nor uses CO as a main feed for methanol production.
Theoretically, the equilibrium concentration of methanol is higher from purely CO
than it is from purely CO2 , simply because Equation (2.3) is more exothermic than
Equation (2.4), as was seen in Chapter 2. Development of a catalyst that is selective
for CO hydrogenation is therefore not completely irrelevant. But it is CO2 , and not
CO, that is more abundant in the environment and by point sources such as industry,
so using CO as the main carbon­source in the feedstock for methanol production
would require an additional pre­processing step of the more abundant CO2 into CO.

• Lower pressure and temperature: Because methanol synthesis from CO2 and
H2 is exothermic (see Equation (2.4)), the lower the temperature, the less CO is
needed to counteract the endothermic rWGS reaction. Unfortunately, direct CO2
hydrogenation scales in such a way with temperature that there is an optimum where
the temperature is high enough for the reaction to proceed, but at the same time low
enough to not be thermodynamically hindered. At the same time, the equilibrium
constant of Equation (2.4) dictates that a higher pressure yields a higher methanol
conversion. In practice, we have to develop not only a selective catalyst, but also a
more active one.

32
 4.1. Proposing a New Cu­X system

4.1 Proposing a New Cu­X system

How do we move away from a catalytic system that has been optimized specifically for
methanol synthesis and has been studied extensively throughout the last decades? The
answer is, we turn to theory. PhD Thomas Smitshuysen originally screened suitable com­
ponents for a X­promoted Cu system using the open quantum materials database (OQMD,
abbrev.) (97, 98). A cut­off formation enthalpy of −3 eV/atom was chosen for the oxide
phase of X. The oxide should be reducible ­ more so than for example SiO2 or MgO, which
are most often used as (irreducable) catalyst supports and have formation enthalpies of
−3.1 eV/atom and −3 eV/atom, respectively. Of the post­transition metals and metal­
loids that were initially screened, the following elements for the X dopant are suggested ­
here listed with the corresponding formation energy of the oxide: Ga (­1.9 eV/atom), Ge
(­1.9 eV/atom), In (­1.7 eV/atom), Sn (­1.7 eV/atom), Sb (­1.7 eV/atom), Bi (­1.5 eV/atom),
Pb (­1.3 eV/atom).

In their recent study of the strong metal support interaction in the Cu­ZnO system, Elnabawy
and coworkers systematically screened the originally proposed oxides, along with several
others, as potential support/promoters in a simple Cu(211) model (99). The (211) stepped
surface is chosen as it most efficiently models the defects that are known to be active on
the commercial Cu/ZnO­based catalyst (28). When accounting for the reducibility of the
oxide under relevant experimental conditions, as well as how favorable it is for the alterna­
tive metal to alloy into the Cu(211) surface layer, they not only discovered the uniqueness
of ZnO, but also concluded that other oxides are either too stable or too readily reducible
under reactions conditions (99). The study leaves us with not many choices but ZnO and
Ga2 O3 if we want to stay in the ’partial wetting’­category. This category is exactly where
we want to be if we want the perfect balance between promoting the Cu surface and not
completely diluting the Cu sites that are responsible for the CO/CO2 adsorption/activation.
Interestingly, the study postulates that if one were to dope a small amount of promoter
instead of relying on it as a oxide support to act as ZnO does for the Cu­ZnO system,
the picture might change. Results yet to be published by the group of Jens Nørskov at
DTU Physics reveal a series of promising catalysts for CO2 hydrogenation within this the­
ory. The difference is now that the wetting behaviour is not taken into account as it has
already been theoretically shown that for example Ge and Sn have a stronger tendency
to completely wet the Cu surface if their oxides are used as a support in a Cu­X system.
Not taking the wetting behaviour into account, postdoc Ang Cao have screened a series
of Cu­X catalysts, all modeled as a Cu(211) surface with a single X atom replacing a Cu
atom in the 211 step site. We thereby move away from relying on the alternative metal
as an oxide support, but instead as a promoter in a single­atom surface alloy. For X=Zn,
the model is in accordance with the study of Behrens et al. that takes into account the
experimental evidence of the state of Zn during reactions conditions (28).

33
Chapter 4. New Catalysts for CO2 Hydrogenation

Figure 4.1: Theoretical methanol TOF on Cu(211) metal surfaces, promoted with a single
X atom, as a function of descriptors E(HCOO*) and E(CO*). Unpublished results by Ang
Cao.

Figure 4.1 showcase the series of promising catalysts for CO2 hydrogenation in a vol­
cano plot. The descriptors for activity are chosen to be the bonding energy of formate
and CO, E(HCOO*) and E(CO*) respectively. The X metal promoter in the Cu­X system
span most of the periodic table, but here, a zoom­in that includes the originally suggested
promoters are shown: Zn, Pd, Ag and Cd (transition metals); Ga, In, Sn and Pb (post­
transition metals); Bi and Ge (metalloids). All Cu­X systems are found to be stable in
the configuration pictured on the right in Figure 4.1. A recent study show that whenever
adsorbate­adsorbate interactions are included, explicitly by including adsorbate specta­
tor species, the theoretical and experimental TOFs for CO2 hydrogenation are in good
agreement (100). In the study of Cu­X catalysts, the same approach is used to include
the sensitivity of the methanol formation rate on Cu(211) to the formate coverage. More
specifically, a threshold value is introduced. Cu­X systems with formate binding energies
that are larger (more negative) than the threshold are found to suffer from poisoning of
formate on the surface. For those specific Cu­X systems, adsorbate­adsorbate interac­
tions are considered. Consequently, the most stable configurations of [Cu­X – adsorbate]
are included in Figure 4.1.
In accordance with literature, Cu and the Zn­doped Cu surface show widely different
turnover frequencies (TOF) for methanol formation from CO2 and H2 , with the CuZn sys­
tem having a higher TOF than that of the pure Cu system. Some of the alternative pro­
moters have already been utilised for the methanol synthesis, most notably In2 O3 , which
in itself or in combination with ZrO2 is reported to be active and highly selective for CO2
hydrogenation (101, 102). Alloys with Ga have also been screened extensively as Pd2 Ga
(103–106), Co­Ga (107), Ni­Ga (108–110), and Ni­Fe­Ga (111). Specifically for Cu­based
catalysts, the oxide Ga2 O3 have been shown to improve methanol selectivity and activity

34
 4.2. Getting to Know the Cu­Ge System

(112). While oxides of Ga and In are well­studied in literature, bimetallic or promoted Cu

systems that incorporate some of the theoretically more promising promoters Sn, Pb, Bi
and Ge are less familiar. To the knowledge of the thesis author, only promotion by Sn
has been investigated for thermal catalysis, with that not even being direct methanol syn­
thesis, but rather DME formation and methanol steam reforming (113–115). The theoret­
ically more active system, Cu­Ge, is yet to be seen described in literature in combination
with CO2 or CO hydrogenation to methanol. Only recently, the binary alloy Cu3 Ge was
proposed as an interesting system in a large screening study for the electrochemical re­
duction of CO2 into multicarbon products (116), while PtGe has been investigated for the
citral hydrogenation reaction (117).

Although there are many requirements for a new catalyst for CO2 hydrogenation, the
first step in fundamental research is the development of a system that first and foremost
preferably only produces methanol. Parallel to that, it is crucial to characterize the material
system well enough to elucidate the active site, after which optimization can take place.
The following sections describe the studies of new material systems for the hydrogenation
of CO2 at low pressures and temperatures. They include preliminary results of Cu­X
materials where X is Sn, Sb, Bi and Pb and the in­depth study of the synthesis and activity
of the Cu­Ge material system, the latter focusing on separating the effects of particle sizes
and promoter effects.

4.2 Getting to Know the Cu­Ge System

The initial study and the following optimization of the synthesis route for a Cu­Ge catalyst
was performed in collaboration with Thomas Smitshuysen and student Daniel Bruun as
part of Daniel’s project in the spring of 2022. Thomas and Daniel synthesized the first
range of Cu­Ge catalysts supported on SiO2 , some of which were subsequently tested
for their activity by the author of this thesis. The combination of SEM and XRD for the
characterization of samples, both as­synthesized and after activity testing (referred to as
post­mortem), enable us to identify crystalline phases and the presence (or the lack) of Cu,
Ge and traces of precursors. In­situ XRD assisted in the identification of crystalline phases
during reduction, synthesis gas atmosphere, and calcination in an oxidizing atmosphere,
to learn about a new material system.

Synthesis and catalytic tests. The Cu/Ge/SiO2 catalysts were prepared by incipient
wetness impregnation as introduced in Section 3.1.1, but using an altered IWI method
as described in the following section. For the impregnation, the following chemicals
were used: 99.9999 % germanium(IV)chloride (Alfa Aesar, 10509), 99.99 % Cu(II)nitrate­
pentahydrate (Sigma Aldrich, 467855), 37 w/w % hydrochloric acid. The dried, impreg­
nated powders were loaded in steel reactors with SiC dilutant. Once loaded in the reactor,
they were reduced in a mixture of H2 and Ar at a given temperature, and catalytic tests
were subsequently performed in H2 and CO2 (ratio 4:1) at 10 bar.

35
Chapter 4. New Catalysts for CO2 Hydrogenation

SEM­EDX. Microscopy was performed by Daniel. A small amount of the as­synthesized

Cu/Ge/SiO2 catalyst was loaded on an Al­stub with double­sided carbon tape.

XRD. Similarly to SEM­EDX, small amounts of as­synthesized and post­mortem catalysts

were taken to be analyzed with XRD. They were loaded on sample holders and analyzed
as described in Sections 3.3.1 and 3.3.2.

In­situ XRD. An as­synthesized sample of CuO/GeO2 /SiO2 was loaded into the in­situ
XRD. A total gas flow of 50 ml/min was used to approximately match the gas velocity
used during activity testing. Diffraction scans were acquired before, during and after each
temperature step.

4.2.1 Optimization of synthesis route for an active Cu­Ge catalyst

In the synthesis of supported catalysts, the use of precursors that decompose to N2 , CO2 ,
O2 and NOx when heated are preferred. It is also preferred to use precursors that are
easily soluble in water or dilute acid, and even better if the precursors can be dissolved
in the same solvent. Purchasable germanium precursors are confined to organic salts
(which are most times highly flammable or insoluble in common solvents) and iodine,
sulphur or chloride salts. Lastly, GeO2 is purchasable and more easy to work with. The
oxide is not easily reducible, as evident from the early in­situ XRD studies.

Preliminary studies of the Cu­Ge system include a sequential incipient wetness impregna­
tion to obtain a Cu/GeO2 /SiO2 catalyst. With IWI as your synthesis route, you are always
limited by the following: The precursors should preferably be soluble in the same (mild)
solvents to ensure a homogeneous precursor solution; The amount of solution used in the
impregnation step must match the volume required to fill the pores of the support. The
reasoning behind it being a sequential IWI route is twofold. Firstly, the precursors of Cu
and Ge can not be dissolved in the same solution. Although Cu nitrate is easily soluble in
water, GeCl4 hydrolyzes to form its corresponding oxide which can potentially precipitate
if there is not enough solvent to properly dissolve it. Consequently, the amount of solution
is limiting, and so is the choice of solvent. As described by Chen et al., GeCl4 is soluble
in dilute HCl ­ and with an increase in HCl concentration, the solubility of the chloride salt
increases (118). If the HCl concentration is too low to begin with, the precursor will hy­
drolyze, and the leaching rate of Ge will decrease. Dissolving GeCl4 in concentrated HCl
could possibly serve as a way of bypassing the oxide phase and instead end up with pure
Ge impregnated on the material. Although reported in literature, it was found in this study
that GeCl4 is in fact not soluble in concentrated HCl.

For the study of the Cu­Ge system, a Cu/SiO2 catalyst with a nominal Cu loading of 17
wt% was used as a zero point to observe the possible promoting or deactivating effect of
Ge. In the following section, it is described why SiO2 was chosen as a support material
for this particular study, and how the synthesis of Cu/Ge/SiO2 was optimized to obtain a
catalyst containing both Cu and Ge.

36
 4.2. Getting to Know the Cu­Ge System

SiO2 vs. Al2 O3 : Choice of support for the Cu­Ge system

Although this is not a complete and thorough study of the possible catalyst supports, it is
still worth presenting the experiments that lay the ground for why SiO2 was chosen as a
support for Cu­Ge. Two of the most commonly used support materials have been studied:
γ­Al2 O3 and SiO2 . These metal oxides have very high melting points and are therefore
stable under most operating conditions. With a Tamman temperature of around 987 K
(714 ◦ C) (119), SiO2 is to be used at lower temperatures than γ­Al2 O3 . This is to avoid the
atoms in the crystal lattice becoming more reactive or susceptible to diffusion. This is not
a problem for methanol synthesis over a Cu catalyst since a) the equilibrium of methanol
from CO2 and H2 is exothermic, and the entropy decreases across the reaction, mean­
ing a low temperature is required for a high yield of methanol, and b) Cu­based catalysts
deactivate over time at high temperatures due to the loss of active surface area through
sintering of Cu (67, 93, 94, 120). The subjective of the following is to screen for an appro­
priate support for the preliminary screening study of the Cu­Ge system. Therefore, neither
of the two Cu­based catalysts have been subjected to thorough characterization through­
out their lifetimes. Additionally, support effects on the methanol TOF is a well­studied
subject, and plenty of studies examine the effect of gas composition, pre­treatment, and
temperature on the TOF.

Figure 4.2: Catalytic mass activity of 17 wt% Cu/γ­Al2 O3 and 17 wt% Cu/SiO2 . Reaction
conditions are 150 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 1.4 × 104 h−1 and a WHSV
of 7.4 × 105 Ncm3 · h−1 · g−1
Cu . Uncertainties are estimated by propagation of error (121).

37
Chapter 4. New Catalysts for CO2 Hydrogenation

Samples of Cu/γ­Al2 O3 and Cu/SiO2 , both with nominal Cu loadings of 17 wt% and cal­
cined on a hot­plate set to 200 ◦ C, were prepared by reduction and subsequently tested
for CO2 ­based methanol synthesis, cycling the temperature from 140 ◦ C up to 230 ◦ C or
just below 300 ◦ C, depending on the catalyst. Figure 4.2 shows the amount of methanol
produced in a CO2 /H2 synthesis gas mixture at 10 bar, as quantified by gas chromatogra­
phy, and plotted as a function of the temperature. It is easy to see why Al2 O3 is commonly
used instead of SiO2 : At 230 ◦ C, the mass activity is more than a factor of 10 higher for
Cu supported on γ­Al2 O3 than for Cu supported on SiO2 . At these conditions, there is a
notable support effect on the mass activity, and surely also the TOF (had this been esti­
mated). The relative activities of the two supports (Al2 O3 > SiO2 ) are in good agreement
with several observations in literature (76, 122–126).

The preparation methods for the two catalysts differ by the initial reduction method: Cu/SiO2
was prepared by reduction in 5 % H2 /Ar (220 ◦ C for 3 hours) before being cooled down to
room temperature, whereas Cu/Al2 O3 was reduced at a lower temperature, but the same
gas composition (170 ◦ C for 3 hours). The samples also differ in terms of experimental
history: The Cu/γ­Al2 O3 was subjected to consecutive runs of reduction, activity tests and
regeneration, before being unloaded from the reactor. Despite the extensive sample his­
tory, the activity of methanol did not change significantly over time, as seen in Figure B.16.
A relevant question is whether the two catalysts differ as much in catalytic mass activity as
they potentially differ in active surface area, with this possibly resulting from the different
pre­treatments? Plenty of studies report on the deactivation of the Cu/ZnO/Al2 O3 catalyst
through the sintering of Cu. Studies report that as CO or H2 O is added, Cu­sintering is
accelerated, and a loss of active surface area is observed (67, 93, 94). Common for these
studies is the use of synthesis gas. What we are looking to explain is the effect of reduc­
tion. A study by Clausen et al. perfectly show the effect of reduction temperature on the
Cu crystallite size (120). The results from their in­situ XRD show a dramatic increase in
crystallite size during reduction: Between 100 −170 ◦ C, the crystallite size increases from
9 nm to approximately 13 nm. Increasing the temperature further to 220 ◦ C, the crystallite
has increased another 1.5 nm. Assuming an initial diameter of 9 nm and assuming that
the total catalytic area of a sample is inversely proportional to the radius of the nanoparti­
cle, the initial loss of catalytic surface area would be 30 % at 170 ◦ C and an additional 10
% when the temperature is increased to 220 ◦ C. For both SiO2 ­supported Cu catalysts,
as well as Al2 O3 ­supported ones, the preparation method has been shown to have an
effect on the active surface area (127). Furthermore, it is found that that Cu/Al2 O3 needs
an additional pre­treatment in pure H2 besides the already applied 5 % H2 reduction to
completely reduce CuO to the active Cu phase (125). To answer the question at the be­
ginning of this paragraph, Thrane et al. answers it well by commenting that the difference
in pre­reduction from one metal­support system to another does not cause the observed
large difference in methanol TOF, even though there is some uncertainty linked to the
determination of the Cu surface area with chemisorption methods (125).

38
 4.2. Getting to Know the Cu­Ge System

Even though an Al2 O3 seems like the obvious choice for support, there is the question of
selectivity. What is not shown in Figure 4.2 is that the 17 wt% Cu/γ­Al2 O3 catalyst pro­
duced fairly high levels of dimethyl ether (DME) from methanol dehydration in a CO2 /H2
atmosphere at 10 bar, a trend that is attributed to the presence of acid sites on the γ­Al2 O3
(76). This is not surprising, as γ­Al2 O3 is used as a catalyst by itself for the dehydration
of formic acid (23). The DME is included in the reported mass activity, as discussed in
connection with Figure B.16. Even though Cu/SiO2 have a shockingly bad mass activity
for CO2 hydrogenation, it was chosen as a starting point as it was more selective towards
methanol formation. Looking at the bright side, a bad starting point is a good starting point
to observe a promotional effect.
Besides the question of selectivity, SiO2 has some properties that makes the character­
ization of the metallic phase easier. Since SiO2 is amorphous, it does not show narrow
distinctive Bragg peaks in the 2θ [30­80]° range in an XRD pattern, as illustrated in Fig­
ure B.25, making the analysis of any metal alloys easier.
The successful route to Cu/Ge/SiO2
The preliminary synthesis of Ge­promoted Cu/SiO2 consisted of the following steps. A 17
wt% Cu/SiO2 sample was impregnated with GeCl4 dissolved in dilute HCl. As the boiling
point of GeCl4 is 83 ◦ C, the impregnated sample was only dried at low temperatures to
avoid evaporation of the GeCl4 precursor. Subsequently, the sample was loaded into a
tube furnace to be reduced in H2 /N2 to remove the chlorides. Doing this, the chlorides
are in theory safely removed and there is no risk of HCl forming in the reactor of the
experimental setup when working with reducing gases. The reduced catalyst was loaded
in a steel reactor and tested for its activity without any additional reduction.

Two catalysts prepared this way were tested ­ one of high Ge loading (Ge:Cu ratio of 0.2)
and one of low loading (Ge:Cu ratio of 0.04). Surprisingly, none of the catalysts were
successful in producing methanol from a H2 ­rich synthesis gas mixture. SEM­EDX of
the reduced samples revealed no presence of Ge. Even an additional sample with a Ge
loading of 30 wt% showed no presence of Ge. However, both of the samples analyzed
with SEM­EDX, grains of particles were found to have a large loading of Cl relative to Cu.
SEM images and EDX spectra for two samples containing 20 and 30 wt% are given in
the Appendix in Figures B.30 and B.31, respectively. The SEM images and EDX spectra
shown in the Appendix are representative of the entire sample. Chlorine is notoriously
known for poisoning metal catalysts on oxide supports by altering their dispersion (128),
and by forming volatile Cu chlorides (12, 129). Therefore, it was no surprise to find that
Cu/Ge catalysts synthesized from a GeCl4 precursor showed no activity in terms of CO2
hydrogenation. The Cl­containing precursor is thought to simply have evaporated during
the treatment in the tube furnace, and with a possible insufficient reduction procedure,
traces of HCl used during impregnation with GeCl4 were left to poison the Cu/SiO2 sample.

39
Chapter 4. New Catalysts for CO2 Hydrogenation

Figure 4.3: PXRD pattern of CuO/GeO2 /SiO2 synthesized via the sequential impregnation
method. Patterns of the as­synthesized GeO2 /SiO2 intermediate and a 17 wt% CuO/SiO2 ,
as well as pure CuO, GeO2 and SiO2 references, are pictured for easy comparison. Spec­
tra are shifted on the y­axis. Cu and Ge nominal loadings of 18 wt% and 28 wt%, respec­
tively.

A catalyst containing both Cu and Ge with little to no trace amounts of Cl were obtained
by applying a sequential incipient wetness impregnation method. Contrary to the unsuc­
cessful impregnation method, GeCl4 was impregnated on SiO2 support as the first step.
GeCl4 is found to evaporate in air outside of a glovebox and is therefore difficult to work
with. By wetting the SiO2 powder with water, we aid in ’capturing’ the GeCl4 by forming
the stable GeO2 (and HCl) more quickly than if GeCl4 were to react with the water vapour
in the air. The formed hydrochloric acid was evaporated by heating, and the PXRD of
the resulting white powder is shown in Figure 4.3. Comparing the diffraction pattern with
that of a pure GeO2 reference is evident of a successful first step of the sequential IWI
synthesis. A Cu precursor was subsequently impregnated on the GeO2 /SiO2 catalyst.
PXRD of the resulting powder is shown in Figure 4.3, along with a reference pattern of
a 17 wt% Cu/SiO2 and pure CuO. The sequential impregnation route proves useful in
creating a CuO/GeO2 /SiO2 catalyst with no chlorides present, useful for testing for CO2
hydrogenation.

40
 4.2. Getting to Know the Cu­Ge System

4.2.2 Activity of a CuO/GeO2 /SiO2 catalyst

A sample of CuO/GeO2 /SiO2 was synthesized by Daniel as described in Section 4.2.1

to give nominal loadings of Cu and Ge of 15 wt% and 11 wt%, respectively. Samples
of the same batch were prepared by reduction in 5 % H2 /Ar for either 3 hours at 220 ◦ C
(open circles) or 6 hours at 220 ◦ C (closed squares). Following an activity test, the sample
reduced for a long period of time was once again reduced, this time at 350 ◦ C for 3 hours
(open squares). In Figure 4.4, the activity of the CuO/GeO2 /SiO2 catalyst after a reduction
procedure is shown. It was tested for CO2 ­based methanol synthesis at 10 bar while
cycling the temperature up and down. Only after a second reduction at 350 ◦ C do we gain
a slightly considerable amount of methanol mass activity in the investigated temperature
range. This activity is however far from that of a 17 wt% Cu/SiO2 , shown with red squares.
Rather than promoting the Cu­based catalyst, Ge poisons the active surface. PXRD of
the post­mortem samples reveal crystalline Cu as well as GeO2 , but no crystalline Ge
phase, indicating that the Cu­GeO2 is not nearly as active for CO2 hydrogenation as pure
Cu, and that the reduction conditions applied here are not nearly enough to reduce GeO2 .
Commenting on the precision of the data in the low temperature range, for mass activities
as low as these, uncertainties estimated by propagation of error tend to be very large.
The uncertainties given in Figure 4.4 are more a measure of the error of the experimental
setup rather than that of the mass activity. Despite the large error bars, there is indication
of an activation after reduction at higher temperature.

Figure 4.4: Catalytic mass activity of a Cu/GeO2 /SiO2 catalyst with nominal loadings of
15 wt% Cu and 11 wt% Ge synthesized using the sequential IWI. Reaction conditions
are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, and a WHSV of 2.4 × 105 Ncm3 · h−1 · g−1 Cu .
Uncertainties are estimated by propagation of error (121).

41
Chapter 4. New Catalysts for CO2 Hydrogenation

4.2.3 In­situ XRD study of CuO/GeO2 /SiO2

In order to get insight into the CuO/GeO2 /SiO2 catalyst, especially during reduction of the
present species, in­situ XRD have been employed. Using XRD, the reduction of the Cu
and Ge oxide phases could be followed globally. The in­situ study is divided into three
parts: Two rounds of reduction, activity testing in CO2 /H2 , and at the end a calcination
at high temperatures. A sample of the catalyst shown in Figure 4.3 was loaded into the
in­situ XRD sample holder. This sample is not representative of the Ge­promoted Cu
catalyst we are searching to investigate, but it can give some useful insight into what is
occurring during reduction of the Cu­Ge system.

Figure 4.5: In­situ XRD of the reduction of CuO/GeO2 /SiO2 in a flow of 5% H2 /He at 1
bar. Diffractograms are shown for each temperature step, and they are the last recorded
at each temperature. The diffractograms are corrected for the sample displacement ob­
served at room temperature in relation to the GeO2 reference. Reference patterns: CuO
(ICSD 16025), GeO2 (ICSD 59642), Cu(GeO3 ) (ICSD 25754), Cu (ICSD 235809), Ge
(ICSD 121532), Cu3 Ge (ICSD 53266).

The first reduction cycle was done in 5% H2 /He at 1 bar. The temperature was increased
in 20 ◦ C or 50 ◦ C steps in the temperature range 160 −550 ◦ C. Figure B.1 shows the
diffractograms taken during the reduction, all recorded while the temperature is held con­
stant. During the reduction, multiple phenomena are observed, and the most important
diffractograms are shown in Figure 4.5. What is not shown here is the small shift in reflec­
tions observed at room temperature, which is due to a small displacement between the
sample holder and the X­ray beam path. The magnitude of the shift was determined by
comparing the diffractogram of the calcined sample with that of a GeO2 reference pattern

42
 4.2. Getting to Know the Cu­Ge System

(ICSD 59642). This reference pattern was found to agree perfectly with the ex­situ PXRD
pattern of the calcined sample of Figure 4.3. The error introduced by such a displacement
is described by ∆2θ = 2·s·cos(θ)·sin(θ)
R·sin(ω) , where s is the displacement from the goniometer
axis, determined to be 0.74 mm, and R is the goniometer radius (240 mm). As the scans
are coupled theta­2theta scans, ω is the same as θ. All of the diffractograms presented in
this in­situ study are corrected for this shift.

The shift observed at elevated temperatures is also due to changes in position of the sam­
ple, especially the height, during heating and cooling cycles, and as the sample volume is
reduced during the initial reduction. The shifts at elevated temperatures are not corrected
for as the purpose of this study is to observe phase changes.

Apparent from the diffractograms in Figure 4.5 is that remains of precursors (present at
low angles) disappear at a temperature of 200 ◦ C. All CuO is reduced to Cu by a tem­
perature of 260 ◦ C. This is a higher temperature than expected, but it might be that the
reduction potential of the gas is not high enough, or due to the non­ideal flow of gases
through the sample holder. It is unfortunately also highly likely that the sample holder
itself and the reaction chamber surrounding the sample holder is subject to reduction as
was found to be the case for subsequent experiments involving this experimental setup.
While Cu is fully reduced at 260 ◦ C, GeO2 is not completely reduced until reaching a tem­
perature above 450 ◦ C. With the reduction of CuO and GeO2 , an intermediate Cu(GeO3 )
phase briefly appears before the stable binary Cu3 Ge phase appears. The analysis of
phases is in general difficult in the region between 40 and 50 ◦ as the peaks of the Cu,
Ge and Cu3 Ge phases overlap in the region, and it is made even more difficult with the
shift happening at high temperatures. It is more easy to follow the formation of the Cu3 Ge
phas in the region around a diffraction angle of 60 ◦ . At 400 ◦ C, a Bragg peak starts to
appear at approximately 59 ◦ , corresponding to the Cu3 Ge phase. At 550 ◦ , GeO2 is not
found, and all Bragg peaks are related to Cu, Ge and Cu3 Ge.

A second treatment in 5% H2 /He at 1 bar, this time up to a temperature of 700 ◦ C, is car­

ried out after testing the sample for CO2 ­based methanol synthesis. The diffractograms of
the calcined sample, along with those of the reduced sample after 550 ◦ C and 700 ◦ C, all
recorded at room temperature, are shown in Figure 4.6. The diffractograms are shifted
upwards for easy comparison, and reference lines for the most important phases are
given. The phases does not change much. Both pure Cu and Ge phases are present
after another reduction, albeit the intensities of the corresponding Bragg peaks have de­
creased and the Cu3 Ge phase experience an equivalent increase. The diffractograms
taken after each of the reductions are shown in Figure B.3 and are not shifted for more
easy comparison of absolute intensities.

43
Chapter 4. New Catalysts for CO2 Hydrogenation

Figure 4.6: PXRD of CuO/GeO2 /SiO2 after reduction. Diffractograms are recorded at
room temperature before reduction (i.e. calcined state), and after reduction at 550 ◦ C and
700 ◦ C. The diffractograms are corrected for the sample displacement observed at room
temperature in relation to the GeO2 reference. Reference patterns: CuO (ICSD 16025),
GeO2 (ICSD 59642), Cu (ICSD 235809), Ge (ICSD 121532), and Cu3 Ge (ICSD 53266).

After several rounds of reduction and activity testing, the sample was calcined at high
temperatures, but using synthetic air (20 % O2 /N2 ) instead of ambient air. The full col­
lection of diffractograms taken at each temperature steps are given in the Appendix in
Figure B.4. While Cu is fully oxidized to bulk CuO at approximately 300 ◦ C, bulk GeO2
as seen in the calcined sample is not visible during heated calcination. Not even after
the removal of the sample from the in­situ holder does crystalline bulk GeO2 form. This
is apparent from the ex­situ PXRD of the sample in Figure 4.7. The calcined state of Ge
is thought to be a mixture of cubic Ge, quartz GeO2 and amorphous GeO2 and shows
how fresh and air­exposed GeO2 (and in general GeOx ) can have very different crystal
structure (130). If the electronic structure of Ge/GeO2 was to be further characterized,
X­ray absorption spectroscopy would have to be used.

The most important observation from the in­situ XRD study of a CuO/GeO2 /SiO2 catalyst
is the behaviour of GeO2 during reducing conditions. It requires much higher temperatures
in a 5 % H2 gas mixture to reduce GeO2 , while CuO is more readily reduced. This helps
to explain the higher reduction temperature needed to increase the activity of the Ge­
promoted Cu/SiO2 catalyst discussed in the previous sections.

44
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

Figure 4.7: Ex­situ PXRD of the calcined Cu/Ge/SiO2 sample after removal from in­situ
XRD. The in­situ XRD patterns is the latest scan, taken after a week of being in the in­situ
sample holder, and is the same as that of Figure B.4. The in­situ diffractogram is cor­
rected for the sample displacement observed at room temperature in relation to the GeO2
reference. References: Ge (ICSD 121532), CuO (ICSD 16025), GeO2 (ICSD 59642).

4.3 Study of Ge Promotion in Cu­Ge/SiO2

In the following, a method of synthesising Cu­Ge catalyst with varying molar ratios of
Cu and Ge will be described. The method was developed in collaboration with Jette
Mathiesen. The Cu­Ge catalysts are named after the theoretical nominal molar fractions
of Cu and Ge, with x being the molar fraction of Ge: Cu(1­x) Gex .

When preparing Cu­Ge catalysts with varying amounts of Ge promoter, the phase diagram
of the binary system comes in handy for later reference. The solution phases, including
liquid, face­centered cubic (FCC), hexagonal close­packed (HCP) and diamond (Ge) were
investigated by Wang et al. (131). They collected their research, which they prove to fit
well with experimental values, in the binary phase diagram in Figure 4.8. The work of
Wang et al. (131), as well as the studies referenced in their paper, only concerns the three
intermetallic compounds ε­Cu0.765 Ge0.235 , θ­Cu0.735 Ge0.265 and η­Cu0.75 Ge0.25 , which all
have larger Ge:Cu ratios than what we want for our study of Cu­Ge catalysts. For low
Ge mole fractions, a binary Cu­Ge system will likely form, and is assumed to be mixed
phases of either the η­phase and HCP phase, or the HCP phase and the FCC phase.
From a crystallographic point of view, Cu(1­x) Gex particles with low Ge mole fractions will
resemble a Cu FCC phase with randomly distributed Ge atoms within the crystal structure
or on the surface of Cu particles.

45
Chapter 4. New Catalysts for CO2 Hydrogenation

Figure 4.8: Phase diagram of the Cu–Ge binary system. Reprinted with permission from
(131). Copyright Elsevier B.V., 2010.

Synthesis and catalytic tests. The Cu(1­x) Gex /SiO2 catalysts were prepared as de­
scribed in Section 3.1.2. For the synthesis of Cu and Cu­Ge nanoparticles, the follow­
ing chemicals were used: 97 % germanium(II)bromide (Sigma Aldrich, 572659), 98 %
copper(II) acetate (Sigma Aldrich), SiO2 (44740, Alfa Aesar). The dried, impregnated
powders were loaded in glass­coated steel reactors with SiC dilutant. Once loaded in the
reactor, catalytic tests were performed in H2 and CO2 (ratio 4:1) with trace amounts of Ar
at 10 bar total pressure.
As a reference, the industrial­grade catalyst Cu/ZnO/Al2 O3 /MgO (45776, Alfa Aesar) has
been used. It will be referred to as CZA.

XRD. Small amounts of as­synthesized and post­mortem Cu(1­x) Gex /SiO2 powders were
taken to be analyzed with XRD. They were loaded on sample holders and analyzed as
described in Sections 3.3.1 and 3.3.2.

SEM. Scanning electron microscopy was performed by PhD Filippo Romeggio (DTU Sur­
fCat). Droplets of the NP suspension were drop­casted on a mechanically cleaned glassy
carbon stub. Powder samples (as­synthesized as well as post­mortem) were loaded on
an Al­stub with double­sided carbon tape. Unless otherwise stated, acquisition and anal­
ysis was done using the AZtec Software from Oxford Instruments.

X­ray total scattering. Samples of as­synthesized and post­mortem Cu(1­x) Gex /SiO2
were loaded into thin kapton tubes. Experiments were conducted at Deutsches Electronen­
Synchrotronen (DESY) in Germany. The sample­to­detector distance and instrumental
broadening was calibrated using a LaB6 standard. Total scattering data were collected in
transmission geometry using a Varex 4343CT detector with a pixel size of 150×150 µm
in the RA­PDF setup mode (90), where a wavelength of 0.207 Å was used. To use for
the step of background subtraction, an empty sample holder, as well as sample holders
with SiO2 and SiC, was measured in the RA­PDF setup. The steps of correction, nor­

46
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

malization and integration of the 2D data, as well as the steps of Fourier transforming the
data to obtain the experimental PDFs and refinement of these, were carried out by Jette
Mathiesen.

TEM. Transmission electron microscopy was performed by PhD Stefan Kei Akazawa
(DTU VISION). For TEM, a small amount of sample was loaded on a lacey­carbon Au
TEM grid 300 mesh. By dropping the TEM grid into the container with powder sample,
gently shaking and tapping the grid lightly after removal from sample container, small
grains of sample are left on the grid. Microscopy was performed on a FEI Titan E­TEM,
and images were acquired with a tension of 300 keV and a dose rate of approximately
50 e− · Å−2 · s−1 .

4.3.1 Synthesis of Cu(1­x) Gex catalysts

The morphology of the nanoparticles formed via a hot­injection wet­synthesis approach
depend on the synthesis parameters. The parameters include, but are not limited to, the
solvent used for the synthesis, the strength of the reducing agent, the oxidation state of
the metal precursor, and the presence of surfactants and stabilizers. Introducing more
than one metal precursor naturally also introduces more complexity to the synthesis. The
road to an optimal set of parameters for synthesising Cu­Ge nanoparticles is an ongoing
project. In the following, the NaBH4 ­based synthesis was used to produce a series of Cu­
Ge nanoparticles with varying ratios of Ge:Cu and varying crystallite sizes. A oleylamine­
based synthesis approach was also shortly investigated, but was not found successful in
producing any distinctive Cu­ or Cu­Ge phases. Preliminary results of this synthesis route
is found in Appendix B.2.
NaBH4 ­based synthesis
The synthesis route of the NaBH4 ­based synthesis follows the route described in sec­
tion 3.1.2. The organic solvent is triethylene glycol (TrEG, abbrev.). Interestingly, a color­
change is visible throughout every step of the synthesis, indicating the formation of metal
ion complexes at every step. With the addition of a reducing agent, the solution immedi­
ately changes to black in color, indicating the formation of nanoparticles.
This synthesis approach has two sides to it: One is the temperature of the precursor­
mixture, another is the injection of the reducing agent. The first part concerns the reducing
potential of the solvent as a function of reaction temperature. Normally, high temperatures
are needed for metals to reduce in the case a reducing agent is absent. In this case, TrEG
also acts as a reducing agent (132) at a reaction temperature of <170 ◦ C, meaning we
eliminate the need for high reduction temperatures. The second part of quickly introduc­
ing the reducing agent ensures a fast reduction of the metal precursors to form metallic
seeds. An advantage of this method is that an instant nucleation of the metal ions is en­
sured by injecting the reducing agent. The formed seeds can then undergo growth, where
the resulting size and shape is determined by the ability of either the solvent or additional
surfactants to direct the seed growth in a controlled manner.

47
Chapter 4. New Catalysts for CO2 Hydrogenation

Whether all of the metal precursor precipitates as nanoparticles naturally depends on

the solubility of the precursors and whether all of it nucleates. When washing the formed
nanoparticles, the solvent is colorless, meaning no precursor was left in solution but rather
had nucleated. Working with the nanoparticles throughout this study, we assume a full
conversion of the metal precursors into metal nanoparticles.

Figure 4.9 shows the diffraction patters of four different Cu(1­x) Gex nanoparticles with
nominal Ge mole fractions of x=0, 0.15, 0.33 and 0.46. Drops of each of the NP suspen­
sions were deposited on a glass plate and diffractograms were obtained after the solvent
had dried. The upper, light blue, diffractogram is pure Cu, where the dominating crystal
phase corresponds well to a Cu­FCC structure (ICSD: 235809, lattice parameter 3.6 Å).
Increasing the amount of Ge to a mole fraction of 0.18 should according to the phase dia­
gram of Figure 4.8 result in a mixture of the η­Cu0.75 Ge0.25 ­phase and a HCP phase. In this
case, the dominating Bragg peak at 43.4° and a less intense, broad peak at approximately
45.6°, indicate the presence of two crystal phases. Again, the Cu­FCC structure (ICSD:
235809) fits well with the first peak, and one of two Cux Gey crystal structures ((Cu5 Ge)0.33
w. ICSD 627677, Cu1.7 Ge0.3 w. ICSD 627675) fits well with the peak at higher angles. It
becomes more apparent, as the mole fraction of Ge increases going from x=0.18 to the
Cu0.52 Ge0.49 nanoparticles all the way to the Cu0.14 Ge0.86 nanoparticles, that the Cux Gey
crystal structures start to dominate over the Cu­FCC phase.

Figure 4.9: XRD patterns of as­synthesized Cu and Cu(1­x) Gex nanoparticles with vary­
ing amounts of Ge, each spectrum shifted on the y­axis. Reference spectra: Cu (ICSD
coll. code 235809), Ge (ICSD coll. code 212532), Cu3 Ge (ICSD coll. code 53266) and
(Cu5 Ge)0.33 (ICSD coll. code 627677).

