Insight in the Outside of catalysts with LEISInsight in the Outside of Catalysts with LEIS...

Insight in the outside of catalysts with LEIS

Citation for published version (APA):Jansen, W. P. A. (2002). Insight in the outside of catalysts with LEIS. Technische Universiteit Eindhoven.https://doi.org/10.6100/IR559530

DOI:10.6100/IR559530

Document status and date:Published: 01/01/2002

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand 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 the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 23. Jul. 2021

https://doi.org/10.6100/IR559530

https://doi.org/10.6100/IR559530

https://research.tue.nl/en/publications/00cc5d15-8093-454b-942e-0ba310fa1c2c

Insight in the Outside of Catalysts with LEIS

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan deTechnische Universiteit Eindhoven, op gezag van deRector Magnificus, prof.dr. R.A. van Santen, vooreen commissie aangewezen door het College voor

Promoties in het openbaar te verdedigen opdonderdag 24 oktober 2002 om 16.00 uur

door

Wilhelmus Paulus Anna Jansen

geboren te ‘s-Hertogenbosch

Dit proefschrift is goedgekeurd door de promotoren:

prof.dr. H.H. Brongersmaenprof.dr. J.W.M. Niemantsverdriet

Copromotor:dr. A.W. Denier van der Gon

Printed at the Universiteitsdrukkerij, Eindhoven University of Technology

CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN

Jansen, W.P.A.

Insight in the Outside of Catalysts with LEIS / W.P.A. Jansen. – Eindhoven:Technische Universiteit Eindhoven, 2002. – Proefschrift.ISBN 90-386-1535-3NUR 913, 916Trefwoorden: oppervlaktefysica / lage-energie ionen verstrooiing / katalyseSubject headings: surface physics / low-energy ion scattering / catalysis

A catalyst is like a mountain guide who takes parties over a pass to the next valley. He guides one party across, then -unchanged-

returns to pick up another group.

Dr. Mills, National Geographic 164, 693 (1983).

The work described in this thesis has been carried out at the Schuit Institute of Catalysis,Department of Applied Physics, Eindhoven University of Technology, The Netherlands. Theinvestigations were supported by the Netherlands Organization for Scientific Research(NWO).

Contents

Introduction 7

1. The influence of compaction and surface roughness on low-energy ionscattering signals 13

2a. Noble metal segregation and cluster size of Pt/Rh/CeO2/γ-Al2O3 automotivethree-way catalysts studied with LEIS 35

2b. Coke deposition on automotive three-way catalysts studied with LEIS 51

3. New insights into the nature of the active phase of VPO catalysts-a quantitative static LEIS study- 61

4. Dynamic behavior of the surface structure of Cu/ZnO/SiO2 catalysts 77

5. A differentially pumped pressure cell for in situ LEIS analysis of catalystsduring reactions 93

6. Surface coverages during CO oxidation over Pt(110) -an in situ LEIS study 109

Summary 121

Samenvatting 125

Publications 128

Acknowledgement / Dankwoord 131

Curriculum vitae 132

7

General introduction

Significance of catalysisThere is no life without the miracles of catalytic reactions in plants, animals, and human

beings. Biocatalysts (enzymes) selectively accelerate over 95% of the processes in livingorganisms and enable reactions to take place within seconds that would otherwise takecenturies [1]. Not only nature applies catalysis, it is also the cornerstone of the chemicalindustry. Approximately 90% of all our chemicals and materials are produced using catalysisat one stage or another [2]! Human application of catalysis started very early on. Forexample, without knowing so, ancient Sumerians applied catalysis many millennia ago toproduce their beer. Only in the 19th century Davy, Faraday and Berzelius, among others,started the systematic study of catalysis. However, even today, many catalytic processes arefound in a trial and error approach. Rational design of catalysts -meaning calculationfollowed by synthesis of the desired catalyst- is not possible yet [3]. To enable rationalcatalyst design in the future we should gain better understanding of catalysis. Moreover, suchknowledge will enable improvement of currently applied catalytic processes.

Catalysis: a surface processLarge-scale petrochemical and commodity chemical industries rely mostly on so-called

heterogeneous catalysis. In this type of catalysis different phases for the catalyst (typicallysolid) and the reactants (liquids or gases) are applied. A heterogeneous catalytic cycle can bedescribed as follows: reactants adsorb on the catalyst surface where they are activated andreact and the cycle is completed when the formed products desorb from the catalyst surface,leaving the (unchanged) catalyst behind for a new catalytic cycle. Hence, all catalytic actiontakes place on the catalyst surface, making knowledge of this surface a key to a betterunderstanding of catalysis.

As Pauli already stated: “Bulk was made by God and the surface by the Devil”. Thecomposition of a surface is generally very different from that of the bulk. Apart fromprocesses such as evaporation and adsorption from the surrounding atmosphere, the surface isoften affected by surface segregation, i.e. enrichment or depletion at the surface of certainatoms or compounds from the bulk. As an example, sub ppm bulk concentrations of Na havebeen shown to dominate the oxygen terminated ZnO surface for annealing temperatures of500 K – 700 K [4,5]. To probe surface phenomena apart from the bulk many surface sensitivetechniques, such as low-energy electron diffraction (LEED), X-ray photoelectronspectroscopy (XPS) and low-energy ion scattering (LEIS) have been developed, turningcatalysis almost into “the noble art of characterization” [6]. The rationale for the existence ofthe presently used multitude of characterizing techniques is that they all have their strengthsand weaknesses, making a combination of them often the winning choice [7]. For instance,different combinations of XPS, LEED, high-resolution electron microscopy (HREM),transmission electron microscopy (TEM), X-ray diffraction (XRD), inductively coupledplasma - atomic adsorption spectroscopy (ICP-AAS) and LEIS have been employed for the


8

studies presented in this thesis. The emphasis in this thesis, however, will be on LEIS atechnique that offers unique possibilities in catalysis research.

LEIS in catalysisIn LEIS the sample is bombarded with noble gas ions of a few keV. The energy of a

noble gas ion that is backscattered by a surface atom is determined by the laws ofconservation of momentum and energy. The energy spectrum of the scattered ions isequivalent to a mass spectrum of the surface atoms (see Figure 1). A detector that is designedto selectively detect ions in combination with the high neutralization probability of the noblegas ions assures that the contribution of deeper layers is negligible. Hence, LEIS selectivelyprobes the outermost atomic layer thus exactly where catalysis takes place. The intensity of apeak is directly proportional to the surface coverage of the corresponding element andtherefore allows for a quantitative analysis of real-life industrial catalysts; in contrast to mostother surface science techniques there is no so-called structure gap.

Figure 1: The inset shows a schematic of the LEIS analysis method: monoenergetic noble gas ions(Ei) are directed onto a sample and scatter back with a final energy (Ef) which depends on the mass ofthe surface atoms of the sample. Energy spectra of the backscattered ions can be interpreted as massspectra of the outermost atomic layer of the sample. The example shows a spectrum of vanadiumphosphorous oxide (VPO) that reveals the presence of V, P, and O in the outermost atomic layer ofVPO.

The LEIS measurements presented in this thesis have been performed on the ERISS andCalipso LEIS set-ups (see Figure 2). These home-built instruments apply highly sensitiveanalyzers to allow static LEIS (i.e. LEIS without damage) [8] and effective neutralizers toprevent charging of rough insulating surfaces that catalysts often have [9]. Moreover, this

0

15

30

45

700 1300 1900 2500Energy [eV]

LEIS

sig

nal [

cnts

./nC

]

O

PV

He + , Ne +

θ E i

E f


9

study benefited from the methodology that has been developed to enable quantitative LEIS[8,10,11]. To learn more about the possibilities with LEIS in catalysis research the reader isreferred to the following reviews [12,13] and the remainder of this thesis.

Figure 2: The Energy Resolved Ion Spectrometer or ERISS LEIS set-up that has been used for mostexperiments described in this thesis.

Scope of this thesisThe aim of the studies described in this thesis is to determine surface structural properties

of different real-life heterogeneous catalysts and to relate these to their catalytic performance.Since this required new instrumentation and new methodology, some studies deal with thedevelopment of the LEIS analysis technique as well. For instance, Chapter 1 presents a studyon the influence of surface roughness and compaction on LEIS. Such knowledge is importantsince heterogeneous catalysts are often supported, powdered catalysts, with extremely highspecific surface areas (super rough surfaces up to 1000 m2/g) that are routinely compacted tofacilitate (LEIS) analysis.

An example of a widely used supported, powdered catalyst is Pt/Rh/CeO2/γ-Al2O3, whichis applied in automotive three-way catalysts to diminish the emission of harmful species incar exhaust gas. To enable modeling and optimization of this catalyst the noble metaldispersion, i.e. the number of noble metal surface atoms divided by the total number of noblemetal atoms, has to be known. However, conventional techniques like CO chemisorption,HREM or TEM produce ambiguous results for ceria supported catalysts. Therefore, a newmethod has been developed to enable determination of dispersion based on LEIS signals(Chapter 2). Besides dispersion, other parameters such as segregation, and the influence of(de)activation and regeneration are studied in relation to the catalytic performance of thiscatalyst as well. It was for example not clear whether the Pt and Rh were mixed or separately


10

present at the surface of the automotive three-way catalyst. Nevertheless such knowledge isessential to understand the catalytic performance and it can lead the way for optimization ofthe catalyst composition. As will be shown LEIS can be used to solve this question (Chapter2). To investigate the formation of carbonaceous deposits on this catalyst the surfacecomposition of fresh and used Pt/Rh/CeO2/γ-Al2O3 are compared. Knowledge of thesedeposits is instrumental to enable kinetic modeling of the catalytic process, and mayeventually lead to improved behavior of the automotive exhaust converter via modelpredictive control (Chapter 2).

Another catalyst showing ample room for improvement is the vanadium phosphorusoxide (VPO) catalyst, a catalyst that is commercially used in the selective oxidation ofn-butane to maleic anhydride, a process with a selectivity of only 58%. In industry it isstandard practice to add an overstoichiometric amount of phosphate during preparation of thiscatalyst, however, nobody knows the fate of this phosphate [14]. To investigate this,quantitative LEIS is used in combination with XRD to determine the surface composition ofthe VPO catalyst and to shed light on the fate of the overstoichiometric amount of phosphate.Moreover, the influence of different treatments is studied to gain insight in patentedactivation and regeneration procedures of the VPO catalyst (Chapter 3).

Treatments are also very important in case of Cu/ZnO based catalysts that are used inmethanol synthesis. The surface composition of Cu/ZnO based catalysts is dynamic anddepends heavily on (pre)treatments. Improvement of the resolution of our LEIS detector incombination with isotopically enriched (supported) 63Cu/68ZnO catalysts enabled us to studythe surface composition of this catalyst upon exposure to various reducing conditions(Chapter 4). In this case, not only information on the surface composition, but also on itsoxidation state is relevant, since both Cu0 and Cu1+ have been proposed to be the active Cuspecies in the Cu/ZnO based catalysts. To address this issue a novel method is presented thatenables selective determination of the oxidation state of metals in the outermost atomic layer.

As already indicated above, the gaseous environment used during e.g. pretreatment orcatalysis may well influence the surface composition and oxidation state of a catalyst.Nevertheless, only a very limited number of LEIS studies have been carried out in situ[15-20]. Like most other surface science techniques, LEIS has almost only been used in ultra-high vacuum (UHV). Hence, there is a large pressure gap between surface science andindustrial catalysis that applies pressures P ≥ 1 atm. And although examination before andafter reaction has greatly contributed to our knowledge, ex situ analysis in UHV remains likestudying life having access to only pre-natal and post-mortem information [21]. Adsorbatesmay completely restructure a catalyst surface and surface phases that are unstable in highvacua may well play a critical role in catalysis [22]. A differentially pumped pressure cell ispresented that partly bridges the pressure gap between analysis and catalysis by enabling insitu LEIS during reactions (Chapter 5). The possibilities of this cell are illustrated by an insitu LEIS study of the CO oxidation over Pt(110) at steady state and in the oscillatory regime(Chapter 6). This study also nicely illustrates, once more, the intimate relationship between


11

the catalyst surface (as monitored using in situ LEIS) and the catalytic performance (asmonitored with a quadrupole mass spectrometer).

The contents of the following chapters have been individually prepared for publication ininternational scientific journals and can be read independently. Some overlap betweenchapters is therefore unavoidable.

References

1. Colen R.E.R., Ph.D. Thesis, Leiden University (1997).2. Bowker M., The basis and applications of heterogeneous catalysis, Oxford University

press, Oxford (1998).3. Schüth F., Schuit Lecture February 18, 2002.4. Brongersma H.H., and Buck T.M., Nucl. Instr. Meth., B149 (1978) 697.5. Creemers C., Hove H., and Neyens A., Recent Dev. In Condens. Mater. Phys., 2 p. 363,

(J.T. Devreese Ed.), Plenum, New York, (1981).6. Rostrup-Nielsen J.R., Top. Catal., 1, 377 (1994).7. Niemantsverdriet J.W., Spectroscopy in Catalysis, an Introduction, VCH

Verlaggeselschaft, Weinheim, (1993).8. Bergmans R., Ph.D. Thesis, Eindhoven University of Technology (1996).9. Leerdam G.C., Ph.D. Thesis Eindhoven University of Technology (1991).10. V.D. Oetelaar L.C.A., Ph.D. Thesis Eindhoven University of Technology (1997).11. Jacobs J.-P., Ph.D. Thesis Eindhoven University of Technology (1997).12. Brongersma H.H., Groenen P.A.C., and Jacobs J.-P., Science of Ceramic interfaces II, (J.

Nowotny, Ed.), Elsevier, New York (1994).13. Taglauer E., Fundamental aspects of heterogeneous catalysis studied by particle beams,

(H.H. Brongersma, and R.A. Van Santen Eds.), NATO ASI B 265, p. 301. Plenum Press,1991.

14. Centi G., Trifirò F., Ebner J.R., and Franchetti V.M., Chem Rev., 88, 55 (1988).15. Brongersma H.H., V.D. Ligt G.C.J., and Rouweler G., Philips J. Res., 36, 1 (1981).16. Kraus A.R., Auciello O., Lamich G.J., Gruen D.M., Schultz J.A., and Chang R.P.H.,

J. Vac. Sci. Technol., A12, 1943 (1994).17. Kraus A.R., Im J., Schultz J.A., Smentkowski V.S., Waters K., Zuiker C.D., Gruen D.M.,

and Chang R.P.H., Thin Solid Films, 270, 130 (1995).18. Lin Y., Kraus A.R., Auciello O., Nishino Y., Gruen D.M., Chang R.P.H., and Schultz

J.A., J. Vac. Sci. Technol., A12, 1557 (1994).19. Fujii Y., Nakajima K., Namuri K., Kimura K., and Mannami M., Surf. Sci., 318, L1225

(1994).20. Pfanzelter R., Igel T., and Winter H., Surf. Sci., 375, 13 (1997).21. Somorjai G.A., Opening address 215th ACS National Meeting, Dallas, Texas (1998).22. Somorjai G.A., Top. Catal., 8, 1 (1999).

13

1The Influence of Compaction and Surface Roughness on

Low-Energy Ion Scattering Signals*

AbstractInvestigation of the surface composition of powders often requires compaction. To study theeffect of compaction on surface analysis, samples have been compacted at various pressuresranging from 0 Pa (i.e. no compaction) up to 2000 MPa. Low-Energy Ion Scattering (LEIS)was used to determine the composition of the outermost atomic surface layer, ScanningElectron Microscopy (SEM) was used to study the surface morphology of SiO2 test samples.Using SEM changes in the morphology due to compaction have been detected. The LEISyield of a compacted powder is found to be independent of the applied pressure during thecompaction in the range of 2 MPa up to 2000 MPa. Analysis of a sub-monolayer Ta2O5 on asilica support shows that the composition of the outermost atomic layer is not changed aftercompaction up to a pressure of at least 300 MPa. If compaction is applied, the absolute LEISyield appears to be independent of the specific surface area of silica supports in the range 50to 380 m2/g. A minor difference in LEIS signals is observed between compacted silicasupports and flat quartz samples. In order to determine the surface roughness factorindependently, and to study the material dependence of the surface roughness factor, angledependent LEIS measurements have been carried out on oxidized silicon, gallium and goldsurfaces. The results on the oxidized silicon confirm the small influence of surface roughnessfor silica particles, whereas the measurements on the more closely packed metallic galliumand gold surfaces indicate a significant surface roughness effect.

* The contents of this chapter has been submitted for publication: W.P.A. Jansen A.Knoester, A.H.J. Maas, P. Schmit, A. Kytökivi, A.W. Denier van der Gon, and H.H.Brongersma, Surface and Interface Analysis

Influence of compaction and surface roughness on LEIS signals

14

1. IntroductionRough, powdered materials are used in all kinds of products, such as paints, car tires,

medicines and catalysts. In the case of supported catalysts the active phase is spread overrough, powdered support materials to obtain thermostability, a maximum surface area andcoarse bodies that are suitable for use in reactors [1]. The way in which an active phase isspread over the support material determines its accessibility and its practical value in terms ofreactivity. Therefore, many studies have been devoted to the determination of the dispersion,wetting and interaction of active species with support materials [1-4].

Most surface science techniques require vacuum conditions. Before analysis, powdersare often compacted to facilitate the introduction of samples into vacuum. Moreover,compaction allows one to obtain a more homogeneous sample thickness, which is especiallyimportant when a technique relies on transmission of signals through the sample, as is thecase in infrared spectroscopy.

When analyzing powders with LEIS, one also has to deal with surface roughness.Surface roughness is not only determined by compaction, but also by the structure of thepowder. In the basic LEIS formula for the LEIS yield (Yi) of element i the so-called surfaceroughness factor R (0<R<1) corrects for signal losses due to surface roughness [5,6]:

Yi = I⋅ Ni ⋅ c ⋅ R ⋅ Pi+

⋅ dσi/dΩ, (1)whereI = the primary ion currentNi = the number density of the surface atoms of element ic = an instrumental factor taking into account the acceptance angle and the

efficiency of the analyzerPi

+ = the ion fraction of projectiles after scattering from element i

dσi/dΩ = the differential cross-section for scattering by atom i

Since Pi+ is generally independent of the matrix, c⋅Pi

+⋅dσi/dΩ will not change whenmeasuring the same element in different samples in the same set-up [5-7]. The current I ismeasured, hence, Yi/(Ni⋅R) can be determined. The Ni of a material can be determined bycomparing the Yi of that material with the Yi of a reference sample with a known Ni and thesame surface roughness. However, it can be very hard to obtain a reference sample with acertain surface roughness. Therefore, the influence of the surface roughness (or specificsurface area) on LEIS, has been studied to obtain values of the surface roughness factor R.

Several earlier studies deal with the influence of the surface roughness on LEIS yields[8-10]. Nelson [8] has presented a model showing that it is not the absolute size of theroughness that plays the major role in shadowing, but rather it is the slope of the roughnessthat is important. Hence, the signal loss due to surface roughness is not directly correlated tothe specific surface area, but to the gradients occurring in the surface. LEIS measurements ongold films deposited on sapphire and on γ-Al2O3 were presented that supported this model[8]. Two other studies report LEIS measurements on different aluminas [9,10]. Both studiesshow the ratio between the different elemental LEIS yields is independent of the surface

Chapter 1

15

roughness. This is very important since the surface composition is derived from this ratio.However, the outcome of both studies with regard to the absolute LEIS yield, from whichatomic densities are derived, is different. Margraf et al. [10] report that the absolute LEISyield strongly increases when going from rough to flat alumina surfaces (after sputtercleaning an increase up to a factor 5 or 6). J.-P. Jacobs et al. [9] observed 1.7 times higherLEIS yields on sputter cleaned sapphire (a flat α-Al2O3 reference) compared to those ofsputter cleaned pellets of pressed rough alumina powders (α-Al2O3 of 5.5 m2/g and γ-Al2O3

of 269 m2/g). In both studies all alumina surfaces were sputter cleaned, this may have causedsignal loss. For instance, R. Cortenraad et al. [11] have observed that ion bombardment of aclose-packed high-melting point material (W) at room temperature can lead to a signaldecrease of approximately 30% due to the sputter induced roughness and disorder. Duringsputtering atoms are removed, consequently sputtering can induce roughness on an atomicscale. Because of the high gradients in this kind of roughness, the signal should decrease dueto shadowing, according to Nelson [8].

Differences due to incomplete charge compensation and possibly sputter induced surfaceroughness, and contamination of the aluminas may have caused the difference between theobserved changes in the LEIS yield in refs. [9] and [10].

In this study we have used a series of silica supports with increasing specific surfaceareas to investigate the influence of compaction and surface roughness on the LEIS yield.Loose and compacted powders of these materials (compaction at pressures ranging from2 MPa up to 2000 MPa) have been analyzed with LEIS to study the outermost atomic surfacecomposition while SEM was used to study the surface morphology. Like alumina, silica is avery important support material in catalysis [1,12]. In contrast to α-Al2O3, which has a Mohshardness of 9 [28], the Mohs hardness being the usual measure for hardness in the mineralworld, silica can be straightforwardly compacted. Moreover, silica can be obtained in ahigher purity than alumina. As far as the authors are aware, no earlier work has beenpublished on the influence of compaction, or specific surface area of silica on the LEIS yield.In addition, we have studied the influence of surface roughness on Ga and Au to evaluate thematerial dependence of the surface roughness factor.

In the case of a supported catalyst with the active phase selectively present in theoutermost atomic layer, a compaction-induced mixing of the outermost and deeper atomiclayers would be detrimental. Therefore, we have also analyzed a model catalyst consisting ofa sub-monolayer Ta2O5 on a silica support.

2. Experimental

2.1 Silica samplesTable 1 gives an overview of the samples that were analyzed during this study. The

specific surface areas of the powders range from 50 m2/g to 380 m2/g. The Aerosil silicapowders are commercially available fumed silicas from Degussa AG. The manufacturer ofthe silica powder with a specific surface area of 185 m2/g is Akzo Düren. A high puritypolycrystalline flat quartz sample (1×1×0.5 cm3) has been used as a reference having a


16

“negligible” specific surface area. AFM analysis of the quartz sample revealed a line patterncorresponding to a polishing treatment with a peak to peak distance of 2 µm. The surfaceroughness between these lines has a root mean square (RMS) of 8 nm. Since the slopescorresponding to the polishing pattern form only a minor part of the surface, the RMS 8 nm isthe relevant surface roughness of this sample regarding LEIS analysis.

Table 1. Overview of the samples used in this study.

Sample Source Formula Specific SurfaceArea (m2/g)

Quartz SiO2 ~10-4

Aerosil Ox-50 Degussa SiO2 50Aerosil 130 Degussa SiO2 130Akzo Düren Akzo Düren SiO2 185Aerosil 380 Degussa SiO2 380

Ta2O5/Aerosil Ox-50* Degussa/FortumOil and Gas Oy

Ta2O5/SiO2 50

* A sub-monolayer Ta2O5 has been deposited on Aerosil Ox-50 using ALD, see section 2.2.

2.2 Ta2O5/SiO2 prepared with ALDA sub-monolayer Ta2O5 was deposited on a silica support (Aerosil Ox-50) with Atomic

Layer Deposition (ALD). A characteristic feature of this method is the controlled build-up ofsurface structures [13,14]. Tantalum oxide has been chosen since it completely wets on silica.Moreover, Ta2O5 is very stable and does not sinter during oxidizing cleaning treatments up to800 K.

The Aerosil Ox-50 was calcined in a muffle furnace at 723 K for 16 hours. The pre-treatment has been continued in the ALD reactor at 473 K for 2 hours. Then, Ta(OC2H5)5 waschemisorbed onto the Aerosil Ox-50 at 473 K followed by a nitrogen purge. Using an airtreatment at 723 K the ligand residues were removed. Since only one chemisorption cycle hasbeen performed, Ta2O5 is selectively present in the outermost atomic layer of the Ta2O5/SiO2.

2.3 CompactionTo study the effect of compaction, samples were compacted at various pressures ranging

from 0 Pa (i.e. no compaction) up to 2000 MPa. The powders were pressed into stainless steelcups (∅ 8.4 mm, depth 0.6 mm). For the analysis of the loose powder we have used a deeperstainless steel cup (∅ 8.4 mm, depth 2.25 mm).

To compact powders, cups filled with loose powder were mounted in a die made of hardsteel (∅ 8 mm). The die should assure a homogeneous transfer of the applied load to thesample. To avoid contamination, the die was cleaned thoroughly with ethanol before eachuse. A 10 kg weight was used to obtain a 2 MPa load. For higher loads a Carver laboratorypress (type 20500-109) was used. The Carver laboratory press is fitted with a load gauge.

Chapter 1

17

2.4 CleaningLEIS selectively probes the outermost atomic layer, hence surface contaminants would

obscure the intrinsic composition and thus have to be removed before analysis.The quartz sample was rinsed in 2-propanol. Prior to the LEIS analysis, the samples were

introduced into the load-lock of the UHV Calipso LEIS set-up. The base pressure of the load-lock is in the low 10-7 mbar range. In the load-lock the samples were cleaned using an oxygenatomic beam, for 2 minutes at an oxygen pressure of 2.2×10-4 mbar. The applied atomic beamsource is an Oxford Applied Research Atom/Radical Beam Source Model MPD21. Thesource operates by means of an electrical discharge generated by inductively-coupled RFexcitation at 13.56 MHz. Oxygen atoms produced by dissociation in the discharge can reachthe vacuum environment along with undissociated molecules via an array of fine holes in thebeam aperture. The electric potentials are such that the currents of ions and electrons thatescape the discharge during operation are less than 1 mA. Since the samples are not directlyinserted into the plasma, the sample temperature remains relatively low (~308 K) duringoperation.

Additional cleaning cycles with the atomic oxygen beam showed no change in thecleanness of the silica and Ta2O5/SiO2 samples, and no impurities were detected by LEIS(both 3 keV He+ and Ne+). In addition, the LEIS yields remained initially constant duringsputtering. Therefore, we assume that hydrogen, carbon, and nitrogen are effectively removedby the chemical cleaning procedure. The lower detection limit for the mass range O to Si hereis estimated to be ~1%, the detection limit for heavier elements is <1%.

2.5 Gallium sphereTo study the LEIS yield of a (hemi)spherical particle as a function of the impact position

on the sphere, a sphere of Ga was studied with a micro-beam. Gallium is a liquid at RT(melting point 302 K). The positioning of a Ga droplet (99.99% Ga from Goodfellow) on astainless steel support in an oxygen lean atmosphere (glove box, 3 ppm O2) leads to a perfectsphere because this shape has minimum surface free energy. The inset in Fig. 7 shows aphoto of the upper half of the studied ∅ 3.04 mm Ga sphere (the lower half is hidden behindthe brim of the stainless steel support). Note that a hemispherical particle with radius r givesthe same LEIS yield as a spherical particle with radius r, since the lower half of the sphericalparticle is invisible for LEIS.

2.6 Low-Energy Ion ScatteringTo determine the surface composition, LEIS experiments have been carried out in the

Calipso LEIS instrument. This ultra high vacuum set-up has a base pressure in the low 10-10

mbar range. The pressure increases to 10-9-10-8 mbar during LEIS experiments due to noblegas influx from the ion source. The primary ions are mass selected, focussed and directedperpendicular to the target. Ions scattered through 145o are accepted by the analyzer. Theanalyzer is a double-toroidal electrostatic analyzer, similar to that of the EARISS set-up,which has been described in more detail elsewhere [15-17]. This analyzer makes veryefficient use of the backscattered particles by measuring (i) simultaneously a considerable


18

part of the energy spectrum of the backscattered particles and (ii) 320o of the azimuthal range.In addition, the ion beam is rastered over an area A=1×1 mm2 during the measurements, thusreducing the ion induced damage even more. The sensitivity of the EARISS-type analyzertogether with an enlarged measuring area (rastered beam), allow us to perform static LEIS,i.e. with low dose and thus negligible damage. All measurements on the silica (supported)samples were, unless otherwise stated, carried out with 3 keV 4He+ ions, using typically4×1014 ions/cm2 per measurement. As a rule of thumb, a 4He dose of 1016 ions/cm2 is requiredto remove one monolayer. Pitts and Czanderna report that a 1 keV 3He+ dose of8.5×1016 ions/cm2 changes the Si:O ratio from 0.500 to 0.56 [18]. During a measurement a200 times lower dose is used, therefore, we do not expect a significant reduction due topreferential sputtering. To prevent charging of the isolating samples, low energetic electrons(10 eV) were flooded from all azimuths over the sample during measurements.

The samples are not only exposed to electrons and ions during the analysis itself, but alsoduring cleaning. The atomic beam source that is used to clean the samples produces very low-energy atoms and as a byproduct thermic ions and electrons in an extremely oxidativeatmosphere. Therefore, reduction of the silica due to preferential sputtering during thecleaning process can be excluded.

To analyze a ∅ 3.04 mm Ga sphere, a 3 keV Ne+ micro-beam has been made by placing a∅ 300 µm aperture in the primary beam line. This aperture is used as an object for the lensethat determines the beam spot at the sample position. To determine the beam size of thefocussed beam at the proper sample height, the micro-beam was deflected over a plate with a∅ 10 µm pinhole. Only ions passing through this pinhole are detected with a channeltron.Synchronization of the flux measurement and the deflection of the beam allow us todetermine the beam profile. Figure 1 shows the beam profile of the FWHM 60 µm beam thatwas used in this study. The manipulator that contains the samples was used to scan thesample under the micro-beam at a constant speed of 25 µm/s, which resulted in a currentdensity of 1.6⋅1014 ions/cm2.

Fig. 1. Beam profile of the FWHM 60 µm 3 keV Ne+ beam that was used to scan a ∅ 3.04 mm Gasphere. The two line scans have been summed over two perpendicular directions.

0

8

16

24

-60 -30 0 30 60

position (µµµµm)

yiel

d (a

rb.u

nits

)

60 µm FWHM

Chapter 1

19

2.7 Scanning Electron MicroscopyPrior to the SEM measurements a thin layer of gold was deposited on the silica

samples to prevent charging. The gold layer was applied with an Emitech K550 sputter coater(2.5 min. 30 mA, in an Au/Ar atmosphere at 10-1 mbar). No gold particles could be observedup to the maximum applied magnification (i.e. 200 000×).

The SEM measurements have been carried out in a SEM-FEG XL30 from Philips.The base pressure of the set-up is 10-6 mbar. All images were taken using 1.5 kV electrons.

3. Experimental Results

3.1 CompactionFigure 2 shows LEIS spectra of Aerosil Ox-50 before and after compaction at various

pressures. After normalisation to the applied ion dose, all LEIS yields are identical withinexperimental error (5%), except for the LEIS yield obtained on a non-compacted Aerosil Ox-50 (0 Pa). The spectrum of the non-compacted sample had to be multiplied by 2.2±0.2 to bescaled with the other spectra. The LEIS yields of compacted silica powders thus appear to beindependent of the applied pressure in the range from 2 up to 2000 MPa. All spectra,including that of the loose Aerosil Ox-50 (0 Pa), show the same Si:O LEIS yield ratio. Hence,the surface composition, which is derived from the Si:O ratio, is not influenced by thecompaction process.

Fig. 2. Spectra showing LEIS yields of Aerosil Ox-50 before and after compaction at variouspressures. The spectra have been shifted relative to each other for clarity, the offset can be seen at thehigh-energy end of the spectra.

0 1000 2000 3000Energy (eV)

0

50

100

LEIS

sig

nal (

Cou

nts/

nC)

O

Si

Fe

2 MPa

60 MPa

300 MPa

2000 MPa

0 Pa

0 Pa * 2.2


20

After compaction at 2000 MPa an iron peak appears in the LEIS spectrum of theAerosil Ox-50 (Fig. 2, identity Fe confirmed using Ne+). The appearance of Fe can beexplained in two ways. Firstly, iron from the die could be left on the sample surface. If thereis any Fe2O3 on the hard steel die this could be released due to heat and friction that occurduring compaction at 2000 MPa. Secondly, with SEM small cracks could be observed insamples that were compacted at 2000 MPa. Since the pills are pressed in stainless steel cups,Fe from the cups may have been detected. Note, the Fe peak corresponds to about only 2% ofa monolayer, the relatively large peak area originates from the high LEIS sensitivity for Fe.

LEIS measurements of a sub-monolayer Ta2O5 on silica show that the Ta LEIS yieldremains constant, within experimental error (5%), going from 2 to 300 MPa. Hence, aftercompaction up to 300 MPa the outermost atomic layer is left intact. After compaction at 2000MPa the Ta2O5/SiO2 sample became heavily contaminated with Fe. The origin of the Fe isprobably the same as in the case of the pure Aerosil Ox-50. Since we could not make an ironfree Ta2O5/SiO2 compaction at 2000 MPa, these results have not been evaluated.

Figure 3a shows an electronmicrograph of loose Aerosil Ox-50 powder with a specificsurface area of 50 m2/g. Figures 3b to 3d show Aerosil Ox-50 powder at the same scale aftercompaction at 2, 300, and 2000 MPa, respectively. After compaction the dimensions of thedark areas between the silica particles are strongly reduced. Also, the scale of surfaceroughness is reduced after compaction. The compaction at 300 MPa (Figure 3c) shows twotypes of roughness. The left-hand side resembles that for the compaction obtained at 2 MPa,the right-hand side resembles the compaction obtained at 2000 MPa. As explained in section2.3, the compaction is prepared by filling a cup with loose powder and compacting thispowder using a die. To a certain extend the filling will be inhomogeneous and the polisheddie has still roughness, therefore, the effective pressure will not be uniform over the sample.This difference in effective pressure can explain why part of the compaction made at theintermediate pressure of 300 MPa resembles that made at 2 MPa and the other part resemblesthat made at 2000 MPa.

Let us now try to correlate the LEIS and the SEM results as a function of the appliedpressure during compaction. Comparison of the electronmicrographs of the loose powder(Fig. 3a) and the compaction at 2 MPa (Fig. 3b) shows that after compaction the large gapsare filled up with material. Material in the large gaps that is present before compaction isapparently partly invisible for LEIS. The signal loss due to this shadowing effect is overcomeafter compaction at 2 MPa. In other words, macro roughness has disappeared. This mayexplain why the LEIS signal increases typically a factor of 2 after compaction.

Chapter 1

21

Fig. 3. Electronmicrographs showing the structure of loose Aerosil Ox-50 powder (3a), Aerosil Ox-50after compaction at 2 MPa (3b), 300 MPa (3c) and 2000 MPa (3d) (magnification 40 000 ×).

Figure 4 shows an electronmicrograph of Aerosil Ox-50 after compaction at 2 MPa ata magnification of 200 000×. The micro morphology of the silica is clearly visible: silicatespecies spheres (i.e. primary particles) that agglomerate into clusters. As an example, one ofthese clusters has been encircled in figure 4. The specific surface area of the silica depends onthe size of the primary particles, and the size and stacking of the clusters. Due to compactionat a high pressure the stacking of the silica clusters can be changed. The space between thesilica clusters becomes evenly filled after compaction at high pressures (see figure 3d). Someclusters seem to be merged. The LEIS measurements however, still give the same Si and Oyields. Analyzing a sub-monolayer Ta2O5 on a silica support, we find the same Ta yield aftercompaction at different pressures. Thus, the outermost atomic Ta2O5 layer remained intactafter compaction up to 300 MPa, although a significant part of the clusters has been pressedtogether at this pressure, as is shown in figure 3c. Since similar LEIS yields were obtained forTa2O5/SiO2 compacted between 2 MPa and 300 MPa, we conclude that the microscalestacking of the silica clusters does not influence the LEIS yield.

