why do chelating ligands improve the activity of nimo hydrotreating

Research Collection

Doctoral Thesis

Why do chelating ligands improve the activity of NiMohydrotreating catalysts?

Author(s): Cattaneo, Riccardo

Publication Date: 2000

Permanent Link: https://doi.org/10.3929/ethz-a-004056334

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-004056334

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

Diss. ETH Nr. 13760

Why do chelating ligands improvethe activity of

NiMo hydrotreating catalysts?

A dissertation submitted to the

Swiss Federal Institute of Technology Zurich

for the degree of

Doctor of Technical Sciences

presented by

Riccardo Cattaneo

Dipl. Chcm. Eng. ETI I Zurich

born on 25 November 1971

from Porlezza (Italy)

Accepted on recommendation of

Prof. Dr. R. Prins, Examiner

Prof. Dr. A. Wokaim, Co-examiner

7urich. 2000

Contents

Chapter 1 Introduction 1

Hydrotreatmg 1

The catalysts 2

Chelating ligands 3

This thesis 4

Literature 6

Chapter 2: X-ray Absorption Spectroscopy 9

X-rays 9

Synchrotron beam lines 11

The photoelectric effect 14

EXAFS 16

Data extraction 21

Data analysis 25

Pros & Cons ofEXAFS 29

Literature 32

Chapter 3: How does the Structure of NiMo/SiCh CKidic Precursors

Influence the Activity of the Final Sulfided Catalysts? 35

Inti oduction 35

Expei iwenttil methoch 36

Sample preparation and characterization 36

Stilfidation and reaction 38

EXAFS measuiements 41

EXAFS anah sis 42

Results 43

Catalytic Activity).

43

Characterization ofNickel 45

i

Characterization ofMo 55

Discussion 60

Conclusions 62

Liteiatuie 63

Chapter 4: A Quick EXAFS Study of the Sulfidation of NiMo/Si02

Hydrotreating Catalysts Prepared with Chelating Ligands 67

Introduction 68

Expei imental methods 68

Sample pi epcn atwn and tests of activity 68

XIFS measin ements 68

hXAFSanahsis 70

Results 72

Molybdenum 72

Nickel 81

Discussion 88

Moh hdenum m the catalyst pi ecm soi s 88

Sulfidation of Molybdenum 89

Chelating ligands and catalytic activity 94

Conclusions 96

Liteiatuie 97

Chaptci 5 Influence of chelating ligands on sulfided TNiMo catalysts:

an EXAFS study 101

Introduction 101

Lxpenmcntal methods 102

Catalysts pi cpai atwn and catnit\ tests 102

FXiFS measw ements 103

EXIFSanahsis 103

Results 104

MoK-edgc 106

u

Ni K-edge

Discussion

Molybdenum

Nickel

Conclusions

Liteiatuie

Chaptei 6' Influence of chelating ligands on

y-AhCh-supported NiMo catalysts

Intioduction

IIDN i eaction netw ai k

/IDS i eaction mechanism

Expei imental

Catalysts pi epai alion

IIDN activity lests

HDS activity test s

N4FS measui ements

ENI PS analysis

Results

Catalytic pei foi mane e

Catalysts chai ac ici nation

Discussion

Moh bdenum

Nickel

Conclusions

Liteiatuie

Chaptei 7 Summary & Conclusions

The oxidic state

The sulfidation

Ihe sulfidic state

Reniai ks & suggestions 165

Acknowledgements 167

Curriculum Vitae 169

iv

Summary

Hydrotreating catalysts are used in the refining industry to remove S, N, O and metals

from petroleum derivatives. The aims of this process are two. On the one hand, the decrease

of SO2 and N(\ emissions during the combustion of fuels which contribute to the

phenomenon of acid rain, on the other hand, the protection of S-sensitive catalysts which

arc used in later stages of the refining process. The materials studied in this work consisted

of sulfided NiMo catalysts supported on S1O2 and y-ALOa. The addition of chelating

ligands during the preparation of these catalysts induces an increase in the activity of the

final catalysts. The used chelating ligands were nitrilotriacetic acid (NTA), ethylene

diamine (EN) and ethylene diamine tctraacetic acid (EDTA). This work aims to

characterize these catalytic materials in the oxidic precursors, during sulfidation and in the

final sulfided state. The information obtained concerning the structural features should

explain the beneficial effect of the ligands on the catalytic activity.

The structure of SiC^-supported catalyst precursors was studied by means of Raman,

UV-Vis and extended X-ray absorption fine structure (EXAFS) spectroscopy. These

techniques revealed that in the absence of chelating ligands Ni interacts strongly with the

support and nickel silicates are formed. The chelating agents form complexes with Ni and

hinder the contact between Ni and the support. Mo is mainly present as polymolybdate

units in the catalyst precursors when no ligands or small ligand amounts are employed. The

concentration of the ligands was \aried in order to observe the effects on Ni and Mo. As

soon as the amount of ligands present in the catalyst precursors is higher than the quantity

needed to complex all Ni, Mo starts to be affected by the ligands and to be present in

smaller units on the support.

The sulfidation processes of Ni and Mo were followed by means of Quick EXAFS

(QEXAFS). The presence of the ligands retards the sulfidation of Ni. The change in the

sulfidation temperature of Ni indicates that the mechanism of sulfidation of Ni changes

when chelating ligands are used. In the catalyst containing Ni and Mo the sulfidation of Mo

takes place in a narrower temperature interval than in the catalyst containing only Mo.

When the concentration of the ligands is increased, the sulfidation of Mo tends to be similar

v

to that in the catalyst prepared in the absence of Ni. According to these observations the

presence of Ni influences the sulfidation of Mo when no ligands are employed. With the

addition of chelating ligands Ni does not interact with the environment, so that Mo and Ni

arc sulfided independently.

EXAFS measurements of the sulfided catahsts suggested the presence of nickel

sulfide clusters in which Ni has either a square pyramidal or an octahedral geometry. The

formation of smaller clusters with similar geometry could be enhanced by the use of

chelating ligands. Therefore, the higher activity observed in catalysts prepared with

chelating ligands could be due to a higher dispersion of Ni. No Ni-Mo proximity could be

detected by EXAFS, what suggests that the Ni-Mo-S model proposed in the past on the

basis of EXAFS data was a wrong interpretation. The interactions between Mo and the

support, formed in the presence of small amounts of ligands, induce the formation of more

regular M0S2 crystallites. When no ligands are employed the static disorder of the M0S2

particles is more pronounced. When larger amounts of NTA and EDTA (molar ratios

NTA:Ni and EDTA:Ni > 1.5) are employed the size of the M0S2 crystallites decreases

because of the formation of Mo complexes in the catalyst precursors.

The hydrodesulfurisation (HDS) activity is strongly affected by changes in the

sulfidation mechanism and final structure of Ni, while the hydrodenitrogenation (HDN)

activity is more dependent on the structure of the M0S2 slabs.

vi

Riassunto

I catahzzatori di idroraffmazione \engono usati nell'industria petrolifera per

eliminare S, N, O e metalli dai derivati del petrolio. Questo processo ha due scopi. Da un

lato la diminuzione delle emissioni di SO2 c di NC\ durante la combustione dei carburanti

derivati dal petrolio. che contribuiscono al fenomeno delle pioggie acide, dall'altro la

protezione dei catahzzatori che vengono impiegati nei successivi trattamenti delle frazioni

del greggio, che spesso vengono disattivati dallo zollb. I catahzzatori studiati in questo

lavoro sono a base di solfuri di Ni e Mo dispersi su Si02 e Y-AI2O3. L'aggiunta di agenti

chelanti durante la preparazione di questi materiali favorisée un aumento dell'attività

catalitica. Gli agenti chelanti utilizzati sono facido nitrilotriacetico (NTA),

retilendiammina (EN) e facido etilendiamminotetracetico (EDTA). Lo scopo di questa tesi

c caratterizzare questi materiali durante Fintera preparazione, partcndo dai precursori

essiccati, per passare alia fase di solforazione e quindi ai catahzzatori veri e propri in forma

solforica. Le informazioni riguardanti la struttura dovrebbero spicgare Faumento

delfattività catalitica conseguito con l'aggiunta dei leganti.

La struttura dei precursori dispersi su silice è stata studiata tramite Raman e

spettroscopie UV-Vis e EXAFS (extended X-ray absorption fine structure). Queste tecniche

hanno rivelato che, in assenza di agenti chelanti. il Ni interagisce con il supporta formando

silicati di nichel. I leganti formano complessi con il Ni e impediscono il contatto fra Ni e

supporte. Se non vengono impiegati leganti 0 ne vengono utilizzate piccole quantità, il

molibdeno c essenzialmente présente sotto forma di polimolibdati nei precursori essiccati.

La concentrazione dei leganti è stata variata per osservare gli effetti su Ni e Mo. Non

appena la concentrazione di legante nel precursore supera la quantità necessaria a

complessare tutto il Ni. il Mo commcia a subire l'effetto dei leganti e a formare unità più

piccole.

II processo di solforazione di Ni e Mo è stato studiato tramite la teenica EXAFS

veloce (QEXAFS). La presenza dei leganti ritarda la solforazione del Ni. Lo spostamento

délia temperature di solforazione del Ni indica che il meccanismo di solforazione del Ni

cambia quando vengono utilizzati gli agenti chelanti. Nel catalizzatore contenente Ni e Mo,

vu

preparato senza leganti, la solforazione del Mo avvicne in un intervallo di temperatura più

stretto rispetto al catalizzatore contenente solo Mo. Alf alimentäre délia concentrazione dei

leganti, la solforazione del Mo diventa simile a quella nel catalizzatore preparato in assenza

di Ni. Questo comportamento dimostra che la presenza del Ni influenza la solforazione del

Mo. I leganti impediscono ogni contatto fra il Ni e Fambiente circostante, in modo che Ni e

Mo vengano solforati indipendentemente.

Le misurazioni EXAFS sui catahzzatori solforati hanno suggerito la presenza di

clusters di solfuro di nichel, nei quali il Ni ha una struttura quadrata piramidale o ottaedrica.

L'uso di agenti chelanti potrebbe favorire la formazione di clusters più piccoli ma di simile

struttura geometrica. Perciô, Felevata attività dei catahzzatori preparati con agenti chelanti

potrebbe essere dovuta a una maggiore dispersione del Ni. Nessun segno di vicinanza fra

Ni e Mo è stato rivelato dai risultati EXAFS nei catalizzatori solforati. Ciô suggerisce che il

modello Ni-Mo-S, proposto in passato sulla base di risultati EXAFS, era una

interpretazione sbagliata. Le interazioni tra le speci di Mo e il supporte, formatesi in

presenza degli agenti chelanti. porta allô sviluppo di particellc di M0S2 più regolari.

Quando, invece non vengono usati i leganti, il disordine statico del disulfuro di molibdeno

è maggiore. L'uso di grandi quantità di N fA e EDTA (rapporte molare NTA:Ni e

EDTANi > 1.5) provoca la formazione di particellc di M0S2 più piccole rispetto agli altri

casi. Questo effetto viene spiegato con la presenza di complessi di Mo nei precursori

essiccati.

L'attività di deidrosolforazione (1ÏDS) è fortemente infiuenzata da cambiamenti del

meccanismo di solforazione e dalla struttura finale del Ni, mentre Fattività di

deidronitrogenazione (IIDN) c maggiormente dipendente dalla struttura delle piastrine di

M0S2.

vin

Chapter 1

Introduction

Hydrotreating

Petroleum is one of the most important raw materials in our society. An enormous

number of products are derived from crude oil. Ehe market sectors which mainly use oil

derivatives are the fuel and the transport industries. Additionally, a large variety of

chemicals tire derived from the processing of crude oil. For these reasons the industrial

processes involved in the refinery of petroleum have been widely developed. The idea

behind oil refining is the conversion of the raw material into clean lighter and more

valuable hydrocarbons. In brief, in oil refining, all the different hydrocarbon crude oil

fractions are separated in a distillation step, according to their boiling point. Before

undergoing other treatments, such as isomerisation. alkylation and cracking, the largest

amount of products obtained from the distillation step are hydrotreatcd. Hydrotreating

refers to a variety of catalytic hydrogénation processes in which unsaturated hydrocarbons

are saturated and S. N, O and metals are remo\ed from molecules which contain these

atoms. Ehis purification process is important for two reasons. The first is the decrease of

the polluting effects of the produced substances. Lower S andN contents induce a decrease

of SO2 and NC\ emissions during the combustion of fuels. These gases contribute to the

phenomenon of acid rain. legislations concerning the S and N emissions in most

Chapter 1

industrialised countries become continuoush stricter because of the awareness of this

problem. The second reason for the importance of hydrotreating is that most catalysts that

are used in further treatments of the oil products do not tolerate sulfur and metals.

Therefore, the performance of hydrotreating catahsts must be steadily improved.

The catalysts

The most commonly used h\ drotrcating catalysts are composed of the sulfides of two

metals supported, usually, on y-ALCX The bimetallic combinations are Ni-Mo, Co-Mo or

Ni-W. Supported CoMo sulfides are excellent in the removal of sulfur

(hydrodcsulfurisation, FIDS) but are slight!} less active in the removal of N

(hydrodenitrogenation, IIDN) and the hjdrogenation of aromatics. Supported NiMo

sulfides, on the other hand, are very good FTDN and hydrogénation catalysts. The NiW

containing catalysts have the highest activity for aromatic hydrogénation at low H2S

pressures [ 1, 2]. Since the catahsts studied during this thesis project were composed of Ni

and Mo, we will refer to these two elements during this general introduction, keeping in

mind that many considerations are \ alid also for the other combinations.

The catalysts are generally produced b\ impregnation of the support with solutions

containing salts of Ni and Mo. The impregnated support is then dried at temperatures

around 120°C. Usually a further heating step to higher temperatures (400-500°C), the so

called calcination, follows the drying procedure. The so obtained catalyst precursors are

then transformed into the actual hydrotreating catalysts by sulfidation at temperatures

around 400°C in a mixture of IE and a molecule containing S. ELS, thiophene, CS2 or

dimethyldisulfide can be used for this. The properties of the final sulfidic catalysts are

dependent on the whole preparation procedure, from the pll and concentration of the

impregnation solution to the sulfidation temperature.

The support empkned during this thesis work was mainly SiCT because it has been

claimed that silica-supported catahsts prepared with chelating ligands can reach activities

superior to alumina supported ones [3. 4]. Moreover silica based catalysts should be less

prone to coke formation because of the kwver acidity of silica [3J. Since coke deposition is

2

Introduction

a major catalyst deactivator in hydrotreating processes, an active and stable silica-based

catalyst would have an important impact on the refining industry.

Chelating ligands

Subject of this thesis was the study of the effects of chelating ligands on the structure

of NiMo catalysts during the whole preparation procedure. Chelating (from the greek chelé:

forked nail, claw) ligands are organic molecules possessing two or more arms with which

they can bind a metal and form a chelate. The chelating ligands used in this work were

nitrilotri acetic acid (NTA), ethylenediamine (EN) and ethylenediamine tetraacetic acid

(EDTA). Their formulae arc depicted in Fig. 1. fhesc compounds arc added to the

impregnating solutions and therefore affect the structure of Ni and Mo during the

preparation. When chelating ligands are used in the preparation of catalysts no calcination

is carried out in order to prevent the combustion of the organic molecules.

The presence of chelating ligands during catahst preparation induces a strong

increase in the hydrotreating activity |3-7|. However, the scientific explanation for this

effect is still unclear.

HO

HO N'"

^y—OHo

NTA EDTA EN

Fig. 1. The chelating ligands used in this woik

40, A0HO, y0

.N

N H7N'NH0

O OH

O OH

j

Chapter I

This thesis

Previous work by our group showed that m a solution containing NEf and Mo6h salts

NTA preferentialh complexes Ni21 [8]. Nevertheless, what consequences this fact has on

the structure of the dned catalysts is still unexplained Chapter 3 of this thesis will deal with

this question.

Another important point is how the chelating molecules influence the sulfidation of

Ni and Mo. Medici and Prms showed that NIA could delay the sulfidation of Ni to higher

temperatures, but no effect was reported on the sulfidation of Mo [9]. In chapter 4 we will

see, in more detail, that the suggestion of Medici and Prins was correct but that also the

sulfidation of Mo is affected by the presence of chelating ligands.

In Chapter 5 the structure of the final sulfidic catahsts is discussed. This topic has

been extremely controversial and se\ ei al models ha\ e been proposed in the past. In general

Mo is considered responsible for the foimation of the active species and Ni as the promoter

agent [2, 10, 11]. There is agreement on the fact that Mo is present on the support as M0S2

crystallites. These crystallites are believed to stack on top of each other and form small

tower-like structures as depicted in Fig 2

Mo

S

Fi«. 2. Layeicd stiuctute of MoS-.

> Y tA «O D

V*\ \*.

^<!, ^ W "A

J?

V

v ~V k

O Y) $

4

Introduction

The most discussed topic concerns the structure ofNi. Since NiMo catalysts are much

more active than Mo-only catalysts, it is supposed that Ni interacts somehow with M0S2.

The most widely accepted model for this interaction suggests that the promoter is situated

on the edges of the M0S2 particles as sketched in fig. 3 [12]. An infrared absorption study

of the absorption of NO molecules on a series of sulfided C0M0/ALO3 catalysts indicated

that this model, the so-called edge-decoration model, is very likely [13]. This work

suggests that the M0S2 edges are covered b\ Co. Nevertheless the exact structure of the

promoter is not known. Extended X-ra\ absorption fine structure (EXAFS) studies

proposed a precise structure for Ni on the IM0S2 edges [5], but we will see that the

interpretation of the data was not totally correct.

Support

Fig. 3. The edge decoration model orNi-Mo-S phase

Another topic that still attracts attention concerns the assignment of the element

responsible for the catahtic acthity in these materials. Unsaturated Mo has been generally

believed to be the active site. Ehe promoting role of Ni was ascribed to an electron donating

effect towards Mo [14J. ITowever, the idea that Ni is the real catalyst is getting more and

more consensus. This idea arises from the observation that cobalt and nickel sulfides

supported on carbon ha\e a higher acti\it\ than MoS^'C [15]. On the contrary, AI2O3-

supported Co and Ni sulfides ha\e a \er\ low FIDS acthity. In the alumina-supported Ni

and Co catalysts, cobalt and nickel interact so strongh with the support that, under normal

sulfidation conditions, the metal ions are not properly sulfided. On the contrary, in carbon-

5

Chapter 1

supported catalysts interactions between the support and Ni or Co absent and they can be

sulfided. From these considerations it has been suggested that Ni is the actual catalyst. In

this model MoS? is considered a secondary support which avoids the Ni support

interactions and enhances its catalytic acth it\.

The results presented in Chapter 5 will show that the presence of chelating ligands

has an important effect on the structure of the final catalysts. This will be also confirmed in

Chapter 6 where similar conclusions will be drawn also for catalysts supported on y-ALOv

Since X-ray absorption spectroscopy was the main tool used to characterize the

investigated catalysts during this work, Chapter 2 gives an introduction to this

spectroscopic technique.

Literature

1. Stanislaus. A., and Cooper, B. FF. Cat Rev -Sei. Eng. 36(1), 75 (1994).

2. Prins, R., Hydrotreating Reactions, in "Handbook of heterogeneous

catalysis", Ertk G.. Knözinger, FI. and Weitkamp, J. Eds., VC1T, Weinheim,

1908 (1997).

3. Thompson, M. S.. 0.181.035. European Patent Application (1986).

4. van Veen, J. A. R., Gerkema. E., van der Kraan, A. M., and Knoester, A.,,/.