At the highest loading of Ge, a peak at 45.2° dominate over the one corresponding to the
Cu­FCC crystal structure. The most intense peak fits well with a Cu3 Ge crystal structure
of the Cu3 Ti type (ICSD 53266). Presence of shoulder peaks to the right of peaks at

48
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

40° and 45.2° indicate the presence of phases corresponding to a (Cu5 Ge)0.33 crystal or
similar. Figures B.6 and B.7 in Appendix B.3 show the Cu0.54 Ge0.46 nanoparticles with
the mentioned reference patterns. Based on the initial stoichiometry in the synthesis, we
assume a 50:50 alloy to form. Besides the binary Cu3 Ge and (Cu5 Ge)0.33 phases, peaks
at 43.4° and 50.4° indicate the presence of a pristine Cu phase. One would expect a pure
Ge phase, but as it is clearly observed in the reference pattern of Ge in Figure 4.9, main
peaks of crystalline Ge is masked out by the presence of both a broad background and
other crystalline phases in the spectra of the unsupported nanoparticles. Nevertheless,
we cannot completely rule out the presence of small Ge crystallites, which would give
rise to a broad diffraction profile. The presence of Cu and Ge was confirmed with SEM­
EDX for (1)­Cu0.85 Ge0.15 . Measurements on different areas of the glassy carbon stub
revealed the presence of both Cu and Ge, and images of areas with no particles were
taken to make sure that no Cu or Ge signals were coming from the glassy carbon stub
itself. Referring to the SEM images in Figure 4.10, we find particles in the approximate
size range 20 −250 nm, with some lone spherical particles and others that have possibly
agglomerated into larger particles, making it difficult to estimate the exact sizes of the
individual particles. The low end of the size range agrees with the findings from XRD.
Using the Scherrer equation and the Cu(111) peak of Figure 4.9, the average size of Cu
crystallites of the unsupported (1)­Cu0.85 Ge0.15 sample is 21 nm.

Figure 4.10: Ex­situ SEM of unsupported (1)­Cu0.85 Ge0.15 nanoparticles on a glassy car­
bon stub. Images were acquired by PhD Filippo Romeggio. The microscope was oper­
ated at 20 kV and with a working distance of 12.9 mm.

Cu(1­x) Gex on a SiO2 support

A total of 40 µL of the unsupported nanoparticles was removed from each of the NP
suspensions before supporting the nanoparticles on SiO2 . This amount corresponds to
less than 1 % of the total NP suspension ­ a loss that, if considered, results in a change of
0.03 wt%. The magnitude of the error this difference can result in is considerably smaller
than the error related to the quantification of products during CO2 hydrogenation.
Once the as­synthesized nanoparticles are supported on SiO2 , the powder is grinded to
ensure a uniform powder and to avoid the presence of any large grains. The powders
are subsequently analyzed with PXRD. Figure 4.11a show Cu0.67 Ge0.33 as­synthesized

49
Chapter 4. New Catalysts for CO2 Hydrogenation

and as supported on SiO2 . Firstly, we notice how the crystal phases are maintained after
working further with the synthesized nanoparticles. As is seen in Figure 4.11b, the same
is true for the four samples presented in Figure 4.9.

(a) XRD of Cu0.67 Ge0.33 . As­synthesized (blue) (b) PXRD patterns of Cu(1 − x)Gex /SiO2 , each
and supported on SiO2 (black). spectrum shifted on the y­axis

Figure 4.11: XRD patterns of SiO2 ­supported Cu(1­x) Gex nanoparticles. Reference spec­
tra: Cu (ICSD 235809), Cu3 Ge (ICSD 53266), (Cu5 Ge)0.33 (ICSD 627677), and Cu1.7 Ge0.3
(ICSD 627675).

A question arising is whether the full amount of synthesized nanoparticles have been
successfully deposited on the support. Naturally, XRD of two different samples cannot
answer this. Comparing the diffractograms simply serve as a way of concluding on the
success of the synthesis method. Microscopy coupled with EDX, or the more exact ICP­
OES method, would be preferred. When estimating the metal loading of the catalysts
for future use, it was assumed that no significant amount of Cu or Cu­Ge nanoparticles
was left on the glass while heating the mixed suspension solution. This will naturally
introduce quite the error if the opposite is true, but for now it is a necessary assumption
to make for when we need to normalize the activity of the individual samples during CO2
hydrogenation.
Proof of concept with X­ray TS PDF
Figure 4.12 shows the experimental PDF of as­synthesized Cu0.67 Ge0.33 /SiO2 in black,
plotted as the reduced pair distribution function, G(r). The sharp and extended oscil­
lations in real space indicate, in accordance with PXRD of the same sample in Fig­
ure 4.11a, a crystalline phase. By a simple model­free analysis of the experimental PDF
of Cu0.67 Ge0.33 /SiO2 , it is clearly seen how the experimental PDF (in black) cannot solely
be described by a FCC Cu structure (in red). Peak position and broadening coincides
with the presence of Ge in either a Cu3 Ge phase (in blue), or another FCC­like Cux−y Gex
phase. Comparing a 1­phase refinement (Figure 4.13a), where only FCC Cu is used, to
a 2­phase refinement with Cu3 Ge and Cu (Figure 4.13b), the presence of Ge is further
illustrated. In the 1­phase fit, there is still significant structural information in the differ­

50
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

ence curve (i.e. the difference between the refined model and the experimental data) that
the FCC Cu model does not account for. This is hinted by the non­periodic oscillations.
Including a second phase, namely a Cu3 Ge phase, we obtain an improved agreement
between the model and the experimental data. This is clearly displayed by the decrease
in Rw value, which indicates the quality of a fit. The respective values are given in Fig­
ures 4.13a and 4.13b as well as in Table B.3. While still not a perfect fit, the 2­phase
refinement manages to model the peak positions, as well as the absolute and relative
peak intensities, better than the 1­phase refinement. For this specific data set, the quality
of the data was not sufficient to estimate crystallite sizes. Instead, it worked as a proof of
concept that demonstrates how PDF refinement is a complementary method to in­house
PXRD.

Figure 4.12: Experimental PDF (black) of as­synthesized Cu0.67 Ge0.33 /SiO2 , and the cal­
culated PDF’s of FCC Cu (red) and Cu3 Ge (blue) models.

(a) 1­phase: FCC Cu. (b) 2­phase: FCC Cu, Cu3 Ge.

Figure 4.13: (a) 1­phase and (b) 2­phase refinement of as­synthesized Cu0.67 Ge0.33 /SiO2 .
The green line is the difference between the model (blue) and experimental data (black).

51
Chapter 4. New Catalysts for CO2 Hydrogenation

4.3.2 CO2 hydrogenation over Cu­based catalysts: Setting the stage

Serving as an introduction to the activity of Cu­based catalysts for CO2 hydrogenation, as
well as serving as a sanity check of the synthesis discussed in the previous section, we
compare the activity of two differently prepared Cu catalysts. In Figure 4.15, the catalytic
activity per mass of metals is shown for two Cu­based catalysts. The black data points
show the catalytic mass activity for a 17 wt% Cu/SiO2 catalyst synthesized via IWI and is
the same data as presented in Figure 4.2. The blue data points show the mass activity of
the 4.3 wt% Cu/SiO2 catalyst synthesized via the hot­injection method. Experiments with
similar values of WHSV are represented in this figure, meaning the experiments are not
necessarily carried out using the same amount of active material or total gas flow over
the catalyst. The reactions conditions for the 17 wt% Cu/SiO2 are written in the figure text
of Figure 4.2. During testing of the 4.3 wt% Cu/SiO2 , a total gas flow of 50 ml/min was
used, meaning this experiment was not carried out using the same GHSV, but the two
experiments have similar total gas flow per mass of active metal. The effect of GHSV and
WHSV was shortly investigated, and the data is shown in Appendix B.6. For a 17 wt%
Cu/SiO2 prepared via IWI, a larger total gas flow (meaning a larger WHSV and GHSV)
yields a higher mass activity for methanol.

Figure 4.14: Mass activity of methanol over SiO2 ­supported Cu catalysts.

Figure 4.15: Catalytic mass activity of 4.3 wt% Cu/SiO2 synthesized via the method de­
scribed in Section 3.1.2 and 17 wt% Cu/SiO2 prepared by IWI. Reaction conditions are
CO2 /H2 in 1:4 at 10 bar, and a WHSV of approximately 8 × 105 Ncm3 · h−1 · g−1
Cu .

From the catalytic mass activity of methanol in Figure 4.15 we see that there is a trend of
higher activity for a catalyst with a lower weight loading, despite having the same amount
of active material (in this case, Cu) in the reactor during testing. PXRD of the post­mortem
samples reveal very different sizes of Cu crystallites, with the high­loading Cu catalyst
having very large crystallites (on average 91 nm) compared to the low­loading with an

52
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

average Cu crystallite size of 33 nm. This trend of lower weight loading producing smaller
nanoparticles is well established (133), and the higher activity for CO2 hydrogenation on
catalysts with smaller nanoparticles can in part be explained by a larger total surface
area, but also by a larger quantity of highly reactive under­coordinated sites (edges and
corners). Comparing the selectivity of the low­loading Cu­based catalyst to the high­
loading one, smaller nanoparticles display a greater selectivity towards methanol as a
product compared to larger nanoparticles. The selectivity is plotted alongside the mass
activity for a small temperature range in Figure B.18. In the work of Barberis et al., the
influence of Cu particle size on the product selectivity is explained as a lower fraction of
sites that favour CO formation on the smaller nanoparticles (134). In support of this, both
CO2 hydrogenation to methanol and the rWGS reaction have been shown to be structure
sensitive on Cu surfaces (28, 29, 34, 40, 135).

Under reaction conditions, the high­loading catalyst show no increase in mass activity
over a temperature cycle. But for the low­loading catalyst, the mass activity of methanol
increases during the first temperature cycle. In Figure 4.16a, the mass activity of the
low­loading (2)­Cu2 /SiO2 catalyst during the first and last temperature cycles of a total
of four cycles i shown. After the first cycle, the very normal trend of increasing activity
as a function of temperature is observed: The production of methanol increases slightly
during the first temperature cycle and stabilizes over time during the subsequent cycles
in CO2 /H2 . While the activity for methanol increases as a function of temperature, CO
dominates as the main product at higher temperatures (Figure 4.16b), why the methanol
selectivity never reaches more than approximately 60 % when at its highest.

(a) Mass activity of methanol during first and last (b) Mass activities of dominant products during
temperature cycles. CO2 hydrogenation.

Figure 4.16: (a) The effect of cycling on the mass activity and (b) the product distribu­
tion of last temperature cycle, of the 4.3 wt% (2)­Cu1 /SiO2 catalyst. Reaction condi­
tions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV
of 8 × 105 Ncm3 · h−1 · g−1
Cu .

53
Chapter 4. New Catalysts for CO2 Hydrogenation

Evidently, the yield of methanol is limited by the production of unwanted side products.
The rWGS reaction seem to have the biggest effect on the selectivity for all of the Cu­
based catalysts investigated here. Besides the rWGS reaction (Equation (2.5)), other
reactions can become limiting factors at high enough temperatures where CO or CO2 can
dissociate. The amounts of CH4 and CH3 CH2 OH produced at high temperatures were
almost always 0.001 % or lower, much lower than the produced amount of methanol, and
believed to be a product of reactions involving CO/CO2 dissociation rather than contam­
ination of Ni, which is a well­known methanation catalyst (136, 137). The formation of
ethanol in a methanol­rich syngas mixture, as well as CO methanation, are both well­
known phenomena on Cu­based catalysts (138, 139).

PXRD of the as­synthesized (2)­Cu1 /SiO2 catalyst, as shown in Figure 4.17b, indicate no
crystalline oxide phases before being subjected to reaction conditions. This is the case
for all Cu­ and binary Cu­Ge catalysts in the study, and so the catalysts have not been
pre­treated under reducing conditions. The activation seen during the first temperature
cycle in Figure 4.16a is also common for all catalysts. The apparent activation might be
due to the removal of a passivating oxide layer, or due to a restructuring into a metallic
phase that is more active for CO2 hydrogenation. The latter have not been investigated
with in­situ methods and is only a theory. Also common for all catalysts investigated here
is that a signal corresponding to ethanol is high during the first GC measurements at the
start of an experiment. Even after drying of the catalyst in a pre­heated oven, there are
still remains of the solvent used during synthesis of the catalysts.

4.3.3 Effect of size and composition in the Cu­Ge System

The question of size effect and support interactions on SiO2 ­supported Cu have been
discussed thoroughly in the literature, but generally for smaller metal particles and larger
size ranges than those investigated here, and spanning different SiO2 supports and meth­
ods of synthesis and pre­treatments (52, 59, 127, 140, 141). It is thought that for sizes
larger than 2 −3 nm, electronic particle size effects are thought not to play a significant role
(142–145). For Cu particles larger than that, it is then a question of the size­dependent
interactions with the support, as well as the structure sensitivity of the hydrogenation of
CO2 to methanol. The structure sensitivity originates from the fact that some facets are
more selective than others for a certain reaction. This will in turn depend on the size of
the nanoparticle: Edges and corners are more frequent with smaller nanoparticles, as
illustrated by reference (146) in Figure 4.25.

In an effort to try and separate the effects of particle size and metal composition, the
study of the Cu­Ge system was approached from two angles. The first is the preparation
of differently sized nanoparticles within the same Cu:Ge ratio, which is discussed in part
I. The second approach is the synthesis of catalysts with different molar ratios of Cu and
Ge, which is discussed in part II. An attempt at combining the two effects is discussed in
part III.

54
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

Part I: Size effects

Two batches of Cu nanoparticles were prepared via the hot­injection synthesis route in
such a manner that differently sized Cu crystallites were obtained. More specifically, the
synthesis was stopped immediately after injection of the reducing agent instead of con­
tinuing to heat the solution while stirring. By stopping the synthesis before the incubation
period, smaller crystallites were obtained. XRD of the resulting nanoparticles, yet to be
supported on SiO2 , are shown in Figure B.8. Fitting the diffractograms to a Lorentzian
distribution yields estimates of the Cu crystallites of 71 nm and 20 nm for (1)­Cu1 and (2)­
Cu1 , respectively. These size estimates are not necessarily relevant for the further study,
but it is worth noting that the size remains approximately the same after the nanoparticles
have been supported on SiO2 : The supported catalysts have approximate Cu crystallites
of 67 nm and 19 nm, as estimated from the black ’pre­’ diffractograms in Figures 4.17a
and 4.17b.

(a) 4.2 wt% (1)­Cu1 /SiO2 catalyst. (b) 4.3 wt% (2)­Cu1 /SiO2 catalyst.

Figure 4.17: PXRD patterns of two Cu/SiO2 catalysts, pre­ and post­ testing of activity.
Reference spectra: Cu (ICSD 235809), SiC dilutant from test reactor.

Figure 4.18a show the mass­normalized activity of methanol of the Cu/SiO2 catalysts of
different Cu crystallite sizes. The data for (2)­Cu1 is the same as shown in Figure 4.16.
(1)­Cu1 was tested under reaction conditions for two consecutive temperature cycles up
to a temperature of 230 ◦ C, while (2)­Cu1 was tested for a longer time and at higher tem­
peratures. When normalizing with the amount of active metal to obtain the mass activ­
ity, the two catalysts perform similarly in the temperature range 140 −230 ◦ C. Despite
a high temperature of 300 ◦ C, the methanol yield does not decrease significantly for the
(2)­Cu1 /SiO2 catalyst. This is despite the production of CO and methane as seen in Fig­
ure 4.16b and the visible increase in Cu crystallite from 19 nm to 33 nm as determined
from Figure 4.17b. Assuming that the crystallites are a measure of the Cu nanoparticle
diameter, this would correspond to a loss of approximately 40 % in catalytic surface area.
The production of methanol is likely not high enough for the system to reach and be lim­
ited by the equilibrium. It is likely that the smaller, and possibly more active, Cu particles
on the (2)­Cu1 /SiO2 catalyst are sacrificed as larger Cu particles grow during the reaction

55
Chapter 4. New Catalysts for CO2 Hydrogenation

conditions. If the catalysts investigated in Figure 4.18a behave the same when subjected
to reaction conditions, this behaviour is supported by the SEM images of (1)­Cu1 /SiO2
in Figure 4.19. It is possible that the large group of particles have agglomerated during
reactions conditions, which naturally means a smaller active surface area.

(a) Methanol mass activity. (b) Surface­averaged mass activity.

Figure 4.18: (a) Catalytic mass activity of methanol during multiple temperature cycles
and (b) the mass activity (left scale) and surface area normalized activity (right scale)
during the last temperature cycle, of (1)­Cu/SiO2 (black) and (2)­Cu/SiO2 (blue). Reaction
conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV
of 8 × 105 Ncm3 · h−1 · g−1Cu . The normalized rate in subfigure (b) is based on the crystallite
size estimated from PXRD of post­mortem samples (from Figure 4.17). Note the two
different scales in this subfigure.

Figure 4.19: Ex­situ SEM of post­mortem (1)­Cu1 /SiO2 on carbon tape. Images were
acquired by PhD Filippo Romeggio. The microscope was operated at 5 kV and with a
working distance of 4.2 mm.

The question is then whether we expect the two catalysts of different mean particle sizes
to perform similarly? And consequently, whether the activity should be normalized using
the amount of active material or using the surface area of the suspected active material.
Whenever activity data is presented, the question of normalization arises. As perfectly
illustrated by the author of reference (37), the activity of the various catalysts designed

56
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

for CO2 hydrogenation presented in literature can be compared in many ways, some
potentially better than others. Analysis of the performance of the catalysts can come with
a large bias if the normalization is not done correctly. Common practice in the literature is
to normalize with the amount of what is thought to be the active material, or the full weight
of the catalyst, support included, to have what is commonly referred to as a ’mass activity’.
The normalization using surface area is similar to the turnover frequency (TOF, abbrev.),
the latter being the most fundamental metric for catalytic activity. Determining the TOF
implies determining the number of catalytic sites present on the sample during reaction
conditions. This is most easily done using a chemisorption experiment that probes the
active surface. In this case, great caution should be taken, especially when using SiO2 as
a support, as the metal­support interaction can bias the result. In the work by Chatterjee
et al., a deviation was found between the area probed by chemisorption methods for a
range of Cu particle sizes between (2.5 −8 nm) on different silica supports (127). This
indicates that there are surface sites that are probed differently depending on the surface
reaction, leading to the conclusion that silica interacts with a large part of the Cu particles.
When chemisorption measurements of the number of surface atoms are not available,
the fraction of surface atoms available to a gas atmosphere can be deduced using the
relations between the dispersion and the specific surface area or the mean particle size.
This naturally implies that we must ignore any possible strong interaction between the
active metal particles and the support. It is preferred to use a surface area based on
a size distribution from a microscopic study of the particles. When not possible, XRD
line broadening analysis proves somewhat useful in that it provides a volume­weighted
average size of the crystallites (147), given that sample broadening is separated properly
from instrumental broadening.

The surface area normalized activity discussed in the following sections is based on the
copper­time yield (CTY, abbrev.), expressed in molMeOH · molCu −1 · hour−1 , and the num­
ber of copper surface atoms, according to the equation: CTY/DCu . Here, DCu is the dis­
persion of surface Cu atoms (i.e. the fraction of surface Cu atoms to the total amount
of Cu atoms). The use of dispersion to normalize activity data and the calculation of is
inspired by the work of Barberis et al. (134), and the calculation based on reference
(148). For the calculation of the fraction of Cu surface atoms a series of assumptions
have been made: Firstly, that the particles are spherical and their surfaces fully acces­
sible. Secondly, that the crystallite size extracted from XRD is directly proportional and
representative of the nanoparticle diameter. And thirdly, that the proportions of the three
low­index planes (111), (100) and (110) of the polycrystalline surface of the FCC metal are
equal. Using published crystal data, the number of atoms per unit area in each of these
planes, as well as the mean number of atoms, can be calculated for an FCC structure with
a lattice parameter of a=3.597 Å (149). With the relations given in reference (148), the
dispersion is calculated as given in Equation (4.1). vCu is the volume occupied by a Cu
atom in the bulk, and aCu is the surface area occupied by a Cu atom on a polycrystalline
surface. For Cu, these values are vCu = 11.83 Å3 and aCu = 6.85 Å2 .

57
Chapter 4. New Catalysts for CO2 Hydrogenation

vCu /aCu
DCu = 6 · (4.1)
dCu

Table 4.1 shows the structural properties of the Cu(1­x) Gex /SiO2 catalysts. Besides the
nominal Cu loading, it gives the Cu crystallite size of the post­mortem catalysts and the Cu
dispersion based on this crystallite size. As the crystallite size might be subject to change
throughout the activity test, the sizes of the post­mortem samples are considered to be the
best representatives of the state of the catalysts during the last temperature cycle. The
sample history is therefore important to mention in connection with the normalized activity
data, as additional cycles of reduction or high­temperature experiments might cause a
further increase in Cu crystallite size.

Normalizing the catalytic mass activity of (2)­Cu1 /SiO2 and (1)­Cu1 /SiO2 given in Fig­
ure 4.18a with the measures for Cu dispersion of Table 4.1 gives us a new picture.
As the crystallite sizes of the post­mortem samples are representative of the last tem­
perature cycle, Figure 4.18b show only the activity of the last temperature ramp in the
range 140 −300 ◦ C. Where the mass activities before were quite similar (left scale, closed
squares), the surface area normalized activity between the two catalysts are now signifi­
cantly more different (right scale, open squares). This is naturally due to the fact that the
fraction of surface to bulk atoms is larger in smaller crystallites compared to large ones,
where most of the mass is contained within the particle instead of outside on the surface.
Assuming the accurate way of normalizing activity data is based on the available surface
atoms, the (2)­Cu1 /SiO2 catalyst seems less active per surface site than the (1)­Cu1 /SiO2
catalyst.

To partly conclude on the size effect on SiO2 ­supported Cu particles: The effect of the
size of the nanoparticle is still a matter of debate, and we should be careful in normalizing
seemingly identical mass activities with the number of surface atoms. The following quote
perfectly illustrates why complementary methods are needed for determining the surface
area responsible for the activity: “Care should be taken not to overlook the presence of a
few large particles which will contribute little to the total active surface but may contribute
significantly to the catalyst active mass and thus lead to overestimated specific surface
areas and underestimated turnover frequencies.” (148).
Part II: Promotional effect of Ge
With the effect of particle size on the activity for CO2 hydrogenation still open for debate,
we turn to the second part of this study: The effort to separate size­ and promotion­
effects. A series of SiO2 ­supported Cu(1­x) Gex catalysts have been prepared with varying
amounts of Ge and a variance in Cu crystallite sizes, for which the properties relevant for
the subsequent section are given in Table 4.1.

58
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

(a) (1)/(2) ­ Cu0.85 Ge0.15 /SiO2 . (b) (1)/(2) ­ Cu0.94 Ge0.06 /SiO2 .

Figure 4.20: PXRD patterns of as­synthesized Cu1­x Gex /SiO2 catalysts. Reference spec­
tra: Cu (ICSD 235809), (Cu5 Ge)0.33 (ICSD 627677), Cu1.7 Ge0.3 (ICSD 627675).

The two Cu0.85 Ge0.15 /SiO2 catalysts with a small variance in Cu crystallite size have
been investigated for their activity and selectivity under conditions similar as used in Sec­
tion 4.3.3. The data in Figure 4.21a show a number of trends: Despite the same nomi­
nal loading of Cu and Ge, there is a difference in the activities for the Cu0.85 Ge0.15 /SiO2
catalysts. Different from the (1)/(2)­Cu1 /SiO2 catalysts, there is a larger spread in the
methanol mass activity for the two catalysts across the investigated temperature range.
Furthermore, both of the catalysts display a behaviour indicative of an activation during
the multiple temperature cycles. The preparation procedure of the two catalysts aimed
at different sizes of Cu crystallites, and as estimated from PXRD (Figure 4.20a), sizes
of 19 nm and 15 nm for the (1)­ and (2)­ catalysts was obtained. It is surprising that de­
spite the seemingly small difference in size, the two catalysts perform very differently in
terms of CO2 hydrogenation. While there is a small difference in the appearance of the
binary Cu­Ge phases of the two catalysts, as evident from the peak above 45°, we should
be careful in concluding that there is more of a less active Cu­Ge alloy present in the
(2)­Cu0.85 Ge0.15 /SiO2 catalyst than in the (1)­ counterpart.

The mass activity normalized with the measure of Cu atoms at the surface is given in
Figure 4.21b. Again, it is only the last temperature ramp of the two experiments that is
shown. The two experiments only differ in the total number of temperature cycles, and
none of them have been tested at temperatures above 230 ◦ C. Although the last point of
the blue data is averaged over a 5 hours step instead of the usual 1.5 hours, no decrease
in mass activity was observed during the 5 hours (see Figure B.19 for reference). After be­
ing subjected to operating conditions of 10 bar and 230 ◦ C in synthesis gas, the diameter
of the Cu crystallites of (1)­Cu0.85 Ge0.15 /SiO2 and (2)­Cu0.85 Ge0.15 /SiO2 have increased
to 23 nm and 20 nm, respectively.
The difference in mass activity from one catalyst to the other remains the same after aver­
aging with the Cu dispersion. The lower activity of (2)­Cu0.85 Ge0.15 /SiO2 is thought to be

59
Chapter 4. New Catalysts for CO2 Hydrogenation

a result of a different catalytic surface than what is found for (1)­Cu0.85 Ge0.15 /SiO2 . This
theory naturally opens up multiple questions about the nature of the catalytic surface that
are to be answered by complementary methods other than XRD.

(a) Methanol mass activity. (b) Surface­averaged mass activity.

Figure 4.21: (a) Catalytic mass activity of methanol during multiple temperature cycles
and (b) the mass activity (left scale, solid squares) and surface area normalized activity
(right scale, open squares) during the last temperature cycle, of (1)­Cu0.85 Ge0.15 /SiO2
(black) and (2)­Cu0.85 Ge0.15 /SiO2 (blue). Reaction conditions are 50 ml/min of CO2 /H2
in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . The
normalized rate in subfigure (b) is based on the crystallite size estimated from PXRD of
post­mortem samples. Note the two different scales in this subfigure.

The behaviour observed for the Cu0.85 Ge0.15 /SiO2 catalysts is not observed for the set
of Cu0.94 Ge0.06 /SiO2 catalysts. Unlike the catalysts investigated in Figure 4.21, the (2)­
catalyst is slightly more active than its (1) counterpart. Although there is quite a big
spread in the Cu crystallite sizes of the as­synthesized catalysts (32 and 23 nm, that is),
the methanol mass activities are not as different as those seen for Cu0.85 Ge0.15 /SiO2 .
Surface­averaged mass activities were not obtained for the set of Cu0.94 Ge0.06 /SiO2 as
the (2)­catalyst was used to investigate the effect of additional reduction. Data for the set
of Cu0.94 Ge0.06 /SiO2 catalysts are given in the appendix in Figure B.24.
Part III: Combining the effects of size and promotion
At this point of the study it is appropriate to combine the effects of size and promotion by
comparing the activity towards the target reaction of the range of catalysts listed in Ta­
ble 4.1. The first takeaway from the table is the variance in Cu crystallite sizes across the
range of catalysts. For all of the (2)­catalysts, the synthesis of the unsupported nanoparti­
cles was terminated as soon as all the reducing agent had been added. That way, smaller
nanoparticles were obtained compared to their counterpart (1)­catalysts without having to
change the concentration of the Cu precursor. The second takeaway is the variance in
Ge loading. As shown in Figures 4.9 and 4.11b, XRD patterns of catalysts with higher

60
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

loadings of Ge show Bragg peaks indicative of a Cu3 Ge phase. As the fraction of Ge de­
creases, the XRD patterns become more dominated by the crystalline Cu phase and and
slight amounts of the (Cu5 Ge)0.33 and Cu1.7 Ge0.3 phases, as evident from Figure 4.20.
This indicates that as the Cu:Ge ratio increases, we move away from a bulk Cu­Ge alloy
to a phase that resembles more and more Cu FCC ­ in line with the predictions from the
Cu­Ge phase diagram in Figure 4.8. The phase diagram does not clearly indicate how
Ge is arranged between the Ge mole fractions of 0 and 0.2, and it might be randomly
distributed Ge atoms on the surface of Cu particles.

In the following, only catalysts that have been subjected to the exact same treatment are
discussed. The ones not discussed in this section have either been treated at lower or
very high temperatures, or have been treated in a pure H2 atmosphere in­between cycles
of reaction conditions. Likewise, the catalysts have only been pre­treated in reaction con­
ditions and not in a H2 ­rich atmosphere before being exposed to syngas. Pre­treatment
of a Cu/ZnO catalyst can significantly change the mass activity of methanol (51, 60), and
there is no reason to believe the same would not be true for a Cu1−x Gex /SiO2 catalyst.
The effect of pre­treatment is therefore taken out of the equation in this screening study.

Table 4.1: Properties of the SiO2 ­supported Cu(1­x) Gex catalysts. The most important
ones are marked with a symbol *.

Sample name Cu loading / wt% dCu / nm a Cu dispersion / % b # Ge ML c

(1)­Cu1 4.2 67 (66) 1.6 ­
* (2)­Cu1 4.3 19 (33) 3.2 ­
* Cu0.67 Ge0.33 3.1 50 (40) 2.6 13
(1)­Cu0.85 Ge0.15 3.7 19 (23) 4.5 2.4
(2)­Cu0.85 Ge0.15 3.6 15 (20) 5.1 1.9
* (1)­Cu0.94 Ge0.06 3.7 32 (48) 2.1 1.7
(2)­Cu0.94 Ge0.06 3.6 23 (50) 2.1 1.2
* Cu0.96 Ge0.04 3.6 55 (75) 1.4 1.7

a Cu crystallite size as determined by PXRD using the Cu(111) diffraction peak of

the as­synthesized and (post­mortem) Cu(1­x) Gex /SiO2 catalysts.
b Cu dispersion (fraction of surface Cu atoms at the surface) is based on the post­

mortem Cu diameter. Here, it is expressed as a percentage.

c Number of Ge monolayers on Cu surface as based on d
Cu of as­synthesized
samples. Assumptions made are explained in­text in Section 4.3.4.

61
Chapter 4. New Catalysts for CO2 Hydrogenation

Figure 4.22 show the mass activity of methanol for the following catalysts: (2)­Cu1 /SiO2
(in blue, same data as given in Figure 4.16); Cu0.67 Ge0.33 /SiO2 (wine); (1)­Cu0.94 Ge0.06 /SiO2
(black); and Cu0.96 Ge0.04 /SiO2 (dark yellow). The data presented in the figures is that of
the last of a total of four temperature cycles in synthesis gas. In Figures 4.22a and 4.22b,
the mass activity normalized to the mass of catalytically active component is shown. When
comparing the intrinsic activity of these catalysts, it is important to compare them in a
range where transport limitations have not set in, or where the thermodynamics not yet
dictate whether conversion is possible (23). This leaves us with no other choice than
to compare catalysts in the regime where the rate increases steeply with temperature,
which is at fairly low temperatures. Whether normalizing with the amount of Cu as in Fig­
ure 4.22a or [Cu+Ge] as in Figure 4.22b, the picture remains roughly the same. Notice
that the scales of the Cu­ and [Cu+Ge]­normalized mass activities are different in terms
of units. There is no particular reason for this other than preference ­ there is no common
way to report on mass activities in the literature. The mass activities for the catalysts with
no or very little amounts of Ge all perform similarly in the entire temperature range. What
stands out is the behaviour of the Cu0.67 Ge0.33 /SiO2 catalyst. Up until a temperature of
220 ◦ C, it performs similarly to the other catalysts. Going past a temperature of 220 ◦ C,
the mass activity starts to increase less steeply until it decreases with temperature. From
Figures B.20 and B.21 it is also evident that the selectivity of methanol is worse com­
pared to (2)­Cu1 /SiO2 , as the Cu0.67 Ge0.33 /SiO2 produce higher levels of methane. The
production of methanol over the catalyst with a high Ge­loading is probably hindered by
a combination of thermodynamics and the production of methane.

62
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

(a) Catalytic mass activity of methanol normal­ (b) Catalytic mass activity of methanol normal­
ized with amount of Cu. ized with amount of Cu+Ge.

(c) Surface­averaged mass activity. (d) Catalytic mass activity in comparison to CZA.

Figure 4.22: (a), (b) Catalytic mass activity and (c) surface­averaged mass activ­
ity of a series of Cu(1­x) Gex /SiO2 catalysts. (d) Comparison of data in (a) with the
industrial­type Cu/ZnO/Al2 O3 /MgO catalyst (CZA, abbrev.). Reaction conditions are
50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of GHSV of 7.7 × 103 h−1 and a WHSV
of 8 × 105 Ncm3 · h−1 · g−1
Cu . No pre­treatment.

63
Chapter 4. New Catalysts for CO2 Hydrogenation

Figure 4.23: PXRD patterns of post­mortem Cu(1­x) Gex /SiO2 catalysts, each spectrum
shifted on the y­axis. The data insert focuses on the peak position of the main Cu(111)
peak. Peaks at 60° are from SiC dilutant.

Normalization of the data in Figure 4.22a with the estimates of the fraction of Cu surface
atoms yields a slightly different picture in Figure 4.22c, although not convincingly so. Nor­
malizing the mass activity with the averaged Cu surface area is not considered fair when
the bulk phases involving Ge are not taken into account or completely ignored in the calcu­
lation of dispersion. It is not correct to assume that all the Cu in the sample is available as
pure Cu when there are such clear Cu­Ge phases present in the diffractograms of the post­
mortem samples, as seen in Figure 4.23. The normalization is more valid for the pure
Cu catalyst, as well as the (1)­Cu0.94 Ge0.06 /SiO2 and Cu0.96 Ge0.04 /SiO2 catalysts. With
XRD as the only characterization tool, it is impossible to assume otherwise. The catalysts
with little Ge resemble the pure Cu catalyst with little to no variations in the Cu(111) peak
position. While there is no change in Bragg peak positions before and after activity testing
for the catalysts of Figures 4.22 and 4.23 with little to no Ge, there is a more apparent
restructuring of Cu0.67 Ge0.33 /SiO2 . The XRD patterns of as­synthesized and post­mortem
samples are shown in Figure 4.17b and Figures B.10 to B.12 with Figure B.10 showing an
increase in the binary Cu1.7 Ge0.3 phase after testing of the Cu0.67 Ge0.33 /SiO2 in syngas.
Although it is not supported by an in­situ study that carefully follows the restructuring of
the Cu­Ge phases under relevant conditions, the difference in ex­situ XRD patterns before
and after CO2 hydrogenation suggests that the trend in activity observed in Figure 4.22
can be due to restructuring.

64
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

In the end, what matters is whether we were successful in producing a catalyst more
active and selective for CO2 hydrogenation than the industrially used Cu/ZnO/Al2 O3 cat­
alyst. Although it is optimized for the CO2 /CO/H2 synthesis gas mixture, it is still a highly
active catalyst when it comes to CO2 hydrogenation to methanol. This is evident from
the supporting information in Figure C.1a. Comparing the mass activity of methanol for
the industrial­grade CZA catalyst with the investigated Cu(1­x) Gex /SiO2 catalysts in Fig­
ure 4.22d, it is perfectly illustrated how the industrially used catalyst have been highly
optimised for this specific reaction. It also illustrates that there is still a long way to go for
the binary Cu­Ge catalyst in terms of further development.
Turning to the surface­averaged mass activity in Figure 4.24, the CZA catalyst still per­
form much better than all of the Cu(1­x) Gex /SiO2 catalysts. In the same way as with
Cu0.67 Ge0.33 /SiO2 , the normalization with the Cu surface fraction might not be accurate
for CZA as ZnO and the nature of the active site is ignored in the estimation of the dis­
persion. Conferring with the PXRD of the post­mortem it once again becomes apparent
why CZA is such a success: The measured Cu crystallite size is roughly 10 nm despite
testing the performance of the catalyst at 300 ◦ C. This observation is consistent with the
observations of the in­situ study of reference (37).

Figure 4.24: Surface­averaged mass activity of a series of Cu(1­x) Gex /SiO2 catalysts
alongside that of the industrial­type Cu/ZnO/Al2 O3 /MgO catalyst. Reaction conditions are
50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of GHSV of 7.7 × 103 h−1 and a WHSV of
8 × 105 Ncm3 · h−1 · g−1Cu . No pre­treatment.

65
Chapter 4. New Catalysts for CO2 Hydrogenation

4.3.4 Discussion on the single­atom surface alloy Cu­Ge system

Without considering the measure of Cu surface dispersion, there is no clear indication of
the promotional effect of Ge in Cu(1­x) Gex /SiO2 catalysts. The ideal case would be to have
obtained the exact same size distribution of Cu nanoparticles across the sample range so
that the effect of Ge promotion could be separated completely from the size­effect. Efforts
to explain the promotion of Ge (or lack of) observed in Figure 4.22 are still in the works.

A first theory, of which there is no evidence yet, is that of Ge completely covering the sur­
face of Cu of the (1)­Cu0.94 Ge0.06 /SiO2 and Cu0.96 Ge0.04 /SiO2 catalysts. This behaviour
is not uncommon as a materials system will always aim at minimizing the surface free
energy, which for alloys means that the component with the lower surface free energy
will segregate to the surface (23). The cubic Cu (space group Fm3m) and Ge (space
group Fd3m), as found in the open­source Materials Project (150), have surface energies
weighted by the Wulff shape’s facet areas of 1.42 J/m2 and 0.91 J/m2 , respectively (151).
Theoretically, Ge is likely to segregate to the surface of Cu if they are alloyed. Estimates
of the number of Ge monolayers on the Cu surface are listed in Table 4.1. The estimates
are made under the assumption that Cu and Ge are alloyed and well­mixed, and without
taking into account their relative crystal structures but rather assuming that the crystal
structure is that of FCC Cu. The estimates illustrate the much important fact that the
smaller the nanoparticle is, the bigger is the ratio between surface area and volume. In
practice, this means that the smaller the nanoparticle, the more Ge is needed to deco­
rate the surface of Cu with the small amount that is predicted by theory to be active for
CO2 hydrogenation. What it also means is that all of the investigated Cu(1­x) Gex catalysts
theoretically have a monolayer or more of Ge on top of the Cu particles.

Returning to theoretical predictions of an alternative Cu­X system, the Cu(211) was used
as a model for the screening study of new promoters. This opens up the question of how
large a fraction of a Cu particle of a certain size is (211). Conferring with the surface
statistics of Figure 4.25, it is evident that the number of under­coordinated sites such as
those found on edges and corners are heavily dependent on the size of the nanoparticle.
Considering the imperfect cubo­octahedral Cu FCC crystal of Figure 4.25b, which has
more defects than the perfect one of Figure 4.25a, one type of edge site make up 15 % of
a 10 nm nanoparticle. This fraction, along with those of the other under­coordinated sites
found at edges and corners, only decreases with increasing nanoparticle size. Likewise,
Barberis et al. report that the number of edge sites (which they refer to as consisting of
(211) and (110) surfaces) peak at around 20 % at a nanoparticle size of 3 nm (134). From
there, it only goes downhill to less than 5 % at 30 nm. If we aim for 1/4 and preferably
less of a monolayer on the surface of a Cu(211) facet, it’s even less promoter than what
is used for the Cu particles used in this study. The morphology of small Cu particles
depend heavily on the reaction conditions and gas composition, as evident from atomically
resolved TEM images of Cu/ZnO model catalysts (59). The Cu/ZnO system is very much
subject to change during reaction conditions, and the Cu/Ge system might also be. For
the future, the gas­dependent surface morphology is worth studying.

66
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

(a) Perfect Cu FCC cubo­octahedron particle. (b) Imperfect Cu FCC cubo­octahedron particle.