Fig. 4. Electronmicrograph showing the structure of Aerosil Ox-50 after compaction at 2 MPa(magnification 200 000 ×). A so called “silica cluster” has been encircled.

200 nm

3b

1 µm

3d

1 µm

3c

1 µm

1 µm

3a


22

As already mentioned, alumina is another important support material. Alumina hasanother micro morphology than silica (more flake like, compared to the spherically shapedsilica). Therefore, we have also analyzed the difference in LEIS yields betweennon-compacted and compacted commercial γ-Al2O3 powder (Akzo/Ketjen 000 1.5E, 205m2/g). The experimental details are comparable to those on silica, except for the fact that thesamples were cleaned by calcination (heating overnight in 0.5 bar of oxygen to 550 K). Formore details on sample preparation and measurement refer to Peeters et al. [4]. If LEISspectra of non-compacted and compacted γ-Al2O3 (compaction at 300 MPa) are compared, adoubling (factor 2.0±0.2) of the absolute LEIS yields is observed for the compacted samples.The Al:O ratio remains constant after compaction.

3.2 Specific Surface AreaTable 2 shows the relative LEIS yields (as determined by integrating the LEIS signals) of

a flat polycrystalline quartz sample and pressed silica powders with different specific surfaceareas ranging from 50 to 380 m2/g. The Si and O LEIS yields in the table have beennormalised to the yields obtained on the quartz sample.

Table 2. Relative LEIS yields of a flat quartz sample and pressed silica powders (300 MPa) with aspecific surface area ranging from 50 to 380 m2/g

Sample Specific surface area(m2/g)

Si O Si:O ratio

Quartz ~10-4 1.00 1.00 1.00Aerosil Ox-50 50 0.80 0.84 0.95

Aerosil 130 130 0.77 0.81 0.95Akzo Düren 185 0.83 0.82 1.01Aerosil 380 380 0.83 0.84 0.99

Let us first discuss the data obtained on the different silica powders. Since we have shownthat compaction up to at least 300 MPa does not change the composition of the outermostatomic layer, all powders have been compacted at this pressure. The LEIS yields that areobtained on the different powders are equal within experimental error. For instance, thesomewhat lower LEIS yields that are found for the Aerosil 130 can easily be ascribed to theuncertainty in the determination of the beam current and/or traces of contaminants that mightstill be present after the cleaning procedure with the atomic oxygen beam. Hence, themeasurements show that LEIS yields on isolating, oxidic powders are reproducible within5%. The results appear to be independent of the manufacturer of the silica samples. The LEISyields on the Aerosil powders from Degussa are e.g. equal to those on the silica powder fromAkzo Düren. Since the absolute LEIS yields remain constant, the surface composition that isderived from the ratio between the Si:O LEIS yields is also independent of the specificsurface area in the range from 50 up to 380 m2/g.

Chapter 1

23

How can we understand the apparent independence of the LEIS yields on the specificsurface area? The following model could explain this. Suppose figure 5 represents the cross-section of a surface that is probed with an ion beam. The spheres in the figure represent thespherical agglomerates of the silica powder. The pattern of the surface roughness in figure 5repeats itself. Therefore, an equal fraction of the surface will be visible for LEIS, irrespectivewhether the length scale of the repetition/figure is of the order of 1 nm (for a high specificsurface area) or of the order of 1 mm (for a low specific surface area). At the upper limit(order 1 mm) the ion beam does not probe complete patterns anymore; the yield may thendiffer from the average yield that represents the complete pattern. The lower limit (order1 nm) stems from the fact that a LEIS signal originates from atom cores, hence fordimensions of the order of 1 nm a surface may not be considered homogeneous anymore. Inpractice, however, the lower limit will only be encountered during sputtering. The sputterprocess can remove single atoms, hence can create roughness on an atomic scale, thus belowthe order of 1 nm. This effect is particularly important for high melting-point materials, suchas the refractories [11]. In other materials diffusion at room temperature is efficient enough tocancel or reduce this sputter induced atomic roughness. Higher temperature treatments, likecalcination, will also eliminate/reduce this type of roughness.

Fig. 5. Schematic representation of the cross-section of a surface, the arrows indicate probing ions.Dependent on the size of the length scale the figure either represents a surface with a high or a lowspecific surface area.

In our case, the SEM results showed primary particle sizes ranging from ~10 nm (380m2/g) to ~60 nm (50 m2/g). These values are in agreement with the primary particle sizesprovided by the manufacturer [19]. Even for a material with an extremely high specificsurface area of 1000 m2/g (e.g. active carbon) the primary size will be ~6 nm, hence more orless homogeneous at an atomic scale. (Note that in the case of active carbon another problemmay influence LEIS analysis as well: the LEIS yield probably decreases due to remaininghydrogen from the synthesis process, rather than due to the surface roughness correspondingto the high specific surface area).

length scale

Probing ions

⇓⇓⇓⇓


24

This model can explain the influence of compaction as well. The upper limit of thesurface roughness scale (~1 mm) plays apparently a role in the case of loose powders.However, after compaction at 2 MPa the roughness is reduced below this scale and the LEISyields have increased by approximately a factor 2 - 2.2. Further compaction at higher loads(up to 2000 MPa) does not change the LEIS yields anymore. After pressing at 2 MPa theroughness is already much smaller than 1 mm and up to 2000 MPa the roughness scaleremains much higher than 1 nm.

If we compare the LEIS yields obtained on the different silica powder samples to thoseobtained on the flat quartz sample we find a significant difference. The LEIS yields for boththe Si and O are about 17% higher than those of a pressed silica powder. Again we see noinfluence on the ratio of the different LEIS yields. The Si:O ratio of 1.02 is consistent withthose obtained on the pressed silica powders. The differences in the LEIS yields obtained onpolycrystalline quartz and on the pressed silica powders, which are fumed silicas, partlyoriginate from the difference in (atomic) density. The density of quartz is 2.65 g/cm3, while afumed silica has only a density of 2.25 g/cm3 [20]. One can therefore expect that the ratiobetween the surface densities is (2.25)2/3 / (2.65)2/3 = 0.89. The ratio of the LEIS yields equals0.83. Hence, the surface roughness factor R = 0.93±0.07. It should be noted that the “flat”quartz sample that has been used for comparison has a RMS 8 nm surface roughness. In orderto understand the effect of the roughness of the “flat” reference sample, the theoretical R, thatwould account for the difference in LEIS yield between atomically flat silica and compactedsilica supports, is calculated below.

4. Modeling the surface roughness factor in LEIS

A probing ion that penetrates a material can be blocked (or physically shielded) on theoutgoing trajectory and will then not contribute to the LEIS signal (as illustrated in Fig. 5).An ion passing by an atom or a cluster of atoms may still become undetectable when itstrajectory is close enough to the material to be neutralized. One of the consequences of theseprocesses is the decrease of the LEIS yield by surface roughness. Earlier models, of Nelson[8] and later Jacobs et al. [9], predict losses up to 50% due to physical shielding whencomparing rough to atomically flat surfaces. As will be shown, these models overestimate thephysical shielding because of their 2-dimensional approach. Moreover, they only reflect theeffect of physical shielding. We will, therefore, present a 3-dimensional model that quantifiesthe changes in physical shielding and neutralization as a result of surface roughness. In thismodel, primary ions that impinge perpendicular onto the target surface and scatter over145°±1° (the acceptance ∆ϑ = 2°) will be considered (the geometry of the Calipso LEIS set-up). To get a more general view, ions impinging perpendicular to the target surface thatscatter over 135°±1° will be modeled as well. The Round Robin experiment of 1998 showedthat high impact angles and scattering angles in the range 135° to 145° are commonly used inLEIS set-ups [22]. The model can be easily extended to other geometries. Hollow cones,having apical angles of 145°±1° and 135°±1°, are used to model the analyzed fraction of thebackscattered ions. SEM micrographs (Figs. 3 and 4) show that a combination of close-

Chapter 1

25

packed spheres can very well represent the silica surface. The spheres correspond to theprimary silica particles and their size depends on the specific surface area of the silica (seealso above). For a certain specific surface area the spheres are equally sized, since the silicasare monodisperse. In section 3.1 the LEIS yield of silica compacts was shown independent onthe microscale stacking of the silica clusters. Therefore, we have assumed a surfaceconsisting of equally sized close-packed spheres (Fig. 6), to model the silica surface.

Fig. 6. Top (6a) and exploded side (6b) views showing a hollow cone and a close- packed sphericalsurface that were used to model backscattered ions and a silica surface, respectively. Cross-sectionsof the cone and the spheres have been calculated at 5° intervals for the range 0° < ω < 180°, asindicated in Fig. 6b. The values 0° < ω < 90°, give the upper limit for the physical shielding. Thosebetween 90° < ω < 180°, that correspond to a free sphere, give the lower limit for the physicalshielding. Regions that are invisible to LEIS are shown in black.

Figure 6a shows the top view of a close-packed layer of equally sized spheres. The figureshows that the top layer covers only 90.7% of the surface. The second layer of spheres, iscompletely physically shielded by the top layer. Therefore, the openness causes a signal lossof 9.3% (the black area in Fig. 6a) when comparing a close-packed spherical surface with aflat surface. Also, not all ions that scatter from the top layer of a rough surface can contributeto the LEIS signal. The percentage of the ions that are blocked by the sphere from which they

6a

φ

6b

A

probing ions

φ

ω


26

scattered depends on the impact point and thus on the angle ω (Fig. 6b). Shielding due toneighboring spheres will be maximal at places where the neighboring spheres are closesttogether. The physical shielding is maximal along lines connecting spheres that touch eachother (e.g. along line A in Fig. 6a). Therefore, the maximum physical shielding has beendetermined by integrating the cross-section of a close-packed spherical surface and a hollowcone over φ (Fig. 6) for different values of ω along A. A minimum value for the physicalshielding can be calculated by ignoring the shielding by neighboring clusters. Therefore, theminimum physical shielding has been determined by integrating the cross-section of a hollowcone with a free sphere over φ for different values of ω. To integrate over ω, the incident ionsare assumed to be homogeneously distributed over the equatorial planes of the spheres. Thisis valid if the dimensions of the raster of the probing ion beam (1×1 mm2 in this case) arelarger than the radii (r) of the spheres that describe the silica surface (for r << 1 mm).

Using this model and hollow cones with apical angles corresponding to a scattering angleof 145±1°, the maximum signal loss due to physical shielding (along line A) is 12.1%, whilethe minimum loss (for a free sphere) is 11.7%. Hence, for a scattering angle of 145°±1° thesignal loss due to physical shielding will be in the range 11.9%±0.2%. For a scattering angleof 135°±1° the signal loss due to physical shielding will be in the range 15.6%±0.7%. Notethese values are significantly lower than the earlier predicted value of 50% physicalshielding. The main difference is due to the 3-dimensional approach.

Besides physical shielding, surface roughness may cause losses due to extra neutralizationas well. Since matrix effects have been shown to be absent for SiO2 [23], the neutralizationcan be quantified using literature numbers that have been obtained on Si. In the case ofsilicon the neutralization process (vc(Si) = 4.8⋅105 m/s [24]) is dominated by the so-calledviolent collision neutralization [24,25]. Since this is a binary process, between a single ionand an atom, it will not be affected by roughness [25]. Besides the violent collisionneutralization, part of the ions is neutralized due to so-called Auger neutralization. The latterneutralization process has a longer range and may, therefore, be influenced by surfaceroughness. The Hagstrum model [25-27], which was developed for metals, takes this intoaccount by calculating the ion fraction (Pi

+ in formula 1) with the reciprocal of the ionvelocities normal to the surface (v⊥) on the ingoing (vi

⊥) and outgoing (vf⊥) ion trajectories:

.)11(, ⊥⊥ +⋅−

+ = fiAugerc vv

v

i eP (2)

The characteristic velocity for the 3 keV He+ Auger contribution to the neutralization(vc,Auger) in silicon can be estimated to be at most vc,Auger = 4×104 m/s [25]. If this value andthe Hagstrum model are applied to calculate the extra neutralization loss, the total signal lossdue to surface roughness (physical shielding + extra neutralization) is 22% or 19% for set-upsusing scattering angles of 135° or 145°, respectively. Hence, the surface roughness factor thatis found theoretically (R = 0.81) is a little lower than the experimental value (being R =0.93±0.07). Realizing that the upper limit of vc,Auger is used to calculate the neutralization, andthat our “flat” reference sample has a RMS 8 nm roughness, the best estimate for the surface

Chapter 1

27

roughness factor is the average of the theoretically and experimentally determined values,which yields R = 0.87±0.06.

Note that the model assumes that the radii (r) of the spheres that describe the silicasurface are in the range 1 nm << r << 1mm. This has been shown valid for all silicas,however, the result does not include sputter-induced roughness and should therefore only beapplied for statically analyzed samples.

5. Dependence of the LEIS signal for a Ga sphere on the tilt angle of the surfaceTo study the LEIS signal of a (hemi)sphererical particle as a function of the impact

position on the sphere, a ∅ 3.04 mm droplet of Ga was scanned under a FWHM 60 µm 3 keVNe+ beam. By collecting the Ga LEIS yield during 1s intervals, while moving the sample at25 µm/s, the LEIS signal could be determined as a function of the angle ω (see Fig. 6b or Fig.7 for the definition of ω). Figure 7 shows the Ga LEIS signal as a function of ω, while theinset shows a scaled photo of the upper half of the analyzed Ga droplet (the lower half of thedroplet is hidden behind the brim of the support). The photo shows clearly the spherical shapeof the upper half of the Ga droplet.

Fig. 7. The Ga LEIS signal of a ∅ 3.04 mm Ga sphere as a function of the angle ω, as determinedwith a FWHM 60 µm Ne+ beam. The solid line represents a fit of the LEIS signal with sin(ω). Theinset shows a scaled photo of the upper half of the analyzed Ga sphere.

-500

0

500

1000

1500

2000

2500

0 0.5 1 1.5 2 2.5 3 3.5 4

Ga

LEIS

sig

nal (

arb.

u.)

.

0 48 70 90 110 132 180

ω

angle ωωωω

-1.5 -1 -0.5 0 0.5 1 1.5

position (mm)


28

It is found that the Ga LEIS signal follows a sin(ω) curve, as is clear from the solid curve inFig. 7. (The remaining LEIS signal outside the sphere is caused by the background of thestainless steel support. The small deviations from sin(ω) for ω ~ 90° are probably caused bymicro-cracks). Integration of sin(ω) over the beam area gives 2/3⋅π⋅r2 (corresponding to R =0.66) If not a single sphere, but a close-packed surface of equally sized spheres is analyzedthe signal will decrease further because of the extra packing (see above). This results in asurface roughness factor of R = 0.60 for a surface of close-packed equally sized spheres.Hence, the surface roughness factor of Ga differs from the previously determined surfaceroughness factor for SiO2 (R(SiO2) = 0.87±0.06.

6. Material dependence of the surface roughness factorSurprisingly, the previously described experiment with Ga yielded a 27% lower value

for R(Ga) than for R(SiO2). Therefore, we have determined the surface roughness factor ofSiO2 in independent experiments. Instead of the flat quartz sample, LEIS signals for anoxidized flat silicon wafer were compared with those for a pressed silica powder. The Si:OLEIS yield ratio of the oxidized Si wafer differs from that of silica (0.79 vs. 1.0), therefore asingle roughness factor for both elements can not be determined in a straightforward manner.It was found that the roughness factor for silicon was 0.71, whereas for the oxygen a highervalue of 0.90 was determined. Since the surface is mainly oxygen terminated, the latter valueshould be the most reliable, and indeed this is close to the previously determined value of0.87±0.06. Next, the LEIS yields of the flat, oxidized Si wafer were measured as a functionof the tilt angle of the sample by rotating the sample over ω (see Figure 8). The measureddependence has been used to calculate the surface roughness factor which yieldsR(SiO2) = 0.93. Note, that in this evaluation the surface roughness factor yields the samevalue for the Si and O signals. Hence, three independent experiments show that R(SiO2) =0.87±0.06.

Similarly to the oxidized Si, the LEIS signal of atomically flat Au on mica has alsobeen measured as a function of the tilt angle of the sample (Fig. 8). Figure 8 shows a cleardifference between the tilt angle dependence of the LEIS yields for oxidized Si and Au.Whereas the effect on the LEIS yield is only minor for oxidized Si, the effect on the Au LEISyield is significant (R(Au) = 0.72±0.07).

Table 3 summarizes the found surface roughness factors. Two values are given for thesurface roughness, the higher value (exclusive loss due to openness) applies for singlespherical particles, or low (metal) loads; the lower value (inclusive loss due to openness)applies for high (metal) loads. As shown above, the openness of a close-packed supportcauses a loss of 9.3% when comparing a monodispersed close-packed spherical surface witha flat surface. Physical shielding from neighboring clusters will further reduce the LEIS yieldof a surface consisting of close-packed clusters. If a scattering angle of 145° is applied, thecomplete LEIS signal loss on a close-packed spherical surface due to openness is 9.8%.

Chapter 1

29

Table 3. Surface roughness factor for different materials as determined in this study.

Material Surface roughness factor(including loss due to openness)

Surface roughness single sphere(exclusive loss due to openness)

Au 0.72±0.07 0.79±0.07Ga 0.60±0.07 0.66±0.07

SiO2 0.87±0.06 0.96±0.06

Fig. 8. Au (diamonds), Si (squares) and O (stars) LEIS yields as function of the tilt angle. The lineindicates sin(ω), the Si and O yields have been determined on an oxidized Si wafer.

It should be noted that for a closed packed sample that contains a mixture of cluster sizes, the9.3% openness may be partially filled up by smaller clusters. This may lead to a few percenthigher LEIS signal.

We conclude that different materials may result in different surface roughness factors.The differences between the various materials are ascribed to differences in the neutralizationprobability and packing of the surfaces. It can be expected that materials containing lightelements and a low packing density (small shadowing effects and low neutralization cross-section by neighbor atoms) result in a high R value, such as found for SiO2. In contrast, metalsurfaces having relatively high Z numbers and exposing closely packed surfaces result in alower R value, like Au or Ga.

0

0.25

0.5

0.75

1

0 30 60 90

Angle from normal incidence (ωωωω )

LEIS

Yie

ld /

LEIS

Yie

ld a

t 00


30

7. Concluding remarksFor both silica and γ-Al2O3 powders the elemental LEIS yield ratios are unchanged by

compaction. The absolute LEIS yields of O and Si in silica, however, increase strongly(typically by a factor of 2.2±0.2) after compaction at 2 MPa. Further compaction up to atleast 2000 MPa did not change the absolute LEIS yield anymore. A similar increase in theLEIS yield (factor 2.0±0.2) was obtained going from loose γ-Al2O3 powder to a compactedone. For reliable quantitative LEIS the measurements should only be performed withcompacted powders. Moreover, compaction facilitates introduction of samples into vacuum,increases the absolute LEIS yield and, therefore, improves the signal to noise ratio.Measurements on a sub-monolayer of Ta2O5 on a silica support confirmed that silicasupported samples can be compacted up to at least 300 MPa without influencing thecomposition of the outermost atomic layer (higher pressures have not been addressed duringour study).

If macro roughness is suppressed by compaction, calibration of the elemental LEISsensitivity becomes independent of the specific surface area of silica in the range 50 to380 m2/g. Earlier studies have shown the same independence for alumina compacts in therange 5.5 – 269 m2/g [9,10]. Hence, in spite of the huge difference in micro morphology(silica has a spherical geometry whereas alumina has a more flake like geometry), the surfaceroughness factor R remains constant when comparing either alumina or silica compacts withdifferent specific surface areas. Hence, compacted powders can be used as reference samplesfor the calibration of the elemental LEIS sensitivity of other compacted powdersindependently of their specific surface area.

If flat reference samples are used to quantify data obtained on compacted powders, thedata should be corrected for surface roughness. When comparing a flat quartz sample withcompacts of silica supported materials the experimentally determined surface roughnessfactor is R(SiO2) = 0.93±0.07. It should be noted that the “flat” quartz sample that has beenused for comparison has a RMS 8 nm surface roughness. In order to understand the influenceof the roughness of the “flat” reference sample, we have modeled the signal differencebetween atomically flat quartz and compacts.

Earlier models [8,9] predict losses up to 50% due to physical shielding when going fromatomically flat to rough surfaces. In those studies the effect was overestimated because of the2-dimensional approach. Our 3-dimensional model shows that the decrease due to physicalshielding is only 11.9%±0.2% for primary ions that impinge perpendicular to the target andscatter over 145°. For other widespread geometries, using ions impinging perpendicular ontothe target surface and scatter angles in the range 135° - 145°, slight deviations (<4%) fromthis value are found.

Besides physical shielding, surface roughness may cause losses due to extra neutralizationas well. Using the Hagstrum model to account for the extra Auger neutralization a surfaceroughness factor R(SiO2) = 0.81 is obtained. When realizing that the upper limit of vc,Auger isused to calculate the neutralization and that our “flat” reference sample has a RMS 8 nmroughness, the most likely value for R is the average value R(SiO2) = 0.87±0.06. One should

Chapter 1

31

note that the model does not include sputter-induced roughness. Hence, the result is onlyvalid for statically analyzed samples.

The Ga LEIS signal of a macroscopic Ga sphere is determined as a function of the tiltangle ω using a microbeam. From this experiment follows that the Ga LEIS yield shows asinusoidal behavior as a function of the tilt angle. Hence, the surface roughness factor for asingle Ga sphere equals R(Ga) = 0.66±0.07 (including losses due to opennessR(Ga) = 0.60±0.07). A similar value is experimentally determined for Au,R(Au) = 0.72±0.07. The different surface roughness factor values are ascribed to differencesin neutralization and packing of the surfaces. In absence of an R-factor determination for aspecific system, the most likely the surface roughness factor for (typically close packed)metal surfaces is ~0.7 and for (more open) oxide surfaces ~0.9

AcknowledgementThe Netherlands Organization for Scientific Research (NWO) and the European ScienceFoundation (ESF) are gratefully acknowledged for their financial support.

References

[1] R.A. van Santen, P.W.N.M. van Leeuwen, J.A. Moulijn, B.A. Averill (Eds.),”Catalysis: An Integrated Approach”, 2nd revised and enlarged edition, ElsevierAmsterdam (1999)

[2] L.C.A. van den Oetelaar, A. Partridge, P.J.A. Stapel, C.F.J. Flipse, H.H. Brongersma,J. Phys. Chem. B102, 9532 (1998).

[3] J. Leyrer, R. Margraf, E. Taglauer, and H. Knözinger, Surf. Sci., 201, 603 (1988).[4] I. Peeters, A.W. Denier van der Gon, M.A. Reijme, P.J. Kooyman, A.M. de Jong, J.

van Grondelle, H.H. Brongersma, and R.A. van Santen, J. Catal., 173, 28 (1998).[5] E. Taglauer, “Fundamental Aspects of Heterogenous Catalysis Studied by Particle

Beams” H.H. Brongersma, and R.A. v. Santen (Eds.), NATO ASI B265, PlenumPress, (1991) p. 301

[6] H.H. Brongersma, P.A.C. Groenen, and J.-P. Jacobs, “Science of Ceramic interfacesII”, J. Nowotny Editor, Elsevier, (1994) p. 113

[7] W.P.A. Jansen, M. Ruitenbeek, A.W. Denier van der Gon, J.W. Geus, and H.H.Brongersma, J. Catal., 196, 379 (2000)

[8] G.C. Nelson, J. of Appl. Phys., 47, 1253 (1976).[9] J.-P. Jacobs, S. Reijne, R.J.M. Elfrink, S.N. Mikhailov, M. Wuttig, and H.H.

Brongersma, J. Vac. Sci. Technol., A 12, 2308 (1984).[10] R. Margraf, H. Knözinger, and E. Taglauer, Surf. Sci., 211/212, 1083 (1989).[11] R. Cortenraad, S.N. Ermolov, B. Moest, A.W. Denier van der Gon, V.G. Glebovsky,

and H.H. Brongersma, accepted for publication in Nucl. Instr. Meth., B174, 173(2001).

[12] Cornils B., W.A. Herrmann, R. Schlögl, and C.-H. Wong (Eds.), “Catalysis from A toZ, a concise encyclopedia”, Wiley-VCH, (2000) p. 526


32

[13] T. Suntola, Mater. Sci. Rep., 4, 261 (1989).[14] E.-L. Lakomaa, Appl. Surf. Sci., 75, 185 (1994).[15] G.J.A. Hellings, H. Ottevanger, S.W. Boelens, C.L.C.M. Knibbeler, and H.H.

Brongersma, Surf. Sci., 162, 913 (1985).[16] G.J.A. Hellings, H. Ottevanger, C.L.C.M. Knibbeler, J. van Engelshoven, and H.H.

Brongersma, J. of Electron Spectrosc. Relat. Phenom. 49, 359 (1989).[17] R.H. Bergmans, A.C. Kruseman, C.A. Severijns, and H.H. Brongersma, Appl. Surf.

Sci., 70/71, 283 (1993).[18] J.R. Pitts, and A.W. Czanderna, Nucl. Instr. Meth., B13, 245 (1986).[19] Aerosil Fumed Silica, Degussa Corporation[20] Handbook of Chemistry and Physics, (D.R. Lide, Ed. In Chief), CRC Press 75th Ed.

(1995) table 4-95[21] H.H. Brongersma, and G.C. v. Leerdam, “Fundamental Aspects of Heterogenous

Catalysis Studied by Particle Beams”, H.H. Brongersma and R.A. v. Santen (Eds.),NATO ASI Series B265, Plenum Press, (1991) p. 283

[22] H.H. Brongersma, M. Carrere-Fontaine, R. Cortenraad, A.W. Denier van der Gon,P.J. Scanlon, I. Spolveri, B. Cortigiani, U. Bardi, E. Taglauer, S. Reiter, S. Labich, P.Bertrand, L. Houssiau, S. Speller, S. Parascandola, H. Ünlü-Lachnitt, and W. Heiland,Nucl. Instr. Meth., B142, 377 (1998).

[23] G.C. van Leerdam, and H.H. Brongersma, Surf. Sci., 254, 153 (1991).[24] Mikhailov S.N., Elfrink R.J.M., Jacobs J.-P., Van den Oetelaar, Scanlon P.J., and

Brongersma H.H., Nucl. Instr. Meth., B93, 148 (1994).[25] R. Cortenraad, A.W. Denier van der Gon, H.H. Brongersma, S.N. Ermolov, and V.G.

Glebovsky, Phys. Rev., B65, 195414 (2002).[26] H.D. Hagstrum, Phys. Rev., 96, 336 (1954).[27] H.D. Hagstrum, “Inelastic ion-surface collisions”, H.H. Tolk and J.C. Tully Eds.,

Academic Press, New York, (1977) p. 1[28] Handbook of Chemistry and Physics, (D.R. Lide, Ed. In Chief), CRC Press 75th Ed.

(1995) table 12-186

35

2aNoble Metal Segregation and Cluster Size of Pt/Rh/CeO2/γγγγ-Al2O3

Automotive Three-Way Catalysts studied with LEIS*

AbstractA method is presented to determine the average cluster size of supported catalysts using low-energy ion scattering (LEIS) data. The method is particularly suited to determine the averagecluster size of atomically dispersed metals and can be used for clusters with diameters up to10 nm. For Pt/γ-Al2O3 quantitative agreement is shown between the average cluster size asdetermined with LEIS and with transmission electron microscopy (TEM). In the case of three-way catalysts, classic methods to determine the cluster size, such as TEM or COchemisorption, often fail or produce ambiguous results because of the presence of ceria.Therefore, LEIS has been applied to determine the average noble metal cluster size of acommercial three-way catalyst. Moreover, LEIS shows that the three-way catalyst containsmixed Pt/Rh clusters, with a surface strongly enriched in Pt after reduction.

* The contents of this chapter has previously appeared in W.P.A. Jansen, J.M.A. Harmsen,A.W. Denier van der Gon, J.H.B.J. Hoebink, J.C. Schouten, and H.H. Brongersma Journal ofCatalysis, 204, 420-427 (2001). Note that minor changes occur, due to new insightconcerning the surface roughness factor in LEIS, as presented in chapter 1.

Three-way catalyst clustersize determination

36

1. IntroductionThree-way catalysts are widely used in new cars for the simultaneous oxidation of

hydrocarbons and carbon monoxide as well as reduction of nitric oxide from the exhaust ofthe engine. This is possible because of the precise lambda sensor, which keeps thecomposition of the exhaust gas near stoichiometric. In reality, the composition oscillates witha frequency of about 1 Hz around the stoichiometric point, thus imposing transient behaviourupon the monolithic converter. After a cold-start of a car, the temperature of the catalyticsurface causes the kinetics of the oxidation and reduction reactions to be too slow to obtainthe required conversion.

A detailed transient kinetic model based on elementary reaction steps would allowoptimisation of the currently used catalysts as well as the control system [1]. A model basedcontroller could be used for optimising conversions based on model calculations. Research atEindhoven University of Technology aims at the construction of such a model [2].

For kinetic modelling, information about the noble metal dispersion is a primaryrequirement. Also for catalyst optimisation and for research on catalyst ageing, informationabout the average cluster size is important. Classic methods of obtaining this type ofinformation, such as CO chemisorption, may fail or give ambiguous results for ceriacontaining catalysts, where huge CO storage capacities have been found compared to similarcatalysts without ceria [3-5]. Holmgren et al. [6] used CO chemisorption at 195 K in order tosuppress the influence of ceria. At this temperature the CO uptake on ceria was stronglysuppressed, but not completely hindered. Apart from the influence of ceria, chemisorptionmay fail because of anomalous average metal co-ordination numbers at highly dispersedsurfaces [7]. The latter is explained by an increased probe-molecule-to-metal stoichiometryfor metal atoms situated at the corners and edges of small metal clusters. Moreover, thepossibility of dissociative CO adsorption has been reported on small Pt particles (1 nm – 3nm) deposited on alumina [8,9]. Electron microscopy (TEM/HREM) is difficult for ceriacontaining catalysts due to the low contrast between the noble metal and the ceria [10].

This paper addresses the possibility to use Low Energy Ion Scattering (LEIS) fordetermining the average noble metal cluster size of a commercial Pt/Rh/CeO2/γ-Al2O3

catalyst. LEIS selectively probes the outermost atomic layer, hence only the surface of acluster is visible. Therefore, the noble metal LEIS yield does not only depend on the totalmetal loading and the specific surface area, but also on the fraction of the atoms in a clusterthat is accessible to LEIS. This fraction depends on the cluster size. Therefore, the averagecluster size of a catalyst can be determined with LEIS, if the total metal loading and thespecific surface area are known. The method has been verified by using a Pt/γ-Al2O3 catalystin order to compare the calculated average cluster size from LEIS signals with the averagecluster size as determined by TEM. After verification, LEIS has been applied to determine theaverage cluster size of the ceria-containing commercial three-way catalyst. Moreover, LEIShas been used to solve the question whether or not the three-way catalyst contains separate Ptand Rh clusters or mixed Pt/Rh clusters.

Chapter 2a

37

2. Experimental

2.1 Catalyst pretreatmentThe Pt/γ-Al2O3, Rh/γ-Al2O3, Pt/Rh/γ-Al2O3 catalysts and a commercial

Pt/Rh/CeO2/γ-Al2O3 catalyst, as used for coating monoliths, have been provided by dmc2

A.G. (Hanau, Germany) as powders with a mean particle diameter of 12 µm. The powdershave been pressed, crushed, and sieved to obtain fractions with pellet diameters between 0.11and 0.15 mm. To enhance isothermicity during kinetic measurements, the catalysts have beendiluted with inert α-Al2O3 (0.15 mm – 0.21 mm) in the ratio of 1.5 gram α-Al2O3 per 1 gramcatalyst material. In order to enable reproducible kinetic experiments, the followingpretreatment has been carried out for all catalysts. The catalyst is heated to 773 K in a steadyflow of He. Then the catalyst is oxidised during 1 hour by a stream containing 25 vol. %oxygen. Next the catalyst is kept under flowing He for 30 minutes in order to purge reversiblyadsorbed oxygen, followed by reduction in a He stream containing 5 vol. % H2 at 773 K for 2hours. Finally, the catalyst is allowed to cool down to reaction temperature under a He stream.

Prior to LEIS analysis, all powder samples have been compacted into pellets at 300MPa. Recently, we have shown that compaction at 300 MPa does not influence the surfacecomposition [11]. LEIS selectively probes the outermost atomic layer, hence surfacecontaminants would obscure the intrinsic composition, and therefore have to be removedbefore analysis of the intrinsic composition is possible. To clean the samples an atomicoxygen beam has been used. The use of atomic oxygen permits very effective cleaning at lowtemperatures (ca. 310 K) [12]. After the oxidative cleaning process, the samples are reducedat 573 K in 20 kPa H2 flowing at 2.6 mmol/min for 10 minutes. The temperature of 573 Kensures reduction of the noble metals while sintering does not occur [12].

After evacuation, hydrogen from the reduction treatment remains on the surface. Thisremaining hydrogen can be selectively removed by very light sputtering, since the sputter ratefor hydrogen is usually 10 to 50 times higher than for other elements [13]. From LEISmeasurements as a function of ion dose, it appeared that hydrogen was fully removed from athree-way catalyst after a He dose of 2×1015 ions/cm2 [12]. Between a He dose of2×1015 ions/cm2 and 10×1015 ions/cm2 the surface composition and the elemental LEIS yieldsof the catalyst remained unchanged. In this study 3 keV Ne ions have been used to allowseparation of Pt and Ce in the catalyst carrier. The sputter rate for Ne is typically 10 timeshigher than for He. Therefore, all presented measurements are carried out using Ne dosesbetween 0.2×1015 ions/cm2 and 1×1015 ions/cm2.


38

2.2 Low-Energy Ion Scattering (LEIS)In LEIS (also known as Ion Scattering Spectroscopy or ISS) experiments, low-energy

noble gas ions are scattered by atoms in the exposed surface. According to the laws ofconservation of energy and momentum, the energy spectrum of the backscattered ions isequivalent to the mass spectrum of the target atoms. The information depth of LEIS is limitedto one atomic layer because of the high neutralisation probability of the noble gas ions.

Earlier studies on Pt/Rh alloys have shown that the LEIS sensitivities for Pt and Rhcan be successfully calibrated using pure Pt and Rh reference samples [14-15]. Moreover,these studies showed that after prolonged sputtering, with either 2 or 3 keV Ne, the bulk ratioof the Pt/Rh alloy is obtained.

A cluster may be damaged due to the dissipated energy of an impinging ion.Molecular dynamics have shown, however, that the backscattered ion is already on its wayback to the detector before this happens [16]. LEIS analysis will therefore not be influenced,as long as the same spot is not probed more than once. Hence, LEIS measurements have to becarried out with a low ion-dose. All LEIS measurements have been performed in the UHVCalipso LEIS set-up. This set-up, which has been developed at the Eindhoven University ofTechnology, is equipped with a sensitive double-toroidal analyser and a large position-sensitive detector. These allow measurement of a large part of the energy spectrumsimultaneously [17]. The very high sensitivity enables to measure noble metal concentrationsdown to some 10 ppm of a monolayer with a relatively low dose. To spread the dose over alarger area, the primary ion beam is rastered over an area of 2 × 2 mm2 during measurements.In the Calipso LEIS set-up the primary ion beam is directed perpendicular towards the target,and ions scattered over 145° with respect to the incoming beam are detected. During theexperiments, the catalyst samples were prevented from charging by flooding with low-energyelectrons from all directions.