Cheni SocChem Commun 1987,1684(1987).

5. Louwers, S. P. A., and Prins. R.. J. Catal. 133, 94 (1992).

6. Inamura, K.. Uchikawa, K., Matsuda. S.. and Akai, Y., Appl. Surf Sei.

121/122.468(1997).

7. Hiroshima. K.. Mochizuchi. T.. Flonma, T., Shimizu, T., and Yamada, M.,

Appl Surf Sei 121/122. 433 (1997).

8. Medici, L.. and Prins, R.. J Catal 163, 28 (1996).

9. Medici. L.. and Prins. R.. J Catal 163, 38 (1996).

10. Prins, R., dc Beer. V. IT. J., and Somorjai, G. A.. Catal. Rev.-Sei. Eng. 31, 1

(1989).

11. Chianelli, R., Daage, M.. and Ledoux. M. T, Adv. Catal. 40. 177 (1994).

6

Introduction

12. Ratnasamy, P., and Sivasankcr. S.. Catal. Rev.-Sei. Eng. 22, 401 (1980).

13. Topsoe, N. Y., and Tops0e. IF. .7 Catal 84, 386 (1981).

14. Harris. S., and Chianelli, R. R„J Catal. 98, 17 (1986).

15. Duchet, J. C, van Oers, E. M., de Beer, V. H. .T., and Prins, R., J Catal 80,

386(1983).

7

Chapter 1

W^,~ "* I

Chapter 2

X-ray Absorption Spectroscopy

X-rays

In 1895 Röntgen discovered a new type of radiation which he called X-rays [1. 2].

For his work on X-ra\s he received the first Nobel Prize in physics in 1901. Röntgen's

achievements were more than pure discover}, he studied the properties of the new rays so

well that he laid the foundations not only for important methods of X-ray detection and for

radiography, but also for the application of X-ra\ absorption to analytical chemistry.

In general. X-ray photons can be produced in two ways. In the first way they

originate when electrons with high velocity are strong!) decelerated, either by a target

material or through bending their path by a magnet, and as a result of a secondary process

from interaction of electrons with atoms. When an electron is stopped or decelerated

Brcmsstrahlung is produced, which consists of an electromagnetic radiation with all sorts of

wavelengths (polychromatic radiation). In the second waw the electron hits an atom and

subsequently an electron is ejected from the atom, creating a core-hole in the atom. Then an

electron from an outer shell can relax to the core-hole and as a result an X-ray quantum is

emitted with a very specific wavelength (monochromatic radiation). As both ways of X-ray

production are the result of the interaction of electrons with matter, it can easily be seen

Chapter 2

that both processes will take place simultaneously. However, as shown in Fig. 1, while

Bremsstrahlung is emitted all the time due to deceleration, the production of

monochromatic X-rays is limited to specific electron energy values and depends on the

target material.

Hot tungstenfilament

(catodc)

5=>

Tai get mateiial

(anode)

E ->

FIC. 1. Electrons arc accelerated towards the target material. Through the impact, X-rays with different

wavelengths are produced. When the electrons have the same energy as the electronic shells of the target

material, X-rays with characteristic wavelengths are produced

Nowadays, synchrotrons are the most powerful X-ray source. In a synchrotron,

charged particles such as electrons or positrons travel under ultra high vacuum with a speed

approaching that of light. When their trajectory is bent by means of strong magnets, X-rays

are emitted tangential to the particle's orbit. Fig. 2 show's why a relativistic speed is needed.

For non-relativistic accelerated charges, the energ} llux is emitted isotropically around the

acceleration. At relativistic energies the radiation pattern becomes sharply peaked in the

direction of motion of the radiating charge. Ehe higher is the energy of the particles, the

A BElectron

orbit

L lectron

orbit

FFG. 2. Radiation emission pattern of electrons in circular motion. Case A: non-relativistic electrons.

Case B: relativistic electrons.

10


more collimated becomes the emitted radiation.

Normally, the synchrotron radiation source is a three-stage machine, comprising a

linear accelerator (linac). a booster synchrotron and a storage ring (Fig. 3). Electrons are

fired from the linear accelerator into the booster s}nchrotron, where they are accelerated to

almost the speed of light before being injected into the storage ring. Ffere the electrons

travel in a vacuum inside a tube around the circumference of the ring and remain stored in

orbit producing synchrotron radiation for 10 to 20 hours. The data presented in this work

were collected at the European Synchrotron

Radiation Facility (ESRF) in Grenoble,

France. and at the Hamburger

S}nchrotronstrahlungslabor (11ASYLAB) in

1 lamburg. Germany. ESRF is, at present, one

of the most powerful synchrotrons in the

world. Its storage ring has an 844 m

circumference where electrons travel with an

energy of 6 GeV and an intensity of 200 mA.

At HASYLAB synchrotron radiation is

emitted from positrons in the storage ring

DORIS. The beam energy is 4.45 GeV and

FIG. 3. Schematic lepresentation of a synchrotron. ^lie intensity 100 niA.

Synchrotron beam lines

Three kinds of devices are used in a S}iichrotron to generate X-rays by bending the

trajectory of the electrons: bending magnets, wigglers and undulators [3]. Bending magnets

consist of a single magnet that induces a uniform curved motion of the charged particles.

Wigglers and undulators are constituted of several magnets that transmit periodical

impulses to the electrons that therefore move in an oscillatory path through the device. The

difference between a wigglcr and an undulator is the amplitude of the oscillations imparted

on the electrons. In a wiggler the radiation from each oscillation is incoherent and can be

approximated as the sum of N separate sources where N is the number of bends. Wigglers

It

Chapter 2

allow the spectrum to be shifted considerabh to higher energy, enabling high-energy

photons to be produced at a more easily achiev able electron energy. In an undulator the size

of the oscillations is smaller and their number is much larger. Therefore, the interference

effects become important and the spectrum becomes a set of several extremely intense

peaks. In case of EXAFS measurements onh bending magnets and wigglers arc used.

Beam lines are built in correspondence with these devices. The produced X-rays are

collected in highly evacuated tubes and further collimatcd by means of very sophisticated

optics. The so obtained beam is polychromatic. Since, during an X-ray absorption

experiment, the energy of the X-rays has to be precisely scanned in a defined interval.

monochromators are used in order to keep onh the radiation with the desired wavelength.

Almost monochromatic light is obtained according to the Bragg relation:

nX-2dsinO (1)

with d the lattice spacing of the crystal used as monochromator, 9 the incident angle, n the

order of reflection, and Xthe desired wavelength. Usually. Si crystals are employed. For X-

raysfrom2to 10 keV Si (111) (d - 3.1350 A) is used. 1 he range between 10and20keVis

covered best by Si (311) crystals (d = 1.6372 A) and above about 20 keV Si (511) (d -

1.0450 Â) is appropriate. Two kinds of monochromators are generally used in X-rays

absorption dedicated beam lines: double cnstal and channel-cut monochromators [4]. In

both cases the beam is reflected twice in order to maintain a horizontal path. In the case of

double crystal monochromators the two crvstals move independently. Usually they are not

exactly parallel: the second one is slightly adjusted in order to remove harmonics (X-rays

with wavelength nz. with n^ 1). A channel-cut monochromator consists of a crystal in

which a precise channel has been incised. LTsing the same principle, the X-ray wavelength

is chosen changing the angle of the monochromator. but this time harmonics have to be

rejected in another way: usually Au or Cr mirrors are employed. By tuning their angle it is

possible to choose a cut-off energ}. after which X-ra} s are no longer reflected.

The X-rays obtained in these \\a\s are used to execute the experiments that are

carried out in an experimental hutch. The most frequently used method to quantify the

absorption of X-ra} s in heterogeneous catahsis is to measure the intensity of the incoming

12


(I0) and transmitted (lt) beams by means of ionisation chambers. Depending on the energy

range, the chambers arc filled with different gases, usually nitrogen, argon, or krypton and

mixtures thereof. Fig. 4 summarises the set-up of an X-ray absorption beam line.

Reference

Sample compound

Synchrotron ^n n —« -— —

Radiation /

Slits V f /Monocfiomator \ /

Ionisation chambers

FIG. 4. Schematic set-up of a Upical X-ray Absorption beam line

In a classical EXAFS experiment the monochromator is rotated to a certain position

and halted. Before collecting data at the corresponding X-ray wavelength, a certain dead

time has to be waited to allow a stabilisation of the mechanical equipment. After that, the

real data collection starts, that amounts to 1 s for data points close to the edge energy and

gradually increases with energ}, reaching 6 s if high k \ allies have to be reached or in case

of very diluted samples. Ehe collection of a spectrum in the classical way takes therefore a

relatively long time. 20-40 min. Since the collection of spectra in shorter times offers

interesting application possibilities, several methods have been developed to collect an

EXAFS spectrum in a much shorter time range.

The Quick EXAFS (QEXAFS) technique is very similar to classical EXAFS with the

difference that the monochromator is constanth moving and that the absorption data are

integrated over small time intervals (e.g. 0.01 s) [5. 6]. Ehe minimal time required to record

a spectrum is about 1 min. Hie shortening of the measuring time is due to the fact that no

dead time is needed between the data points. Hie drawback of QEXAFS is a slight loss of

data quality in comparison with classical EXAFS data.

An even faster wa} to collect EXAFS data consists in the use of dispersive optics,

where a polychromatic beam is produced and focussed on the sample by means of a bent

crystal (polychromator) [7. 8j (Fig. 5). The detection is carried out by means of a position

sensitive detector that is calibrated with an energ}-position correlation. In this way the

spectrum can be theoretical!} recorded in milliseconds, because the full data set is recorded

13

Chapter 2

simultaneously, but the signal must often be accumulated for several seconds in order to

obtain a good signal to noise ratio. This method requires a complex set up and a different

data treatment. Moreover onh' a limited data range can be recorded, usually 400-600 eV

ADemazriithîiition minci // _—r~:

,/ /f~-fr

//,' S- Etfe;f:"e

Wrticjl focn;in2 urn M

7->

'

/

Ta-etf !nnhi.-i^

P >:ifi< n ;er_citr*e detector

FIG. 5. Principle of dispersive energ} EXAFS (DEXAFS). Set-up of the DEXAFS beam line ID24 at ESRF

(fromhttp:/'www esrf fr/e\p facilities 1Ü24TD24 html)

after the edge, which corresponds to a k-range (s. next sections) of 10-13 Â"1.

Thephotoelectric effectAfter having described how X-rays are produced and brought to the sample, the basic

principles about the interaction of X-rays with matter will be discussed. When X-rays hit a

material two main processes can occur. Hiev can be absorbed by the atoms composing the

material or they can be scattered, and if none of these processes takes place, they pass

through the material as transmitted beam. When the monochromatic photon beam passes

through the sample, its incident intensity h will be decreased by a specific amount which is

determined by the absorption capacity of the material and the path length that the X-rays

have covered:

dl =•- jj, I0 dx (2)

14


The linear absorption coefficient/./ is a function of the photon energy. Integration of the

attenuation in intensity o\ er the total thickness of the sample x results in Lamberf s law:

I0 - lt e^E^ (3)

During an X-ray absorption experiment, Iq and the intensity of the transmitted beam I, are

measured before and after the sample, respectively. The absorbance of the sample is

therefore calculated bv:

MX=

ta|j; (4)

O

60

O

The degree of absorption is a function of the energ} of the photons. When the energy of the

X-rays that hit a sample is increased, a gradual decrease of the absorption is observed due

to the decrease of the ionisation cross-section of atoms with energy. ITowever, when the X-

ra\s ha\e enough energy to excite and

knock an electron out of its orbit a sharp

increase in absorption will be observed.

Therefore, the absorption profile over a

large range is a decreasing line with

jumps at the energies corresponding to

the binding energies of the electrons of

the various Orbitals, as depicted in Fig.

6. If this excitation energy corresponds

to the energy of an electron belonging to

an s core level, the absorption increase is

called a A-edge. if the energy

corresponds to a p type core level, the

absorption increase is called an L edge.

Iog(Energ\)

FIG. 6. Profile of the re!ati\e absorption cross-

section over a larçe energy tance

15

Chapter 2

EXAFS

It is interesting to closely observe how an absorption spectrum around the edge is

structured. In Fig. 7 the absorption spectrum at the K- edge of Mo in M0S2 is plotted. In

general, the interval about 50 eV before the edge is called the pre-edge region. The edge

region extends until about 20 eV after the edge. For an isolated atom the shape of the

absorption profile after the edge is regularly decreasing, whereas for an atom surrounded by-

neighbours a series of oscillations are present until 1000-1500 eV after the edge. The shape

of these wiggles is dependent on the kind, number and distance of neighbouring atoms. The

principle of EXAFS. which is an acronym for Extended X-ray Absorption Fine

Structure, is to extract information about the surrounding of the investigated species

through a mathematical processing of the absorption data. It is also possible to obtain

qualitative structural information

from the absorption spectrum in

the first 50 eV right after the

edge. This region is called X-ray

Absorption Near Edge Structure

^"^(XANES) 19, 10]. The physical

principles that explain the

Pre-edgen _

presence of a fine structure in

the absorption data will be given

21500 in this section, together with the

basic formulae.

When the incoming X-ray

beam has enough energy to

1 0

EXAFS

XAXES

05

1t«l 20020 20060

-0 1

19500 20000 20500 21000

X-ray energy | eV]

FIG. 7. X-rav absorption spectrum of \foS^

excite an electron of the sample to a vacant state or to the continuum, the kinetic energy of

the produced photoelectron will be:

Ep = hv-F, (5)

\6


where hv is the energy of the X-rays and E0 the energy necessary to kick out the electron

from its orbit. Flowever, the removed photoelectron can be considered as a spherical wave

expanding around the absorbing atom. Its wav elength X is defined as:

2/t

(6)

with the wav c v ector k87TTT1

l h2 J (hv-E0) M15123AyïA:_I,>s[eVj-E0LeVl (7)

where m is the electron mass and h the Planck's constant.

Once the X-rays have surmounted the edge energ}. the removed electron has enough

kinetic energy to be excited and expand towards the continuum. Flowever. when

neighbouring atoms are present, the photoelectric wave is backscattered towards the

original absorbing atom and the outgoing and backscattered waves interfere, as show in

Fig. 8. The wavelength A. of the outgoing wave is dependent on the X-ray energy and, since

during an EXAFS experiment the energ} of the X-ra} s is increased constantly, the

wavelength of the photoelectron decreases continuously. The interference between the two

waves is alternatively constructive or destructive depending on the X-ray energy. The

absorption of X-rays is close!} related to the electronic state of the atom. Therefore, when

the outgoing and backscattered waves interfere constructively the absorption coefficient

becomes larger. When the} are out of phase it becomes smaller. The nature of the

hv

,A

nv

//•nv A>

FTG. 8. X-tays excite an election of the absotbei atom, that is emitted as a cnculai photoelection-wave

Neighbouimg atoms backscattet the outcoimtig \\a\e tow aids the absotber Outgoing and backscattered

waves mtei feie

17

Chapter 2

neighbouring atoms (backscatterers) influences the backscattered wave and consequently

the absorption of X-ra} s. For this reason in case of an isolated atom no fine structure after

the edge is observed. One of the factors that led to the development of EXAFS was the

possibility to obtain precise information about the nature, the number and the distance of

neighbouring atoms. Although position and amplitude of the oscillations are connected to

the distance and number of neighbours, it is necessary to transform the data to achieve the

exact results, as will be explained in the next section.

Eo understand how the X-ra} absorption is related to the electronic aspects of this

process it is helpful to consider the formulae that describe the dependence of the absorption

coefficient from the X-ray energ} according to the dipole approximation [11]:

H- 4 N 7t2 e2 - |<T'j| 11 TV2 P(hr) (8)

Here is N the number of atoms of one kind per unit volume, co the radial frequency of the

X-ray photon with energy hco. |T',X the wave function of the photoelectron in its initial state

(the core state) and 'fl'\)\ the final wave function of the photoelectron. p(Ef) is the density of

final states, Ef the energy of the outgoing electron. FI is the photon-electron interaction

ITamiltonian.

The interference between outgoing and backscattered electron wave gives rise to the

sinusoidal variation of u versus the energ}. We can see that u depends on \fV\)\, lvF-r)| and

p(Ef). |VF,)| is fixed at a given core level; p(Ft) accounts for a monotonous contribution;

only bFf)| can therefore be responsible for the sinusoidal oscillation of the absorption

coefficient. )vFf)| is given by the sum of the outgoing wave and the backscattered wave, due

to the backscattering action of each neighbouring atom. The fine structure or EXAFS

function is defined as the normalized oscillator} part of u and is given by:

18


X(E)-bo

(9)

p0 being the smoothly varving part of p which corresponds to the absorption by an isolated

atom, that is the absorption coefficient if the phenomenon of backscattering would not

exist. The EXAFS signal so defined is normalized and contains information per absorbing

atom.

The most commonly used formula to describe the EXAFS phenomenon is derived

from the generally accepted short-range single-electron single-scattering theory [12-15].

The meaning of single and multiple scattering is shown in Fig. 9. The mentioned theory

assumes that multiple scattering has no significant effect on the EXAFS function. This is

hvA

nvB

hv

4

c

FIG. 9. The produced photoelectron can experience \aiious scatteimg patterns: A. Single scattering,

B Multiple scattering, 0, Foiwaid scatteimg.

generally valid if one considers that the length of the path in multiple scattering tends to

cancel out the involved waves. It has, nevertheless, to be noted that in case of

approximately collinear arrangement of two neighbouring atoms, multiple scattering

becomes important due to a forward scattering effect that enhances the transmission of the

wave (Fig. 9). According to the short-range single-electron single-scattering theory, % can

be expressed as a function of k as a wave function composed of an amplitude and an

oscillatory term:

Shells

X(k)- X Aj(k) sm5j(k)i

(10)

19

Chapter 2

The index j represents the different coordination shells. Each jn-shell is made up of N,

identical atoms at a distance R, from the absorber atom. Ehe amplitude A,(k) contains the

coordination number (N) and the Debye Waller factor a that accounts for thermal and static

disorder [16]:

Aj(k)=is02E1(k)e-2k:öl2e-2R./^ (11)

S0 is the amplitude reduction factor due to man} bod} effects, which takes Into account the

relaxation of the central atom and the multielectron excitation [17]. F,(k) is the

backscattering amplitude which depends on the scattering power of each neighbouring

atom. The term e"~ '"'

,where X represents the mean free path of the photoelectron, takes

into account the inelastic losses due to the finite lifetime of the excited state (the hole

lifetime and the photoelectron lifetime).

The sine argument §j(k) is:

8,(10-

2kR, + (I>,(k) (12)

One period of the sine function is completed when the ratio of double distance 2R,, between

absorber and backscatterer. and photoelectron wavelength has increased by one. This is the

meaning of the expression sin(2kR,). Fhe photoelectron suffers additional energy dependent

phase shifts in the absorber and in the backscatterer. They are summarised in the scattering

phase shift 0,(k).

In general we can sav that the EXAFS signal is determined bv the backscattering&

Ni TD n /I \

amplitude E,(k), multiplied b} the factor =; and decreased by the terms e~~ i^

,

k lvj~

e ', S0" and the sinusoidal factor which is a function of interatomic distances, and

phase shift d>,(k).

20


Data extraction

After an introduction about the physical principles of EXAFS wc will discuss how

the experimentally collected data have to be treated in order to obtain the data that will be

used for the analysis. Processing of the data presented in this work was performed by

means of the program XDAP (Version 2.2) [18]. Ihe explanation is therefore based on the

procedures used in this program.