Figure 4.25: Surface statistics of a Cu FCC (a) a perfect cubo­octahedron and (b) an
imperfect cubo­octahedron crystal as a function of nanoparticle diameter. Assuming Cu
atomic diameter of 2.55 Å. Data used for the plot is from (146).

Figure 4.26: Types of crystals considered for Figure 4.25. (a) is a perfect cubo­
octahedron, (d) is an imperfect cubo­octahedron Cu FCC crystal. Reprinted with per­
mission from (152).

The preliminary experiments presented here pose several new questions, which we want
to address with a combination of ICP, microscopy and spectroscopy:

• Firstly, we want to investigate whether Ge and Cu are present in the weight loadings,
but more importantly the ratios, that we set out to synthesize. The unsupported
nanoparticle suspension comes in handy as the nanoparticles are easily separated
from the solvent and can subsequently be dissolved in the preferred solution for a
specific ICP method. For weight loadings specifically, scanning electron microscopy
coupled with X­ray analysis is widely used and is, as opposed to ICP, not nearly an
as destructive method.

• To determine what is at the surface of the nanoparticles, surface­sensitive meth­

ods as opposed to methods that probe the bulk of a material are preferred. While
the XPS and ISS methods are not the most suitable for powders, it can be done if
extra care is taken to the sample surface. A big issue is the charging of the SiO2
support, and circumventing this means either flooding a reaction chamber with Ar

67
Chapter 4. New Catalysts for CO2 Hydrogenation

to discharge the sample, or to use a conductive support, such as carbon, instead of

SiO2 . For this specific point, the reader is referred to the XPS spectra of a selected
few Cu(1­x) Gex /SiO2 powder samples in Figures B.14 and B.15. Almost no signal
is obtained from Cu on SiO2 support, and no signal of Ge. Instead of studying the
powder sample, better results would probably be achieved by drop­casting the un­
supported Cu(1­x) Gex nanoparticles on a UHV compatible substrate or on a single
crystal.

• While XRD is a great first approach at revealing the crystal structure, total scattering
is better for probing local structure. For even better local structure information of
particles instead of crystallites, transmission electron microscopy is preferred. Size
and morphological distributions from a TEM study will also give a better estimate of
the surface area instead of relying on the single value from XRD to give an estimate
for an average radius of particles.

As discussed in earlier sections, there are many parameters to choose from in the hot­
injection wet­synthesis approach. As this study is only preliminary, the parameters used
here are still being optimized. The most important goal of the future is to control the
particle size distribution across a series of Cu(1­x) Gex catalysts. It was found especially
difficult to obtain small particle sizes for a Cu(1­x) Gex sample with a very small loading of
Ge, as evident from the properties of Cu0.96 Ge0.04 /SiO2 in Table 4.1. It was necessary to
scale up the amount of Cu to be able to work with such small amounts of Ge that was
needed for that given Cu:Ge ratio. For small and narrow size distributions of nanopar­
ticles, surfactants are usually needed (153). The use of surfactants can however block
the active sites of the catalyst, and so, a lot of time, chemicals and heat must be used
to remove them (154). On the other hand, when not using surfactants, high­viscosity
solvents are required instead (155). That only introduces more time­consuming steps of
washing to recover the nanoparticles. Evidently, there are quite a few parameters left to
optimize to effectively synthesize Cu­Ge nanoparticles with a specific set of properties
via the hot­injection method. A different method than synthesizing the Cu­Ge nanoparti­
cles via a mixture of the precursors might be to aim for a core­shell structure of Cu and
Ge, with Ge constituting the shell around small Cu nanoparticles. By annealing the core­
shell nanoparticles in various atmospheres and temperatures, it is possible to control the
degree of mixing (156).

Ideally, the Cu(1­x) Gex /SiO2 catalyst should be studied under relevant conditions using a
combination of methods that probe the whole sample (bulk probe) or a small part of it
(local probe). The Swiss­Norwegian BM31 beamline at the European Synchrotron Ra­
diation Facility (ESRF, abbrev.) have the instrumentation that allows for time and space
resolved studies (157): Here, it is currently possible to perform XAS (EXAFS/XANES),
XRD and X­ray total scattering/PDF separately or (almost) simultaneously under differ­
ent experimental conditions. Combining these three methods can give structural infor­
mation of the short­range environment (limited to first few coordination shells) around
selected atom species (XAS), yield atomic structure information from bulk structural anal­

68
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

ysis (XRD), while simultaneously elucidating the atomic structure of nanosized materials
that lack long­range atomic order (PDF analysis of X­ray TS data).
X­ray total scattering and HR­TEM study of (1)­Cu0.85 Ge0.15 /SiO2
In the following subsection, the preliminary findings from total scattering data and TEM
imaging of the post­mortem (1)­Cu0.85 Ge0.15 /SiO2 are discussed. PDF analysis of total
scattering data was done by Jette Mathiesen, and the acquisition and subsequent analysis
of (HR­)TEM images was done by PhD Stefan Akazawa.

(a) (b)

(c)

Figure 4.27: PDF refinement of (a) as­synthesized and (b) post­mortem (1)­
Cu0.85 Ge0.15 /SiO2 . Green lines display the difference between data (black) and Cu model
(blue). In (c), the experimental PDFs are compared to the calculated PDFs of Cu (red)
and Cu3 Ge (blue).

The refined parameters of the PDF analysis of X­ray total scattering data are given in the
Appendix in Table B.4, and a walk­through of the background subtraction done for this
sample is illustrated in Figure B.26. Comparing the experimental PDFs in Figure 4.27c

69
Chapter 4. New Catalysts for CO2 Hydrogenation

with the calculated model of Cu and Cu3 Ge, the presence of a FCC Cu structure is re­
vealed for both the as­synthesized sample as well as the post­mortem one. The nanopar­
ticle seemingly does not undergo any significant changes in local structure. The slight
broadening of peaks might relate to Ge incorporation, but refinement of the experimental
PDFs yields low Rw factors by only considering the FCC Cu model. As seen in Fig­
ures 4.27a and 4.27b, the refined model manages to explain most peak positions for
both the as­synthesized (Figure 4.27a) and the post­mortem (Figure 4.27b) sample of (1)­
Cu0.85 Ge0.15 /SiO2 . While unexplained features of the experimental PDFs might be due
to the presence of Ge that cannot be explained by incorporating a Cu3 Ge model, they
are also likely to be a result of residual noise. A 2­phase refinement using an additional
Cux Gey model (Cu3 Ge, (Cu5 Ge)0.33 or Cu1.7 Ge0.3 , the latter two indicated in Figure 4.20a)
did not yield a better model for the experimental PDF.

In an attempt to gain local information about the catalyst, transmission electron microscopy
was applied. Ex­situ TEM of the post­mortem sample of (1)­Cu0.85 Ge0.15 /SiO2 , given in
the Appendix in Figure B.28, show that the sample consists of large and thick particles in
the range >200 nm. This is supported by SEM­EDX that showed large, agglomerated Cu
particles. Figure B.27 in the Appendix show the SEM­EDX image of a representative par­
ticle. Converging the beam in TEM mode on the thick particles produced a diffraction pat­
tern, indicating that the particles do contain crystalline structure. Obtaining high­resolution
TEM images of the particles proved difficult since lattice fringes were only vaguely visible
at the very edges of the particles, and not in the center of the particles. This is possibly
due to thick and overlapping particles. To identify the crystal phase, a clear image of a
separate crystal, with the crystal being in the zone­axis, is needed. A TEM image showing
an area of thick, largely crystalline particles and the Fast­Fourier transforms (FFTs, ab­
brev.) of the HR­TEM images at the edge of large, agglomerated particles, are discussed
in Appendix B.9. Crystal phase identification of the crystalline area imaged at the edge
of agglomerated particles was complicated by the possible presence of multiple possible
phases of Cux Oy and Cux Gey .

Figure 4.28 show a 12 nm particle on the amorphous SiO2 support that is observed to
lie in zone­axis, making it more fitting for crystal phase identification. The FFT of the
particle reveals two sets of 6­fold symmetrical spots for which there are two possible
crystal structures and zone­axis orientations for observing these spots: FCC [111] and
hexagonal [001]. To match the imaged crystal to any possible materials there might be
in the sample, the reciprocal lengths of the relevant materials are overlayed as circles
for both the FCC [111] case (upper right image of Figure 4.28) and for the hexagonal
[001] case (lower right image of Figure 4.28). In the lower left image of Figure 4.28, the
reciprocal lengths of the peaks for the SiC dilutant (as determined by PXRD) are shown.
Some of the SiC XRD peaks are close to the FFT spots, but the SiC system have not been
studied further in terms of symmetry and expected reflections. For the case of possible
Cux Oy and Cux Gey materials, none of the considered hexagonal or FCC crystal structures
match the FFT spots completely.

70
 4.3. Study of Ge Promotion in Cu­Ge/SiO2

Figure 4.28: HR­TEM image (top left) and FFT of a single particle observed in the post­
mortem sample of the (1)­Cu0.85 Ge0.15 /SiO2 sample. The FFTs were obtained after ap­
plying a Hanning window over the real­space HR­TEM image. The circles in the FFTs
represent reciprocal distances of relevant crystal structures and reference materials. Top
right: FCC crystal structures. Bottom left: SiC dilutant measured with in­house XRD.
Bottom right: Hexagonal crystal structures. HR­TEM image was acquired and analyzed
by PhD Stefan Akazawa.

The PDF analysis of total scattering data, along with the analysis of (HR)TEM images,
indicates just how difficult it may be to locate, let alone explain the state of, Ge in the
Cu(1−x) Gex /SiO2 catalysts with low Ge loading. Especially for (HR­)TEM, we need a sin­
gle crystal that lies in the zone­axis in order to properly identify the crystal structure. For
future studies, in­situ HR­TEM would be an especially interesting technique to apply to
obtain fundamental insight into the Cu­Ge material system. The unsupported Cu(1−x) Gex
nanoparticles can easily be drop­casted on a suitable reactor that can be loaded into the
microscope. By carefully evaluating the bulk and surface responses while the nanoparti­
cles are exposed to heating/cooling cycles and relevant gas compositions, while prefer­
entially performing both HR­TEM and STEM­EDX, we can gain valuable insight into the
nature of any active surface sites during catalysis.

71
Chapter 4. New Catalysts for CO2 Hydrogenation

4.4 Preliminary Studies of Possible Promoters for Cu­X/C

The preliminary results in this section are part of a thesis project that the author of this
thesis co­supervised. The catalysts were prepared by PhD Thomas Smitshuysen and
MSc student Mikkel Stensberg, who also tested the catalysts for activity.

Synthesis and catalytic tests. The Cu/X/C catalysts were prepared by incipient wetness
impregnation as introduced in Section 3.1.1. For the impregnation, the following chem­
icals were used: 98 % Cu(II)formate­tetrahydrate (Alfa Aesar, A18569), Bi(III)acetate,
Sn(II)acetate, Sb(III)acetate, Pb(II)acetate­trihydrate, PBX51 carbon black (Cabot Cor­
poration). All samples were synthesized with a resulting nominal loading of 5 wt% Cu and
a X:Cu ratio of 0.05. The dried, impregnated powders were loaded in steel reactors with
SiC dilutant. Once loaded in the reactor, they were calcined in 20 % O2 /Ar and subse­
quently reduced in 20 % H2 /Ar at 250 ◦ C, both at 1 bar. Catalytic tests were performed in
H2 and CO2 (ratio 4:1) with trace amounts of Ar at 10 bar.

Figure 4.29: Catalytic mass activity of a series of Cu/X/C catalysts. Reaction conditions
are CO2 /H2 in 1:4 at 10 bar, and a WHSV of 1.3 × 104 Ncm3 · h−1 · g−1 Cu for the Cu/X/C
−1 −1
catalysts and 2 × 10 Ncm · h · gCu for the Cu/ZnO/Al2 O3 /MgO catalyst.
5 3

Figure 4.29 show the preliminary activity measurements of the alternative Cu­X catalysts,
all supported on carbon black support, in comparison to an industrial­type CZA catalyst.
The WHSV per weight of Cu is different for the CZA and the Cu­X/C catalysts as it was
discovered very late in the screening process that approximately 60 % of the gas was lost
before reaching the plug flow reactor. These results should therefore be interpreted with

72
 4.5. Conclusion on New Catalysts for CO2 Hydrogenation

great care. The data for the Cu­X/C catalysts in Figure 4.29 are treated with the assump­
tion that these experiments were done under similar gas flow conditions. Interestingly,
the Sn­promoted Cu/C catalyst outperforms all of the other X­promoted catalysts in terms
of methanol mass activity in the entire temperature range. Compared to the pure Cu cat­
alyst, the Pb­promoted Cu/C performs worse. None of the carbon­supported catalysts
perform as well as the CZA catalyst, which shows that there is plenty of optimization left
to do.

Unfortunately, not much useful information can be retrieved from PXRD patterns of the
used catalysts as they are dominated by Cu oxide phases and peaks originating from the
SiC dilutant. What can be retrieved are estimates of Cu crystallite sizes. As evident from
the narrow Bragg peaks of the Cu phase, the average Cu crystallite sizes are quite large
for all of the used samples, ranging from approximately 80 to 30 nm. XRD patterns of
used samples and relevant references are given in Figure D.3, along with the crystallite
size estimates.

4.5 Conclusion on New Catalysts for CO2 Hydrogenation

New Cu­X catalytic systems for the thermal hydrogenation of CO2 have been investigated.
The Cu(211) facet, which is known to be active for methanol formation, with loading of X
small enough to form a single­atom surface alloy, were predicted to be active ­ some more
than others, and some more so than the well­known Cu­Zn system. The low loading of
promoters is generally a very important goal, and in connection with this it is worth noting
that elements such as Sn, Sb and Pb might be more viable substitutions than Bi and more
so than Ge, as predicted by Laursen et al. (96). As the question of viability have not been
considered here, the study of new Cu­X material systems for CO2 hydrogenation was
focused on the most promising promoter, namely Ge.

CuO/GeO2 ­based catalysts were successfully synthesized by sequential incipient wet­

ness impregnation. Rearranging the steps of impregnation successfully resulted in a cat­
alyst containing both CuO and GeO2 , with no traces of Cl present, that also showed cat­
alytic activity in terms of CO2 hydrogenation. Although the CuO/GeO2 /SiO2 investigated
in this part of the study had a nominal loading of Ge much larger than what is theoretically
required to form a single­atom surface alloy of Ge on Cu, it proved to be a useful first go
at the Cu­Ge system.
Pre­treatment and subsequent activity testing for CO2 ­based methanol synthesis at 10
bar was done for a catalyst synthesized via the optimal sequential IWI procedure. Even
though the SiO2 ­supported Cu catalyst was found to have very little activity in terms
of methanol formation, it outperformed the CuO/GeO2 ­based catalyst. Pre­treatment at
higher temperatures resulted in a small increase in activity for CuO/GeO2 /SiO2 , but still
far from that of Cu/SiO2 . Using in­situ XRD and studying the formation of phases through­
out the reduction of a high­loading CuO/GeO2 /SiO2 sample, the sequential reduction of
Cu­ and Ge oxides, along with the formation of bulk Cu3 Ge, was observed. The high

73
Chapter 4. New Catalysts for CO2 Hydrogenation

temperatures needed for reducing GeO2 helps to explain the low methanol mass activity
observed during activity testing. While it is now established that CuO and GeO2 needs dif­
ferent reduction potentials, it is important to highlight that the reduction procedure needed
for GeO2 is thought to be detrimental to Cu, as Cu is prone to sintering at high tempera­
tures. This should be taken into consideration if GeO2 is to be used as an intermediate in
the formation of a Ge­promoted Cu catalyst.

Switching gears to study, with the intention of separating, the effects of particle size and
Ge promotion, Cu(1­x) Gex /SiO2 catalysts with varying nominal loadings of Ge were syn­
thesized following a hot­injection wet­synthesis approach. By tuning the synthesis pa­
rameters, different mean Cu particle sizes were obtained within the same Ge/Cu nominal
ratio. While a Cu­Ge bulk alloy phase was clearly visible in both PXRD and X­ray total
scattering data of the samples with the highest Ge loading investigated (Cu0.67 Ge0.33 ), it
proved difficult to determine the state of Ge in the samples of lower Ge­loading.
The activity of two Cu­based catalysts can be separated based on an estimate of the
number of surface Cu atoms, i.e. the dispersion, but it is not considered as straight for­
ward for Ge­promoted Cu catalysts. Although the surface­averaged mass activities of the
range of Cu(1­x) Gex /SiO2 are slightly different on a linear scale in the temperature range
260 −300 C, it is important to point out that the mean Cu particle sizes vary across the
range of catalysts. An estimated measure of the number of Ge monolayers on the sur­
face of Cu indicates at least one monolayer Ge or more for all of the samples, despite the
low Ge loading, simply due to the large variance in Cu particle size. Assuming that Ge
segregates to the surface of Cu at low loadings, combined with the fact that the fraction of
under­coordinated sites such as Cu(211) decrease with increasing particle size, means
that the number of Ge­promoted Cu(211) sites is likely very small for the catalysts studied
here. This is possibly a reason for the lack of promotion in methanol mass activity for the
Cu(1­x) Gex /SiO2 catalysts doped with small amounts of Ge.

74
Chapter 5

Carbon­supported Cu/Zn as a
Model System
Two things are generally accepted for the co­precipitated Cu/ZnO/Al2 O3 catalyst. One
is that Cu is the main catalytically active component, and the other is the importance of
ZnOx as a promoter (18). While there is a strong consensus on the importance of ZnOx
in Cu catalysts, the exact role of ZnOx and the structure of the active site is still widely
investigated in literature. The origin of the Cu­ZnO synergy is classical metal­promoter
structure and can serve as a model for metal­support systems to come, consequently
making the Cu­ZnO system a very important one to fully understand. The role of Zn is
complicated by its dual role as an electronic and structural promoter. What makes the
task of revealing the state of Zn during working conditions even more complicated is that
additional components such as MgO, ZrO2 , TiO2 , Ga2 O3 or CeO2 often are added to the
commercial Cu/ZnO/Al2 O3 catalyst to fight e.g. reduced catalytic surface area (158, 159).
Although it proves beneficial for the performance, adding these components adds more
complexity to an already complex system since having the multiple oxidic components
that might work as both structural and electronic promoters makes fundamental studies
on the nature of the active site during CO2 hydrogenation challenging. Specifically for
the commercial methanol catalysts, a large fraction of the ZnOx promoter is thought to be
present as spectator species (69, 74, 76, 127, 160–163). Especially this aspect inspired
the type of model system considered in this study, namely a series of Cu nanoparticles,
with an increasing amount of Zn, supported on an inert carbon material. The reasoning
behind the increasing amount of Zn comes down to the desire for separating the electronic
and structural promotion that Zn/ZnO evidently has on Cu.

This kind of model for the oxide­supported Cu/Zn catalyst is not new to the field. The
research group of J. Nakamura have extensively studied the activity of Cu­ZnO systems
with varying amounts of Zn (27, 41, 45–47, 55, 61, 164). Most of these studies are
of Zn­decorated Cu single­ or polycrystalline crystals, while some are of co­precipitated
Cu­Zn. Common for all is that a clear optimum for the activity is seen, and as the mol­
% or Zn coverage increases, the activity decreases sharply after the optimum. A next
step in the direction of elucidating the role of Zn in the commercial Cu/ZnO/Al2 O3 catalyst
is the introduction of a support. The group of P. E. de Jongh have published several
interesting results the past years, all from investigations of Cu and Cu/Zn nanoparticles
on graphitic carbon supports (52, 64, 134, 165). Most recently, it was confirmed that
by using a weakly interacting graphitic carbon support, and combining it with a small Zn
loading, it is possible to avoid the large fraction of Zn spectator species normally found

75
Chapter 5. Carbon­supported Cu/Zn as a Model System

when using an oxide support and large Zn loadings. That way, the promotional phase
of Zn that is in close contact with Cu could be investigated using time­resolved in­situ
X­ray absorption spectroscopy (XAS, abbrev.). By studying the X­ray absorption near
edge structures (XANES, abbrev.) of Cu and Zn, it was revealed that the Zn is present as
reduced Zn atoms on the Cu under reaction conditions (64).

Carbon materials are a popular choice for support because they can be fabricated in
different physical shapes and forms, as well as be chemically functionalized and/or deco­
rated with metallic nanoparticles. Carbon­based supports are also no stranger to thermal
methanol synthesis (52, 64, 134, 165–169). The nanostructured carbon materials span
from zero dimensions (NPs), to one dimension (carbon nanotubes), and further to two
and three dimensions (graphene sheets and mesoporous carbon). The review by Lam et
al. highlights some of the key advantages of carbon supports for catalytic applications,
with some of them being tailored pore size distribution and the low price (170). One of
the highlighted points is the stability of the structure at temperatures as high as 1000 K
­ except if oxygen is present (> 500 K) or if the structure is used for hydrogenation reac­
tions (> 700 K). This point is not something that have been investigated in this study, but
it should be taken into account when considering high­temperature data or for the future
optimization of the synthesis of carbon­supported nanoparticles. For this study, the most
important advantages of the carbon support are the large specific surface area and high
porosity and the excellent electron conductivity, the latter making the carbon support a
great choice for characterization by X­ray photoelectron spectroscopy and microscopy.

In this study, carbon­supported Cu/(Zn) nanoparticles with varying amounts of Zn are

proposed as a model for the industrial methanol catalyst. By using a weakly interacting
support such as carbon, one can (in theory) focus on the Cu­Zn interaction without having
to worry about strong metal­support interactions. The idea behind the model is threefold.
By using low metal loadings, we can hopefully avoid sintering while maintaining small
nanoparticle sizes, while a step­wise increase in the loading of the promoting species will
allow us to observe the promotional effect. If the optimal amount of Zn is determined, the
chemical state of the relevant fraction of Zn before, during, and after reaction conditions
can be determined. Ideally, the interaction between Cu and Zn during reaction conditions
can also be established, which at the end would prove the nature of the active site of the
Cu­Zn catalyst.
The following sections describe the study of carbon­supported Cu/(Zn) nanoparticles and
the effect of Zn on the conversion of synthesis gas to methanol. Catalysts are referred
to as Cu/Zn, where the usage of ’Zn’ is not an implication of the chemical state of Zn
during reaction conditions. The use of ’Zn’ instead of ’ZnO’ refers to the fact that Zn is
considered as a promoter, and not a support. The majority of the study has been on the
effect of Cu particle size on the activity and how the combination of X­ray diffraction and
scanning TEM methods can be used to determine particle sizes in the nanosize range.

76
Synthesis and catalytic tests. The CuZn/C catalysts were prepared by incipient wetness
impregnation as introduced in Section 3.1.1. For the impregnation, the following chem­
icals were used: 98 % Cu(II)formate­tetrahydrate (Alfa Aesar, A18569), Zn(II)formate­
dihydrate (Santa Cruz Biotechnology, sc­280201), PBX51 carbon black (Cabot Corpora­
tion). The dried powders were calcined in static air at 220 ◦ C before being loaded in either
glass reactors with no dilutant or glass­lined steel reactors with SiC dilutant. Once loaded
in the reactor, they were reduced in H2 /Ar at a given temperature, and catalytic tests were
subsequently performed in CO2 /CO/H2 in 1:1:5 at 1 bar. Unless otherwise stated, the
reduction is as follows: 2 % H2 /Ar at 175 ◦ C for 3 hours followed by 30 minutes at 240 ◦ C
in 100 % H2 . The reported elemental compositions (weight loadings, Zn/Cu ratios) are
based on the nominal metal loadings of the final catalyst, assuming metallic Cu and Zn.

SEM/HAADF­STEM. HAADF­STEM coupled with EDX, as well as SEM­EDX, was per­

formed by PhD Mads Lützen. SEM­EDX was performed on calcined and post­mortem
powder samples loaded on conductive carbon tape on Al stubs. A Quanta FEG micro­
scope equipped with detectors for both imaging and EDX was used for this purpose. The
microscope was operated at an acceleration voltage of 30 keV using a working distance
(WD) of 10 mm.
HAADF­STEM was carried out on a Titan Analytical 80­300 ST microscope operating at
300 keV and a current of 150 pA. For sample preparation, a small amount of post­mortem
sample was dispersed in ethanol and drop casted on a lacey­carbon Au TEM grid.
The analysis of all spectra was done using the AZtec Software from Oxford Instruments
and ImageJ in the Fiji distribution (171, 172). Size­distribution were estimated using the
HAADF­STEM images by outlining the particles. For this, two measures for the diameter
was extracted, and the average diameter (assuming the nanoparticles are spherical) was
used to determine the volume­averaged and surface­averaged particle sizes. Approx­
imately 200 individual particles across different locations on each of the samples were
measured to determine the size­distributions.

XRD. Small amounts of as­synthesized, reduced and post­mortem samples were taken
to be analyzed with XRD. They were loaded in between two pieces of kapton foils in a
dedicated sample holders.

X­ray total scattering. Samples of post­mortem Cu(Zn)/C were loaded into thin kapton
tubes. Experiments were conducted at Deutsches Electronen­Synchrotronen (DESY) in
Germany. The sample­to­detector distance and instrumental broadening was calibrated
using a CeO2 standard. Total scattering data were collected in transmission geometry us­
ing a Perkin­Elmer detector with a pixel size of 300×300 µm in the RA­PDF setup mode
(90), where a wavelength of 0.122 Å was used. The 2D data were corrected, normalized
and integrated using Fit2D, and Fourier transformed using xPDFsuite to obtain the PDFs
(173–175). Background subtraction was performed by subtracting the scattering signals
obtained from the empty sample holder and from a sample of PBX51 carbon black sup­
port that had undergone the same pre­treatment and activity measurement as the Cu/Zn
catalysts. PDFgui was used for the PDF refinement (176).

77
Chapter 5. Carbon­supported Cu/Zn as a Model System

Supporting methods: N2 O­RFC and BET. Reactive frontal chromatography by nitrous

oxide (N2 O­RFC, abbrev.), performed at the Institute of Chemical Engineering, DTU, was
used to titrate Cu areas on a selected few catalysts after in­situ reduction. Brunauer­
Emmett­Teller (BET, abbrev.) surface areas were determined of the untreated carbon
black support and a series of calcined CuZn/C catalysts with a nominal Zn/Cu ratio of
0.06. Specific for BET, N2 physisorption measurements were performed at −196 ◦ C using
a Micromeritics ASAP 2020 V6.01 apparatus. Prior to the physisorption measurements,
the samples were outgassed in a N2 flow. The specific surface area were calculated by
applying the Brunauer­Emmet­Teller equation to the N2 adsorption data in the range of
p/p0 =0.20 to 0.30 of the pressure ratio (177).

5.1 Effect of Zn Content on Syngas Conversion

The metal weight loading and elemental Cu/Zn composition was validated for selected cat­
alysts by inductively coupled plasma optical emission spectroscopy (ICP­OES, abbrev.)
and scanning electron microscopy combined with energy­dispersive X­ray spectroscopy
(SEM­EDX, abbrev.). ICP­OES was performed for two catalysts by Konrad Herbst at
Topsøe. The results shown in Table 5.1 indicate very little deviation from the expected
nominal weight loadings.
PhD Mads Lützen analyzed selected samples (both calcined and post­mortem) with SEM­
EDX, and the relation between the expected and the analyzed Zn/Cu ratio is shown in
Figure 5.1. The deviation becomes somewhat larger at higher Zn/Cu ratios, while the
relation between the expected and measured composition fits almost perfectly at low Zn
loadings. Analyzing the SEM­EDS images, the Cu and Zn wt% were very different from
the nominal (expected) metal loadings. Referring to the ICP­OES analysis, this is not
expected to be the case, which indicates that the number of investigated particle grains
are not enough to give an accurate global composition. The weight loadings estimated
from EDS data are instead used to compare the measured Zn/Cu ratio to the expected
one. There does not seem to be a systematic error for this synthesis, and so the small
discrepancies observed in Figure 5.1 were not followed any further. The samples will be
referenced using the nominal composition as expected from the weighing of chemicals,
especially as the study is not aimed at optimizing the weight­loading of the catalysts, but
rather observe the promotional effect of Zn. All catalysts are named CuZn­X/C, in which
X represents the Zn/Cu ratio as based on the nominal weight loading.

Table 5.1: Nominal weight loadings (from weighing of chemicals) and measured weight
loadings (from ICP­OES) of selected carbon­supported CuZn catalysts. ICP­OES was
measured by Konrad Herbst at Topsøe.
Nominal loading Analysis result
5.01 wt% Cu; 0.58 wt% Zn 4.94 wt% Cu; 0.57 wt% Zn
5.05 wt% Cu; 0.72 wt% Zn 5.08 wt% Cu; 0.72 wt% Zn

78
 5.1. Effect of Zn Content on Syngas Conversion

Figure 5.1: Expected (from nominal weight loadings) and measured (as determined by
SEM­EDS by PhD Mads Lützen) Zn/Cu ratios of selected catalysts in their calcined (black)
and postmortem (red) states. Line going through origo is inserted to guide the eye.

The activity in the methanol synthesis reaction of the carbon­supported catalysts with
increasingly larger amounts of Zn to Cu were investigated at 1 bar, in synthesis gas
(CO2 /CO/H2 in 1:1:5), at temperatures below 240 ◦ C and at low conversion. A synthe­
sis gas feed of CO2 /CO/H2 was chosen to mimic industrially relevant conditions (178–
180). Although studies in pressure regimes that are industrially relevant (20 −100 bar)
are the key to overcome the pressure gap, investigations of the CuZn/C model catalysts
were done at 1 bar. The 1 bar limit is chosen so that the observations made in this study
can more easily be compared with previous and future (quasi) in­situ or operando studies
where the pressure is usually limited to 1 bar or less. The 1 bar limit is not necessarily
detrimental as the dynamic behaviour of Cu/ZnO catalysts have been observed at low
pressures (59, 181). Segregation trends in SiO2 ­supported Cu/ZnO have been found to
be similar at both high and low pressures (76), indicating that experiments performed at
low pressures, while using realistic gas conditions, are also useful for elucidating surface
and bulk trends.
Before testing any of the Cu(Zn)/C catalysts, it was ensured that no activity was observed
for empty glass reactors or glass­lined steel reactors, nor for the untreated carbon black
support, or for a mixture of SiO2 and SiC. It is therefore safe to say that only the Cu(Zn)/C
catalysts contribute to the activity.

79
Chapter 5. Carbon­supported Cu/Zn as a Model System

Figure 5.2 shows how increasing the ZnOx content maximizes the stability in a H2 /CO2 /CO
feed. Comparing the activity of CuZn­0.31/C in Figure 5.2a with that of CuZn­1.5/C in Fig­
ure 5.2b, it is evident how the higher loading of Zn of the CuZn­1.5/C catalyst not only helps
to increase the activity, and especially so at high temperatures, but also promotes the sta­
bility. During testing of the CuZn­0.31/C catalyst with 1.5 wt% Zn, the methanol formation
is limited at the highest temperatures investigated. While there is no sign of deactivation
during the first half of a second temperature cycle, a stability test at 240 ◦ C for 10 hours
show a continuous decline in the volumetric concentration of methanol ­ as seen in Fig­
ure 5.2a. Increasing the amount of ZnOx to gain 12 wt% of Zn in the final catalyst helps
to stabilize the carbon­supported catalysts as the mass activity over a 10­hour period at
reaction conditions does not decline. The observation is in line with that of Dalebout et
al., who ascribe the trend to a limited CuOx particle growth with a higher Zn loading (64).

(a) CuZn­0.31/C (b) CuZn­1.5/C

Figure 5.2: Volumetric concentration of methanol during a temperature cycle followed by

a 10­hour stability test at 240 ◦ C for CuZn/C catalysts with Zn/Cu ratios of (a) 0.31 and (b)
1.5. Reaction conditions are 100 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of
6 × 105 Ncm3 · h−1 · g−1Cu (3 × 10 Ncm · h
4 3 −1 · g−1 ).
cat

Figure 5.3 shows the mass activity of methanol for the Cu(Zn)/C catalysts as a function
of the amount of Zn. These catalysts were all tested in glass reactors and reduced under
the same conditions. While the reaction conditions are the same for the data shown in
Figure 5.3, some of the catalysts was subjected to a longer testing period and/or higher
temperatures, which will affect the post­mortem state of the catalysts. To compare the
intrinsic activity of the catalysts, the activity measurements are generally presented at a
temperature of 220 ◦ C or below.

80
 5.1. Effect of Zn Content on Syngas Conversion

Figure 5.3: Catalytic mass activity of methanol on carbon­supported Cu and CuZn parti­
cles as a function of Zn loading, using glass reactors. Reaction conditions are 100 ml/min
of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of 6 × 105 Ncm3 · h−1 · g−1 Cu . Note the log­
arithmic scale. Lines are inserted to guide the eye, and uncertainties are estimated by
propagation of error (121).

For some of the tested catalysts of Figure 5.3, there is a increase in methanol mass activity
during the cooling part of the temperature cycle compared to the heating part. The cat­
alysts in question were those of the lowest Zn­loading, CuZn­0.05/C and CuZn­0.13/C,
with the first one experiencing the largest increase in activity. The first is found in Fig­
ure F.1, the second in Figure 5.4b. A similar trend is observed by Kuld et al. who ascribe
it to a delay in the accumulation of active species (51). This is not unlikely considering
that the activity of these two catalysts are quite low compared to the rest of the catalyst
series. Although it is yet to be investigated, the higher mass activity after the first half of
the cycle can also be due to a further reduction of the metallic species. While CuO/C is
fully reduced to Cu with the conditions applied here (as evident from the in­situ studies of
Cu/C in Appendix F.3), we are yet to observe the reduction of ZnO to Zn in the CuZn/C
catalysts with low Zn loadings. It is a known phenomenon that the state of the Zn at the
surface is dependent on the pre­treatment and the reduction potential of the gas mixture
(50, 51). Due to the slight difference in activities when heating and cooling for some of
the tested catalysts, data points in Figure 5.3 are extracted from the temperature ramp
downs.
For Cu(Zn)/C catalysts tested in glass reactors, the amount of zinc required to obtain
maximum mass activity corresponds to Zn/(Cu + Zn) fractions above 0.1 (following the red
curve for activities at 220 ◦ C and taking into account the estimates of error). Not only is
this quite similar to previous reports of Cu(Zn) nanoparticles supported on carbon (52, 64),

81
Chapter 5. Carbon­supported Cu/Zn as a Model System

the optimal Zn loadings used here are also slightly lower than the optimal loading found
for the commercially used catalyst and other oxide­based Cu catalysts (28, 51, 61, 66, 76,
182–185), as shown in Table 5.2. Interestingly, the optimal composition of Cu and Zn is
different from one study to another for the same support (compare those of references (52)
and (28)), likely due to differences in synthesis and pre­treatment. All of the compositions
reported in the table are, however, quite close to that of the industrial­type purchasable
methanol catalyst, indicating that all of the catalysts in theory are suitable models for the
commercial catalyst. Despite this, it is argued that oxide supports interacts more strongly
with the metallic species, making carbon support more suitable to study the interaction
between Cu and Zn under relevant conditions.

Table 5.2: Optimal Cu/Zn compositions in literaturea

Catalyst Investigated (optimal) composition Reference

Cu/Zn/C Zn/(Zn+Cu) = 0.1­0.2b This study
Cu/Zn/HSAG, Cu/ZnO/(Al2 O3 ) Zn/(Zn+Cu) = 0.05, 0.1 (52)
Cu/ZnOx /C Zn/(Zn+Cu) = 0.15­0.25 (64)
Cu/ZnO/Al2 O3 Zn/(Zn+Cu) = 0.25­0.3 (28)
Cu/Zn Cu/ZnO = 50/50 (61)c
Cu/Zn Zn/(Zn+Cu) = 0.2 (185)
Cu/ZnO Zn/(Zn+Cu) = 0.3 (182)c
Zn/Cu(111) ΘZn = 0.2 (47)c
ZnO/Cu(111) ΘZnO = 0.2 (32, 68)c

a If several Cu/Zn compositions are reported, the optimal is listed here. For refer­
ence: CZA catalyst from Alfa Aesar (CuO/ZnO/Al2 O3 /MgO=63.5/24.7/10.1/1.3
wt%) corresponds to Zn/(Zn+Cu)=0.28.
b High limit from experiment with glass­lined steel reactors.

c CO /H feed.
2 2

As briefly mentioned, the catalysts tested to create Figure 5.3 differ in their sample history.
Besides the catalysts that were subjected to a stability test (Figure 5.2), the CuZn­0.13/C
catalyst was subjected to two consecutive temperature cycles with the second one be­
ing at higher temperatures. During this second temperature cycle, the methanol mass
activity follows the temperature very closely, due to the increased kinetics at higher tem­
peratures. This is apparent from both the quantitative signals from the gas chromatograph
(Figures 5.4a and 5.4b), as well as from the qualitative signals measured by the QMS
(Figure 5.4c). Not only does the QMS show the signals of atomic masses 28, 31 and 44,
coming primarily from CO, methanol and CO2 . It also shows the signal of atomic mass 18,
being primarily H2 O. The same trend seen for methanol is also seen for water, although it
is less visible at lower temperatures, as the background level is much higher for the H2 O
signal than it is for the methanol signal. Increasing the temperature all the way to 300 ◦ C
diminished the methanol yield significantly, and the activity is not regained during the last

82
 5.1. Effect of Zn Content on Syngas Conversion

ramp down in temperature. This behaviour is very similar to that of the industrial­type
CZA catalyst, which has been tested in different feed gas compositions. Evident from the
QMS signals during testing in a CO2 /CO/H2 gas feed, the CZA catalyst also produces
water, although less to than during CO2 hydrogenation (Figures C.2b and C.2c). Whether
supported on carbon or an oxide, the CuZn system is thought to have limited production
of methanol at high temperatures due to the limitation of thermal dynamics. Furthermore,
the fact that the mass activity of methanol is not regained is indicative of the deactivation
through sintering of Cu, caused by a combination of high temperatures and presence of
water. This is a well­established fact for the Cu/ZnO­based methanol catalyst (67, 93, 94,
120).

(a) CuZn­0.13/C: Volumetric concentration. (b) CuZn­0.13/C: Mass activity.

(c) CuZn­0.13/C: QMS signal.

Figure 5.4: (a) Volumetric concentration and (b) mass activity of methanol during two
consecutive temperature cycles for CuZn­0.13/C. (c) is the QMS signal and temperature
during the second temperature cycle. Reaction conditions are 100 ml/min of CO2 /CO/H2
in 1:1:5 at 1 bar, with a WHSV of 6 × 105 Ncm3 · h−1 · g−1
Cu .

83
Chapter 5. Carbon­supported Cu/Zn as a Model System

5.2 A Question of Surface Area or Zn Promotion

Recent findings by Barberis et al. indicate that the CO2 hydrogenation reaction, and more
specifically the methanol selectivity, is influenced by Cu particle size (134). Considering
this, it is possible that part of the trend in Figure 5.3 is due to Cu surface area, and some
is due to Zn promotion. Although the catalysts of Figure 5.3 were subjected to slightly
different testing procedures, all of the samples were characterized in their post­mortem
state to analyze the morphology of the particles, as well as the particle sizes. In an attempt
to separate the size effects and the promotional effect of Zn, the particle sizes of the
catalysts in their post­mortem states are investigated using a combination of methods
that are inherently bulk (PXRD), local (STEM), and bulk­averaged but sensitive to local
structure (PDF analysis of total scattering data).