2.3 Transmission Electron MicroscopyTransmission electron microscopy (TEM) has been performed using a Philips CM

30 T electron microscope with a LaB6 filament as the electron source. The TEM was operatedat 300 kV. Samples were mounted on a micro-grid carbon polymer supported on a coppergrid by placing a few droplets of a suspension of ground sample in ethanol on the grid,followed by drying at ambient conditions.

Chapter 2a

3. Results

3.1 Transmission Electron MicroscopyFigure 1 shows a TEM micrograph of Pt/γ-Al2O3. Throughout the sample many small

metal clusters can be observed. Most clusters are about 0.5 nm - 3 nm in diameter with a fewclusters up to about 7.5 nm present as well. When TEM is applied to Pt/Rh/CeO2/γ-Al2O3, themetal particles are obscured from proper imaging due to their low contrast with the ceriaparticles [10].

Figure 1: TEM Micrograph of Pt/γ-Al2O3. Throughout the samobserved. Most clusters are about 0.5 – 3 nm, with a few large

3.2 Upper limit of the Pt surface atom density at the Pt/γ-AA first calculation concerns the maximum possi

attained if all platinum in the Pt/γ-Al2O3 would be mono-outermost atomic layer. Using the metal weight fraction aof the Pt/γ-Al2O3 as summarised in table 1, the maximum area is:

=

metMweightmetal

densityatomsurface

where Nav is Avogadro’s number, and Mmetal is theg/mol). The number of Pt atoms per unit surface area foselectively present in the outermost atomic layer, is thus catalyst. Using the bulk atomic density of pure Pt (6.58×6.4×1016 Pt atoms equals 9.7×10-13 m3. Assuming the (6.58×1028)2/3 = 1.6×1019 atoms/m2, the maximum Pt surf

7.5 nm

39

ple many small metal clusters can ber clusters present as well.

l2O3 surfaceble Pt surface atom density that isatomically spread and present in thend the BET specific surface area (σ)number of Pt atoms per unit surface

,Nav

σ××

al

fraction(1)

molar mass of the metal (MPt = 195r mono-atomically spread Pt that iscalculated as 6.4×1016 Pt atoms/m2

1028 atoms/m3 [19]), the volume ofsurface atom density of Pt equalsace atom density for the Pt/γ-Al2O3


40

is 6.4×1016/1.6×1019, which equals a coverage of 4.0×10-3 ML. A monolayer (ML) here isdefined as 1.6x1019 atoms/m2.

Table 1. Physical characteristics of the examined catalysts.

Pt/γ-Al2O3 Rh/γ-Al2O3 Pt/Rh/γ-Al2O3 Pt/Rh/CeO2/γ-Al2O3

Pt g/g catalyst (ICP)1) 3.98×10-3 - 3.98×10-3 3.98×10-3

Rh g/g catalyst (ICP)1) - 7.9×10-4 7.9×10-4 7.9×10-4

σ (m2/g) (BET)1) 193 193 193 157ϑPt (ML) (LEIS) (1.1±0.1)×10-3 - (0.8±0.1)×10-3 (1.0±0.1)×10-3

Relative ϑRh (LEIS) - 3.2±0.3 1.000 -Average d (nm) (LEIS) 1.3±0.2 - 2.1±0.32) 2.1±0.3

d (nm) (TEM) 0.5– 3 few 7.5 - - -1) Inductively coupled plasma (ICP) and BET analysis of the catalyst materials are described in [18]. 2) This value for d is obtained when assuming the surface composition of the clusters of the Pt/Rh/γ-Al2O3 equals that of the Pt/Rh/CeO2/γ-Al2O3.

3.3 Pt surface atom density at the Pt/γ-Al2O3 surface for reduced Pt clustersPlatinum in Pt/γ-Al2O3 is not mono-atomically spread over the outermost atomic

layer, but present in clusters. Since the Pt/γ-Al2O3 has been reduced prior to the LEISanalysis, the Pt atoms are in a metallic state. Metallic Pt atoms will tend to cluster because oftheir high surface free energy [20-22]. TEM pictures of reduced Pt on an oxidic support showspherically shaped Pt clusters, see figure 1 and reference [23]. For molten Pt droplets on anoxidic carrier, the contact angle is approximately 115º [22]. Figure 2 shows a schematic ofsuch a cluster with radius r, and contact angle α. The volume of clusters with contact angle αequals cα×π×r3, where cα is a geometric factor. Thus for α = 90º (hemisphere) and α = 180º(complete sphere), cα equals 2/3 and 4/3, respectively. Figure 3 shows cα as a function of α.

Figure 2: Schematic illustration of a supported particle with contact angle α, radius r and diameterD, adapted from Van den Oetelaar et al. [24]. In LEIS, the surface area β is projected to Avis.

αAvis

β r

d

Chapter 2a

41

Figure 3: The geometric factor cα as a funtion of the contact angle α, the volume of a sphericallyshaped cluster equals cα×π×r3.

If we define Vcluster as the total volume of all clusters on a m2 catalyst and N as thenumber of clusters present on 1 m2 catalyst support, then:

.π 3rcNVcluster ×××= α (2)

As shown in the previous section, Vcluster equals 9.7×10-13 m3 for the Pt/γ-Al2O3

catalyst. Only part of the atoms in these clusters will be visible because of shielding by theneighbouring surface atoms, because the surface is not atomically smooth [11,24]. Thesurface roughness experiments from Jansen et al. show that the effective surface coverage ofspherically shaped clusters as detected by LEIS, Avis as represented in figure 2, is given by[11]:

.π 2rRAvis ××= (3)

Where R is the surface roughness factor ranging from ~0.7 (for metals) to ~0.9 for low Znumber materials exposing relatively open surfaces. In estimating the surface roughnessfactor, the losses due to openess should not be taken into account, due to the low metalloading of the Pt/γ-Al2O3 catalyst (0.398 wt % Pt), which ensures that the Pt clusters are farapart. For the Pt clusters we assume that the surface roughness factor for Au (R = 0.79±0.07)should be used because the metals have almost the same structure, Z-number and latticeconstant.

Experimentally it was found that the Pt LEIS yield for the diluted Pt/γ-Al2O3 was only0.044±0.004% of that of a sputter cleaned pure polycrystalline Pt sample. If corrections forthe dilution of the catalyst with α-Al2O3 and the sputter induced roughness on the Pt reference

0.67

0.83

1.00

1.17

1.33

90 120 150 180

αααα

c α ααα


42

sample [25] are included, a Pt surface coverage (ϑ) corresponding to ϑ =N×Avis=(1.1±0.1)×10-3 ML is found for pure Pt/γ-Al2O3. Using this value for ϑ, the average clusterradius r of the Pt/γ-Al2O3 is given by:

,

1

2

3

R

fractionweightmetalc

rNrNr metal

ϑρσ

ππ α

×

×

=××××= (4)

where ρmetal is the volumetric density (ρPt = 21.45×106 g/m3 [19]). The average clusterdiameter (d) of the Pt clusters in Pt/γ-Al2O3 as determined from the LEIS measurementsequals d = 1.3±0.2 nm. The error margin in d is determined by the 10% error margins in bothϑ and R .

Another important parameter that can influence the validity of the average particlesize as determined with LEIS is the particle size distribution. Let us first consider an exampleof a Gaussian particle size distribution since this type of distribution occurs frequently [26].Analysing a Gaussian distribution centred round 3 nm with a full width half maximum of2nm, an average particle size of 2.7 nm would be found using LEIS. Hence, in the case of aGaussian particle size distribution LEIS gives a representative value for the average particlesize. Evaluating a distribution which is a combination of very small (e.g. 0.4 nm) and largerparticles (e.g. 6 nm), LEIS may give an underestimation of the average cluster size. Forinstance, LEIS would give an average cluster size of only 0.8 nm for a catalyst containingequal amounts of 0.4 nm clusters and 6 nm clusters. In this case, however, microscopictechniques would have failed as well, since 0.4 nm particles cannot be resolved with HREMor TEM. Hence, a combination of microscopic techniques and LEIS may give extrainformation on the particle size distribution. LEIS probes a statistical average over severalmm2 catalyst, whereas microscopic techniques probe only minute areas. Moreover, theaccuracy of LEIS increases with increasing dispersion, since LEIS selectively detects atoms inthe outermost atomic layer [27]. Therefore, very small clusters can be determined with thehighest possible accuracy, whereas microscopic methods are limited by a minimum clustersize visibility criterion.

In the case of the Pt/γ-Al2O3 catalyst TEM measurements showed cluster sizes in therange 0.5 - 3 nm with a few up to 7.5 nm. Hence, the average cluster size as determined withthe LEIS measurements (i.e. 1.3±0.2 nm) is well within the range of the TEM measurements.In the case of the Pt/Rh/CeO2/γ-Al2O3 three-way catalyst TEM is not feasible since the ceriacarrier obscures proper imaging of the metal particles [10]. Both the Pt/γ-Al2O3 and thePt/Rh/CeO2/γ-Al2O3 are synthesized in a similar way. Hence, for the latter a homogeneousparticle size distribution, where the average cluster size given by LEIS is accurate, can beexpected as well.

3.4 LEIS analysis of Pt/Rh/CeO2/γ-Al2O3

Figure 4 shows LEIS spectra of Rh/γ-Al2O3 (thin dashed line), Pt/Rh/γ-Al2O3 (thickdashed line), and the Pt/Rh/CeO2/γ-Al2O3 three-way catalyst (thick solid line) measured with

Chapter 2a

43

3 keV Ne+. Apparently, the Rh/γ-Al2O3 and the Pt/Rh/γ-Al2O3 are contaminated with Ce (~ 2weight %). The Rh LEIS peak is on the low-energy side of the huge Ce peak. The figureshows that the Rh surface coverage in the Pt/Rh/CeO2/γ-Al2O3 catalyst will be indiscerniblefrom the Ce background of the catalyst carrier. In order to enable the determination of the Rhsurface atom density and possible Pt segregation, we have also measured Pt/Rh/γ-Al2O3 andRh/γ-Al2O3. The absence of the huge Ce peak allows Rh detection in both the Pt/Rh/γ-Al2O3

and Rh/γ-Al2O3 samples. Although all catalysts contain equal Rh weight %, the Rh surfaceatom density of the Rh/γ-Al2O3 is 3.2±0.3 times higher than that of Pt/Rh/γ-Al2O3. Hence, ifthe catalysts would contain separate Pt and Rh clusters, the Rh cluster volume in Rh/γ-Al2O3

would have to be ( ) 2/32.3 = 5.7 times smaller than in Pt/Rh/γ-Al2O3. Since the Rh/γ-Al2O3 andthe Pt/Rh/γ-Al2O3 have been prepared in a similar way, a huge difference in cluster volume isnot likely. The observed difference in the Rh surface atom density can be explained, however,by mixed Pt/Rh clusters, where part of the Rh in Pt/Rh/γ-Al2O3 is covered with Pt. This isvery likely since Pt segregation has been shown for single crystalline Pt/Rh alloys of variouscompositions after annealing in vacuum or reduction at elevated temperatures, [14, 15, 28].Hence, the catalyst does not contain many separate Pt and Rh clusters, but mainly mixedPt/Rh clusters. A Pt/Pd/Rh/CeO2/Al2O3 automotive catalyst, manufactured by W.R. Grace,contained mixed noble metal clusters as well [26]. Only bi- and tri-metallic particles wereobserved, in that particular study, using analytical electron microscopy, which was feasiblebecause this catalyst contained only 2% Ce.

Figure 4: LEIS spectra of Rh/γ-Al2O3 (thin dashed line), Pt/Rh/γ-Al2O3 (thick dashed line), andPt/Rh/CeO2/γ-Al2O3 (solid line), measured with 3 keV Ne+. The intense Ce peak masks the presence ofRh in the three-way catalyst. The small Ce peak in Rh/γ-Al2O3 and Pt/Rh/γ-Al2O3 is due tocontamination. All catalysts contained equal Rh weight %.

Experimentally it was found that the Pt LEIS yield for the diluted Pt/Rh/CeO2/γ-Al2O3

catalyst was only 0.040±0.004% of that of a sputter cleaned pure polycrystalline Pt sample. Ifcorrections for the dilution with α-Al2O3, the surface roughness of the compacted powder

1300 1800 2300Energy (eV)

0

10

LEIS

Sig

nal (

Cou

nts/

nC)

Pt

Ce

Rh


44

[11] and the sputter induced roughness on the Pt reference sample [25] are applied, a Ptcoverage corresponding to ϑ = (1.0±0.1)×10-3 ML is found for the outermost atomic layer ofthe Pt/Rh/CeO2/γ-Al2O3 catalyst.

In the next section the surface Pt to Rh ratio of the catalyst will be calculated from thebulk composition using thermodynamics and LEIS data previously obtained on a singlecrystalline Pt0.25Rh0.75. Moreover, the calculated Rh depletion will be compared to themeasured depletion on Pt/Rh/γ-Al2O3.

3.5 Influence of segregation in Pt/Rh/CeO2/Al2O3 on the determination of the average clustersize

As explained above, part of the Rh in the mixed Pt/Rh clusters appeared to be coveredby Pt. To a lesser extent Pt will be covered with Rh. The amount of Rh at the surface of thePt/Rh/CeO2/γ-Al2O3 catalyst remains below the LEIS detection limit (figure 4). However, toenable determination of the average diameter of the Pt/Rh clusters, the Rh coverage has to beadded to the visible Pt fraction in the three-way catalyst. Therefore, the surface compositionwill be calculated using bulk values and thermodynamics.

The final treatment of the Pt/Rh/CeO2/γ-Al2O3 catalyst before LEIS analysis is a 10min. reduction at 573 K. Assuming the surface composition after this treatment to be equal tothe equilibrium composition at this temperature in vacuum, the following Pt : Rh ratio can beexpected for the bulk Pt0.725Rh0.275 three-way catalyst:

.exp

×∆−×

=

TRG

XX

XX

bulkRh

Pt

surfaceRh

Pt (5)

A segregation energy (∆G) of -6 kJ/mol for Pt has been determined experimentally for singlecrystalline (410) Pt0.25Rh0.75 after annealing in vacuum at 573 K [15]. Using this value, a (Pt :Rh)surface ratio of 9.28 is found for the bulk Pt0.725Rh0.275 three-way catalyst. This ratiocorresponds to Pt0.903Rh0.097, i.e. 9.7% of the average cluster surface area consists of Rh. Ifthis amount is added to the observed Pt (ϑPt = (1.0±0.1)×10-3 ML), the noble metal surfacecoverage of the Pt/Rh/CeO2/γ-Al2O3 catalyst corresponds to ϑ = (1.1±0.1)×10-3 ML. As hasbeen shown for the Pt/γ-Al2O3 catalyst, the average cluster diameter can be determined usingϑ. Using ρRh = 12.41×106 g/m3 [19] and the noble metal weight fractions and BET specificsurface area as summarised in table 1, the average cluster diameter of the Pt/Rh clusters in thePt/Rh/CeO2/Al2O3 three-way catalyst is 2.2±0.3 nm.

One should note that there can be less segregation in very small clusters than in bulksamples [29,30]. To investigate whether the segregation is limited in the three-way catalysts,the noble metal dispersion (D), i.e. the ratio of the surface atoms and the total number ofatoms, has to be determined. The inter-atomic distances in Pt0.725Rh0.275 are 0.28 nm [31],hence the average cluster diameter of 2.2±0.3 nm corresponds to approximately 1.9⋅102 atomsper cluster. When approximating the spherically shaped clusters with a cubo-octahedron

Chapter 2a

45

composed of 201 atoms, 122 of these atoms will be exposed to the surface [32]. Hence, thedispersion of the clusters is D = 0.61.

Due to the limited number of atoms in the cluster, 61% of the atoms in the noblemetal clusters are exposed to the surface, the segregation of one species leads to a depletionof that species in the bulk of the cluster, which has been neglected in equation (5). In order toenable correction for the depletion of the bulk the following relation between the surface andbulk composition should be used:

( ) ( ) ( ) ibulkisurfacei XXDXD =×−+× 1 , (6)

where Xi is the overall fraction of element i in the cluster (for the Pt/Rh/CeO2/Al2O3 three-way catalyst clusters XPt = 0.725 and XRh = 0.275). Combining equations (5) and (6) givesPt0.824Rh0.176 for the surface composition and Pt0.571Rh0.428 for the bulk composition of thePt/Rh/CeO2/Al2O3 three-way catalyst clusters. Note that the found bulk composition issignificantly different from the overall cluster composition (i.e. Pt0.725Rh0.275).

The presence of step edges, however, will increase the Pt surface enrichment. Thesurface enrichment has been calculated for Pt surface atoms that are missing 4 neighbours.However, at step edges the Pt atoms are missing up to 8 neighbours. Measurements of Moestet al. have shown the validity of the broken bond model in explaining the segregationbehaviour of Pt/Rh [15]. Therefore, ∆G is taken to be directly proportional to the number ofmissing neighbours. Van Hardeveld and Hartog predict 24 atoms missing 6 neighbours, 36atoms missing 5 neighbours, 6 atoms missing 4 neighbours, and 56 atoms missing 1neighbour in a 201 atoms sized cubo-octahedron (more spherical clusters give similar results)[32]. Using these numbers, correction for the enhanced segregation at step edges gives for thePt0.725Rh0.275 a surface composition of Pt0.870Rh0.130.

Since the formation of RhO2 is thermodynamically more favourable than that of PtO2

[19], the metal support interaction will further deplete the surface in Rh and enhance Ptsegregation. Therefore, the obtained Pt0.870Rh0.130 might still be an overestimation of theactual Rh surface percentage. If the cluster – support interface is assumed to consist entirelyof Rh oxide, due to the more favourable metal – support interaction, segregation would yielda Pt0.886Rh0.114 cluster surface.

The Pt0.886Rh0.114 corresponds to a Rh depletion of 114/275 = 0.41, which is inreasonable agreement with the experimentally determined value, which equals 0.31±0.06.Using Pt0.886Rh0.114 for the surface concentration, equation (4) gives d = 2.1±0.3 nm for theaverage cluster diameter of the Pt/Rh/CeO2/Al2O3 three-way catalyst. Hence, the agreementfound between the experimentally and thermodynamically determined Rh depletion indicatesonce more the validity of the average noble metal cluster size as determined with LEIS.

3.6 General applicability of LEIS to determine the size of clusters with different shapesIn this study the average cluster sizes of two reduced catalysts having clusters with

contact angles α = 115° have been determined. The presented method can also be used to


46

determine the size of clusters with other contact angles. Even the size of hemisphericallyshaped clusters (α = 90°), which are encountered for oxidised clusters [20, 24], can bedetermined. Therefore, figure 3 gives cθ as a function of α and, as explained in [24], thevisible fraction ϑ does not change for 90° ≤ α ≤ 180°. Hence the method can be used for allkinds of pseudo-spherically shaped clusters.

4. ConclusionsA method has been presented to determine the average cluster size of supported

catalysts using LEIS data. For Pt/γ-Al2O3 agreement has been shown between the averagecluster size as determined with this method, and with TEM. LEIS has been successfullyapplied to determine the average noble metal cluster size of a ceria supported commercialthree-way catalyst. The average noble metal cluster size of the investigated Pt/Rh/CeO2/γ-Al2O3 commercial three-way catalyst is 2.1±0.3 nm.

LEIS analysis of Pt/Rh/γ-Al2O3 and Rh/γ-Al2O3 showed on the Pt/Rh/γ-Al2O3 afterreduction at 573 K a Rh surface depletion of a factor 0.31±0.06 compared to the overallcomposition. This is in reasonable agreement with thermodynamic calculations yielding afactor of 0.41 when assuming 2.1±0.3 nm clusters that are reduced at 573 K. Hence, thepresence of a significant number of pure Rh clusters on the similarly preparedPt/Rh/CeO2/γ-Al2O3 catalyst is very unlikely. The Pt/Rh/CeO2/γ-Al2O3 catalyst, therefore,contains mixed Pt/Rh clusters that are strongly enriched in Pt after reduction. Since theaverage cluster size as determined with LEIS has been used for the thermodynamicalcalculations, the agreement found with the experimentally determined Rh depletion gives anextra indication of the validity of the determined cluster size.

The high sensitivity of LEIS for metals permits cluster size determination down tometal loadings of some 10 ppm in the outermost atomic layer of the surface. Moreover, theaccuracy of LEIS increases with increasing dispersion, since LEIS selectively detects atoms inthe outermost atomic layer. Hence, very small clusters can be determined with the highestpossible accuracy, whereas microscopic methods require a minimum cluster size. Moreover,the average cluster size as determined with LEIS reflects a statistical average over severalmm2 catalyst, whereas microscopic techniques probe only minute areas.

AcknowledgementFinancial support for the study was given by the Dutch Technology Foundation (STW). Theauthors are grateful to dmc2 A.G. (Hanau, Germany) for providing the catalysts. Dr. P.J.Kooyman of the National Centre for High Resolution Electron Microscopy (Delft, TheNetherlands) is gratefully acknowledged for performing the electron microscopyinvestigations.

Chapter 2a

47

References

[1] Balenovic, M., Backx, A.C.P.M., and Hoebink, J.H.B.J., SAE-paper 2001-01-0937[2] Hoebink, J.H.B.J., Harmsen, J.M.A., Balenovic, M., Backx, A.C.P.M., and

Schouten, J.C., preprints CAPoC5, 1, 225 (2000) to be published in Topics in Cat.(2001).

[3] Rogemond, E., Essayem, N., Frety, R., Perrichon, V., Primet, M., Chevrier, M.,Gauthier, C., and Mathis, F., J. Catal. 186(2), 414 (1999).

[4] Holles, J.H., Switzer, M.A., and Davis, R.J., J. Catal. 190(2), 247 (2000).[5] Maunula, T., Ahola, J., Salmi, T., Heikki, H., Härkönen, M., Luoma, M., and

Pohjola, V.J., Appl. Catal. B12, 287 (1997).[6] Holmgren, A., Andersson, B., and Duprez, D., Appl. Catal. B22(3), 215 (1999).[7] Kip, B.J., Duivenvoorden F.B.M., Koningsberger D.C., and Prins, R., J. Catal. 105,

26 (1987).[8] Winkelmann, F., Wohlrab, S., Libuda, J., Bäumer, M., Cappus, D., Menges, M., Al-

Shamery K., Kuhlenbeck H., and Freund H.-J., Surf. Sci. 307-309, 1148 (1994).[9] Klimenkov, M., Nepijko, S., Kuhlenbeck H., Bäumer, M., Schlögl R., and Freund

H.-J., Surf. Sci. 391, 27 (1997).[10] Fajardie, F., Tempere, J-F., Manoli, J-M., Touret, O., and Djéga-Mariadassou, G.,

Catal. Lett. 54, 187 (1998).[11] Jansen, W.P.A., Knoester, A., Maas, A.J.H., Schmit, P., Denier van der Gon, A.W.,

and Brongersma, H.H., to be published.[12] Harmsen, J.M.A., Jansen, W.P.A., Hoebink, J.H.B.J., Schouten, J.C., and

Brongersma, H.H., to be published.[13] Bergmans, R., PhD. Thesis Eindhoven University of Technology The Netherlands,

p. 70. (1996).[14] Beck, D.D., DiMaggio, C.L., and Fisher G.B., Surf. Sci. 297, 293 (1993).[15] Moest, B., Wouda, P.T., Denier van der Gon, A.W., Langelaar, M.C., Brongersma,

H.H., Nieuwenhuys, B.E., and Boerma, D.O., Surf. Sci., 473, 159 (2001).[16] Den Otter, W.K., Brongersma, H.H., and Feil, H., Surf. Sci. 306, 215 (1994).[17] Ackermans, P.A.J., van der Meulen, P.F.H.M., Ottevanger, H., van Straten, F.E. and

Brongersma, H.H., Nucl. Instr. Meth. B35, 541 (1988).[18] Campman, M., PhD. Thesis Eindhoven University of Technology, The Netherlands,

p. 38. (1996).[19] Handbook of Chemistry and Physics, (D.R. Lide Ed in Chief) p. 4-21. CRC Press

(1995).[20] Diebold, U., Pan, J.-M., and Madey, T.E., Surf. Sci. 331-333, 845 (1994).[21] Clausen, B.S., Schiøtz J., Gråbæk, L., Ovessen, C.V., Jacobsen, K.W., Nørskov,

J.K., and Topsøe, H., Topics in Cat. 1, 367 (1994).[22] Wetting and Work of Adhesion in Oxide-Metal Systems, Ceramic Microstructures,

Control at the Atomic Level, (Tomsia A.P., and Glaeser, Ed.) p. 65-82 PlenumPublishing Corporation (1998).


48

[23] Bernal, S., Calvino, J.J., Cauqui, M.A., Gatica, J.M., Larese C., Péreze Omil, J.A.,and Pintado, J.M., Catal. Today 50, 175 (1999).

[24] Van den Oetelaar, L.C.A., Partridge, A., Stapel, P.J.A., Flipse, C.F.J. andBrongersma, H.H., J. Phys. Chem. B102, 9532 (1998).

[25] R. Cortenraad, S.N. Ermolov, B. Moest, A.W. Denier van der Gon, V.G. Glebovsky,H.H. Brongersma, Nucl. Instr. Meth. B 174, 173 (2001).

[26] Kim, S., and D’Aniello, M.J. Jr., Appl. Catal. 56, 23 (1989).[27] Jacobs J.-P., Lindfors L.P., Reintjes J.G.H., Jylhä O. and Brongersma H.H., Catal.

Lett. 25, 315 (1994).[28] Beck, D.D., DiMaggio, C.L., and Fisher G.B., Surf. Sci. 297, 303 (1993).[29] Williams, F.L., and Nason, D., Surf. Sci. 45, 377 (1974).[30] Van den Oetelaar, L.C.A., Nooij, O.W., Oerlemans, S., Denier van der Gon, A.W.,

and Brongersma, H.H., J. Phys. Chem. B102, 3445 (1998).[31] Gmelin Handbuch der Anorganischen Chemie, p. 831 Springer Verlag Berlin,

(1951).[32] Hardeveld, R. van, and Hartog, F., Surf. Sci. 15, 189 (1969).

51

2bCoke deposition on automotive three-way catalysts studied with

LEIS*

AbstractIn order to study carbonaceous deposits LEIS measurements have been performed on freshand used commercial three-way catalysts (Pt/Rh/CeO2/γ-Al2O3). The catalysts showed asignificant decrease in the platinum surface concentration of approximately 50% after use inhydrocarbon oxidation at cold start conditions, and could be completely regenerated. Theselectivity towards the outermost atomic layer of LEIS allowed a one to one correlationbetween the surface Pt concentration as detected with LEIS and that obtained by kineticmodelling. This supports recent assumptions on selective deactivation. Conventional surfacescience techniques such as XPS, SIMS or AES would have yielded ambiguous results on thesurface composition since their probing depths are not limited to the outermost atomic layer.

* The contents of this chapter has previously appeared in J.M.A. Harmsen, W.P.A. Jansen,J.H.B.J. Hoebink, J.C. Schouten, and H.H. Brongersma, Catalysis Letters, 74, 133-137(2001).

LEIS study of coke on three-way catalysts

52

1. IntroductionNowadays, most cars are supplied with a monolith converter, which contains a so-

called three-way catalyst to diminish the emission of harmful components present in the carexhaust gas [1]. The catalyst converts carbon monoxide, uncombusted hydrocarbons, andoxides of nitrogen to carbon dioxide, water, and nitrogen. After a cold start, however, thetemperature of the catalyst is too low for an effective conversion (viz., T <600 K), resulting inemissions of toxic pollutants. The rates of conversion of the reactants are then completelydetermined by the intrinsic chemical reaction kinetics. Therefore, for a further optimisation ofthe exhaust gas converter and the control system, currently an air/fuel ratio based feedbackcontroller, a full understanding of the process is necessary, notably the kinetics of thereactions involved. The influence of the driver, as well as the delay of the sensor/controllersystem cause steep and frequent changes in the composition of the exhaust gas from the carengine [2]. This indicates the need for modelling of the transient kinetics. Transient kineticmodelling, based on elementary reaction steps, yields a detailed model, capable of predictingaccurately the conversions of all components over the automotive catalyst. Such a model canbe used to improve the performance of the monolith converter via model predictive control[3].

During the cold-start period, which will take generally 1 – 2 minutes, the engine isnormally run at rich conditions in order to ensure its operation. This will result in a largeexcess of carbon monoxide and unburned hydrocarbons in the exhaust gas that partly adsorbonto the catalytic surface. Partial decomposition of these adsorbed components during themonolith heat-up will lead to the presence of carbonaceous deposits on the catalytic surface,which will block active sites of the catalyst, until the deposits can be burned off. The latterwill not occur until the end of the cold-start period, at T ≈ 600K. Carbon deposits from CO onPt/CeO2 and Rh/CeO2 catalysts have been imaged by means of TEM [4], and also ethylene isnotorious for depositing carbonaceous species on noble metal catalysts [5-8]. This means thatdeposition of carbonaceous species contributes significantly to the emissions during the cold-start period. Recent kinetic studies [9,10] on ethylene and acetylene oxidation in theframework of automotive catalysis endorsed the involvement of carbonaceous species,notably the selective deactivation of noble metal sites with respect to oxygen adsorption. Theintention of this study is to provide an independent, quantitative confirmation of the decreaseof noble metal sites, as estimated from the recent kinetic studies [9,10]

A real exhaust gas contains more than hundred different hydrocarbons [11-13].Ethylene and acetylene have been chosen to represent these hydrocarbons. Ethylene, with arelatively large abundance in an exhaust gas [13] represents the hydrocarbons, which can beoxidised quite easily, among which aromatics [14,15]. Acetylene was chosen mainly becauseof both its large inhibitive effect and high concentration after a cold engine start, as discussedrecently [10].

In this study Low-Energy Ion Scattering (LEIS) has been used to investigatecarbonaceous deposits on a commercial three-way catalyst with a characteristic noble metalloading of 0.48 weight %. Using LEIS the elemental composition of the outermost atomiclayer of a catalyst can be determined in a quantitative way. Hence, if the catalyst surface is

Chapter 2b

53

partly covered with carbon this will both be detected by observing the presence of carbon andby measuring the disappearance of the underlying material. Therefore, LEIS can revealwhether the carbon is deposited on the carrier material or on the noble metals. Techniqueslike XPS, Auger Electron Spectroscopy (AES) or Secondary Ion Mass Spectroscopy (SIMS),do not provide this information since their probing depths are not limited to the outermostatomic layer which leads to ambiguous results [16]. For instance -assuming θ = 45°, Mg KαX-rays, inelastic mean free path of 3.07 nm for Pt4f in C [17], and a carbon monolayerthickness of 0.2 nm [18]- a surface that is totally covered with 1 atomic layer of carbon stillgives about 91% of the Pt XPS signal of a clean Pt sample. While a Pt surface that is only halfcovered with 2 monolayers of C would give the same 91% Pt signal. Therefore, XPS cannotbe used to determine the part of the Pt surface that is covered with C. The XPS signal of the Catoms can also not be used, since C can also be on the support. Hence, quantification of thecarbon coverage does not make sense. The same kind of ambiguity will be encountered whenusing AES or SIMS. Moreover, matrix effects will further complicate the analysis when usingSIMS. The information depth of LEIS, however, is limited to 1 atomic layer [19,20] andallows quantification of the noble metal area loss. Moreover, the high sensitivity of LEIStowards noble metals enables detection down to concentrations of some 10 ppm of amonolayer. This opens a unique possibility to quantitatively verify the kinetic modelling. Ingeneral it will thus be possible with LEIS to study the nucleation site of carbonaceousdeposits.

The noble metal surface areas of fresh and used commercial catalysts as detected withLEIS, have been compared with the corresponding values as estimated from kineticmodelling of the acetylene and ethylene oxidation at cold start conditions, to verify theoutcome of the latter [9,10].

2. Experimental

2.1. Catalyst (pre-) treatmentA commercial Pt/Rh/CeO2/γ-Al2O3 catalyst, as used for coating monoliths, has been

provided by dmc2 A.G. (Hanau, Germany) as a powder with a mean particle diameter of 12µm. The powder is pressed, crushed, and sieved to obtain a fraction with pellet diametersbetween 0.11 and 0.15 mm. To enhance isothermicity during kinetic measurements in a fixedbed reactor, 0.92 g of catalyst is diluted with 1.4 g of inert α-Al2O3 (0.15 mm – 0.21 mm).

In order to enable reproducible kinetic experiments the following pre-treatment iscarried out. The catalyst is heated to 773 K in a steady flow of He. Then the catalyst isoxidised during 1 hour by a stream containing 25 vol. % oxygen. Next the catalyst is keptunder flowing He for 30 minutes in order to purge reversibly adsorbed oxygen, followed byreduction in a He stream containing 5 vol. % H2 at 773 K for 2 hours. Finally, the catalyst isallowed to cool down to reaction temperature under a He stream.


54

2.2. Kinetic experimentsDifferent batches of the thus pre-treated Pt/Rh/CeO2/γ-Al2O3 catalyst have been used

in transient ethylene oxidation experiments at cold start conditions (393-443 K), and intransient acetylene oxidation experiments at cold start conditions (503-543 K). During thetransient experiments, the catalyst material is alternately exposed to He flows with either0.15 kPa hydrocarbons or 0.5 kPa O2 at frequencies between 0.05 and 0.25 Hz for a period of2 weeks. More detail on the transient experiments can be found in [9] (on ethylene oxidation)and [10] (on acetylene oxidation).

2.3. LEIS experimentsIn LEIS (also known as Ion Scattering Spectroscopy or ISS) experiments, low-energy

noble gas ions are scattered by atoms in the exposed surface. According to the laws ofconservation of energy and momentum, the energy spectrum of the back-scattered ions isequivalent to the mass spectrum of the target atoms. The information depth of LEIS is limitedto one atomic layer, because of the high neutralisation probability of the noble gas ions.

Prior to LEIS analysis the catalyst samples have been compacted into pellets at300 MPa. Recently, we have shown that compaction at 300 MPa does not influence thesurface composition [21]. LEIS selectively probes the outermost atomic layer, hence surfacecontaminants would obscure the intrinsic composition. Therefore, the contaminants have tobe removed before analysis of the intrinsic composition is possible. To clean the fresh, and toregenerate the used three-way catalyst samples, an atomic oxygen beam has been applied. Theuse of atomic oxygen allows very effective cleaning at low temperatures (ca. 310 K). Afterthe oxidative cleaning, the catalysts are reduced at 573 K in 20 kPa H2 flowing at2.6 mmol/min for 10 minutes.

After evacuation, hydrogen from the reduction treatment remains on the catalystsurface. This remaining hydrogen can be selectively removed by sputtering, since the sputterrate for hydrogen is usually 10 to 50 times higher than that for other elements [22]. FromLEIS measurements as a function of ion dose, it appeared that hydrogen was removed upon aHe dose of 2×1015 ions/cm2. Between a He dose of 2×1015 ions/cm2 and 10×1015 ions/cm2,the surface composition of the catalyst remained constant. The sputter rate for Ne is typically10 times higher than for He. Therefore, all presented measurements are carried out usingeither a He dose between 2×1015 ions/cm2 and 9×1015 ions/cm2 or a Ne dose between0.2×1015 ions/cm2 and 1×1015 ions/cm2.