1 3

1 1

09

07

05

03

01

-0 1

-0 3[

-

19500

A

20000 20500 21000

X-ra} energ} [eV]

21^09

B

1 3

1 1

09

07

s05

I 03

01

-0 1

-0 3

19500 20000 20500 21000 21500

X-ray energ} [eV]

FIG. 10. Preedge subtraction' \. Preedge approximation.

B. Data aftei preedge subti action

The spectra obtained during a

measurement, the so-called raw data,

are usually plotted as absorption data

versus energy. The aim of processing

these data is to subtract the

monotonically decreasing

background (p0) due to the decrease

in absorption with the energy and

isolate the real EXAFS signal arising

from the interaction of the

photoelectron with the neighbouring

atoms. Before p0 can be determined,

the pre-edge has to be subtracted

(Fig. 10). The pre-edge is

approximated by a Victorccn

empirical form [19]:

V(EV E+co (13)

where Ci. C| and cu are fitting parameters. The pre-edge is extrapolated and subtracted from

the whole data range. After the pre-edge subtraction, the edge energy has to be defined. The

inflection point (maximum in the first deriv ativ e) in the edge region is generally considered

as a systematic choice for the edge (fig. 1 1). At this point, the signal of the isolated atom p0

21

Chapter 2

1 2

1

08

06

04

02

A

0

19970 19990 20010 20030 20050

X-ray energ} [eV]

05

>

Q

0 09

0 05 -

0 01

B

has to be determined and

subtracted from the data (Fig. 12).

The parametcriscd technique used

by XDAP to carry out this step is a

smoothing spline algorithm [20].

which is a polynomial function

defined over a series of intervals.

The ends of every interval are tied

together so that the function and

some derivatives (usually the first

and the second one) are made

continuous across the ends. The

calculation of the background is

controlled with the following

expression:

-0 03

19970 19990 20010 20030 20050

X-ra} energy [eV]

FTG. 11. Edge energy determination A. Edge legion.

B. Derivative of the absorption m the edge region.

V

1

S,- ux(E,)f-W, kj2

ye1 j

<SM (14)

Fhe value of SM is defined by the

user. W is a weighting parameter,

usually 0.075. S, is the value of the background that is varied in order to satisfy the

equation (14). The two following criteria have to be fulfilled in choosing SM: no

oscillations with frequencies similar to the EXAFS spectrum should be removed and low

frequency oscillations should not be included in the data. Both conditions can be easily

checked by observing the first derivative of the background that is more sensitive to the

oscillatory behaviour than its primitive function. Checking the Fourier transformation is

also extremch useful: when a too high SM v alue is chosen, signals below 1 Â start to grow

very fast.

22


Before %(k) can be obtained, the total absorption as obtained from experiments has to

be normalized per absorber atom. The spectrum obtained from background removal

corresponds to Au. ~ \i~ u0. According to equation (9)

Ii-Mox-

Eo

1 2

1 1

1

09

0 8 '

0 500 1000

X-rav energ} [eV]

0 1

M

-0 1

13

A

B

i{

Wave vector [X ]

FIG. 12. Backgtound subtiaction A. Absoiptton data and

calculated backgtound (dotted line), B. /_ data and

dcuvative of backgtound (dotted line)

Au must be divided by u,0. Ideally

one would like to measure the

absorption coefficient in the

absence of neighbouring atoms,

without any fine structure

oscillation. It is easy to understand

1500 that in practice this is not possible.

Ihe first solution is to divide the

absorption coefficient by the just

calculated background.

Unfortunately this is not always

acceptable. The background usually

decreases with increasing energy;

however, the negative slope can

vary strongly depending on the

experimental condition (materials in

the sample, gases in the ion

chambers). This could distort the

obtained results that are based, as

will be shown later, on comparing

spectra of the sample under anal} sis and spectra of reference compounds. Therefore, the

data were normalized with:

X(h)Au

(15)

Chapter 2

where \l{ is the calculated background absorption coefficient at 50 eV above the edge

(normalization by edge jump). This is allowed as long as the same procedure is applied in

the reference spectra with which the experimental spectra are compared. When using

theoretical references (see section "Data anal} sis*') [21], one has to consider whether it is

better to normalize the data b\ the edge jump, by the calculated background or by a linearly

or exponential!} decreasing function. Normalization b\ the background can produce large

errors, but, on the other hand, it takes into account the decay in absorption with energy.

Generally, the error induced by the normalization with the edge jump is small, but the

decrease in the atomic cross-section with energ} is neglected.

Once the x data are obtained, they are expressed as a function of the wave vector k,

using equation (7) to perform the transformation of E |eV| into k [A" ]. A further treatment

is thereafter needed in order to produce the data that are used to perform the actual analysis.

Applying a Fourier Transformation (FT) to the x data was first suggested by Sabers, Lytle

and Stern, who realised that this process allowed to obtain an estimation of the distance of

the neighbouring atoms [22]. In fact, the Fourier frans formation consists in applying the

complex formula:

1 .kma\ :^i ,

FT-p(r>-j= J X(k) eUvl dk (16)

As result of this complex function one obtains a real (Re) and an imaginary (lm) part. The

magnitude of the function, or the absolute part, is obtained with the following equation:

Absolute part - y/lm2 ~ Re^ (17)

Plotting the absolute part of the transform versus the distance r\ a figure is obtained with

peaks at the distance f roughly corresponding to the interatomic distances between

absorbing and backscattering atoms, r* is about 0.2-0.5 A shorter than the actual distance

due to the phase factor m the sine term of the yjk) function, eq. (10).

24


The Fourier Transform can be taken w ith different k-weighting (multiplication by k ):

FT---pn(0=-?==1 Ikniak"x(k)ei2k,'dk (18)\J2ti k>"i"

This option can be used to distinguish between high and low Z scatterer atoms [23]. An

element with low mass (like ox} gen) will scatter mainly at lower k-values, while high mass

elements (like mohbdenum) will scatter significantly at higher k-values. In fact, the peak in

the FT corresponding to the heav} atom will grow more, relatively to that corresponding to

the light atom, when going from the k - to the k -weighted FF. Because of the non-linearity

of the phase shift as a function of k. the position of a light scatterer peak in the FF will shift

to larger distances going from k to V weighting. This effect is much less pronounced with

heavy atoms. In general a k'-weighted transform is more advisable because it minimises

chemical effects on the EXATS signal (e. g. small changes in E0 values) and makes the

oscillations more uniform by compensating for the natural decrease of the signal due to the

1/k factor present in eq. (11) and to the backscattering amplitude, which falls off

approximately as 1/k".

Data analysisThe analysis of EXAFS data is usual!} based on equations (10) to (12) that combined

give the following expression:

ShellsN.

X (k) - X —L S02 F,(k) c"2k-cV c"2Ri X{k)sin(2kR, + fl>,(k)) (19)

j k Rp

Chapter 2

The aim of the analysis is to reproduce the data obtained experimentally using this

equation, i.e. to find out the structural information of each shell around the absorber atom.

Considering the fact that an EXAFS spectrum can be composed of one to five shells, the

first step is to concentrate on the single shells, starting from the ones at the lowest

distances. For this purpose a window has to be chosen in the FT spectrum for the selected

shell (Fig. 13). Ehe analysis should deliver the coordination number N, the distance from

the absorber R and the Delwe-Waller factor o" for each single shell. Among these

2RA,parameters also the quantities So, F, e"

,and <I> arc unknown. To circumvent the

complexity of a system with such a large number of variables, the use of reference

compounds was adopted. For a given pair of absorber and backscatterer, a reference

compound is taken and its FXAFS spectrum is either measured under the same conditions

as the investigated sample, or simulated by means of specialised programs. The code Feff

(Version 7), developed at the FAiversity of Washington [24], was used in this work to

produce the spectra of reference compounds and simulate EXAFS spectra of compounds

with a structure known from crystallographic studies.

Absolute part

(magnitude)

Imaginary part

6 7 8

FIG. 13. k'-weighted Fourier transformation of MoS? I he dashed rectangle represents the window isolating

the first shell (Mo-S).

<

30

15

3

O

dJ

-15

'5

-30 \i.

3 4

RlAl

26


Since the coordination number and the distance of the chosen shell in the reference

compound are known, its wave function is expressed as:

N r

X.er (k) - 7-rH A.ei (k) sin(2kRiel + <I>„IVU) (20)kKief

In this expression the term Alcl is composed of:

A,el<k) - S02 Fiel(k) e^cf e-2R«^k) (21)

From equation (20) the redefined amplitude AIC| and the phase shift <Flcf can be calculated

and used in the equation that describes the spectrum of the investigated shell and the

function to be minimised becomes:

Xevp (10 " HT^ Ad (1C) c"2k'2A°2 sin(2k'Rshcll + <Die(fk')) (22)[v Kshell

The transferability of A1Cf and <I>1C( has been shown for compounds with the same absorber-

scatterer pair [25. 26], and even for absorbers or backscatterers which arc neighbours in the

periodic table [27. 28]. The parameters that have to be optimised arc written in bold

characters. Ao~ is the Debye-Waller factor relative to the one of the reference compound. It

is not possible to calculate direct!} <r of the sample because in the expression for the

amplitude of the reference compound A1C| the Delwe-Waller factor of the reference is

contained. It can be noticed that the calculated term of eq. (22) is a function of k' instead of

k. k" is the photoelectron wave vector corrected for the difference in inner potential

between the sample and the reference compound (AE()):

2mck^irAEo (23)

27

Chapter 2

AEo is the fourth parameter that is optimised during the minimisation process that consists

of anon-linear least squares fit procedure.

For each shell four parameters are optimised, fhe number of variables is, however,

limited by the k-range used in the x(k) data and by the R-range of the fitted window in the

FT spectrum, fo calculate the number of free parameters (degrees of freedom) the Nyquist

theorem is used |29]:

P-^^.l (24)

This series of formulae stays behind the fitting procedure. The practical development

involves the production of a reference file for each shell that has to be fitted, the guessing

of a starting value for the parameters that have to be optimised and the judgement of the fit

produced by the computer. Each of these steps has great importance. Ehe chosen reference

should be as similar as possible to the studied absorber-backscattercr pair. The choice of the

starting values for the iterative process has an influence on the results. It is therefore useful

to try different starting values in case no convergence is obtained. The result of the fits has

to be judged firstly optically, comparing the simultaneous plot of the experimental and

calculated data. This should be done for the x and for the FT data. Since N and <f arc

strongly correlated, it is possible to obtain several solutions for the same spectrum with

different N and a" values. It is. therefore, always useful to compare the data using different

weighting factors. A good fit should give similar results using k and k' as weighting

factor. It is also very important to consider the statistical errors calculated from the fit to

have a more accurate idea about the quality of the fit. This step consists in checking the

standard deviation for each parameter and consider the variance of the fit j 30].

28


Pros & Cons ofEXAFS

To conclude this chapter some comments have to be added in order to give an

overview on advantages and disadvantages of the discussed spectroscopic technique and to

add some word of advice that are connected to the interpretation of the data presented in

this thesis.

The three main features that made FXAFS such a useful characterisation technique,

particularly in heterogeneous catalysis, are the following:

• Since the edge energy is typical for even element, this technique is element

specific; it allows to concentrate on a single element even if it is mixed with other

elements in a compound.

• The information obtained from EXAFS is due to the fact that the electrons excited

by the X-tays are backscattered by the neighbouring atoms. Since the mean free

path of such electrons is usually smaller than 10 A. EXAFS provides the

structural parameters concerning the local structure around the absorber atom.

• Because of the high penetrating capacity of X-rays through matter, the results

attained with EXAFS are an average of all species of the chosen elements present

inside and on the surface of the studied material. It is therefore a bulk technique.

A further advantage which was utilized during this thesis is the possibility of time

resolved experiments. Quick FXAES and Dispersive EXAFS allow to follow how a

reaction proceeds in situ. The quality of the data is usually reduced in comparison with

classical EXAFS. but a lot of useful information can be attained.

Among the disadvantages of EXAFS there is the usually limited beam time that is

available at synchrotron sources and the high sensitiv it} to experimental procedures. Fhe

samples have to be as (macroscopicalh) homogeneous as possible: the presence of

inhomogeneities or small holes provokes big experimental errors. Another factor that has to

be avoided, in order to eliminate the presence of glitches and irregularities in the data, is the

presence of higher harmonics in the X-ray beam that hit the sample. This can be obtained

by using proper harmonic rejection s} stems.

29

Chapter 2

As far as the interpretation of the data is concerned, it is always important to keep in

mind that the parameters obtained from the fits contain some error, partially due to

experimental problems, partiall} to the insufficient statistics (low signal to noise ratio). The

experimental errors can be almost eliminated by means of a proper sample and beam line

preparation. The statistical errors are usually reduced averaging several scans of the same

sample, typically 3-5. Another aspect that has to be considered when Interpreting EXAFS

data is the underestimation of the coordination number. Ehe reason for the false estimation

is a disorder in the sample either due to thermal vibration (c. g. [31]) or to structural

disorder (e. g. |32]). Fo reduce the thermal disorder samples are usually cooled to liquid

nitrogen or liquid helium temperatures. Ehe static disorder is. on the contrary, a feature that

can not be reduced. To understand the effect of the static disorder on the value obtained for

the coordination number, one has to keep in mind that every absorber-backscatterer pair

produces a contribution that, summed with the other ones, will produce the oscillation

observed in an EXAES spectrum. Every contribution consists of a sinusoidal oscillation

that arises from the term sin(2kR \ O(k)) in equation (12). The phase shift <D(k) is

dependent only on the kind of backscatterer atom. An absorber atom with two

backscatterers of the same kind but with slightly different distances will produce two

oscillations with different periods. The amplitude of the curve obtained by the sum of these

oscillations will reach a maximum when the single oscillations overlap as depicted in Fig.

14. If the overlapping of the two sinus functions takes place in the experimentally

measurable k-range. the estimation of the coordination number will be correct. lfy on the

contrary, this maximum is outside the accessible k-range. an underestimation of the

coordination number will occur. Since for a small difference in R. the maximum will be at

high k values, this factor has to be taken into account for low Z-backscattcrers. This is due

to the fiict that for light elements the maxima of the oscillation are reached at relatively low

(< 3 A"1) and high (> 10 A"1) k-values. ITowever, in the region k <^ 3 A"1, the signal is still

influenced by edge effects and is therefore excluded from the used data, whereas in the

region k > 10 A"1 low Z-elements scatter ven weakly and the signal at high k values is low,

so that no information can be obtained from this latter region, neither.

30


B=sin(2 k 2)

-1 5

2

-2 5

25

2

05

0

-0 5

I

-1 5

-2

-2 5

10

A-B

15 '

FIG. 14. The sum of two sinus functions with different periods

Chapter 2

Literature

1. Liebhafsky. 11. A., Pfeiffer, FI. G., Winslow, E. H., and Zcmany, P. D., "X-rays,

electrons, and anah tical chemistry". John Wiley & Sons, New York ( 1972).

2. Agarwal. B. K.. "X-ray Spectroscop} ". Springer-Verlag. New York (1991).

3. Raoux, D., Introduction to synchrotron radiation and to the physics of storage rings,

/'// "Neutron and synchrotron radiation for condensed matter studies". Springer-

Verlag, Berlin. 37 (1993).

4. Freund. A., X-Ray optics for synchrotron radiation, in "Neutron and synchrotron

radiation for condensed matter studies". Springer-Verlag. Berlin, 88 (1993).

5. Frahm. R.. Physica B 158, 342 (1989).

6. Frahm. R.. Rev Sei Instriim 60(7). 2515 (1989).

7. Shido. f., and Prins. R.. Current Opinion in Solid State and Material Science 3 (4),

330(1998).

8. Fontaine, A., Interaction of X-rays with matter: X-ray absorption spectroscopy. /';;

"Neutron and synchrotron radiation for condensed matter studies", Springer-Verlag,

Berlin. 353 (1993).

9. Kosugi. N.. Theory and Anah sis of XANES. in "X-Ray Absorption Fine Structure

for Catalysts and Surfaces". Iwasawa. Y. Eds.. World Scientific. Singapore, 59

(1996).

10. Durham, P. J.. Theory of XANES. in "X-Ray Absorption", Koningsberger, D. and

Prins, R. Eds.. Wiley. New York. 53 ( 1988).

11. Bethe, FF. and Salpeter. F.. "Quantum Mechanics of One and Two Electron

Systems". Springer-Verlag, Berlin (1957).

12. Lee. P. A.. Teo. B. K.. and Simsons. A. L.. J.im Chem Sue. 99, 3856 ( 1977).

13. Fee, P. A., and Pendry, J. B„ Phys Rev B 11, 2795 (1975).

14. Stern. E. A., rim Rev B 10. 3027 ( 1 974).

15. Stern. E. A.. Savers, D. E.. and Lytic. F. \V„ Pin s Rev B IE 4836 (1975).

16. Fco, B. K., "EXAFS: Basic Principles and Data Analysis", Springer-Verlag. Berlin.

28(1986).

32


Stern. E. A., Eheory of EXAFS. in "X-Ray Absorption". Koningsberger, D. and

Prins, R. Eds., Wiley. New York. 37(1988).

Vaarkamp, M., Dring, F, Oldman. R. T. Stern, E. A., and Koningsberger, D. C,

Phys Rev. 5 50.7872(1994).

Savers, D. E.. and Bunker, B. A.. Data analysis, in "X-Ray Absorption",

Koningsberger. D. and Prins, R. Eds.. Wiley. New York, 215 (1988).

Cook, J. W. ,k, and Savers, D. E.,./ Appl Phys. 52. 5024 (1981).

Haskek D., Ravel, B., Newville, M.. and Stern, E. A., Physica B 209-209, 151

(1995).

Sayers. D. E.. Stern. E. A., and Lytle. F, W.. Phvs. Rev Lett. 27, 1204 (1971).

Vaarkamp. M.. Cat. Poday 39. 271 (1998).

Ankudinov. A. L., and Rehr, J. T, Phys Rev B 56. R1712 (1997).

Bunker. B. A., and Stern. F, A.. Phys. Rev. B 27. 1017 (1983).

Citrin. P. IT.. Eisenberger. P.. and Kincaid. B, M., Phys. Rev. Lett. 22, 3551 (1976).

Lengeler. B., J. Phys (Paris) 47, 75 (1986).

Teo. B. K.. and Lee. P. A.../. .4?;?. Client Soe. 101. 2815 (1979).

Brigham, E. O., "The Fast Fourier Fransform", Prentice Hall, FTiglcwood Cliffs,

New Jersey. (1974).

Vaarkamp. M., Linders. .1. C, and Koningsberger. D. C, Physica B 209 (1-4), 159

(1995).

Rockenberger. J., froger. L., Komowski. A.. Vossmeyer, E., Eychmüller, A.,

Feldhaus. T. and Weller, IF. J. Phys Chem B 101. 2691 ( 1997).

Shido. T., and Prins. R.. J. Phys. Chem B 102, 8426 (1998).

33

Chapter 2

i

i

Chapter 3

How does the Structure of NiMo/Si02 Oxidic

Precursors Influence the Activity of the Final

Sulfided Catalysts?

Introduction

The most stable Mo species on the silica surface, which has a point of zero charge at

pH 2, are polymolybdates [Mo-024.m(OIl)mj(A"mV [1. 2], although Raman studies suggested

the presence of interactions with the support and the formation of molybdosilicic acid |2.

3J. Nickel can interact with silica to form Ni phy llosilicates [4, 5]. This was. however,

seldom observed for Ni/SiCE samples prepared by the classical incipient wetness

impregnation method |6].