5.2.1 Bulk­averaged information: PXRD analysis

Figure 5.5: XRD patterns of post­mortem Cu(Zn)/C catalysts with increasing amount of
Zn (upper to lower patterns). Data insert is focused on the Cu(111) peak of the Cu/C
and first 3 CuZn/C catalysts, and asterisks are used to highlight the samples that have
undergone a different activity testing. XRD pattern of carbon black support is pictured
for easy comparison. Reference spectra: Cu (ICSD 235809, red), CuO (ICSD 16025,
orange), Zn (ICSD 25286, blue) and ZnO (ICSD 26170, grey).

Figure 5.5 show the XRD patterns of the Cu(Zn)/C catalysts from Figure 5.3. The diffrac­
tograms are taken of the post­mortem samples that have been exposed to air. The pres­
ence of ZnOx in increasingly larger amounts, from CuZn­0.05/C all the way to the CuZn­
1.5/C catalyst, apparently limits the Cu particle sizes as evident from the broadening of
the Cu(111) main peak. While crystalline Cu are visible for all samples, with the exemption
of CuZn­1.5/C, the presence of Zn is less obvious. The main peak of HCP Zn is difficult

84
 5.2. A Question of Surface Area or Zn Promotion

to distinguish from that of FCC Cu in the region above a diffraction angle of 40°, but the
presence of a broad diffraction profile together with a narrow peak just above 2θ 35° points
to the presence of anisotropically shaped crystallites. ZnO commonly has a hexagonal
wurtzite crystal structure which have a tendency to form anisotropic crystals (186).
A slight shift of the Cu(111) peak is observed for Cu/C and the first three CuZn/C catalysts.
As argued by (64), the small downshift of the Cu(111) peak, in this case from 2θ 43.4° to
43.3°, can suggest CuZn alloy formation. While it is not supported by neither STEM nor
X­ray total scattering of the first three samples of low Zn loadings, STEM­EDX images in
the following section suggest the possibility of CuZn alloying at higher Zn loadings.

Estimation of the Cu crystallite size is complicated by the diffuse peaks originating from
the carbon support, especially as the Cu(111) peak width and intensity decreases with
the increase in Zn content. Using the HighScore Plus software package to analyse the
diffraction patterns, a broad background was added (on top of the existing one) to simulate
the diffuse peaks of the carbon support. Estimates of the mean Cu crystallite sizes as
determined from PXRD with the Scherrer equation using this approach are collected in
Table 5.3.
In using the Scherrer equation (Equation (3.6)), we assume that the broadening of the
diffraction profile are only due to limitations of the crystalline domain. Two definitions of the
peak profile width can be used: the full width at half maximum intensity (FWHM), referred
to as β 1/2 , and the integral breadth β i (total area under the line profile divided by the line
intensity at maximum). The latter not only takes better into account the small crystallites
that gives rise to the ’wings’ of a line profile such as those seen in Figure 5.5, it also gives
the volume­weighted average size (147, 148, 187), which can also be extracted from
complementary particle studies. While the measure of β i might be the preferred choice, its
determination requires very careful evaluation of the tails of the peak and the background.
The careful evaluation of the background is, as mentioned above, difficult for carbon­
supported nanoparticles, which is why the crystallite sizes from PXRD are considered
rough estimates. The average size using the integral breadth, β i , as a measure of peak
width is lower than when using the FWHM, β 1/2 . Rietveld refinement of the diffraction
patterns of (a)­Cu/C and CuZn­0.05/C shown in Figure 5.5 reveals smaller mean sizes of
the Cu crystallites. For the un­promoted (a)­Cu/C, Rietveld refinement yielded estimates
of 36 nm and 25 nm using β 1/2 and β i , respectively. Rietveld refinement is somewhat of
an improvement over using the Scherrer equation, since the Rietveld refinement includes
considering the contribution from the carbon support, taking into account the instrument
broadening, and considering a set of structural models. For the CuZn­0.05/C sample,
refinement using models for Cu, Zn and their most common oxides yielded size estimates
of 38 nm (FWHM) and 27 nm (integral breadth), respectively. Supporting information for
the Rietveld refinements are found in Appendix F.6.

As expected from visual inspection of the XRD patterns in Figure 5.5, the calculated mean
Cu crystallite size decreases with increasing Zn loading. Pure Cu on carbon support show
a mean Cu crystallite size of 60 nm. At the two highest Zn­loadings investigated, it was

85
Chapter 5. Carbon­supported Cu/Zn as a Model System

not possible to estimate the mean nanoparticle size of Cu. The trend of ZnOx supporting
the formation of small Cu nanoparticles is found in similar studies of the CuZn system, and
perfectly illustrates the structural promotion of ZnOx as a support (52, 64, 182). While the
trend seen for dCu is to be expected, the mean sizes are larger than expected if compared
to similar studies of carbon­supported Cu with similar Cu loadings (52, 134).
As discussed in literature of Pd/C, a diffraction line profile that is tall (i.e. sharp and in­
tense peaks) with extended wings corresponds to a wide distribution of crystallite sizes,
whereas broad and flat line profiles corresponds to narrow size distributions (148, 188).
The diffraction line profiles of Figure 5.5 indicate that both cases are found for this collec­
tion of Cu(Zn)/C samples.

5.2.2 Local information: HAADF­STEM coupled with elemental analysis

In Figures 5.6a to 5.6d, HAADF­STEM images of used CuZn/C catalysts show visible
nanoparticles of high contrasts compared to the ’cloudy’ support, in various sizes depend­
ing on the nominal Zn/Cu ratio. The images are representative of how all of the images
of post­mortem samples look, with particles evenly spread across the support material.
Most notably is the large particles images in the CuZn­0.05/C sample. Here, we observe
several particles in the 100 −200 nm range are, suspected to be large agglomerated or
sintered particles of Cu (Figure 5.6a). Although chemical mapping was not performed
for this particular sample, it is thought unlikely to be ZnOx since the nominal Zn loading
is low in this sample. Besides this sample, particles imaged using HAADF­STEM were
generally estimated to be in the 10 −20 nm range.

It is noted that the scale bar for the CuZn­0.05/C sample is on the scale of 200 nm, i.e.
a factor of 10 compared to the remaining images, illustrating the large size distribution of
particles found in CuZn­0.05/C. While the large particles may contribute little to the total
active surface area, they contribute significantly to the catalytic active mass. They have
not been included in the STEM­derived size distributions as they will lead to overestimated
specific surfaces. The size distributions based on the HAADF­STEM images are given in
the Appendix in Figure F.4. They were determined as described in the beginning of this
chapter: Individual particles were outlined by hand, and two diameters were measured
to give an average diameter under the assumption that the particle are spherical. This is
naturally not always the case, and so an average error estimate of the average nanoparti­
cle diameter is given with the particle sizes in Figure F.4. Only a few STEM­EDX images
were obtained for each sample, hence the HAADF­STEM images obtained separately
from the STEM­EDX images were used instead. From Figures 5.7 to 5.9 and support­
ing STEM­EDX images in the Appendix (Figures F.5 to F.7), it is clear that Zn particles
were somewhat larger and less dispersed compared to the Cu particles. The volume­ and
surface­averaged particle sizes of Table 5.3 therefore exclude any larger particles thought
to be Zn. For future studies, more care must be taken to separate Cu and Zn particles via
STEM­EDX, to achieve a better estimate of size estimates for Cu and Zn ­ especially if
the two species are separate particles and have widely different particle size distributions.

86
 5.2. A Question of Surface Area or Zn Promotion

In Figures 5.7 to 5.9, the STEM­EDX images of post­mortem samples of CuZn­0.31/C,

CuZn­0.62/C and CuZn­1.0/C are shown. The CuZn­1.0/C catalyst had several 10 −20 nm
nm­sized ZnOx particles due to the high total metal loading. The Zn species were at these
high loadings less uniformly distributed over the support surface as seen in representa­
tive STEM­EDX images in Figure F.5b. Single Cu particles, clearly crystalline, were also
observed for this particular sample (Figure F.5a). For all three samples investigated us­
ing chemical mapping, we find Cu and Zn species within the same particle (Figures 5.9
and F.7) or in close proximity to each other (Figures 5.7 and 5.8). The EDX mapping in
Figure 5.9 indicates CuZn alloy formation at higher Zn loadings.

(a) CuZn­0.05/C. (b) CuZn­0.13/C.

(c) CuZn­0.62/C. (d) CuZn­1.0/C.

Figure 5.6: Representative HAADF­STEM images of (a) CuZn­0.05/C (note the x10 scale
bar in this sub­figure), (b) CuZn­0.13/C, (c) CuZn­0.62/C, and (d) CuZn­1.0/C on which
particle distribution of Table 5.3 and Figure F.4 are based on. Images were acquired by
PhD Mads Lützen.

87
Chapter 5. Carbon­supported Cu/Zn as a Model System

Figure 5.7: STEM­EDX image of Cu and Zn particles in close proximity in the post­mortem
(after CO2 /CO/H2 , and exposed to air) sample of CuZn­0.31/C. Images were acquired by
PhD Mads Lützen.

Figure 5.8: STEM­EDX image of single particles in the post­mortem sample of CuZn­
0.62/C where Cu and Zn are in close proximity. Images were acquired by PhD Mads
Lützen.

Figure 5.9: STEM­EDX image of a single Cu/Zn particle in the post­mortem sample of
CuZn­1.0/C. Notice the larger scale bar of 25 nm. Images were acquired by PhD Mads
Lützen.

88
 5.2. A Question of Surface Area or Zn Promotion

5.2.3 Bulk­averaged, but sensitive to local structure: X­ray TS/PDF

To complement the mean crystallite sizes determined via PXRD, we make use of X­ray
total scattering to circumvent the problem of having a broad background coming from the
support. Total scattering data are rich in nanoscale structural information, and with proper
refinement of a model ­ much like what is done in Rietveld refinement ­ both global and
local structure can be extracted. As introduced in Section 3.4, the PDF represents a his­
togram of all interatomic distances in a sample. From the PDF, the following can (ideally)
be extracted: The peak positions tell us about the interatomic distances between pairs of
atoms in the sample; The peak intensity is related to the number of pairs of a certain type
as well as the atomic number of the elements in the pair; The peak width tells us about
the positional disorder of atoms in the structure.
In the following, the refinement is kept to a minimum, and modeling is focused on the
’model­free’ analysis of total scattering data. Partly because it is a time­consuming task,
but also because the data set used here is that of the samples from Figure 5.3 that, as dis­
cussed earlier, have different sample histories. Extensive refinement was not considered
relevant for this collection of samples.

(a) (b)

Figure 5.10: Total scattering data of post­mortem Cu(Zn)/C samples: (a) reduced total
scattering data structure functions and (b) reduced atomic pair distribution functions.

As the first step, the total scattering data of all the samples given in Table 5.3 are presented
in both reciprocal space as F(Q) and real space as G(r) in Figure 5.10. Sharp scattering
peaks in reciprocal space in Figure 5.10a are observed for the first three samples, namely
(a)­Cu/C, CuZn­0.05/C and CuZn­0.13/C. This suggests large metallic nanoparticles, in
accordance with the PXRD patterns of Figure 5.5. The Zn­rich samples shows broad
oscillations in Q­space, which is typical for structures without long­range periodic order.
Looking at the total scattering data in real space in Figure 5.10b, the similarity of the
PDFs in the low r region tells us that locally, the structure of Cu(Zn)/C across the Zn/Cu
ratio range is the same.

89
Chapter 5. Carbon­supported Cu/Zn as a Model System

Model­free analysis of two of the PDFs of Figure 5.10b, namely the (a)­Cu/C and the
CuZn­1.5/C catalysts, are given in Figure 5.11. Comparing the PDFs to calculated PDFs
of models reveal that all peaks of the (a)­Cu/C sample are explained by FCC Cu, as
expected from PXRD of the same sample, and additionally a CuO phase. Calculated
PDFs of Cu, CuO, and Cu2 O, along with the experimental PDF of (a)­Cu/C, is given in
the Appendix. The first peak at 2.53 Å is associated with the nearest Cu neighbour, the
next peak at 3.62 Å with the second­nearest, the peak at 4.4 Å with the third­nearest, and
so on. The peak positions of the FCC Cu model, as well as the most dominating ones of
CuO, are highlighted in Figure 5.11a. For the highest Zn/Cu ratio investigated here, the
experimental PDF of CuZn­1.5/C can be explained by comparing it to calculated PDFs of
Cu, CuO, and ZnO. The position of the most prominent peaks of the calculated HCP Zn
PDF is indicated to show that while one peak at approximately 8.5 Å might be contributed
to crystalline Zn, the rest of the peaks are not present.

(a) (a)­Cu/C (b) CuZn­1.5/C

Figure 5.11: Model­free analysis of the experimental PDFs of (a) (a)­Cu/C and (b) CuZn­
1.5/C with indicators for Cu and Zn references, and their respective oxides, taken from
calculated PDFs.

Using the knowledge from the model­free analysis of pure Cu/C and Zn­rich CuZn/C sam­
ples, it is clear from the collection of experimental PDFs in Figure 5.10b that at higher Zn
loadings, the Cu­ and Zn­ oxides start to dominate the PDF. While the PDFs indicate
separate Cu­ and Zn­ oxide particles, there is also no evidence of bulk CuZn alloy. This
observations is consistent with the findings of the STEM study in which Cu was found to
be dispersed over the support, while Zn was less dispersed. As the oxide phases domi­
nate the PDFs, these are not useful for determining the state of Zn throughout the series
of catalysts. For the specific study of the surface interactions of Cu(Zn) nanoparticles on
a support, in­situ difference PDF (d­PDF) would without a doubt be more suitable than
ex­situ PDF studies of post­mortem samples (189). This method can be used to study
the structure of the active catalysts and structural changes occurring during reactions
conditions by for example studying the local structure of Cu (89).

90
 5.2. A Question of Surface Area or Zn Promotion

Figure 5.12: Experimental ex­situ PDFs of post­mortem samples of Cu(Zn)/C.

Figure 5.12 shows the same PDFs as given in Figure 5.10b but in an r­range from 1.5 Å to
70 Å. Zooming out to examine the long­range order of the Cu(Zn)/C samples, it is revealed
that PDF peaks extend further out in r­space for samples with low Zn/Cu ratio, indicating
structures that have long­range order. Without refining the PDFs, we see an average long­
range structure that varies from an estimated 6 nm for (a)­Cu/C to approximately 2 nm for
CuZn­1.5/C. In line with PXRD and STEM studies the mean particle size decrease as a
function of Zn loading. While total scattering/PDF directly measures a structure it also
probes the whole sample, which is why the method gives an average measure of the
nanoparticle size. Without further refinement of the individual phases that gives rise to
the total scattering pattern, it is not possible to conclude on the individual sizes of Cu and
Zn from the PDFs. Examples of refined PDFs are found in Appendix F.5. The one­phase
fit of (a)­Cu/C gives an average nanoparticle size of 6.3 nm, and using only one phase
for FCC Cu gives a fairly good fit with an Rw factor of approximately 0.2 with only minor
features left to be fitted by a CuO phase. The refined parameters for this sample are given
in Table F.1. One should be careful with the interpretation of Rw ­values as only highly
ordered, crystalline models, such as LaB6 or CeO2 , can be modeled to give Rw values of
1­5 %. Fits of disordered materials such as those studied here are often reported with Rw
values above 15 % (89). Hence, it is important to do a visual inspection of the fit, as any
large features in the difference curve indicate that something is missing from the model.

PDF is sensitive to the crystallite size, which is smaller than the particle size that is esti­
mated by STEM. Comparing the particle size distributions obtained from HAADF­STEM
with the crystallite sizes estimated from X­ray total scattering, there is an obvious discrep­
ancy ­ sizes from X­ray TS are consequently smaller than those from STEM. The crystallite

91
Chapter 5. Carbon­supported Cu/Zn as a Model System

size may be smaller than the nanoparticle size if the particle is polycrystalline, or if there
is a significant rearrangement of the atoms at the surface of the particle resulting in the
formation of an amorphous layer (186). If the latter is the case, the particle size extracted
from the PDF patterns will be smaller than the actual size, since structural rearrange­
ments at the surface will cause the oscillations of the PDF to decay at shorter distances
(186, 190–192). Nonetheless, the HAADF­STEM images shown in Figures 5.6c and 5.6d
show particles of sizes smaller than the size distributions indicate, some of which closer
to the crystallite domains estimated from the experimental PDFs of Figure 5.12. This is
simply because the size distributions are based on particles where one can with certainty
distinguish between and ’cloudy’ support and the boundary of the particle. The smaller
the particle, the more difficult it was to obtain a clear measure of the diameter.

92
 5.2. A Question of Surface Area or Zn Promotion

5.2.4 Separating the effects of surface area and Zn promotion

The results from the studies of post­mortem samples of Cu(Zn)/C catalysts from Sec­
tions 5.2.1 to 5.2.3, along with the nominal Cu and Zn weight loadings, are compiled in
Table 5.3. Average Cu crystallite sizes of the catalysts in their post­mortem states have
been determined from the X­ray diffractograms. By using the integral breadth instead of
the FWHM as a measure of peak width, we arrive at the volume­weighted average crys­
tallite size, which is more in line with the STEM­derived volume­averaged number also
given in the table. Particle sizes extracted from PDF analysis of X­ray total scattering
data, which are a measure of the crystallite domain rather than a measure of the parti­
cle size, are considerably smaller than the volume­ and surface­averaged particle sizes
derived from HAADF­STEM study.

Table 5.3: Properties of selected Cu(Zn)/C catalysts used in experiments with glass reac­
tors. Asterisks are used to highlight the samples that have undergone a different activity
testing.

Sample name Cu / Zn // wt% a dCu // nm b Particle size // nm c dV , dS // nm d

(a)­Cu/C 5/­ 60 (39) <6 ­
CuZn­0.05/C 5 / 0.26 64 (39) < 5.5 39 ± 24, 29 ± 19
* CuZn­0.13/C 5 / 0.63 48 (28) < 5.5 ­
* CuZn­0.31/C 4.9 / 1.5 20 (11) < 4.5 14 ± 5, 13 ± 4.9
CuZn­0.62/C 4.9 / 3.0 7.8 (6.2) <4 15 ± 7, 13 ± 6.7
CuZn­1.0/C 4.8 / 4.9 ­ <3 12 ± 4, 11 ± 3.5
* CuZn­1.5/C 4.6 / 7.2 ­ <2 ­

a Nominal loading of the final catalyst assuming metallic Cu and Zn.

b Mean Cu crystallite size as determined by PXRD using β 1/2 and (β i ) of the Cu(111)
diffraction peak of the post­mortem catalyst.
c Estimated from X­ray TS PDF. A measure of the crystallite domain.

d HAADF­STEM­derived volume­averaged (d ), and surface­averaged particle sizes

V
(dS ). Error estimates for the volume­ and surface­averaged particle sizes, ± sV and
± sS respectively, are calculated as the standard deviation of the population (σ N −1 ).

As suggested by the preceding studies of Cu(Zn)/C catalysts with varying amounts of

Zn, the optimal Zn loading is reached at a Zn/(Zn+Cu) ratio above 0.1. A combination of
XRD, X­ray total scattering and TEM reveals that smaller Cu particles are stabilized by an
increasing amount of Zn, suggesting that the trend observed in Figure 5.3 might be a com­
bination of size­ and promoter effects. In Figure 5.13, the (Cu) surface­averaged mass
activity at a reaction temperature of 220 ◦ C is given as a function of the Zn/(Zn+Cu) ratio.
Similarly to what is seen for the mass activity, the surface­averaged activity increases
steeply as a function of Zn loading until an optimum is reached. After the optimum at
Zn/(Zn+Cu)=0.1, the activity decreases with increasing Zn loading. As discussed in Sec­
tion 4.3.3 where this method of normalizing the mass activity with the dispersion of surface

93
Chapter 5. Carbon­supported Cu/Zn as a Model System

Cu atoms was first introduced, the value of dispersion is based on the measure of the Cu
crystallite size which in turn can be dependent on the treatment during activity testing.
Unfortunately, two of the samples that give rise to the trend in Figure 5.13 were treated
differently during activity testing, which at the end might have changed the trend in a way
that is yet to be uncovered.

Normalization of the methanol mass activity with the measure of surface Cu atoms yields a
trend that is similar to the observations of Dalebout et al. (64) and van den Berg et al. (52).
Both of the two studies report similar Cu particles across their respective Zn/(Zn+Cu) se­
ries, albeit reported with fairly large standard deviations. While Dalebout et al. report that
the highest methanol TOF was obtained for carbon­supported catalysts with Zn/(Zn+Cu)
molar fractions between 0.15 and 0.25 (64), taking the upper and lower error estimates
into account, the optimal molar fraction is more that of 0.05, which is more in line with the
previous study of the same group (52). Comparing the two studies with the observations
made in this thesis indicates that the optimal Zn/(Zn+Cu) might depend on the Cu particle
sizes obtained across the Zn/(Zn+Cu) series, but also the amount of Zn promoter.

Figure 5.13: Surface­averaged mass activity of selected Cu(Zn)/C catalysts at 220 ◦ C, all
from activity testing in glass reactors. Reaction conditions are those stated in Figure 5.3.
Mass activities are normalized by the mean Cu crystallite size obtained from PXRD.

5.3 Modifications to the Synthesis and Testing of Cu(Zn)/C

The experimental setup used to evaluate the activity and stability of catalysts underwent
several rounds of renovations throughout this project. In connection with this, the practice
of testing catalysts was also changed from using glass reactors to using glass­lined steel
reactors. The change was made as the type of connection that is used between the glass

94
 5.3. Modifications to the Synthesis and Testing of Cu(Zn)/C

reactor and the steel tubing of the rest of the setup is prone to leak. Although there where
no signs of leaks for the data in Figure 5.3, a change to steel reactors was necessary to
establish a trust­worthy and reproducible procedure for future testing of catalysts.
An important reason for making changes to the experimental setup is what lays behind
the trend seen for the activity during two consecutive rounds. In Figure 5.4a, and more
obviously when quantified in Figure 5.4b, we see an increase in mass activity of methanol
over two temperature cycles. Initially, one might think this is due to additional promotion
if the active site changes over time. Looking further into the experimental parameters, it
was discovered that the injection pressure was different for the two temperature cycles,
despite not removing the glass reactor from the setup in between cycles, nor changing
the total mass flow. At the time of this experiment, the injection pressure was measured
at the front of the GC. The renovation of the experimental setup is described in detail
in Appendix E.2, but the main conclusion is this: Normalising the FID signals with the
pressure measured on the back of the GC, which represent the pressure at the FID better
than the pressure measured at the front of the GC, yields a more accurate measure for
the methanol signal. Adding a second pressure measurement effectively enabled us to
separately normalise TCD (reactants) and FID (products) signals.

Figure 5.14: Catalytic mass activity of methanol on carbon­supported Cu and CuZn parti­
cles as a function of Zn loading, using glass­lined steel reactors. Reaction conditions are
50 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of 6 × 105 Ncm3 · h−1 · g−1 Cu . Note
the logarithmic scale. Lines are inserted to guide the eye, and uncertainties are estimated
by propagation of error (121).

After establishing an experimental procedure for the testing of carbon­supported Cu(Zn)

particles that is reproducible, the entire series of Cu(Zn)/C catalysts with an increasing

95
Chapter 5. Carbon­supported Cu/Zn as a Model System

Zn/Cu ratio was reproduced. The data shown in Figure 5.14 are all from tests using
glass­lined steel reactors, which are less prone to leakage as there are no glass­to­steel
connections. As the steel reactors have a smaller inner diameter, the Cu(Zn)/C sample
amount was reduced by 50 % compared to the experiments with glass reactors. The
flow of gas was matched to the sample amount to maintain the WHSV per weight. The
GHSV was also maintained by diluting the catalyst sample with inert SiC dilutant. Com­
paring the data from two separate experiments using glass and glass­lined steel reac­
tors in Figure F.2, the methanol mass activity of the CuZn­0.06/C catalyst is considered
quite similar for the two experiments, especially when considering the overlapping error
estimates. However, this is not the case for all samples across the Zn/(Zn+Cu) range.
While the trend of an increasing activity as a function of Zn loading is still evident from
experiments involving steel reactors, the apparent optimal amount is shifted to a higher
Zn/(Zn+Cu) ratio of approximately 0.2. The difference in mass activities between the ex­
periments involving glass reactors and those involving glass­lined reactors is highlighted
in Figure 5.15a. Here, the mass activities at 220 ◦ C from Figures 5.3 and 5.14 are plotted
together for easy comparison. While the CuZn/C samples show similar activities above
a Zn/(Zn/Cu) ratio of 0.2, that is not the case below this ratio. Here, the data from Fig­
ure 5.3 seem to be outliers when compared to the data obtained from experiments with
glass­lined steel reactors. While the difference might be due to differences in the Cu­Zn
surface or due to Cu surface area, it might as well be setup­specific and not related to the
CuZn/C samples.

As discussed in Section 4.3.4, the optimal case to study a promotional effect is to obtain
similar size distributions across a range of X­promoted Cu nanoparticles. That way, it
would be possible to completely separate the effect of a promoter from the effect of surface
area. With conventional incipient wetness impregnation, Cu particle sizes of different
sizes were obtained. If all of the Zn in the CuZn/C catalysts were to segregate to the
surface of the Cu particle, varying Cu particle sizes across the Zn/(Zn+Cu) ratio means
that the Zn coverage at the surface does not increase in the stepwise manner we hope
for, as it is shown by Kuld et al. that the coverage of Zn at the surface of Cu varies with
particle size (51). This would make it difficult to observe the effect of Zn coverage that
Nakamura et al. reports (41).

In another effort to separate the size­effect from the promotional effect, a sequential incip­
ient wetness impregnation (seq. IWI, abbrev.) method has been applied. A 5 wt% Cu/C
sample prepared via conventional IWI (conv. IWI, abbrev.) was reduced in 50 % H2 at
220 ◦ C. PXRD of the as­reduced Cu/C indicated a mean Cu crystallite size of 3.3 nm. A
Zn precursor was subsequently impregnated on the reduced Cu/C sample with the result­
ing powder subsequently being prepared and tested as described at the beginning of this
chapter. Three samples were prepared following the sequential IWI. Figure 5.15a com­
pares the series of Cu(Zn)/C catalysts prepared via conv. IWI and seq. IWI. The activity
of the samples prepared via seq. IWI generally follows the trend of the conv. IWI samples,
with the exception of the catalysts with a nominal Zn/Cu ratio of 0.06. PXRD of the first

96
 5.3. Modifications to the Synthesis and Testing of Cu(Zn)/C

four post­mortem Cu(Zn)/C samples reveal a bigger spread in mean Cu crystallite size
for the conv. IWI samples than for the seq. IWI samples. Figure F.17 and Figure F.16
in the Appendix show the PXRD patterns with the corresponding Cu crystallite sizes as
determined using the FWHM with the Scherrer equation.

(a) (b)

Figure 5.15: Methanol mass activity of Cu(Zn)/C catalysts. Sub­figure (a) shows the
activity of Cu(Zn)/C catalysts as a function of Zn/(Zn+Cu) ratio at a temperature of 220 ◦ C.
The data for ’Conv. IWI, glass’ is the same as that in Figure 5.3, and that for ’Conv. IWI,
steel’ is from Figure 5.14. Sub­figure (b) shows the activity during a temperature cycle
for the two CuZn­0.06/C catalysts prepared via conventional IWI and sequential IWI, both
tested in glass­lined steel reactors.

Figure 5.15b shows the methanol mass activities of the (b)­ and (c)­CuZn­0.06/C cata­
lysts during a single temperature cycle. Here, (b)­ refers to the conventional IWI, and (c)­
refers to the sequential IWI. The difference in methanol mass activities is more distinct
on a linear scale, and it is clearly seen that the (c)­CuZn­0.06/C catalyst outperforms (b)­
CuZn­0.06/C, especially during the ramp down in temperature. The diffraction line profiles
of the post­mortem samples indicate wide size distributions, just like what is seen for the
samples in Figure 5.5, and the mean crystallite sizes are 43 nm for (b)­CuZn­0.06/C and
75 nm (c)­CuZn­0.06/C. BET specific surface areas of the catalysts in their as­synthesized
(calcined) states are 959 ± 18 m2 /g and 1004 ± 18 m2 /g, respectively. Normalization of
the methanol mass activity with the measure for Cu dispersion (as based on PXRD size
estimates) does not change the picture of Figure 5.15b, hence it is not shown here. The
specific surface areas are so similar in value, especially when the uncertainty is taken
into consideration, that the trends in Figure 5.15 are not thought to be a matter of surface
area only.
A hypothesis explaining the enhanced methanol mass activity when applying the sequen­
tial IWI route relates to the role of formate and its relation to Cu surfaces. It is largely
accepted that methanol is formed over a formate­pathway on the industrially relevant
Cu/ZnO/Al2 O3 catalyst, and that the hydrogenation of formate is the rate­determining step

97
Chapter 5. Carbon­supported Cu/Zn as a Model System

during CO2 hydrogenation (27, 33–35, 55). Later studies by Nielsen et al. support this by
suggesting that the coverage of surface formate species is representative of the true rate
of the catalytic cycle in Cu­catalyzed methanol synthesis (193). From this, it is hypothe­
sized that by impregnating the Zn formate on Cu particles, the Zn is preferably impreg­
nated on the Cu surfaces instead of on the support, since Cu is known to stabilize formate
species. If found to be true, this effectively helps to disperse the Zn more so than for the
conventional IWI synthesis. The effect would naturally be largest for the CuZn/C catalysts
of very low Zn loading, why future studies of the sequential impregnation method should
cover more samples in the range below a Zn/(Zn+Cu) ratio of 0.2. For these samples, it
would be interesting to investigate the dispersion and size distribution of Zn with STEM­
EDX. If the hypothesis is correct, we would expect to see a change in the dispersion of
Zn.
The modifications to the conventional IWI method of preparing Zn­promoted Cu/C cat­
alysts are preliminary, and so besides studying the dispersion of Zn, future studies are
aimed at elucidating the effect of calcination temperature, the size distribution of Cu, and
the state of ZnOx at the optimal Zn loading. Supported nanoparticles are prepared us­
ing a range of different techniques in the literature, why there is many different routes to
take in optimizing the synthesis of Cu(Zn)/C (194). Some studies report a loss of surface
area during the calcination procedure (195, 196), hence a milder calcination procedure
might be beneficial for the carbon­supported Cu(Zn) particles investigated here. We also
note that by carefully controlling the drying and calcination steps, homogeneous size dis­
tributions can be obtained, specifically with the use of an N2 atmosphere (197–199). In
addition to this, research groups have reported on the idea of support functionalization to
retard Ostwald ripening, which is one reason for catalyst deactivation, both for SiO2 and
various carbon supports (140, 200, 201). A combination of functionalization of the carbon
support and control of the drying/calcination steps would be highly beneficial for the study
of Zn promotion in Cu(Zn)/C samples, as it would be more easy to separate the effect of
Cu surface area from that of Zn promotion in samples with homogeneous size distribution
across the Zn/(Zn+Cu) range.

5.4 Outlook and Future Work

The Cu/ZnO and Cu/ZnO/Al2 O3 systems are the most intensively studied catalyst systems
by in­situ methods (39, 40, 56–58, 202–204), while in­situ studies of carbon­supported Cu
and CuZn particles relevant for the present study are fewer (64). The post­mortem state
of a catalyst that has been exposed to air is not the optimal state to relate to the activity of
the catalyst, and so it is worth investigating how the state of the catalyst changes during
the different stages of operation. During this PhD, a 5 wt % Cu/C catalyst was used
to study the effect of reduction and synthesis gas on the state of carbon­supported Cu
particles using the in­situ XRD as described in Section 3.3.2. As the study turned out,
the XRD with in­situ capabilities was not suitable for this kind of study as the oven itself
showed catalytic activity during activity testing in synthesis gas. The in­situ study of Cu/C
is discussed in­depth in Appendix F.3 in the Appendix. The work of optimizing the in­situ

98
 5.5. Conclusion on Carbon­supported Cu/Zn as a Model System

XRD for studies involving reducing gases is ongoing, but beyond the scope of this thesis.

Cu is regarded as the active metal for CO2 hydrogenation based on the linear relation­
ship between the methanol activity and the specific Cu surface area (18, 205, 206). The
specific Cu surface area is therefore an important parameter for the methanol activity and
would be fitting to use to normalize activity data as those given in Figures 5.3 and 5.14.
The chemisorption methods H2 ­TPD and N2 O­RFC, as described by Chinchen et al. (207)
and Muhler et al. (204), are usually applied to obtain this information. Based on the inter­
action between a probing molecule and the sample surface, and assuming the interaction
is highly selective, the methods provide estimates of the Cu dispersion. The current con­
sensus seems to be that both methods should be used when estimating the specific Cu
surface area, especially for oxide­supported and/or Zn­containing Cu catalysts, and that
for Zn­free, carbon­supported Cu catalysts, the N2 O­RFC and H2 ­TPD methods demon­
strate good correspondence (127, 208). N2 O­RFC measurements done at DTU Chemical
Engineering gives an estimate of 5.4 m2 /g for a 5 wt% Cu/C catalyst in the reduced state,
which is in the same range as those measured for Zn­free Cu­based catalysts (163, 193,
209, 210).

Future studies of the carbon­supported Cu and CuZn nanoparticles should focus on in­
vestigating the left side of the Zn/(Zn+Cu) row where the optimum is not reached. While
there is still plenty of work to be done in optimizing the catalyst preparation, the reduction
procedure and the gas feed mixture, the most important point is to elucidate the state of
Zn before, during, and after the optimal Zn loading has been reached. Instead of focusing
on the apparent best Zn loading, it would be highly interesting to investigate the role of Zn
across the range of Zn loadings to establish whether or not all of the ZnO is systematically
reduced to Zn, and incorporated in the Cu surface as a surface alloy, until a maximum is
reached, and the rest of the ZnO is present as spectator species.
To bridge the pressure gap and display the relevance of surface­sensitive experiments for
the high­pressure regime, quasi in­situ XPS and ISS characterization in combination with
a high­pressure reactor cell (HPC) should be done, much like Divins et al. has done for
SiO2 supported CuZn particles (76), but using a weakly interacting carbon support.

5.5 Conclusion on Carbon­supported Cu/Zn as a Model

System
Catalysts of Cu(Zn) nanoparticles on mesoporous carbon were prepared to better sepa­
rate the electronic and structural effect of the ZnO promoter. With only modest amounts
of Zn promoter, the methanol formation for the CuZn/C was significantly faster than for
a Cu/C catalyst in a CO2 /CO/H2 feed. With a stepwise increase in the loading of Zn
compared to Cu, which was kept at 5 wt%, we were able to see a promotional effect
up until a certain Zn weight loading. The highest mass activity for experiments using
glass reactors were obtained for a carbon­supported catalyst with a Zn/(Zn+Cu) fraction
of 0.1. PXRD of the catalyst samples after they had been tested in a CO2 /CO/H2 feed
revealed clear differences in the Cu crystallite sizes, indicating the role of ZnO as a struc­

99
Chapter 5. Carbon­supported Cu/Zn as a Model System

tural promoter. However, the broad X­ray diffraction profile of the mesoporous carbon
support combined with the broad diffraction profile of nanosized particles complicates the
determination of crystallite sizes. To circumvent this issue, we combined three ex­situ
methods: Conventional PXRD, HAADF­STEM imaging combined with energy­dispersive
X­ray spectroscopy, and PDF analysis of X­ray total scattering data. All three methods
confirmed a decrease in Cu crystallite size and Cu particle size distribution with an in­
creasing Zn loading. PDF analysis of total scattering data revealed crystallite domains
down to approximately 2 nm for the CuZn/C sample with a Zn/Cu ratio of 1.5. For the
CuZn/C catalysts on the left side of the optimal Zn/(Zn+Cu) ratio, we estimate crystallite
domains of approximately 6 nm. Where PXRD greatly overestimates the mean Cu crys­
tallite size compared to X­ray total scattering, STEM serves as an intermediate between
the two methods. STEM imaging furthermore revealed both large (possibly sintered) and
smaller, more dispersed, Cu nanoparticles on the post­mortem sample of CuZn­0.05/C.
The trends observed across the range of Cu(Zn)/C catalysts show that a combination
of bulk­averaged methods (XRD, X­ray total scattering) and methods that probe locally
(STEM, X­ray total scattering) is needed to properly elucidate the morphology and parti­
cle size distributions of Cu and Zn. Investigations of the range of Cu(Zn)/C samples while
using an optimized experimental procedure revealed a similar trend across the range of
samples, but with a different optimal fraction of Zn to Cu, indicating that the experiments
with glass reactors were outliers.
As the Cu particle sizes differ across the Cu(Zn)/C sample range, the observed trend in
mass activities is thought to be a combination of promotional and size effect. A differ­
ent synthesis for the carbon­supported Cu(Zn) catalyst was therefore investigated. Us­
ing a pre­treated 5 wt% Cu/C samples as a precursor for the CuZn/C catalyst, a series
of Cu(Zn)/C samples with low Zn/Cu ratios were prepared using a sequential IWI syn­
thesis. PXRD of the post­mortem samples prepared via sequential IWI revealed quite
similar Cu crystallite sizes for the Zn­promoted samples. Additionally, they showed en­
hanced methanol mass activities compared to those prepared via conventional IWI, which
is thought to be due to a better dispersion of Zn on the Cu surface, consequently promoting
the effect of Zn.

100
Chapter 6

Optimizing a UHV Chamber for

Surface­Sensitive Studies
The following chapter scratches the surface of the task that is reviving a UHV setup that
has been out of use for some time. Although the setup did not contribute with data to the
previous chapters, getting to know the workings of it and optimizing it for future use has
taken up a lot of my time the past years. The challenging, but rewarding, journey it has
been working with this setup and what I have learned from it deserves to be described
in its own chapter. Now, a foundation for more experiments has been established, with
experimental procedures in place to support the testing and characterization of supported
nanoparticles.
This chapter would not have been possible without the practical help and guidance of the
SurfCat floormanagers. Student Mathias Ribergaard Vinther was the main contributor
when it came to implementing the new positioning system. Olivia Sloth is thanked for
her day­to­day practical help, and the newest team­member on team Munin, Jonathan
Schouenborg, is acknowledged for wanting to take over and optimize the setup.

The experimental setup, originally named the parallel screening setup, is invented by
Department of Physics and Department of Chemical Engineering, DTU, and was first
described in literature in 2004 (211). At its early days, it was used for the simultane­
ous electron­beam deposition of metals (up to four metals) which were then individually
treated and analyzed using a range of (surface sensitive) methods. The setup, as it was
originally build, is described in­depth in the 2004 paper, why we in the following only will
go into detail with the modifications made.
The setup is a combined high­pressure cell­ultrahigh vacuum (HPC­UHV, abbrev.) sys­
tem consisting of three parts: the sample preparation chamber, the surface analysis
chamber, and the high pressure cell. The sample is mounted at the end of a manipulator
arm which allows the sample to be moved throughout the three parts of the setup. An ultra­
high vacuum lock specially designed for this setup ensures that the preparation chamber
and the surface analysis chamber will remain at a pressure in the UHV range when the
high­pressure cell is filled with gas at atmospheric pressure. A knife on the manipulator
arm ensures that this is also the case when the sample has reached the high­pressure
cell at the end of the setup. The advantage of the setup is the possibility for alternating
between preparation and surface analysis of a catalyst sample in ultrahigh vacuum, and
testing of the catalytic activity at high pressure. All this is done without exposing the cat­
alyst sample to ambient air, effectively making the measurements quasi in­situ. One of
the research papers that came out of the setup includes, but is definitely not limited to,

101
Chapter 6. Optimizing a UHV Chamber for Surface­Sensitive Studies

the 2014 paper by Kuld et al., who showed that the chemical state of Zn and Cu depends
on the pre­treatment conditions, which, combined with quantification of the chemical sur­
face, suggested a surface­alloy opposed to a ZnO overlayer (50). Based on this idea,
a model for the reduction and segregation of Zn onto a Cu NP was formulated and cor­
related to the activity of a Cu/ZnO/Al2 O3 catalyst, showing that an increased reduction
potential during the pre­treatment produces more Cu­Zn surface alloy (51). These two
studies demonstrate the importance of testing catalysts under realistic conditions, while
maintaining the ability to study them with surface science techniques both prior to and
after relevant reaction conditions, to get as close to studying the workings of the catalyst
while in operation.