All LEIS measurements have been performed in the UHV Calipso LEIS set-up. Thisset-up, which has been developed at the Eindhoven University of Technology, is equippedwith a sensitive double-toroidal analyser and a large position-sensitive detector. These allowto measure a large part of the energy spectrum simultaneously [23]. Since the sensitivity ofthis set-up is about a factor 1000 higher than that of conventional LEIS, it is now possible todetect as little as some 10 ppm of a monolayer Pt on a supported catalyst.

To analyse the surface composition of the three-way catalyst before and after use inthe ethylene and acetylene oxidation, 3 keV 4He ions have been used. Moreover, 3 keV Neions have been used to allow separation of Pt from the Ce in the catalyst carrier.

Chapter 2b

55

Unfortunately, Rh cannot be observed since the expected Rh surface concentration is ca. 10times lower than the Pt surface concentration because of segregation [24]. Moreover, the Rhpeak is on the low-energy side of the huge Ce peak. Hence a small Rh peak is indiscerniblefrom the Ce background of the catalyst carrier. The beam was rastered over an area of 1 × 1mm2 during measurements. In the Calipso LEIS set-up the primary ion beam is directedperpendicular towards the target, and ions scattered over 145° with respect to the incomingbeam are detected. During the experiments the catalyst samples were prevented from chargingby flooding with low-energy electrons from all directions.

3. Results

3.1. LEIS analysis of the catalyst carrierFigure 1 shows LEIS spectra of a fresh (solid line) and a used (dotted line)

Pt/Rh/CeO2/γ-Al2O3 three-way catalyst. The spectra have been measured with a 3 keV 4Hedose of 6×1015 ions/cm2, and are therefore selectively representing the outermost atomic layerof the catalyst. From this figure, it can be seen that the catalyst surface exposes not only largeamounts of Ce, O, and Al, but also a significant amount of F. Additional measurements on thesupport materials revealed that the F originates from the γ-Al2O3 support material. Like othersurface contaminants, F will obscure underlying catalyst constituents. Since the F contentdiffers from sample to sample, it hampers exact quantification of the composition of thecatalyst support. However, the Al peak area is about twice as high as the Ce peak area and Ceis much heavier and therefore more easily detected than Al. The present amount of F cannotexplain this difference, hence there is much more Al in the surface than Ce. This is inagreement with the bulk analysis of the diluted three-way catalyst being 11 weight % CeO2

and 89 weight % γ-Al2O3 [25].

Figure 1. Two LEIS spectra obtained with 3 keV 4He on a fresh (thick solid line) and a used (dottedline) three-way catalyst.

700 1700 2700Energy (eV)

0

10

20

30

LEIS

Sig

nal (

Cou

nts/

nC)

O

F

Al

Ce


56

3.2. LEIS analysis of the Pt surface area after acetylene oxidationTo allow separate determination of Pt and Ce in the LEIS spectrum, 3 keV Ne ions

have been used. Figure 2 shows the Pt LEIS signals for a fresh three-way catalyst (dottedline), the catalyst after use in the acetylene oxidation (thick solid line), and the catalyst afterregeneration following use in the acetylene oxidation (thin solid line). Prior to the LEISmeasurements, the fresh catalyst has been cleaned using atomic oxygen, whereupon it wasreduced at 573 K. As has been described in the previous section, these treatments have beencarried out to reveal the intrinsic catalyst surface. To regenerate the catalyst after use in theacetylene oxidation, the same treatment has been applied. The figure shows the Pt signaldecreased 50%±5% after use in the acetylene oxidation at cold start conditions. This decreasein Pt signal can be explained in two ways, either half of the Pt is covered by some othermaterial, or severe sintering took place during use. An eightfold volume increase of the Ptparticles due to sintering is needed to explain a 50% signal decrease. This is highlyimprobable during use at 543 K, especially since both the fresh and the used catalyst havebeen exposed to 773 K during the pre-treatment. As shown in figure 2, the Pt surfaceconcentration could be completely restored after regeneration.

Figure 2. Three LEIS spectra obtained with 3 keV Ne on a fresh (dotted line), a used (thick solid line)and a regenerated (thin solid line) three-way catalyst. The figure shows a drop in the Pt surface areaafter use in acetylene conversion at cold start conditions. After regeneration the Pt surface area isfully restored.

LEIS data are generally independent of the chemical environment (i.e. no matrixeffects) [19,20]. Therefore, the sensitivity for Pt can be calibrated against a reference samplewith a well-known Pt surface atomic density. The Pt surface concentration of the three-waycatalyst has been calibrated against a sputter cleaned pure polycrystalline Pt sample. Fromthis, it was learned that the Pt surface concentration in the fresh three-way catalyst equalsapproximately 0.040% of a monolayer. However, this concentration is obtained on a three-

1850 1950 2050 2150Energy (eV)

0.0

0.6

1.2

LEIS

Sig

nal (

Cou

nts/

nC)

Pt

Chapter 2b

57

way catalyst that has been diluted with inert α-Al2O3. If correction for the dilution is applied,a Pt surface concentration of 0.098% of a monolayer for the pure fresh three-way catalyst isfound. If 50% of this amount is covered with carbonaceous species, this cannot be detected inthe carbon LEIS signal, since the sensitivity for carbon is too low. However, the highsensitivity for Pt allows detection of the Pt surface concentration loss.

The Pt surface concentration can be entirely restored during regeneration, or cleaning.This proves that no sintering took place during use. The loss in Pt signal is due to a 50%±5%coverage of the Pt with carbonaceous species, deposited during acetylene oxidation. Thesecarbonaceous species can be effectively removed using atomic oxygen and a consecutivereduction at 573 K.

3.3. Comparison between LEIS analysis and kinetic modellingLEIS analysis reveals that three-way catalysts that have been used in ethylene

oxidation at cold start conditions show a 51%±5% decrease in Pt surface concentration. Table1 summarises the relative Pt surface concentrations found with LEIS and compares them tothose obtained with kinetic modelling by Harmsen et al. [9,10]. The experimental errormargin for LEIS analysis is mainly caused by the determination of the Pt peak area. Sinceboth peak area determinations have an accuracy of 7%, a 10% error margin is estimated forthe peak area ratio of the fresh and the used catalyst.

For the data obtained by kinetic modelling, the error estimation is more difficult,because the statistical data from transient regression analysis is unreliable, due to the fact thatsuccessive measurements in time are not independent. The kinetic modelling results revealeda 60% and a 40% reduction of the Pt surface concentration in case of acetylene and ethylene,respectively. This is of the order of magnitude of the 50% reduction found in the LEISexperiments. This value of 50% Pt surface reduction has been used in model simulations ofthe ethylene and acetylene conversion. It was found that this value of 50% could also welldescribe the experimental data. This indicates an uncertainty of approximately 0.1 for thekinetic modelling results in the normalised Pt surface concentration. This error margin isgiven in table 1. Consequently, the observed 50±5% or 51±5% decrease in Pt surfaceconcentration after use in respectively acetylene or ethylene oxidation, as found with LEIS,agree quantitatively with the noble metal area loss found in kinetic modelling.

Table 1. The Pt surface concentration derived from LEIS and Kinetic modelling normalised to thoseof a fresh three-way catalyst.

LEIS Kinetic ModellingFresh three-way catalyst 1 1

Three-way catalyst used inacetylene conversion

0.50±0.05 0.4±0.1

Three-way catalyst used inethylene conversion

0.49±0.05 0.6±0.1

Regenerated three-way catalyst 1.1±0.1 -


58

4. ConclusionsLEIS analysis shows a Pt surface concentration of about 0.098% of a monolayer in a

fresh commercial three-way catalyst. After use in either acetylene or ethylene oxidation atcold start conditions, a decrease of 50±5% of the Pt surface concentration is observed withLEIS. The Pt surface concentration can be completely restored by exposing the used catalystto atomic oxygen and a subsequent reduction at 573 K. Therefore, the observed Pt loss is notdue to sintering, but due to deposition of carbonaceous species during the hydrocarbonoxidation. The observed loss in Pt surface area with LEIS agrees quantitatively with the noblemetal area loss as found in kinetic modelling of acetylene and ethylene oxidation, being60±10% and 40±10%, respectively [9,10]. As such, the current results support earlierassumptions on selective catalyst deactivation with respect to the oxygen adsorption. Hence,this study shows that LEIS can be successfully applied to quantitatively investigatecarbonaceous deposits on commercial supported three-way catalysts. This kind of informationis believed to be unique for surface analysis.

AcknowledgementFinancial support for the study was given by the Dutch Technology Foundation (STW). Theauthors are grateful to dmc2 A.G. (Hanau, Germany) for providing the catalyst.

References[1] M. Shelef, R.W. McCabe, Catal. Today 62 (2000) 35.[2] T. Kirchner, G. Eigenberger, Chem. Eng. Sci., 51 (1996) 2409.[3] M. Balenovic, A.J.L. Nievergeld, J.H.B.J. Hoebink, A.C.P.M. Backx, SAE-paper

1999-01-3623.[4] S. Bernal, J.J. Calvino, M.A. Cauqui, J.M. Gatica, C. Larese, J.A. Perez Omil, J.M.

Pinado, Catal. Today 50 (1999) 175.[5] P. Berlowitz, C. Merigis, J.B. Butt, H.H. Kung, Langmuir 1 (1985) 206.[6] R.I. Kvon, E.A. Ivanov, A.I. Boronin, React. Kinet. Catal. Lett. 65 (1998) 381.[7] R.I. Kvon, A.I. Boronin, S.K. Shaikhutdinov, R.A. Buyanov, Appl. Surf. Sci. 120

(1997) 239.[8] L. Kepinski, Catal. Today 50 (1997) 237.[9] J.M.A. Harmsen, J.H.B.J. Hoebink, J.C. Schouten, Ind. Eng. Chem. Res. 39 (2000)

599.[10] J.M.A. Harmsen, J.H.B.J. Hoebink, J.C. Schouten, Chem. Eng. Sci., 56(6) (2001)

2019.[11] S. Kubo, Y. Yamamoto, S. Kizaki, S. Yamazaki, T. Tanaka, K. Nakanishi, SAE paper

932706, 1993[12] M.C. Drake, R.M. Sinkevitch, A.A. Quader, K.L. Olson, Th.J. Chapaton, SAE paper

962075, 1996[13] R. Impens, Stud. Surf. Sc. Catal., 30 (1987) 11.[14] B. Mabilon, D. Durand, and Ph. Courty, Stud. Surf. Sc. Catal., 96 (1995) 181.[15] J.M. Bart, A. Pentenero, M.F. Prigent, ACS Symposium Series, 495 (1992) 42.

Chapter 2b

59

[16] A.G. Sault, Catal. Lett., 29 (1994) 145.[17] S. Tanuma, C.J. Powell, D.R. Penn, Surf. Interface Anal., 11 (1988) 577.[18] M.P. Seah, W.A. Dench, Surf. Interface Anal. 1 (1979) 2.[19] H.H. Brongersma, P.A.C. Groenen, J.-P. Jacobs, in Science of Ceramic Interfaces II,

Ed. J. Nowotny (Elsevier 1994) p. 113.[20] E. Taglauer, in Fundamental Aspects of Heterogeneous Catalysis Studied by Particle

Beams, Eds. H.H. Brongersma and R.A. van Santen, NATO ASI B 265 (Plenum Press1991) p. 301.

[21] W.P.A. Jansen, A. Knoester, A.J.H. Maas, P. Schmit, A.W. Denier van der Gon, H.H.Brongersma, to be submitted

[22] R. Bergmans, PhD Thesis, Eindhoven University of Technology The Netherlands,(1996) p. 70

[23] P.A.J. Ackermans, P.F.H.M. van der Meulen, H. Ottevanger, F.E. van Straten, H.H.Brongersma, Nucl. Instr. Meth. B35 (1988) 541.

[24] W.P.A. Jansen, J.M.A. Harmsen, J.H.B.J. Hoebink, J.C. Schouten, and H.H.Brongersma, submitted for publication

[25] M. Campman, PhD Thesis, Eindhoven University of Technology The Netherlands,(1996) p. 38

61

3New Insights into the Nature of the Active Phase of VPO

Catalysts-a Quantitative Static LEIS Study-*

AbstractThe selectivity of currently used VPO catalysts towards maleic anhydride is 58% and onlyafter 2⋅103 h optimum activity is reached. To improve this knowledge on the active phase ishighly desirable. Therefore, the atomic composition of the outermost atomic layer ofequilibrated VPO catalysts was studied with low-energy ion scattering (LEIS). Using externalreference samples, absolute numbers for the atomic densities of both P and V could bedetermined. VPO catalysts prepared in either aqueous or organic media showed largeamounts of carbonaceous surface species. From catalysts prepared in an aqueous medium thecarbonaceous species could be removed using a mild oxidation treatment withoutsignificantly changing the average valence of the vanadium atoms in the catalyst as evidencedfrom XPS. After this cleaning treatment the surface concentration of vanadium (at/cm2)agrees well with that expected for a vanadyl pyrophosphate structure. However, the surfacephosphorous concentration is twice as high as that in vanadyl pyrophosphate, leading to asurface P/V ratio of 2.0±0.2. This shows that VPO catalysts may be terminated by a distortedvanadyl pyrophosphate structure, where the excess amount of phosphorous is positionedbetween the vanadyl units and the phosphate groups. The catalyst surface is also compared toVOPO4 phases. A significant contribution of a phase such as αΙΙ−VOPO4 could explain theobserved surface P/V ratio of 2.0.

* The contents of this chapter has previously appeared in W.P.A. Jansen, M. Ruitenbeek,A.W. Denier van der Gon, J.W. Geus and H.H. Brongersma, Journal of Catalysis, 196, 379-387 (2000).

New insights into the active phase of VPO catalysts

62

IntroductionVanadium Phosphorous Oxide (VPO) catalysts are commercially applied for the

selective oxidation of n-butane to maleic anhydride. Maleic anhydride and its derivativesmaleic acid and fumaric acid are produced with an annual worldwide capacity of almost 1⋅106

tons [1,2]. The average annual demand growth is estimated to be about 4% [3]. Thesenumbers together with the fact that the current yield to maleic anhydride of an equilibratedcatalyst amounts to about 58% indicate that improvement of the process is of great economicand environmental interest. These improvements require a thorough knowledge of thecatalytic and structural properties of the active V-P-O phase. The bulk of VPO catalysts isassumed to consist of vanadyl pyrophosphate, (VO)2P2O7, which has a stochiometric P/Vratio of 1. It is generally mentioned in the literature, however, that a prerequisite for optimumcatalytic performance of VPO catalysts is the addition of an overstoichiometric amount ofphosphate during the preparation. The best catalytic performance has been reported forcatalysts of a preparative P/V ratio of 1.1 [2]. Assuming that the bulk indeed consists ofstoichiometric vanadyl pyrophosphate, the excess phosphate must either be present at thesurface or be poorly bound to the surface and removed during the synthesis procedure. Toinvestigate the catalyst surface, several groups have employed XPS. However, univocalresults have not been obtained [4-14]. Generally, the experimental P/V ratio ranges from 1.5to 3, which is much higher than the stoichiometric bulk value of vanadyl phosphates, i.e. 1.0.

Coulston and co-workers have calculated the dependency of the P/V ratio on themorphology of vanadyl pyrophosphate [14]. Their results indicate that exposure of differentcrystallographic planes of vanadyl pyrophosphate will result in maximum P/V ratios: 1.015(021 plane), 0.986 (001 plane), 1.098 (100 plane, phosphate termination), and 1.136 (100plane, pyrophosphate termination) [14]. These numbers indicate that only a limited amount ofexcess phosphate can be bound at the surface of vanadyl pyrophosphate.

Delichere et al. have argued that the high experimental P/V ratios are the result ofvanadium vacancies in the (amorphous) surface layers of vanadyl pyrophosphate [13]. Forthis reason, they proposed a model in which the composition of the VPO catalysts isdescribed by the general formula [(VIV

(1-5x)VV4x◊x)O]2P2O7. In this formula ◊ is a vanadium

vacancy. Nevertheless, this is not sufficient to explain the experimental P/V ratios of 1.5-3,and other effects have to be taken into account as well.

More phosphate could be present at the surface when different surface VPO phasesare formed. Morishige et al. proposed that the high experimentally observed P/V ratio is theresult of the presence of an amorphous metaphosphate phase on the surface [10]. VO(PO3)2

would be formed upon thermal pre-treatment of VO(H2PO4)2, present in the VO(HPO4)⋅0.5H2O precursor as a contaminant [7,10]. Sananes and co-workers have shown that this poorlyactive VO(PO3)2 phase has a surface P/V ratio of 4.0, which is much higher than thegenerally observed values for vanadium pyrophosphates [15,16].

The quantitative XPS analysis [4-14] makes use of the photo-ionization cross-sections, which have been calculated for a very large number of metal oxides by Scofield[17]. These values have proven, however, to be inadequate for the analysis of VPO XPS data[12-14]. Therefore, several authors determined experimental sensitivity factors for vanadium

Chapter 3

63

and phosphorous using VPO glasses [12,18-20]. For these compounds, it is assumed that thesurface P/V ratio is the same as that of the bulk. With this approach, experimental surfaceP/V ratios of 1.1 to 1.4 were determined for bulk VPO catalysts [12,18-20]. Since the surfaceP/V ratio of VPO glasses may also be the result of phosphorous segregation from the bulk,the assumption that the surface P/V ratio is equal to the bulk P/V ratio is not obvious for thesematerials.

To overcome this problem, Coulston et al. have used five organometallic complexescontaining vanadium and phosphorous atoms as standard materials [14]. Phosphoroussegregation is not expected in these clusters, however, it has not been proven that the ligandsare stable under UHV conditions. Nevertheless, an experimental P/V ratio of 1.1 was foundfor a bulk VPO catalyst with 10% excess phosphorous [14]. This amount is in agreement withthe predictions of Coulston et al., based on a model of phosphate groups terminating thevanadyl pyrophosphate surfaces [14].

Since the surface P/V ratios that have been determined with XPS are neither univocalnor selectively dealing with the outermost surface layer, we have applied static low-energyion scattering (LEIS) to establish the surface composition of bulk VPO catalysts. In contrastto XPS, which has a penetration depth of typically 10 monolayers, LEIS is a technique thatselectively probes the outermost atomic layer of a (catalyst) surface. This offers theopportunity to study the surface P/V ratio VPO catalysts more precisely than with XPS, whichaverages out the surface and subsurface composition of the samples.

Three earlier LEIS studies of VPO catalysts have been reported [12,13,49]. UsingLEIS, Richter et al. determined a surface P/V ratio of ~2-3 [12,49]. However, theirquantification is based on comparison with VPO glasses, and therefore may have beendistorted because of possible phosphorous segregation in these glasses, as mentioned above.Moreover, a very high ion current has been applied in this study (2700 nA/cm2); hence, thecatalyst surface may have been damaged. Preferential sputtering can change the surfacecomposition.

A second LEIS study of VPO catalysts has been reported by Delichere et al., whofound a surface P/V ratio of 2.4 [13]. However, they suggested this value could be anoverestimation since carbon, which they observed with XPS, may have preferentially maskedvanadium. Indeed, we also found large amounts of carbon deposits on the VPO catalysts.Using a mild oxidation treatment in air, without significantly changing the average valence,we were able to remove the carbonaceous species from VPO catalysts prepared in an aqueousmedium. A comparison of contaminated and cleaned VPO showed that carbon is indeedpreferentially masking vanadium. Therefore, carbon has to be removed before the intrinsicsurface can be analyzed. In Eindhoven, the ERISS analyzer is used, which is orders ofmagnitude more sensitive than conventional LEIS equipment. Hence, we could perform ouranalyses in a static mode, thus without damaging the surface. In addition, we appliedcalibration of the elemental sensitivity for both phosphorous and vanadium against externalreference samples. This allowed a quantitative analysis of the surface of a cleaned VPOcatalyst and determination of the surface P/V ratio.Methods


64

Catalyst preparationBulk VPO was prepared in a mixture of organic solvents following a procedure

described by Katsumoto et al. [21]. 15 gram of V2O5 was reduced for 16 hours at 393 K in 60ml of 1:1 (v/v) i-butanol/cyclohexanol mixture. After the mixture was cooled to roomtemperature, 21 g of 85% o-H3PO4 in 30 ml i-butanol was added to the dark greensuspension, resulting in a P/V ratio of 1.1. After this mixture was refluxed for 6 hours ablue/green suspension was obtained, which was filtered and dried in a nitrogen flow for 12hours at 398 K.

In another method the bulk VPO was prepared in an aqueous medium following aprocedure described by Centi et al. [22]. After reduction of 15 g V2O5 for 16 hours in boilingconcentrated HCl (37%), 85% o-H3PO4 was added to the dark blue V(IV) solution until a P/Vratio of 1.1 was established. After 2 more hours of refluxing, the solution was concentrated byevaporation of the water. A blue, viscous syrup was obtained, which was dried in a nitrogenflow for 10 hours at 393 K.

XRD showed that both preparation methods resulted in the formation of the samehemihydrate precursor phase.

Finally, the VPO catalysts were obtained by thermal pre-treatment of the hemihydratephase in a micro reactor (50 ml/min N2, 723 K, 16 hours). All catalysts were tested andequilibrated in the n-butane oxidation reaction for more than 100 hours. They will be referredto as VPO-org and VPO-aq for the first and second of the above mentioned preparationmethods respectively.

Both VPO-org and VPO-aq show a selectivity of about 60% towards maleic anhydrideover a broad conversion range. However, VPO-org is more active than VPO-aq (conversion45% versus 17% respectively at 723 K, [37]. The difference in conversion can be explainedby the difference in specific surface area. The specific surface area (BET method) wasdetermined to be 9 m2/g for VPO-org and, 3 m2/g for VPO-aq, respectively [37].

Sample pre-treatmentBefore analysis the VPO powders were pressed (312 MPa) into stainless steel cups.

Measurements comparing powders to pressed powders showed no influence of pressing onthe surface composition.

In the UHV chamber of the ERISS LEIS set-up samples can be heated in-situ up to1250 K, and the temperatures higher than 430 K can be measured within 3 K accuracy withan Impac IP 120 pyrometer. A loadlock allows treatments of samples in various gases up to1 bar at temperatures between room temperature and 773 K without exposure to air. In theloadlock a chromel/alumel-couple is used to measure the temperature of the sample. Aftercarrying out the treatment(s), the samples are evacuated and inserted in the UHV were LEIScan be applied.

Chapter 3

65

LEISLow Energy Ion Scattering (LEIS) has been used to characterize the surface

composition of the catalysts. In LEIS a beam of mono-energetic noble gas ions is scattered byatoms of the target surface [23,24]. According to the laws of conservation of energy andmomentum, the energy of the backscattered ions is characteristic for the mass of the atomsfrom which they scattered. Hence, the energy spectrum of the backscattered ions can beinterpreted as the mass spectrum of the surface atoms. The information depth of LEIS islimited to one monolayer, since ions penetrating the outermost layer, will be almost allneutralized and, therefore not detected. As a consequence of the ion bombardment, atoms aresputtered from the surface. This allows for the determination of the in-depth distribution ofthe atoms by measuring during prolonged sputtering. However, intrinsic information on theoutermost surface layer should be obtained before sputtering has damaged this layer, thususing a low dose.

The LEIS experiments were carried out in the ERISS set-up (Energy Resolved IonScattering Spectrometer). This ultra high vacuum set-up has a base pressure in the low 10-10

mbar range which increases to 10-9-10-8 mbar during LEIS experiments due to noble gasinflux. It is equipped with an ion source (Leybold type IQE-12/38) and an electrostaticanalyzer. The primary ions are mass selected, focussed and directed perpendicular to thetarget. Ions scattered over 145o are accepted by the analyzer. The analyzer is a double-toroidalelectrostatic analyzer, similar to that of the EARISS set-up, which has been described in moredetail elsewhere [25-27]. This analyzer makes a very efficient use of the backscatteredparticles by measuring simultaneously a considerable part of the energy spectrum of thebackscattered particles and 320o of the azimuthal range. In addition, the sample manipulatorcan be scanned under the ion beam during measurements, effectively increasing the beamspot and thus reducing the ion dose even more. The EARISS-type analyzer together withscanning measurements, allows us to perform static LEIS, i.e. with negligible damage.Measurements can be carried out using only 1013-1014 ions/cm2 while typically a 3He dose of1016 ions/cm2 is required to remove one monolayer. Because 3He has the lowest sputter rate,we have used 3 keV 3He for this study. To prevent charging of insulating samples, such asVPO, low- energetic electrons (10 eV) can be flooded from all azimuths over the sampleduring measurements.

CalibrationSince in LEIS the scattering events can be considered as binary collisions between the

impinging ions and the target surface atoms, LEIS data are generally independent of thechemical environment of the scattering element (no matrix effects). This allows calibration ofthe sensitivity for a certain element against external reference samples with a well-knownsurface atomic density.

The sensitivity for V was calibrated against a high-purity polycrystalline vanadiummetal sample. The bulk of this vanadium sample was very pure since it was grown via thezone-melting method. Moreover, it was annealed at 1850 K for 2 hours in order to removecontamination. Then, the sample was inserted into the UHV of the ERISS and sputter cleaned


66

at room temperature. The signal obtained upon sputter cleaning was used as the referencesignal. Measurement of the complete LEIS spectrum proved the vanadium surface was indeedclean. Assuming an atomic radius of 0.135 nm [35] and exposure of the most densely packedsurface plane [36], the measured vanadium signal should represent 1.43⋅1015 atoms/cm2.However, from experience with tungsten we know the signal drops to 75% upon sputtercleaning due to sputter-induced roughness. Therefore, the V signal is assumed to correspondto 0.75×1.43⋅1015 atoms/cm2; see Table 1.

In the case of P we could not use the pure element for calibration because it is notstable in vacuum. The well-defined GaP (110) surface, on the other hand, is stable in vacuumand is restored upon heating to 873 K [28,29]. In GaP (110), P atoms in the outermost layerare not shielded since they protrude somewhat from the surface [29]. Therefore, the P signalwas calibrated against GaP (110). A set-up combining low-energy electron diffraction(LEED) and LEIS was used to confirm that sputter/anneal cycles produced a clean GaP (110)surface. This set-up was described in detail by Cortenraad et al. [30]. After several sputtercycles followed by 7 minutes of annealing at 873 K a sharp 1 x 1 LEED pattern, characteristicfor GaP (110), was obtained. In-situ LEIS measurements proved the surface to be clean.Subsequently, we used the same treatment in ERISS and measured the LEIS signal of GaP(110) for calibration of the elemental sensitivity for P. Using a unit mesh of 5.405 Å times3.864 Å, we calculated a phosphorous density of 4.76⋅1014 atoms/cm2 for the GaP (110)surface [29].

In addition to polycrystalline vanadium and GaP (110), we used V2O5 powder as areference for the oxidic shielding. The vanadyl groups in vanadyl pyrophosphate and V2O5

have a similar structure [31,39]. In both cases the vanadium atoms are octahedrallysurrounded by oxygen. The oxygen atoms therefore partly mask vanadium atoms.Comparison of the vanadium signal in VPO to that in V2O5 makes it possible to determinewhether the vanadium signal is consistent with a vanadyl pyrophosphate structure. The V2O5

powder was pressed in a stainless steel cup (312 MPa). To dehydrate and clean the surface,the V2O5 was oxidized (10 minutes, 200 mbar O2 at 773 K) prior to LEIS analysis. Thespecific surface area of the V2O5 powder was 20 m2/g; hence, the difference in surfaceroughness compared to VPO can be neglected.

Table 1: Absolute atomic densities (in 1015 at/cm2) in the outermost atomic layer for VPO-aq andreference samples. All data were obtained with a 3He dose of 0.1⋅1015 ions/cm2. VPO-aq was cleanedby a 1 hour oxidation treatment in 200 mbar O2 at 573 K.

V poly-crystalline

V2O5 GaP (110) Equilibrated VPO-aqafter transport

through air

EquilibratedVPO-aqcleaned

P - - 0.476 0.029 0.17V 1.09 0.19 - 0.0099 0.086P/V - - - 2.9±0.2 2.0±0.2

Chapter 3

67

To analyze the LEIS spectra, we have used a linear background subtraction and simplyintegrated the peak content. For each element, the energy range that was used in theintegration was kept constant for all spectra.

XPSXPS spectra were obtained with a VG Clam 2 spectrometer equipped with a Mg Kα

source and a hemispherical analyser connected to a single-channel detector. Measurementswere performed at 20 eV pass energy. The background pressure during data acquisition waskept below 10-8 mbar. The samples were crushed and mounted on an iron stub carrying anindium film.

The average valence of the vanadium in the catalyst can be calculated from XPSmeasurements according to the empirical formula reported by Coulston et al. [14]:

AV = 13.82 – 0.68⋅[O(1s)-V(2p3/2)] (1)

In this formula, O(1s) and V(2p3/2) are the absolute values of the electron binding energies. Adifference of 0.1 eV in the peak positions already results in a difference in the AV values of0.07. In general, the resolution of an XPS spectrometer amounts to 0.25 eV and, hence, theuncertainty in the AV is ±0.2.

ResultsLEIS analysis VPO

LEIS measurements of equilibrated VPO-org and VPO-aq catalysts, which weretransported through air after catalytic tests, show clear surface peaks corresponding tovanadium, phosphorous and oxygen, see filled circles in fig. 1. This proves all three elementsare present in the outermost surface of VPO (applied 3He dose 7⋅1013 ions/cm2). Moreover acarbon peak can be seen; since the sensitivity of LEIS decreases with decreasing mass, thispeak corresponds to a surface that is for more than 80% covered by carbon, as describedbelow. Carbon might preferentially mask either vanadium or phosphorous and has to beremoved to measure the intrinsic surface P/V ratio.

Surface contamination can be removed either by sputtering or by exposing the surfaceto a cleaning treatment. Since the surface composition may change due to preferentialsputtering, a cleaning treatment is preferred. Of course, one has to be aware of the fact thatthe surface may also be changed during a cleaning treatment. Therefore, the catalyst was atfirst only heated in He up to 573 K; however, this treatment did not remove the carbonaceousspecies. This is in agreement with Delichere et al., reporting no significant changes incatalytic activity of VPO upon heating [13]. Because thermal treatment failed in removingcarbonaceous species, the catalysts were heated in oxygen (1 hour in 200 mbar O2 at 573 K).This temperature was chosen to stay below 623 K, where XRD measurements indicated aphase transition for VPO in air [33]. Because of the mild treatment and the small amount ofcatalyst (typically 10-2 g per sample), heat effects are not expected to play an important role.To check experimentally whether the oxidation treatment did not change the oxidation state


68

of the catalyst, its average valence (AV) was determined before and after the treatment. UsingXPS, we found for equilibrated VPO-org and VPO-aq catalysts, which were transportedthrough air after catalytic tests, values ranging from 4.0±0.2 to 4.1±0.2, respectively. This isin agreement with the literature where AV-values of 4.02 to 4.3 have been reported for VPOcatalysts [34]. Upon oxidation treatment in 200 mbar O2 at 573 K, we measured an AV of4.3±0.2. Hence, the differences in AV are within the experimental error of the XPSmeasurements. The average valence state of the surface vanadium atoms is not significantlychanged upon the oxidation treatment.

Figure 1: LEIS spectra of VPO-aq (dose: 1⋅1014 ions/cm2 3 keV 3He). Comparison of VPO-aq beforecleaning (filled circles) and after 1 hour 200 mbar O2 at 573 K (open circles) shows both V, P and Oare present in the outermost atomic layer of VPO. Due to the removal of carbonaceous compoundsthe V, P and O contribution dramatically increase upon cleaning.

The LEIS spectrum, indicated by the open circles in Figure 1, shows no morecarbonaceous species are detected on VPO-aq catalysts after an oxidation treatment in200 mbar O2 at 573 K for 1 hour. LEIS spectra of VPO-org catalysts, on the other hand, stillshow an unknown amount of carbon on the surface upon this treatment. Since carbon couldonly be removed from the VPO-aq catalyst, we continued our study with this catalyst.

Figure 1 shows LEIS spectra of VPO-aq before (filled circles) and after (open circles)cleaning. From this figure it can be seen that the V, P, and O signals strongly increased uponthe cleaning treatment. However, the vanadium signal increased most dramatically.

Removal of carbonaceous species resulting in a recovery of active sites might explainthe beneficial effects of a mild oxidation treatment on the catalytic properties of VPOcatalysts as reported by Contractor et al. [44]. They observed enhanced activity and surfacearea without loss of selectivity upon calcination of VPO. Calcination is patented for both

0

15

30

45

700 1600 2500

Energy [eV]

LEIS

sig

nal [

cnts

./nC

] C

O

PV

Chapter 3

69

regeneration and activation of VPO catalysts for pressure, temperature, and contact timeranges comprising the conditions used in this study [44,46]. Table 1 shows that only 15% ofthe vanadium and phosphorous atomic sites is visible to LEIS before cleaning. Because thedata in this table have been collected with a 3He dose of only 1⋅1014 ions/cm2 they selectivelyrepresent the outermost atomic layer. The atomic densities of the powder samples have beencorrected for their surface roughness compared to those of the flat polycrystalline V and GaP(110) reference samples as explained before. Because VPO is a powder, part of the LEISsignal is lost due to physical screening. Recent measurements show that the LEIS signal ratioof flat metal samples and metal clusters on pressed powders is 1:0.79. There is no differencein the LEIS signal for pressed powders with different specific surface areas ranging from 50to 380 m2/g [32]. Both P and V are calibrated against a flat reference sample; hence, thesurface roughness correction factor (1.3) is equal for P and V. As a consequence, thedetermination of the P/V ratio is not influenced. Because the VPO-aq samples have a specificsurface area of 3 m2/g, the surface roughness correction factor is equal for different catalystsamples; hence, differences in P and V atomic densities between different catalyst samplesare not affected either.

When the data are corrected for differences in surface roughness, the intrinsicphosphorous surface density in VPO-aq is 1.7⋅1014 atoms/cm2 and for vanadium 0.86⋅1014

atoms/cm2 which gives a surface P/V ratio of 2.0±0.2. Upon an oxidation treatment in 200mbar O2 at 573 K, the phosphorous and vanadium signal increased six- and nine-fold,respectively. The relative increase for vanadium is higher than that for phosphorous, whichshows that carbon preferentially masks vanadium, as suggested by Delichere et al. [13].Therefore, determination of the surface P/V ratio on a contaminated catalyst yields a too highvalue. The surface P/V of our contaminated VPO-aq catalyst was 2.9±0.2. Carbon deposits,preferentially masking vanadium, explain this high P/V ratio.