The addition of chelating ligands. such as nitrilotriacetic acid (NTA) or

ethylenediamine tetraacetic acid (EDEA). to the impregnating solution has a beneficial

effect on the catalytic activity of NiMo/SiCE catahsts [7-12]. fhe resulting catalysts are

even more active than alumina-supported catalysts. Ehe improved activity has been

ascribed to a better dispersion of nickel and molybdenum on the support in the catalyst

precursors and. in the case of NTA. to the delayed sulfidation of Ni [12]. Both factors

would enhance the formation of the so-called Ni-Mo-S type II phase, which, at present, is

considered to be the active phase in hydrotreating catalysts [9, 13, 14].

Previous work in our group concentrated on the effect of NTA on Ni and Mo in the

impregnating solutions and in the catal} st precursors |1]. It was found that NTA

Chapter 3

preferentially complexes Ni and causes a delà} in its sulfidation [1, 12]. Eo extend this

work, in this chapter the effect of other chelating ligands on NiMo/SiCE catalyst precursors

is reported, especially ethylenediamine (EN), and a more detailed structural description of

the metal ions on the support is given. The structure of the catalyst precursors is correlated

with the catalytic activity of the final sulfided catalysts in the hydrodcsulfuri/ation of

thiophene at atmospheric pressure. The explanation of the structure of the oxidic precursors

helps, therefore, to understand the factors that Influence the formation of active sites on the

support during sulfidation. In particular, the structural changes in the catalyst precursors are

revealed by means of extended X-ray absorption fine structure (EXAFS) and by UV-VIS

and Raman spectroscopy, as a function of the concentration of two chelating ligands,

cthylenencdiamine (EN) and nitrilotriacctic acid (NTA). The information attained with this

work is essential to understand how Ni and Mo behave during the sulfidation process. As

the next chapters of this thesis will show, a mechanistic explanation of the sulfidation of

NiMo/SiOo is possible only after having a clear idea about the structure of the catalyst

precursors.

Experimental methods

Sample preparation and characterization

All catalysts were prepared by pore volume impregnation with a solution containing

the metal salts and the chelating ligands. The solution was prepared by dissolving the

organic complcxing ligand in 15 ml of aqueous ammonia (25%, Flukapuriss. p.a.), adding

3.60 g M0O3 (25 mmol Fluka puriss. p.a.) and heating the solution to 80°C in order to

dissolve MoO, [1], After cooling to about 60°C. 2.18 g Ni(N(>,)v6(IEO) (7.5 mmol, Fluka

purum p.a.) were added and dissolved. At this point the solution turned from uncoloured to

the colour typical for the formed Ni complex, that varied from blue to dark violet

depending on the concentration and the kind of ligand employed. Ehe pH was then adjusted

to 8.0 with a 2 M UNO-, solution and the solution was diluted to 25 ml with deionized

water.

36

Structure of Precursors and Catalytic Activity

Several chelating ligands were tested. The investigated catalyst precursors arc listed in

Tabic 1. In all cases, the support was SiCE (C560 Chemie Uetikon), which had a particle

size of 125-250 um, a BEE surface of 565 m2/g. a BEE pore volume of 0.83 ml/g. and

which had been dried overnight at 120°C prior to impregnation. After impregnation the wet

powder was again dried overnight at 120°C with a heating rate of 2°C/min. To avoid the

destruction of the metallorganic complexes in the catalyst precursors before the sulfidation,

no calcination was carried out at higher temperatures.

TABLE 1. NiMo'SiCE catalysts prepared with different chelating ligands (molar ratio

Ni:Mo:ligand=0.3:1:0.3)

Catalyst Loading Loading nL/Ni log [Ki(NiL)j

Ni [%] Mo I%]

NiMo 1.4 7.5

NiMo-nitrilotriacctic acid (NTA) 1.3 7.2 4 ll.5

NiMo-EDEA 1.3 7.0 6 18.0

NiMo-cthy lenediamine (EN) 1.3 6.9 2 7.5

Ni-Mo-diethylenetriamine (D f) 1.3 7.0o

j 10.7

NiMo-tricthylcnctetraamme (ET) 1.3 7.2 4 14.0

NiMo-tetraethylenepentamine (TP) 1.3 6.9 5 17.8

NiMo-pentaethy lenehexamine (PIT) 1.3 7.3 6 19.3

NiMo-18-crown-6 1.3 7.3 .1

NiMo-formic acid 1.3 7.1 l 0.5

NiMo-citric acid 1.3 *7 1 4 4.2

UV-VIS reflectance measurements were carried out on a Perkin Elmer Lambda 16

spectrometer, equipped with an integration sphere that allowed measurements in the

reflection mode, using pure SiCE as a reference. Solutions were measured in 1 cm. quartz

cells. The NK concentration in the anah zed solutions was 0.03 mol/1, therefore the

impregnating solutions had to be diluted.

37

Chapter 3

The Raman measurements were performed on a Bruker Equinox 55 spectrometer

equipped with an F'E-Raman-Modul FRA 106. with a Nd:YAG laser (1064 nm). a CaF2

beamsplitter, and a Ce detector (Ü418-S) cooled by liquid nitrogen. About 10 mg of the

samples were pressed into a hole (diameter 2 mm. depth 1.5 mm) on an aluminium disk

(diameter 1 cm). The number of accumulated spectra for each sample was 4096.

Sulfidation and reaction

The sulfidation and thiophene HDS reaction occurred at atmospheric pressure in the

apparatus represented in Fig. 1. The presence of two parallel reactors allowed the

simultaneous sulfidation of a catalyst in one reactor and the catalytic testing of another one

in the other reactor, "fhe reactor contained in a quartz tube (inner diameter 13 mm. outer

diameter 16 mm, length 350 mm) equipped with a quart/ frit in the middle. A layer of

about 20 mm quartz wool was pressed over the frit. 100 mg of the catalyst precursors were

diluted and mixed with 1 g dried SiCE and the obtained powder was poured in the reactor

on the quartz wool with some additional quartz wool (20 mm) on it. Ehe quartz reactor was

heated by means of ovens consisting of an insulated cUindrical copper-bronze heating

block, into which four high power heating cartridges (220 V, 160 W each) were fitted. The

oxidic precursors were sulfided at 400°C (heating rate 6°C7min) for 2 h with a mixture of

10% ILS in FE (Messer Griesheim 3.0) that (lowed through the reactor from the beginning

of the heating process. Ehe activity of all catalysts was tested in the hydrodesulfurization of

thiophene at 400°C [12|. 'fhe feed, consisting of 3% thiophene in EL, was obtained by

bubbling PL through a series of four thiophene saturators that were cooled to 2°C. The

contact between the catalyst and the mixture FEThiophene had to be controlled very

carefully, 'fhe flow rate was always increased vet"} slowly from 0 to 75 mEs. The flow rates

of ELS and IE were regulated by means of Brooks thermal mass flow controllers. Ehe

products of the desulfurisation reaction were conducted directly to a gas Chromatograph

(Hewlett Packard FTP 5890) and analyzed on line. The temperature of the gas

Chromatograph was kept constant to 40°C and a good separation of the products was

obtained (in order of appearance 1-butène, n-butane. trans-2-butene. cis-2-butcne and

thiophene). The sample was injected in the column by means of a six-way-valve and a 125

pi loop, 'fhe column was a WCO f fused silica 0.25mmx 50 m capillary column with a CP-

38

Sti uctui e of Pi ecursors and Catalytic Activity

SIL 5CB coating A flame ionization dctectoi (FID) was used to analyze the composition of

the outcommg gas Ehe data were collected even 20 mm with a Nelson 900 Analytical

intelligent mteiface (Peikm Elmei), fiom which they were transferred to a personal

computet, wheie the data anah sis was petfoimed with the piogiam Turbochiom Veision

6 0 2 (Peikm Elmei) fhe tesults obtained E; h aftei the fust measiuement weic used foi

the determination of the kinetic paiameteis

h s:

Ile

II

NHL)t i

ne

ri\

c i, 11,

s I

HC

Btooks theimal mass

flow contiollei

f I Filter

On-oft \al\e

rc

I nie metenne; \al\e

Back pi essuie icgulatoi

Ihiee-/fout-vva\ valves

I IG 1 Plow scheme ol the tl lopiiciie hAdiodcsullutizition

1 Biooks the mill in iss llow conti olsi

2 I hcimostit (set it 2 C) with I thiophene sit ii itois

3 Oven loi icictoi

l Icmpei iluie eontiolkr

") Ois ehiomtto^nph IIP ->S)0

6 Intel (ice Pctlun I Imel )0H

/ Petsonil compute)

S Bubble llow nutet

(1 (low I lempetiluie 1 mlie it<u Ceoitiolle)

19

Chapter 3

The reaction rate was described with a first order expression in the thiophene partial

pressure pu,. The reaction rate constant was calculated assuming an ideal steady-state plug

flow reactor behaviour [J 5] with the formula:

k=irMrâJ (,)

where F0,, is the molar thiophene flow at the reactor inlet. W the catalyst weight and X the

thiophene conversion defined as:

AT1PTjA-p|

X=l-— M -——^ (2)P?h AIh+lACt

where p^ is the initial thiophene partial pressure. Aj, is the thiophene chromatographic

peak area and 2jA^ is the sum of the contributions of 1 -butène, n-butanc and eis- and

trans-2-butcne. If one takes Into account that the response factor of the FID built in the GC

is not the same for all species, the mass balance is in good agreement with this evaluation

of the conversion 116].

In order to directly compare catalysts with slightly different metal loadings, another

reaction rate was calculated based on the Mo loadine:

IE -k-|^-| (3)'nui,,

' v

where is the Mo loading in the oxide catal} st precursor.-tflsupp

40


EXAFS measurements

EXAFS spectra of several compounds were measured at the Ni and Mo A-edges. Ehe

data were collected at the Swiss Norwegian Beam Fine (SNBL, BM1) at the European

Synchrotron Radiation Facility (ESRF), Grenoble. France, fhe electron energy and ring

current were 6 GeV and 130-200 niA, respectiveh. At SNBL the incident X-rays are

monochromated by a Si (111) channel-cut monochromator and harmonics arc rejected by a

gold-coated mirror positioned at an angle of 7 inrad to the beam. The X-ray intensity is

monitored by ionizing chambers and the estimated resolution is 1 eV at the Ni A-edge and 2

eV at the Mo A-edge. For measurements at the Ni A-edge. the chambers for measuring In

and E were 17 and 31 cm long, and were filled with pure nitrogen and a mixture of argon

and nitrogen (Ar/AE - 40A0). respectively. Fhe chambers used for measurements at the Mo

A-edge were 17 and 62 cm long and filled with pure argon and a mixture of krypton and

nitrogen (Kr/Ni - 25/75), respectively. Some of the data were collected at the General

Purpose Italian Line for Diffraction and Absorption (GILDA, ID8) at the ESRF [17].

Ehe Ni A-edge spectra were divided into six regions: 7780-8300 eV (pre-edge). 8300-

8370 eV (edge region) and the four post-edge regions (3-6.5, 6.5-10, 10-15, 15-17 A"1)

between 8370 and 9433 eV. Ehe collection times for the data points in each scan region

were 1, 1, 2, 3. 4 and 4 s. respectively. As far as the Mo A-eclge was concerned, the

collection times for each data point in a scan were 1, 2. 2, 3 and 3 s for the intervals 19454-

19954 eV (pre-edge), 19954-20164 eV (edge region) and the three 6.5-10, 10-15, 15-21 Â"1

post-edge regions between 20164 and 21684 eV. respectiveh. Fhe distance between the

post-edge data points was determined so that the difference in their k values was smaller

than 0.05 A"1, Five scans were averaged for the Ni A-edge and three for the Mo A'-edgc.

Catalyst samples were pressed into self-supporting wafers and mounted in an in situ

EXAFS cell [18]. fhe thickness of the samples was chosen to adjust the total absorption to

u.x •=- 4 for the Ni A-edge (low Ni concentration) and the edge lump to 1 for the Mo A-edge.

Ehe sample was cooled to liquid nitrogen temperature for all measurements.

41

Chapter 3

EXAFS analysis

The program XDAP (version 2.1.5) was used to analyze and fit the data [19. 20]. The

pre-edge background was approximated by a modified Victorcen curve and the background

was subtracted using a cubic spline routine. The spectra were normalized by the edge jump.

The k'-weighted, and in some cases k1 -weighted EXAFS functions were Fourier

transformed and fitted in R-space. Ehe free parameters were interatomic distance,

coordination number. Debye-Waller factor and the correction of the edge energy. The

errors of the parameters were statistically estimated using the random errors of the observed

data. The goodness of fit was calculated for every model from the k- and R-space fit range

and the number of free parameters [21].

Reference spectra were calculated using FelT7 [22] for several cluster models that

will be discussed in more detail later. Crystallographic data were obtained from the

Inorganic Crystal Structure Database (ICSD-CRYS'llN) and the Cambridge Structure

Database (CSD).

42


Results

Catalytic Activity

The activity of catahsts prepared with various chelating ligands (Table 1) was

investigated. The ligands can be divided in three groups: the chelating amines, such as

ethylene diamine (EN), diethylenc triamine (DT), triethylcne tetraamine (IT), tetracthylene

pentamine (TP) and pentaethylene hexamine (PH), the combined amino-acetic acid-

containing complexes such as NTA and ED fA, the organic formic and citric acids, and

the crown ether 18-crown-6 (CT). All amines form very stable Ni complexes [23 j. In a Ni-

CE complex reported in the literature, Ni coordinates three of the six oxygen atoms of the

organic complex |24j. Ehe stability constants of the Ni complexes of formic acid and citric

acid are lower than those of the amines (Table I) [23]. Attempts to use oxalic acid were

abandoned because of the low solubility of nickel oxalate.

The results of the hydrodesulfurization measurements of thiophene at atmospheric

pressure are presented in Fig. 2. fhe Ni:Mo:ligand molar ratio was 0.3:1:0.3 in all catalyst

<4

CO

M

^4

00

p

0.12

0.1

0 08

0.06 -

CO

OO

0 04

CD

0.02 -

0

^P „O^ „V A^ n<0 (ZT M <<y *-sN>

$ XÎV ^ xs# K®^ ^ # X „A&

r>>

XP <?#

FIG. 2. Thiophene I IDS activ ity of Xi\lo SiO cataly sts prepared w ith different chelating ligands.

43

Chapter 3

precursors. Catalysts prepared with NEA and EDEA show the highest rate constants. Ehe

ni/Ni column in Table 1 gives the number of coordinating ligand atoms available per Ni

atom, taking into account all potential coordinating atoms and that the Ni/ligand ratio is 1

for all samples. The number of theoretically coordinating atoms of CE was set at three as

suggested in [24].

-àCO

M

oo

CD

00

Q

0.12 i .

*

;EDTA ,'

O,

0 1

0 0c fJ'

0.06

0 04

0.02 -

EN

NTA

ligand/Ni molar ratio

FIG. 3. Thiophene HDS activity of NiMo/Si02 catalysts prepared with different amounts ol

ethylenediamine and nitrilotriacetic acid (NEMo^O.S: l molar ratio).

The concentration of the complexing agents was varied for two catalysts in order to

follow their catalytic activity as a function of the composition of the coordination spheres

around the two metals. NfA and FN were chosen because of their different structures and

because of what is known about their complexing properties. In Fig. 3, the thiophene ITDS

rate constants of the obtained catahsts are plotted as a function of the ligand to Ni ratio. For

NEA a dramatic increase in catalytic activity is observed for the ligand to Ni ratio between

0 and 1. A maximum is reached between NTA'Ni ~ 1.5 and 2; the catalytic activity

decreases for higher NEA/Ni ratios. "The activity of catalysts prepared with EN is generally

lower but increases at first rapidly, then more gradually until EN/Ni = 4.

44


Characterization ofNickel

EV-VTS Spectroscopy. UV-VTS spectra of the impregnating solutions and of

several dried catalyst precursors were collected to investigate the coordination ol the NE

ions. Replacing coordinating oxygen atoms with nitrogen atoms shifts the VIS absorption

bands in the region between 500 and 700 nm. attributed to the JA2g(F) to Tig(F) transition,

to lower wavelengths [25]. Table 2 lists the wavelengths of this absorption band for

aqueous solutions of NkNCEEAfFliO) with different amounts of EN. The solutions were

prepared In the presence and absence of NEE. Ihe absorption maxima of both solutions

shift to lower wavelengths with increasing EN concentration. Those species that, according

to thermodynamics, should be present in aqueous Ni-EN solutions were calculated from the

stability constants of the Ni2' complexes using an extended version of the program Spex [1,

26. 27]. In the solution with EN/Ni ^ E 63% of Ni"' should be present as

[Ni(EN)(HA))t]2 20% as [Xi(ENTh(FI20h]2 and 17% as ENi(lEO)6]2+. Hence, in passing

from the solution without EN to the solution with FNXti =• 1, about one third of the oxygen

atoms of FLO coordinated to Ni are replaced with nitrogen atoms of EN. This explains the

large difference in the wavelength of the absorption bands of the two solutions. In the

9 i

aqueous solution with EN/Ni = 2, about 85% of Ni" should be present as

[^(ENEOEOEEE ,whereas for EN/Ni - 3. |\i(EN),|2 corresponds to 98% of all Ni2"

present in the solution.

TABLE 2. UV-VIS absorption bands of Ni-EN complexes in water and ammonia solutions

EN:Ni In H^O In ammonia solution

|nm] (8 M, pH 8.0) [nm]

0 723 608

1 618 595

2 565 562

3 545 546

45

Chapter 3

TABLE 3. UV-VIS absorption bands of the impregnating solutions (pH-8.0, Ni:Mo=0.3:l)

and of the dried catalyst precursors

9

10

9

8

7

6

5

4

3

2

1

EN:Ni Impregnating Dried catalyst

molar ratio solutions [nm] precursors [nm]

0 620 708

0.16 594 709

0 66 583 710

1.33 557 659

530 585

6.66 523 549

A

V

300

EN Ni=3 3

ENNi=1 33

EN NfO66

EN Ni=6 6

No ligand

400 500 600 700

Wavelength [nm]

^

800

FIG. 4. UV-VIS spectia of NiMoFN impiegnatmg •solutions containing diffeient amounts ot

ethylenediamine. ]Nf|,M- 0 0~î mol I. pi 1-8

46


1 he absorption maxima of the ammonia solutions have lower wavelengths than those

of the aqueous solutions for EN/Ni ~ 0 and 1 but are practically the same for higher EN

concentrations. This is due to the fact that when no EN is present in the solution, Ni2"1 is

present as 45% [NiCNHOe]2 -45% [Ni(NIl ATEO)^ and 10% rNi(NH3)4(FI20)2]2E When

KN is added to the solution, mixed Ni-EN-NFE,-I EO complexes are formed.

Ehe spectra of some NiMoEN impregnating solutions (Fig. 4) show that an increasing

EN concentration affects the Ni"1 environment. In the regions around 850, 500-650 and

300-400 nm, the absorption bands shift to higher energies with increasing EN

concentration. Ehe absorption maxima for the bands m the region between 500 and 650 nm

are listed in fable 3. Ammonia is present m the solutions; thus, the absorption maxima

should be compared with those of the ammonia solutions in fable 2. Ehe spectrum of the

0 07

0 06

s

0 05

0 04

0 03

0 02

0 01

ENNi=6 66

\ l

\\

ENNi=3 33

ENNi=1 33

ENN!=0 66

A"'aJ»

^.