6.1 Setting the stage

The literal backbone of the HPC­UHV setup is the motion of the manipulator in the x, y and
z directions. Motion was originally controlled by Oriental α­STEP stepper motors which in
turn were controlled by a LabVIEW program. Limit and home switches were mounted on
the manipulator and read by the original LabVIEW motion control program. Upon learning
that one of the two motor drivers for controlling the motion in the z­direction was out of
order, and that no repairs could be made to the driver, nor a new one was available, it was
decided that the existing motors and drivers should be replaced. To ease the transition, the
specifications of the new stepper motors were kept the same as the old. In practice, that
meant purchasing closed loop stepper motors with resolutions of 0.36°/step with similar
holding torques, ranging from 0.3 Nm for the y­axis and 2.0 Nm for the x­axis.

Figure 6.1: The data and command structure of the positioning system of the HPC­UHV
setup. Communication with the Raspberry Pi is through a socket server sending and
receiving positioning data and alarms.

102
 6.1. Setting the stage

The positioning information is managed by the built­in ABZO sensor of the stepper motor,
ensuring that the information is retained in case of unexpected power outages. As the
motor drivers are of the built­in controller type, there is no need to perform a return­to­
home operation before restarting positioning operations after the power has been turned
off normally or unexpectedly. The home position has only been reset once since installing
the new motors, which was at the time of installation. One of the most important aspects
of the motor driver is the possibility of executing items such as motor speed, positioning,
and acceleration/deceleration via either I/O, Modbus RTU/RS­485, or FA network, where
Modbus RTU/RS­485 is the most commonly used type of communication at SurfCat. The
system configuration made for combining stepper motors, the built­in controller type driver
and the master controller is roughly sketched out in Figure 6.1. The drivers are supplied
with the same AC power supply, and for of the four each motor/driver sets, an individual
24 VDC power supply is used. The drivers are linked in series to enable control of all
of them with one master. While the software supplied with the motor/driver set has all of
the functions necessary for positioning within the setup, the software can only control one
motor at the time. As the positioning of the manipulator in the z­direction requires that two
motors move in unison at the exact same speed to avoid twisting the bellow surrounding
the manipulator arm, the supplied software was traded for a self­made driver and graphical
user interface (GUI, abbrev.). The GUI allows control of the motors by sending commands
to the Raspberry Pi which in turn sends commands to the motor drivers. The GUI has
two tabs, with the main one shown in Figure 6.2: One to control the stepper motors and
receive messages, and one to change the motor parameters and preset different locations
within the HPC­UHV system.

Figure 6.2: The main tab of the GUI used to control the stepper motors on the parallel
screening setup.

With a new positioning system also comes a new coordinate system for the HPC­UHV
setup. While the coordinate system remains the same (as indicated in Figure 6.3), the
absolute coordinates are subject to change. Based on the coordinates reported for the
old stepper motor system, estimates for each of the positions throughout the three parts of

103
Chapter 6. Optimizing a UHV Chamber for Surface­Sensitive Studies

the setup were put into the motor/driver software with the goal of determining the optimal
positions for analysis by XPS and ISS, for sputtering of the sample surface, and for safe
positioning when the sample is in the HPC.

Figure 6.3: Schematic of the combined HPC­UHV setup showing the two horizontal
planes (X and Z). The third plane, the Y­direction, goes out of the plane, towards the HSA.
Annotated and reprinted with permission from (211). Copyright 2004 AIP Publishing.

6.2 Back to basics

The equipment for XPS and ISS is aligned in such a manner that it focuses onto the same
point when the sample is turned 90° to face the hemispherical energy analyzer (HSA,
abbrev.). The HSA used for XPS and ISS measurements is a SPECS Phoibos 100 with
single channel detector. X­rays are generated by a SPECS RQ 20/38 X­ray gun with a
dual Al/Mg anode that has been recently re­coated. A PHI model 04­303 ion gun is placed
at a 130° scattering angle for ISS.
X­ray generation requires the dissipation of a large amount of power by the anode, and the
anode must therefore be water cooled. As a safety precaution, the X­ray source control
does not deliver high voltage if a sufficient water flow is not delivered at the X­ray gun.
For the purpose of having a high enough water flow of approximately 7 L/min, as well
as having a water temperature low enough for the cooling of the anode, a PolyScience
liquid­to­liquid cooler (referred to as a chiller) was installed before any operation of the
X­ray gun was attempted.
Since the X­ray gun had recently been re­mounted after service of the anode, and the X­
ray source control had been replaced, each of the anodes were identified anew. Survey
spectra using either the Al anode (Kα 1486.6 eV) or the Mg anode (Kα 1253.6 eV) were
obtained, and the anodes were subsequently identified from the C 1s peak.

104
 6.2. Back to basics

Figure 6.4: First test sample after replacing the stepper motors on the parallel screening
setup. In the sample cup, specialized for powder samples, two metals foils of Cu and Au
overlap, and a Ag wire is placed across the two foils. The coordinate system relates to
the motion of the manipulator arm.

To roughly sketch out the optimal positions for measuring XPS and ISS, a combination of
Cu and Au foils with a Ag wire going across the foils, was loaded in the cup­like sample
holder. The three metals are easily separated with both XPS and ISS. The sample holder
and the combination of metal foils and wire is shown in the picture and a schematic in
Figure 6.4. The coordinate system of the setup is split into X, Y and Z directions. The X­
direction denotes the horizontal motion along the setup and is the route the sample holder
takes when going through the three parts of the setup from the HPC, through the analysis
chamber, and all the way to the preparation chamber. The Y­direction is the second axis
that denotes motion horizontally to the setup, moving the manipulator towards and away
from the HSA. The X­ and Y­directions are indicated in Figure 6.4. The last direction, the
Z­direction, is the vertical motion of the manipulator arm, as indicated in Figure 6.3. In
Figure 6.4, this direction would go out of the plane of the paper, towards the reader.

Figures 6.5a, 6.6a and 6.7a show the detailed scans obtained in the Ag 3d, Au 4d, and
Cu 2p primary XPS regions. Plots of the absolute intensities of chosen peaks as a func­
tion of the position of the sample are given in Figures 6.5b, 6.6b and 6.7b. The first two
figures follow selected peaks in the Ag 3d and Au 4d XPS regions while the sample is
moved in the X­direction (Figure 6.5b) and the Z­direction (Figure 6.6b). From these, an
obvious optimum for both of the planes is seen, giving an estimated optimal (X, Z) coor­
dinate of (354, 12)mm. The optimum for Cu is estimated to be at a Z position of 11 mm,
which agrees with the Cu foil being placed above the Au foil. Moving the sample towards
a more negative value in the Z­direction will consequently expose more of the Cu foil to
the X­ray beam. Consequently, a first estimate of the optimal position for measuring XPS
is at (X, Z)=(354, 11.5)mm.

105
Chapter 6. Optimizing a UHV Chamber for Surface­Sensitive Studies

A peak at 348.4 eV is apparent in the main XPS region of Ag 3d and Au 4d. The peaks
follow those of all other peaks during the movement of the sample, but the absolute in­
tensity of the peak is higher further away from the optimum, i.e. the center of the sample.
This is true for both the movement in the Z­ and the X­direction. It is therefore thought
to be a contaminant on the stainless steel sample holder, although neither Fe, Ni, nor Cr
(the main components of stainless steel) are obvious in the detailed spectrum, nor the
survey spectrum, of the Cu/Au/Ag sample.

(a) Normalized scans. (b) Absolute peak intensity of chosen peaks.

Figure 6.5: XPS of the Cu/Au/Ag sample. Detailed scans in the region 325 −380 eV as
a function of position in the X­direction, with a starting position of (X, Y, Z)=(348, ­3, 4)
mm. XPS spectra obtained using Al Kα X­rays (12 kV, 20 mA), a pass energy of 40 eV,
and an energy step of 0.4 eV. Pictured spectra are averages of 20 consecutive scans.

(a) Normalized scans. (b) Absolute peak intensity of chosen peaks.

Figure 6.6: XPS of the Cu/Au/Ag sample. Detailed scans in the region 325 −380 eV as
a function of position in the Z­direction, with a starting position of (X, Y, Z)=(356, ­3, 4)
mm. XPS spectra obtained using Al Kα X­rays (12 kV, 20 mA), a pass energy of 40 eV,
and an energy step of 0.4 eV. Pictured spectra are averages of 20 consecutive scans.

106
 6.2. Back to basics

(a) Normalized scans. (b) Absolute peak intensity of Cu 2p peak.

Figure 6.7: XPS of the Cu/Au/Ag sample, with focus on Cu 2p. Detailed scans in the
region 920 −940 eV as a function of position in the Z­direction, with a starting position
of (X, Y, Z)=(356, ­3, 4) mm. XPS spectra obtained using Al Kα X­rays (12 kV, 20 mA), a
pass energy of 40 eV, and an energy step of 0.4 eV. Pictured spectra are averages of 20
consecutive scans.

The sample was loaded when the entire setup was ventilated, and the sample was not
subjected to thorough chemical cleaning before loading. Nevertheless, it served its pur­
pose well enough for estimating the optimal position for XPS measurements. Figure 6.8
show a survey spectrum obtained with the sample positioned in the optimal (X, Y, Z) po­
sition. The primary XPS regions are annotated, and all but one peak can be explained by
the presence of Cu, Au, Ag, C or O. The data insert show the main peak in the C 1s region,
showing that there is an approximately 0.5 eV shift in the peak position. However, it de­
pends which binding energy the C­C component is set to. In this figure, it is set to 284.2 eV
as described by the Thermo Fisher Scientific handbook on XPS and Auger, which gener­
ally refers to reference (212) of Briggs and Seah. The small shift towards higher binding
energies may be indicative of oxidized and/or contaminated metal surfaces, but it may
also be due to differential charging resulting from the rough surfaces the metal foils and
metal wire have may have since they weren’t mechanically cleaned.
The shape of the spectra presented here is common for X­rays produced in a twin­anode
source. This kind of source produces broad X­ray line widths, as well higher background
due to bremsstrahlung radiation. In comparing the spectrum of Figure 6.8 taken with the
non­monochromated dual anode X­ray source to that of Figures B.14 and B.15, which is
measured with a monochromated X­ray source, the difference in background and overall
spectrum quality is clear. While the use of the dual anode X­ray source enables selection
of the most appropriate anode for a specific analysis, some of the quality is sacrificed
when the source is not monochromated. Obtaining good XPS spectra on the combined
HPC­UHV setup comes at the cost of time, as detailed scans have to be averaged over
20 or more consecutive scans to obtain a good resolution.

107
Chapter 6. Optimizing a UHV Chamber for Surface­Sensitive Studies

Figure 6.8: XPS survey (average of 3 consecutive scans) of the Cu/Au/Ag sample in the
energy region 0 −1050 eV at the position (X, Y, Z)=(354, ­0.3, 11.5)mm. Obtained using
Al Kα X­rays (12 kV, 20 mA), a pass energy of 100 eV, and an energy step of 1 eV. The
primary XPS regions of metals and contaminants are indicated, and the arrow indicates
unknown contaminant.

6.3 Outlook: Future work and possibilities

The combination of metal foils and wire proved as a useful sample for testing the newly
installed stepper motors and for mapping out the coordinate system of the HPC­UHV
setup. The next step is to determine whether it is possible to analyze the exact same
area with XPS and ISS, and consequently hard code into the stepper motor driver the
kind of route the sample needs to take going to and from XPS/ISS. For this, a sample with
a small dot of metal in the center of the sample holder should be used, which has been
cleaned thoroughly to ensure a clean and smooth surface, will be the best basis for the
further optimizations.

The preceding sections presents only a small part of the capabilities of the combined
HPC­UHV setup. Although not part of this thesis, efforts have been made to optimize the
position for He­based ion scattering spectroscopy (ISS, abbrev.) analysis of the Cu/Au/Ag
sample, as well testing Ar­ISS as a way of sputtering the sample in between measuring
XPS spectra. A main issue with the ion gun on the HPC­UHV setup is that the set beam
voltage does not correspond to the measured voltage, consequently complicating the

108
 6.3. Outlook: Future work and possibilities

analysis of ISS spectra. Future efforts should be aimed at installing an external measure
of voltage, instead of depending on the (at times unreliable) analog control unit. If the
energy and mass of the primary ion is known (Ep ,Mp ), as well as the scattering angle θ,
the only unknown variable is the mass of the surface atom, Ms . Via the governing equation
of ISS, this can be determined using the measured energy of the scattered ion, Es :

 √(
)2 2
 cos(θ) + Ms
− sin2 (θ)

Es Mp
=  (6.1)
Ep  Ms
1+ M 
p

Given that only charged particles are detected in ISS, any noble gas ions that have ’picked
up’ an electron and therefore will be scattered as neutral atoms will not be detected. This
is where the exceptional surface sensitivity and the strength of ISS lies: Only impinging
ions that are backscattered by the outermost atomic layer of a surface, thereby still having
a charge, are detected. For future studies of surface phenomena, the combination of XPS
and ISS will be extremely useful, and especially so for determining the chemical state of
the relevant fraction of Zn in the model CuZn/C catalyst investigated in Chapter 5.

109
Chapter 7. Conclusion

Chapter 7

Conclusion
As we work towards a CO2 ­neutral future where we need to phase out conventional fossil
fuels, methanol stands out. Not only does it have the potential to be used as a fuel in itself,
the industry behind it also have a potential to ’go green’. Whether we want to produce
renewable methanol from alternative feedstock, or want to embrace the ’methanol econ­
omy’, we need to: 1) Give the catalyst that is conventionally used for industrial methanol
synthesis a makeover that includes optimization of the long­term stability and the selec­
tivity, and 2) formulate new catalytic systems that are active and selective for the low­
pressure, low­temperature process of CO2 hydrogenation.
In this thesis, research have been presented within these two areas. More specifically,
it investigated a Cu/Zn/C model system for the industrial­type Cu/ZnO/Al2 O3 catalyst, as
well as suggested and investigated new catalysts for CO2 hydrogenation.

7.1 Carbon­supported Cu(Zn)

In this study, a model catalyst for the industrial­type Cu/ZnO/Al2 O3 methanol catalyst was
investigated. The use of an inert carbon material as the support, combined with the low
fractions of Zn compared to Cu in the model Cu(Zn)/C catalyst, is thought to help in re­
vealing the speciation of the active fraction of the ZnO promoter under relevant conditions.
Inspired by the early works of of Nakamura et al. (27, 41, 45), the Zn/Cu fraction was
increased stepwise. As the fraction of promoter increased, a promotional effect was ob­
served: The mass activity of methanol consistently increased as a function of increased
Zn loading, until an optimal Zn/(Zn+Cu) fraction was reached at Zn/(Zn+Cu)≈0.2. This
optimal fraction was observed after establishing a reliable experimental procedure for test­
ing Zn­promoted Cu/C catalysts.
A thorough investigation of the range of Cu(Zn)/C catalysts using conventional PXRD,
HAADF­STEM combined with energy­dispersive X­ray spectroscopy for chemical map­
ping, and PDF analysis of synchrotron­based X­ray total scattering data, indicated that
the promotional effect might be partly due to surface area and partly due to Zn promotion.
Though based on post­mortem samples of activity tested Cu(Zn)/C samples, all three
characterization techniques were used to conclude that the Cu particle size decreases
with increasing Zn content. XRD yields an average of the Cu crystallite size that is con­
sistently larger than the average particle size determined with HAADF­STEM. The issue of
determining the crystallite size of nanosized particles with conventional PXRD was solved
by performing PDF analysis of X­ray total scattering data. The PDFs indicate similar local
structure, but the oxides of Cu and Zn dominate the PDFs of the post­mortem samples.
By instead investigating the PDF in a large r­range, information of the long­range order

110
 7.2. New Methanol Catalysts

can be extracted. Estimates of crystallite sizes range between 6 nm for the pure Cu/C to
2 nm for the CuZn/C sample with the highest Zn loading.
Separating the effect of active surface area by normalizing mass activity data with a mea­
sure of the surface Cu atoms does not change the picture: There is still an apparent
optimal Zn/(Zn+Cu) fraction. The promotional role of Zn is yet to be determined for this
CuZn/C model. But with an established procedure for testing the activity of Zn­promoted
Cu/C catalysts, and the knowledge gained from determining particle sizes using methods
that probe either local or bulk environments, combined with the preliminary studies of a
different synthesis route to Zn­promoted Cu/C particles, the foundation has been laid, and
the state of the relevant fraction of Zn promoter seems just around the corner.

7.2 New Methanol Catalysts

Working towards methanol synthesis from CO2 ­rich gas feeds, new Cu­based catalysts
promoted by an atom X were studied theoretically. The theoretical studies indicate that
we obtain a catalyst highly active for CO2 hydrogenation by promoting a Cu(211) facet
with a single atom of the promoter X. In this part of the study, the most promising alterna­
tive promoter was studied.
When using an organic precursor for Cu and a chloride­containing precursor for Ge, a
sequential incipient wetness impregnation was needed, with the order of which the pre­
cursors were impregnated onto the support proving especially important in order to form a
stable mized oxide phase: The successful route to a CuO/GeO2 /SiO2 meant firstly form­
ing a stable GeO2 /SiO2 phase, onto which the Cu precursor could be impregnated to form
the resulting CuO/GeO2 /SiO2 phase.
Cu/SiO2 catalysts generally show very little activity for CO2 hydrogenation, but this is
considered a good starting point to observe a promotional effect of Ge. However, the pro­
motional effect was never observed for a 17 wt% Cu / 11 wt% Ge catalyst. Surprisingly,
the pure Cu/SiO2 proved a better catalyst. The lack of activity for the CuO/GeO2 /SiO2 was
explained by an in­situ XRD study showing the sequential reduction of firstly CuO, and
then GeO2 . Only after reaching a temperature of 400 ◦ C, a binary Cu3 Ge phase starts
to form. The catalyst was never tested for CO2 hydrogenation at temperatures this high,
and a reduction at 350 ◦ C would, according to the in­situ XRD studies, only help to form
a Cu(GeO3 ) phase, which is not thought to be active for CO2 hydrogenation.
When a CuO/GeO2 /SiO2 catalyst did not prove useful for studying the promotional ef­
fect of Ge, SiO2 ­supported Cu(1­x) Gex catalysts with varying molar fractions of Ge were
successfully prepared following a hot­injection wet­synthesis approach. Additionally, for
each Ge molar fraction, different sizes of Cu particles were obtained. This part of the
study attempted at separating the effect of active surface area from that of Ge promo­
tion. While the activity of two Cu­based catalysts could be separated based on an es­
timate of the surface Cu atoms, this approach did not yield consistent results for the
Ge­promoted catalysts. With the large variance in Cu particle sizes across the range
of tested Cu(1­x) Gex /SiO2 catalysts, the theoretical amount of Ge on the surface (assum­
ing Ge segregates to the surface of Cu) stays the same across the range ­ even with the

111
Chapter 7. Conclusion

varying molar fractions of Ge. In fact, it is far more Ge at the surface than we need to
reach the theoretically active Ge promoted Cu(211) surface.
For large molar fractions of Ge, a Cu3 Ge phase is clearly visible with conventional PXRD.
A 2­phase PDF refinement of X­ray total scattering data, including a pure Cu and a Cu3 Ge
phase, complements the findings from PXRD. However, for the theoretically more inter­
esting catalysts of very low Ge molar fractions, it proved difficult to determine the state of
Ge with both PXRD, from PDF analysis of total scattering data, and lastly with HR­TEM.
There is still much to be optimized for the Cu­Ge system. Most important is the control of
the particle size distributions by tuning the parameters of the hot­injection wet­chemical
synthesis. Just as important is it to establish a procedure to characterize these kind of
systems with very low loading of promoters.

112
Appendix A
Supporting data for X­ray methods
A.1 PXRD

Table A.1: In­situ XRD instrument broadening of the XRD peaks measured on a Si stan­
dard sample.
Position (◦ 2θ) FWHM (◦ 2θ)
28.4388 0.0754
47.2965 0.0834
56.1171 0.0819
69.1235 0.0938
76.3707 0.0914
88.0254 0.1010
94.9480 0.117
106.7050 0.126
114.0909 0.144

Table A.2: Ex­situ XRD instrument broadening of the XRD peaks measured on a Si stan­
dard sample.
Position (◦ 2θ) FWHM (◦ 2θ) Integral breadth (◦ 2θ)
28.4041 0.0881 0.12203
47.2587 0.1043 0.14482
56.0767 0.1124 0.15824
69.0797 0.1265 0.18195
76.3252 0.1242 0.18254
87.9774 0.1332 0.20001
94.8992 0.1442 0.21209

113
Appendix B. Supporting data for Cu and CuGe catalysts

Appendix B
Supporting data for Cu and CuGe
catalysts
B.1 Supporting data for in­situ XRD study of CuO/GeO2 /SiO2

Figure B.1: In­situ XRD of the first reduction of CuO/GeO2 /SiO2 in a flow of 5% H2 /He at 1
bar. Diffractograms are shown for each temperature step, and they are the last recorded at
each temperature. Temperature range for this reduction is 160 −550 ◦ C. Lower and upper
diffractograms are long end­scans taken at 25 ◦ C. The diffractograms are corrected for
the sample displacement observed at room temperature in relation to the GeO2 reference.

114
 B.1. Supporting data for in­situ XRD study of CuO/GeO2 /SiO2

Figure B.2: In­situ XRD of the second reduction of CuO/GeO2 /SiO2 in a flow of 5% H2 /He
at 1 bar. Diffractograms are shown for each temperature step, and they are the last
recorded at each temperature. Temperature range for this reduction is 300 −700 ◦ C.
Lower and upper diffractograms are long end­scans taken at 25 ◦ C. The diffractograms
are corrected for the sample displacement observed at room temperature in relation to
the GeO2 reference.

115
Appendix B. Supporting data for Cu and CuGe catalysts

Figure B.3: In­situ X­ray patterns of CuO/GeO2 /SiO2 after reduction at 550 ◦ C and 700 ◦ C.
Diffractograms are taken at 25 ◦ C and are not shifted for easy comparison of intensities.
The diffractograms are corrected for the sample displacement observed at room tempera­
ture in relation to the GeO2 reference. Reference patterns: Cu (ICSD 235809), Ge (ICSD
121532), and Cu3 Ge (ICSD 53266).

116
 B.1. Supporting data for in­situ XRD study of CuO/GeO2 /SiO2

Figure B.4: In­situ XRD of the calcination of Cu/Ge/SiO2 in a flow of 20% O2 /N2 at 1 bar.
Diffractograms are shown for each temperature step, and they are the last recorded at
each temperature. Temperature range for the calcination is 200 −550 ◦ C. Lower and up­
per diffractograms are long end­scans taken at 25 ◦ C. The diffractograms are corrected
for the sample displacement observed at room temperature in relation to the GeO2 refer­
ence.

117
Appendix B. Supporting data for Cu and CuGe catalysts

B.2 Early synthesis studies

For the synthesis of Cu and CuGe nanocrystals, two approaches were investigated, both
being hot­injection wet­synthesis approaches as generally described in Section 3.1.2. The
early studies of the synthesis route are described in the following: A mixture of oleylamine
(OLA abbrev.), Ge(II)Br2 precursor and Cu(II) acetate (Cu(ac)2 abbrev.) and surfactants
if needed, is heated to Temp1 while stirring for 30 minutes in a RBF, either in ambient air
or under a N2 flow. The solution is heated to Temp2, and alternatively heated to Temp3,
before being cooled to room temperature, and the mixture is treated as described in sec­
tion 3.1.2 to precipitate the nanoparticles. The synthesis routes as presented in Table B.1
and Figure B.5 were always prepared with a 1:1 concentration of metal ion precursors.

Table B.1: Synthesis parameters used in oleylamine­based synthesis of CuGe nanopar­

ticles.
Dataset Surfactants N2 (Y/N) Temperature route
V.50 Tri­n­octylphospine + oleic acid N [110 → 235 → 260] ◦ C
V.51 None Y [110 → 235 → 260] ◦ C
V.52 None N [110 → 235 → 260] ◦ C
V.54 Tri­n­octylphospine + olecic acid Y [170 → 220 → RT] ◦ C

Figure B.5: XRD pattern of four individual oleylamine­based synthesis products. Refer­
ence patterns are: Cu (ICSD 235809), Cu3 Ge (ICSD 53266), (Cu5 Ge)0.33 (ICSD 627677),
and Cu1.7 Ge0.3 (ICSD 627675).

118
 B.3. XRD spectra of unsupported Cu and CuGe nanoparticles

B.3 XRD spectra of unsupported Cu and CuGe nanoparticles

Figure B.6: XRD pattern of as­synthesized Cu0.54 Ge0.46 nanoparticles and reference pat­
terns of Cu (ICSD 235809), Cu3 Ge (ICSD 53266), (Cu5 Ge)0.33 (ICSD 627677), Cu1.7 Ge0.3
(ICSD 627675), Cu5 Ge2 (ICSD 87340), and Ge (ICSD 121532).

Figure B.7: Zoom­in of diffraction pattern in Figure B.6.

119
Appendix B. Supporting data for Cu and CuGe catalysts

Figure B.8: XRD pattern of as­synthesized (1)­Cu1 and (2)­Cu1 .

120
 B.3. XRD spectra of unsupported Cu and CuGe nanoparticles

The diffractograms like those shown in Figure 4.9 are subjected to further analysis. Fitting
the data to a Lorentzian, as illustrated in Figure B.9, the Cu crystallite sizes of the inves­
tigated nanoparticles are estimated and given in Table B.2. The analysis is based on the
Cu(111) peak, while using the Scherrer equation with a K­value of 0.94 and a diffractome­
ter broadening of 0.06925 ◦ . The latter is determined at UCPH using a LaB6 standard and
a K­value of 0.94.

Figure B.9: XRD pattern of as­synthesized (1)­Cu1 nanoparticles and best fits.

Table B.2: Size estimates of Cu crystallites of the investigated nanoparticles. Measure­

ments done at UCPH.
Sample name dCu (nm)
(1)­Cu1 71
(2)­Cu1 20
Cu0.67 Ge0.33 44
Cu0.82 Ge0.18 19
(1)­Cu0.85 Ge0.15 21
(2)­Cu0.85 Ge0.15 18
(1)­Cu0.94 Ge0.06 37
(2)­Cu0.94 Ge0.06 22
Cu0.96 Ge0.04 44

121
Appendix B. Supporting data for Cu and CuGe catalysts

B.4 PXRD of Cu1−x Gex /SiO2 ­ before and after activity testing

Figure B.10: PXRD patterns of as­synthesized and post­mortem Cu0.67 Ge0.33 /SiO2 , each
spectrum shifted on the y­axis. Reference patterns: Cu (ICSD 235809), Cu3 Ge (ICSD
53266), Cu1.7 Ge0.3 (ICSD 627675).

Figure B.11: PXRD patterns of as­synthesized and post­mortem Cu0.94 Ge0.06 /SiO2 , each
spectrum shifted on the y­axis. Reference pattern: Cu (ICSD 235809).

122
 B.4. PXRD of Cu1−x Gex /SiO2 ­ before and after activity testing

Figure B.12: PXRD patterns of as­synthesized and post­mortem Cu0.96 Ge0.04 /SiO2 , each
spectrum shifted on the y­axis. Reference pattern: Cu (ICSD 235809).

123
Appendix B. Supporting data for Cu and CuGe catalysts

B.5 XPS of selected Cu1−x Gex /SiO2

Figure B.13: XPS of adhesive carbon tape.

Figure B.14: XPS of as­synthesized Cu0.82 Ge0.18 /SiO2 powder sample pressed on adhe­
sive carbon tape.

124
 B.5. XPS of selected Cu1−x Gex /SiO2

Figure B.15: XPS of as­synthesized Cu0.96 Ge0.04 /SiO2 powder sample pressed on adhe­
sive carbon tape.

125
Appendix B. Supporting data for Cu and CuGe catalysts

B.6 Supporting data for activity of Cu and CuGe­based

catalysts

Figure B.16: Catalytic mass activity of 17 wt% Cu/γ­Al2 O3 prepared by IWI. Reaction
conditions are 150 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 1.4 × 104 h−1 and a WHSV
of 7.4 × 105 Ncm3 · h−1 · g−1
Cu . Uncertainties are estimated by propagation of error (121).

Figure B.16 show the methanol activity of Cu/γ­Al2 O3 as a function of temperature over
multiple temperature cycles. Prior to the first temperature cycle, the catalyst was reduced
in 5 % H2 /Ar at 170 ◦ C for 3 hours. In between the first and second cycles, the catalyst
was regenerated in 100 % H2 at 170 ◦ C for 3 hours. After a second cycle of CO2 hydro­
genation, the sample was subjected to another reduction, although this time in 5 % H2 /Ar
at 350 ◦ C for 3 hours. If the production of DME from methanol dehydration is accounted
for, the mass activity of methanol seems unchanged throughout the different cycles of
reductions and reaction conditions. If only methanol is included in the mass activity, there
is a slight decrease in activity over the multiple cycles, as is expected for Cu at such high
temperatures and multiple cycles.

126
 B.6. Supporting data for activity of Cu and CuGe­based catalysts

Figure B.17: Catalytic mass activity of 17 wt% Cu/SiO2 prepared by IWI. Reac­
tion conditions are CO2 /H2 in 1:4 at 10 bar. Open squares have a WHSV of
7.4 × 105 Ncm3 · h−1 · g−1
Cu ; Closed squares have a WHSV of 2.4 × 10 Ncm · h
5 3 −1 · g−1 .
Cu
Post­mortem Cu crystallite sizes from PXRD are 91 nm and 101 nm for the open­ and
closed­squares, respectively. Uncertainties are estimated by propagation of error (121).

Figure B.18: Catalytic mass activity and methanol selectivity of 4.3 wt% Cu/SiO2 syn­
thesized via the method described in Section 3.1.2 and 17 wt% Cu/SiO2 prepared by
IWI. Reaction conditions are CO2 /H2 in 1:4 at 10 bar, and a WHSV of approximately
8 × 105 Ncm3 · h−1 · g−1
Cu .

127
Appendix B. Supporting data for Cu and CuGe catalysts

Figure B.19: Volumetric concentration of methanol during repeated temperature cycles

of (2)­Cu0.85 Ge0.15 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar,
GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . Black squares represent
individual GC injections.

Figure B.20: Volumetric concentration of products (methanol, methane, CO) during re­
peated temperature cycles of (2)­Cu1 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2
in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . Dots
represent individual GC injections.

128
 B.6. Supporting data for activity of Cu and CuGe­based catalysts

Figure B.21: Volumetric concentration of products during repeated temperature cycles of

Cu0.67 Ge0.33 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV
of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . Dots represent individual GC
injections.

Figure B.22: Volumetric concentration of products during repeated temperature cycles of

Cu0.96 Ge0.04 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV
of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . Dots represent individual GC
injections.

129
Appendix B. Supporting data for Cu and CuGe catalysts

Figure B.23: Volumetric concentration of products during repeated temperature cycles of

Cu/ZnO/Al2 O3 /MgO. Reaction conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV
of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1Cu . Dots represent individual GC
injections.

(a) Methanol mass activity. (b) Effect of reduction on mass activity.

Figure B.24: (a) Catalytic mass activity of methanol during multiple temperature cycles of
(1)­Cu0.94 Ge0.06 /SiO2 (black) and (2)­Cu0.94 Ge0.06 /SiO2 (blue) and (b) the mass activity
of (2)­Cu0.94 Ge0.06 /SiO2 before and after reduction in 5 % H2 /Ar at 220 ◦ C for 3 hours.
Reaction conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and
a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .

130
 B.7. PXRD of Cu­based catalysts on different supports

B.7 PXRD of Cu­based catalysts on different supports

Figure B.25: XRD pattern of used 17 wt% Cu/SiO2 and Cu/γ­Al2 O3 . Reference pattern of
Cu (ICSD 235809). Cu crystallite sizes as determined with the Scherrer equation: 91 nm
for Cu/SiO2 , and 123 nm for Cu/γ­Al2 O3 .

131
Appendix B. Supporting data for Cu and CuGe catalysts

B.8 PDF refinement of selected Cu(1­x) Gex /SiO2 catalysts

The altered background subtraction and PDF refinements of the total scattering data pre­
sented in the following are carried out by Jette Mathiesen. For all refinements, the follow­
ing parameters were used: Qinstr. = 22; Qmax = 17; Qmin = 0.4.

Table B.3: Refined parameters for the as­synthesized Cu0.67 Ge0.33 /SiO2 using 1 phase
(FCC Cu) or 2 phases (FCC Cu and Cu3 Ge).
Refined parameter 1­phase 2­phase
Rw / % 59.06 32.19
ScaleCu 0.1 0.06
δ2 6.17 3.75
aCu / Å 3.63 3.62
UCu / Å2 1.26 × 10−2 6.97 × 10−3
Crystallite size, Cu / nm 63.82 75.05
ScaleCu3 Ge ­ 0.09
aCu3Ge / Å ­ 5.26
bCu3Ge / Å ­ 4.22
cCu3Ge / Å ­ 4.50
UCu / Å2 ­ 9.06 × 10−3
UGe / Å2 ­ 8.94 × 10−3
Crystallite size, Cu3 Ge / nm ­ 121.31

The final phase composition of the 2­phase refinement of as­synthesized Cu0.67 Ge0.33 /SiO2
is estimated as Cu3 Ge : Cu = 62.15 % : 37.85 %.

Specifically for (1)­Cu0.85 Ge0.15 /SiO2 , the background subtraction is not straightforward
due to the presence of both SiC, SiO2 and kapton in ratios that are different from the pure
backgrounds. This is observed in Figure B.26a. The background subtraction was per­
formed by separating the kapton contribution from the measured background containing
SiC and SiO2 (Figure B.26b). Lastly, a complete background is created by individually
scaling each component to account for variations in concentrations of SiC and SiO2 and
variations in packing densities (Figure B.26c). Figure B.26d illustrates the importance of
correctly defining the background and the contribution the background has to the noise of
the experimental PDF. When comparing the black curve containing the measured [SiO2 +
kapton] background with the yellow one, which uses the individually scaled background,
it is seen that noise dominates the local structure of the PDF without scaling. Better local
analysis in the low r­range (< 5 Å) can be performed if a scaled background is applied.

Table B.4: Refined parameters for the as­synthesized and post­mortem samples of (1)­
Cu0.85 Ge0.15 /SiO2 .
Refined parameter As­synthesized Post­mortem
Rw / % 19.97 14.64
ScaleCu 0.09 0.18
δ2 3.23 4.04
aCu / Å 3.62 3.62
UCu / Å2 8.20 × 10−3 7.91 × 10−3
Crystallite size, Cu / nm 59.01 90.59

132
 B.8. PDF refinement of selected Cu(1­x) Gex /SiO2 catalysts

(a) (b)

(c) (d)

Figure B.26: Experimental PDF of post­mortem (1)­Cu0.85 Ge0.15 /SiO2 with (a) no back­
ground subtraction and with the individual backgrounds, and (c) the background from
individually scaling each of the background contributions. Sub­figure (b) shows the sub­
traction of kapton and (d) illustrates how the background contributes to the noise in the
experimental PDF.

133
Appendix B. Supporting data for Cu and CuGe catalysts

B.9 SEM­EDX and (HR­)TEM of (1)­Cu0.85 Ge0.15 /SiO2

Figure B.27: Ex­situ SEM­EDX of large, agglomerated particle in the post­mortem sample
of the (1)­Cu0.85 Ge0.15 /SiO2 catalyst. Image was acquired and analyzed by PhD Filippo
Romeggio.

The following TEM and HR­TEM images were acquired and analyzed by PhD Stefan
Akazawa. Lattice fringes were observed at the edge of thick, agglomerated particles.
Such an area is indicated by a red box in Figure B.28. A HR­TEM image of the area is
shown in Figure B.29. In the HR­TEM, lattice fringes are vaguely visible. The FFT of the
area show high­intensity spots in k­space with lengths between 4 1/nm and 5 1/nm, which
corresponds to lattice plane distances between 2.5 −2 Å. Circles indicating k­values for
a series of materials expected to be in the sample are imposed on the FFT image. The
crystal information for Cu­containing materials are obtained from ICSD (88), while that of
SiC is obtained from in­house XRD of the pure SiC dilutant. The spots in the FFT cannot
be clearly identified as any one of the suggested materials, consequently complicating
the crystal phase identification.

134
 B.9. SEM­EDX and (HR­)TEM of (1)­Cu0.85 Ge0.15 /SiO2

Figure B.28: TEM image of post­mortem sample of (1)­Cu0.85 Ge0.15 /SiO2 catalyst showing
thick, largely crystalline, particle. TEM image was acquired by PhD Stefan Akazawa.

Figure B.29: HR­TEM image and corresponding FFT of the region displayed by a red
box in Figure B.28. Middle: FFT of HR­TEM image with rings indicating k­space values
for each of the labelled materials (SiC, Cux Oy , and Cux Gey ). Right: Close up of up a
single peak in the FFT image. HR­TEM image was acquired and analyzed by PhD Stefan
Akazawa.

135
Appendix B. Supporting data for Cu and CuGe catalysts

B.10 SEM images for early synthesis studies

Figure B.30: SEM(­EDX) of a Cu­based catalysts containing nominal Ge loading of 20

wt%. EDX spectrum shows the presence of Si (from support material), Cu and Cl, but no
Ge. Image was acquired and analyzed by Daniel Graungaard Bruun.

Figure B.31: SEM(­EDX) of a Cu­based catalysts containing nominal Ge loading of 30

wt%. EDX spectrum shows the presence of Si (from support material), Cu and Cl, but no
Ge. Image was acquired and analyzed by Daniel Graungaard Bruun.

136
Appendix C
Supporting data for the
Cu/ZnO/Al2O3/MgO catalyst
The following experiments of the Alfa Aesar Cu/ZnO/Al2 O3 /MgO catalyst were done as
a part of the SRP project carried out by Olivia H. Nikolajsen from Gefion gymnasium.
The experiments and subsequent data treatment were carried out by the author of this
thesis as they were also beneficial for ongoing projects. The experiments were done in
connection with optimizing the experimental setup discussed in Section 3.2.1 to run at
pressures higher than 1 bar. These experiments were carried out with a high gas velocity
to elucidate the intrinsic activity of the catalyst. The mass activity is higher when using
CO2 instead of the industrially used mixture of CO2 and CO. Since CO2 has a 20 times
higher intrinsic rate for the formation of methanol compared to CO, replacing some of the
CO2 with CO will result in a lesser rate for the formation of methanol (178).

(a) Dependence of mass activity on CO2 /CO/H2 (b) Dependence of pressure in CO2 /CO/H2 gas
gas mixture gas mixture. 10 bar. mixture. 2.5 , 5 and 10 bar.