According to TPR experiments, VPO reduction in 100 mbar H2 starts at temperatureshigher than 623 K [37]. Since we wanted to clean the VPO without reducing it, VPO-aq hasbeen treated in hydrogen (1 hour, 200 mbar H2 at 573 K). Upon this treatment a low atomicdensity of P and V is found. This can be partly explained by hydrogen adsorbed on thecatalyst during the reductive treatment. The adsorbed hydrogen may shield the V and/or P inthe catalyst. Upon a 3He dose of 1-1.5 ⋅1015 ions/cm2, the P and V atomic density are 52 %and 36 % of those in the cleaned sample. At such doses hydrogen is largely sputtered away,since it has typically a 10-50 times higher sputter-rate than other elements [38]. At low doses,when hydrogen still partly covers the catalyst, a P/V ratio of 3.6±0.2 was found. Upon a 3Hedose of 1.5⋅1015 ions/cm2, the P/V ratio decreased to 2.6±0.2. Probably a considerable part ofthe catalyst surface is still covered with contamination upon our reduction treatment. WhenVPO-aq is reduced upon calcination, initially about 70% of the atomic density is found.Again, the P and V atomic densities increase upon low dose, which indicates hydrogencoverage. Upon removal of hydrogen by sputtering, the atomic densities and P/V ratio tend tothose of catalysts that were only calcined. The observed surface composition is thus similarupon our calcining treatment (1 hour 200 mbar O2 at 573 K) and upon this treatment followed


70

by a reductive treatment (1 hour 200 mbar H2 at 573 K). This indicates the treatments have nostrong influence on the surface composition.

Surface structure of VPOThe vanadyl groups in vanadyl pyrophosphate and V2O5 have similar primary

structure. Both in vanadyl pyrophosphate and V2O5 the vanadium is octahedrally surroundedby oxygen atoms [31,39]. The nearest neighbour oxygen atoms partly shield vanadium atoms.For instance, 50 % of the surface vanadyl groups is terminated by oxygen atoms (transconfiguration of the vanadyl dimers). Therefore, vanadium atoms in these groups are neitheraccessible for catalysis nor visible to LEIS. Because of the similar nearest neighbourscomparison of the vanadium signal in VPO to that in V2O5 makes it possible to determinewhether the vanadium signal is consistent with a vanadyl pyrophosphate structure. Usinglattice parameters of V2O5 we calculated a total vanadium atomic density of 9.8⋅1014 at/cm2 inV2O5 for the c-plane (110) [31]. Single crystal data of vanadyl pyrophosphate give avanadium atomic density of 4.6⋅1014 at/cm2 for the c-plane (110) [39]. Since the nearestneighbours and therefore the chemical accessibility and the visibility for LEIS of thevanadium atoms are similar in vanadyl pyrophosphate and V2O5, the vanadium signal ofvanadyl pyrophosphate should correspond to 47% of that in V2O5; thus assuming thatphosphates do not shield the vanadium because they are not nearest neighbours.

The vanadium signal of VPO-aq (after 1 hour in 200 mbar O2 at 573 K) equaled 44%of that in V2O5 (see also fig. 2); hence, assuming a vanadyl pyrophosphate structure we miss3% of the vanadium atoms in the outermost layer. The 3% difference is hardly significant butcould be explained by vacancies, as suggested by Delichere et al. [13]. However, the missingvanadium may also be shielded by strongly bound carbonaceous species, which are notremoved from VPO-aq during the cleaning treatment (200 mbar oxygen at 573 K for 1 hour).A third explanation for the missing 3% vanadium atoms is coverage by excess phosphorous.

While the surface concentration of vanadium is in reasonable agreement with thatexpected for vanadyl pyrophosphate, the situation for phosphorous is completely different.We find 8.9⋅1014 at/cm2 of phosphorous if we assume a similar oxygen shielding forvanadium and phosphorous. The actual oxygen shielding of phosphorous in a vanadylpyrophosphate structure is expected to be rather higher than lower in comparison to that ofvanadium. Since the phosphorous and vanadium atomic densities are equal in vanadylpyrophosphate, only 4.6⋅1014 at/cm2 phosphorous are expected. Hence, there is about a factorof 2 more phosphorous than that in the surface of the ideal vanadyl pyrophosphate structure.Both the surface P/V ratio and the phosphorous atomic density thus disagree with a VPOcatalyst that is terminated with a vanadyl pyrophosphate structure.

However, one could imagine a vanadyl pyrophosphate structure having an extraamount of phosphates on top. LEIS analysis showed that, in VPO, in comparison to vanadylpyrophosphate, only 3% of the vanadium atoms exposed to the surface is missing. Hence, thehuge excess amount of phosphorous cannot reside on top of vanadium oxide groups,exposing their vanadium atoms to the surface. In that case the phosphorous atoms wouldshield about all the vanadium atoms.

Chapter 3

71

Figure 2: LEIS spectrum of cleaned VPO-aq (open circles) compared to that of V2O5 (solidline). In order to scale the vanadium signals the V2O5 has been multiplied with 0.44. The vanadiumatomic density in V2O5 is 9.8⋅1014 ions/cm2, hence 44% of this signal corresponds to 4.3⋅1014 ions/cm2

of octahedral surrounded V atoms on the catalyst surface.

If the extra amount of phosphorous would be on top of other phosphorous atoms, thephosphorous atoms would shield each other. Hence, the high phosphorous atomic density asdetermined with LEIS cannot be explained by stacking phosphorous on top of phosphorous inthe outermost surface layer. Furthermore, oxygen-terminated vanadium oxide and phosphategroups, which are already protruding from the surface, are unlikely positions to incorporatethe excess amount of phosphorous. Hence, the best places that can account for the excessamount of phosphorous are probably in between the vanadyl octahedrons and phosphategroups. Since this structure is very open, it leaves about 43% of the c-plane (110) of thevanadyl pyrophosphate structure for the excess amount of phosphorous. The incorporation ofextra phosphates in the vacancies would highly distort the vanadyl pyrophosphate structure.

This would be in agreement with previous studies that already indicated the vanadylpyrophosphate structure is not continued at the surface of VPO catalysts [40,41,45]. In anESR study Ruitenbeek et al. found evidence for the presence of an amorphous VPO phase[40]. HREM measurements of Guliants et al. showed a 15 Å amorphous overlayerterminating the VPO catalyst surface [41]. A distorted vanadyl pyrophosphate structure maycorrespond to an amorphous signal in ESR and HREM.

The overstoichiometric amount of phosphate used in the catalyst synthesis is enoughto cover the catalyst surface without disturbing the bulk vanadyl pyrophosphate structure. TheVPO-aq has a specific surface area of 3 m2/g and, therefore, exposes 1 in 280 atoms to itssurface and the VPO-org with a specific surface area of 9 m2/g exposes 1 in 93 atoms to itssurface. Hence, the 10% excess phosphate, used in the catalyst synthesis, would even beenough to form 9 (for the VPO-org) or 28 (for the VPO-aq) monolayers of pure phosphate.

0

15

30

45

700 1600 2500

Energy [eV]

LEIS

sig

nal [

cnts

./nC

]

O

P

VX 0.44


72

In the preceding we explained our LEIS data in terms of a (VO)2P2O7 structure withadditional phosphate groups. In the literature VPO catalysts are often considered as a mixtureof (VO)2P2O7 with different VOPO4 phases [47,48]. In regard to the total atomic phosphorousand vanadium atomic densities, the (VO)2P2O7 and VOPO4 phases are rather comparable. Forinstance, slabs of (001) α−VOPO4 and (201) β−VOPO4 phases match closely with (100)(VO)2P2O7 and the crystallographic misfits are 2 % and 3 %, respectively, for α−VOPO4 andβ−VOPO4 [47]. Any solid mixture obtained at the steady state with stoichiometriccomposition between VPO5 and VPO4.5 can be pictured. The frameworks of γ−VOPO4 and(VO)2P2O7 are made of the same elementary units (pairs of octahedra). Only smalldisplacements of the atoms laying in the (010) γ−VOPO4 are needed to incorporate V5+

species in (VO)2P2O7 (or the reverse, some V4+ in γ−VOPO4) by the local formation ofmicrodomains with the same structural unit [47].

In some of the VOPO4 phases half of the phosphate groups are phosphorousterminated and the other half is oxygen terminated, like in (VO)2P2O7. However, inαΙΙ−VOPO4 all of the phosphate groups expose their phosphorous atom to the surface.Therefore, the surface P/V ratio of αΙΙ−VOPO4 is 2, instead of 1 like in (VO)2P2O7. Hence, asignificant contribution of a phase such as αΙΙ−VOPO4 could explain the surface P/V ratio andthe phosphorous atomic density as found in LEIS on the VPOaq catalyst. In regard to thevanadium atoms, the nearest neighbours and the orientation of the vanadium atoms towardthe surface are the same in αΙΙ−VOPO4 and (VO)2P2O7. Hence, the vanadium atomic densityin αΙΙ−VOPO4 is only 14% higher than that in (VO)2P2O7 because the αΙΙ−VOPO4 is moreclosely packed. Moreover, it should be noticed that the AV of the vanadium atoms, asdetermined with XPS, is in the range 4 – 4.3, whereas in αΙΙ−VOPO4 the AV is 5. Since XPSaverages layers, it is not to be excluded that part of the outermost surface layer of the catalystmight be αΙΙ-VOPO4 alike.

Evidently, there are differently bound phosphates in the VPO catalysts studied. Thesemay have different catalytic activity. Moreover, the required phosphoric acid addition duringcatalysis may selectively replenish a single type of phosphates.

ConclusionsUsing the LEIS technique, we have found that equilibrated VPO catalysts, prepared in

both aqueous and organic media, are largely covered by carbonaceous species upon exposureto air after catalytic tests. Thermal treatment in an inert atmosphere or vacuum up to 573 Kdoes not remove these species. Only upon mild oxidation treatment (200 mbar O2 at 573 K)are carbon deposits removed from the VPO-aq catalysts. This is interesting since calcinationis patented for both activation and regeneration of VPO catalysts for pressure, temperatureand contact time ranges comprising the conditions of our oxidative treatment [44,46]. XPSshowed this treatment does not significantly change the average valence of the catalyst. Aftera mild oxidation treatment, the average valence of the vanadium in the catalyst was 4.3±0.2.No more carbonaceous species are detected on VPO-aq catalysts upon this treatment. VPO-org catalysts on the other hand, still contain carbon upon the oxidation treatment. VPOcatalysts have a platelike morphology, therefore, they can accommodate carbon in the bulk

Chapter 3

73

[42,43]. In the VPO-org catalysts, carbon from the organic medium may have been leftbetween these layers during synthesis. During the cleaning treatment this carbon maysegregate and replenish the removed carbon at the VPO-org surface. Since VPO-org andVPO-aq have an equal activity m-2 we assume the same active site for both catalysts andtherefore continued our study with VPO-aq.

Upon the cleaning treatment, the total surface atomic densities of P and V increasedsix- and nine-fold respectively. Since the vanadium atomic density increased most, carbondeposits are thus preferentially shielding surface vanadium atoms. Therefore, this feature canexplain the higher surface P/V ratios, which were earlier observed with LEIS [12,13]. Thecleaned VPO-aq catalyst exposed 1.7⋅1014 P atoms/cm2 and 0.86⋅1014 V atoms/cm2,corresponding to a surface P/V ratio of 2.0±0.2. A comparison between the vanadium signalsof a cleaned VPO-aq catalyst and V2O5 showed that only 3% of the vanadium atoms ismissing in the outermost surface layer compared to that of a vanadyl pyrophosphate structure.The missing vanadium atoms may be explained by vacancies, shielding due to eitherremaining contamination or excess phosphorous or simply a different structure. Assuming asimilar oxygen shielding of vanadium and phosphorous in a vanadyl pyrophosphate structure,we have found a huge excess of phosphorous. In comparison to a vanadyl pyrophosphatestructure the surface P/V ratio of 2.0±0.2 should not be explained in terms of missingvanadium, but in terms of excess phosphorous. The excess amount of phosphorous may bepositioned between the vanadyl units and the phosphate groups. This would result in adistorted structure at the surface of a cleaned VPO-aq catalyst that differs from the vanadylpyrophosphate structure. If the distorted vanadyl pyrophosphate is the actual active phase inthe selective oxidation of n-butane to maleic anhydride, the excess of phosphate is differentlybound to the catalyst and might change under catalytic conditions. It would be worthwhile toinvestigate whether the phosphoric acid addition during catalysis may selectively replenish asingle type of phosphates.

As shown in the previous section, the observed surface P/V ratio of 2.0 can also beexplained by a significant contribution of a phase such as αΙΙ−VOPO4.

AcknowledgementsS. Ermolov and V. Glebovsky from the Chernoglovka Institute for Solid State Physics

(Russia) are acknowledged for providing a high purity polycrystalline vanadium sample.R. Cortenraad is kindly acknowledged for performing the LEED measurements. L. Coulierfrom the Schuit Institute of Catalysis (TUE) is kindly acknowledged for performing the XPSmeasurements. This work was supported by NWO.

References[1] Felthouse T.R., Burnett J.C., Mitchell S.F., and Mummey M.J. in “Kirk-Othmer

Encyclopaedia of chemical technology”, 4th edition volume 15, p. 893, John Wiley &Sons Inc. 1995.

[2] Centi G., Trifirò F., Ebner J.R., and Franchetti V.M., Chem. Rev. 88, 55 (1988).[3] Chem. Week, June 4 p. 37 (1997).


74

[4] Bastians P., Genet M., Daza L., Acosta D., Ruiz P., Delmon B., in “New developments inselective oxidation by heterogeneous catalysts”, (P. Ruiz and B. Delmon Eds.) p. 267Elsevier, Amsterdam, 1992.

[5] Harrouch Batis N., Batis H., Ghorbel A., Vedrine J.C., and Volta J.C., J. Catal. 128, 248(1991).

[6] Abon M., Béré K.E., Tuel A., and Delichere P., J. Catal. 156, 28 (1995).[7] Cornaglia L.M., Caspani C., Lombardo E.A., Appl. Catal. 74, 15 (1991).[8] Garbassi F., Bart J., Tassinari R., Vlaic G., and Labarde P., J. Catal. 98, 317 (1986).[9] Satsuma A., Hattori A., Futura A., Miyamoto A., Hattori T., and Murakami Y., J. Phys.

Chem. 92, 2275 (1988).[10] Morishige H., Tamaki J., Miura N., and Yamazoe N., Chem. Lett. 1513 (1990).[11] Stefani G., Budi F., Fumagalli C., and Succiu G.D., in “New developments in Selective

Oxidation”, (G. Centi, F. Trifiro Eds.) p. 537 Elsevier, Amsterdam, 1990.[12] Richter F., Papp H., Götze Th., Wolf G.U., and Kubias B., Surf. Interface Anal. 26, 736

(1998).[13] Delichere P., Béré K.E., and Abon M., Appl. Catal. A: General 172, 295 (1998).[14] Coulston G.W., Thompson E.A., and Herron N., J. Catal. 163, 122 (1996).[15] Sananes M.T., Hutchings G.J., Volta J.C., J. Chem. Soc. Chem. Commun. 243 (1995).[16] Sananes M.T., Hutchings G.J., Volta J.C., J. Catal. 154, 253 (1995).[17] Scofield J.H., J. Electron Spectrosc. Rel. Phen. 8, 129 (1976)[18] Shimoda T., Okuhara T., Misono M., Bull Chem. Soc. Jpn. 58, 2163 (1985).[19] Okuhara T., Misono M., Catal. Today 16, 61 (1993).[20] Kubias B., Richter F., Papp H., Krepel A., Kretschmer A., in “Proceedings of the third

world congress on oxidation catalysis”, (R.K. Graselli, S.T. Oyama, A.M. Gaffney, J.E.Lyons Eds.) p. 461. Elsevier, Amsterdam, 1997.

[21] Katsumoto K., and Marquis M., U.S. patent 4,132,670 (1979).[22] Centi G., Garbassi C., Manenti I., Riva A., and Trifiro F., in “Prepartion of Catalysts

III”, (B. Delmon, P.A. Jacobs, and P. Grange Eds.) p. 543. Amsterdam, 1983.[23] Taglauer E., in “Fundamental Aspects of Heterogeneous Catalysis Studied by Particle

Beams” (H.H. Brongersma and R.A. van Santen, Eds.), NATO ASI B 265 p. 301.Plenum Press, 1991.

[24] Brongersma H.H., Groenen P.A.C., and Jacobs J.-P., in “Science of Ceramic InterfacesII” (J. Nowotny Eds.) p. 113. Elsevier, 1994.

[25] Hellings G.J.A., Ottevanger H., Boelens S.W., Knibbeler C.L.C.M., and BrongersmaH.H., Surface Science 162, 913 (1985).

[26] Hellings G.J.A., Ottevanger H., Knibbeler C.L.C.M., Van Engelshoven J., andBrongersma H.H., Journal of Electron. Spectrosc. Relat. Phenom. 49, 359 (1989).

[27] Bergmans R.H., Kruseman A.C., Severijns C.A., and Brongersma H.H., AppliedSurface Science 70/71, 283 (1993).

[28] Okano A., Thoma R.K.R., Williams G.P., and Williams R.T., Physical Review B, 52(20), 14789 (1995).

Chapter 3

75

[29] Lee B.W., Ni R.K., Masud N., Wang X.R., Wang D.C., and Rowe M., Journal ofVacuum Science and Technology, 19 (3), 294 (1981).

[30] Cortenraad R., Denier van der Gon A.W., Brongersma H.H., Gärtner G., andManenschijn A., Applied Surface Science 146, 69 (1999).

[31] Fiermans L., Clauws P., Lambrecht W., Van den Broucke L., and Vennik J., Phys. Stat.Sol. 59 A, 485 (1980).

[32] Jansen W.P.A., Knoester A., Maas A.J.H. Schmit P., and Brongersma H.H., submitted.[33] Overbeek R.A., Versluijs-Helder M., Warringa P.A., Bosma E.J., and Geus J.W., Stud.

Surf. Sci. Catal. 82, 183 (1994).[34] Abon M., Volta J.C., Applied Catalysis: A General 157, 173 (1997).[35] Kittel C., in “Introduction to Solid State Physics”, p. 76. John Wiley & Sons 1986.[36] Leerdam G.C., Jacobs J.-P. and Brongersma H.H., Surf. Sci. 268, 45 (1992).[37] Overbeek R.A., Thesis, Utrecht University, 1994.[38] Bergmans R., Thesis p. 70. Eindhoven University of Technology, 1996.[39] Nguyen P.T., Hoffman R.D., and Sleight A.W., Mater. Res. Bull. 30, 1055 (1995).[40] Ruitenbeek M., Barbon A., Van Faassen E.E., and Geus J.W., Catlysis Letters 54, 101

(1998).[41] Guliants V.V., Benziger J.B., Sundaresan S., Wachs I.E., Jehng J.M., and Roberts J.E.,

Catalysis Today 28, 275 (1996).[42] Busca G., Cavani F., Centi G., and Trifiro F., J. Catal. 99, 400 (1986).[43] Sola G.A., Pierini B.T., and Petunchi J.O., Catal. Today 15, 537 (1992).[44] Contractor R.M., Horowitz H.S., Patience G.S., and Sullivan J.D., US Patent 5,895,821

(1999).[45] Ruitenbeek M., Van Dillen A.J., Barbon A., Van Faassen E.E., Koningsberger D.C.,

and Geus J.W., Catalysis Letters 55, 133 (1998).[46] Contractor R.M., US Patent 5,021,588 (1991).[47] Bordes E., Catal. Today 1, 499 (1987).[48] Hutchings G.J., Christopher J.K., Sananes-Schulz M.T., Burrows A. and Volta J.C.,

Catal. Today 40, 273 (1998)[49] Richter F., Papp H.,Wolf G.U., Götze Th., and Kubias, Fresenius J. Anal. Chem., 365,

150 (1999)

77

4Dynamic Behavior of the Surface Structure of Cu/ZnO/SiO2

Catalysts*

AbstractIn order to grasp the dynamic behavior of the surface composition of Cu/ZnO based catalysts,the surface atomic densities of 63Cu and 68ZnO have been determined separately with staticLEIS on 63Cu/68ZnO/SiO2 catalysts. Our data show that the methanol synthesis activity andsurface composition of 63Cu/68ZnO/SiO2 depend strongly on the reduction temperaturebetween 473 – 673 K. The catalyst surface is strongly enriched in ZnO under methanolsynthesis conditions. The degree of oxidation of the Cu species in the outermost atomic layerof the Cu/ZnO/SiO2 surface has been determined by performing LEIS in combination withadsorptive decomposition of N2O. The observed oxidation behavior of the Cu species in thereduced catalyst surface differs clearly from that of pure metallic Cu and can only beexplained in terms of the formation of Cu(I)/ZnO with oxygen vacancies. The oxygenvacancy concentration is shown to be clearly affected by the reducing agent (being 5%CO/5% CO2/90% H2 or pure H2).

* The contents of this chapter has previously appeared in W.P.A. Jansen, J. Beckers, J.C. v.d.Heuvel, A.W. Denier v.d. Gon, A. Bliek and H.H. Brongersma, Journal of Cataysis, 210,229-236 (2002).

Dynamic behavior of the surface of Cu/ZnO/SiO2

78

1. IntroductionCu/ZnO based catalysts are commercially applied in many industrial processes such as

the production of methanol from synthesis gas, the low temperature shift reaction, and thesynthesis of fatty alcohols from methyl esters. The annual worldwide methanol productionalone amounts to 30⋅106 tons [1], hence, the optimization and understanding of these catalystsis of great interest. Although extensive studies have been carried out for more than twodecades, controversies remain concerning the nature of the active site in methanol synthesisand the role of ZnO. Surface structures like Cu on top of ZnO [2-4], Cu(I)/ZnO [5,6], ZnO ontop of Cu [7,8], mixed Cu-O-Zn [9], and Cu-Zn alloys [10-12] have been proposed for theCu/ZnO based catalysts.

Part of the controversy may originate from the frequently reported dynamic behavior ofthe Cu/ZnO system, see e.g. [3,10,13]. It has been speculated that under reaction conditions,part of the ZnO in the catalyst may be reduced and segregate to the surface [12]. Spencershowed that the formation of an α-brass is thermodynamically favored under reactionconditions [14-16]. He also suggested that the presence of traces of CO2 could significantlychange the surface concentration of zinc, which could explain part of the experimentaldiscrepancies [16]. To probe the assumed segregation processes, selective information on theoutermost atomic layer is essential to avoid averaging of the surface and subsurfacecompositions.

The commonly used method to determine the Cu metal surface area in Cu/ZnO catalystsis by adsorptive decomposition of nitrous oxide [17], according to the reaction:

2Cus + N2O → (Cus)2O + N2, (1)

where Cus denotes a Cu surface site. In spite of its common use, this method entails afundamental problem to select experimental conditions that lead to complete monolayercoverage of Cu with oxygen, while excluding sub-surface oxidation of Cu. Moreover, theremoval by N2O of oxygen vacancies that are possibly present in ZnO would lead to anoverestimation of the Cu metal area when using adsorptive decomposition of nitrous oxide.When Cu-Zn alloys (brasses) are formed, this method is no longer suited to characterize theCu surface area and these surfaces may be overestimated with as much as 100% [18].

We have applied low-energy ion scattering (LEIS) to determine the Cu(O) and Zn(O)surface atom densities in the outermost layer, thus exactly where catalysis takes place. Sincethe natural Cu and Zn isotopes have overlapping masses, a separate analysis of Cu and Zn ishampered. In a previous study we have shown, however, that isotopic enrichment allows theseparate detection of 63Cu and 68Zn in 63Cu/68ZnO catalysts with LEIS [19]. At that time theLEIS measurements showed a considerable amount of Pb on the surface of the 63Cu/68ZnOcatalyst. It was concluded that this Pb originated from the 68ZnO raw material and segregatedduring catalyst reduction to the surface. In contrast, no impurities were detected on thesurface of a 63Cu/68ZnO/SiO2 catalyst. This seemingly contradictory behavior was attributedto the higher dispersion of Cu and Zn in the 63Cu/68ZnO/SiO2 catalyst or interaction of Pbwith the support. As no Pb was detected on the surface of 63Cu/68ZnO/SiO2 catalysts, these

Chapter 4

79

were used in the present study to investigate the dynamic behavior of the Cu/ZnO system,following reductive treatments with various gas mixtures and at various temperatures.

In principle, LEIS merely provides information on the atomic composition, and not on theoxidation state. The LEIS yield of a pure metal, however, is typically ~2 to 5 times higherthan that of the corresponding metal oxide, due to shielding of the metal by the oxygen in theoxide. Using this difference, information can be obtained on the oxidation behavior/state ofthe Cu surface species from LEIS measurements before and after N2O decomposition. Thenovel method of combining N2O decomposition and LEIS has two advantages. Firstly, LEISenables the separate analysis of the Cu and Zn concentrations. Secondly, LEIS assuresselective information on the outermost atomic layer. Hence, the oxidation behavior of the Cusurface species and the reliability of adsorptive decomposition of N2O in assessing the Cusurface area on Cu/ZnO catalysts could be determined.

2. Experimental2.1 Catalyst preparation

The synthesis procedure of the isotopically enriched 63Cu/68ZnO/SiO2 catalysts was thesame as is described in our previous study [19]. Isotopically enriched [63Cu]-Cu was obtainedfrom Campro Scientific with a chemical and isotopical purity of >99.875 wt % and>99.87 wt %, respectively. Isotopically pure 68ZnO was obtained from Alfred Hempel GmbHand was chemically and isotopically >98.8 wt % pure. Both Cu and ZnO were converted intotheir nitrate forms by dissolution in 65% nitric acid (Acros Chimica, p.a.).

A 63Cu/68ZnO/SiO2 (12.5±0.4 wt % Cu, 5.15±0.15 wt % Zn: as determined withinductively coupled plasma – atomic emission spectroscopy, 95% confidence interval)catalyst was prepared by homogeneous precipitation as described elsewhere [20].Subsequently, the precipitation mixture was aged for 140 min. after which it was filtered andflushed with demineralized water. The catalyst was dried overnight in air at 363 K.

The dried catalyst samples were pressed (at 250 MPa) into cavities (∅ 2 mm) in aluminadisks (∅ 9.9 mm). We have determined that pressing up to at least 300 MPa does notinfluence the surface composition of a silica-supported catalyst. Each disk contains twocavities, hence, two identically prepared catalysts. The catalyst was calcined in a flow of2 cm3⋅s-1 of dry air at a temperature of 750 K for 12 h. Subsequently, the catalyst was reducedfor 1 h at either 473, 573, or 673 K in a flow of 2 cm3⋅s-1 hydrogen (Praxair, 99.999% pure).The heating rate in all cases was 72 K⋅h-1. The reduction procedures correspond to thosedescribed in refs [3,13,20,21,34] (and references therein), where remarkable changes incatalyst activities for acetone hydrogenation, ester hydrogenolysis, and methanol synthesisare reported depending on the temperature of the reductive pretreatment. Before transferringthe samples to the LEIS set-up, these were passivated at 363 K in a flow of 2 cm3⋅s-1

1% N2O/99% Ar.In the LEIS set-up, the catalysts were re-reduced in a pretreatment chamber at the

previously applied reduction temperature, either in 5% CO/5% CO2/90% H2 or in purehydrogen at atmospheric pressure for 15 min. In order to prevent subsequent changes in thesurface composition, samples were always cooled below 373 K in the reactant mixtures


80

before evacuation was started. Both in situ measurements [7] and kinetically calculated timescales [15] show that in this way changes in the surface composition may successfully beprevented. N2O decomposition was carried out in the LEIS pretreatment chamber for 45 min.at 363 K in 5% N2O/95% He at atmospheric pressure.

The pure 63Cu sample did not receive ex situ reduction or calcination and was onlyreduced in the LEIS pretreatment chamber (30 min. in 1 atm. H2 at 573 K). To investigate theshielding of Cu by oxygen atoms for 4 keV Ne+ scattering, N2O decomposition was carriedout on a sputter cleaned 63Cu sample in the LEIS pretreatment chamber (45 min. in 1 atm.5% N2O/95% He at 363 K). A pure 68ZnO sample was analysed after calcination (30 min. in1 atm. O2 at 573 K) and after a subsequent reduction (15 min. in 1 atm. H2 at 573 K).

2.2 LEISThe LEIS experiments were carried out using the ERISS LEIS set-up. In this set-up a

beam of mono-energetic noble gas ions (He+ or Ne+ with energies of 3 and 4 keV,respectively) is directed perpendicularly onto the target. The energy distribution of thebackscattered ions is analysed for a fixed scattering angle (145°) with a double toroidalelectrostatic analyser. The analysis is similar to that of the EARISS set-up, which has beendescribed in more detail elsewhere [22,23]. This analyser makes very efficient use of thebackscattered ions by measuring simultaneously a considerable part of the energy spectrumcontained in 320° of the azimuthal angle. Using this type of analyser measurements can becarried out using only 1013 ions/cm2. For metallic Cu the 4 keV Ne+ sputter yield is 3atoms/ion [24]. When the same yield is assumed for ZnO and the number of atoms in thesurface of the catalyst amounts to 1.3⋅1015 at/cm2 (average of Cu [25] and ZnO [26]), a doseof 0.4⋅1015 Ne+/cm2 corresponds to the removal of one atomic layer. The 3 keV He+ sputteryield is 0.25 atoms/ion [24], resulting in a dose of 5.2⋅1015 He+/cm2 for the removal of oneatomic layer. Hence, the ERISS allows us to perform static LEIS, i.e., with negligibledamage. Note that during the previous study -using the NODUS set-up- [19], ~102 higherdoses had to be used. In the present study both low doses (~1013 ions/cm2) and higher doses(~1015 ions/cm2) have been applied, the latter to obtain depth profiles. Depth profiles of theatomic compositions of the different catalysts have been obtained by fitting the areas of theCu and Zn peaks in the 63Cu/68ZnO/SiO2 LEIS spectra, to reference spectra measured on pure63Cu, 63CuO and 68ZnO as a function of the applied dose. The depth profiles have been fittedassuming a combination of uncovered CuO and ZnO, and CuO and ZnO that were eithershielded by hydrogen or by each other. The sputter-rates of H, CuO, and ZnO were keptconstant in all fits.

2.3 Surface oxidation state evaluated with LEIS and adsorptive N2O decompositionIn principle, LEIS provides only information on the atomic densities of the atoms present

in the surface, and not on their oxidation state. One cannot tell whether the detected oxygenatoms belong to copper- or to zincoxide. However, the oxygen atoms will (partly) shield themetal atom in an oxidized metal. Therefore, the 4 keV Ne+ Cu LEIS yield of a metallic Cureference sample is a factor of 5 higher than that of oxidized copper formed after exposing

Chapter 4

81

the Cu metal to N2O (30 min. treatment in 1 atm. 5% N2O/95% He at 363 K). Similarly, the 4keV Ne+ LEIS yield of Zn is a factor of 3.7 higher than that of ZnO. Hence, comparison ofthe Cu and Zn LEIS yields of a 63Cu/68ZnO/SiO2 catalyst before and after exposition to N2O,under the conditions as mentioned above, can give information on the degree of oxidation ofthe Cu and Zn atoms in the catalyst surface. Aluminum disks, containing two catalystsamples each, have been analyzed with LEIS before and after exposing them to N2O. In thisway two further identically prepared and pretreated catalysts could be compared to obtaininformation on the degree of oxidation of the surface species.

3 Results3.1 Oxidation state of the Cu and Zn in the Cu/ZnO/SiO2 catalyst

It is crucial to asses the degree of oxidation of the copper and zinc atoms in theCu/ZnO/SiO2 catalyst in order to comprehend the structure-activity relation. As explainedbefore, the combination of LEIS measurements before and after N2O chemisorption may beused to address this issue. Figure 1 shows the Cu and Zn LEIS signals of reduced63Cu/68ZnO/SiO2 (hydrogen at 473 K) before (open markers) and after (solid markers)subsequent exposure to N2O.

Figure 1. The Zn (triangles) and Cu (squares) atomic density of Cu/ZnO/SiO2 as a function of 4 keVNe+ dose. The solid data points have been obtained after reduction at 473 K in pure H2, the open datapoints have been obtained after the sample was subsequently exposed to N2O at 363 K.

The figure shows that both the Cu and Zn LEIS yields did not change significantly (i.e. lessthan 4%). The Cu and Zn LEIS yields of 63Cu/68ZnO/SiO2 reduced in hydrogen at 673 K,decreased about 20% upon N2O decomposition. Hence, the oxidation behavior of the Cu andZn species in the surface of a reduced 63Cu/68ZnO/SiO2 catalyst is different from that of pure,

0

2

4

6

0 1 2 3Dose (1015 Ne+/cm2)

LEIS

Sig

nal (

Cnts

/nC)


82

metallic Cu and Zn where a reduction by a factor of 3.7 - 5 results from N2O exposure. Fromthis we can conclude that less than 5% (after reduction at 473 K) or 25% (after reduction at673 K) of both the Cu and Zn species at the surface is metallic. The minor effect of the N2Odecomposition on the LEIS yields of reduced 63Cu/68ZnO/SiO2 can only be explained by thepresence of (partly) oxidized Cu and Zn that may well be present in the form of Cu(I)/ZnO[5,6]. The presence of oxygen vacancies [27] may explain the observed 20% decrease of theCu and the Zn signals after the N2O treatment of catalysts, which were previously reduced at673 K. We have used CuO and ZnO to calibrate surface area percentages of the (partly)oxidized Cu and Zn species in the catalyst surface. Earlier LEIS studies have shown thevalidity of this approach, since copper oxide [28] and zinc oxide [29] show no matrix effects.

3.2 Dynamic behaviorThe surface atomic densities of ZnO and CuO have been determined with LEIS for

63Cu/68ZnO/SiO2 reduced in pure H2 at 473, 573 and 673 K. Figure 2 shows the CuO andZnO percentages of the total surface area as a function of the applied 4 keV Ne+ dose. Asdescribed above, a dose of 0.4⋅1015 ions/cm2 corresponds to the removal of approximately onemonolayer (ML). Hence, the data shown in Fig. 2 represent surface area percentages fordepths of 0 − 7 ML.

Figure 2a-c. The ZnO (open triangles) and CuO (solid squares) atomic densities as a function of4 keV Ne+ dose for Cu/ZnO/SiO2 after reduction in pure H2 at 473 K (2a), 573 K (2b), and 673 K (2c).All data have been fitted using the same, constant values for the H, CuO and ZnO sputter-rates (solidlines).

Regardless of the applied reduction temperature, the data show a sharp increase inboth the CuO and ZnO atomic densities for doses up to ~0.25⋅1015 ions/cm2. This sharp

0

1

2

3

4

5

6

0 1 2 3Dose (1015 Ne+/cm2)

Surfa

ce A

rea

(%)

473 K

0

1

2

3

4

5

6

0 1 2 3Dose (1015 Ne+/cm2)

673 K

0

1

2

3

4

5

6

0 1 2 3Dose (1015 Ne+/cm2)

573 K

Chapter 4

83

increase may be explained by the presence of residual hydrogen (probably ashydroxylgroups) from the reductive treatment that remained on the surface after theevacuation treatment. Therefore, the depth profiles of the Cu (/Zn) LEIS yields have beenfitted assuming a combination of uncovered CuO and ZnO, and CuO and ZnO that wereeither shielded by hydrogen or by each other. All fits are based on the same, constant valuesfor the sputter-rates of H, CuO, and ZnO. From the fits, represented by the solid lines in Fig.2, it can be deduced that hydrogen has a 13 times higher sputter yield than ZnO. This is inagreement with a reported 10 to 50 times higher sputter rate for hydrogen compared to that ofother elements [30].