A //

No ligand

^v,VMAAfv,Af

400 500 600 700 800

Wav elencth I nm]

FIG. 5. Reflectance UV-VIS spectra of dned \i\fo SiO cataKst ptecuisois piepaied with diffeient

amounts of ethv lenedtamine

47

Chapter 3

impregnating solution without EN has a maximum at 620 nm, whereas the Ni-NFE solution

has a maximum at 608 nm. Ehe impregnating solutions with the highest EN/Ni ratios (3.33,

6.66) show absorption bands with particularh low wavelengths. Ehe presence of Mo

explains this shift of the maxima, because it is the only parameter that is different to the

ammonia solutions. 'Fhe reflectance UV-VIS spectra of the dried catalyst precursors are

presented in Fig. 5. Mo compounds on the support absorb very strongly below 500 nm.

covering all signals arising from nickel in this region. Nevertheless, it is possible to

distinguish an absorption band for each spectrum. The absorption maxima arc listed in

Table 3. From the data and the spectra, it is clear that, for the catalyst precursor without a

ligand and for catalyst precursors with EN Ni = 0.16 and 0.66, almost no Ni-amine

complexes were present after drying. Some EN was apparently removed from the Ni2h ions

in the catal}st precursors with higher ligand concentrations (see Tabic 3), because the

wavelengths of the absorption maxima for the dried catahst precursors with EN/Ni 1.33,

3.33 and 6.66 are higher than those for Ni" complexes containing the same number of

amine ligands. Ehe values for Ni-EN complexes in aqueous solutions (Eable 2) suggest that

the dried catalyst precursors with EN/Ni - 1.33. 3.33 and 6.66 contain less than 1. fewer

than 2 and slightly fewer than 3 EN molecules per Ni"" ion. respectively.

NiXAS of NiMoEN/SiCE. To follow the change in structure of the catalyst

precursors as a function of the amount of ligand used during the preparation, we compared

fhe Fourier transformed yjkfk' functions measured at the Ni A-edge for five different

catalyst precursors: NiMo SiCE and four NiMoEN'SiCE compounds with EN/Ni = 0.16,

0.66. 3.33 and 6.66.

The Ni A-edge Fourier transformed x(k)-k' functions are shown in Fig. 6 (k-range:

3-14 À"1). Ehe results of the fits are listed in Eable 4. In the NiMo/Si02 spectrum, there are

two main peaks at 1.6 and 3 A (uncorrected for phases). The first is ascribed to the oxygen

atoms that surround the Ni" ions. The reference for oxvgen was produced with the Feff

code using a Ni-0 distance of 2.12 A. The first shell of the catalyst precursors prepared

with EN was analyzed with references for Ni-0 and Ni-N that were calculated with the

Feff code, with distances and coordination numbers taken from the crystallographic data

of the complex Ni(EN)2-2EEO [28]. Ehe first shell of the Ni2 ions of the catalyst precursors

48


0 12 3 4 5 6

R[A|

FIG. 6. Absolute pairs of the formet tiansforms of the \'i x-edge ySf)k EXAFS of NiMoAtCA catalyst

piecursois as a iunction ol the amount of ethylenediamine used dutmg the prepaiation (k-range 3-14 Ä ')

with EN/Ni = 0.16 was fitted using an oxygen contribution only, since EIV-VIS showed

that no EN molecule is surrounding Ni in the catah st precursor (sec Eable 3). In contrast,

for the catalyst precursors with 1 N'Ni - 0 66 and 3 33. both the nitrogen and the oxygen

contribution were used for the fitting. Nevertheless, a relativel} low Ni-N coordination

number (0.2) was obtained. For the EN'NT - 3 33 catalyst precursor, it was 3.7, whereas

fitting with oxygen gave a coordination number of 3 2. Ebese results agree with the

UV-VIS measurements that rev ealed Ni to be surrounded, on average, by fewer than one

and by two FN molecules for the catahst precursors with ENAh - 0.66 and 3.33,

respectiveh. Moreover, in the same spectrum, the features of a Ni-EN coordination become

visible. Ihe peak at 2.5 A (not phase-corrected) arises from the scattering of the carbon

atoms of EN. as was seen in a Eefi7 simulated spectrum (Fig. 7) of the complex

49

Chapter 3

Ni(EN)2-2IEO (28], from which we calculated the Ni-C reference. Ehe small peak at

2.15 A does not belong to a real distance, it is caused by an interference of the signals of

the nitrogen and oxygen atoms in the first shell. Ihe Ni-C coordination number of 2.1 for

the catalyst precursor with FN/Ni= 3.33 is. however, lower than expected; since it should

equal the Ni-N coordination number (3.7). Ehe greater distance from the central atom may

lead to an enhanced mobility of the carbon atoms and a resulting underestimation of the

coordi nation number.

TABLE 4. Structural parameters resulting from the Ni A-edge Fourier-filtered k1-weighted

EXAFS functions of the dried NiMoEN'SiCE catahst precursors prepared with different

amounts of ethvlencdiamine

Catalyst

precursor

Shell Ncomd, Ä|A] A<r

[1(E\42]

AEo [eV] Goodness R-range

of fit pA]

NiMo 0 4.4 1.99 -2.50 7.8

Si 4.5 3.35 -2.91 4.7

EN:Ni=O.I6 C) 6.8 2.00 3.19 5.3

Si 3.0 3.35 -3.85 -4.8

EN:Ni-0.66 N 0.2 1.82 -6.73 3.0

0 6.4 2.03 3.83 3.7

Si 1.5 3.37 -3.63 -7.2

EN :Nh 3.3 3 N 3.7 2.03 1.10 0.8

0 3.2 2.11 1.18 5.3

c 2.1 2.82 4.24 11.9

SI 0.8 2>.J / -3.46 _7,7

EN:Ni-6.66 N 5.5 "2.13 0.51 -1.5

c 8.0 2.92 4.17 3.1

3.50 0.60-3.60

0.82 .00-3.6

1.08 1.00-3.4

1.50 1.00-3.40

2.09 1.20-2.90

50

Structure ofPrecursors and Catalytic Activity

8

6

R|A1

FIG. 7. Fouiiet transform of the yyk) E £\\VS of»KENT 2H>0 simulated by Fcft7

In the case of the catalyst precursor with ENNi ~- 6.66. fitting with the Ni-0

reference yielded a coordination number of onh 4.1. With the Ni-N reference, in contrast, a

coordination number of 5.5 was obtained, closer to the value of 6 for the expected

octahedral geometry around nickel. Ehe degrees of freedom were not sufficient to perform

a fit using the Ni-N and the Ni-0 contributions simultaneous!}. fhe second shell was fitted

with a Ni-C contribution, but the resulting coordination number was 8, which is too large

even for a Ni(EN)3 complex. Moreover, the features of the fitting curve suggested that an

additional contribution overlaps with that of Ni-C. Several attempts (Ni-Ni. Ni-Si. Ni-Mo)

were made to identify the element responsible for the second contribution but with no

success. From the UV-VIS spectra, the average number of EN molecules around Ni in the

catalyst precursor with EN'Ni - 6.66 should be slightly smaller than 3. It is not possible to

calculate the exact number of EN molecules coordinated by Ni by means of EXAFS,

because the N and O atoms in the first shell can not be distinguished, even though it is clear

51

ChapIer 3

that nitrogen is the main contribution and the number of carbon atoms resulting from the fit

is overestimated.

As far as the second shell is concerned, comparing the spectrum of NiMo/Si02 with

the spectra of the catahst precursors with FN Ni - 0.16 and 0.66, there is a clear decrease

in the intensif} of the signal at 3.05 Ä (not phase corrected). This is assumed to be

generated by a Si shell of the support. Clause et al [29] reported on Ni silicates in their

Ni/Si02 compounds that were prepared by incipient wetness impregnation as well as by ion

exchange. In their EXAFS spectra they observed signals mainly at 2.8 A (not phase-

corrected) that they attributed to a combined contribution of Ni and Si. ITowever, they

measured all samples at room temperature. In contrast, in our spectrum, judging from the

distance from the central atom, the peak at 3.05 A is thought to arise exclusively from a Si

contribution. Since the measurements were carried out at liquid nitrogen temperature, the

contributions of the thermal vibrations are drasticalh reduced, enabling the detection of a

cleaner signal of the backscatterer atoms. The reference for the Si shell was calculated with

the Feff code using a Ni-Si distance of 3.3 A. A fit with a Ni-Mo coordination was

attempted, but the F-tcst showed that there is a 85% probability that the Ni-Si model

describes the system better than the Ni-Mo shell. In the catalyst precursor with

EN/NI = 0.16. the Ni-Si coordination number decreases from 4.5 to 3.0. The signal of the

Ni-Si shell becomes smaller, but also broader, with increasing amounts of EN, but it is still

present at EN/Ni - 3.33. until it disappears at the highest FN concentration.

52


NiXASofNiMoNTA/Si02. The ^-weighted Fourier transformed EXAFS

functions of NiMoN fA/Siü2 catalyst precursors with NEA/Ni = of 0.25, 0.66. 1.5. 3.0 and

6.66 (Fig. 8) show that N fA strongly affects the first and second shells around Ni. The k-

range of the presented data is 3-14 Â"1 for the catalyst precursor without a ligand as well as

for the precursors with NTA Ni = 0.25, 1.5. 3 0. 6.66. while It is 3-10 Â"1 for the catalyst

precursor with NEA/Ni = 0.66. Ehe second signal at 2.3 A (not phase corrected) arises from

the carbon atoms of NTA. The results of the fits (Table 5), with the exception of the

precursor catalyst with NTA/Ni = 1.5, clearly show that the Ni-C coordination number

increases gradually with increasing amount of NFA. For the catalyst precursor with

NTA/Ni - 0.25 the signal was too weak to be fitted. On the contrary, for NTA/Ni - 1.5. the

Ni-C coordination number obtained by the fit was too large (9.0), as in the case of the

catalyst precursor prepared with ethylenediamine with EN/Ni = 6.66 (sec above). For

. H

r-i

cd

3'

5 a. u.

NlANi=6 66

NTA Nf3 0

NTA Ni=1 5

NTA Ni=0 66

NTA Ni=0 25

No ligand

R[w]

FIG. 8. Absolute parts ot the foutier ttanstoims of the \'t x-edge y.(k) k' EXAFS of NtMo/SiO;

catalyst precursors as a function of the amount ot nitttlottiacetic acid used dining preparation de¬

range 3-14 A k except for X" fA \u - 0 66 k-range 3-10 A ')

53

Chapter 3

NTA/Ni = 0.66. the average Ni-C coordination number is only 0.9, while it increases to

almost 5 and 6 when the ratio is 3.33 and 6.66. respectively. In fact, if it is assumed that all

NTA present binds to Ni and that Ni is complexed by the three acetate arms of the chelating

agent, then the Ni-C coordination number should be 3.9 for NTA/Ni- 0.66 and 6 for

NTA/Ni =- 1. Two explanations for such a low experimentally observed Ni-C coordination

number are that the nickel ions exchanged some of the ligand bonds for interactions with

the support and are coordinated by only two ol' the three acetate arms and by the nitrogen

atom of the NTA molecules, or that a fraction of the nickel is present on the support in

another structure. When NTA/Ni - 6.66. the Ni-C coordination number is about 6,

suggesting that all Ni atoms are completely complexed by one N fA molecule.

The Ni-Si contribution at 3.05 Ä (not phase corrected) is still present for the catalyst

precursor with the lowest NTA concentration but is not detected at higher NTA

concentrations.

TABLE 5. Structural parameters resulting from the Ni A-edge Fourier-filtered k -weighted

EXAFS functions of the dried NiMoNTA/Si02 catalyst precursors prepared with different

amounts of Nl'A

Catalyst Shell Ncooni. R [Â] ArT AEo Goodness R-range

precursor IH)"3Â2] [eV] of fit [Ä]

3.50 0.60-3.60

1.69 1.00-3.60

4.33 1.00-3.00

4.1 1.00-2.70

1.62 1.00-2.70

2.75 "0.60-2.80—

54

NlMO 0 4.4 1.99 -2.50 7.8

Si 4.5 3.35 -2.91 4.7

NTA:Ni=0.250 5.9 2.02 0.53 4.0

Si 3.1 3.35 -4.19 -4.0

NTA:NE4),660 5.8 2.05 1.44 2.9

c 0.9 2.86 -13.82 9.5

NTA:Ni=-E500 6.0 2.04 1.45 1.8

c 9.0 2.83 2.03 -2.0

NTA:Nr=3.000 5.7 2.04 0.84 4.1

c 4.8 2.87 -0.78 3.6

NTA:NU6.66 0 5.1 2.05 0.70 2.8

C 5.9 2.84 1.45 8.0


Characterization ofMo

Laser Raman Spectroscopy. Raman spectra were recorded for dried catalyst

precursors. The spectra of Si02-supported Mo. NiMo. and NiMoEN (EN/NI - 0.16, 0.66)

are presented in Fig. 9. Catalyst precursors containing NTA and higher EN concentrations

could not be investigated b\ Raman, because the samples were destroyed by the laser beam.

Ehe spectra show common features arising from silica (bands at 485, 217 cm" ) and from

the NO/ ion (710 cm"1). The bands at 973, 955 and 616 cm" in the spectrum of the sample

containing onh Mo correspond in wave number to (Emolybdosilicic acid [3], a Eeggin type

structure consisting of 12 octahedral MoO(1 units surrounding a Si04 tetrahedron [30].

However, the high intensity of the band at 955 cm"1 and the presence of the band at 367 cm"

1and the shoulder at 238 cm"1 suggest that poh moh bdates are also present on Si02. The

mentioned Raman bands are consistent with octahedral!} coordinated polymolybdates

NiMo,

EN:Ni=0.66

NiMo

EN:Ni=0.16

NiMo onlv

Mo onh

1000 900 800 700 600 500 400 300 200

Wave number (cnf

FIG. 9. Raman spectra of SiO-supported catalyst precursois

Chapter 3

interacting weakly with the silica surface, similar in structure to the aqueous

isopolymolybdate anions M07O946" and MogCEf,4" [2, 31]. Several aqueous

isopolymolybdate anions composed of edge sharing MoOe octahedra display an intense

Mo-0 symmetric stretching vibration from 943 to 965 cnf' [32, 33]. Moreover, the weak

shoulders at 900 and 315 cm"1 evidence the presence of molybdates in this supported

material.

A band at 961 cm"1 is visible in the spectrum of the NiMo/Si02 sample, similar to the

955 cm"1 isopolymolybdate band in the Mo-only sample. Nevertheless, the signals at 905

1 ^

and 315 cm arc characteristic for MoOf" 133-36]. As a consequence, in the absence of a

ligand, Mo Is present on the support as a mixture of polymolybdate clusters and isolated

MoCEf~

molecules.

In the spectra of catalyst precursors prepared with EN a shoulder around 900 cm"1,

typical of MoOf2" was also observed. The two bands at 960 and 940 cm"1 are, nevertheless,

the most intense signals. It is difficult to sa} whether both bands arise from polymolybdate

species, or just one of them, fhe weak signal at 360 cm"1 is a further proof that such

clusters are present. Polyanions in aqueous solutions and supported on silica show a band

around 960 cm"1 [2, 31]. whereas for crystalline (NEE)6MoA02p4IFO and (NH0&Mo2O7 the

band shifted to 937 cnf' |37|. Ehis suggests that two kinds of Mo polyanions are present in

the samples prepared with EN: one which forms weak interactions with Si02, the other

which is like the crystalline compound and is not influenced by the support.

Mo XAS of NiMoEN/SiOi. Fig. 10 shows the Fourier transformed k'-weighted

EXAFS functions of NiMo'Si02 and NiMoEN SiCE catalyst precursors with EN/Ni - 0.16,

0.66, 3.33 and 6.66. Ehe results of the fits are presented in Table 6. We did not fit the Mo-0

shells, because the overlapping Mo-0 contributions at différent distances cause a reduction

in the signals, with consequent underestimation of the coordination numbers (38]. The

signal at 3 Â, due to the backscattering of the first Mo-Mo shell, was fitted with 0.9 Mo-Mo

contributions for the catahst precursor without a ligand. The Mo-Mo coordination number

increases to 1.3 for the catahst with FN Ni - 0.66. Ehe Mo-Mo contribution is more

intense and narrower for the catahst precursor prepared without a ligand than for the

56


3 a. el

^s

ai

u

tS:

\

ENNi=6 66

ENNF3 33

EN Ni=0 66

ENNi=0 16

R[A|

FTG. 10. Absolute parts of the I ouner transforms of the \lo A-edge y(k) C EXAFS of NiMo/SiCE catalyst

precursors as a function of the amount of ethylenediamine used dui mg preparation (k-range 3-17 A ')

catalyst precursors prepared with EN; the Mo-Mo coordination number, however, is

comparable. Ehis is a sign of a better defined distance distribution of the Mo-Mo shell. Ehe

signal splits into two parts in the spectrum of the catahst precursor with EN/Ni - 6.66,

suggesting that the structure of the second shell in this compound is more complex than in

the other catalyst precursors. A fit with two Mo-Mo contributions was carried out and gave

0.5 and 0.4 Mo neighbour atoms at a distance of 3.16 and 3.42 A. respectively, typical Mo-

Mo distances in Mo-CEf" clusters. A similar split was observed in the k3-weighted EXAFS

spectrum of (NFEEMo-^i simulated with the Eefi7 code. The reference for the Mo-Mo

shell was taken from the experimental Mo-Mo EXAFS contribution in MoS2.

57

Chapter 3

TABLE 6. Structural parameters resulting from the Mo K-edgt Fourier-filtered

k -weighted EZXAFS functions of the dried NiMoEN/Si02 catalyst precursors prepared with

different amounts of ethvlenediamine: nearest Mo shell

Catalyst Shell JNoohI, R[Al Act2 AE() Goodness R-range

precursor I10"3Â21 [eV] of fit [Â]

NiMo Mo 0.9 3.29 -0.81 6.0 2.39 2.70-3.30

ENfNfi-0.66 Mo 1.3 3.30 0.90 7.2 2.35 2.70-3.30

EN:Ni-3.33 Mo E0 J.JO 0.33 3.0 3.68 2.75-3.35

EN:Na6.66 Mo 0.5 3.16 -1.65 15.6 4.09 2.60-3.60

Mo 0.4 3.42 -4.20 -3.9

Mo XAS of NiMoNTA/SiCE. Ehe features of the Mo A-edge Fourier transformed

l</-weighted EXAFS functions of NiMoNEA'SiCA catalyst precursors, plotted in Fig. 11,

demonstrate how Alo is affected by the presence of the chelating ligand. To understand the

changes caused by the increasing NTA concentrations, it must be taken into account that

the formation of the [MoO^NTA)]'" complex is possible only when NTA is present at

NTA/Ni ratios higher than one. because Nl'A is much more strongly bonded by Ni21 than

by Mo61 [1|.

The Mo-Mo contribution at 3 Ä (not phase-corrected) is reduced drastically when the

NTA/Ni ratio is increased from 0.66 to 1.5. \s can be seen from the spectra of the catalyst

precursors with NEA/Ni - 0.25 and 0.66, Mo does not react to the presence of the chelating

agent, and some polymolybdate is still present on the support. In the three cases in which a

Mo-Mo contribution at 3 A is observed, the Mo-Mo coordination number is always around

1.1. In the spectra of the three catalyst precursors with NTA/Ni ratios higher than I. a new

weak signal around 2.8 A is observed. Its intensify was too low to carry out a significant fit,

however.

58

S/; ucture of Precursors and Catalytic Activity

5 a. u.