Figure C.1: Mass activity of methanol of an industrial­grade Cu/ZnO/Al2 O3 /MgO catalyst

during (a) consecutive activity tests in CO2 /CO/H2 of varying ratios and (b) in CO2 /CO/H2
at varying pressures. Reaction conditions are a total gas flow of 150 ml/min, 10 bar, a
GHSV of 2.3 × 104 h−1 and a WHSV of 2.7 × 105 Ncm3 · h−1 · g−1 Cu . Pre­treated via re­
◦
duction in 5 % H2 /Ar at 220 C for 3 hours. Experimental history: Reduction, CO2 /CO/H2 ,
CO2 /H2 , and lastly CO/H2 .

137
Appendix C. Supporting data for the Cu/ZnO/Al2 O3 /MgO catalyst

(a) Reduction in 5 % H2 /Ar. (b) Gas feed: CO2 /CO/H2 .

(c) Gas feed: CO2 /H2 . (d) Gas feed: CO/H2 .

Figure C.2: QMS signal and temperature during (a) reduction and consecutive activity
tests of the industrial­grade Cu/ZnO/Al2 O3 /MgO catalyst. Sub­figures (b) through (d) are
in different gas feeds. Reaction conditions are those described in the figure text for Fig­
ure C.1.

Personal side note: Olivia reached out to me in the winter of 2022 after having attended a
lecture I held at her school a year before which, in her own words, made a great impression
on her. She wanted to write her main project in the subjects of Power­to­X and catalysis,
and she wanted to carry out experiments at a research institution. There was no doubt in
my mind when I invited her to visit DTU. I will always strive to inspire and empower young
women and individuals from minorities to follow their passion and to pursue a career in
science. I am very proud to have been a part of Olivia’s project that she (unsurprisingly)
got the highest grade for.

138
Appendix D
Supporting data for CuX catalysts

Figure D.1: Methanol mass activity of a series of Cu/X/C catalysts. Normalized with
the amount of Cu. Reaction conditions are CO2 /H2 in 1:4 at 10 bar, and a WHSV of
1.3 × 105 Ncm3 · h−1 · g−1
Cu .

Figure D.2: Methanol mass activity of a series of Cu/X/C catalysts. Normalized with the
amount of Cu and promoter. Reaction conditions are CO2 /H2 in 1:4 at 10 bar, and a WHSV
of 1.3 × 105 Ncm3 · h−1 · g−1
Cu .

139
Appendix D. Supporting data for CuX catalysts

Figure D.3: PXRD pattern of post­mortem Cu/X/C catalysts (X=Sn, Bi, Pb, Sb) and ref­
erence patterns of Cu (ICSD 235809), CuO (ICSD 16025), Cu2 O (ICSD 52043), and SiC
dilutant.

Cu crystallite size using the Scherrer equation and the Cu(111) peak: Cu/Sn/C (76 nm);
Cu/Sb/C (60 nm); Cu/Pb/C (60 nm); Cu/Bi/C (34 nm); Cu/C (85 nm).

140
Appendix E
Supporting information for
experimental setup
E.1 Data treatment
The python­based code for treating data from the plug flow reactor setup has many au­
thors and is lastly updated by PhD Thomas Smitshuysen. It is written specifically to be
used with the setup. The flow of the ’GC PARSER’ data treating algorithm is in short:
1. The algorithm loads the necessary data sets containing temperature, pressure, mass
flows and GC spectra.
2. Functions are loaded from dedicated python scripts. One contains the functions
needed for integration, plotting functions, etc., and another contains the manual
settings that can be subject to change. These include the conversion factors for
each of the gases linked to the setup, type of background, and the fitting parameters,
the latter including where to start and end the integration, as well as a setting for
subtracting another peak.
3. GC spectra are numerically integrated, and with the use of conversion factors are
converted to volumetric concentrations.
4. Along the way, the GC injections are synchronised with temperature, reactor pres­
sure and gas flow, to be used for the conversions from one unit to another, to be
used as requirements for later ordering into time­ and temperature­resolved data
files, or to simply be loaded into data files. These data files makes it easy for the
user to spot deviations as the experiment progresses as practically everything is
saved.
The next three figures represent the process described above. The reaction conditions
for the experiment are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and
a WHSV of 8 × 105 Ncm3 · h−1 · g−1
Cu .

141
Appendix E. Supporting information for experimental setup

Figure E.1: Examples of the numerical integration of GC peaks. The peaks are pictured
during the same time of a 80­hour long experiment while testing the (1)­Cu0.94 Ge0.06 /SiO2
catalyst.

Figure E.2: Integrated FID and TCD signals. The data is from an a 80­hour long exper­
iment using the (1)­Cu0.94 Ge0.06 /SiO2 catalyst. Points are individual GC injections, each
of them integrated as illustrated in Figure E.1.

142
 E.1. Data treatment

Figure E.3: Volumetric concentration of methanol during repeated temperature cycles

while testing the (1)­Cu0.94 Ge0.06 /SiO2 catalyst. Points are the individual GC injections
from Figure E.2, converted into volumetric concentrations using the calibrations factors
specific for each gas.

Table E.1: Table of retention times and calibration factors for conversion of integrated data
to concentrations.
Species Detector Retention time Conversion factor [mol­% · min · V−1 · bar−1 ]
H2 TCD 7.4 min 314.0 ± 0.1
Ar TCD 7.8 min 4.555 ± 0.001
CO2 TCD 4.4 min 3.701 ± 0.001
CO TCD 9.2 min 4.373 ± 0.001
CH4 FID 0.9 min 0.2959
C2H4 FID 1.7 min 0.2959/1.97
C2H6 FID 2.2 min 0.2959/1.99
C3H6 FID 8.0 min 0.2959/3.00
C3H8 FID 8.2 min 0.2959/3.00
C4Hx FID 9.7 −10.0 min 0.2959/4.00
DME FID 8.8 min 0.2959/1.10938
MeOH FID 9.20 min 0.44217
EtOH FID 10.6 min 0.2959/1.59

All of the TCD species, along with methanol and methane, have been calibrated by PhD
Thomas Smitshuysen, the latest being methane and methanol (as of 2022). DME was
calibrated by the author of reference (85). The rest is extrapolated using CH­group quan­
tity with data from reference (213).
The conversion factors are determined by measuring the response (i.e. the raw signal)
of the TCD and FID detectors at a certain volume percentage of a given species. The
volumetric concentration of a species can easily be controlled as the setup is equipped

143
Appendix E. Supporting information for experimental setup

with mass flow controllers to control the flow of a gas, as well as Ar to dilute a pure gas, for
example CH4 . Dilution is used whenever a calibration gas mixture is not available. These
calibration factors are therefore a direct and setup­specific measure of the sensitivity of
the GC towards a certain species. It has been shown that by dividing with the pressure
on the GC injection line, a better linear fit between the detector signal and the volume
percentage is obtained (37), why the conversion factor is in units of concentration (%) per
detector intensity (min · V−1 ) per pressure (bar−1 ).
Using the factors in Table E.1, the concentration of a species in a flow of gas is easily
calculated. As the GC removes a sample of gas from the outlet of the reactor, and the
GC sample valve and the tube always contains the same volume of gas, the magnitude
of that gas will be proportional to the number of moles with a factor. Assuming all gases
to be ideal, we can write:

Fgas · psample · Vsample

Idetector = Fgas · ngas = Fgas · [gas] · nsample = · [gas] (E.1)
R · Tsample

Here, the unknown is [gas], the molar concentration of the given gas. Rearranging the
above equation to isolate [gas], and with all the constant values contained in the conver­
′
sion factor, Fgas , we get:

R · Tsample ′ Idetector
[gas] = Idetector = Fgas (E.2)
Fgas · psample · Vsample psample

Here, psample is not the pressure in the reactor, but the pressure on the GC line going
from the outlet of the reactor. Recent findings show that there is quite a pressure drop
over the GC. After pressure measurement was added in front of both of the detectors to
monitor this pressure drop, the most important species measured with the FID detector
was re­calibrated (methane and methanol). The signal from the FID and TCD detectors
are therefore divided by the corresponding pressures on the gas line going to the GC. If a
line pressure for a given GC injection is not available, the pressure is set to the pressure
from the last injection or to 1 bar, ensuring that no GC injections are lost just because a
pressure gauge was faulty.

144
 E.2. Setup optimizations

E.2 Setup optimizations

Slight variations in the packing of the catalyst sample bed, coupled with the tendency of
carbon support to pack tightly within the reactor, would tend to cause trouble in keeping the
reaction conditions the exact same across experiments involving Cu(Zn)/C. The pressure
in the reactor, calculated as the mean value between the pressures measured on the
back­side and the front­side of the reactor, would tend to fluctuate between 1.2 bar and
1.5 bar, as well as increase during an experiment, when combining glass reactors and
200 mg of Cu(Zn)/C (undiluted). Investigations of the effect of the reactor pressure on the
methanol mass activity showed minor increases in mass activity compared to the large
increases observed at low Zn loadings in Figure 5.3. The slight increase of approximately
17 % in mass activity is seen in Figure E.4.

Figure E.4: Effect of reactor pressure during methanol synthesis in a CO2 /CO/H2 gas
feed. Pressure on the back­side of the GC set to 850 mbar.

While attempting to mimic the conditions found when using glass reactors, but using an
industrial­type CZA catalyst operating at equilibrium conditions instead (in what in reality
is a methanol calibration experiment), it was found that the pressure­normalized FID de­
tector signals did not make sense. While operating at equilibrium conditions, the methanol
signal should be constant with a changing flow. A large amount of the catalyst (5 g) was
used to effectively ’drive’ the reaction into equilibrium at 1 bar, and while testing the activity
of methanol in a CO2 /H2 flow at different total flows and temperatures (conditions shown
in Figure E.5a), the following was found: The integrated TCD signals (Figure E.5b) show
CO2 and H2 going up and down with flow, as determined by the experimental procedure,
and CO following due to the rWGS. The normalized signal of Figure E.5d shows CO in­
creasing with temperature until equilibrium is reached at 220 ◦ C, where the volumetric
concentration stabilizes. The integrated FID signals of Figure E.5c show methanol vary­
ing with temperature and pressure. However, when normalising with the pressure at the
GC, the volumetric concentration of methanol decrease as a function of temperature (Fig­
ure E.5d). While the normalized TCD signals are normal for a reaction in equilibrium, the
FID signals does not indicate equilibrium at any point ­ for the CZA catalyst, especially
using 5 g of it, the CO2 hydrogenation should be in equilibrium. Adding a second gauge at

145
Appendix E. Supporting information for experimental setup

the outlet of the GC revealed a large pressure drop over the GC, most likely caused by the
reduction in tube size ­ it goes down to 1/16’ tubing inside the GC. Calibrating products
measured at the FID detector with the pressure measured on the back side of the GC
(resembling the pressure at the FID more so than the pressure measured at the front of
the GC) gave a better result for the methanol signal.

(a) Experimental course. (b) Integrated TCD signals.

(c) Integrated FID signals. (d) Volumetric concentration.

Figure E.5: CO2 hydrogenation using 5 g CZA catalyst with different flow rates and tem­
peratures (a). Pressure is set to 1 bar and GC back­pressure set to 200 mbar to simulate
conditions from experiments using glass reactors. Subfigures (b), (c) are integrated TCD
signals (CO, CO2 and H2 ) and FID signals (methanol), and subfigure (d) is the volumetric
concentrations. Experiment performed by PhD Thomas Smitshuysen.

146
 E.2. Setup optimizations

Figure E.6: On­line GC­sampler, Agilent 7890A. The gas is lead in by the TCD sampling
(’Sample # 1 in’) and exits again by the FID sampling (’Sample # 2 out’). The schematic
is annotated and reprinted from (85).

147
Appendix F. Supporting data for Cu(Zn)/C catalysts

Appendix F
Supporting data for Cu(Zn)/C
catalysts
F.1 Activity data

(a) CuZn­0.05/C: Volumetric concentration. (b) CuZn­0.05/C: Mass activity.

Figure F.1: (a) Volumetric concentration and (b) mass activity of methanol during a tem­
perature cycle for CuZn­0.05/C. Reaction conditions are 100 ml/min of CO2 /CO/H2 in
1:1:5 at 1 bar, with a WHSV of 6 × 105 Ncm3 · h−1 · g−1
Cu .

148
 F.1. Activity data

Figure F.2: Mass activity of methanol over a CuZn­0.62/C catalyst. Activity is tested in
glass reactors (black, 200 mg sample and 100 ml/min total gas flow) and glass­lined steel
reactors (blue, 100 mg sample, SiC dilution, and 50 ml/min total gas flow). Uncertainties
are estimated by propagation of error (121).

Uncertainties are quite large when estimated by propagation of error. To determine the
error of the measurements, repeated experiments have been performed using the glass­
lined steel reactors and a selected set of conditions. For this purpose, two batches of
CuZn/C catalyst were synthesized via conventional incipient wetness impregnation to give
a nominal Zn/Cu ratio of 0.62.

Figure F.3: Error estimation using two batches of CuZn/C (Zn/Cu=0.62). Reac­
tion conditions are 100 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of
6 × 105 Ncm3 · h−1 · g−1
Cu .

149
Appendix F. Supporting data for Cu(Zn)/C catalysts

F.2 Representative microscopy

(a) CuZn­0.05/C. (b) CuZn­0.31/C.

(c) CuZn­0.62/C. (d) CuZn­1.0/C.

Figure F.4: HAADF­STEM size distributions of CuZn/C catalysts post activity testing. Note
the much larger x­axis used for CuZn­0.05/C.

150
 F.2. Representative microscopy

(a) Cu nanoparticle. (b) Zn nanoparticle.

Figure F.5: STEM­EDX images of separate (a) Cu and (b) Zn particles in the post­mortem
sample of CuZn­1.0/C post activity testing in syngas at 1 bar, both on a 10 nm scale.

Figure F.6: STEM­EDX image of CuZn­0.31/C

Figure F.7: STEM­EDX image of CuZn­0.62/C

151
Appendix F. Supporting data for Cu(Zn)/C catalysts

F.3 In­situ XRD study of Cu/C

As­synthesized and calcined samples of a 5 wt% Cu/C was loaded into the in­situ XRD.
A total gas flow of 50 ml/min was used to match the gas velocity used during activity
testing. Diffraction scans were acquired before, during and after each temperature step.
Reduction procedure and synthesis gas mixture was the same as those used during the
catalytic tests.

(a) (b)

Figure F.8: In­situ XRD study of the state of a Cu/C catalyst during (a) reduction. Sub­
figure (b) shows long scans taken of the sample in different states (as­synthesized, re­
duced, post CO2 /CO/H2 , and after being exposed to ambient air). In sub­figure (b), the
diffractograms are shifted on the y­axis.

Figure F.8a shows the effect of the reduction applied for all Cu(Zn)/C catalysts used during
this PhD project. 10­minute scans were obtained during the entire reduction of the 5 wt
% Cu/C catalyst while focusing on the area around the Cu(111) peak. During the first
two short scans, no significant changes to the Cu(111) peak is observed. At the 3rd and
4th scans, which is 30 and 40 minutes into the 3 hour reduction in 2 % H2 at 175 ◦ C,
the Cu(111) peak has increased in intensity. While the reduction continues there is no
significant change in the peak width or intensity. Scans taken during cooling of the sample
in 100 % He are also included in Figure F.8a, but no visible change in peak profile is
apparent. At this point, the mean Cu particle size is determined to be 2.5 nm.
Before and after each step of the in­situ experiment, long scans were obtained. These are
shown in Figure F.8b. Surprisingly, there seems to be no change in the diffraction patterns
after activity testing in synthesis gas. What is more surprising is the high QMS signal
corresponding to mass 15 (CH4 ) observed during activity testing of the reduced Cu/C
sample. As seen from Figure F.9a, m/z 15 increases and decreases step­wise, following
the temperature of the sample. This is not a product of Ni contamination, as was seen
for a previous experiment in Figure F.9b, during which the Ni contamination was clearly
visible in the diffractograms (see Figure F.10). While it was thought that the CH4 formation
was due to contamination of the in­situ sample holder, methane formation was still found
for a chemically/thermally cleaned sample holder (see Figure F.9c). Recent studies of
the Anton Paar furnace, which are not shown here, indicate severe contamination of the
furnace. The effect that the formation of CH4 during the activity test has on the Cu surface
is yet to be established.

152
 F.3. In­situ XRD study of Cu/C

(a) Cu/C sample activity. (b) Cu/C with Ni contamination, activity.

(c) In­situ XRD holder activity.

Figure F.9: QMS signal during activity testing in CO2 /CO/H2 of (a) 5 wt % Cu/C catalyst,
(b) a Cu/C catalyst contaminated with Ni, and (c) the in­situ sample holder.

153
Appendix F. Supporting data for Cu(Zn)/C catalysts

Figure F.10: Ni contamination of a Cu/C catalyst during activity testing in syngas.

154
 F.4. Example of total scattering data: CuZn­0.05/C

F.4 Example of total scattering data: CuZn­0.05/C

(a) (b)

(c)

Figure F.11: (a) Total scattering data, (b) reduced total scattering data structure function
and (c) PDF for a carbon­supported CuZn sample.

155
Appendix F. Supporting data for Cu(Zn)/C catalysts

F.5 PDF refinement of selected Cu(Zn)/C catalysts

Table F.1 lists the refined parameters from the PDF refinements of (a)­Cu/C and CuZn­
0.05/C using a one­phase and two­phase approach, respectively: Unit cell parameters
(a, b, c); the scale factor (controlling the overall intensity of the PDF); a parameter that
accounts for correlated atomic motion (δ 2 ); and a parameter related to the spherical par­
ticle diameter.
A one­phase model describes well the experimental data of (a)Cu/C with only slight de­
viations thought to be due to the presence of CuO that have not been accounted for in
the refined model. While the two­phase model is able to correctly describe most of the
positions of the experimental PDF peaks of CuZn0.05/C, along with the peak widths, the
intensity of the peaks are not correctly described in the low r­range. This is thought to be
due to an incorrect inclusion of the instrumental damping of the PDF. This might be due
to incorrect refinement of sample damping (which is due to the finite size of the nanopar­
ticles), or not including a correct value for the instrumental damping of the PDF.

Figure F.12: Comparison of experimental PDF of (a)­Cu/C with calculated PDFs of Cu,
CuO and Cu2 O. Model of FCC Cu is used in the subsequent refinement.

156
 F.5. PDF refinement of selected Cu(Zn)/C catalysts

Figure F.13: 1­phase PDF refinement of post­mortem (a)­Cu/C. The blue curve is cal­
culated from a structural model (FCC Cu, ICSD coll. code 235809), which is refined to
give the best possible fit to the experimental PDF (black). The green line displays the
difference between the model and experimental data. Refined parameters are listed in
Table F.1.

Table F.1: Refined parameters for the post­mortem sample of (a)­Cu/C using a one­phase
approach.
Refined parameter 1­phase: FCC Cu
Rwp / % 18.82
ScaleCu 0.086
δ2 2.87
aCu / Å 3.59
Crystallite size, Cu / nm 6.26

157
Appendix F. Supporting data for Cu(Zn)/C catalysts

Figure F.14: 2­phase PDF refinement of post­mortem CuZn­0.05/C. The blue curve is re­
fined to give the best possible fit to the experimental PDF (black). The green line displays
the difference between the model and experimental data. Refined parameters are listed
in Table F.2.

Table F.2: Refined parameters for the post­mortem sample of CuZn­0.05/C using a two­
phase approach.
Refined parameter 2­phase: Cu + CuO
Rwp / % 40.58
ScaleCu 0.042
δ2 3.79
aCu / Å 3.59
ScaleCuO 0.025
aCuO / Å 4.70
bCuO / Å 3.50
cCuO / Å 5.09
Particle size / nm 6.56

158
 F.6. Rietveld refinement of selected Cu(Zn)/C catalysts

F.6 Rietveld refinement of selected Cu(Zn)/C catalysts

Rietveld refinement was performed by PhD Jonathan Schouenborg using the Topas soft­
ware.

Figure F.15: Rietveld refinement of post­mortem samples of (a) (a)­Cu/C and (b) CuZn­
0.05/C. The dashed lines indicate the position of the fundamental parameters peaks used
to describe the carbon support.

159
Appendix F. Supporting data for Cu(Zn)/C catalysts

F.7 PXRD of post­mortem Cu(Zn)/C catalysts

Figure F.16: XRD patterns of post­mortem Cu(Zn)/C catalysts with increasing amount of
Zn (upper to lower patterns). Samples prepared via sequential IWI. Reference spectra:
PBX51 (in­house measurement), Cu (ICSD 235809, red), Zn (ICSD 25286, blue) and SiC
dilutant (ICSD 19706, dashed black).

Figure F.17: XRD patterns of post­mortem Cu(Zn)/C catalysts with increasing amount
of Zn (upper to lower patterns). From catalytic testing in glass­lined steel reactors.
Reference spectra: PBX51 (in­house measurement), Cu (ICSD 235809, red), Zn (ICSD
25286, blue) and SiC dilutant (ICSD 19706, dashed black).

160
 List of Figures

List of Figures
1.1 Changes in global surface temperature relative to 1850–1900: History and
causes of warming. Reprinted with permission. Copyright IPCC (2). . . . . 1
1.2 Volumetric (in MJ/L) and gravimetric (in MJ/kg) energy density of selected
energy carriers. Data points are taken with permission from (11). Copyright
2014 Elsevier. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Growth in primary chemical production in the ’Net Zero Scenario’ over the
year 2000­2030 as reported by the International Energy Agency (IEA). Coloured
dashed lines past the year 2021 represent the projected growth. Data set
is from (13), https://www.iea.org/reports/chemicals. License: CC BY 4.0. . . 4

2.1 A simple potential energy diagram of a heterogeneous catalytic reaction. . 6

2.2 Equilibrium methanol profiles. (a) Equilibrium concentration with a feed of
25/75 CO/H2 , 10/10/80 CO/CO2 /H2 and 25/75 CO2 /H2 . For the latter, the
rWGS is included or excluded by either considering CO as a product (’incl.
rWGS’) or excluding it (’excl. rWGS’). Sub­figure (b) shows the equilib­
rium concentration in a CO/CO2 /H2 feed as a function of temperature and
pressure. Simulated using HSC 7 (19). . . . . . . . . . . . . . . . . . . . . 9

3.1 Schematic of the experimental setup. Modified with permission from (37). . 17
3.2 Illustration of the Bragg law. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 A scattering event illustrating the relation between incident and reflected
beams and their corresponding wave vectors, ki and kf , and the scattering
vector Q. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4 An X­ray scattering experiment. The principle of powder diffraction is illus­
trated: More than one crystal is at the correct orientation (angled θ to the
incident beam) to satisfy Bragg’s law, resulting in more than one cone of
diffracted rays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5 Events happening as a results of the interaction between a primary electron
beam and a sample in an electron microscope. . . . . . . . . . . . . . . . . 26
3.6 Schematic diagram of image formation in a transmission electron micro­
scope. Modified version, reprinted with permission from (91). Copyright
John Wiley and Sons, 2008. . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.7 The photoemission and Auger processes resulting from absorption of an
X­ray photon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.8 The mean free path (nm) of electrons as a function of their kinetic energy
(eV). Reprinted with permission from (23). Copyright John Wiley and Sons,
2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.1 Theoretical methanol TOF on Cu(211) metal surfaces, promoted with a sin­
gle X atom, as a function of descriptors E(HCOO*) and E(CO*). Unpub­
lished results by Ang Cao. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Catalytic mass activity of 17 wt% Cu/γ­Al2 O3 and 17 wt% Cu/SiO2 . Re­
action conditions are 150 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of
1.4 × 104 h−1 and a WHSV of 7.4 × 105 Ncm3 · h−1 · g−1 Cu . Uncertainties are
estimated by propagation of error (121). . . . . . . . . . . . . . . . . . . . . 37

161
List of Figures

4.3 PXRD pattern of CuO/GeO2 /SiO2 synthesized via the sequential impregna­
tion method. Patterns of the as­synthesized GeO2 /SiO2 intermediate and
a 17 wt% CuO/SiO2 , as well as pure CuO, GeO2 and SiO2 references, are
pictured for easy comparison. Spectra are shifted on the y­axis. Cu and
Ge nominal loadings of 18 wt% and 28 wt%, respectively. . . . . . . . . . . 40
4.4 Catalytic mass activity of a Cu/GeO2 /SiO2 catalyst with nominal loadings
of 15 wt% Cu and 11 wt% Ge synthesized using the sequential IWI. Reac­
tion conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, and a WHSV of
2.4 × 105 Ncm3 · h−1 · g−1
Cu . Uncertainties are estimated by propagation of
error (121). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.5 In­situ XRD of the reduction of CuO/GeO2 /SiO2 in a flow of 5% H2 /He at
1 bar. Diffractograms are shown for each temperature step, and they are
the last recorded at each temperature. The diffractograms are corrected
for the sample displacement observed at room temperature in relation to
the GeO2 reference. Reference patterns: CuO (ICSD 16025), GeO2 (ICSD
59642), Cu(GeO3 ) (ICSD 25754), Cu (ICSD 235809), Ge (ICSD 121532),
Cu3 Ge (ICSD 53266). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.6 PXRD of CuO/GeO2 /SiO2 after reduction. Diffractograms are recorded at
room temperature before reduction (i.e. calcined state), and after reduc­
tion at 550 ◦ C and 700 ◦ C. The diffractograms are corrected for the sample
displacement observed at room temperature in relation to the GeO2 refer­
ence. Reference patterns: CuO (ICSD 16025), GeO2 (ICSD 59642), Cu
(ICSD 235809), Ge (ICSD 121532), and Cu3 Ge (ICSD 53266). . . . . . . . 44
4.7 Ex­situ PXRD of the calcined Cu/Ge/SiO2 sample after removal from in­
situ XRD. The in­situ XRD patterns is the latest scan, taken after a week
of being in the in­situ sample holder, and is the same as that of Figure B.4.
The in­situ diffractogram is corrected for the sample displacement observed
at room temperature in relation to the GeO2 reference. References: Ge
(ICSD 121532), CuO (ICSD 16025), GeO2 (ICSD 59642). . . . . . . . . . . 45
4.8 Phase diagram of the Cu–Ge binary system. Reprinted with permission
from (131). Copyright Elsevier B.V., 2010. . . . . . . . . . . . . . . . . . . . 46
4.9 XRD patterns of as­synthesized Cu and Cu(1­x) Gex nanoparticles with vary­
ing amounts of Ge, each spectrum shifted on the y­axis. Reference spectra:
Cu (ICSD coll. code 235809), Ge (ICSD coll. code 212532), Cu3 Ge (ICSD
coll. code 53266) and (Cu5 Ge)0.33 (ICSD coll. code 627677). . . . . . . . . 48
4.10 Ex­situ SEM of unsupported (1)­Cu0.85 Ge0.15 nanoparticles on a glassy car­
bon stub. Images were acquired by PhD Filippo Romeggio. The micro­
scope was operated at 20 kV and with a working distance of 12.9 mm. . . . 49
4.11 XRD patterns of SiO2 ­supported Cu(1­x) Gex nanoparticles. Reference spec­
tra: Cu (ICSD 235809), Cu3 Ge (ICSD 53266), (Cu5 Ge)0.33 (ICSD 627677),
and Cu1.7 Ge0.3 (ICSD 627675). . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.12 Experimental PDF (black) of as­synthesized Cu0.67 Ge0.33 /SiO2 , and the
calculated PDF’s of FCC Cu (red) and Cu3 Ge (blue) models. . . . . . . . . 51
4.13 (a) 1­phase and (b) 2­phase refinement of as­synthesized Cu0.67 Ge0.33 /SiO2 .
The green line is the difference between the model (blue) and experimental
data (black). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.14 Mass activity of methanol over SiO2 ­supported Cu catalysts. . . . . . . . . 52

162
 List of Figures

4.15 Catalytic mass activity of 4.3 wt% Cu/SiO2 synthesized via the method de­
scribed in Section 3.1.2 and 17 wt% Cu/SiO2 prepared by IWI. Reaction
conditions are CO2 /H2 in 1:4 at 10 bar, and a WHSV of approximately
8 × 105 Ncm3 · h−1 · g−1Cu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.16 (a) The effect of cycling on the mass activity and (b) the product distribution
of last temperature cycle, of the 4.3 wt% (2)­Cu1 /SiO2 catalyst. Reaction
conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1
and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . . . . . . . . . . . . . . . . . . . . 53
4.17 PXRD patterns of two Cu/SiO2 catalysts, pre­ and post­ testing of activity.
Reference spectra: Cu (ICSD 235809), SiC dilutant from test reactor. . . . 55
4.18 (a) Catalytic mass activity of methanol during multiple temperature cycles
and (b) the mass activity (left scale) and surface area normalized activity
(right scale) during the last temperature cycle, of (1)­Cu/SiO2 (black) and
(2)­Cu/SiO2 (blue). Reaction conditions are 50 ml/min of CO2 /H2 in 1:4 at
10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .
The normalized rate in subfigure (b) is based on the crystallite size esti­
mated from PXRD of post­mortem samples (from Figure 4.17). Note the
two different scales in this subfigure. . . . . . . . . . . . . . . . . . . . . . . 56
4.19 Ex­situ SEM of post­mortem (1)­Cu1 /SiO2 on carbon tape. Images were
acquired by PhD Filippo Romeggio. The microscope was operated at 5 kV
and with a working distance of 4.2 mm. . . . . . . . . . . . . . . . . . . . . . 56
4.20 PXRD patterns of as­synthesized Cu1­x Gex /SiO2 catalysts. Reference spec­
tra: Cu (ICSD 235809), (Cu5 Ge)0.33 (ICSD 627677), Cu1.7 Ge0.3 (ICSD 627675). 59
4.21 (a) Catalytic mass activity of methanol during multiple temperature cycles
and (b) the mass activity (left scale, solid squares) and surface area normal­
ized activity (right scale, open squares) during the last temperature cycle,
of (1)­Cu0.85 Ge0.15 /SiO2 (black) and (2)­Cu0.85 Ge0.15 /SiO2 (blue). Reaction
conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1
and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . The normalized rate in subfigure
(b) is based on the crystallite size estimated from PXRD of post­mortem
samples. Note the two different scales in this subfigure. . . . . . . . . . . . 60
4.22 (a), (b) Catalytic mass activity and (c) surface­averaged mass activity of
a series of Cu(1­x) Gex /SiO2 catalysts. (d) Comparison of data in (a) with
the industrial­type Cu/ZnO/Al2 O3 /MgO catalyst (CZA, abbrev.). Reaction
conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of GHSV of
7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . No pre­treatment. . 63
4.23 PXRD patterns of post­mortem Cu(1­x) Gex /SiO2 catalysts, each spectrum
shifted on the y­axis. The data insert focuses on the peak position of the
main Cu(111) peak. Peaks at 60° are from SiC dilutant. . . . . . . . . . . . 64
4.24 Surface­averaged mass activity of a series of Cu(1­x) Gex /SiO2 catalysts
alongside that of the industrial­type Cu/ZnO/Al2 O3 /MgO catalyst. Reac­
tion conditions are 50 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of GHSV
of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu . No pre­treatment. 65
4.25 Surface statistics of a Cu FCC (a) a perfect cubo­octahedron and (b) an
imperfect cubo­octahedron crystal as a function of nanoparticle diameter.
Assuming Cu atomic diameter of 2.55 Å. Data used for the plot is from (146). 67
4.26 Types of crystals considered for Figure 4.25. (a) is a perfect cubo­octahedron,
(d) is an imperfect cubo­octahedron Cu FCC crystal. Reprinted with per­
mission from (152). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

163
List of Figures

4.27 PDF refinement of (a) as­synthesized and (b) post­mortem (1)­Cu0.85 Ge0.15 /SiO2 .
Green lines display the difference between data (black) and Cu model
(blue). In (c), the experimental PDFs are compared to the calculated PDFs
of Cu (red) and Cu3 Ge (blue). . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.28 HR­TEM image (top left) and FFT of a single particle observed in the
post­mortem sample of the (1)­Cu0.85 Ge0.15 /SiO2 sample. The FFTs were
obtained after applying a Hanning window over the real­space HR­TEM
image. The circles in the FFTs represent reciprocal distances of relevant
crystal structures and reference materials. Top right: FCC crystal struc­
tures. Bottom left: SiC dilutant measured with in­house XRD. Bottom
right: Hexagonal crystal structures. HR­TEM image was acquired and an­
alyzed by PhD Stefan Akazawa. . . . . . . . . . . . . . . . . . . . . . . . . 71
4.29 Catalytic mass activity of a series of Cu/X/C catalysts. Reaction conditions
are CO2 /H2 in 1:4 at 10 bar, and a WHSV of 1.3 × 104 Ncm3 · h−1 · g−1 Cu for
−1 −1
the Cu/X/C catalysts and 2 × 10 Ncm · h · gCu for the Cu/ZnO/Al2 O3 /MgO
5 3

catalyst. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

5.1 Expected (from nominal weight loadings) and measured (as determined by
SEM­EDS by PhD Mads Lützen) Zn/Cu ratios of selected catalysts in their
calcined (black) and postmortem (red) states. Line going through origo is
inserted to guide the eye. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.2 Volumetric concentration of methanol during a temperature cycle followed
by a 10­hour stability test at 240 ◦ C for CuZn/C catalysts with Zn/Cu ratios of
(a) 0.31 and (b) 1.5. Reaction conditions are 100 ml/min of CO2 /CO/H2 in
1:1:5 at 1 bar, with a WHSV of 6 × 105 Ncm3 · h−1 · g−1 Cu (3 × 10 Ncm · h
4 3 −1 · g−1 ). 80
cat
5.3 Catalytic mass activity of methanol on carbon­supported Cu and CuZn par­
ticles as a function of Zn loading, using glass reactors. Reaction con­
ditions are 100 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of
6 × 105 Ncm3 · h−1 · g−1Cu . Note the logarithmic scale. Lines are inserted to
guide the eye, and uncertainties are estimated by propagation of error (121). 81
5.4 (a) Volumetric concentration and (b) mass activity of methanol during two
consecutive temperature cycles for CuZn­0.13/C. (c) is the QMS signal
and temperature during the second temperature cycle. Reaction condi­
tions are 100 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of
6 × 105 Ncm3 · h−1 · g−1Cu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.5 XRD patterns of post­mortem Cu(Zn)/C catalysts with increasing amount
of Zn (upper to lower patterns). Data insert is focused on the Cu(111) peak
of the Cu/C and first 3 CuZn/C catalysts, and asterisks are used to highlight
the samples that have undergone a different activity testing. XRD pattern of
carbon black support is pictured for easy comparison. Reference spectra:
Cu (ICSD 235809, red), CuO (ICSD 16025, orange), Zn (ICSD 25286, blue)
and ZnO (ICSD 26170, grey). . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.6 Representative HAADF­STEM images of (a) CuZn­0.05/C (note the x10
scale bar in this sub­figure), (b) CuZn­0.13/C, (c) CuZn­0.62/C, and (d)
CuZn­1.0/C on which particle distribution of Table 5.3 and Figure F.4 are
based on. Images were acquired by PhD Mads Lützen. . . . . . . . . . . . 87
5.7 STEM­EDX image of Cu and Zn particles in close proximity in the post­
mortem (after CO2 /CO/H2 , and exposed to air) sample of CuZn­0.31/C.
Images were acquired by PhD Mads Lützen. . . . . . . . . . . . . . . . . . 88

164
 List of Figures

5.8 STEM­EDX image of single particles in the post­mortem sample of CuZn­

0.62/C where Cu and Zn are in close proximity. Images were acquired by
PhD Mads Lützen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.9 STEM­EDX image of a single Cu/Zn particle in the post­mortem sample of
CuZn­1.0/C. Notice the larger scale bar of 25 nm. Images were acquired
by PhD Mads Lützen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.10 Total scattering data of post­mortem Cu(Zn)/C samples: (a) reduced total
scattering data structure functions and (b) reduced atomic pair distribution
functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.11 Model­free analysis of the experimental PDFs of (a) (a)­Cu/C and (b) CuZn­
1.5/C with indicators for Cu and Zn references, and their respective oxides,
taken from calculated PDFs. . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.12 Experimental ex­situ PDFs of post­mortem samples of Cu(Zn)/C. . . . . . . 91
5.13 Surface­averaged mass activity of selected Cu(Zn)/C catalysts at 220 ◦ C,
all from activity testing in glass reactors. Reaction conditions are those
stated in Figure 5.3. Mass activities are normalized by the mean Cu crys­
tallite size obtained from PXRD. . . . . . . . . . . . . . . . . . . . . . . . . 94
5.14 Catalytic mass activity of methanol on carbon­supported Cu and CuZn par­
ticles as a function of Zn loading, using glass­lined steel reactors. Reaction
conditions are 50 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of
6 × 105 Ncm3 · h−1 · g−1Cu . Note the logarithmic scale. Lines are inserted to
guide the eye, and uncertainties are estimated by propagation of error (121). 95
5.15 Methanol mass activity of Cu(Zn)/C catalysts. Sub­figure (a) shows the
activity of Cu(Zn)/C catalysts as a function of Zn/(Zn+Cu) ratio at a tem­
perature of 220 ◦ C. The data for ’Conv. IWI, glass’ is the same as that in
Figure 5.3, and that for ’Conv. IWI, steel’ is from Figure 5.14. Sub­figure
(b) shows the activity during a temperature cycle for the two CuZn­0.06/C
catalysts prepared via conventional IWI and sequential IWI, both tested in
glass­lined steel reactors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

6.1 The data and command structure of the positioning system of the HPC­UHV
setup. Communication with the Raspberry Pi is through a socket server
sending and receiving positioning data and alarms. . . . . . . . . . . . . . . 102
6.2 The main tab of the GUI used to control the stepper motors on the parallel
screening setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.3 Schematic of the combined HPC­UHV setup showing the two horizontal
planes (X and Z). The third plane, the Y­direction, goes out of the plane, to­
wards the HSA. Annotated and reprinted with permission from (211). Copy­
right 2004 AIP Publishing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.4 First test sample after replacing the stepper motors on the parallel screen­
ing setup. In the sample cup, specialized for powder samples, two metals
foils of Cu and Au overlap, and a Ag wire is placed across the two foils.
The coordinate system relates to the motion of the manipulator arm. . . . . 105
6.5 XPS of the Cu/Au/Ag sample. Detailed scans in the region 325 −380 eV
as a function of position in the X­direction, with a starting position of (X,
Y, Z)=(348, ­3, 4) mm. XPS spectra obtained using Al Kα X­rays (12 kV,
20 mA), a pass energy of 40 eV, and an energy step of 0.4 eV. Pictured
spectra are averages of 20 consecutive scans. . . . . . . . . . . . . . . . . 106

165
List of Figures

6.6 XPS of the Cu/Au/Ag sample. Detailed scans in the region 325 −380 eV
as a function of position in the Z­direction, with a starting position of (X,
Y, Z)=(356, ­3, 4) mm. XPS spectra obtained using Al Kα X­rays (12 kV,
20 mA), a pass energy of 40 eV, and an energy step of 0.4 eV. Pictured
spectra are averages of 20 consecutive scans. . . . . . . . . . . . . . . . . 106
6.7 XPS of the Cu/Au/Ag sample, with focus on Cu 2p. Detailed scans in the
region 920 −940 eV as a function of position in the Z­direction, with a start­
ing position of (X, Y, Z)=(356, ­3, 4) mm. XPS spectra obtained using Al
Kα X­rays (12 kV, 20 mA), a pass energy of 40 eV, and an energy step of
0.4 eV. Pictured spectra are averages of 20 consecutive scans. . . . . . . . 107
6.8 XPS survey (average of 3 consecutive scans) of the Cu/Au/Ag sample in
the energy region 0 −1050 eV at the position (X, Y, Z)=(354, ­0.3, 11.5)mm.
Obtained using Al Kα X­rays (12 kV, 20 mA), a pass energy of 100 eV, and
an energy step of 1 eV. The primary XPS regions of metals and contami­
nants are indicated, and the arrow indicates unknown contaminant. . . . . . 108