After the ZnO coverage has increased, because of the hydrogen removal, it starts todecrease while the CuO coverage keeps increasing. Hence, the surface of 63Cu/68ZnO/SiO2 isenriched in ZnO after reduction. This is true even for reduction at 473 K, the extend of theenrichment in ZnO depends strongly on the reduction temperature. Fig. 3 shows the surfacecomposition (i.e. CuO and ZnO surface percentages deduced from fits for zero dose and inthe absence of H) of 63Cu/68ZnO/SiO2 after reduction in H2 at 473, 573 and 673 K. Note thatthe measurements before and after exposure to N2O demonstrate that following reduction at473 K less than 4% of the Cu surface species is metallic (see section 3.1). Hence, the surfaceconcentration of metallic Cu after reduction at 473 K in pure H2 is less than 0.2%. Afterreduction at 673 K, the surface concentration of metallic Cu is 3⋅102 ppm.

After reduction at 673 K about 96% of the cluster surface area is ZnO terminated andalmost no CuO is exposed (see Fig. 3). This is in agreement with thermodynamics [19,31].The low CuO content in the outermost layer also explains the absence of a sharply increasingCuO surface concentration as a function of dose upon reduction at 673 K, see Fig. 2c.Following reduction at this temperature CuO is mainly covered with ZnO, instead ofhydrogen. In contrast, the fits show no significant shielding of ZnO by CuO, regardless of theapplied reduction temperature.

Figure 3 The CuO (black bars), ZnO (white bars) and total CuO+ZnO (grey bars) surface atomicdensities as derived from the fits of the LEIS data after reduction in pure H2. Note the decrease in thetotal cluster surface area because of sintering.

0

2

4

6

8

473 573 673

Reduction Temperature (K)

Surf

ace

Are

a (%

)


84

Both Figs. 2 and 3 show a decrease of the total cluster area with increasing reductiontemperatures. Reference [32] shows that such a decrease can be related to sintering. Hence,after reduction at 573 K and especially after reduction at 673 K, the clusters have partlysintered. TEM results confirm this, a combination of small (~3 nm) and large (~10 nm)clusters is observed on a catalyst reduced at 673 K, whereas a catalyst reduced at 473 Kexposes only small clusters (~3 nm).

After reduction at 673 K the clusters have (partly) sintered and the overall ZnOconcentration is high enough to cover the entire cluster surface. The LEIS measurementsshow that indeed 96% of the cluster surface area is covered with Zn species after reduction at673 K. In thermodynamic equilibrium one would expect the ZnO enrichment even higherafter reduction at 473 K than after reduction at 673 K. However, the overall ZnOconcentration in the catalyst can cover at most 55% of the cluster surface area if the metal isspread over 3 nm clusters. After our 15 min. reduction treatment only 44% of the clustersurface area is covered with Zn species. Hence, the concentration is kinetically determined.After prolonged exposure to H2 at 473 K a cluster surface concentration of ~55% may beexpected.

During methanol synthesis, Cu/ZnO catalysts are not exposed to pure hydrogen, butto CO/(CO2)/H2 mixtures. To study the effect of a methanol synthesis gas, we have alsoanalyzed a 63Cu/68ZnO/SiO2 catalyst pretreated in 5% CO/5% CO2/90% H2, a gas mixturethat was also used by Grunwaldt et al. from Haldor Topsøe [12] to mimic industrially appliedsynthesis gas mixtures. In figure 4 the CuO and ZnO percentages of the total surface area arepresented as a function of Ne+ dose following treatments in 5% CO/5% CO2/90% H2 (soliddata points) and in pure hydrogen (open data points) at 573 K. Following reduction in5% CO/5% CO2/90% H2 the CuO (squares) and ZnO (triangles) surface areas are 11% lowerthan the corresponding signals after reduction in pure H2.

Figure 4. The ZnO (triangles) and CuO (squares) atomic densities the open data points were obtainedafter reduction at 573 K in pure H2, the solid data points after reduction in 5% CO/5% CO2/90% H2.

0

1

2

3

4

0 1 2 3Dose (1015 Ne+/cm2)

Surf

ace

Are

a (%

)

Chapter 4

85

Exposure to N2O following reduction in H2 leads to a drop in the CuO and ZnO LEIS yieldsof less than 4% (after reduction at 473 K) and 20% (after reduction at 673 K) we assume thatthe difference of 11% (reduction in 5% CO/5% CO2/90% H2 vs. pure H2 at 573 K) is likelycaused by removal of the oxygen in the catalyst surface by CO2 [33]. Except for thisdifference, the treatments in 5% CO/5% CO2/90% H2 and pure H2 produce very similarsurfaces. In both cases one observes a surface that is enriched in ZnO, the ZnO/Cu ratio uponlow ion doses (reflecting the surface conditions) is much higher than the bulk value. From themeasurements it may be concluded that a pretreatment in pure hydrogen leads to a slightlyhigher ZnO enrichment than pretreatment in 5 % CO/5% CO2/90% H2. However, thedifference is small with respect to the earlier discussed temperature effect.

4 Surface structure modelsRecently, Poels and Brands reviewed literature, catalytic tests, XRD, FT-IR and

preliminary LEIS analyses of the Cu/ZnO/SiO2 catalyst [13]. The results relevant for thispaper are summarised here. Table 1 gives an overview of the results of catalyst activity testsconcerning acetone hydrogenation, ester hydrogenolysis, and methanol synthesis [13,34]. Byway of comparison, the activity of unpromoted Cu/SiO2 and Cu/MnOx/SiO2 are also shown.

Table 1. Summary of the activity data of relevant Cu based catalysts, with reduction temperatures of573 K and 673 K [13,34].

Cu/SiO2 Cu/ZnO/SiO2 Cu/MnOx/SiO2

methyl acetatehydrogenolysis 573 K

(673 K)10%

(10%)20%

(48%)20%

(27%)methyl palmitate

hydrogenolysis 573 K(673 K)

18%(18%)

40%(66%)

48%(52%)

methanol synthesisCO/H2 = 1/2 573 K(CO/H2 = 1/2 673 K)

0 nmol s-1 g-1 cat.(0 nmol s-1 g-1 cat.)


-

methanol synthesisCO/H2 = 1/3 573 K(CO/H2 = 1/3 673 K)



-

Unpromoted Cu/SiO2 has a modest activity, which essentially remains constant for allreactions as the reduction temperature is increased from 573 K to 673 K. In contrast,Cu/ZnO/SiO2 catalysts demonstrate a remarkable activity increase for higher catalystsreduction temperatures. Therefore, Poels and Brands proposed that the Cu – ZnO interfaceplays a crucial role in catalysing these reactions. After reviewing literature and experimentalevidence they reported two possible models that explain the enlarged interface after hightemperature reduction. The two models suggest either (1) migration of partly reduced ZnO on


86

top of Cu or (2) reversible formation of flat epitaxial Cu particles upon high temperaturereduction.

Three arguments would speak against model (1). For one, a treatment in pure N2

during 2 h of the reduced catalyst at a reduction temperature of 673 K partly reverses theobserved increase in activity upon reduction. The reversibility of this was demonstrated by asubsequent repeated second reduction in H2 at 673 K, resulting in exactly the same activity aswas observed for a single reduction in hydrogen. This reversibility was consideredinconsistent with alloy formation as the promoting effect. However, ppm levels of O2 due toleakage or other contaminants in the N2 feed may well explain the observed deactivationduring the 2h inert treatment. Similar to N2O decomposition, trace amounts of O2 may act topassivate the catalyst. Such a passivation, however, can be completely reversed bysubsequent reduction without any loss of activity [35]. The observed catalytic activity may beexplained as follows: during either the “inert” or N2O treatment the catalysts gets deactivateddue to partial oxidation, and during the subsequent reduction Cu(I)/ZnO with oxygenvacancies is formed again and the catalyst is regenerated.

A second argument against model (1) concerns the reported similarity between theactivity as a function of reduction temperature of Cu/ZnO/SiO2 and Cu/MnOx/SiO2 catalysts[13]. It should be noted however, that the increase in activity of the Cu/ZnO/SiO2 is muchmore profound than that of the Cu/MnOx/SiO2 (Table 1).

A third argument against model (1) is formed by the absence of a peak shift in XRDthat would accompany alloy formation [13]. However, it should be noted that XRD is a bulktechnique that selectively detects crystalline phases. Hence, amorphous or surface phasesaccompanying segregation, as shown by the LEIS measurements, would not be detected bythis technique. Note that also EXAFS and even XPS may similarly average out segregationbecause of the probing depths of these techniques [36]. In conclusion, there remain no seriousarguments against model (1).

The present LEIS measurements confirm and extend earlier preliminary LEIS data,demonstrating that the surface of the Cu/ZnO/SiO2 catalyst is strongly enriched in ZnO uponreduction, in agreement with model (1). The ZnO:CuO ratio increases by a factor of 7 whenthe reduction temperature is increased from 573 K to 673 K. Moreover, there are strongindications for the presence of oxygen vacancies following reduction at 573 K and 673 K,whereas these are completely absent after reduction at 473 K. This may well explain theobserved increase in methanol synthesis activity of the Cu/ZnO/SiO2 catalyst in CO2/H2 andCO/H2 mixtures that increase from 0 nmol s-1 g-1 cat to 45 nmol s-1 g-1 cat and from20 nmol s-1 g-1 cat to 50 nmol s-1 g-1 cat, respectively.

Model (2), the so-called Yurieva model, involves the reversible formation of flatepitaxial Cu particles upon high temperature reduction from the fraction of the Cu dissolvedin the mixed catalyst precursor phase. Using XRD and transmission electron microscopy(TEM) Yurieva et al. [37] observed the formation of flat epitaxial Cu particles on top of a Cu-Zn mixed oxide phase after a treatment under flowing hydrogen at 533 K. Formation of theseepitaxial Cu particles correlated with enhanced hydrogenation activity. These authors showedthat a maximum methanol formation rate is obtained for a catalyst containing about equal

Chapter 4

87

molar amounts of Cu and ZnO. In their study, the XRD and TEM analyses were performedon surfaces of Cu/ZnO and Cu/ZnO/Al2O3 catalysts with very low Cu:ZnO ratios (8% : 92%and 15% : 75%, respectively). Hence, one may question to what extent the active phase ispresent on the barely active catalyst that was investigated by Yurieva et al. The strongestargument in favour of the Yurieva model is the agreement between their model and theadsorptive N2O decomposition data indicating an increasing Cu area after reduction.However, Bernt et al. [18] have clearly proven that, in the case of reduced Cu/ZnO basedsystems, this technique may very well produce misleading results. Our present resultsconfirm this. While the weight increase upon N2O chemisorption indicates a higher Cu0

surface area upon reduction, the LEIS data suggest the reverse. Following reduction thecatalyst surface is strongly enriched in ZnO. Moreover, combined N2O chemisorption andLEIS measurements show that the oxidation behavior of the Cu in the catalyst is clearlydifferent from that of pure metallic Cu. According to our measurements metallic Cu coversonly 3⋅102 ppm of the Cu/ZnO/SiO2 surface area (i.e. less than 1% of the cluster surface area)after reduction in H2 at 673 K. Hence, the contribution of metallic Cu (platelets) is verylimited at the surface of the most active catalyst. As indicated in refs. [13,19] the formation ofepitaxial Cu platelets is also highly improbable from a thermodynamic point of view, sincethe surface free energy of Cu is much higher than that of ZnO [31].

In conclusion: model (1), the migration of ZnO on top of Cu followed by theformation of a (partly) oxidized Cu in a Cu(I)/ZnO surface with oxygen vacancies present,seems the best hypothesis for formation of active sites of the Cu/ZnO/SiO2 catalyst. Since thesolubility of CuO in ZnO is limited (4-6 % (w/w)), the Cu will be largely present as Cu1+ [5].Below the (partly) oxidised surface Cu0 may well be present, as schematically indicated inFig. 5, which shows a model of the surface structure derived from all measurements. Forcomparison a low (473 K) and a high temperature (673 K) reduced catalyst are shown (Fig.5a and 5b, respectively). As discussed in section 3.2 the ZnO surface concentration is stillkinetically determined after our reduction at 473 K. Prolonged reduction at this temperaturewould lead to complete ZnO segregation and cause a ZnO cluster surface concentration of~55%. The observed weight increase in adsorptive N2O decomposition could then beexplained by the oxidation of the subsurface Cu atoms and the oxygen vacancies in thesurface. According to [17,38] the adsorptive N2O decomposition method is not necessarilysurface sensitive, the oxidation of subsurface Cu by N2O is thus very well possible.Termination of Cu metal atoms by a (partly) oxidised over-layer is also favoured bythermodynamics [19,31]. The valence states of the subsurface Cu and Zn are not known.However, only the presence of subsurface Cu in the metallic state can explain the resultsobtained with adsorptive N2O decomposition. The maximum solubility of Cu in ZnO is 2%[39]. Assuming that also the solubility of ZnO in Cu is small, the presence of subsurface-metallic-Cu would require reduction of most of the subsurface Zn species as well. Hence, thepresent model explains the catalytic tests and is in agreement with thermodynamics,adsorptive N2O decomposition, LEIS measurements and quasi in situ TEM measurements[40].


88

Figure 5a,b. Model of the surface structure of Cu/ZnO/SiO2 after low (473 K) and high (673 K)temperature reduction as determined with LEIS. One should note that the cluster shape represents aneducated guess, it was not determined.

5. ConclusionsBoth the catalytic activity and the surface composition of a 63Cu/68ZnO/SiO2 catalyst

depend strongly on the applied reduction temperature in the range 473 K – 673 K. LEIS datashow that the catalyst surface is enriched in ZnO following reduction at 473 K. For reductionat 573 K the ZnO enrichment becomes more prominent and finally for reduction at 673 Kvirtually all Cu species in the catalyst surface are covered with ZnO. The ZnO surfaceenrichment of the 63Cu/68ZnO/SiO2 catalyst appears relatively insensitive towards thereducing agent (being either CO/CO2/H2 or pure H2).

The oxidation of the Cu species in the catalyst surface has been determined using anovel method that is based on the difference in the LEIS yields between metallic Cu and Cuoxide. By combining N2O decomposition and LEIS it is demonstrated that the Cu LEIS yieldof a reduced 63Cu/68ZnO/SiO2 catalyst decreases by less than 4% (after reduction at 473 K)and by 20% (after reduction 673 K) upon a N2O decomposition treatment at 363 K. The CuLEIS yield of a metallic 63Cu reference sample decreases by a factor of 5! Moreover, the Znyield decreases also by less than 4%, and 20% after reduction at 473 K and 673 Krespectively, suggesting a similar metal shielding by the oxygen atoms. Hence, the oxidationbehavior of Cu at the Cu/ZnO/SiO2 surface is clearly different from that of pure metallic Cu.This shows that the adsorptive decomposition of N2O is not a straightforward manner todetermine the Cu0 surface area of reduced Cu/ZnO catalysts. However, the combination ofLEIS and the adsorptive decomposition of N2O allows the separate determination of thedegrees of oxidation of the Cu and Zn species in the outermost atomic layer of the catalystsurface.

Cu/ZnO/SiO2 reduced at 473 K

SiO2 support

Outermostatomic layer:ZnO 0.42Cu1+ 0.54Zn0 0.02Cu0 0.02

Subsurface:Cu:Zn = 9

Cu/ZnO/SiO2 reduced at 673 K

Outermostatomic layer:ZnO 0.77Cu1+ 0.03Zn0 0.19Cu0 0.01

Subsurface:Cu:Zn = 9

SiO2 support

Fig 5a Fig 5b

Chapter 4

89

The present results clearly show that ZnO segregates after reduction. The observedoxidation behavior of the Cu in the reduced Cu/ZnO/SiO2 can only be explained in terms ofthe formation of Cu(I)/ZnO, along with oxygen vacancies. Since the methanol synthesisactivity increases dramatically after reduction as well, these results strongly support theories[5,6,9,11,14-16] suggesting that Cu1+ is the active phase. Oxygen vacancies may also explainthe fact that both the CuO and ZnO LEIS yields are ~11% lower after reduction at 573 K in5% CO/5% CO2/90% H2 compared to reduction at 573 K in pure H2.

It should be noted that our measurements leave hardly any room for the presence ofmetallic Cu at the surface. After reduction at 673 K (resulting in the highest activity in themethanol synthesis being 50 nmol s-1 g-1 cat. for our Cu/ZnO/SiO2 catalyst) the metallic Cusurface concentration is 3⋅102 ppm (i.e. less than 1% of the cluster surface area).

AcknowledgementsThe authors wish to gratefully acknowledge E. Timmerman and R. Rumphorst for their helpin improving the energy resolution of the ERISS analyser to enable separate detection of 63Cuand 68Zn. Dr. P.J. Kooyman of the National Centre for High-Resolution Electron Microscopy(Delft, The Netherlands) is gratefully acknowledged for performing the electron microscopyinvestigations. The Netherlands Organization for Scientific Research (NWO) is gratefullyacknowledged for financial support.

References1. American Methanol Institute, http:\\www.methanol.org (2002).2. Waugh K. C., Catal. Today, 15, 51 (1992).3. Yurieva T., Plyasova L.M., Makarova O.V., and Krieger T.A., J.Mol. Catal., A113,

455 (1996).4. Yurieva T., Catal. Today, 51, 457 (1999).5. Klier K., Adv. Catal. 31, 243 (1982).6. Ponec V., Surf. Sci. 272, 111 (1992).7. Fujitani T., and Nakamura J., Appl. Cat., A191, 111 (2000).8. Nakamura J., Nakamura I., Uchijima T., Kanai Y., Watanabe T., Saito M., and

Fujitani T., J. Catal., 160, 25 (1996).9. Spencer M.S., Top. Cat., 8, 259 (1999).10. Topsøe N.-Y., and Topsøe H., Top. Cat., 8, 267 (1999).11. Fujitani T., Nakamura I., Uchijima T., and Nakamura J., Surf. Sci., 383, 285 (1997).12. Grunwaldt J.-D., A.M. Molenbroek, Topsøe N.-Y., Topsøe H., and B.S. Clausen,

J. Catal., 194, 452 (2000).13. Poels E.K., and Brands D.S., Appl. Cat., A191, 83 (2000).14. Spencer M.S., Surf. Sci., 192, 323 (1987).15. Spencer M.S., Surf. Sci., 192, 329 (1987).16. Spencer M.S., Surf. Sci., 192, 336 (1987).17. Skrzypek J., Slozyński J., and Ledakowicz S., in “Methanol Synthesis, Science and

Engineering”, Polish Scientific Publishers, Warszawa (1994) p. 45.


90

18. Berndt H., Briehn V., and Evert S., Appl. Cat., A86, 65 (1992).19. Viitanen M.M., Jansen W.P.A., Van Welzenis R.G., and Brongersma H.H., J.

Phys.Chem, B103, 6025 (1999).20. Brands D.S., Poels E.K., Krieger T.A., Makarova O.V., Weber C., Veer S., and Bliek

A., Catall. Lett., 36, 175 (1996).21. Brands D.S., Poels E.K., and Bliek A., Stud. Surf. Sci. Catal., 101, 1085 (1996).22. Hellings G.J.A., Ottevanger H., Knibbeler C.L.C.M., Van Engelshoven J., and

Brongersma H.H., Surf. Sci., 162, 913 (1985).23. Bergmans R.H., Kruseman A.C., Severijns C.A., and Brongersma H.H., Appl. Surf.

Sci., 70/71, 283 (1993).24. Matsunami N., Yamamurra Y., Itikawa Y., Itoh N., Kazumata Y., Miyagawa S.,

Morita K., Shimizu R., and Tawara H., Atomic Data and Nuclear Data Tables, 31, 1(1984).

25. Oetelaar, L.C.A. Van de, Ph.D. Thesis, Einhoven University of Technology, 1997.26. Evans, J.W., Wainwright M.S., Bridgewater A.J., and Young D.J., Appl. Catal., 7, 75

(1983).27. Ovesen C.V., Clausen B.S., Schiøtz J., Stoltze P., Topsøe H., and Nørskov J.K., J.

Catal., 168, 133 (1997).28. Mikhailov S.N., Elfrink R.J.M., Jacobs J.-P., Oetelaar L.C.A. Van de, Scanlon P.J.,

and Brongersma H.H., Nucl. Instr. Meth., B93, 145 (1994).29. Leerdam G.C.Van, and Brongersma H.H., Surf. Sci., 254, 153 (1991).30. Bergmans R., Ph.D. Thesis Eindhoven University of Technology The Netherlands,

(1996) p. 70.31. Overbury S.H., Bertrand P.A., and Somorjai G.A. Chem. Rev., 75(5), 547 (1975).32. Jansen W.P.A., Harmsen J.M.A., Denier van der Gon A.W., Hoebink J.H.B.J., and

Brongersma H.H., J. Catal., 204, 420 (2001).33. Chinchen G.C., Spencer M.S., Waugh K.C., and Whan D.A., J.Chem.Soc., Faraday

Trans. 1, 83, 2193 (1987).34. Brands D.S., Ph.D. Thesis, Amsterdam University (1998).35. V.d. Scheur F. Th, Brands D.S., V.d. Linden B., Oude Luttikhuis C., Poels E.K., and

Staal L.H., Appl. Cat., A116, 237 (1994).36. Jansen W.P.A., Ruitenbeek M., Denier van der Gon A.W., Geus J.W., and

Brongersma H.H., J. Catal., 196, 379 (2000).37. Yurieva T.M., Plyasova L.M., Kriger T.A., Zaikovskii V.I., Makarova O.V., and

Minyukova T.P., React. Kinet. Catal. Lett., 51, 495 (1993).38. Luys M.J., V. Oefelt P.H., Brouwer W.G.J., Pijpers A.P., and Scholten J.F.F., Appl.

Catal. 46, 161 (1989).39. Klenov D.O., Kryokova G.N., and Plyasova L.M., J. Mater. Chem., 8, 1665 (1998).40. Castricum H.L., Ph.D. Thesis Amsterdam University (2001) p. 103.

93

5A differentially pumped pressure cell for in situ LEIS analysis of

catalysts during reactions*

AbstractA differentially pumped pressure cell has been developed to enable in situ low-energy ionscattering (LEIS) analysis of catalysts during chemical reactions. The cell is fully compatiblewith an electrostatic analyzer and is, therefore, very well suited to study rough, highlydispersed catalysts. The pressure cell is a continuous flow cell (space velocity 1.2⋅103 s-1) andallows observation of dynamic surface reactions with various, perfectly mixed gases at fullycontrolled partial pressures and temperatures up to 800 K. The design decreases the pressuregap by three orders of magnitude. This offers ample opportunities such as the determinationof specific adsorption sites, surface coverages of different species during reactions, andinformation on growth processes, poisoning or chemically induced segregation. As anexample an in situ LEIS study of the CO oxidation over Pt is presented. By combining in situLEIS and quadrupole mass spectroscopy both the surface and gas composition could bemonitored during this reaction.

* The contents of this chapter has previously appeared in W.P.A. Jansen, A.W. Denier vander Gon, G.M. Wijers, Y.G.M. Rikers, P.W. van de Hoogen, J.A.M. De Laat, T.M. Maas,E.C.A. Dekkers, P. Brinkgreve, and H.H. Brongersma, Review Scientific Instruments, 73,354-361 (2002).

Pressure cell for ion scattering

94

1. IntroductionMany of the traditional techniques that can more selectively probe the nature of the

active site of catalyst surfaces, like XPS, secondary ion mass spectroscopy (SIMS), Augerelectron spectroscopy (AES) and low-energy ion scattering (LEIS) have mostly been appliedunder ultra high vacuum (UHV) conditions. Although examination before and after reactionusing these techniques has greatly contributed to our knowledge, it is difficult to probenucleation sites and transient phases in static studies of post-reacted catalysts. Adsorbatesmay completely restructure a catalyst surface and surface phases that are unstable in highvacua may well play a critical role in catalysis. Hence, dynamic studies of catalysts are key toa deeper understanding of mechanisms of reactions, surface structural evolution and theprocesses of activation and deactivation [1,2].

A very suitable technique to study catalysts is LEIS [3-6]. The high neutralizationprobability of the probing low-energy noble gas ions in combination with an electrostaticanalyzer (ESA) limits the information depth of this technique to a single atomic layer [4,5].Hence, using LEIS selective information on the gas – surface interface, exactly the placewhere catalysis takes place, can be obtained. In situ application of this technique offers ampleopportunities, such as the determination of specific adsorption sites, surface coverages ofdifferent species during reactions, and information on growth processes, poisoning orchemically induced segregation. Nevertheless, only a very limited number of LEIS studies[7-12] have been carried out in situ. Except for ref. [7], where the reaction of Br and O withW was studied, the purpose of all these studies was monitoring growth processes and theyapplied time-of-flight (TOF) instead of an ESA. For LEIS analysis of rough highly dispersedcatalysts an ESA is much more appropriate than a TOF [13].

Here we present a differentially pumped pressure cell to enable in situ LEIS analysisof catalysts during reactions with an ESA. The pressure cell is physically compatible with theexisting ERISS LEIS set-up [6] and does not restrict its operation. The design decreases thepressure gap by three orders of magnitude. With only minor modifications the cell can also beused for other surface science techniques such as SIMS and XPS. The subsequent sectionsdeal with the restrictions for environmental LEIS and the design features with respect tovacuum system and reactor design. Experimental results are added to illustrate possibilities ofenvironmental LEIS with the differentially pumped pressure cell.

2. Design differentially pumped pressure cell2.1 Limitations for environmental LEIS

The signal attenuation in a gas layer, which limits the maximum environmentalpressure in which LEIS can be successfully applied, depends on the type of gas layer and theprobing ions. As a rule of thumb, the cross-section for signal loss (σ) is proportional to

E1 of the probing ions (where E is the primary energy of the probing ions). Lin et al. [14]performed detailed studies of the interaction of low-energy ion beams with gases. For 10 keVHe+ ions in Ar gas they experimentally determined a cross-section for signal loss ofσ = 1.3⋅10-19 m2. In Fig. 1 we show the attenuation of the LEIS signal as a function of the gaspressure and path length of the ions in the reactor for σ = 1.3⋅10-19 m2 (note that the remaining

Chapter 5

95

LEIS signal equals 1-attenuation, as plotted in ref. [13]). Using this graph it is easy todetermine the limiting pressure for environmental LEIS analysis for a specific set-up, whichis determined by the path length of the ions and the signal attenuation of the ions that isacceptable in a specific experiment. For a typical path length of 0.1 m and an attenuation of90% this would result in 1⋅10-2 mbar. This presents the absolute limit possible for a givenpath length and acceptable attenuation. In practice, however, the pressure limit for LEIS isgenerally imposed by the maximum operating pressure of the multi-channel plates orchanneltrons that are included in most LEIS detectors [13]. The maximum operating pressureof channeltrons and multi-channel plates is in the low 10-4 mbar range [15], however, at suchpressures there is a serious risk for discharge. Therefore pressures below 10-6 mbar are used inpractice.

Fig. 1: The LEIS signal loss as a function of Ar pressure and path length of 10 keV He+ ions throughthe vacuum system, as calculated using the cross-section for ion-gas interactions determined by Linet al. [14]. The contour lines show constant signal attenuation, with the attenuation factor as shownin the graph.

Chemical reactions have successfully been studied at pressures in the 10-8 mbar rangeusing, LEED and AES [17], however, real-life catalysis takes place at pressures ≥ 1 bar (thedifference being referred to as the so-called pressure gap). To minimize the pressure gap, ahuge pressure gradient has to be accomplished between the sample and the detector. Thereactor and the detector cannot be separated by use of a window, however, since any windowwould completely neutralize the probing ions and obstruct LEIS analysis. Another way ofaccomplishing a pressure gradient is to use a nozzle to dose gases on the sample. A nozzlecould increase the pressure at a sample in a LEIS set-up by 1 or 2 orders of magnitude incomparison to the environment with the detector. However, a nozzle also introduces extremepressure gradients over the sample (~10% over ∅ 1 mm). To avoid strong pressure gradientsover the sample differentially pumping has been used here instead of a nozzle.

0.02

0.100.25

0.50

0.750.90

0.99

-4 -3 -2 -1 0

0.1

0.2

0.3

0.4

0.5

10-4 10-3 10-2 10-1 10-0

Path

leng

th (m

)

Pressure (mbar)


96

2.2 Vacuum system2.2.1 General principle of the differentially pumped pressure cell

Table 1 summarizes the objectives and the specifications of the differentially pumpedpressure cell, which is schematically shown in figure 2.

Fig. 2: Schematic view of the rotational symmetric LEIS analyzer and the differentially pumpedpressure cell. In this text, the phrase “(differentialy pumped) pressure cell” refers to the ensemblethat is enveloped by the dashed line, the so-called “reactor” is shaded.

Table 1. The objectives and specifications of the differentially pumped pressure cell.

• The pressure cell should be fully compatible with the existing UHV LEIS set-up that uses adouble toroidal analyzer. This requires the cell to be bakable, retractable, and in agreement withthe geometry of the ERISS LEIS detector.

• Pressures P ≤ 1⋅10-3 mbar at the sample in the reactor of the pressure cell correspond to pressuresP ≤ 1⋅10-6 mbar in the surrounding UHV

• At the center of the sample ∆P ≤ 1% over ∅ 3 mm• Computer controlled positioning with ±0.01 mm accuracy in X,Y,Z direction• Admittance of 2 perfectly mixed reactant gases with fully controlled partial pressure ratios• Effective suppression of wall reactions in the reactor of the pressure cell• Real time monitoring of the gas composition in the reactor during in situ LEIS• During in situ LEIS, temperatures in the range RT ≤ T ≤ 800 K, with stability better than ±0.6 K• Flashing up to 1500 K without breaking UHV before starting in situ LEIS• Thermometry during in situ LEIS with an absolute accuracy ±2 K and a relative accuracy ±0.1 K

ion beam

gas inlet

Analyzer

1 cm

scattered ions

Quadrupole Massspectrometer

bellows(B1)

heater reactorthermocouple

Analyzer

sample

Turbo Pump

scattered ions

pressure cell

bellows (B2)

Chapter 5

97

A bellows (bellows b1 in Fig. 2) mounted on a standard conflat 160 flange connectsthe pressure cell to the main vessel and allows the cell to be placed between the manipulatorcontaining the sample and the analyzer, and to retract it when it is not needed. A secondbellows (b2) allows the manipulator to move upward, contacting the bellows and therebysealing the reactor from the UHV environment. A gas inlet is provided to introduce reactantgases into the reactor. Holes and circular slits are provided to allow entrance and exit of theincident and scattered ions, respectively. A turbomolecular pump provides differentialpumping to the area directly above the reactor. A quadrupole mass spectrometer has beeninstalled in the UHV chamber for real-time monitoring of chemical reactions, and thesampleholder and the interior of the reactor are coated with titanium nitride to suppress wallreactions. In the reactor, samples can be heated and the sample temperature can be measuredusing either a thermocouple or fluoroptic thermometry.

2.2.2 Differential pumpingFigures 3a to 3c show cross-sections of the differentially pumped pressure cell and the

sampleholder, the numbers in parentheses refer to numbers in these figures. The sample (3) isplaced in the reactor. Since neither the incoming nor the outgoing ions should be blocked, thepressure cell has holes (∅ 1 mm) in the center for the incoming beam and conical circularslits at θ = 145° for the outgoing ions (see Fig. 2). The widths of the conical circular slits arechosen in accordance with the dimensions of the maximum incoming beam diameter (∅1 mm) and the acceptance of the analyzer (2°). The ESA uses only 320° of the azimuthalangle range, its dead angle has been used for a fluoroptic temperature probe (8) and theattachment of the central part of the circular slits. The requirement that the pressure over a ∅ 3 mm sample should be constant within 1% imposes, given the described geometry, aminimum height of 4 mm for the reactor. Given the 4 mm height, the most efficient way toobtain a gradient of 3 orders of magnitude is to use two restrictions in series. Starting fromthe sample and using 3 mm thick material for the first restriction (7), the conductance of thefirst beam hole C1 is 3⋅10-2 l⋅s-1 and the conductance of the first circular slit C2 is 1.4 l⋅s-1.Assuming an effective pump rate of 100 l⋅s-1 in the surrounding UHV, the pressure would besome 70 times lower at the detector than at the sample. To further reduce the pressure asecond restriction (9) has been applied. Using 8 mm material for the second restriction, theconductance of the second beam hole C3 is again 3⋅10-2 l⋅s-1. The conductance of the secondcircular slit C4 is 5.2 l⋅s-1. This is much higher than that of C2 since the second slit has a largerradius. The thickness of the first restriction (3 mm) and the height of the intermediate spacebetween the two restrictions (7 mm) have been chosen in a way to optimize the total pressuredrop. For the same reason, the tubing to the pump in the intermediate space is widenedoutside the circular slit in order to maximize the effective pump rate. In this way a 52 l⋅s-1

turbomolecular pump (Hycone 60) provides an effective pump rate of 15 l⋅s-1 between the tworestictions. This, in combination with an effective pump rate of ~100 l⋅s-1 in the UHV resultsin a pressure drop of 3 orders of magnitude between the sample and the UHV, provided theleakage from the bottom of the pressure cell into the UHV is negligible (<< 1.4 l⋅s-1).


98

Fig. 3: Cross-sections of the differentially pumped pressure cell (Figs. 3a and 3c) and thesampleholder (Fig. 3b). (1) manipulator; (2) sampleholder; (3) sample; (4) heat and electricalrestriction (alumina); (5) bellows allowing alignment and air tight sealing of the pressure cell; (6)gas inlet; (7) 1st restriction with beam hole for incoming and circular slit for outgoing ions; (8)fluoroptic fiber; (9) 2nd restriction with beam hole for incoming and circular slit for outgoing ions;(10) UHV; (11) analyzer; (12) bellows allowing positioning of the pressure cell; (13) UHV vessel;(14) 160 CF connecting the pressure cell to the UHV LEIS set-up; (15) guidance/alignment; (16)micro valve gas inlet; (17) connection with gas manifold; (18) approach switches; (19)thermocouple; (20) frame; (21) holder; (22) heater.

Chapter 5

99

2.2.3 SampleholderTo minimize the leakage from the bottom of the pressure cell into the UHV, a special

sampleholder has been designed. Figure 3b shows a schematic of this sampleholder, whichallows accommodation of samples up to ∅ 8 mm. The sampleholder consists of two parts, aholder (21) with a hole in its center to allow electron-beam heating (see section 2.3.3) and aframe (20) that can be sealed over a sample or a cup containing a (powder)sample. The frameitself can be sealed airtight to the holder using the elasticity of its three tabs. Both the holderand the frame have been fabricated from stainless steel and are coated with titanium nitride.As will be explained in section 2.3.2, this coating suppresses wall reactions. By pressing thetopside of the frame onto the alumina ring (4), at the bottom of bellows (5) the bottom of thepressure cell becomes sealed (leakage ~0.04 l⋅s-1) from the UHV. Note the leakage from thebottom of the pressure cell into the UHV is <<1.4 l⋅s-1.

2.2.4 Positioning and alignment of the differentially pumped pressure cellThe pressure cell can be retracted to the edge of the UHV chamber using bellows (12)

to allow compatibility with the existing LEIS set-up and to enable fast sample transferwithout breaking UHV. When the pressure cell has been retracted to the edge of the UHVchamber, samples can be transferred in and out the LEIS set-up via a loadlock, thus withoutbreaking the UHV. If the manipulator containing the sample (1) is positioned at its lowestheight, the pressure cell can be positioned between the sample and the analyzer. After thepressure cell has been positioned under the analyzer, the manipulator height can be increasedto seal the pressure cell from the UHV by pressing bellows (5). In this way the sample heightis put in the focal plane of the analyzer, while the bellows allow the accommodation ofdifferent sample heights.