NTA Ni=6 66

NTA Nf3 00

NTANi=1 50

NTA NfO 66

NTA NfO 25

No ligand

R|A]

II. Absolute patts of the fouttet ttanstoims ot the \lo A-edge y(k) E EXAFS of NiMo'Si02

st piecuisots as a function otthe amount of mttilottiacettc acid used duttng piepaiation (k-iange 3-

)

5

Chapter 3

Discussion

The beneficial effect of chelating ligands during wet catalyst preparation on FIDS

catalytic activit} is not limited to mixed amino-acetic acid-containing complexes such as

NTA and EDTA. Amines, crown ethers and organic acids also improve activity. However,

ligands with the largest number of coordinating atoms (NTA. EDTA, PT, citric acid) have

the strongest effect. The ligand/Ni ratio was 1 in all cases. For NTA and EDTA this means

that Ni. but not Mo. is coordinated by the ligands [I ]. fhe Mo A-edge EXAFS showed that

this is also true for EN (Fig. 10). This is likely to be the case for the other amine ligands as

well, since Ni forms very stable complexes with these ligands (Table 1). These

observations indicate that better protection of the nickel ions in the catalyst precursors

enables the preparation of more active catalysts. EAr-\TS results showed that, in the

absence of ligands, the ammonia molecules in the first coordination shell of Ni arc

substituted by oxygen atoms. Ehis is due to the evaporation of NEE, during the drying

procedure, leaving behind Ni ions that are exclusive!} coordinated by oxygen atoms from

H20, OH", 0~" or SiO" [1 [. Ehis was confirmed by the Ni A-edge EXAFS data, which

clearly showed a Ni-Si interaction. In fact, the Ni-Si coordination number of 4.5 suggests

that Ni is cither adsorbed on the surface or built in the Si02 frame in a tetrahedral position,

since the Ni-0 coordination number is also close to 4,

For the dried EN-eontaining catalyst precursors. UV-VIS and EXAFS showed that

EN could be removed from the Ni"' ions as well, the EXAFS curve fittings indicated that

Nr'

has no carbon neighbour in the catalyst precursors when EN/Ni = 0.16 and 0.66. about

two carbon neighbours when EN Ni - 3.33 and probably six carbon neighbours in the dried

catalyst precursor with the highest EN amount. The reason for the decrease in EN

molecules coordinated to Ni" after the drying procedure is again the exchange of ligand

molecules for oxygen atoms of the support. Ronneviot et al. observed the same

phenomenon for Ni/Si02 samples prepared from [\T(N1 FEI"4" by i°n exchange [25].

Burattin el al used the deposition precipitation method to deposit Ni2^ on Si02 [4, 5J.

According to these authors, the Ni(II) hvdroaqua complexes close to the silica surface can

react with the silanol groups via an h} droh tic adsorption:

60


Si-OH+[Ni(El20)TOET)2] +± [Si-0-Ni(IEO)4(OH)]+ FI20 (4)

They suggest that a Ni phyllosilicate layer forms on the support with subsequent

stacking. We concluded from EXAFS spectroscopy that, in the absence of ligands, nickel is

present almost exclusively as nickel silicate. EN inhibits the formation of nickel silicates,

because the chelating diamine molecules prohibit hydrolytic adsorption on the support.

Nevertheless. UV-VIS spectroscopy demonstrates that a fraction of EN coordinated to Ni

is removed during drying and is substituted by oxygen atoms of the support. Ehe EN/Ni

ratio must be substantially higher than 6 to ensure that all the Ni is coordinated by three EN

molecules.

NTA behaved different]} to EN. Ni-Si interactions were observed only at the lowest

NEA concentration, whereas in catalyst precursors containing larger amounts ofN FA, Ni is

present as isolated ions on the support. Ehis can be explained by the fact that EN has only

two binding sites, whereas NFA has four and forms much more stable complexes with Ni

than does EN, protecting Ni better and hindering interactions with the support. For this

reason, the activity of catahsts prepared with N EV Is much more dependent on the ligand

concentration for NEA/Ni ratios below 1. FA en for this chelating ligand the amount needed

to keep all Ni fully coordinated by NTA is eonsiderabl} higher than the expected ligand/Ni

ratio of 1. The formation of [Ni(NTA)2|4" would also be possible with larger amounts of

NTA. Flowever, the complex w ith only one ligand molecule seems to be the species present

on the support, since in (Ni(NTA)2]4" Ni would have eight neighbouring carbon atoms,

whereas the NI-C coordination number obtained from the fit of the spectrum of the

precursor catahst with NfA'Ni = 6.66 was onh 5.9.

'Fhe addition of ligand molecules causes the complexation of the Ni ions, and the

decrease in the number of Ni support connections. We could show that, as far as silica is

concerned, the weaker the Ni-SiCE interactions in the catalyst precursor are, the higher the

catalytic activity of the eventual catahst is. In fact, the absence of links with the support

enables better mobilit} of Ni on Si02 so that, during sulfidation. it can easily move closer

to the MoS2 and form the Ni-Mo-S phase, fhe presence of the ligand. meanwhile, delays

the sulfidation ofNi and subsequent formation of NES? [ 12. 39 J.

61

Chapter 3

Raman spectroscopy indicates the presence of different Mo-species, polymolybdates

and isolated M0O4A Ewo kinds of polymolybdates seem to be present on the support, one

which is like the crystalline compound and is not influenced by the support, the other which

forms weak interactions with Si02. EXAfS revealed no evidence of a Mo-support

interaction, as was indicated by Raman spectroscopy. EN does not have a significant effect

on the structure of the Mo compounds. NTA. on the other hand, has a dramatic effect on the

structure of Mo when it is present at NTANi ratios higher than 1. Mo then forms the

complex [MoOANTA)jA Since the Ni/Mo molar ratio is 0.3/1, all Mo will be complexed

by NEA only when NTA/Ni = 4.3. We conclude from the Mo A-edge spectrum of the

catalyst precursors with NTA/Ni - 1.5 that the first effect of NEA is that polymolybdate

clusters are not formed, since a Mo-Mo contribution was not observed.

Ehe catalytic activity of catalysts prepared with NTA decreases for NTA/Ni ratios

higher than 2. At this NTA concentration, about one third of the Mo atoms should be

complexed with NTA. The further decrease in catahtic activity corresponds to the gradual

complexation of Mo, suggesting that NTA has a negative effect on the performance of the

catalysts when it interacts with Mo. In fact, the presence of NTA can delay the formation of

MoS2 crystallites, so that at the temperature at which the MoS2 crystallites are formed, Ni

has already reacted to NES2 and can not join the MoS2 edges and constitute the active sites

[391.

Conclusions

Catahtic tests in the lndrodesulfurization of thiophene at atmospheric pressure

showed that a wide variety of organic ligands have a beneficial effect on the catahtic

activity of NiMo catalysts supported on SiO?. This observation already demonstrates that

the improvement in catahtic activity is closeh connected with the effect that the used

organic molecules have on Ni. because all chosen catahsts form stable complexes with Ni

but not all of them interact with Mo. Ehis Iwpothesis is reinforced by the fact that the

activity of catalysts prepared with EN and NTA exhibited a significant improvement with

ligand/Ni ratios between 0 and 4 for FN and from 0 to 1.5 for N FA, whereas a decrease in

the acthity was observed for higher ligand concentrations. Amounts of chelating ligands

62


higher than EN/Ni - 4 and NTA/Ni = 1.5 do not to affect the structure of Ni any more, but

only that of Vfo as E/XAFS measurements showed.

The increase in activit} is explained by the elimination of Ni-support interactions

obtained b} the addition of the ligands. In the case of catalyst precursors prepared with EN.

EXAFS reveals a Ni-Si shell until EN/NI - 3.33. whereas NEA eliminates such interactions

already at NTA/Ni lower than 0.66. For these reasons the activity profile of catalysts

prepared with EN shows a slower increase in respect to the ligand concentration. Moreover,

UV-VIS spectroscopy showed that part of the FN is removed from the Ni environment

during the drying procedure.

Raman spectroscopy and Mo A-cdge EXAFS both showed that Mo is present on the

support as a mixture of MoOf" and polymolybdate clusters. Raman spectroscopy suggests,

furthermore, the presence of polymolybdate clusters interacting with the support. The

observation of two kinds of polyanions should not to be underestimated because it is

important for the interpretation of the results describing the sulfidation mechanism of the

catalysts that will be discussed in the next chapter. The presence of the chelating agents,

especially of EN. could influence and enhance the formation of one of the two

polymolybdates and therefore change the sulfidation behaviour of Mo. Fourier-filtered Mo

A-edge EXAFS spectra of the samples containing different amounts of EN and NfA

showed that EN has hardly an} effect on Mo. whereas for NTA concentrations higher than

I [MoOsINTA)]'" is formed, fhe formation of this complex has a negative effect on the

performance of the catalysts as the decrease in catahtic activity of catalysts with NTA/Ni

ratios higher than 2 showed. Ihe reason for this negative influence has to be searched in the

changes of the sulfidation behaviour of Mo and in the consequences on the structure of the

sulfided catalysts. Ihese factors will be considered, among others, in chapters 4 and 5.

Literature

1. Medici. E.. and Prins. R„ ./ Catal 163. 28 (1996).

2. Williams, C. C, Ekerdt, J. G„ Jehng. E. Hardcastle. F. D., Turck, A. M.. and

Wachs, f. R...7 Pins Chem 95. 8781 (1991),

63

Chapter 3

3. Banares, M. A.. JIu. H.. and Wachs, 1. F..,/ Catal. 155, 249 (1995).

4. Burattin. P.. Che. M.. and Louis, C. J. Phys Chem /J 101. 7060 (1997).

5. Burattin. P., Che, M., and Louis, C,./ Phys. Chem. B 102, 2722 (1998).

6. Louis, C, Cheng, Z. X.. and Che, M..,/ Phys. Chem. 97, 5703 (1993).

7. Thompson. M. S... European Patent Application, 1986.

8. van Veen, E A. R., Gerkema, E., van der Kraan, A. M., and Knocster, A., J. Chem.

Soc, Chem. Commun. 1987, 1684 (1987).

9. Eouwers, S. P. A., and Prins, R., J. Catal 133, 94 (1992).

10. Hiroshima. K.. Mochizuchi, T., Elonma. T., Shimizu, T.. and Yamada, M., Appl.

Surf Sei. 121/122. 433 (1997) and Chapter 4 of the present thesis.

11. Inamura, K., Uchikawa. K., Matsuda. S.. and Akai. Y.. Appl. Surf. Sei. 121/122, 468

(1997).

12. Medici, L., and Prins. R.../. Catal. 163. 38 (1996).

13. Candia. R., Sorenson. O.. Villadscn. E. Eopsoe. N. Y., Clausen, B. S., and Eopsoe,

FF, J. Phys Chem. 95. 123(1991).

14. Eijsboufs, S.. Appl. Catal. .1158, 53 (1997),

15. Levenspicl. O., "Chemical Reaction Engineering", .lohn Wiley and Sons. New York

(1972).

16. Dietz. W. A.. J. Gas Chrom 5, 61 (1967).

17. Pascarelli. S., Boscherini. F., D'Acapito. F., Hardy, E, Meneghini, C, and Mobilio,

S.,,Z Synchrotron Rad 3. 147(1996).

18. Kampers. F. W. IE. Maas. E. M. E. van Grondelle, E, Brinkgreve, P.. and

Koningsberger, D. C, Rev. Sei lustrum. 60. 2fc>35 (1989).

19. van Zon. J. B. A. D.. Koningsberger. D. C. van 't Blik. EI. F. E. and Sayers, D. E.. J.

Chem. Phys. 82.5742(1985).

20. Kirim, P. S., van Zon. F. B. M., Koningsberger, D. C, and Gates, B. C, J. Phys.

Chem. 94.8439(1990).

21. Lytic, F. W.. Sayers. D. E.. and Stern. E. A.. Physica B ÏS8. 701 (1989).

22. Zabinsky, S. !.. Rehr. J. E. Ankudinov, A., Albers, R. C. and Eller, M. E. Phys,

Rev. B 52, 2995(1995).

23. "Ginelin Handbook of Inorganic Chemistrv ". Springer Verlag (1989).

64


24. Larson. S. B., Simonsen, S. H., Ramsden. J. N., and Lagowski. J. E, Acta Cryst.

C45. 161 (1989).

25. Bonneviot, L., Legendre. 0., Kcrmarec, M., Olivier, D., and Che, M., ./. Colloid

Interface Sei. 134, 534 (1990).

26. Martelk A. E., and Motekaitis, R. E. "fhe Determination and Use of Stability

Constants", VCII, New York (1988).

27. Smith. R. M„ and Martelk A. E., "Critical Stability Constants", Plenum Press, New

York (1989).

28. Garcia-Granda. S.. and Gomez-Beltran. F.. Acta Cryst. C40, 1145 (1984).

29. Clause, O., Kcrmarec. M.. Bonneviot. L.. Villain. F., and Che, M., ,/. Am. Chem.

Soc. 114,4709(1992).

30. Greenwood. N. N.. and Earnshaw. A.. "Chemistry of the Elements", Pergamon

Press. New York (1986).

31. Williams, C. C, Ekerdt. J. CE. Jehng. E, Ilardcastle. F. D., and Wachs. I. E.. J. Phys.

Chem. 95.8791 (1991).

32. Griffith. W. P.. and Lesniak, P. J. B..,/. Chem Soc. A.1066 (1969).

33. Eytko. K. Ik. and Schonfeld, B. Z.. Z Xa/wforsch 30B. 471 (1975).

34. Gonzales-Vilchez. F.. and Griffith, W. P.. J Chem Soc. Dalton Trans. 1972. 1416

(1972).

35. Icziorowski. FE. and Knözinger, FF../ Phys Chem. 83, 6642 (1980).

36. Payen. E.. Grimblot. E. and Kasztelan. S.. ,7 Phvs. Chem. 91, 6642 (1987).

37. Ilardcastle. F. D.. and Wachs. I. E.. J. Raman Spec!/: 21, 683 (1990).

38. Kisfaludi. G.. Leyrer. E. Enö/inger. IE. and Prins. R., J. Catal. 130, 192 (1991).

39. Cattaneo. R.. Weber, Hi.. Shido, E„ and Prins. R., J. Catal. 191, 225 (2000),

65

Chapter 3

, //

66

Chapter 4

A Quick EXAFS Study of the Sulfidation of

NiMo/Si02 Hydrotreating Catalysts Prepared with

Chelating Ligands

Introduction

To optimize the reactivity of the CoMo and NiMo hydrotreating catalysts used in the

removal of S and N atoms from petroleum derivatives, a better understanding of the

processes involved in the production of such catahtic materials is needed. This chapter

reports a study of a fundamental step in the preparation of hydrotreating catalysts:

presulfiding. Presulfiding is carried out before catalysts are used, in hydrotreating

reactions, in order to convert the oxidic catalyst precursor in the final sulfided catalyst. The

information presented in the previous chapter about the catalyst precursors will be the basis

of the discussion about the mechanisms involved in sulfidation.

Several studies have been made of this pretreatment process using various techniques

such as temperature-programmed sulfidation [1], laser Raman spectroscopy [2]. extended

X-ray absorption fine structure (EXAFS) [3. 4]. transmission electron microscopy [5] and

surface science techniques [6J. Presulfiding consists in heating the catalyst precursor to

400°C in the presence of H2S. Because of these conditions and the dynamics of the system,

a reliable study must be carried out //? situ. In this work. Quick EXAFS was used to

characterize the various steps in the sulfidation ofMo and NT.

Chapter 4

Despite the fact that the first goal of this work was an investigation of the influence of

chelating ligands on the sulfidation temperature of Mo and Ni, during the analysis of the

data important information was obtained also on the sulfidation mechanism of Mo and

partly of Ni in absence of chelating agents. In fact, a complete understanding of the effects

of chelating ligands on the sulfidation of the employed metals is only possible after

clarifying the basic steps in the preparation of the unmodified catalysts.


Sample preparation and tests ofactivity

All the catalysts were prepared on an Si02 carrier. Ehe support material (C560,

Chemie Uetikon; particle size: 125-250 um. BEE surface: 565 m2/g, BEE pore volume:

0.83 ml/g) was dried at 120°C overnight prior to impregnation. Ehe catalyst precursors

were prepared by pore volume co-impregnation of the SiCE support with MoO-, (Fluka

puriss. p. a.) andNi(NO/h'6H20 (Fluka puriss. p. a.), as described in Chapter 3.

Hie oxidic precursors were sulfided for 2 h at 400°C (heating rate 6°C/min) with a

mixture of 10% ELS in IE (Messer Griesheim 3.0). Ihe activity of all the catalysts was

tested in fhe hydrodesulfurization of thiophene at 400°C. Ehe feed (3% thiophene in H2)

was obtained by bubbling IE through a series of four thiophene saturators that were cooled

to 2°C. Ehe product stream was analyzed on line with an HP5890 gas Chromatograph. The

sulfidation and thiophene FIDS reactions occurred at atmospheric pressure in the apparatus

presented inFig.l of Chapter 3.

XAFS measurements

The Quick EXAFS and classical EXAFS measurements were carried out at the XI

(RÖMO II) beam line at HASYLAB (Hamburg. German}), where the monochromator is

equipped with three parallel mounted pairs of Si(ll 1), Si(311), and Si(511) crystals. Ehe

energy of the beam line ranges from 6 to 70 keV [7]. Ihe X-ray energy is regulated by a

continuously moving two crystal monochromator [8|. fhe first crystal was detuned with

respect to the second one to eliminate higher orders of diffraction in the transmitted beam

68

Sulfidation of NiMo/Si02 catalystsfollowed hi QEXAFS

[9], so that the outeoming X-ray had 60% of the original intensity. The Si (311) and the Si

(111) crystals were used for the Mo and Ni A-edges. respectively. The accumulation time

was about 0.2 s/stcp at the Mo A-edge and about 0.4 s'step at the Ni A-edgc. The k-rangcs

used for the analysis of the data were 3-17 A"1 for the Mo and 3-12 A"1 for the Ni A-edge

(Fig. 1). The catahst samples were pressed into self-supporting wafers and mounted in an

in situ EXAFS cell [10|.

-6

3 5

Mo X-edge

13 15 17

M

45

25

05

-1 5

-3 5

Ni ÄT-edge

55

k[À-'l

FIG. 1. x(k) Ê data foi the Mo and \i x-edges measttied at station ROMO II. HASYLAB in Quick

EXAFS mode (spectia collected attct sulfidation at 400 C)

69

Chapter 4

The thickness of the samples was chosen to adjust the edge jump to 1 for the Mo

A'-edgc and the total absorption to px= 4 for the Ni A-edge (low Ni concentration). First,

two spectra of the fresh samples in an He atmosphere were collected. The samples were

then sulfided in situ during data collection. A stream of 10% H2S/H2 flowed through the

cell while it was being heated to 400°C. Fhe heating rate was 6°C/min for the

measurements carried out at the Mo K-edgc and 3°C/min for those at the Ni A-edge. This

difference was chosen in order to increase the accumulation time at the Ni A-edge and,

thus, to obtain reasonabh good data, since the Mo loading was 7 wt %, while the Ni

loading was only 1.3 wt %. fhe catalytic performance of the catalysts was checked using

both heating rates: no difference in activit} was found. A comparison of the data for the

two heating rates was carried out and onh an improvement of the data quaht} was

observed. Moreover, an isothermal experiment at 150°C showed that transport limitations

can be neglected for the order of time of our measurement.