B.1 In­situ XRD of the first reduction of CuO/GeO2 /SiO2 in a flow of 5% H2 /He
at 1 bar. Diffractograms are shown for each temperature step, and they are
the last recorded at each temperature. Temperature range for this reduction
is 160 −550 ◦ C. Lower and upper diffractograms are long end­scans taken
at 25 ◦ C. The diffractograms are corrected for the sample displacement
observed at room temperature in relation to the GeO2 reference. . . . . . . 114
B.2 In­situ XRD of the second reduction of CuO/GeO2 /SiO2 in a flow of 5%
H2 /He at 1 bar. Diffractograms are shown for each temperature step, and
they are the last recorded at each temperature. Temperature range for
this reduction is 300 −700 ◦ C. Lower and upper diffractograms are long
end­scans taken at 25 ◦ C. The diffractograms are corrected for the sam­
ple displacement observed at room temperature in relation to the GeO2
reference. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
B.3 In­situ X­ray patterns of CuO/GeO2 /SiO2 after reduction at 550 ◦ C and 700 ◦ C.
Diffractograms are taken at 25 ◦ C and are not shifted for easy comparison of
intensities. The diffractograms are corrected for the sample displacement
observed at room temperature in relation to the GeO2 reference. Refer­
ence patterns: Cu (ICSD 235809), Ge (ICSD 121532), and Cu3 Ge (ICSD
53266). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
B.4 In­situ XRD of the calcination of Cu/Ge/SiO2 in a flow of 20% O2 /N2 at 1
bar. Diffractograms are shown for each temperature step, and they are the
last recorded at each temperature. Temperature range for the calcination
is 200 −550 ◦ C. Lower and upper diffractograms are long end­scans taken
at 25 ◦ C. The diffractograms are corrected for the sample displacement
observed at room temperature in relation to the GeO2 reference. . . . . . . 117
B.5 XRD pattern of four individual oleylamine­based synthesis products. Refer­
ence patterns are: Cu (ICSD 235809), Cu3 Ge (ICSD 53266), (Cu5 Ge)0.33
(ICSD 627677), and Cu1.7 Ge0.3 (ICSD 627675). . . . . . . . . . . . . . . . . 118
B.6 XRD pattern of as­synthesized Cu0.54 Ge0.46 nanoparticles and reference
patterns of Cu (ICSD 235809), Cu3 Ge (ICSD 53266), (Cu5 Ge)0.33 (ICSD
627677), Cu1.7 Ge0.3 (ICSD 627675), Cu5 Ge2 (ICSD 87340), and Ge (ICSD
121532). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
B.7 Zoom­in of diffraction pattern in Figure B.6. . . . . . . . . . . . . . . . . . . 119
B.8 XRD pattern of as­synthesized (1)­Cu1 and (2)­Cu1 . . . . . . . . . . . . . . 120

166
 List of Figures

B.9 XRD pattern of as­synthesized (1)­Cu1 nanoparticles and best fits. . . . . . 121
B.10 PXRD patterns of as­synthesized and post­mortem Cu0.67 Ge0.33 /SiO2 , each
spectrum shifted on the y­axis. Reference patterns: Cu (ICSD 235809),
Cu3 Ge (ICSD 53266), Cu1.7 Ge0.3 (ICSD 627675). . . . . . . . . . . . . . . 122
B.11 PXRD patterns of as­synthesized and post­mortem Cu0.94 Ge0.06 /SiO2 , each
spectrum shifted on the y­axis. Reference pattern: Cu (ICSD 235809). . . . 122
B.12 PXRD patterns of as­synthesized and post­mortem Cu0.96 Ge0.04 /SiO2 , each
spectrum shifted on the y­axis. Reference pattern: Cu (ICSD 235809). . . . 123
B.13 XPS of adhesive carbon tape. . . . . . . . . . . . . . . . . . . . . . . . . . . 124
B.14 XPS of as­synthesized Cu0.82 Ge0.18 /SiO2 powder sample pressed on ad­
hesive carbon tape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
B.15 XPS of as­synthesized Cu0.96 Ge0.04 /SiO2 powder sample pressed on ad­
hesive carbon tape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
B.16 Catalytic mass activity of 17 wt% Cu/γ­Al2 O3 prepared by IWI. Reaction
conditions are 150 ml/min of CO2 /H2 in 1:4 at 10 bar, GHSV of 1.4 × 104 h−1
and a WHSV of 7.4 × 105 Ncm3 · h−1 · g−1 Cu . Uncertainties are estimated by
propagation of error (121). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
B.17 Catalytic mass activity of 17 wt% Cu/SiO2 prepared by IWI. Reaction condi­
tions are CO2 /H2 in 1:4 at 10 bar. Open squares have a WHSV of 7.4 × 105 Ncm3 · h−1 · g−1Cu ;
−1 −1
Closed squares have a WHSV of 2.4 × 10 Ncm · h · gCu . Post­mortem
5 3

Cu crystallite sizes from PXRD are 91 nm and 101 nm for the open­ and
closed­squares, respectively. Uncertainties are estimated by propagation
of error (121). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
B.18 Catalytic mass activity and methanol selectivity of 4.3 wt% Cu/SiO2 syn­
thesized via the method described in Section 3.1.2 and 17 wt% Cu/SiO2
prepared by IWI. Reaction conditions are CO2 /H2 in 1:4 at 10 bar, and a
WHSV of approximately 8 × 105 Ncm3 · h−1 · g−1Cu . . . . . . . . . . . . . . . . 127
B.19 Volumetric concentration of methanol during repeated temperature cycles
of (2)­Cu0.85 Ge0.15 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2 in
1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .
Black squares represent individual GC injections. . . . . . . . . . . . . . . . 128
B.20 Volumetric concentration of products (methanol, methane, CO) during re­
peated temperature cycles of (2)­Cu1 /SiO2 . Reaction conditions are 50 ml/min
of CO2 /H2 in 1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1
Cu .
Dots represent individual GC injections. . . . . . . . . . . . . . . . . . . . . 128
B.21 Volumetric concentration of products during repeated temperature cycles
of Cu0.67 Ge0.33 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2 in 1:4
at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .
Dots represent individual GC injections. . . . . . . . . . . . . . . . . . . . . 129
B.22 Volumetric concentration of products during repeated temperature cycles
of Cu0.96 Ge0.04 /SiO2 . Reaction conditions are 50 ml/min of CO2 /H2 in 1:4
at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .
Dots represent individual GC injections. . . . . . . . . . . . . . . . . . . . . 129
B.23 Volumetric concentration of products during repeated temperature cycles
of Cu/ZnO/Al2 O3 /MgO. Reaction conditions are 50 ml/min of CO2 /H2 in 1:4
at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .
Dots represent individual GC injections. . . . . . . . . . . . . . . . . . . . . 130

167
List of Figures

B.24 (a) Catalytic mass activity of methanol during multiple temperature cycles
of (1)­Cu0.94 Ge0.06 /SiO2 (black) and (2)­Cu0.94 Ge0.06 /SiO2 (blue) and (b)
the mass activity of (2)­Cu0.94 Ge0.06 /SiO2 before and after reduction in 5 %
H2 /Ar at 220 ◦ C for 3 hours. Reaction conditions are 50 ml/min of CO2 /H2 in
1:4 at 10 bar, GHSV of 7.7 × 103 h−1 and a WHSV of 8 × 105 Ncm3 · h−1 · g−1 Cu .130
B.25 XRD pattern of used 17 wt% Cu/SiO2 and Cu/γ­Al2 O3 . Reference pattern
of Cu (ICSD 235809). Cu crystallite sizes as determined with the Scherrer
equation: 91 nm for Cu/SiO2 , and 123 nm for Cu/γ­Al2 O3 . . . . . . . . . . . 131
B.26 Experimental PDF of post­mortem (1)­Cu0.85 Ge0.15 /SiO2 with (a) no back­
ground subtraction and with the individual backgrounds, and (c) the back­
ground from individually scaling each of the background contributions. Sub­
figure (b) shows the subtraction of kapton and (d) illustrates how the back­
ground contributes to the noise in the experimental PDF. . . . . . . . . . . . 133
B.27 Ex­situ SEM­EDX of large, agglomerated particle in the post­mortem sam­
ple of the (1)­Cu0.85 Ge0.15 /SiO2 catalyst. Image was acquired and analyzed
by PhD Filippo Romeggio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
B.28 TEM image of post­mortem sample of (1)­Cu0.85 Ge0.15 /SiO2 catalyst show­
ing thick, largely crystalline, particle. TEM image was acquired by PhD
Stefan Akazawa. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
B.29 HR­TEM image and corresponding FFT of the region displayed by a red
box in Figure B.28. Middle: FFT of HR­TEM image with rings indicating k­
space values for each of the labelled materials (SiC, Cux Oy , and Cux Gey ).
Right: Close up of up a single peak in the FFT image. HR­TEM image was
acquired and analyzed by PhD Stefan Akazawa. . . . . . . . . . . . . . . . 135
B.30 SEM(­EDX) of a Cu­based catalysts containing nominal Ge loading of 20
wt%. EDX spectrum shows the presence of Si (from support material), Cu
and Cl, but no Ge. Image was acquired and analyzed by Daniel Graun­
gaard Bruun. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
B.31 SEM(­EDX) of a Cu­based catalysts containing nominal Ge loading of 30
wt%. EDX spectrum shows the presence of Si (from support material), Cu
and Cl, but no Ge. Image was acquired and analyzed by Daniel Graun­
gaard Bruun. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

C.1 Mass activity of methanol of an industrial­grade Cu/ZnO/Al2 O3 /MgO cat­

alyst during (a) consecutive activity tests in CO2 /CO/H2 of varying ratios
and (b) in CO2 /CO/H2 at varying pressures. Reaction conditions are a to­
tal gas flow of 150 ml/min, 10 bar, a GHSV of 2.3 × 104 h−1 and a WHSV
of 2.7 × 105 Ncm3 · h−1 · g−1Cu . Pre­treated via reduction in 5 % H2 /Ar at
220 ◦ C for 3 hours. Experimental history: Reduction, CO2 /CO/H2 , CO2 /H2 ,
and lastly CO/H2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
C.2 QMS signal and temperature during (a) reduction and consecutive activ­
ity tests of the industrial­grade Cu/ZnO/Al2 O3 /MgO catalyst. Sub­figures
(b) through (d) are in different gas feeds. Reaction conditions are those
described in the figure text for Figure C.1. . . . . . . . . . . . . . . . . . . . 138

D.1 Methanol mass activity of a series of Cu/X/C catalysts. Normalized with

the amount of Cu. Reaction conditions are CO2 /H2 in 1:4 at 10 bar, and a
WHSV of 1.3 × 105 Ncm3 · h−1 · g−1 Cu . . . . . . . . . . . . . . . . . . . . . . . 139
D.2 Methanol mass activity of a series of Cu/X/C catalysts. Normalized with the
amount of Cu and promoter. Reaction conditions are CO2 /H2 in 1:4 at 10
bar, and a WHSV of 1.3 × 105 Ncm3 · h−1 · g−1 Cu . . . . . . . . . . . . . . . . . 139

168
 List of Figures

D.3 PXRD pattern of post­mortem Cu/X/C catalysts (X=Sn, Bi, Pb, Sb) and
reference patterns of Cu (ICSD 235809), CuO (ICSD 16025), Cu2 O (ICSD
52043), and SiC dilutant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

E.1 Examples of the numerical integration of GC peaks. The peaks are pictured
during the same time of a 80­hour long experiment while testing the (1)­
Cu0.94 Ge0.06 /SiO2 catalyst. . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
E.2 Integrated FID and TCD signals. The data is from an a 80­hour long ex­
periment using the (1)­Cu0.94 Ge0.06 /SiO2 catalyst. Points are individual GC
injections, each of them integrated as illustrated in Figure E.1. . . . . . . . 142
E.3 Volumetric concentration of methanol during repeated temperature cycles
while testing the (1)­Cu0.94 Ge0.06 /SiO2 catalyst. Points are the individual
GC injections from Figure E.2, converted into volumetric concentrations
using the calibrations factors specific for each gas. . . . . . . . . . . . . . . 143
E.4 Effect of reactor pressure during methanol synthesis in a CO2 /CO/H2 gas
feed. Pressure on the back­side of the GC set to 850 mbar. . . . . . . . . . 145
E.5 CO2 hydrogenation using 5 g CZA catalyst with different flow rates and
temperatures (a). Pressure is set to 1 bar and GC back­pressure set to
200 mbar to simulate conditions from experiments using glass reactors.
Subfigures (b), (c) are integrated TCD signals (CO, CO2 and H2 ) and FID
signals (methanol), and subfigure (d) is the volumetric concentrations. Ex­
periment performed by PhD Thomas Smitshuysen. . . . . . . . . . . . . . . 146
E.6 On­line GC­sampler, Agilent 7890A. The gas is lead in by the TCD sampling
(’Sample # 1 in’) and exits again by the FID sampling (’Sample # 2 out’).
The schematic is annotated and reprinted from (85). . . . . . . . . . . . . . 147

F.1 (a) Volumetric concentration and (b) mass activity of methanol during a
temperature cycle for CuZn­0.05/C. Reaction conditions are 100 ml/min of
CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of 6 × 105 Ncm3 · h−1 · g−1 Cu . . . . 148
F.2 Mass activity of methanol over a CuZn­0.62/C catalyst. Activity is tested in
glass reactors (black, 200 mg sample and 100 ml/min total gas flow) and
glass­lined steel reactors (blue, 100 mg sample, SiC dilution, and 50 ml/min
total gas flow). Uncertainties are estimated by propagation of error (121). . 149
F.3 Error estimation using two batches of CuZn/C (Zn/Cu=0.62). Reaction con­
ditions are 100 ml/min of CO2 /CO/H2 in 1:1:5 at 1 bar, with a WHSV of
6 × 105 Ncm3 · h−1 · g−1Cu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
F.4 HAADF­STEM size distributions of CuZn/C catalysts post activity testing.
Note the much larger x­axis used for CuZn­0.05/C. . . . . . . . . . . . . . . 150
F.5 STEM­EDX images of separate (a) Cu and (b) Zn particles in the post­
mortem sample of CuZn­1.0/C post activity testing in syngas at 1 bar, both
on a 10 nm scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
F.6 STEM­EDX image of CuZn­0.31/C . . . . . . . . . . . . . . . . . . . . . . . 151
F.7 STEM­EDX image of CuZn­0.62/C . . . . . . . . . . . . . . . . . . . . . . . 151
F.8 In­situ XRD study of the state of a Cu/C catalyst during (a) reduction. Sub­
figure (b) shows long scans taken of the sample in different states (as­
synthesized, reduced, post CO2 /CO/H2 , and after being exposed to ambi­
ent air). In sub­figure (b), the diffractograms are shifted on the y­axis. . . . 152
F.9 QMS signal during activity testing in CO2 /CO/H2 of (a) 5 wt % Cu/C catalyst,
(b) a Cu/C catalyst contaminated with Ni, and (c) the in­situ sample holder. 153
F.10 Ni contamination of a Cu/C catalyst during activity testing in syngas. . . . . 154

169
List of Figures

F.11 (a) Total scattering data, (b) reduced total scattering data structure function
and (c) PDF for a carbon­supported CuZn sample. . . . . . . . . . . . . . . 155
F.12 Comparison of experimental PDF of (a)­Cu/C with calculated PDFs of Cu,
CuO and Cu2 O. Model of FCC Cu is used in the subsequent refinement. . . 156
F.13 1­phase PDF refinement of post­mortem (a)­Cu/C. The blue curve is cal­
culated from a structural model (FCC Cu, ICSD coll. code 235809), which
is refined to give the best possible fit to the experimental PDF (black). The
green line displays the difference between the model and experimental
data. Refined parameters are listed in Table F.1. . . . . . . . . . . . . . . . 157
F.14 2­phase PDF refinement of post­mortem CuZn­0.05/C. The blue curve is
refined to give the best possible fit to the experimental PDF (black). The
green line displays the difference between the model and experimental
data. Refined parameters are listed in Table F.2. . . . . . . . . . . . . . . . 158
F.15 Rietveld refinement of post­mortem samples of (a) (a)­Cu/C and (b) CuZn­
0.05/C. The dashed lines indicate the position of the fundamental parame­
ters peaks used to describe the carbon support. . . . . . . . . . . . . . . . 159
F.16 XRD patterns of post­mortem Cu(Zn)/C catalysts with increasing amount
of Zn (upper to lower patterns). Samples prepared via sequential IWI.
Reference spectra: PBX51 (in­house measurement), Cu (ICSD 235809,
red), Zn (ICSD 25286, blue) and SiC dilutant (ICSD 19706, dashed black). 160
F.17 XRD patterns of post­mortem Cu(Zn)/C catalysts with increasing amount
of Zn (upper to lower patterns). From catalytic testing in glass­lined steel
reactors. Reference spectra: PBX51 (in­house measurement), Cu (ICSD
235809, red), Zn (ICSD 25286, blue) and SiC dilutant (ICSD 19706, dashed
black). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

170
 List of Tables

List of Tables
4.1 Properties of the SiO2 ­supported Cu(1­x) Gex catalysts. The most important
ones are marked with a symbol *. . . . . . . . . . . . . . . . . . . . . . . . . 61

5.1 Nominal weight loadings (from weighing of chemicals) and measured weight
loadings (from ICP­OES) of selected carbon­supported CuZn catalysts.
ICP­OES was measured by Konrad Herbst at Topsøe. . . . . . . . . . . . . 78
5.2 Optimal Cu/Zn compositions in literature . . . . . . . . . . . . . . . . . . . . 82
5.3 Properties of selected Cu(Zn)/C catalysts used in experiments with glass
reactors. Asterisks are used to highlight the samples that have undergone
a different activity testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

A.1 In­situ XRD instrument broadening of the XRD peaks measured on a Si

standard sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
A.2 Ex­situ XRD instrument broadening of the XRD peaks measured on a Si
standard sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

B.1 Synthesis parameters used in oleylamine­based synthesis of CuGe nanopar­

ticles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
B.2 Size estimates of Cu crystallites of the investigated nanoparticles. Mea­
surements done at UCPH. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
B.3 Refined parameters for the as­synthesized Cu0.67 Ge0.33 /SiO2 using 1 phase
(FCC Cu) or 2 phases (FCC Cu and Cu3 Ge). . . . . . . . . . . . . . . . . . 132
B.4 Refined parameters for the as­synthesized and post­mortem samples of
(1)­Cu0.85 Ge0.15 /SiO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

E.1 Table of retention times and calibration factors for conversion of integrated
data to concentrations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

F.1 Refined parameters for the post­mortem sample of (a)­Cu/C using a one­
phase approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
F.2 Refined parameters for the post­mortem sample of CuZn­0.05/C using a
two­phase approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

171
References

References
(1) IPCC, Climate Change 2021: The Physical Science Basis. Contribution of Working
Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate
Change; Cambridge University Press: Cambridge, United Kingdom and New York,
NY, USA, 2021; Vol. In Press.
(2) IPCC In Climate Change 2021: The Physical Science Basis. Contribution of Work­
ing Group I to the Sixth Assessment Report of the Intergovernmental Panel on Cli­
mate Change, Masson­Delmotte, V., Zhai, P., Pirani, A., Connors, S., Péan, C.,
Berger, S., Caud, N., Chen, Y., Goldfarb, L., Gomic, M., Huang, M., Leitzell, K.,
Lonnoy, E., Matthews, J., Maycock, T., Waterfield, T., Yelekçi, O., Yu, R., and
Zhou, B., Eds.; Cambridge University Press: Cambridge, United Kingdom and
New York, NY, USA, 2021, pp 3–32.
(3) NOAA National Centers for Environmental Information State of the Climate: Global
Climate Report for 2022, 2023.
(4) Keeling, C. D., Bacastow, R. B., Bainbridge, A. E., Ekdahl, C. A., Guenther, P. R.,
Waterman, L. S., and Chin, J. F. S. (1976). Atmospheric carbon dioxide variations
at Mauna Loa Observatory, Hawaii. Tellus 28, 538–551.
(5) Friedlingstein, P., Jones, M. W., O’Sullivan, M., Andrew, R. M., Bakker, D. C.,
Hauck, J., Le Quéré, C., Peters, G. P., Peters, W., Pongratz, J., Sitch, S., Canadell,
J. G., Ciais, P., Jackson, R. B., Alin, S. R., Anthoni, P., Bates, N. R., Becker, M.,
Bellouin, N., Bopp, L., Chau, T. T. T., Chevallier, F., Chini, L. P., Cronin, M., Cur­
rie, K. I., Decharme, B., Djeutchouang, L. M., Dou, X., Evans, W., Feely, R. A.,
Feng, L., Gasser, T., Gilfillan, D., Gkritzalis, T., Grassi, G., Gregor, L., Gruber,
N., Gürses, Ö., Harris, I., Houghton, R. A., Hurtt, G. C., Iida, Y., Ilyina, T., Lui­
jkx, I. T., Jain, A., Jones, S. D., Kato, E., Kennedy, D., Goldewijk, K. K., Knauer,
J., Korsbakken, J. I., Körtzinger, A., Landschützer, P., Lauvset, S. K., Lefèvre,
N., Lienert, S., Liu, J., Marland, G., McGuire, P. C., Melton, J. R., Munro, D. R.,
Nabel, J. E., Nakaoka, S. I., Niwa, Y., Ono, T., Pierrot, D., Poulter, B., Rehder,
G., Resplandy, L., Robertson, E., Rödenbeck, C., Rosan, T. M., Schwinger, J.,
Schwingshackl, C., Séférian, R., Sutton, A. J., Sweeney, C., Tanhua, T., Tans,
P. P., Tian, H., Tilbrook, B., Tubiello, F., Van Der Werf, G. R., Vuichard, N., Wada,
C., Wanninkhof, R., Watson, A. J., Willis, D., Wiltshire, A. J., Yuan, W., Yue, C.,
Yue, X., Zaehle, S., and Zeng, J. (2022). Global Carbon Budget 2021. Earth Sys­
tem Science Data 14, 1917–2005.
(6) Smalley, R. E. (2005). Future global energy prosperity: The terawatt challenge.
MRS Bulletin 30, 412–417.
(7) Olah, G. A. (2005). Beyond Oil and Gas: The Methanol Economy. Angewandte
Chemie International Edition 44, 2636–2639.
(8) Olah, G. A., Goeppert, A., and Prakash, G. K. S. (2009). Chemical Recycling of
Carbon Dioxide to Methanol and Dimethyl Ether: From Greenhouse Gas to Re­
newable, Environmentally Carbon Neutral Fuels and Synthetic Hydrocarbons. The
Journal of Organic Chemistry 74, 487–498.
(9) Olah, G. A. (2013). Towards Oil Independence Through Renewable Methanol
Chemistry. Angewandte Chemie International Edition 52, 104–107.
(10) Goeppert, A., Czaun, M., Jones, J.­P., Surya Prakash, G. K., and Olah, G. A.
(2014). Recycling of carbon dioxide to methanol and derived products – closing
the loop. Chemical Society Reviews 43, 7995–8048.

172
 References

(11) Ehteshami, S. M. M., and Chan, S. H. (2014). The role of hydrogen and fuel cells to
store renewable energy in the future energy network – potentials and challenges.
Energy Policy 73, 103–109.
(12) Bozzano, G., and Manenti, F. (2016). Efficient methanol synthesis: Perspectives,
technologies and optimization strategies. Progress in Energy and Combustion Sci­
ence 56, 71–105.
(13) Sanchez, D. P., Collina, L., Levi, P., and Hodgson, D. Chemicals, tech. rep., Paris:
IEA, 2022.
(14) Laidler, K. J. (1996). A glossary of terms used in chemical kinetics, including reac­
tion dynamics (IUPAC recommendations 1996). Pure and Applied Chemistry 68,
149–192.
(15) IUPAC, Compendium of Chemical Terminology.
(16) Patart, M. (1925). French Patent 540, 343 (Aug. 1921). Badische Anilin und Soda
Fabrik, DR Patents 41.5, 686; 441, 433; and 462, 837 (1923) and US Patents
1,558, 559 and 1,569,775 (1923). Lormand, C. Ind. Eng. Che 17.
(17) Lormand, C. (1925). Industrial Production of Synthetic Methanol. Industrial and
Engineering Chemistry 17, 430–432.
(18) Sehested, J. (2019). Industrial and scientific directions of methanol catalyst de­
velopment. Journal of Catalysis 371, 368–375.
(19) Roine, A. HSC Chemistry 7, 2011.
(20) Frolich, P. K., Fenske, M. R., Taylor, P. S., and Southwich, C. A. (1928). Catalysts
for the Formation of Alcohols from Carbon Monoxide and Hydrogen. Industrial and
Engineering Chemistry 20, 1327–1330.
(21) Storch, H. H. (1928). Behavior of zinc oxide and zinc oxide­chromium oxide cata­
lysts in the decomposition and synthesis of methanol. Journal of Physical Chem­
istry 32, 1743–1747.
(22) Cryder, D. S., and Frolich, P. K. (1929). Catalysts for the Formation of Alco­
hols from Carbon Monoxide and Hydrogen: IV­Decomposition and Synthesis of
Methanol by Catalysts Composed of Zinc and Chromium Oxides. Industrial and
Engineering Chemistry 21, 867–871.
(23) Chorkendorff, I., and Niemantsverdriet, J. W., Concepts of Modern Catalysis and
Kinetics; Wiley�VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2003.
(24) Liu, G., Willcox, D., Garland, M., and Kung, H. H. (1984). The rate of methanol
production on a copper­zinc oxide catalyst: The dependence on the feed compo­
sition. Journal of Catalysis 90, 139–146.
(25) Chinchen, G. C., Denny, P. J., Parker, D. G., Spencer, M. S., and Whan, D. A.
(1987). Mechanism of methanol synthesis from CO2/CO/H2 mixtures over cop­
per/zinc oxide/alumina catalysts: use of 14C­labelled reactants. Applied Catalysis
30, 333–338.
(26) Muhler, M., Törnqvist, E., Nielsen, L. P., Clausen, B. S., and Topsøe, H. (1994). On
the role of adsorbed atomic oxygen and CO2 in copper based methanol synthesis
catalysts. Catalysis Letters 25, 1–10.
(27) Nakamura, I., Fujitani, T., Uchijima, T., and Nakamura, J. (1996). A model catalyst
for methanol synthesis: Zn­deposited and Zn­free Cu surfaces. Journal of Vacuum
Science & Technology A 14, 1464–1468.
(28) Behrens, M., Studt, F., Kasatkin, I., Kühl, S., Hävecker, M., Abild­Pedersen, F.,
Zander, S., Girgsdies, F., Kurr, P., Kniep, B.­L., Tovar, M., Fischer, R. W., Nørskov,
J. K., and Schlögl, R. (2012). The Active Site of Methanol Synthesis over Cu/ZnO/Al2O3
Industrial Catalysts. Science 336, 893–897.

173
References

(29) Studt, F., Behrens, M., Kunkes, E. L., Thomas, N., Zander, S., Tarasov, A., Schu­
mann, J., Frei, E., Varley, J. B., Abild­Pedersen, F., Nørskov, J. K., and Schlögl,
R. (2015). The Mechanism of CO and CO2 Hydrogenation to Methanol over Cu­
Based Catalysts. ChemCatChem 7, 1105–1111.
(30) Yang, Y., Mei, D., Peden, C. H. F., Campbell, C. T., and Mims, C. A. (2015).
Surface­Bound Intermediates in Low­Temperature Methanol Synthesis on Cop­
per: Participants and Spectators. ACS Catalysis 5, 7328–7337.
(31) Wilkinson, S. K., Van De Water, L. G., Miller, B., Simmons, M. J., Stitt, E. H.,
and Watson, M. J. (2016). Understanding the generation of methanol synthesis
and water gas shift activity over copper­based catalysts – A spatially resolved
experimental kinetic study using steady and non­steady state operation under
CO/CO2/H2 feeds. Journal of Catalysis 337, 208–220.
(32) Kattel, S., Ramírez, P. J., Chen, J. G., Rodriguez, J. A., and Liu, P. (2017). Ac­
tive sites for CO2 hydrogenation to methanol on Cu/ZnO catalysts. Science 355,
1296–1299.
(33) Kunkes, E. L., Studt, F., Abild­Pedersen, F., Schlögl, R., and Behrens, M. (2015).
Hydrogenation of CO2 to methanol and CO on Cu/ZnO/Al2O3: Is there a common
intermediate or not? Journal of Catalysis 328, 43–48.
(34) Grabow, L. C., and Mavrikakis, M. (2011). Mechanism of Methanol Synthesis on
Cu through CO2 and CO Hydrogenation. ACS Catalysis 1, 365–384.
(35) Askgaard, T. S., Nørskov, J. K., Ovesen, C. V., and Stoltze, P. (1995). A Kinetic
Model of Methanol Synthesis. Journal of Catalysis 156, 229–242.
(36) Schwiderowski, P., Ruland, H., and Muhler, M. (2022). Current developments in
CO2 hydrogenation towards methanol: A review related to industrial application.
Current Opinion in Green and Sustainable Chemistry 38, 100688.
(37) Smitshuysen, T. E. L. Investigation of Alloy Catalysts for Conversion of Sustain­
able Energy, Ph.D. Thesis, Lyngby: Technical University of Denmark, 2022.
(38) Rasmussen, P. B., Kazuta, M., and Chorkendorff, I. (1994). Synthesis of methanol
from a mixture of H2 and CO2 on Cu(100). Surface Science 318, 267–280.
(39) Rasmussen, P. B., Holmblad, P. M., Askgaard, T., Ovesen, C. V., Stoltze, P.,
Nørskov, J. K., and Chorkendorff, I. (1994). Methanol synthesis on Cu(100) from
a binary gas mixture of CO2 and H2. Catalysis Letters 1994 26:3 26, 373–381.
(40) Yoshihara, J., and Campbell, C. T. (1996). Methanol Synthesis and Reverse Water–
Gas Shift Kinetics over Cu(110) Model Catalysts: Structural Sensitivity. Journal of
Catalysis 161, 776–782.
(41) Nakamura, J., Nakamura, I., Uchijima, T., Watanabe, T., and Fujitani, T. (1996).
Model studies of methanol synthesis on copper catalysts. Studies in Surface Sci­
ence and Catalysis 101, 1389–1399.
(42) Nerlov, J., and Chorkendorff, I. (1998). Promotion through gas phase induced sur­
face segregation: Methanol synthesis from CO, CO2 and H2 over Ni/Cu(100).
Catalysis Letters 54, 171–176.
(43) Behrens, M., Zander, S., Kurr, P., Jacobsen, N., Senker, J., Koch, G., Ressler, T.,
Fischer, R. W., and Schlögl, R. (2013). Performance Improvement of Nanocata­
lysts by Promoter­Induced Defects in the Support Material: Methanol Synthesis
over Cu/ZnO:Al. Journal of the American Chemical Society 135, 6061–6068.
(44) Spencer, M. S. (1987). α­brass formation in copper/zinc oxide catalysts: II. Dif­
fusion of zinc in copper and α­brass under reaction conditions. Surface Science
192, 329–335.

174
 References

(45) Nakamura, J., Nakamura, I., Uchijima, T., Kanai, Y., Watanabe, T., Saito, M., and
Fujitani, T. (1996). A Surface Science Investigation of Methanol Synthesis over a
Zn­Deposited Polycrystalline Cu Surface. Journal of Catalysis 160, 65–75.
(46) Nakamura, J., Uchijima, T., Kanai, Y., and Fujitani, T. (1996). The role of ZnO in
Cu/ZnO methanol synthesis catalysts. Catalysis Today 28, 223–230.
(47) Fujitani, T., Nakamura, I., Uchijima, T., and Nakamura, J. (1997). The kinetics and
mechanism of methanol synthesis by hydrogenation of CO2 over a Zn­deposited
Cu(111) surface. Surface Science 383, 285–298.
(48) Kurtz, M., Wilmer, H., Genger, T., Hinrichsen, O., and Muhler, M. (2003). Deac­
tivation of Supported Copper Catalysts for Methanol Synthesis. Catalysis Letters
86, 77–80.
(49) Hansen, J. B., and Højlund Nielsen, P. E. In Handbook of Heterogeneous Catal­
ysis, 2008, pp 2920–2949.
(50) Kuld, S., Conradsen, C., Moses, P. G., Chorkendorff, I., and Sehested, J. (2014).
Quantification of Zinc Atoms in a Surface Alloy on Copper in an Industrial­Type
Methanol Synthesis Catalyst. Angewandte Chemie International Edition 53, 5941–
5945.
(51) Kuld, S., Thorhauge, M., Falsig, H., Elkjær, C. F., Helveg, S., Chorkendorff, I.,
and Sehested, J. (2016). Quantifying the promotion of Cu catalysts by ZnO for
methanol synthesis. Science 352, 969–974.
(52) van den Berg, R., Prieto, G., Korpershoek, G., van der Wal, L. I., van Bunningen,
A. J., Lægsgaard­Jørgensen, S., de Jongh, P. E., and de Jong, K. P. (2016). Struc­
ture sensitivity of Cu and CuZn catalysts relevant to industrial methanol synthesis.
Nature Communications 7, 13057.
(53) Kattel, S., Ramírez, P. J., Chen, J. G., Rodriguez, J. A., and Liu, P. (2017). Re­
sponse to Comment on ”Active sites for CO2 hydrogenation to methanol on Cu/ZnO
catalysts”. Science 357, 8210.
(54) Nakamura, J., Fujitani, T., Kuld, S., Helveg, S., Chorkendorff, I., and Sehested, J.
(2017). Comment on ”Active sites for CO2 hydrogenation to methanol on Cu/ZnO
catalysts”. Science 357.
(55) Nakamura, J., Choi, Y., and Fujitani, T. (2003). On the Issue of the Active Site
and the Role of ZnO in Cu/ZnO Methanol Synthesis Catalysts. Topics in Catalysis
2003 22:3 22, 277–285.
(56) Clausen, B. S., Schiøtz, J., Gråbæk, L., Ovesen, C. V., Jacobsen, K. W., Nørskov,
J. K., and Topsøe, H. (1994). Wetting/non­wetting phenomena during catalysis:
Evidence from in situ on­line EXAFS studies of Cu­based catalysts. Topics in
Catalysis 1, 367–376.
(57) Ovesen, C. V., Clausen, B. S., Schiøtz, J., Stoltze, P., Topsøe, H., and Nørskov,
J. K. (1997). Kinetic Implications of Dynamical Changes in Catalyst Morphology
during Methanol Synthesis over Cu/ZnO Catalysts. Journal of Catalysis 168, 133–
142.
(58) Grunwaldt, J. D., Molenbroek, A. M., Topsøe, N. Y., Topsøe, H., and Clausen, B. S.
(2000). In Situ Investigations of Structural Changes in Cu/ZnO Catalysts. Journal
of Catalysis 194, 452–460.
(59) Hansen, P. L., Wagner, J. B., Helveg, S., Rostrup­Nielsen, J. R., Clausen, B. S.,
and Topsøe, H. (2002). Atom­Resolved Imaging of Dynamic Shape Changes in
Supported Copper Nanocrystals. Science 295, 2053–2055.
(60) Vesborg, P. C. K., Chorkendorff, I., Knudsen, I., Balmes, O., Nerlov, J., Molen­
broek, A. M., Clausen, B. S., and Helveg, S. (2009). Transient behavior of Cu/ZnO­
based methanol synthesis catalysts. Journal of Catalysis 262, 65–72.

175
References

(61) Fujitani, T., and Nakamura, J. (1998). The effect of ZnO in methanol synthesis
catalysts on Cu dispersion and the specific activity. Catalysis Letters 56, 119–
124.
(62) Topsøe, N.­Y., and Topsøe, H. (1999). On the nature of surface structural changes
in Cu/ZnO methanol synthesis catalysts. Topics in Catalysis 8, 267–270.
(63) Sano, M., Adaniya, T., Fujitani, T., and Nakamura, J. (2002). Formation Process
of a Cu−Zn Surface Alloy on Cu(111) Investigated by Scanning Tunneling Mi­
croscopy. The Journal of Physical Chemistry B 106, 7627–7633.
(64) Dalebout, R., Barberis, L., Totarella, G., Turner, S. J., La Fontaine, C., de Groot,
F. M. F., Carrier, X., van der Eerden, A. M. J., Meirer, F., and de Jongh, P. E.
(2022). Insight into the Nature of the ZnOx Promoter during Methanol Synthesis.
ACS Catalysis, 6628–6639.
(65) Schott, V., Oberhofer, H., Birkner, A., Xu, M., Wang, Y., Muhler, M., Reuter, K.,
and Wöll, C. (2013). Chemical Activity of Thin Oxide Layers: Strong Interactions
with the Support Yield a New Thin­Film Phase of ZnO. Angewandte Chemie In­
ternational Edition 52, 11925–11929.
(66) Lunkenbein, T., Schumann, J., Behrens, M., Schlögl, R., and Willinger, M. G.
(2015). Formation of a ZnO Overlayer in Industrial Cu/ZnO/Al2O3 Catalysts In­
duced by Strong Metal–Support Interactions. Angewandte Chemie International
Edition 54, 4544–4548.
(67) Lunkenbein, T., Girgsdies, F., Kandemir, T., Thomas, N., Behrens, M., Schlögl, R.,
and Frei, E. (2016). Bridging the Time Gap: A Copper/Zinc Oxide/Aluminum Oxide
Catalyst for Methanol Synthesis Studied under Industrially Relevant Conditions
and Time Scales. Angewandte Chemie International Edition 55, 12708–12712.
(68) Palomino, R. M., Ramírez, P. J., Liu, Z., Hamlyn, R., Waluyo, I., Mahapatra, M.,
Orozco, I., Hunt, A., Simonovis, J. P., Senanayake, S. D., and Rodriguez, J. A.
(2018). Hydrogenation of CO2 on ZnO/Cu(100) and ZnO/Cu(111) Catalysts: Role
of Copper Structure and Metal–Oxide Interface in Methanol Synthesis. The Jour­
nal of Physical Chemistry B 122, 794–800.
(69) Zabilskiy, M., Sushkevich, V. L., Newton, M. A., and van Bokhoven, J. A. (2020).
Copper–Zinc Alloy­Free Synthesis of Methanol from Carbon Dioxide over Cu/ZnO/­
Faujasite. ACS Catalysis 10, 14240–14244.
(70) Laudenschleger, D., Ruland, H., and Muhler, M. (2020). Identifying the nature of
the active sites in methanol synthesis over Cu/ZnO/Al2O3 catalysts. Nature Com­
munications 11, 3898.
(71) Amann, P., Klötzer, B., Degerman, D., Köpfle, N., Götsch, T., Lömker, P., Rame­
shan, C., Ploner, K., Bikaljevic, D., Wang, H. Y., Soldemo, M., Shipilin, M., Good­
win, C. M., Gladh, J., Stenlid, J. H., Börner, M., Schlueter, C., and Nilsson, A.
(2022). The state of zinc in methanol synthesis over a Zn/ZnO/Cu(211) model
catalyst. Science 376, 603–608.
(72) Holse, C., Elkjær, C. F., Nierhoff, A., Sehested, J., Chorkendorff, I., Helveg, S., and
Nielsen, J. H. (2015). Dynamic Behavior of CuZn Nanoparticles under Oxidizing
and Reducing Conditions. The Journal of Physical Chemistry C 119, 2804–2812.
(73) Spencer, M. S. (1999). The role of zinc oxide in Cu/ZnO catalysts for methanol
synthesis and the water­gas shift reaction. Topics in Catalysis 8, 259–266.
(74) Schumann, J., Eichelbaum, M., Lunkenbein, T., Thomas, N., Álvarez Galván,
M. C., Schlögl, R., and Behrens, M. (2015). Promoting Strong Metal Support Inter­
action: Doping ZnO for Enhanced Activity of Cu/ZnO:M (M = Al, Ga, Mg) Catalysts.
ACS Catalysis 5, 3260–3270.