Both the pressure cell and the sample are positioned using computer controlledmanipulators. The position of the manipulator containing the samples (1) can be positioned inall directions within 0.01 mm. The pressure cell can be automatically moved along a singleaxis from the edge of the UHV chamber to its working position between the sample and theanalyzer, and vice versa. When the pressure cell approaches its working position within0.5 mm, its precise position is fine tuned to within ±0.01 mm using a so-called approachswitch (18) (Baumer Electric type IWRM 12I9501). The position at the edge of the UHVchamber is less critical and is controlled with an end-switch. The movement of the pressurecell has been aligned to the existing set-up by tilting the guidance of the pressure cell (15).

In the ERISS LEIS set-up, secondary electrons that are produced by the primary ionbeam can be imaged, since the primary ion beam can be electrostatically deflectedsynchronous with imaging of the secondary electrons. Using secondary electron imaging, thesample position can be visualized when the pressure cell is retracted to the edge of the UHVchamber. When the pressure cell is positioned between the sample and the analyzer, thesecondary electrons show the position of the beam hole in the cell. This allows one to checkthe alignment of the sample and the pressure cell with a spatial resolution of some 10 µm.


100

2.3 Reactor design

2.3.1 Gas handlingOf essential importance for environmental LEIS is of course the transport of gases in

and out of the differentially pumped pressure cell. Molecular flow conditions assure perfectmixing of different gases. Therefore, a gas manifold has been designed that allows mixing oftwo reactant gases under molecular flow conditions (Fig. 4). To obtain molecular flowconditions from the point where the different gases meet, the different reactant gases are firstput in separate reservoirs (sphere A and B in Fig. 4). The spheres can be evacuated down to~10-2 mbar using a rotary pump, they can be filled using adjustable needle valves. Thepressure in each sphere can be measured gas independently with an absolute accuracy of4⋅10-2 mbar using capacitance diaphragm gauges (Varian CeramiCel VCMT12TAA). Eachsphere is connected to a ∅ 4 mm tube that is led over a cold trap (liquid nitrogen) towards amicro-valve. This home-built micro-valve (16) consists of a pinhole ∅ 15 µm that can beopened/closed by hand. The pinhole has been sunken in a plate to avoid deformation duringclosing. The small dimensions of the pinhole, together with the dimensions of the tubingbetween the pinhole and the pressure cell, assure molecular flow if the pressure in eachsphere is kept below 20 mbar. Then, gases are perfectly mixed and concentration gradientsare absent in the pressure cell. Pressures of 20 mbar in the spheres correspond to 3⋅10-4 mbarat the sample.

Fig. 4: Schematic of the gas manifold that allows the introduction of two different gases into thepressure cell at different pressures and for different pressure ratios. Via the bellows the manifold isconnected to the gas inlet of the pressure cell at (17) in Fig. 3c.

SphereB

SphereA

cold trap

Chapter 5

101

The pressure in the reservoirs can be measured easily, however, direct measurement ofthe pressure in the reactor containing the sample (3) is very difficult. By assuring molecularflow conditions in all parts of the gas inlet, the partial pressure ratio in the pressure cellbecomes equal to the pressure ratio of the reservoirs in the gas manifold [16]. The absolutepressure in the reactor as a function of the pressure in the reservoir spheres has beencalculated. The uncertainty in this calculation is largely determined by the uncertainty in thesizes of the pinholes in the microvalves in the gas inlet. The ratio between the calculated andmeasured value of this size appeared to be a factor 1.6. To minimize the error in thecalculations of the pressure in the reactor the experimentally determined size is used.Therefore, the calculated pressure in the reactor is expected to be reliable up to at least afactor 1.6.

To assure a stable pressure in the reactor of the pressure cell during experiments, boththe volume ratio and the pressure ratio between the spheres and the reactor are chosenrelatively high, 5⋅103 and ~105, respectively. Therefore, only 0.4% of the gas in the spheres isused per hour, which keeps the pressure drop during experiments at an insignificant level.

To obtain a homogeneous gas inlet into the reactor of the pressure cell, the gas is ledinto the reactor via a circular ring containing four holes. For homogeneity, the conductance ofthese holes (each 2.3⋅10-3 l⋅s-1) is 6 times lower than the conductance of the circular ring, andthe holes are divided over the φ direction in a way that corrects for the pressure drop over thering (at φ = 30°, 126°, 214°, and 298°, respectively). The gas outlet is determined by theleakage through the beam hole and the slit for the outgoing ions (effective pump rate 1.5 l⋅s-1).At 4 mm height, the total solid angle of these openings is constant within 1% over ∅ 3 mmaround the center of the sample. Together with the molecular flow conditions in the reactorthis results in pressure variations below 1% over the central ∅ 3 mm. The volume of thereactor is 1.2 cm3, hence the reactor has a very high refresh-rate (space velocity of 1.2⋅103 s-1).

2.3.2 Monitoring chemical reactions and suppressing wall reactionsOutside the pressure cell in the UHV, there is a quadrupole mass spectrometer (QMS

Balzers Prisma 200). Approximately 25% of the gases leaving the pressure cell leaks, undermolecular flow conditions, into the UHV. When, after saturation, adsorption and desorptionat the walls are in equilibrium, the ratios of the partial pressures of the different gases in thepressure cell can be measured using the QMS in the UHV. Since the dimensions of the UHVchamber are < 1 m and the average velocity of gas molecules at room temperature is typically~4⋅102 m⋅s-1, the gas concentrations measured in the QMS represent instantaneously theconcentration ratios in the pressure cell.

The most important problem that remains with regard to the quantification of thereactant and product gases is that of wall reactions [17]. Unless otherwise mentioned, theentire pressure cell has been made from stainless steel. To suppress wall reactions, thesampleholder and the interior of the reactor surrounding the sample have entirely been coatedwith titanium nitride, except for a 1 mm thick alumina ring (∅ 12.5 mm). The relatively inertalumina ring (4), that is attached to the base of bellows (5), provides a heat barrier andelectric isolation between the sampleholder and the pressure cell. The titanium nitride coated


102

surface will oxidize very fast, however, after oxidation titanium nitride is very inert, like gold,another material that is commonly applied to suppress wall reactions. However, titaniumnitride is much more durable and stable than gold. Gold molecules are rather mobile and maytherefore end up covering the sample. The use of TiN also reduces adhesion and preventsseizing.

It is also important to retract the pressure cell during sputtercleaning, otherwisematerial will be deposited on the interior of the pressure cell. Such material would probablybe active and cause undesirable background rates.

2.3.3 Heating and thermometryBy radiating the back of a sample from a Ta spiral (∅ 0.4 mm) underneath the

sampleholder (22), the sample temperature can be increased up to ~1300 K (note the filamentis mounted outside the reactor, in the UHV). Moreover, electron-beam heating can be appliedsince the voltage of the sample can be increased to +1 keV with respect to the rest of the set-up including the Ta spiral. Using electron-beam heating, temperatures up to 2500 K can beeasily reached. However, if the pressure cell is placed over the sample, the temperatures haveto be kept below 800 K because of the bellows (5). Since the pressure cell can be retractedand repositioned without breaking the UHV, electron-beam heating can be used for in situcleaning and restructuring (single crystal) metal samples before starting a measurementsession using the pressure cell.

The sample temperature can be determined using a K-type thermocouple that ispressed by a spring into a conical hole in the sampleholder (19). The difference between thetemperature as determined using the thermocouple and the actual sample temperature hasbeen calibrated using an in situ remote fluoroptic thermometer (8) (Luxtron 712). A fiberopticprobe has been mounted in the pressure cell (8) to allow in situ use of the remote fluoropticthermometer. The fluoroptic thermometer has an absolute accuracy of ±2 K, and stabilitybetter than ±0.1 K over 1 hour. In order to prevent electrical field disturbances during LEIS,the all silica probe (∅ 0.34 mm) of this thermometer has been jacketed with a titanium nitridecoated stainless steel tube (∅ 0.5 mm) that sticks out 1 mm. To probe the temperature closeto the sample center (at 2.7 mm) without physically blocking the in- and outgoing ions, theprobe has been placed in the dead angle of the ESA, bent under an angle of 45° towards thesurface. The only disadvantage of fluoroptic thermometry is that this method requires that aphosphorescent paste is put on the sample. However, calibration showed a reproducibledifference between the sample temperature measured using the fluoroptic thermometer andthe temperature as determined using the thermocouple (Fig. 5). Therefore, the fluoropticthermometer is only used to calibrate the thermocouple reading. The thermocouple providesthe input for a PID controller (Eurotherm 900 EPC) that is used to control the heating andsample temperature in the temperature range ~300 K < T < 800 K. In practice, a sample-temperature stability of ±0.6 K is obtained (typical 98% interval over several hours asdetermined with the fluoroptic thermometer). Figure 5 shows the relation between the realsample temperature (as determined by the fluoroptic thermometer) and the thermocouplesignal. The solid diamonds are measured in 2⋅10-4 mbar O2, the open circles in UHV, hence,

Chapter 5

103

the presence of gases in the pressure cell does not significantly influence the absolutetemperature, or the stability of the temperature.

Fig. 5: The relation between the sample temperature as obtained with a thermocouple pressed intothe sampleholder and the actual sample temperature as determined with a remote fluoropticthermometer. The open circles have been measured in UHV, the solid diamonds have been measuredin an oxygen atmosphere (PO2 = 2⋅10-4 mbar).

2.3.4. Spatial and time resolutionTwo important parameters with regard to the monitoring of chemical reactions are of

course both the spatial and time resolution of the detection system. Concerning the timeresolution: it takes typically a few minutes to collect a complete LEIS spectrum. If one is onlyinterested in the presence of one element, the time scale can be significantly reduced to about0.01 s by monitoring only the intensity at the top of a LEIS peak versus time. Any processrequiring faster analysis would need significant improvements of instrumentation.

The spatial resolution of LEIS is inherently limited by the damage of the samplesurface by the ion beam. The limitation is determined by the ion dose by which the surfacebecomes damaged and by the sensitivity of the analyzer and detector. In ref. [13] it has beenshown that for an EARISS type analyzer, which is applied in the ERISS LEIS set-up, theminimum allowed beam diameter is 3 µm if a dose of 1⋅1015 ions/cm2 is allowed. Any smallerbeam size would damage the surface significantly before the analysis is complete, thusresulting in non-surface specific information. To realize different beam spot diameters,apertures ranging from 3 mm down to 5 µm can be inserted in the beamline. The apertures areused as object for the lens that determines the beam spot diameter at the target and enablefocusing to a beam diameter of less than 5 µm at the sample surface.

400

450

500

550

450 500 550 600

TLuxtron (K)

T the

rmoc

oupl

e (K

)


104

2.4 A pressure cell for XPS and SIMSSo far the pressure cell has only been described in combination with a LEIS set-up,

however, the principle can also be used in combination with, e.g., XPS or SIMS. To do so,the holes for the incoming and outgoing particles should be adapted to the geometry of theXPS/SIMS set-up. Since these holes determine the pressure gradient, the height of the reactormay have to be changed as well. For the rest, the design can be used as it is. A calculation forthe geometry of a ESCA200 –a wide spread XPS analyzer from Gammadata AB, formerlyScienta– shows that a decrease of the pressure gap of 3 orders of magnitude is feasible.

3. First resultsTo check the alignment of the pressure cell, LEIS measurements have been performed

comparing a gold sample outside the pressure cell in UHV and the same sample inside thepressure cell in an argon atmosphere (at various pressures up to 2.4⋅10-4 mbar). Since the inertsample and the inert gas will not react, the LEIS signals should remain constant. The LEISspectra in Figure 6 show that indeed both the peak energy, the FWHM of the peak, and thegold yield/nC (from which the gold surface concentration is determined) remain constantwithin the experimental errors (<6 eV, <12 eV and <2%, respectively). Hence, differences inLEIS spectra between samples kept in UHV and in gaseous environments -up to totalpressures of at least 2.4⋅10-4 mbar- can be completely ascribed to changes at the gas - surfaceinterface due to the interaction with the gaseous environment. Regarding Fig. 1 and thepathlength in the pressure cell, signal losses due to the gaseous environment are expected tobecome significant at pressures P ≥ 1⋅10-3 mbar.

Fig. 6: 3 keV 4He+ LEIS spectra of Au measured in UHV (1⋅10-10 mbar) and in Ar at 8⋅10-5 mbar and2.4⋅10-4 mbar.

0

125

250

375

500

2300 2600 2900

Energy (eV)

LEIS

yie

ld (c

nts/

nC)

Au

Chapter 5

105

The following experiment illustrates the possibilities of environmental LEIS with thepressure cell. In this experiment O2 (PO2 = 8⋅10-5 mbar) and CO (PCO = 3⋅10-5 mbar) areconverted to CO2 over Pt at 524 K. Before starting the reaction, the Pt has been cleaned by so-called vacuum oxy-polishing (1 hour at 1073 K in PO2 = 1.5⋅10-7 mbar) followed by flashingin UHV up to ~1400 K. The 3 keV 3He+ LEIS spectrum taken before the reaction (solid lineFig. 7a) shows that in this way clean Pt is obtained. A second LEIS spectrum (open circlesFig. 7a), was measured during reaction. The QMS signals of O2, CO and CO2 in Fig. 7b showthat the CO2 production dies within ~15 minutes. Comparison of the LEIS spectra takenbefore and during reaction shows that the Pt gets poisoned by Ni. The Ni poisoning can beexplained by reaction with Ni(CO)4 which is formed if CO gas gets in contact with stainlesssteel (gas bottle and tubing). However, if the cold trap in the gas inlet (Fig. 4) is used, theNi(CO)4 is effectively removed from the CO gas.

Figs. 7a,b: On the left (7a), 3 keV 3He+ LEIS spectra obtained before (solid line) and during the COoxidation (open circle) over Pt. The LEIS measurements clearly show that the Pt gets poisoned by Ni.On the right (7b), QMS spectra showing the concentration of masses 28 AMU (CO), 32 AMU (O2)and 44 AMU (CO2). After ~15 minutes the CO2 production has virtually stopped because ofpoisoning of the Pt.

Figures 8a and 8b show the LEIS and QMS signals under the same reaction conditionswhen the CO gas is let over a cold trap. Now the conversion remains constant for more than 1hour (53%) and the LEIS spectra do not show any sign of poisoning by Ni.

The experiments described show that the differentially pumped pressure cell enablesin situ LEIS analysis. The pressure cell allows for in situ LEIS at 10-3 mbar, while thepressure at the multi-channel plates is kept at 10-6 mbar. If the multi-channel plates are usedin the low 10-4 mbar range (risking a discharge) the pressure cell allows for in situ LEIS at10-1 mbar. In spite of the short pathlength in the reactor (1 cm) a pressure of 1·10-1 mbarcorresponds to signal losses of 93% or 45% in the gas layer for 10 keV He+ and Ne+,

0

200

400

600

500 1500 2500

Energy (eV)

LEIS

yie

ld (a

.u.)

O

Ni

Pt

CaSix2

0E+0

1E-8

2E-8

3E-8

-5 0 5 10 15 20Time (min.)

I QM

S (A

)

O2

CO

CO2


106

respectively. Hence, even though the pressure cell allows one to perform in situ LEIS thereremains a pressure gap between LEIS and catalysis. Although the pressure cell presented herereduces this pressure gap by three orders of magnitude, the gap between the LEIS analysisand industrial relevant catalysis conditions is still very large. However, even at pressuresP ≤ 10-3 mbar there are already ample opportunities for in situ LEIS that can considerablycontribute to our knowledge of catalysis such as, the determination of specific adsorptionsites, surface coverages of different species during reactions, and information on growthprocesses, poisoning or chemically induced segregation.

Figs. 8a,b: On the left (8a) 3 keV 3He+ LEIS spectra obtained before (solid line) and during the COoxidation (open circles) over Pt. The absence of Ni in the spectra confirms the effectiveness of thecold trap. On the right (8b) QMS spectra showing the concentration of masses 28 AMU (CO), 32AMU (O2) and 44 AMU (CO2). Using a cold trap to remove Ni(CO)4 a constant CO2 production(conversion 53%) is obtained.

AcknowledgementsFinancial support of this work by the Netherlands Organization for Scientific Research(NWO) is gratefully acknowledged.

0E+0

1E-8

2E-8

3E-8

-10 0 10 20 30 40 50 60

I QM

S (A

)

O2

CO2

CO0

200

400

600

500 1500 2500Energy (eV)

LEIS

yie

ld (a

.u.)

O

Pt

Ca

x2

Chapter 5

107

References

1. Somorjai G.A., Topics in Cat., 8, 1 (1999).2. Gai P.L., Topics in Cat., 8, 97 (1999).3. Menon P.G., Appl. Cat. A207 N2 (2001).4. Taglauer E., in “Fundamental Aspects of Heterogeneous Catalysis Studied by Particle

Beams” (H.H. Brongersma and R.A. van Santen, Eds.), NATO ASI B 265 p. 301.Plenum Press, 1991.

5. Brongersma H.H., Groenen P.A.C., and Jacobs J.-P., in “Science of Ceramic InterfacesII”, (J. Nowotny Eds.) p. 113. Elsevier, 1994.

6. Jansen W.P.A., Ruitenbeek M., Denier van der Gon A.W., Geus J.W. and BrongersmaH.H., J.Catal., 196, 379 (2000).

7. Brongersma H.H., Ligt G.C.J. van der, and Rouweler G., Philips J. Res., 36, 1, (1981).8. Kraus A.R., Auciello O., Lamich G.J., Gruen D.M., Schultz J.A. and Chang R.P.H., J.

Vac. Sci. Technol., A12, 1943 (1994).9. Kraus A.R., J. Im, Schultz J.A., Smentkowksi V.S., Waters K., Zuiker C.D., Gruen

D.M., and Chang R.P.H., Thin Solid Films, 270, 130 (1995).10. Lin Y., Kraus A.R., Auciello O.H., Nishino Y., Gruen D.M., Chang R.P.H., and J.A.

Schultz, J. Vac. Sci. Technol., A12, 1557 (1994).11. Fujii Y., Nakajima K., Namuri K., Kimura K., M. Mannami, Surf. Sci., 318, L1225

(1994).12. Pfanzelter R., Igel T., Winter H., Surf. Sci., 375, 13 (1997).13. Denier van der Gon A.W., Cortenraad R., Jansen W.P.A., Reijme M.A. and Brongersma

H.H., Nucl. Instr. And Meth., B161-163, 56 (2000).14. Lin Y., Kraus A.R., Krauss R.P.H., Chang R.P.H., Auciello O.H., Gruen D.M. and

Schultz J.A., Thin Solid Films, 253, 247 (1994).15. Philips Photonics, Electron multipliers, Philips Export B.V. (1990)16. Rapp D. and Englander-Golden P., J. Chem. Phys., 43, 1464 (1965).17. Blakely D.W., Kozak E.I., Sexton B.A. and Somorjai G.A., J. Vac. Sci. Technol., 13,

1091 (1976).

109

6Surface coverages during CO oxidation over Pt(110)

-An in situ LEIS study*

AbstractThe CO-oxidation over Pt(110) has been investigated under steady state and oscillatory

conditions with a combination of in situ low-energy ion scattering and quadrupole massspectroscopy. This allowed real time monitoring of the CO and oxygen surface coverages, theamount of free Pt surface sites and the gas concentrations of the reactants and products.Experiments determining the CO and oxygen coverages as a function of temperature forO2/CO ratios varying between 1 and 18 (total pressure 1.1⋅10-4 mbar) show that the onset ofthe steady state CO2-production is coupled to a critical CO coverage at which both the COand oxygen sticking probability change dramatically. The critical coverage corresponds tothat required for the start of the Pt(110)-(1×2) to Pt(110)-(1×1) transition. Experiments underoscillatory conditions show that this critical coverage also plays a key role in the ignition ofthe oscillatory CO2-production.

* The contents of this chapter has been submitted for publication: W.P.A. Jansen, Y.G.M.Rikers, A.W. Denier van der Gon, and H.H. Brongersma, Surface Science

Surface coverages during CO oxidation over Pt(110)

110

1. IntroductionNumerous investigations on the platinum catalysed CO oxidation have been

performed since the classical work by Langmuir [1]. Despite the apparent simplicity of thebasic mechanism, the reaction often exhibits complex behavior. For instance Engel and Ertl[2] determined that the CO2 production is suddenly ignited under conditions that presumablycorrespond to a dramatic change in the sticking probability of CO and O2 on Pt. Besides aconstant steady state CO2 production the reactant rate may also oscillate. The first reports onthe oscillatory behavior of the CO oxidation date from the early seventies [3,4]. Oscillatorybehavior of the CO oxidation has been observed on polycrystalline Pt and even on supportedcatalysts, however, the most extensive studies have been carried out on well-defined Pt(100)and Pt(110) single crystals [5]. To study the oscillatory behavior, changes in work functionshave been monitored in situ with Kelvin probes [6-8] to probe the adsorption of oxygen andCO. Furthermore, many studies have applied differentially pumped mass spectrometers tomonitor gas concentrations [7,8]. Photo-Emission Electron Microscopy (PEEM) [9-11],Ellipso-Microscopy for Surface Imaging (EMSI) [11,12] and Reflection AnisotropyMicroscopy (RAM) [11,12] have been successfully used to investigate pattern formation forthe oscillatory CO oxidation over Pt(110). Low-Energy Electron Diffraction (LEED)[6-8,13], and Auger Electron Spectroscopy (AES) [7,8] have been used to study the Pt(110)surface to elucidate whether changes in the catalyst surface (phase) play a role in the reaction.However, these techniques could not be used successfully to follow oscillations on Pt(110)because the probing electrons disturb the oscillations [14].

None of the aforementioned in situ studies provided quantitative information on theCO covered fraction of the Pt (θCO) during the oscillatory CO oxidation on Pt(110). HoweverEiswirth and Ertl have indirectly deduced that the oscillations occur near the completion(θCO ≈ 50 % of a monolayer) of the CO induced (1×2)→(1×1) structural transformation of thesurface [6]. A monolayer being defined as the surface atom density of the Pt(110) (1×1)structure, i.e. 0.92⋅1015 atoms/cm2 [15]. Their conclusion is based on LEED studies duringthe completion of the CO-induced (1×2)→(1×1) transition on a Pt(110)-surface in theoscillatory regime, however, in the absence of oscillations.

Recently, we have constructed a differentially pumped pressure cell that allows in situlow-energy ion scattering (LEIS) during catalytic reactions [16]. In this paper we use this cellto study the CO and oxygen coverages of Pt(110) during both steady state and oscillatory COoxidation. At the same time, a quadrupole mass spectrometer was used to monitor the gasconcentrations of the reactants and products. The experiments give new, quantitative resultson the O and CO coverage during CO oxidation on Pt(110).

2. Experimental2.1 Cleaning/activation Pt(110)

A Pt(110) sample (∅ 8 mm and 1 mm thick) with a miscut angle of 0.11°, asevidenced from X-Ray Röntgen diffraction, was used for the experiments. The (110)orientation of the sample was verified at the start and end of the experiments with Lauediffraction. The Pt(110) was cleaned and activated by oxy-polishing treatment, which was

Chapter 6

111

carried out by heating the crystal in 2⋅10-7 mbar O2 at 1000 K for 1h. This enhances oxygen-induced segregation of bulk impurities to the Pt surface, and carbonaceous impurities areburnt off in the form of CO2. After heating in oxygen, a PtO overlayer is formed thatdecomposes at temperatures T > 1270 K [17]. To remove the PtO overlayer which formsduring the oxypolishing, the Pt is flashed repeatedly in UHV at T ~ 1470 K (see picture frontcover) as determined with a pyrometer, subsequently to the oxygen-treatment. Thistemperature is also high enough to order the Pt(110) single crystal.

2.2 In situ LEISLEIS (also known as Ion Scattering Spectroscopy (ISS)) [18,19] has been used as the

main analysing technique. Normally this technique is solely applied in UHV. However, witha differentially pumped pressure cell LEIS can be performed in gaseous environments up topressures of ~10-3 mbar. Based on the cross-sections for signal attenuation in CO, oxygen andAr [20] and the experiments with a Au target in Ar gas, the gas layer is not expected tosignificantly affect the LEIS signals up to ~10-3 mbar in our set-up [16].

The set-up is schematically shown in Fig. 1. A bellows (bellows b1 in Fig. 1) mountedon a standard conflat 160 flange connects the pressure cell to the main vessel and allows thecell to be placed between the manipulator containing the sample and the analyser, and toretract it when it is not needed. A second bellows (b2) allows the manipulator to moveupward, contacting the bellows and thereby sealing the reactor from the UHV environment.A gas inlet is provided to introduce reactant gases into the reactor.

Fig. 1 schematic representation of the LEIS analyser and the differentially pumped pressure cell. Thepart that is enveloped by the dashed line is the differentially pumped pressure cell; b1 and b2 arebellows; the shaded area above the sample is the reactor; Fibre denotes a fluoroptic fibre that can beused to probe the sample temperature.

Ion beam

Gas inlet

Analyser

1 cm

Scattered ions

Quadrupolemassspectrometer

b1

Heater

Thermocouple

Analyser

sample

Turbo Pump

Scattered ions

Pressure cell

b2

Fibre

Ion Source


112

Since molecular flow conditions are applied in the entire gas inlet, gases in the pressure cellare perfectly mixed. The pressure and volume ratios between the gas reservoirs and the gasconsumption assure that the pressure drop in the reservoirs, and hence in the cell, are smallerthan 0.4% h-1. Holes and circular slits are provided to allow entrance and exit of the incidentand scattered ions respectively. A turbomolecular pump provides differential pumping to thearea directly above the reactor. The sampleholder and the interior of the reactor are coatedwith titanium nitride to suppress wall reactions. Samples can be heated up to 800 K when thepressure cell is in use, and up to 2500 K when the pressure cell is retracted. The sampletemperature can be measured using either a thermocouple or fluoroptic thermometry. Thethermocouple provides input for a PID controller (Eurotherm 900 EPC) that is used to controlthe heating and sample temperature in the temperature range ~300 K < T < 800 K. In this waya sample temperature stability of ±0.5 K is obtained. In this study the LEIS yield of Pt hasbeen monitored with a time resolution of ~0.1 s.

It is generally accepted [2,21] that a CO molecule is chemisorbed through the carbonatom, the molecular axis being perpendicular to the platinum (110) surface plane. Thisimplies that one cannot distinguish between adsorbed CO and O with LEIS, since in bothcases only the oxygen atoms would be visible. To solve this problem experiments have beencarried out using normal C16O and isotopically labelled 18O2. Since LEIS can separate signalsof masses 16 and 18 this allows one to monitor adsorbed CO separately frommolecularly/atomically adsorbed oxygen species.

To enable quantitative determination of oxygen surface coverages, the oxygen and PtLEIS peak areas of an oxygen saturated Pt(110) surface (1.38⋅1015 O-atoms/cm2) weremeasured (i.e. 1½ monolayers see e.g. [17]). Similarly the O and Pt LEIS peak areas of COsaturated Pt(110) (0.92⋅1015 CO-atoms/cm2, i.e. 1.0 monolayers [15]) were determined. Thesesurfaces were obtained by cleaning Pt(110) (see section 2.1) and subsequently dosing ofeither O2 or CO (at 391 K and PO2 or PCO = 5.4⋅10-5 mbar) in the pressure cell. Using LEISthe absence of contamination and the state of saturation have been verified. For quantificationof the intermediate CO and O coverages, LEIS peak areas have been compared with thesereference spectra, while assuming the intensity scales with the coverage see e.g. [19].

2.3 Monitoring chemical reactionsA quadrupole mass spectrometer (QMS Balzers Prisma 200) is mounted in the UHV,

outside the pressure cell. Approximately 25% of the gases leaving the pressure cell leaks,under molecular flow conditions, into the UHV. When, after saturation, adsorption anddesorption at the walls are in equilibrium, the ratios of the partial pressures in the pressurecell can be measured using the QMS. Since the dimensions of the UHV chamber are <1 mand the average velocity of gas molecules at room temperature is ~4⋅102 m/s the measuredgas concentrations in the UHV represent instantaneously the gas concentrations in thepressure cell.

Chapter 6

113

3 Results

3.1 Steady state CO oxidationThe CO2-production on a Pt(110)-surface has been studied as a function of

temperature and gas composition. The CO partial pressure (PCO) was varied between5·10-6 and 5·10-5 mbar, while the O2 pressure (PO2) was adjusted to keep the total pressure at1.1·10-4 mbar (hence 1.0 ≤ PO2 / PCO ≤ 18). Figure 2 shows that for each composition thereaction rate increases rapidly above a certain so-called ‘step temperature’ Tstep. At Ptotal =1.1⋅10-4 mbar the step temperature scales with log(PCO). Alternate measurements below andabove Tstep show that the reaction can be reversibly stopped and ignited. Hence, no permanentpoisoning occurs. At temperatures above Tstep, the CO2-production decreases slightly withincreasing temperature, because the adsorption of CO and O2 decreases with increasingtemperature [22].

Fig. 2 The CO2-production as a function of Pt(110) temperature for various PCO (5⋅10-6 mbar ≤ PCO ≤5⋅10-5 mbar) balanced with O2 to 1.1⋅10-4 mbar total pressure. The “step temperature” thatcorresponds to the onset of the CO2 production scales with log(PCO).

The ‘step behavior’ is also known from literature (see e.g. ref. [2]) where it issuggested to be linked with a change in the surface structure and as a consequence thesticking probability. To investigate this, LEIS spectra have been taken at temperatures aroundTstep for the gas compositions mentioned above and Ptotal = 1.1⋅10-4 mbar. As an example, Fig.3 shows 3 keV 3He+ LEIS spectra obtained around Tstep, at 456 K and 459 K for PCO =2.7·10-5 mbar and PO2 = 8.1·10-5 mbar. The LEIS spectra show that below Tstep oxygen isabsent (ϑo remains below the detection limit, see Table 1), whereas 16O (hence CO) ispresent. LEIS is not very sensitive to C, but the complete absence of a C signal in the LEISspectra, both for the reference CO system as well as for the surfaces just below Tstep, confirms

0

20

40

60

350 450 550Pt(110) temperature (K)

CO

2 pro

duct

ion

(%)


114

that, at the addressed temperatures and pressures, CO is chemisorbed through the carbonatom, with the oxygen atoms directed outward from the platinum surface plane.Measurements at different pressures, and hence different step temperatures (see table 1 forconditions), show that at Tstep the critical CO coverage ( c

COϑ ) is 23±2% of a monolayer. Atthis c

COϑ , any further temperature increase causes a drop of the CO coverage from 23±2% to3±1% of a monolayer while the oxygen coverage increases from below the detection limit of1.5%, to 25±2% of a monolayer. Simultaneously the CO2-production is ignited as observedfrom the QMS signal. Table 1 summarizes the surface coverages 1.5 K below and 1.5 Kabove Tstep.

The concentration of uncovered Pt determined with LEIS just above Tstep is 3.2 ± 0.5times higher than just below Tstep. This ratio stems not only from the changes in the totalcoverage, but also from the difference in shielding of the Pt LEIS signal between CO and Oadatoms. CO adatoms reside at on top sites on Pt (110) and therefore shield the Pt much moreeffectively than O adatoms that adsorb at so-called valley sites [23]. The difference inadsorption sites is responsible for the difference between the oxygen LEIS yieldscorresponding to CO and O respectively. Using the reference spectra described in section 2.2the reported CO and O coverages have been corrected for these differences. Themeasurements show that the onset of the CO2 production corresponds undoubtedly to adramatic change from 23±2% of a monolayer CO to 25±2% of a monolayer O. Only asurface that changes from one configuration into another can explain this result.

Fig. 3 Two 1.5 keV 3He+ LEIS spectra taken during C16O oxidation with 18O2. One spectrum isdetermined just below at 456 K (solid curve) and one just above at 459 K (dotted curve) the Tstep.

0

40

80

120

550 700 850

Energy (eV)

LEIS

sig

nal (

cnts

/nC

)

18O

16O

456 KX 1/2

16O

459 K

(CO)

(O2)

Chapter 6

115

Table 1 The CO and O surface coverages 1.5 K below and 1.5 K above Tstep (371 K ≤ Tstep≤ 490 K and 5.1⋅10-6 mbar ≤ PCO ≤ 5.0⋅10-5 mbar while PO2 was varied to maintain Ptotal at1.1⋅10-4 mbar).

Temperature θO (% of a monolayer) θCO (% of a monolayer)Tstep – 1.5 K < 1.5 % 23 ± 2 %Tstep + 1.5 K 25 ± 2 % 3 ± 1 %

3.2 Oscillatory behavior CO oxidationTo obtain oscillatory behavior, cleaned Pt(110) was allowed to cool down to the

desired reaction temperature of 525.2 K. At the reaction temperature oxygen was led in (Po =5.7⋅10-5 mbar) and the sample was allowed to stabilise for 1 h. Then, CO was introduced instepwise increasing pressures until self-sustained kinetic oscillations were observed.

Figure 4 shows the oxygen, CO and CO2 partial pressures and the Pt LEIS signal, ameasure for the amount of free Pt surface area, as a function of time. The figure shows thatthe oscillatory behavior starts with spikes. We attribute these spikes to short flips of thesurface structure to another state. Such transitions can be triggered by surface defects [14]. Att = 310 min the CO partial pressure is increased to 5.6⋅10-5 mbar, resulting in a 17% decreaseof the Pt LEIS signal and a continuous oscillatory behavior. After a further increase of thePCO to 5.8⋅10-5 mbar at t = 351 min (not shown in Fig. 4) the oscillations levelled outirreversibly. The latter PCO is 6.0% higher than the PCO at which the spikes started to occurand only 3.0% higher than the pressure at which the continuous oscillatory behavior started.Hence, our observations confirm that Pt(110) has a very narrow oscillatory region. It shouldbe noted that switching the 3He+ probing ion beam on and off does not influence the spikes orthe oscillatory behavior. Experiments at different currents showed that the measured surfacecoverages and QMS signals are not influenced if the fluence is kept below 1⋅1013 ions·s·cm-2.

The QMS measurements show that the reaction, the spikes and the oscillationsproceed according to the stoichiometry of the CO oxidation reaction (hence CO : O2 : CO2 =2 : 1 : -2). Moreover, it can be observed that the CO2 production is in phase with the amountof free Pt surface sites. The peak to peak variation of the oscillations of the Pt concentrationis only 14±4 % and the Fourier transformed frequency spectrum shows no sharp peak. Thepresence of several frequencies and the 14% peak to peak variation indicate that the observedoscillations have a local character. For a surface that oscillates completely synchronized theconcentration of free Pt surface sites could vary at maximum between 0 % and 100 % of amonolayer, hence, the peak to peak variation demonstrates that minimally 14±4 % of thesurface oscillates. Since a CO saturated Pt surface has a 92% lower Pt LEIS yield than a cleanPt surface, a 17% drop of the amount of free Pt surface sites at the start of the oscillationsindicates participation of minimal 18±3% of a monolayer. This, together with the measuredchanges in the overall CO coverage from below the detection limit of 1.5% (before the startof the oscillations) up to <4±1% during the oscillations, shows that the maximum COcoverage in the oscillatory region is 24±7%.