For the classical EXAFS measurements, the samples were cooled to room

temperature after they had reached the desired intermediate sulfidation temperature. Once

at room temperature, the H2S still present in the cell was replaced by Fie by flushing for 10

min; the cell was cooled to liquid nitrogen temperature prior to the EXAFS measurement.

EXAFS analysis

The program XDAP (version 2.2.2) was used to analyze and fit the data [11]. The

pre-edge background was approximated b} a modified Victorccn curve, and the

background was subtracted using a cubic spline routine. I he spectra were normalized by

the edge jump. Ihe k'-weighted EXAFS functions were Fourier transformed and. in the

case of EXAFS spectra measured at liquid nitrogen temperature, fitted in R space. The

quality of the fit was estimated from the value of the goodness of the fit (sv~). Fhe zp value

takes the number of free parameters into account and is used to determine whether the

addition of new parameters would make sense. It was calculated with the formula:

2^ V-'WVS XmoJJ. ~ X.\ni

,,.

Cv > L-(1)

NPTS(v-A/„i() ^ [

OeV j

70

Sulfidaiion 0/N1M0/S1O2 catalysis followed In QEXAFS

where %modei and Xe\p are the model and experimental EXAFS functions respectively, gcxp is

the error in the experimental data (assumed to be 0,01 for each data point), v is the number

of independent data points in the fit range, and NPES is the actual number of data points in

the fit range. Although the absolute value of cx has no meaning, sv enables a comparison of

the goodness of the fit in the same spectra and at the same k weighting with different

parameters, a smaller value of e.x meaning a better fit.

Apart from AloS2 (references for Mo-S. Mo-Mo), reference spectra were calculated

using the Feft7 code [12, 131. Crystallographic data were obtained from the Inorganic

Crystal Structure Database (TCSD-CRYSTTN).

71

Chapter 4

Results

Molybdenum

The 1 IDS activities of catalysts prepared with NTA and EN are shown m Fig. 2. NTA

has a more pronounced effect than EN, and a maximum in catalytic activity is observed

between molar ratios NFA.Ni - 1.5 and 2. For catahsts prepared with EN, however, the

highest activity is obtained at EN:Ni = 4. EDI A has an effect similar to that of NEA at the

concentration studied.

0 12

CO

g 0 08 è

rt 0 06

ÖO

t 004

ta

0 02

EDTA »

EN

NTA

3 4 5

Liçand/Ni molar ratio

FÏG. 2. Thiophene HDS activity of NtVto Si CE catahsts ptepaied with ethylenediamine tettaacetic acid

and different amounts of ntttilotttacettc acid and ethv lenediamme (molai ratio Ni Mo = 0 3)

Ehe Mo A-edge XANES spectra, measured during the sulfidation of the catalyst

without a ligand. are presented in l ig. 3. Mo A-edge XANES has been subject of several

studies, according to which the first three components m the near-edge spectrum of oxidic

molybdenum should be assigned to 1 s —> 4d. Is-a 5s and Is -> 5p excitations |14]. Fhe

first transition is allowed in a tetrahedral field, but is forbidden in an octahedral field. This

selection rule breaks down when the octahedral s}mmetry is distorted. The shapes of the

72

Sulfidation of Vi\fo SiO catalysts followed In QEXAFS

near-edge absorption functions for all the samples can be divided, qualitatively, into three

types, all of which correlate with the extent of reduction. In the first type (Fig. 3, 41-

I72°C), the shape is ver} similar to that of the oxide catahsts. As sulfiding progresses, the

pre-edge feature fades and the edge moves towards lower energies (Fig. 3. 209-279°C).

Finally, after extensive treatment, the near edge absorption attains a shape which is

essentially identical with that of MoS2.

A more accurate description of the sulfidation mechanism of Mo is, however,

obtained from the Fourier transformed spectra of the QEXAFS data. Fig. 4 shows the

sulfidation process of Mo in the sample without a ligand (Ni:Mo - 0.3), as monitored by a

series of Quick EXAFS scans.

.2V-»On

Om

X3a

"O<u

"cd

£o

19980 20010 20040 20070

X-ray energy [eV]

FIG. 3. Mo Ä'-edge XANFS spectta measttied dining the sulfidation of NiMo SiCE (no ligand) plotted as a

junction of the sulfidation tempeiatuie

7-}

Chaptei 4

M

ST

[T. ^

0

et

CTQ

O

I—I

<!

o

C/5

CD

CTQI—i .

O

P

o"P

00

ELSa

W Pn1n> 1—i

CTQ O>i • 3O

Pr-f.

CD

CD

O

RE\1

FIG. 4. Tonner ttanstoims of the Mo x-edge k-weighted Quick tYAFS functions meastiied

dining the sulfidation of NiMo SiO- contamina no ligand 1 he lout sulfidation legions ate shown

(Heating täte 6 C min ktanse AI 7 \')

-74

Sulfidation of X/Mo/Si02 catalysis followed by QEXAFS

While the first spectrum is that of the untreated sample, the others were obtained at

increasing temperature upon treatment with FES. Ehe numbers next to the spectra denote

the average temperature during the scans.

The sulfidation process of molybdenum can be divided into four regions. The first

region is the temperature range in which molybdenum is present in the oxidic form. Fig. 4

shows that the distorted environment of Mo remained unchanged up to 110°C. While the

signals of the Mo-0 shells are between 0.5 and 1.9 A (not phase corrected), an Alo-Mo

contribution due to the presence of polymoly bdate anions can be seen at 3 Ä (not phase

corrected). In the second region, the longest Mo-0 shell as well as the Mo-Mo shell are

eliminated, while the shorter Mo-0 shell is still present; a signal around 2 Â (not phase

corrected), due to the presence of sulfur, is detected. This points to the co-existence of Mo-

O and Mo-S shells with the Mo-0 shell corresponding to Mo=Ot groups. The Mo-S signal

at 1.8 A (not phase corrected) has two maxima. A simulation with the Fefl7 program

showed that this signal can be caused by interference between an Mo-0 shell at 1.7 to 1.8 Ä

and a sulfur shell at 2.45 to 2.5 A. Eherefore. wc consider the peak at 1.8 Â not to be due to

a new Alo species. At the end of the second region a new signal at 2.5 Â (not phase

corrected) appears at a sulfidation temperature of 225°C. Ehis signal belongs to an

intermediate product which dominates the third region and contains Mo-S and Mo-Mo

contributions only. Ehe final product of the sulfidation process, MoS2, starts to form at a

temperature of 3 15°C and corresponds to the fourth region in Fig. 4. Ehe spectra measured

at temperatures higher than 315°C correspond to MoS2. m agreement with low-temperature

EXAFS measurements. Compared to a spectrum measured at liquid nitrogen temperature,

the Mo-Mo signal at 3 A (not phase corrected) has a lower intensif}. The shoulder on the

left side of the Mo-Mo shell in the catalyst sulfided at 400°C Is not due to the presence of

an additional signal but to interference of the Mo-S and the Mo-Mo shells, as shown by

spectra simulated w ith the Eef17 program.

The occurrence of four regions during the sulfidation process was common to all our

catalysts. Flowever. the temperatures at which the transitions from one region fo the other

take place were different, fhe onh exception was the sample with a molar ratio of

NTA:Ni - 6.66. Ehe sulfidation profile of Mo in this catalyst is shown in Fig. 5.

Molybdenum starts to sulfide at room temperature, i.e.. the first region is practically absent.

75

Chapter 4

f—i

400

400eu

400 CfQy—i. »

400 OP

400 i—i

400<

400 'à400

o

400oo

400 p

365^T!

350 £2.

340o

320

CP

C6

310 3CTQ AS

295o

Cù

275 p &h-H p

260 1 1

1—(

245o

225 o

210

195

180

160£0

145 CD

125 O

110P

95>—1

75

45

fresh Reeionl

R|A]

F1G. S. Fourier transforms of the Mo xAdge k-weighted Quick FXAFS functions measured

during the sulfidation of NiMo(N fA) SiCK molat îatto \ 1 \ Ni ~ 6 66 Mo starts to sulfide at room

temperature. (Heating rate 6A mm, k tange 3-17 \ ')

76

Sulfidation of NiMo/SiOî catalystsfollowed by QEXAFS

The Mo-S shell at 1.8 Â and an additional signal at 2.3 Â (not phase corrected)

already appears in the spectrum measured at 45°C. Ehe signal at 2.3 Â, which was observed

only during the sulfidation of this catalyst, is present until 210°C. At temperatures above

210°C. a peak appears at 2.5 Ä, which belongs to the intermediate product of the third

region (vide supra). The third region is present up to 375°C, and MoS2 crystallites are

formed only at about 400°C.

350

wJ3

300<DM

3

3 250S-<

1)

200

3

Ö 150or-<

"5s100

CO50

I third recion

second region

_>. o w ^ CO CD T_ (O CO ^r

i

CD

c 1 o II II CD II CDCN

n CD

of

oz

II

Z

<

z

<h-

z

<

CDII

z

Z

<

O

II

zz

z

zUJ

CDII

z

h- z Z < Q z UJ z

z H- UJ UJ UJ

FIG. 6. Temperature ranges for the second and third regions of sulfidation of Mo on SiCA-supported

catalysts of different composition (molar ratio Ni:]Vlo - 0.3 for catalysts with and without ligands).

Fig. 6 shows a comparison of the sulfidation temperature ranges of the second

(coexistence of AIo-0 and Mo-S shells) and third (presence of the signal at 2.5 Ä) regions

in the investigated catalyst samples, as obtained from the Mo A-edge Quick EXAFS

spectra. In the catalyst which contains Mo only, the temperature range covered by the

second and the third regions is relativeK large. Fhe situation changes completely when

nickel is added to molybdenum, fhe addition of Ni retards the reaction of Mo with FES. so

that it starts at 125°C but ends at a slightly lower temperature. NEA has a significant effect

on the sulfidation profile onh w hen the N TA:Ni molar ratio is higher than 1. In the extreme

case, i.e. a molar ratio of N fA:Ni = 6.66, sulfidation takes place over most of the

Chapter 4

temperature range. Ihe presence of EN induces the same sulfidation behaviour of Mo.

Although in the catalyst with the molar ratio 1 N:Ni - 0.66 no significant influence can be

seen, additional amounts of EN cause a marked broadening of the sulfidation interval of

molybdenum. In addition, it is evident that, m presence of EN, the first phase of the

sulfidation ends at much lower temperature, whereas the second phase is distributed over a

larger temperature range. For catalysts with molar ratios of NFA:Ni - 3. EDTA:Ni ~ 1. and

EN:Ni ~ 2.3. the sulfidation features of Mo are v ery similar.

To obtain more information about the structure of Mo in the second and third regions,

we measured classical EXAFS spectra of some of our catahsts at liquid nitrogen

temperature, after sulfidation at intermediate temperatures. Fig. 7 shows the k1-weighted

Fourier-transformed x(k) functions of the Ni\lo/Si02 catalysts without ligands after

sulfidation at 140. 170 and 290°C. It is difficult to compare the spectra shown in Fig. 7 and

the corresponding QEXAFS spectra measured at similar sulfidation temperatures, because

the latter were measured in a temperature range of about 15°C around the temperatures

given in Figs. 2 and 3. Nevertheless, several similarities can be observed, fhe spectrum

pp

A,

S

03

3

H

4 a. u

0 12 3 4

R[Ä]

FTG. 7. Absolute parts of the Mo Aedge k -weighted I XAFS functions of the NiMo'SiCA catalyst

containing no ligand, sulfided at thtee diffeient tempeiatutes (k tange AI7 Â"1. mcasuted at liquid

nitrogen tcmpctatuie)

78

Sulfidation of ht\!o'St02 catalysis folloM ed by QEXAFS

measured at 140°C (Fig. 7) clearly shows the co-existence of Mo-0 and Mo-S shells. While

a reasonable fitting of this spectrum was not possible because of the high number of

overlapping shells and the low intensities of the signals, the results of the fits of the other

two spectra are presented in Eable 1. In both cases, the Mo-S distance is 2.49 A, which is

greater than the corresponding MoS2 distance of 2.42 A. Fhe Mo-S coordination number is

relatively low (4.6) in the sample sulfided at 170°C, whereas it is extremely high (8.2) for

the catalyst sulfided at 290°C. In both cases, the goodness of the fit has a relatively low

value, thus indicating a good fit. Similar observations (i.e. high coordination numbers of

Alo in intermediate sulfidation regions) have been reported by other groups [5] and can be

explained by the ov erlapping of Mo-0 and Mo-S shells or b} a different Mo valence.

As far as the second shell is concerned, we did not consider the possibility that it

originates from Ni. because the same signal is present in the QEXAFS spectrum of a

catalyst containing only molybdenum. 'Flic same shell is better defined in the spectrum of

TABLE E Structural parameters resulting from the Mo A-edge Fourier-filtered

IE-weighted EXAFS functions of two Si Unsupported catalysts (Figs 7 and 8) sulfided at

di fièrent temperatures

Catalyst Suif. T Shell Ncooul R Ao2 [10"3 AEo Goodness

[°C] [A] A2! [eV| of fit

NiMoA).3:l 170 Mo-S 4.6 2.49 4.06 0.90 0.13

Mo-Mo 0.4 2.82 1.09 0.42

290 Mo-S 8.2 2.49 6.02 0.57 0.39

Mo-Mo 1.4 2.76 1.17 6.53

NiMoNTA- 245Mo-S 8.0 2.49 7.87 0.06 0.27

0.3:1:0.3Mo-Mo 1.4 i jß 1.97 8.82

270 Aio-S 5.4 2.40 6.37 5.32 0.35

Mo-Mo 0 6 2.79 0.30 7.89

79

Chapter 4

R[Ä]

FIG. 8. Absolute parts of the Mo x-edge ^'-weighted LXArS functions of the "NiMo(NTA)/Si02 catalyst

with the molar ratio NTA Ni _ 1. sulfided at three different temperatures (k range: 3-J7 Â"1, measured at

liquid nitrogen temperature)

the NTA:Ni -- 1 catahst sulfided at 270°C (Fig. 8). For this reason, we compared its

backscattering phase with that of the Mo-Mo reference extracted from the spectrum of

crystalline MoS2. Since the two phases overlap almost completely as a lunction of k, we

concluded that the signal at 2.5 A (not phase corrected) arises from a Mo-Mo shell. The

Mo-Mo distances, obtained from the fits of the different spectra (Table 1), range from 2.76

to 2.82 A.

Classical EXAFS spectra of intermediates after sulfiding the NEA:Ni - 1 catalyst

sample at 200, 245. and 270°C are presented in Fig. 8. The spectrum measured at 200°C

could not be fitted because of problems with the positions of the Mo-0 shells. From the two

other spectra, we obtained Mo-S coordination numbers of 8 (245°C) and 5.4 (270°C). while

the Mo-S distance in the latter case (2.40 A) is very close to that in MoS2.

80

Sulfidation of NiMo/SiO- catalysis followed In QEXAFS

Nickel

To stud} the sulfidation of the promoter, wc measured QEXAFS spectra at the Ni K-

edge. Since fhe loading of Ni was only about E3 wt %, the collection of data with a good

signal to noise ratio was not easy. For this reason, we used a lower heating rate of 3°C/min

rather than 6°C/min as for the Mo measurements. Fig. 9 presents two series of XANES

spectra measured during the sulfidation of the catalyst without a ligand and the catalyst

with EDEA NT -" I It is cas} to distinguish sulfide and oxide environments by measuring

the Ni A-edge XANES spectra, because the} show a strong white line when atoms of the

second row. such as nitrogen. ox}gen or fluorine, are present in the first coordination

sphere of Ni Due to the non-ionic character of the Ni-S bond, this feature is absent for

8300 8350 8400 8450 8"*00 8350 8400

X-ray energy [eV] X-ray energy [eV]

Fig. 9. Ni A-edge XANFS spectia ineasiiied dunna the sulfidation ol (A) NiMoAiCE (no ligand) and

(B) NiMoEDTA SiCA (FDFA Ni~T) and plotted as a function ot the sulfidation tempetatuie

81

Chapter 4

P

COr-t-

O

pc-f

CD

B

p

CD

O

R[À]

F1G. 10. Fourier transforms of the Ni A-edge k-weighted Quick EXAFS functions measured

during the sulfidation ofNiMo SiCV containing no ligand (heating rate SXA'min, k range: 3-12 A"1).

82

Sulfidation ofXiMo SiO: catalysts follow cd by QEXAFS

sulfided systems [15]. A comparison of the two series of spectra clearly shows that Ni is

sulfided at a higher temperature in the catal} st prepared with EDEA. Also in the case of Ni

we mainly used the Fourier transformed QEXAFS spectra to draw conclusions about the

changes in structure. The XANES features show, nevertheless, that Ni is completely

sulfided after treatment at 400°C.

* i 'X -r-

Fig. 10 shows a series of Ni A-edge Fourier transformed k -weighted QEXAFS

spectra of the sulfidation of the NiMo/Si02 catal} st. Ni-0 and Ni-Si shells are found at 1.65

and 2.9 A (not phase corrected), respectiveh. in the spectrum of the fresh catalyst. There is

no significant change in the spectra up to a sulfidation temperature of 90°C. The spectrum

measured at 105°C no longer shows an Ni-Si contribution. The transition of nickel in the

oxidic form to nickel in the sulfidic form is revealed by an increase of about 0.1 A in the

distance of the first shell. A first minor shift is observed between 50 and 125°C, but

replacement of ox}gen by sulfur took place between 125 and I40°C. As far as can be

monitored by QEXAFS. the structure of the first shell remains unchanged up to 400°C.

Fig. 11 shows our Ni A-edgc QEXAFS results of the NiMo EN:Ni r 4 catalyst during

sulfidation. The spectrum of the fresh catahst shows the features of the Ni(ENE complex,

while a Ni-Si shell is not present. Ehe occurrence of Ni-support interactions is prevented by

the presence of the ligand as described in chapter 3 and in réf. 116]. A shift of the first shell

occurred during the measurement of the first spectrum after the beginning of the

sulfidation. Ni is. therefore, surrounded by sulfur after sulfidation at room temperature, and

an apparent change in the structure of the first shell is not found at higher sulfidation

temperatures.

83

Chapter 4

P

cd

oo

pj—*

Cl

O

pr-t-

CD

3

O>-i

far-+-

c

o

o

O

R [Â]

FIG. 11. Fourier transforms ofthe Nh E-edge kAwtghted Quick EXAFS functions measured during

the sulfidation ofNiMo(EN)'Si02 with the molar ratio EN Ni~ 4. (Heating rate 3A7min. k range: 3-12 Â"1).

84

Sulfidation ofNih Jo Si02 catalysts folIon cd bv QEXAFS

Fig. 12 helps to clarify the effect of NEA on Ni during sulfidation. From the spectra

of Fig. 12 it is visible how the shift of the first shell takes place at a much higher value than

in the catalyst containing no ligand (Fig. 10). The data presented in this picture evidence, in

addition, that some interesting features about the second shell can be extracted from Ni K-

edge QEXAFS. In the spectra of the catahst in the oxidic state and sulfided until 210°C,

the presence of the Ni-C shell Is confirmed by the signal at 2.3 Â (not phase corrected).

p

d

r—1

ST

C/3

pI—J

tri

H-* »

O

P(-+

CD

BA3CD<-i

CD

O

n

R[A]

FIG. 12. Fourier transforms of the Nt Aedge k -weighted Quick EXAFS functions measured

during the sulfidation of NiVloNTA SiO-. with the molar tatio NT\Ni = 3 (Heating late 3°C/mm. k

range 3-12 ÀA

85

Chapter 4

Flowever. in the following 5 spectra a shoulder at 2.5 Â is observed. A more careful

observation of the previous two figures shows that this signal is also present in the spectra

of the catalyst prepared with EN (Fig. IE 205-240°C) and in the figure of the catalyst

prepared in the absence of ligands (Fig. 10. 140-175°C). even though it is less pronounced.