176
 References

(75) Pandit, L., Serrer, M.­A., Sara�i, E., Boubnov, A., and Grunwaldt, J.­D. (2022).
Versatile in situ/operando Setup for Studying Catalysts by X­Ray Absorption Spec­
troscopy under Demanding and Dynamic Reaction Conditions for Energy Storage
and Conversion. Chemistry ­ Methods 2, e202100078.
(76) Divins, N. J., Kordus, D., Timoshenko, J., Sinev, I., Zegkinoglou, I., Bergmann, A.,
Chee, S. W., Widrinna, S., Karslıoğlu, O., Mistry, H., Lopez Luna, M., Zhong, J. Q.,
Hoffman, A. S., Boubnov, A., Boscoboinik, J. A., Heggen, M., Dunin­Borkowski,
R. E., Bare, S. R., and Cuenya, B. R. (2021). Operando high­pressure investiga­
tion of size­controlled CuZn catalysts for the methanol synthesis reaction. Nature
Communications 12, 1435.
(77) Kandemir, T., Girgsdies, F., Hansen, T. C., Liss, K.­D., Kasatkin, I., Kunkes, E. L.,
Wowsnick, G., Jacobsen, N., Schlögl, R., and Behrens, M. (2013). In Situ Study
of Catalytic Processes: Neutron Diffraction of a Methanol Synthesis Catalyst at In­
dustrially Relevant Pressure. Angewandte Chemie International Edition 52, 5166–
5170.
(78) Divins, N. J., Angurell, I., Escudero, C., Pérez­Dieste, V., and Llorca, J. (2014).
Influence of the support on surface rearrangements of bimetallic nanoparticles in
real catalysts. Science 346, 620–623.
(79) Clausen, B. S., and Topsøe, H. (1991). In Situ high pressure, high temperature
XAFS studies of Cu­based catalysts during methanol synthesis. Catalysis Today
9, 189–196.
(80) Grandjean, D., Pelipenko, V., Batyrev, E. D., Van Den Heuvel, J. C., Khassin,
A. A., Yurieva, T. M., and Weckhuysen, B. M. (2011). Dynamic Cu/Zn interaction in
SiO2 supported methanol synthesis catalysts unraveled by in situ XAFS. Journal
of Physical Chemistry C 115, 20175–20191.
(81) Van Den Berg, M. W., Polarz, S., Tkachenko, O. P., Kähler, K., Muhler, M., and
Grünert, W. (2009). Dynamical changes in the Cu­ZnO x interaction observed in
a model methanol synthesis catalyst. Catalysis Letters 128, 49–56.
(82) Großmann, D., Klementiev, K., Sinev, I., and Grünert, W. (2017). Surface Alloy
or Metal–Cation Interaction­The State of Zn Promoting the Active Cu Sites in
Methanol Synthesis Catalysts. ChemCatChem 9, 365–372.
(83) Bansode, A., Guilera, G., Cuartero, V., Simonelli, L., Avila, M., and Urakawa, A.
(2014). Performance and characteristics of a high pressure, high temperature cap­
illary cell with facile construction for operando x­ray absorption spectroscopy. Re­
view of Scientific Instruments 85, 084105.
(84) Martin, O., Mondelli, C., Cervellino, A., Ferri, D., Curulla­Ferré, D., and Pérez­
Ramírez, J. (2016). Operando Synchrotron X­ray Powder Diffraction and Modulated­
Excitation Infrared Spectroscopy Elucidate the CO2 Promotion on a Commer­
cial Methanol Synthesis Catalyst. Angewandte Chemie International Edition 55,
11031–11036.
(85) Sharafutdinov, I. Investigations into low pressure methanol synthesis, Ph.D. The­
sis, Technical University of Denmark, 2013.
(86) Holzwarth, U., and Gibson, N. (2011). The Scherrer equation versus the ’Debye­
Scherrer equation’. Nature Nanotechnology 2011 6:9 6, 534–534.
(87) Degen, T., Sadki, M., Bron, E., König, U., and Nénert, G. (2014). The HighScore
suite. Powder Diffraction 29, S13–S18.
(88) FIZ Karlsruhe Inorganic crystal structure database ICSD.
(89) Christiansen, T. L., Cooper, S. R., and Jensen, K. M. (2020). There’s no place like
real­space: elucidating size­dependent atomic structure of nanomaterials using
pair distribution function analysis. Nanoscale Advances 2, 2234–2254.

177
References

(90) Chupas, P. J., Qiu, X., Hanson, J. C., Lee, P. L., Grey, C. P., and Billinge, S. J.
(2003). Rapid­acquisition pair distribution function (RA­PDF) analysis. Journal of
Applied Crystallography 36, 1342–1347.
(91) Datye, A. K., Hansen, P. L., and Helveg, S. In Handbook of Heterogeneous Catal­
ysis, 2008, pp 803–833.
(92) Bowker, M. (2019). Methanol Synthesis from CO2 Hydrogenation. ChemCatChem
11, 4238–4246.
(93) Fichtl, M. B., Schlereth, D., Jacobsen, N., Kasatkin, I., Schumann, J., Behrens, M.,
Schlögl, R., and Hinrichsen, O. (2015). Kinetics of deactivation on Cu/ZnO/Al2O3
methanol synthesis catalysts. Applied Catalysis A: General 502, 262–270.
(94) Sun, J. T., Metcalfe, I. S., and Sahibzada, M. (1999). Deactivation of Cu/ZnO/Al2O3
Methanol Synthesis Catalyst by Sintering. Industrial and Engineering Chemistry
Research 38, 3868–3872.
(95) Prieto, G., Shakeri, M., De Jong, K. P., and De Jongh, P. E. (2014). Quantita­
tive relationship between support porosity and the stability of pore­confined metal
nanoparticles studied on CuZnO/SiO2 methanol synthesis catalysts. ACS Nano
8, 2522–2531.
(96) Laursen, A. B., Sehested, J., Chorkendorff, I., and Vesborg, P. C. (2018). Avail­
ability of elements for heterogeneous catalysis: Predicting the industrial viability
of novel catalysts. Chinese Journal of Catalysis 39, 16–26.
(97) Saal, J. E., Kirklin, S., Aykol, M., Meredig, B., and Wolverton, C. (2013). Materials
design and discovery with high­throughput density functional theory: The open
quantum materials database (OQMD). JOM 65, 1501–1509.
(98) Kirklin, S., Saal, J. E., Meredig, B., Thompson, A., Doak, J. W., Aykol, M., Rühl,
S., and Wolverton, C. (2015). The Open Quantum Materials Database (OQMD):
assessing the accuracy of DFT formation energies. npj Computational Materials
2015 1:1 1, 1–15.
(99) Elnabawy, A. O., Schimmenti, R., Cao, A., and Nørskov, J. K. (2022). Why ZnO is
the Support for Cu in Methanol Synthesis? A Systematic Study of the Strong Metal
Support Interactions. ACS Sustainable Chemistry and Engineering 10, 1722–1730.
(100) Cao, A., Wang, Z., Li, H., Elnabawy, A. O., and Nørskov, J. K. (2021). New insights
on CO and CO2 hydrogenation for methanol synthesis: The key role of adsorbate­
adsorbate interactions on Cu and the highly active MgO­Cu interface. Journal of
Catalysis 400, 325–331.
(101) Martin, O., Martín, A. J., Mondelli, C., Mitchell, S., Segawa, T. F., Hauert, R.,
Drouilly, C., Curulla­Ferré, D., and Pérez­Ramírez, J. (2016). Indium Oxide as
a Superior Catalyst for Methanol Synthesis by CO2 Hydrogenation. Angewandte
Chemie International Edition 55, 6261–6265.
(102) Sun, K., Fan, Z., Ye, J., Yan, J., Ge, Q., Li, Y., He, W., Yang, W., and Liu, C. J.
(2015). Hydrogenation of CO2 to methanol over In2O3 catalyst. Journal of CO2
Utilization 12, 1–6.
(103) Bonivardi, A. L., Chiavassa, D. L., Querini, C. A., and Baltanás, M. A. (2000).
Enhancement of the catalytic performance to methanol synthesis from CO2/H2
by gallium addition to palladium/silica catalysts. Studies in Surface Science and
Catalysis 130, 3747–3752.
(104) Fiordaliso, E. M., Sharafutdinov, I., Carvalho, H. W. P., Grunwaldt, J.­D., Hansen,
T. W., Chorkendorff, I., Wagner, J. B., and Damsgaard, C. D. (2015). Intermetallic
GaPd2 Nanoparticles on SiO2 for Low­Pressure CO2 Hydrogenation to Methanol:
Catalytic Performance and In Situ Characterization. ACS Catalysis 5, 5827–5836.

178
 References

(105) García­Trenco, A., White, E. R., Regoutz, A., Payne, D. J., Shaffer, M. S., and
Williams, C. K. (2017). Pd2Ga­Based Colloids as Highly Active Catalysts for the
Hydrogenation of CO2 to Methanol. ACS Catalysis 7, 1186–1196.
(106) Fiordaliso, E. M., Sharafutdinov, I., Carvalho, H. W. P., Kehres, J., Grunwaldt,
J.­D., Chorkendorff, I., and Damsgaard, C. D. (2019). Evolution of intermetallic
GaPd2/SiO2 catalyst and optimization for methanol synthesis at ambient pres­
sure. Science and Technology of Advanced Materials 20, 521–531.
(107) Singh, J. A., Cao, A., Schumann, J., Wang, T., Nørskov, J. K., Abild­Pedersen, F.,
and Bent, S. F. (2018). Theoretical and Experimental Studies of CoGa Catalysts
for the Hydrogenation of CO2 to Methanol. Catalysis Letters 148, 3583–3591.
(108) Studt, F., Sharafutdinov, I., Abild­Pedersen, F., Elkjær, C. F., Hummelshøj, J. S.,
Dahl, S., Chorkendorff, I., and Nørskov, J. K. (2014). Discovery of a Ni­Ga catalyst
for carbon dioxide reduction to methanol. Nature Chemistry 6, 320.
(109) Sharafutdinov, I., Elkjær, C. F., Pereira de Carvalho, H. W., Gardini, D., Chiarello,
G. L., Damsgaard, C. D., Wagner, J. B., Grunwaldt, J.­D., Dahl, S., and Chork­
endorff, I. (2014). Intermetallic compounds of Ni and Ga as catalysts for the syn­
thesis of methanol. Journal of Catalysis 320, 77–88.
(110) Gallo, A., Snider, J. L., Sokaras, D., Nordlund, D., Kroll, T., Ogasawara, H., Ko­
varik, L., Duyar, M. S., and Jaramillo, T. F. (2020). Ni5Ga3 catalysts for CO2 re­
duction to methanol: Exploring the role of Ga surface oxidation/reduction on cat­
alytic activity. Applied Catalysis B: Environmental 267, 118369.
(111) Smitshuysen, T. E. L., Nielsen, M. R., Pruessmann, T., Zimina, A., Sheppard,
T. L., Grunwaldt, J.­D., Chorkendorff, I., and Damsgaard, C. D. (2020). Optimiz­
ing Ni−Fe−Ga alloys into Ni2FeGa for the Hydrogenation of CO2 into Methanol.
ChemCatChem 12, 1–10.
(112) Medina, J. C., Figueroa, M., Manrique, R., Rodríguez Pereira, J., Srinivasan, P. D.,
Bravo­Suárez, J. J., Baldovino Medrano, V. G., Jiménez, R., and Karelovic, A.
(2017). Catalytic consequences of Ga promotion on Cu for CO2 hydrogenation to
methanol. Catalysis Science & Technology 7, 3375–3387.
(113) Cai, M., Subramanian, V., Sushkevich, V. V., Ordomsky, V. V., and Khodakov,
A. Y. (2015). Effect of Sn additives on the CuZnAl–HZSM­5 hybrid catalysts for
the direct DME synthesis from syngas. Applied Catalysis A: General 502, 370–
379.
(114) Abrokwah, R. Y., Deshmane, V. G., and Kuila, D. (2016). Comparative perfor­
mance of M­MCM­41 (M: Cu, Co, Ni, Pd, Zn and Sn) catalysts for steam reforming
of methanol. Journal of Molecular Catalysis A: Chemical 425, 10–20.
(115) Deshmane, V. G., Owen, S. L., Abrokwah, R. Y., and Kuila, D. (2015). Meso­
porous nanocrystalline TiO2 supported metal (Cu, Co, Ni, Pd, Zn, and Sn) cata­
lysts: Effect of metal­support interactions on steam reforming of methanol. Journal
of Molecular Catalysis A: Chemical 408, 202–213.
(116) Li, J., Halldin Stenlid, J., Tang, M. T., Peng, H. J., and Abild­Pedersen, F. (2022).
Screening binary alloys for electrochemical CO2 reduction towards multi­carbon
products. Journal of Materials Chemistry A 10, 16171–16181.
(117) Vilella, I. M., de Miguel, S. R., and Scelza, O. A. (2008). Pt, PtSn and PtGe cat­
alysts supported on granular carbon for fine chemistry hydrogenation reactions.
Journal of Molecular Catalysis A: Chemical 284, 161–171.
(118) Chen, Y., Zhou, J., Zhang, L., Peng, J., Li, S., Yin, S., Yang, K., and Lin, Y. (2018).
Microwave­assisted and regular leaching of germanium from the germanium­rich
lignite ash. Green Processing and Synthesis 7, 538–545.
(119) Ross, J. R. (2012). Catalyst Preparation. Heterogeneous Catalysis, 65–96.

179
References

(120) Clausen, B. S., Steffensen, G., Fabius, B., Villadsen, J., Feidenhans’l, R., and
Topsøe, H. (1991). In situ cell for combined XRD and on­line catalysis tests: Stud­
ies of Cu­based water gas shift and methanol catalysts. Journal of Catalysis 132,
524–535.
(121) Taylor, J. R., An introduction to error analysis ­ The study of uncertainties in phys­
ical measurements, 2nd; University Science Books: 1996.
(122) Fujitani, T., Saito, M., Kanai, Y., Kakumoto, T., Watanabe, T., Nakamura, J., and
Uchijima, T. (1994). The role of metal oxides in promoting a copper catalyst for
methanol synthesis. Catalysis Letters 1994 25:3 25, 271–276.
(123) Saito, M., and Murata, K. (2004). Development of high performance Cu/ZnO­
based catalysts for methanol synthesis and the water­gas shift reaction. Catalysis
Surveys from Asia 8, 285–294.
(124) Burch, R., Golunski, S. E., and Spencer, M. S. (1990). The role of copper and zinc
oxide in methanol synthesis catalysts. Journal of the Chemical Society, Faraday
Transactions 86, 2683–2691.
(125) Thrane, J., Kuld, S., Nielsen, N. D., Jensen, A. D., Sehested, J., and Christensen,
J. M. (2020). Methanol­Assisted Autocatalysis in Catalytic Methanol Synthesis.
Angewandte Chemie International Edition 59, 18189–18193.
(126) Lam, E., Corral­Pérez, J. J., Larmier, K., Noh, G., Wolf, P., Comas­Vives, A.,
Urakawa, A., and Copéret, C. (2019). CO2 Hydrogenation on Cu/Al2O3: Role
of the Metal/Support Interface in Driving Activity and Selectivity of a Bifunctional
Catalyst. Angewandte Chemie International Edition 58, 13989–13996.
(127) Chatterjee, R., Kuld, S., van den Berg, R., Chen, A., Shen, W., Christensen, J. M.,
Jensen, A. D., and Sehested, J. (2019). Mapping Support Interactions in Copper
Catalysts. Topics in Catalysis, DOI: 10.1007/s11244-019-01150-9.
(128) Grant, A. W., Ranney, J. T., Campbell, C. T., Evans, T., and Thornton, G. (2000).
The influence of chlorine on the dispersion of Cu particles on Cu/ZnO(0001) model
catalysts. Catalysis Letters 2000 65:4 65, 159–168.
(129) El’tsov, K. N., Zueva, G. Y., Klimov, A. N., Martynov, V. V., and Prokhorov, A. M.
(1991). Reversible coverage­dependent Cu + Clads → CuCl transition on Cu(111)/Cl2
surface. Surface Science 251­252, 753–758.
(130) Zhao, J., Yang, L., McLeod, J. A., and Liu, L. (2015). Reduced GeO2 Nanoparti­
cles: Electronic Structure of a Nominal GeOx Complex and Its Stability under H2
Annealing. Scientific Reports 2015 5:1 5, 1–10.
(131) Wang, J., Jin, S., Leinenbach, C., and Jacot, A. (2010). Thermodynamic assess­
ment of the Cu–Ge binary system. Journal of Alloys and Compounds 504, 159–
165.
(132) Rodrigues, T. S., Zhao, M., Yang, T. H., Gilroy, K. D., da Silva, A. G., Camargo,
P. H., and Xia, Y. (2018). Synthesis of Colloidal Metal Nanocrystals: A Compre­
hensive Review on the Reductants. Chemistry – A European Journal 24, 16944–
16963.
(133) Che, M., and Bennett, C. O. (1989). The Influence of Particle Size on the Catalytic
Properties of Supported Metals. Advances in Catalysis 36, 55–172.
(134) Barberis, L., Hakimioun, A. H., Plessow, P. N., Visser, N. L., Stewart, J. A., Van­
degehuchte, B. D., Studt, F., and Jongh, P. E. d. (2022). Competition between
reverse water gas shift reaction and methanol synthesis from CO2: influence of
copper particle size. Nanoscale, DOI: 10.1039/D2NR02612K.
(135) Yoshihara, J., Parker, S. C., Schafer, A., and Campbell, C. T. (1995). Methanol
synthesis and reverse water­gas shift kinetics over clean polycrystalline copper.
Catalysis Letters 31, 313–324.

180
 References

(136) Nørskov, J. K., Abild­Pedersen, F., Studt, F., and Bligaard, T. (2011). Density
functional theory in surface chemistry and catalysis. Proceedings of the National
Academy of Sciences 108, 937.
(137) Andersson, M. P., Abild­Pedersen, F., Remediakis, I. N., Bligaard, T., Jones, G.,
Engbæk, J., Lytken, O., Horch, S., Nielsen, J. H., Sehested, J., Rostrup­Nielsen,
J. R., Nørskov, J. K., and Chorkendorff, I. (2008). Structure sensitivity of the metha­
nation reaction: H2­induced CO dissociation on nickel surfaces. Journal of Catal­
ysis 255, 6–19.
(138) Medford, A. J., Lausche, A. C., Abild­Pedersen, F., Temel, B., Schjødt, N. C.,
Nørskov, J. K., and Studt, F. (2014). Activity and Selectivity Trends in Synthesis
Gas Conversion to Higher Alcohols. Topics in Catalysis 57, 135–142.
(139) Lausche, A. C., Medford, A. J., Khan, T. S., Xu, Y., Bligaard, T., Abild­Pedersen,
F., Nørskov, J. K., and Studt, F. (2013). On the effect of coverage­dependent
adsorbate–adsorbate interactions for CO methanation on transition metal sur­
faces. Journal of Catalysis 307, 275–282.
(140) Van Den Berg, R., Parmentier, T. E., Elkjær, C. F., Gommes, C. J., Sehested, J.,
Helveg, S., De Jongh, P. E., and De Jong, K. P. (2015). Support Functionaliza­
tion To Retard Ostwald Ripening in Copper Methanol Synthesis Catalysts. ACS
Catalysis 5, 4439–4448.
(141) Karelovic, A., Galdames, G., Medina, J. C., Yévenes, C., Barra, Y., and Jiménez,
R. (2019). Mechanism and structure sensitivity of methanol synthesis from CO2
over SiO2­supported Cu nanoparticles. Journal of Catalysis 369, 415–426.
(142) Chen, H. W., White, J. M., and Ekerdt, J. G. (1986). Electronic effect of supports
on copper catalysts. Journal of Catalysis 99, 293–303.
(143) Cuenya, B. R. (2010). Synthesis and catalytic properties of metal nanoparticles:
Size, shape, support, composition, and oxidation state effects. Thin Solid Films
518, 3127–3150.
(144) Crampton, A. S., Rötzer, M. D., Ridge, C. J., Schweinberger, F. F., Heiz, U., Yoon,
B., and Landman, U. (2016). Structure sensitivity in the nonscalable regime ex­
plored via catalysed ethylene hydrogenation on supported platinum nanoclusters.
Nature Communications 2016 7:1 7, 1–12.
(145) Stakheev, A. Y., and Kustov, L. M. (1999). Effects of the support on the morphol­
ogy and electronic properties of supported metal clusters: modern concepts and
progress in 1990s. Applied Catalysis A: General 188, 3–35.
(146) Van Hardeveld, R., and Hartog, F. (1969). The statistics of surface atoms and
surface sites on metal crystals. Surface Science 15, 189–230.
(147) Stokes, A. R., and Wilson, A. J. (1942). A method of calculating the integral breadths
of Debye­Scherrer lines. Mathematical Proceedings of the Cambridge Philosoph­
ical Society 38, 313–322.
(148) Bergeret, G., and Gallezot, P. In Handbook of Heterogeneous Catalysis, Ertl, G.,
Knözinger, H., Schüth, F., and Weitkamp, J., Eds., 2008, pp 738–765.
(149) Davey, W. P. (1925). Precision Measurements of the Lattice Constants of Twelve
Common Metals. Physical Review 25, 753–761.
(150) Jain, A., Ong, S. P., Hautier, G., Chen, W., Richards, W. D., Dacek, S., Cholia,
S., Gunter, D., Skinner, D., Ceder, G., and Persson, K. A. (2013). Commentary:
The Materials Project: A materials genome approach to accelerating materials
innovation. APL Materials 1, 011002.
(151) Tran, R., Xu, Z., Radhakrishnan, B., Winston, D., Sun, W., Persson, K. A., and
Ong, S. P. (2016). Surface energies of elemental crystals. Scientific Data 2016
3:1 3, 1–13.

181
References

(152) Borodziński, A., and Bonarowska, M. (1997). Relation between Crystallite Size
and Dispersion on Supported Metal Catalysts. Langmuir 13, 5613–5620.
(153) Li, D., Wang, C., Tripkovic, D., Sun, S., Markovic, N. M., and Stamenkovic, V. R.
(2012). Surfactant removal for colloidal nanoparticles from solution synthesis: The
effect on catalytic performance. ACS Catalysis 2, 1358–1362.
(154) Quinson, J., Mathiesen, J. K., Schröder, J., Dworzak, A., Bizzotto, F., Zana, A.,
Simonsen, S. B., Theil Kuhn, L., Oezaslan, M., Jensen, K. M., and Arenz, M.
(2020). Teaching old precursors new tricks: Fast room temperature synthesis of
surfactant­free colloidal platinum nanoparticles. Journal of colloid and interface
science 577, 319–328.
(155) Wang, Y., Ren, J., Deng, K., Gui, L., and Tang, Y. (2000). Preparation of Tractable
Platinum, Rhodium, and Ruthenium Nanoclusters with Small Particle Size in Or­
ganic Media. Chemistry of Materials 12, 1622–1627.
(156) Ghosh Chaudhuri, R., and Paria, S. (2012). Core/shell nanoparticles: Classes,
properties, synthesis mechanisms, characterization, and applications. Chemical
Reviews 112, 2373–2433.
(157) Van Beek, W., Safonova, O. V., Wiker, G., and Emerich, H. (2011). SNBL, a dedi­
cated beamline for combined in situ X­ray diffraction, X­ray absorption and Raman
scattering experiments. https://doi.org/10.1080/01411594.2010.549944 84, 726–
732.
(158) Graciani, J., Mudiyanselage, K., Xu, F., Baber, A. E., Evans, J., Senanayake, S. D.,
Stacchiola, D. J., Liu, P., Hrbek, J., Sanz, J. F., and Rodriguez, J. A. (2014). Highly
active copper­ceria and copper­ceria­titania catalysts for methanol synthesis from
CO2. Science 345, 546 LP –550.
(159) Arena, F., Mezzatesta, G., Zafarana, G., Trunfio, G., Frusteri, F., and Spadaro, L.
(2013). How oxide carriers control the catalytic functionality of the Cu–ZnO system
in the hydrogenation of CO2 to methanol. Catalysis Today 210, 39–46.
(160) Robinson, W. R., and Mol, J. C. (1991). Support effects in methanol synthesis over
copper­containing catalysts. Applied Catalysis 76, 117–129.
(161) Zabilskiy, M., Sushkevich, V. L., Palagin, D., Newton, M. A., Krumeich, F., and
van Bokhoven, J. A. (2020). The unique interplay between copper and zinc during
catalytic carbon dioxide hydrogenation to methanol. Nature Communications 11,
2409.
(162) Prieto, G., Zečević, J., Friedrich, H., De Jong, K. P., and De Jongh, P. E. (2012).
Towards stable catalysts by controlling collective properties of supported metal
nanoparticles. Nature Materials 2012 12:1 12, 34–39.
(163) Pandit, L., Boubnov, A., Behrendt, G., Mockenhaupt, B., Chowdhury, C., Jelic,
J., Hansen, A.­L., Saraçi, E., Ras, E.­J., Behrens, M., Studt, F., and Grunwaldt,
J.­D. (2021). Unravelling the Zn­Cu Interaction during Activation of a Zn­promoted
Cu/MgO Model Methanol Catalyst. ChemCatChem, DOI: 10.1002/CCTC.202100692.
(164) Nakamura, I., Fujitani, T., Uchijima, T., and Nakamura, J. (1998). The synthesis of
methanol and the reverse water­gas shift reaction over Zn­deposited Cu(100) and
Cu(110) surfaces: comparison with Zn/Cu(111). Surface Science 400, 387–400.
(165) Dalebout, R., Visser, N. L., Pompe, C. E. L., de Jong, K. P., and de Jongh, P. E.
(2020). Interplay between carbon dioxide enrichment and zinc oxide promotion of
copper catalysts in methanol synthesis. Journal of Catalysis 392, 150–158.
(166) Wang, J., Lu, S.­m., Li, J., and Li, C. (2015). A remarkable difference in CO2
hydrogenation to methanol on Pd nanoparticles supported inside and outside of
carbon nanotubes. Chemical Communications 51, 17615–17618.

182
 References

(167) Díez­Ramírez, J., Sánchez, P., Rodríguez­Gómez, A., Valverde, J. L., and Do­
rado, F. (2016). Carbon Nanofiber­Based Palladium/Zinc Catalysts for the Hydro­
genation of Carbon Dioxide to Methanol at Atmospheric Pressure. Industrial &
Engineering Chemistry Research 55, 3556–3567.
(168) Liang, X.­L., Xie, J.­R., and Liu, Z.­M. (2015). A Novel Pd­decorated Carbon Nanotubes­
promoted Pd­ZnO Catalyst for CO2 Hydrogenation to Methanol. Catalysis Letters
145, 1138–1147.
(169) Zhang, Q., Zuo, Y.­Z., Han, M.­H., Wang, J.­F., Jin, Y., and Wei, F. (2010). Long
carbon nanotubes intercrossed Cu/Zn/Al/Zr catalyst for CO/CO2 hydrogenation to
methanol/dimethyl ether. Catalysis Today 150, 55–60.
(170) Lam, E., and Luong, J. H. T. (2014). Carbon Materials as Catalyst Supports and
Catalysts in the Transformation of Biomass to Fuels and Chemicals. ACS Cataly­
sis 4, 3393–3410.
(171) Schneider, C. A., Rasband, W. S., and Eliceiri, K. W. (2012). NIH Image to ImageJ:
25 years of image analysis. Nature Methods 2012 9:7 9, 671–675.
(172) Schindelin, J., Arganda­Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch,
T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., Tinevez, J. Y., White, D. J.,
Hartenstein, V., Eliceiri, K., Tomancak, P., and Cardona, A. (2012). Fiji: an open­
source platform for biological­image analysis. Nature Methods 2012 9:7 9, 676–
682.
(173) Hammersley, A. P., Svensson, S. O., Hanfland, M., Fitch, A. N., and Häusermann,
D. (2006). Two­dimensional detector software: From real detector to idealised im­
age or two­theta scan. https://doi.org/10.1080/08957959608201408 14, 235–248.
(174) Hammersley, A. P. (2016). FIT2D: a multi­purpose data reduction, analysis and
visualization program. Journal of Applied Crystallography 49, 646–652.
(175) Yang, X., Juhas, P., Farrow, C. L., and Billinge, S. J. L. (2014). xPDFsuite: an
end­to­end software solution for high throughput pair distribution function trans­
formation, visualization and analysis. DOI: 10.48550/arxiv.1402.3163.
(176) Farrow, C. L., Juhas, P., Liu, J. W., Bryndin, D., Boin, E. S., Bloch, J., Proffen, T.,
and Billinge, S. J. (2007). PDFfit2 and PDFgui: computer programs for studying
nanostructure in crystals. Journal of Physics: Condensed Matter 19, 335219.
(177) Brunauer, S., Emmett, P. H., and Teller, E. (1938). Adsorption of Gases in Multi­
molecular Layers. Journal of the American Chemical Society 60, 309–319.
(178) Sahibzada, M., Metcalfe, I. S., and Chadwick, D. (1998). Methanol Synthesis from
CO/CO2/H2over Cu/ZnO/Al2O3at Differential and Finite Conversions. Journal of
Catalysis 174, 111–118.
(179) Klier, K., Chatikavanij, V., Herman, R. G., and Simmons, G. W. (1982). Catalytic
synthesis of methanol from COH2: IV. The effects of carbon dioxide. Journal of
Catalysis 74, 343–360.
(180) Lee, J. S., Lee, K. H., Lee, S. Y., and Kim, Y. G. (1993). A Comparative Study
of Methanol Synthesis from CO2/H2 and CO/H2 over a Cu/ZnO/Al2O3 Catalyst.
Journal of Catalysis 144, 414–424.
(181) Koitaya, T., Yamamoto, S., Shiozawa, Y., Yoshikura, Y., Hasegawa, M., Tang, J.,
Takeuchi, K., Mukai, K., Yoshimoto, S., Matsuda, I., and Yoshinobu, J. (2019).
CO2 Activation and Reaction on Zn­Deposited Cu Surfaces Studied by Ambient­
Pressure X­ray Photoelectron Spectroscopy. ACS Catalysis 9, 4539–4550.
(182) Tisseraud, C., Comminges, C., Belin, T., Ahouari, H., Soualah, A., Pouilloux, Y.,
and Le Valant, A. (2015). The Cu–ZnO synergy in methanol synthesis from CO2,
Part 2: Origin of the methanol and CO selectivities explained by experimental stud­

183
References

ies and a sphere contact quantification model in randomly packed binary mixtures
on Cu–ZnO coprecipitate catalysts. Journal of Catalysis 330, 533–544.
(183) Le Valant, A., Comminges, C., Tisseraud, C., Canaff, C., Pinard, L., and Pouilloux,
Y. (2015). The Cu–ZnO synergy in methanol synthesis from CO2, Part 1: Origin of
active site explained by experimental studies and a sphere contact quantification
model on Cu+ZnO mechanical mixtures. Journal of Catalysis 324, 41–49.
(184) Herman, R. G., Klier, K., Simmons, G. W., Finn, B. P., Bulko, J. B., and Kobylin­
ski, T. P. (1979). Catalytic synthesis of methanol from COH2: I. Phase composi­
tion, electronic properties, and activities of the Cu/ZnO/M2O3 catalysts. Journal
of Catalysis 56, 407–429.
(185) Günter, M. M., Ressler, T., Bems, B., Büscher, C., Genger, T., Hinrichsen, O.,
Muhler, M., and Schlögl, R. (2001). Implication of the microstructure of binary
Cu/ZnO catalysts for their catalytic activity in methanol synthesis. Catalysis Letters
71, 37–44.
(186) Ingham, B. (2015). X­ray scattering characterisation of nanoparticles. Crystallog­
raphy Reviews 21, 229–303.
(187) Sehested, J., Gelten, J. A., and Helveg, S. (2006). Sintering of nickel catalysts:
Effects of time, atmosphere, temperature, nickel­carrier interactions, and dopants.
Applied Catalysis A: General 309, 237–246.
(188) Smith, W. L. (1972). Crystallite sizes and surface areas of catalysts. Journal of
Applied Crystallography 5, 127–130.
(189) Chapman, K. W., Chupas, P. J., and Kepert, C. J. (2005). Selective recovery
of dynamic guest structure in a nanoporous Prussian blue through in situ X­ray
diffraction: A differential pair distribution function analysis. Journal of the Ameri­
can Chemical Society 127, 11232–11233.
(190) Petkov, V., Gateshki, M., Choi, J., Gillan, E. G., and Ren, Y. (2005). Structure of
nanocrystalline GaN from X­ray diffraction, Rietveld and atomic pair distribution
function analyses. Journal of Materials Chemistry 15, 4654–4659.
(191) Petkov, V., Peng, Y., Williams, G., Huang, B., Tomalia, D., and Ren, Y. (2005).
Structure of gold nanoparticles suspended in water studied by x­ray diffraction
and computer simulations. Physical Review B ­ Condensed Matter and Materials
Physics 72, 195402.
(192) Bedford, N., Dablemont, C., Viau, G., Chupas, P., and Petkov, V. (2007). 3­D
Structure of Nanosized Catalysts by High­Energy X­ray Diffraction and Reverse
Monte Carlo Simulations: Study of Ru. Journal of Physical Chemistry C 111,
18214–18219.
(193) Nielsen, N. D., Jensen, A. D., and Christensen, J. M. (2020). Quantification of For­
mate and Oxygen Coverages on Cu Under Industrial Methanol Synthesis Condi­
tions. Catalysis Letters 150, 2447–2456.
(194) Munnik, P., de Jongh, P. E., and de Jong, K. P. (2015). Recent Developments in
the Synthesis of Supported Catalysts. Chemical Reviews 115, 6687–6718.
(195) Munnik, P., Wolters, M., Gabrielsson, A., Pollington, S. D., Headdock, G., Bitter,
J. H., De Jongh, P. E., and De Jong, K. P. (2011). Copper nitrate redispersion
to arrive at highly active silica­supported copper catalysts. Journal of Physical
Chemistry C 115, 14698–14706.
(196) Beerthuis, R., de Rijk, J. W., Deeley, J. M., Sunley, G. J., de Jong, K. P., and
de Jongh, P. E. (2020). Particle size effects in copper­catalyzed hydrogenation of
ethyl acetate. Journal of Catalysis 388, 30–37.

184
 References

(197) Munnik, P., Krans, N. A., De Jongh, P. E., and De Jong, K. P. (2014). Effects
of drying conditions on the synthesis of Co/SiO2 and Co/Al2O3 fischer­tropsch
catalysts. ACS Catalysis 4, 3219–3226.
(198) Munnik, P., De Jongh, P. E., and De Jong, K. P. (2014). Control and impact of
the nanoscale distribution of supported cobalt particles used in fischer­tropsch
catalysis. Journal of the American Chemical Society 136, 7333–7340.
(199) Wolters, M., Munnik, P., Bitter, J. H., De Jongh, P. E., and De Jong, K. P. (2011).
How NO affects nickel and cobalt nitrates at low temperatures to arrive at highly
dispersed silica­supported nickel and cobalt catalysts. Journal of Physical Chem­
istry C 115, 3332–3339.
(200) Carmo, M., Linardi, M., and Poco, J. G. R. (2009). Characterization of nitric acid
functionalized carbon black and its evaluation as electrocatalyst support for direct
methanol fuel cell applications. Applied Catalysis A: General 355, 132–138.
(201) Shinde, V. M., Skupien, E., and Makkee, M. (2015). Synthesis of highly dispersed
Pd nanoparticles supported on multi­walled carbon nanotubes and their excellent
catalytic performance for oxidation of benzyl alcohol. Catalysis Science & Tech­
nology 5, 4144–4153.
(202) Clausen, B. S., Lengeier, B., and Rasmussen, B. S. (1985). X­ray absorption
spectroscopy study of Cu­based methanol catalysts. 1. Calcined state. Journal
of Physical Chemistry 89, 2319–2324.
(203) Frost, J. C. (1988). Junction effect interactions in methanol synthesis catalysts.
Nature 1988 334:6183 334, 577–580.
(204) Muhler, M., Nielsen, L. P., Törnqvist, E., Clausen, B. S., and Topsøe, H. (1992).
Temperature­programmed desorption of H2 as a tool to determine metal surface
areas of Cu catalysts. Catalysis Letters 14, 241–249.
(205) Baltes, C., Vukojević, S., and Schüth, F. (2008). Correlations between synthesis,
precursor, and catalyst structure and activity of a large set of CuO/ZnO/Al2O3
catalysts for methanol synthesis. Journal of Catalysis 258, 334–344.
(206) Kurtz, M., Bauer, N., Büscher, C., Wilmer, H., Hinrichsen, O., Becker, R., Rabe, S.,
Merz, K., Driess, M., Fischer, R. A., and Muhler, M. (2004). New synthetic routes
to more active Cu/ZnO catalysts used for methanol synthesis. Catalysis Letters
92, 49–52.
(207) Chinchen, G. C., Hay, C. M., Vandervell, H. D., and Waugh, K. C. (1987). The
measurement of copper surface areas by reactive frontal chromatography. Journal
of Catalysis 103, 79–86.
(208) Fichtl, M. B., Schumann, J., Kasatkin, I., Jacobsen, N., Behrens, M., Schlögl,
R., Muhler, M., and Hinrichsen, O. (2014). Counting of Oxygen Defects versus
Metal Surface Sites in Methanol Synthesis Catalysts by Different Probe Molecules.
Angewandte Chemie International Edition 53, 7043–7047.
(209) Nielsen, N. D., Smitshuysen, T. E. L., Damsgaard, C. D., Jensen, A. D., and Chris­
tensen, J. M. (2021). Characterization of oxide­supported Cu by infrared measure­
ments on adsorbed CO. Surface Science 703, 121725.
(210) Nielsen, N. D., Jensen, A. D., and Christensen, J. M. (2020). The roles of CO
and CO2 in high pressure methanol synthesis over Cu­based catalysts. Journal
of Catalysis, DOI: https://doi.org/10.1016/j.jcat.2020.11.035.
(211) Johansson, M., Hoffmann Jørgensen, J., and Chorkendorff, I. (2004). Combined
high­pressure cell–ultrahigh vacuum system for fast testing of model metal alloy
catalysts using scanning mass spectrometry. Review of Scientific Instruments 75,
2082–2093.

185
References

(212) Briggs, D., and Seah, M. P., Practical Surface Analysis: Auger and X­ray Photo­
electron Spectroscopy; John Wiley and Son: Chichester, 1990; Vol. 1.
(213) Schofield, K. (2008). The enigmatic mechanism of the flame ionization detector:
Its overlooked implications for fossil fuel combustion modeling. Progress in Energy
and Combustion Science 34, 330–350.

186