116

Figs. 4a,b Overview (4a at top) and close-up (4b at bottom) of the reactant and product partialpressures (upper three curves, left y-axis) and the Pt-LEIS signal (curve at the bottom, right y-axis) asa function of time during the oscillatory CO oxidation. At t = 310 min the PCO is increased to5.62⋅10-5 mbar, at t = 325 min PCO is further increased to 5.71⋅10-5 mbar. The temperature and PO2

are kept constant at 525.2 K and 5.68⋅10-5 mbar, respectively.

1.E-05

3.E-05

5.E-05

314 316 318 320Time (min)

Part

ial p

ress

ure

(mba

r)

2.5E+05

4.5E+05

6.5E+05

LEIS

sig

nal (

cnts

/nC

)

Pt-peak

CO2

O2

COT = 525.2 K

1.E-05

3.E-05

5.E-05

295 310 325 340Time (min)

Part

ial p

ress

ure

(mba

r)

2.5E+05

4.5E+05

6.5E+05

LEIS

sig

nal (

cnts

/nC

)

Pt-peak

CO2

O2

COT = 525.2 K

Chapter 6

117

4. Discussion and conclusions

In agreement with literature [2,21] in situ LEIS shows that CO is adsorbed on Pt(110)through the carbon atom, the molecular axis being perpendicular to the platinum surfaceplane. In addition, in situ LEIS allows to determine the reactant coverages and the amount offree surface area during reaction. This has been used to verify the assumption of Engel andErtl [2] that the abrupt ignition of the CO oxidation corresponds to a change of the surfacestructure and with it the sticking probabilities of CO and oxygen. Our experiments show thatthe CO2 production ignites at a critical CO coverage of 23±2% of a monolayer. Upon ignitionof the reaction the CO coverage drops to 3±1% of a monolayer and the oxygen coverageincreases from below the detection limit of < 1.5% of a monolayer up to 25±2% of amonolayer. The fact that these processes occur within a sharp temperature window of lessthan 3 K together with the earlier reported critical CO coverage of 20% for the change fromthe Pt(110)-(1×2) into the Pt(110)-(1×1) structure (e.g. [23,24]), indicate that the start of thelatter phase transition is responsible for this.

The surface coverages could also be monitored during the oscillatory CO oxidation overPt. Hence, in contrast with the probing electron beams used in various electron spectroscopies[14] the probing ions used in LEIS do not disturb the oscillatory behaviour, provided that theion fluence is kept below 1⋅1013 ions·s·cm-2. During the oscillatory CO oxidation the amountof free Pt surface sites is in phase with the measured CO2 production. Determination of theCO coverage during the oscillatory reaction shows that the upper limit for CO coverageduring the oscillatory CO oxidation is less than 24±7% of a monolayer instead of 50% of amonolayer as assumed previously [6]. Hence, our quantitative measurements support that,similarly to the steady state situation, the onset of the oscillatory CO2 production may well beconnected to the start of the phase transition of Pt(110)-(1×2) into Pt(110)-(1×1).

AcknowledgementsThe authors wish to gratefully acknowledge S. Carabineiro and B. Nieuwenhuys for theiradvice concerning establishing oscillatory behavior. The Netherlands Organization forScientific Research (NWO) is acknowledged for financial support.

References1. Langmuir I., Trans. Faraday Soc., 17, 607 (1922).2. Engel T. and Ertl G., Adv. In Cat., 28, 1, (1979).3. Hugo P., Ber. Bunsenges. Phys. Chem., 74, 121, (1970).4. Beusch H., Fieguth P., and Wicke E., Chem.-Ing.-Tech., 44, 445 (1972).5. Ertl G. Adv. in Cat., 37, 213, (1990).6. Eiswirth M. and Ertl G., Surf. Sci., 177, 90, (1986).7. Eiswirth M., Moller P., Wetzl K., Imbihl R., and Ertl G., J. Chem. Phys., 90, 1,

(1989).8. Imbihl R., Ladas S., and Ertl G., J. Vac. Sci. & Techn. A, 6(3), 877, (1988).


118

9. Nettesheim S., Von Oertzen A., Rotermund H.H., and Ertl G., J. Chem. Phys., 98, 12,(1993).

10. Eiswirth M., Bär M., Rotermund H.H., Physica D, 84, 40, (1995).11. Rotermund H.H., Haas G., Franz R.U., Tromp R.M. and Ertl G., Applied Physics,

A61(6), 569, (1995).12. Dicke J., Rotermund H.H., and Lauterbach J., J. Opt. Soc. Amer., A17(1), 135, (2000).13. Falta J., Imbihl R., Sander M., and Henzler M., Phys. Rev. B, 45(12), (1992).14. Rotermund H.H., Surf. Sci. Rep., 29, 265, (1997).15. Jackman T.R., Davies J.A., Jackson D.P., Unertl W.N., and Norton P.R., Surf. Sci., 120,

389, (1982).16. Jansen W.P.A., Denier van der Gon A.W., Wijers G.M., Rikers Y.G.M., Van de

Hoogen P.W., De Laat J.A.M., Maas T.M., Dekkers E.C.A., Brinkgreve P., andBrongersma H.H., Rev. Sci. Instr., 73, 354 2002).

17. Ducros R., and Merill R.P., Surf. Sci., 55, 227 (1976).18. Taglauer E., in “Fundamental Aspects of Heterogeneous Catalysis Studied by Particle

Beams”, (Eds. H.H. Brongersma and R.A. van Santen) NATO ASI B Plenum NewYork, Vol. 265, p. 301, (1991).

19. H.H. Brongersma, P.A.C. Groenen, and J.-P. Jacobs, in “Science of Ceramic InterfacesII” (Editor J. Nowotny) Elsevier, p. 113. (1994).

20. Lin Y., Kraus A.R., Krauss R.P.H., Chang R.P.H., Auciello O.H., Gruen D.M. andSchultz J.A., Thin Solid Films, 253, 247 (1994).

21. Van Santen R.A., and Gelten R.J., Lectures Surface Chemistry, Lecture Notes, 15,(1996).

22. Kapteijn F., Moulijn J.A., Van Santen R.A. and Wever R., in “Catalysis an integratedapproach”, (eds. R.A. van Santen, P.W.N.M. van Leeuwen, J.A. Moulijn and B.A.Averill) Elsevier, Amsterdam p.81 (1999).

23. Imbihl R., in “Fundamental Aspects of Heterogeneous Catalysis Studied by ParticleBeams”, (Eds. H.H. Brongersma and R.A. van Santen) NATO ASI B Plenum NewYork, Vol. 265, (1991).

24. Hoffmann P., Bare S.R., and King D.A., Surf. Sci., 117, 245 (1982).

121

Summary

Catalysts and the products made with them are omnipresent. Virtually all polymersare made with a catalytic process, as are a multitude of other chemicals ranging from fuels tocommodity chemicals and from pharmaceuticals to pesticides. Besides, most of ourautomotive exhaust gases are cleaned by a catalyst. Catalysts selectively accelerate chemicalreactions without being consumed themselves. For a large group of catalysts, calledheterogeneous catalysts, a catalytic cycle can be described as follows: reactants adsorb on thecatalyst surface where they are activated and react and the cycle is completed when theformed products desorb from the catalyst surface, leaving the (unchanged) catalyst behind fora new catalytic cycle. It may be clear that the surface of the catalyst plays a key role in this.To obtain a better understanding of the catalytic action several studies relating the surfacecomposition to catalytic performance have been carried out for this thesis. A great part of thepresented surface structural information was obtained using the LEIS analysis technique.Since new methodology and new instrumentation had to be developed in order to obtain thisinformation part of the thesis is devoted to these developments.

The first Chapter deals with the influence of surface roughness and compaction onLEIS. Since supported catalysts, which are powders with specific surface areas up to as highas 1000 m2/g, are frequently used in catalysis knowledge of this influence is very important.For this study Ta2O5/SiO2 with Ta2O5 exclusively present in the outermost atomic layer wasprepared using atomic layer deposition. By comparing LEIS analyses of this material aftercompaction at various pressures, we could show that the routinely applied compaction ofpowders does neither affect the composition of the outermost surface layer nor its LEISanalysis in the addressed range from 0 Pa to 300 MPa. In the past, people had alreadyevaluated the influence of surface roughness on LEIS, however, their results did not strokewith ours. The effect was overestimated because of incomplete neutralization and a 2dimensional approach in modeling. Both our experiments and our 3 dimensional modelingapproach show that LEIS signals represent the projection of oxidic particles onto theirequatorial plane. This means that in the case of silica, the most widely used catalyst supportbesides alumina, the effect of surface roughness is minor. The surface roughness factor Rcomparing compacted silica powders and atomically flat silica equals R = 0.87±0.06 whenusing a LEIS setup with a scattering angle of 145°. For other widespread geometries, usingscattering angles in the range 135° – 145°, slight deviations (<4%) from this value are found.For Ga and Au, two metals exposing closed packed metal surfaces, we have shown that theLEIS signal shows a more or less sinusoidal behavior as a function of the tilt angle of theanalyzed surface. Hence, the influence of surface roughness on LEIS signals is higher formetals (RAu = 0.72 and RGa = 0.60) than for silica (RSiO2 = 0.87)

The relation between the LEIS signal of a spherical particle and a flat surface hasbeen used to develop a method to determine cluster sizes based on LEIS signals (Chapter 2).When the total metal load and the specific surface area of the support of a catalyst are known(from e.g. AAS-ICP and BET analysis) one can determine the average cluster radius from the

Summary

122

“missing” LEIS signal. The method was verified by comparing the cluster size of, similarlysynthesized, Pt/γ-Al2O3 with TEM and with the new LEIS based method, in this case bothmethods agreed on an average cluster size of 1.3±0.2 nm.

This LEIS based method has been applied to determine the noble metal cluster size ofa Pt/Rh/CeO2/γ-Al2O3 catalyst, a material used as washcoat for automotive exhaust catalysts.For this catalyst conventional methods for cluster size determination, such as COchemisorption, HREM and TEM, could not be used because of the presence of ceria. UsingLEIS, we determined an average cluster size of 2.1±0.3 nm for the Pt/Rh/CeO2/γ-Al2O3.

Not only the cluster size, but also the surface composition of the Pt/Rh/CeO2/γ-Al2O3,has been determined. From this it could be learned that the catalyst contained mixed Pt/Rhclusters, with a Pt enriched surface. Moreover, the amount of carbonaceous deposits uponcold start conditions has been quantitatively determined which enabled modeling of thecatalyst performance during cold start conditions (Chapter 2).

Another catalyst that appeared to suffer heavily from carbonaceous deposits is thevanadium phosphorous oxide (VPO) catalyst that is used in the synthesis of maleic anhydride(Chapter 3). Knowledge on the active phase of this catalyst is highly desirable since theselectivity of currently used VPO catalysts towards maleic anhydride is 58% and it takes upto 2½ months to reach optimum activity. Carbonaceous species were found to preferentiallycover the vanadium surface species in this catalyst. Therefore, an erroneous surface structurefor this catalyst was determined in earlier studies that only addressed the phosphorus tovanadium ratio. A mild calcination of the VPO (200 mbar O2 at 573 K for 30 min) allowed usto selectively remove the carbonaceous species without changing the average valence state ofthe vanadium as verified with XPS. This is an interesting result since calcination is patentedfor both activation and regeneration of VPO catalysts for pressure, temperature and contacttime ranges comprising the conditions of our oxidative treatment. Comparison of the surfacevanadium and phosphorous concentrations (as determined with LEIS) with crystallographicdata (as determined with XRD) allowed us to determine the surface structure of the VPOcatalyst. The surface of VPO has the same vanadium surface concentration as vanadylpyrophosphate. However, the surface cannot be explained by the generally assumed vanadylpyrophosphate phase, since it contains a huge excess of phosphate, which gives rise to asurface P/V ratio of 2.0±0.2. The observed surface concentrations and P/V ratio may either beexplained by an excess amount of phosphorous positioned between the vanadyl units and thephosphate groups of a heavily distorted vanadyl pyrophosphate or by a significantcontribution of a phase such as αΙΙ−VOPO4.

Similar to the case of the VPO catalyst, the surface structure of Cu/ZnO basedcatalysts, which are used in methanol synthesis, is subject to much debate. There are tworeasons for this, first of all the surface structure of this catalyst changes depending on theapplied environment, secondly Cu and Zn are very similar and therefore difficult todistinguish for many characterizing techniques, including LEIS. However, after improvingthe mass resolution of our LEIS detector, application of isotopically enriched (supported)63Cu/68ZnO catalysts allowed separate detection of Cu and Zn with LEIS. This opened the

Summary

123

possibility to study the dynamic behavior of Cu/ZnO based catalysts (Chapter 4). Depthprofiles show that the catalyst surface is enriched in ZnO under methanol synthesisconditions. Not only information on the surface composition, but also its oxidation state isrelevant, for instance both Cu0 and Cu1+ have been predicted to be the active Cu species inthe Cu/ZnO based catalyst. To address this issue we have developed a novel method thatenables the selective determination of the oxidation state of the metals in the outermostatomic layer by combining LEIS before and after adsorptive N2O decomposition. Applicationof this method –which uses the fact that the LEIS yield of a metal is typically 2 to 5 timeshigher than that of the corresponding metal oxide- shows that the Cu surface species are(partly) oxidized and that the ZnO contains oxygen vacancies. Our measurements providestrong proof for an earlier proposed model, which suggests that the methanol synthesisactivity is attributed to Cu(I)/ZnO with oxygen vacancies. The oxygen vacancy concentrationis shown to increase with the reduction temperature. Besides reduction in 5% CO/5%CO2/90% H2 is shown to yield a much lower oxygen vacancy concentration than reduction inpure H2 at the same temperature.

The Cu/ZnO based catalyst is a clear example of the influence that gaseousenvironments can have on the surface composition of a catalyst. It is generally accepted thatadsorbates may completely restructure a catalyst surface and surface phases that are unstableunder high vacua may well play a critical role in catalysis. Therefore, we have built adifferentially pumped pressure cell (Chapter 5) that enables in situ LEIS during reactions (upto 10-3 mbar) and that partly bridges the pressure gap between LEIS (applying UHV) andindustrial catalysis (applying pressures P ≥ 1 atm.). Using in situ LEIS in combination withquadrupole mass spectroscopy (QMS) we have monitored the CO oxidation over Pt(110)both at steady state and in the oscillatory regime (Chapter 6). The use of C16O andisotopically enriched 18O2 enabled us to separately monitor adsorbates concentrations and thefree Pt surface concentration with LEIS, while the gaseous reactants and products weresimultaneously monitored with QMS. The activity of the CO oxidation is linked torestructuring of the Pt surface since the reaction ignites simultaneously with a dramaticincrease in the oxygen sticking probability at a critical CO coverage of 23%±2% of amonolayer.

125

Samenvatting

Katalysatoren en de producten die ermee worden gemaakt vinden we overal. Zoworden bijvoorbeeld bijna alle polymeren, brandstoffen, geneesmiddelen en pesticidengemaakt met behulp van katalysatoren. Daarnaast worden de meeste auto uitlaatgassenkatalytisch gereinigd. Katalysatoren versnellen selectief chemische reacties zonder daarbijzelf te worden opgebruikt. Voor een grote groep katalysatoren, namelijk de heterogenekatalysatoren, kan een katalytische cyclus als volgt worden beschreven: Reactantenadsorberen op het katalysator oppervlak waar ze worden geactiveerd en met elkaar reageren,tenslotte verlaten de gevormde reactie producten via desorptie het katalysator oppervlak, datonveranderd achterblijft, gereed voor een volgende katalytische cyclus. Het mag duidelijkzijn dat het oppervlak van de katalysator een sleutelrol speelt. Dit proefschrift beschrijftenkele studies waarin de oppervlakte compositie van verschillende katalysatoren isgerelateerd aan de katalytische activiteit om een beter begrip van de katalyse te krijgen. Voordeze studies is een groot deel van de informatie over de katalysator oppervlakken verkregenmet de LEIS analyse techniek. Hiervoor is nieuwe methodologie en instrumentatieontwikkeld en een deel van dit proefschrift is daarom aan deze ontwikkelingen gewijd.

Het eerste hoofdstuk gaat over de invloed van oppervlakte ruwheid en hetsamenpersen van poeders (compactie) op LEIS. Omdat gedragen katalysatoren, dit zijnpoeders met specifieke oppervlakken tot wel 1000 m2/g, vaak worden gebruikt in de katalyseis kennis van deze invloed van groot belang. Voor deze studie is met atomic layer depositionTa2O5/SiO2 gemaakt, waarin Ta2O5 alleen aanwezig is in de buitenste atoomlaag. Door LEISanalyses van dit materiaal te vergelijken na compactie bij verschillende drukken, isaangetoond zowel de compositie van de buitenste atoomlaag als de LEIS analyse niet wordenbeïnvloed door de routinematig gebruikte compactie van poeders met drukken tot 300 MPa.In het verleden had men al de invloed van oppervlakte ruwheid op LEIS onderzocht, echterdeze resultaten verschilden van de onze. Het effect werd destijds overschat vanwegeincomplete neutralisatie tijdens de LEIS experimenten en een 2 dimensionale aanpak tijdenshet modelleren. Zowel onze experimenten als ons 3 dimensionaal model laten zien dat LEISsignalen van oxidische sferische deeltjes het aantal deeltjes in het equatoriale vlak van dezedeeltjes representeren. Dit betekent dat in het geval van silica, naast alumina de meestgebruikte drager voor katalysatoren, het effect van oppervlakte ruwheid minimaal is. Deoppervlakte ruwheidsfactor R tussen silica poeders en atomair vlak silica is 0.87±0.06wanneer een LEIS opstelling met een strooihoek van 145° wordt gebruikt. Voor andere veelgebruikte geometrieën, met strooihoeken tussen de 135° en 145°, varieert R slechts < 4 %.Voor Au en Ga (twee metalen met een dichtgepakt oppervlak) hebben we laten zien dat deLEIS signalen een sinusoïdaal gedrag als functie van de tilt hoek van het geanalyzeerdeoppervlak laten zien. Hieruit volgt dat voor metalen de invloed van oppervlakte ruwheid ophet LEIS signaal groter is (RAu = 0.72±0.07 en RGa = 0.60±0.07) dan voor silica (RSiO2 =0.87±0.06).

De ontdekte relatie tussen het LEIS signaal van een sferisch deeltje en een vlakoppervlak is gebruikt om een methode te ontwikkelen om met behulp van LEIS

Samenvatting

126

clustergrootten te bepalen (Hoofdstuk 2). Wanneer de totale metaal belading en het specifiekeoppervlak van de katalysator bekend zijn (uit bijv. AAS-ICP en BET metingen) kan degemiddelde clustergrootte worden bepaald aan de hand van het “ontbrekende” LEIS signaal.De geldigheid van de methode is onderzocht door de clustergrootte van, op dezelfde wijzegesynthetiseerd, Pt/γ-Al2O3 te bepalen met zowel TEM als met de nieuwe op LEISgebaseerde methode, in dit geval gaven beide methoden een gemiddelde clustergrootte van1.3±0.2 nm.

De op LEIS gebaseerde methode is vervolgens gebruikt voor de bepaling van degemiddelde clustergrootte van de edelmetaal clusters van een Pt/Rh/CeO2/γ-Al2O3

katalysator, een materiaal dat voor auto uitlaatgas katalysatoren wordt gebruikt. Voor dezekatalysator kunnen vanwege de aanwezigheid van ceria, conventionele methoden om declustergrootte te bepalen, zoals CO chemisorptie, HREM en TEM niet worden gebruikt. MetLEIS bepaalden we een gemiddelde clustergrootte van 2.1±0.3 nm voor het Pt/Rh/CeO2/γ-Al2O3. Niet alleen de clustergrootte maar ook de oppervlakte samenstelling van de clustersvan het Pt/Rh/CeO2/γ-Al2O3 is bepaald. Hieruit blijkt dat de katalysator gemengde Pt/Rhclusters bevat met een Pt verrijkt oppervlak. Bovendien is gekeken naar de koolstof afzettingna koude start condities (Hoofdstuk 2).

Een andere katalysator die zwaar bleek te lijden onder koolstof depositie is devanadium fosfaat oxide katalysator (VPO) katalysator die wordt gebruikt voor de synthesevan maleine anhydride (Hoofdstuk 3). Kennis van de actieve fase van deze katalysator is zeerwenselijk omdat de selectiviteit van de huidige generatie VPO katalysatoren voor maleineanhydride slechts 58% bedraagt en het tot 2½ maand kan duren voordat de optimale activiteitwordt bereikt. Koolstof blijkt op deze katalysator preferentieel het vanadium af te dekken.Daarom zijn in het verleden, tijdens studies waarin alleen de fosfor/vanadium verhoudingwerd bepaald, verkeerde oppervlakte structuren voor deze katalysator gevonden. Een mildecalcinatie van VPO (200 mbar O2 bij 573 K gedurende 30 min.) blijkt selectief koolstof vanhet oppervlak te verwijderen, zonder dat de gemiddelde valentie toestand wijzigt. Dit laatsteis geverifieerd met XPS. Gezien het feit dat calcinatie onder de gebruikte temperatuur en gasconcentraties is gepatenteerd voor zowel activering als regeneratie van VPO katalysatoren isdit een interessant resultaat. Een vergelijking van de oppervlakte concentraties van vanadiumen fosfor (zoals bepaald met LEIS) en kristallografische gegevens (bepaald met XRD)maakte het mogelijk om de oppervlakte structuur van de VPO katalysator te bepalen. Hetoppervlak van de VPO katalysator blijkt dezelfde vanadium concentratie als de vanadylpyrofosfaat structuur te bezitten. Echter het katalysator oppervlak kan nooit deze structuurhebben, omdat het een grote overmaat aan fosfaat bevat, de oppervlakte P/V verhouding is2.0±0.2. De gemeten oppervlakte concentraties en P/V verhouding worden ofwel veroorzaaktdoor een overmaat aan fosfaat groepen tussen vervormd vanadyl pyrophosphaat of door eensignificante bijdrage van een αII achtige VOPO4 fase.

Zoals in het geval van de VPO katalysator is er ook veel debat over de oppervlaktestructuur van Cu/ZnO katalysatoren die worden gebruikt voor methanol synthese. Hiervoorzijn twee oorzaken, de oppervlaktestructuur van de Cu/ZnO katalysator verandert wanneer dekatalysator wordt blootgesteld aan een andere atmosfeer en daarnaast lijken Cu en Zn erg op

Samenvatting

127

elkaar waardoor ze moeilijk te onderscheiden zijn voor de meeste analysetechnieken. Ditlaatste geldt ook voor LEIS, echter sinds we de massa resolutie van de detector hebbenverbeterd is het mogelijk om in (gedragen) katalysatoren die zijn gesynthetiseerd metisotopisch verrijkt 63Cu/68ZnO het Cu en Zn te scheiden met LEIS. Dit opende demogelijkheid om het dynamisch gedrag van Cu/ZnO katalysatoren te onderzoeken(Hoofdstuk 4). Diepteprofielen laten zien dat het katalysatoroppervlak verrijkt is aan ZnOonder methanol synthese condities. Niet alleen informatie over de oppervlakte samenstelling,maar ook over de oxidatietoestand is van belang, bijvoorbeeld zowel Cu0 als Cu1+ zijnvoorspeld als de actieve Cu component voor de Cu/ZnO katalysator. Om ook deoxidatietoestand te kunnen onderzoeken hebben we een methode ontwikkeld die het mogelijkmaakt om selectief informatie te verkrijgen over de oxidatietoestand van de buitensteatoomlaag. Deze methode die berust op LEIS metingen voor en na N2O titratie -en gebruikmaakt van het feit dat de LEIS opbrengst van metalen typisch een factor 2 tot 5 hoger is voormetalen dan voor het korresponderende metaal oxide- laat zien dat het Cu aan het katalysatoroppervlak deels geoxideerd is en dat het ZnO zuurstof vacatures bezit. Onze metingenonderschrijven een eerder voorgesteld model, waarin de methanol synthese activiteit wordttoegeschreven aan Cu(I)/ZnO met zuurstof vacatures. De metingen laten zien dat de zuurstofvacature concentratie toeneemt als functie van de reductie temperatuur. Tevens blijkt datbehandeling in 5% CO/5% CO2/90% H2 tot een veel lagere zuurstof vacature concentratieleidt dan reductie in zuiver H2 bij dezelfde temperatuur.

De Cu/ZnO katalysator illustreert duidelijk dat de atmosfeer waaraan een katalysatorwordt blootgesteld een grote invloed kan hebben op de oppervlaktesamenstelling. Het isalgemeen geaccepteerd dat adsorbaten een katalysatoroppervlak volledig kunnenherstructureren en het is heel goed mogelijk dat fasen die onder hoog vacuüm niet stabiel zijneen essentiële rol spelen tijdens katalyse. Om dit te kunnen onderzoeken hebben we eendifferentieel gepompte drukcel gebouwd (Hoofdstuk 5) die het mogelijk maakt om in situLEIS te doen tijdens reacties (tot 10-3 mbar) en zo de zogenaamde “pressure gap” tussenLEIS (in UHV) en industriële katalyse (bij drukken P ≥ 1 atm.) deels overbrugt.Gebruikmakend van in situ LEIS in combinatie met quadrupool massa spectroscopie (QMS)hebben we de CO oxidatie over Pt(110) onder steady state condities en in het oscillerendregime onderzocht (Hoofdstuk 6). Het gebruik van C16O en isotopisch verrijkt 18O2 maakt hetmogelijk om zowel de adsorbaat concentraties als de concentratie onbedekt Pt oppervlak tebepalen met in situ LEIS, terwijl tegelijkertijd gasvormige reactanten en produkten kunnenworden gevolgd met de QMS. Onze metingen laten zien dat de reactie tegelijk met eendramatische toename van de zuurstof plakkans bij een kritische CO bedekking van 23%±2%van een monolaag ontsteekt. Dit duidt erop dat de aktiviteit van de CO oxidatie samenhangtmet de fase van het Pt oppervlak.

128

Publications

Chapter 1• Jansen W.P.A., Knoester A., Maas A.H.J., Schmit P., Kytökivi, Denier v.d. Gon, and

Brongersma H.H., The influence of compaction and surface roughness on low-energy ionscattering signals, Surface and Interface Analysis, submitted.

Chapter 2a• Jansen W.P.A., Harmsen J.M.A., Denier v.d. Gon A.W., Hoebink J.H.B.J., Schouten J.C.,

and Brongersma H.H., Noble metal segregation and clustersize of Pt/Rh/CeO2/γ-Al2O3

automotive three-way catalysts studied with LEIS, Journal of Catalysis, 204, 420-427(2001).

Chapter 2b• Harmsen J.M.A., Jansen W.P.A., Hoebink J.H.B.J., Schouten J.C., and Brongersma H.H.,

Coke deposition on automotive three-way catalysts studied with LEIS, Catalysis Letters,74, 133-137 (2001).

Chapter 3• Jansen W.P.A., Ruitenbeek M., Denier v.d. Gon A.W., Geus J.W., and Brongersma H.H.,

New insights into the nature of the active phase of VPO catalysts – A quantitative LEISstudy, Journal of Catalysis, 196, 379-387 (2000).

Chapter 4• Viitanen M.M., Jansen W.P.A., V. Welzenis R.G., and Brongersma H.H., Cu/ZnO and

Cu/ZnO/SiO2 catalysts studied by low-energy ion scattering, Physical Chemistry, B103,6025-6029 (1999).

• Jansen W.P.A., Beckers J., Heuvel J.C. v.d., Denier v.d. Gon A.W., Bliek A., andBrongersam H.H., Dynamic behavior of the surface structure of Cu/ZnO/SiO2 catalysts,Journal of Catalysis, 210, 229-236 (2002).

Chapter 5• Denier v.d. Gon A.W., Cortenraad R., Jansen W.P.A., Reijme M.A., and Brongersma

H.H., In situ analysis by low-energy ion scattering, Nuclear Instruments and Methods,B161-163 56-64 (2000).

• Jansen W.P.A., Denier v.d. Gon A.W., Wijers G.M., V.D. Hoogen P.W., De Laat J.A.M.,Maas T.M., Dekkers E.C.A., Brinkgeve P., and Brongersma H.H., Ontwerp en bouw vaneen differentieel gepompte drukcel voor oppervlakte onderzoek onder reactiecondities,Nevacblad, 38 nr. 4, 83-87 (2000) (Dutch).

Publications

129

• Rikers Y.G.M., Jansen W.P.A., Wijers G.M., Denier v.d. Gon A.W., and BrongersmaH.H., LEIS tijdens Katalyse, Nevacblad, 39 nr. 3, 75-80 (2001) (Dutch).

• Jansen W.P.A., Denier v.d. Gon A.W., Wijers G.M., Rikers Y.G.M., V.D. Hoogen P.W.,De Laat J.A.M., Maas T.M., Dekkers E.C.A., Brinkgeve P., and Brongersma H.H., Adifferentially pumped pressure cell for in situ LEIS analysis of catalysts during reactions,Review Scientific Instruments, 73, 354-361 (2002).

Chapter 6• Jansen W.P.A., Rikers Y.G.M., Denier v.d. Gon A.W., and Brongersma H.H., Surface

coverage during the oscillatory CO oxidation over Pt(110) -An in situ LEIS and QMSstudy, Surface Science, submitted.

Other subjects• König R., Herrmannsdörfer T., Riese D. and Jansen W., Magnetic properties of Ag sinters

and their possible impact on the coupling to liquid 3He at very low temperatures, Journalof Low Temperature Physics, 106, 581-590 (1997).

• V.D. Linde S.C., Jansen W.P.A., De Goeij J.J.M., V. IJzendoorn L.J., and Kapteijn F., In-target production of [15O]nitrous oxide by deuteron irradiation of nitrogen gas, AppliedRadiation and Isotopes, 52, 77-85 (2000).

• Ермолов С.Н., Глевский В. Г, Янсен B., Маркин С.Н., Бронгерсма Х.Х.,ИССЛЕДОВАНИЕ ПОВЕРХНОСТИ БИКРИСТАЛЛОВ МОЛИБДЕНА МЕТОДОМРМН, Извесмця Академцц наук серия физическая, in press (Russian).

• Ermolov S.N., Jansen W.P.A., Markin S.N., Glebovsky V.G., and Brongersma H.H., Thesurface of Mo bicrystals studied by Low-Energy Ion Scattering, Surface Science, 512,221-228 (2002).

• Brongersma H.H., Gildenpfennig A., Denier van der Gon A.W., Van De Grampel R.D.,Jansen W.P.A., Knoester A., Laven J., and Viitanen M.M., Insight in the outside: Newapplications of low-enery ion scattering, Nuclear Instruments and Methods in Physics,B190, 11-18 (2002).

131

Acknowledgement / Dankwoord

Enkele mensen wil ik in het bijzonder bedanken voor hun bijdrage aan detotstandkoming van dit proefschrift en voor de plezierige tijd die ik met hen de afgelopen vijfjaar heb gehad. Natuurlijk in de eerste plaats mijn promotor Hidde Brongersma, voor zijnstimulerend enthousiasme waarmee hij me heeft ingewijd in de wereld van de chemie en deionenverstrooiing en voor de kansen die hij me geboden heeft zowel tijdens de promotie alsbij Calipso. Mijn copromotor Arnoud Denier van der Gon wil ik bedanken voor de talrijkeleerzame discussies en de puntjes op de i (van silica). Hans Niemantsverdriet, fijn dat je mijntweede promotor wilde zijn en dat ik gebruik kon maken van je encyclopedische kennis vande katalyse. Verder dank ik al mijn collega’s van de groep Fysica van Oppervlakken enGrenslagen voor de fijne samenwerking, de prettige sfeer en de vele uitstapjes.

Tijdens mijn promotie had ik het geluk deel te mogen nemen aan veel interessantesamenwerkingsprojecten met verschillende universiteiten. Matthijs bedankt voor de prettigesamenwerking en je expertise op het gebied van VPO. Jurriaan en Han bedankt voor talloosveel Cu/Zn kats. Jan en Jozef het was leuk om met jullie samen te werken op het gebied vanauto uitlaatgas katalysatoren, een onderzoek dat we deden op de Calipso. Net als het werkaan Ta2O5/SiO2; Jos, Arie, Anja, Minna, Theo en andermaal Hidde bedankt voor de gebodenkansen en de prettige samenwerking.

Inmiddels ben ik erachter gekomen dat high-tech apparatuur zich onderscheidt doorhet feit dat er regelmatig aan moet worden gesleuteld om alle mogelijkheden (inclusief demogelijkheden waar het apparaat niet voor was gebouwd) te kunnen benutten. Gelukkigstonden Gerard, Piet, Jos en vele anderen GTDers altijd klaar. Zonder jullie vakkennis wasdit proefschrift een stuk dunner geweest! Ook de computer (Erwin) en elektrotechnische(Wijnand en Rein) ondersteuning verdient een pluim, Rein ik zal de ERISS niet meeraanzetten.

Ik heb één afstudeerder gehad, maar dat was wel een goede. Yuri, bedankt voor jebijdragen aan de CO oxidatie metingen, steady state en oscillerend.

In the group, I have always enjoyed the presence of a multitude of enthousiasticguests. It was nice to cooperate with you. In particular I would like to thank Sergeij Ermolovand Vadim Glebovsky for the pleasant cooperation on refractories. I truly enjoyed workingwith you.

Tot slot wil ik beide ouders en mijn vrouw en paranimf Viola bedanken omdat ze eraltijd waren en voor al het plezier dat we samen hebben beleefd en zullen beleven.

132

Curriculum vitae

Wim Jansen werd op 29 oktober 1973 geboren te ‘s-Hertogenbosch. Na het behalen van hetVWO diploma aan het Maurick College te Vught startte hij in September 1992 met de studieTechnische Natuurkunde aan de Technische Universiteit Eindhoven (TUE). Tijdens dezestudie liep hij stage aan de Universtät Bayreuth in de groep van prof.dr. F. Pobell en hijstudeerde af onder begeleiding van prof.dr.ir. J.J.M. de Goeij met als onderwerp de synthesevan 15O-[N2O] en het gebruik hiervan in positron emission profiling (PEP) onderzoek aankatalysatoren. In augustus 1997 werd het ingenieursdiploma behaald en aansluitend hieraanbegon hij als onderzoeker in opleiding in de vakgroep Fysica van Oppervlakken enGrenslagen van de TUE onder leiding van prof.dr. H.H. Brongersma. Tijdens dit onderzoek,ondersteund door de Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO), isde compositie van katalysator oppervlakken onderzocht en gerelateerd aan katalytischeactiviteit. Met name is gebruik gemaakt van de LEIS techniek welke ook verder ontwikkeldis tijdens dit onderzoek. Het werk hieraan resulteerde in dit proefschrift. Tijdens zijnpromotie werkte Wim in 2000 part-time voor Calipso B.V., waar hij commercieel katalyseonderzoek verrichtte. Sinds juli 2002 is hij werkzaam als klinisch fysicus in opleiding in decluster Erasmus MC Daniel - LUMC – MCH Westeinde.

Date post:	26-Feb-2021
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

Insight in the Outside of catalysts with LEISInsight in the Outside of Catalysts with LEIS...

Documents