Since this signal is even observed when no chelating ligands are employed, we can exclude

that it arises exclusively from a Ni-C contribution. On the other hand, the complete absence

of aNi-C shell cannot be assumed when ligands are emplo}ed. Similar signals were already

reported for partiaux sulfided NiMo/SiOi [4J and CoMo/C |17] catalysts and were fitted

w ith a Ni-Ni shell, suggesting the presence of an intermediate stage in the sulfidation of Ni.

Ehis intermediate product precedes the final product of the sulfidation that can be

recognised from the weak signal at 2.8 Â (phase uncorrected) present in the three presented

Ni A-edge QEXAFS spectra. Ehis latter signal can be studied quantitatively only by means

of classical EXAFS. as will be done in the next chapter.

400

o 350 i

o1 1

C-( 300 '

3-h->

KÎT-H 250 !

aB8 200

do

150"3Td

kP 1003en

50

1

y

l.iI

Intermediate

product

Shift of first

shell

FIG. 13. Sulfidation temperature of \t in StCA-supported catalysts of different composition the

black regions represent the temperatures at which the shift of the first shell is observed in the QFXAFS

spectra The white rectangles correspond to the temperature interval where the intermediate product

(shoulder at 2 3 4) is obsetved (molar ratio \i Mo - 0 3 for catahsts with and without ligands)

86

Sulfidation ofNiMo/Si02 catalystsfollowed In QEXAFS

Figs. 10 to 12 show that the substitution of the Ni-0 shell by a Ni-S shell takes place

in a relatively small, well-defined temperature range. Ehis step is the first one in the

sulfidation of Ni. .A second step is revealed from the obtained data and corresponds to a

rearrangement of neighbouring nickel sulfide particles either interacting with the support or

forming small nickel sulfide clusters of unknown nature. The sulfidation temperatures of

our catalysts are presented in Fig. 13. In the figure two regions are plotted for each catalyst,

one corresponding to the shift of the first shell, the other to the observation of the shoulder

at 2.5 A denoting the intermediate sulfidation product. In the catalyst containing only Ni

supported on SKX sulfidation begins at room temperature and continues in a relatively

wide range of temperature. Ehe presence of an intermediate sulfidation product in the Ni-

only catalyst can not be clearly determined, fhe addition of Mo causes a delay in the

sulfidation of Ni. In the catalyst with a molar ratio of NTA:Ni = 0.25 (not shown in Fig.

13), NEA has no effect on the sulfidation of Ni, while for NTA:Ni ^ 1, a continuous

increase in the sulfidation temperature is observed. Ehe catalyst prepared with EDTA

represents the extreme case, where nickel is sulfided only at around 350°C. The presence of

EN seems to broaden the sulfidation interval of Ni. because the N atoms in the coordination

sphere around Ni are substituted at ver\ low temperatures, whereas the intermediate

product is observed at much higher temperatures.

87

Chapter 4

Discussion

In this section we will first consider the structure of the molybdenum species present

in the oxidic catalyst precursors and propose a sulfidation mechanism based on QEXAFS

and EXAFS measurements as well as on the underlying chemistry. Ehen, the effects of the

ligands on Mo and Ni as well as on the activit} will be discussed.

Molybdenum in the catalyst precursors

Analysis of NiMo/SiOi catalyst precursors by means of Raman spectroscopy and

EXAFS showed that Alo is present on SiO: as polyanions composed with seven or eight

Mo atoms (e. g. [Mo70:i]6" or [MogOoô]4") [16. 18]. The presence of such clusters is due to

the depolymerization of Mod at the pFl used in the preparation of the catalyst precursors

119). If the structure of these compounds had been well-ordered, with equal Mo-Mo

distances, then the theoretical Mo-Mo coordination number of the shell at 3.3 Â obtained

by EXAFS would have been about 3.4. Flowever. the Mo-Mo distances in these

polymolybdates are not well-ordered [16]. Elms, the EXAFS functions for the different Mo

atoms interfere so that the averaged signal is lower than expected (see Chapter 2 "Pros &

Cons of EXAFS""). For this reason, the Mo-Mo coordination number obtained from the fits

is only about 1 (sec Chapter 3, Eable 6), and the signal at 3 Ä (not phase corrected) in the

EXAFS and QEXAFS spectra has a relatively low intensity. A comparison of the EXAFS

spectra of [AI07O24]6" or [MosO^r,]4" (simulated by Feff7 on the basis of the Mo-Mo

distances as determined by means of crystallographic data), with the spectra of our catalyst

precursors confirmed this model. Depending on the chemical nature of the bond, the Mo-0

distances in the [M07O24]6" or [MoAAd]*"anions can v ary by several tenths of an Angstrom.

The shortest Mo-0 bonds are those of the terminating Mo=0 units (about 1.7 Â), while the

Mo-0 bonds in the bridging Mo-O-Alo functions are longer. Eheir length can vary from

1.75 to 2.6 Â [20. 21]. Such a bond length distribution can be noticed for the fresh catalyst

in the multitude of Mo-0 signals between 0.5 and 1.8 A (Fig. 4).

88

Sulfidation ofNiMo/St02 catalysts followed by QEXAFS

Sulfidation ofMolybdenum

QEXAFS spectra were measured from room temperature to 400°C. It is well known

that an increase in thermal vibrations causes a decrease in the EXAFS signal, which can

lead to a false estimation of the coordination number [22]. Nevertheless, we compared the

QEXAFS spectra with spectra measured at liquid nitrogen temperature for catalysts

sulfided at 400°C for 30 min. Ehe results of the fits show that the Dcbye-Waller factor was

about one order of magnitude larger for the QEXAFS spectra. The coordination number for

the Mo-S shell was about 10% smaller, whereas the Mo-Mo coordination number was

about 20% larger. The spectra of the third and fourth regions were also fitted. A

comparison of the results of the fits for the various catalysts showed that there are no

significant differences.

The analysis of QEXAFS data will, therefore, be based on a qualitative discussion of

the presence and disappearance of various signals but not on a quantitative analysis. For

quantitative analyses, classical EXAFS spectra will be used. We did not interpret the

changes in the relative amplitudes of the different Mo-0 signals, because an increase in

temperature and small changes in the bond distances can cause relatively large variations in

the amplitudes. This is seen in the first five spectra in Fig. 4. It cannot be determined from

the features of these spectra, whether the number of terminal or bridging oxygen atoms is

changing.

The Mo-S signal can clearly be distinguished in the spectrum collected at a

sulfidation temperature of 125°C. This signal has two maxima; these should not be

interpreted as two real signals but as only one. since Feff7 simulations showed that the first

one is the result of interference between Mo-0 and Mo-S shells.

Mo-O and Mo-S shells are observed simultaneously during the second region of the

sulfidation process. To interpret these observations correctly, the experimental set-up must

be considered. Samples were pressed into wafers, of 1 to 1.5 mm thickness, and then

mounted in our EXAFS cell. This thickness induces an inhomogeneous sulfidation of the

sample, as was observed for samples in which sulfidation was interrupted during the

temperature ramp. Fhe outer layers were already sulfided, as shown by the brown colour of

the surface, while the inner part was still in a previous sulfidation state as shown by its

89

Chapter 4

yellow colour. The inner part of the sample was. therefore, sulfided later than the surface of

the sample. This phenomenon is not important for samples that are treated for long periods

of time under the same conditions; it must, however, be considered when measuring

QEXAFS, because changes are observed in periods of few minutes. The question is

whether the delay between the sulfidation of the surface and the inner part of the sample is

significant compared to the time taken to register a QEXAFS spectrum.

Because of the time delay between sulfidation of the outer and inner parts of the

wafer, assignment of the four observed sulfidation regions to four transition periods must

be done with care before concluding that four different regions are present. It is clear that in

the beginning the sample is oxidic and that it is in the sulfidic state by the end of the

sulfidation process. There is no doubt that a different transition product forms in region

three, because the peak at 2.5 A (phase uncorrected) is not observed at the initial oxidic

stage or in the final sulfidic state. Does a first transition product start to form at the end of

the first period, which is transformed into the second transition product (with the 2.5 Ä

peak) during the third period? A simpler explanation would be that only three actual

regions exist during temperature-programmed sulfidation and that the spectra measured in

the second region consist of mixtures of the oxidic and the transition regions with the peak

at 2.5 A. In that case, however, the spectra obtained from sulfidation between room

temperature and 280°C should be linear combinations of the spectrum of the fresh sample

and of the spectrum of the transition product with the signal at 2.5 Â (Fig. 4, 280°C).

To determine whether the spectra from room temperature to 280°C are linear

combinations or not. several intermediate spectra were analyzed. Fig. 14 shows a

comparison of the spectrum measured around 170°C with the best fit of a linear

combination of the oxidic and transition regions. Ihe best fit was obtained with 73% of the

oxidic region and 27% of the third region. Ibis figure shows that the Mo-0 shells are well

fitted, while a rough fit only is obtained for the Mo-S signal. Moreover, the signals at 2.5

and 3 A, present in the calculated spectrum and which arise from Mo-Mo sheiks, are absent

in the measured spectrum. Hie peak at 3 A, due to pol} molybdate species, may be absent at

170°C because of increased thermal vibrations induced by the higher temperature than that

at which the oxidic sample was measured (room temperature). Flowever, the peak at 2.5 A

is visible in spectra measured at higher temperatures (280°C) and should, therefore, also be

00

Sulfidation ofMMo/Si02 catalystsfoilowed by QEXAFS

present at 170°C if some of the molybdenum had already reacted to the transition state. The

same differences in experimental and linear fit were observed for other spectra between

room temperature and 280°C. Therefore, the spectra measured in the second region in

R [Ä]

FIG. 14. Comparison of the Mo Aedge Fourier transforms of the QEXAFS function of the

catalyst containing no ligand sulfided at 170°C (solid line) and a fit calculated with 73% of the

spectrum in the oxidic state and 27% of the spectrum of the catalyst sulfided at 280°C (dotted line).

particular cannot be explained by a simple transition of the oxidic precursor to the transition

region (with the peak at 2.5 A); the existence of another transition product must be

assumed.

The transformation of the fresh catahst precursor to the first transition product might

occur as follows. The terminall}' bonded oxygen atoms are the most exposed structural

features of the polymolybdate species and. therefore, are accessible for a reaction with FES.

They are the first to be substituted by sulfur upon reaction with FES, according to

[-Mo=0] +1ES -> {-Mo-S} -t-11:0 (2)

91

Chapter 4

Details of this reaction step arc given in [23]. Ehe replacement of terminally bonded

oxygen (0~") by the larger sulfur (S ") has structural consequences. As mentioned above,

polymolybdate anions show a variety of bonding distances, fhis is due to the irregularity of

the octahedral [MoOß] building blocks and to the way in which they arc interconnected.

When oxygen is replaced b\ sulfur, structural distortions cause the long Mo-0 bonds to

break. This process, which converts bridging Mo-O-Mo functions into terminal Mo=0

units, is quite common and can even be predicted by a simple force field based molecular

modelling calculation. If bond breaking takes place at the edges of the anions, mononuclear

molybdate ions. M0O4"\ are expelled from the M07 particle, as can be seen in our QEXAFS

spectra. As soon as the Mo-S signal emerges, the Mo-Alo shell at 3 A, which is due to the

presence of polymolybdate species, decreases in intensity (Fig. 4, 110-155°C). Hardly any

change is noticed in the short Mo-0 shells in the QEXAFS spectra. Ehe expected loss in

intensity due to the replacement of Mo-_0 by Mo=S functions is compensated by the

generation of new terminally bonded oxygen atoms from bridging atoms and the formation

of molybdate anions, which contain only short Mo-0 bonds due to considerable d7t-p7t

bonding contributions. Moh bdenum coordinates 0 and S simultaneously, and consequently

a molybdenum oxysulfide is the intermediate product of the sulfidation. Above 195°C. the

intensity of the Alo-0 and Mo-S signals changes dramatically, i.e., the Mo-0 signal

decreases and the A1o-S signal increases, .fudging from the small side band at 2.5 Â. the

third region starts to form at about 245°C.

At 260°C, all the Mo is present on the support as the intermediate product of the third

region before it reacts further to MoS:. Our EXAFS measurements (Eable 1) show that the

intermediate product has Aio-\lo distances between 2.7 and 2.8 Â. This indicates that Mo is

present in a reduced state, i.e.. either in a A5 or +4 oxidation state. While the bonding

distance of Mo(IV)-Mo(IV) lies within this interval (2.73 A in the case of the [Molv3Si3]2"

cluster anion [24]), MofV)-Alo(V) bonds are usualh slight!}' longer than 2.8 Ä (2.82 Â in

the complex anion |MoV:Si:]"" [251). fhis shows how difficult it is to describe structural

properties of the intermediate region based on EXAFS data alone. However, based on

structural and inorganic chemistix. an approximate description of the intermediate structure

can be made.

92

Sulfidation ofNiMo/St02 catalysis followed by QEXAFS

The initial reduction of molybdenum proceeds via an internal redox process between

Mo(VI) and S"" as a consequence of the oxygen-sulfur exchange explained above. Two

adjacent {-Mo64=S2"} functions convert into a {-Mo5'~(S2)2TMoD,~} unit, which corresponds

to oxidation of sulfur (2S~" -> S2"

+ 2e") and reduction of molybdenum (2Mo(Vl) + 2e" ->

2Mo(V)). A detailed discussion of this reaction step can be found in Refs. [23] and [26].

This redox process has a profound influence on the breaking of the Mo-0 bonds and,

therefore, on the generation of new Mo-=0 entitles, lipon reduction, the Mo-Mo distances

become shorter, and the interconnection of Mo centres by oxygen links is weakened, what

facilitates the formation of new reactive Mo=0 sites. Temperature and the increasingly

destabilized oxysulfide structure enable all the oxygen to be replaced by sulfur, while Mo is

further reduced to the 4 f state. This mechanistic view is in line with our QEXAFS data,

which indicate that bridging oxygen atoms are no longer present in the intermediate product

at 280°C but are eliminated before the formation of M0S2. An analogous mechanism

applies in the sulfidation of the expelled molybdate species. We expect that the O-S

exchange reaction starts at higher temperatures than in the case of the aggregated species.

Ehis would explain why the short Mo-0 shell is still present in the QEXAFS spectra while

the Mo-S shell is already quite well developed. Ehe replacement of two oxygen ligands by

sulfur, followed by the redox process, leads to an intermediate similar to {(S:)Mo(fV)0).

The same chemistry and analogous types of intermediates are involved in the

decomposition reactions of the complexes (NIEVyMoOoS: [23] and (NEE)2MoS4 [27]. This

intermediate is coordinatively unsaturated and, therefore, unstable. Stabilization is achieved

by aggregation with other intermediate species or with the multinuclear oxysulfide. The

rapid formation of Mo(IV) rather than Mo(V) is also favoured by the presence of a second

reducing agent (Hi) and the structural stability is dependent on the coordination number.

While the metal-ligand coordination number of Mo is 6 in the initial oxide (Mo(VI) -

octahedral coordination) and in the final MoS: (Mo(IA) - trigonal prismatic coordination),

Mo(V) favours lower coordination numbers. For this reason, we conclude that the Mo-Mo

distances of 2.7 to 2.8 Â in our EXAFS spectra are mainly due to the presence of Mo(IV)-

Mo(lV) bonds.

93

Chapter 4

Similar EXAFS spectra have already been reported [4, 28, 29]. In some cases, high

Mo-S coordination numbers and high Mo:S stoichiometry. together with short Mo-Mo

distances, led to the conclusion that MoSi is an intermediate during sulfidation. Based on

the structural model for amorphous MoSi of Cramer et al. [28]. it was suggested that these

MoS^-like sulfidation intermediates are Mo(V)-containing materials, in which chain-like

Mo^+ centres are interconnected by S~" and Sy" species [30]. Flowever, it has been shown

that amorphous MoS;, is a compound of Mo(IV) and, in agreement with the chemical

behaviour of Mo(IV), the arrangement of the metal centres is not chain-like [27]. We do not

exclude the possible existence of an MoSv-like material in intermediate stages of the

sulfidation process. Ehis would be in line with our measurements and would also fit the

mechanistic description of the sulfidation reaction given above. Flowever, the existence of

stable, i.e., observable, Mo-S intermediates of Mo(V) is very unlikely.

The transformation of the third MoSMike product to the final M0S2 is a relatively

fast process and takes place in the narrow temperature range from 280 to 315°C. This is so,

because the transition to MoS: does not require further reduction of the metal centres. A

discussion of the decomposition of amorphous M0S1 to microcrystalline M0S2 can be

found in Ref. [27].

Chelating ligands and catalytic activity

A comparison of the catalyst containing only Ni and Mo and the other catalysts

shows that the ligands tend to broaden the sulfidation interval of Mo (Fig. 6). Towering the

temperature at which Mo begins to be sulfided could be explained by the fact that smaller

particles, such as the complexes formed between Alo and NfA or Mo, are attacked more

easily by FES than are the larger polymoh bdate units. In contrast, the delay in terminating

the sulfidation shows that the presence of ligands influences the nature of the intermediate

product that precedes MoS:. Otherwise. MoS: would start to form at the same temperature

in all samples, if temperature dependence of the reaction from the third region to MoS: is

assumed. 'Fhe effect of the ligands on the intermediate may be related to its dispersion on

the support as well as its aggregation state. A quantitative analysis can. however, not be

carried out on the basis of the QEXAFS data.

94

Sulfidation ofNiMo/Si02 catalysts followed bv QEXAFS

The sulfidation of Ni is dramatically affected by the presence of the chelating ligands.

By means of QEXAFS we could detect three regions during the sulfidation of Ni: the

oxidic state, an intermediate region characterized by the presence of a shoulder at 2.3 Ä

(not phase corrected) and the final region whose structure still has to be clarified and will

be discussed in Chapter 5. The passage to the intermediate region is usually preceded by the

substitution, in the first coordination sphere, of 0 or N ligands by S. In the case of the

catalyst containing no ligands and those prepared with NTA and FDTA. the O-S or N-S

exchange is more or less direct!} followed b} the formation of the intermediate product. In

the presence of EN. on the contrary, the two processes take place at extremely different

temperatures. All three ligands have in common the fact that, in conclusion, they all delay

the termination of the sulfidation of Ni in comparison with the catalyst prepared without

chelating agents.

A relationship with the activity data is. therefore, certainly connected with the effect

of the organic ligands on the sulfidation of Ni. Maximum activity ( sec Fig. 2) is observed

for catalysts with NTA:Ni - 1.5 to 2, FN:Ni = 4, and EDTA:Ni = 1. Catalysts with

NTA:Ni - 3. 6.6 and EN:Ni = 6.66 have lower activities, in spite of the fact that Ni is

sulfided at higher temperatures in these catalysts, fhis observation suggests that the

beneficial influence of the delay in the sulfidation of Ni is counterbalanced by an effect of

the ligands on Mo. The fact that the highest concentration of NTA and EN

(ligand:Ni = 6.66) causes a significant decrease in activity can be explained by the

excessive broadening of the sulfidation region for Mo (Fig. 6). In the case of NEA, this

effect is due to the fact that, at such a high NEA concentration, not only Ni but also Alo is

complexed by NEA with the formation of [MoO}(NTA)[A the sulfidation of which

apparently starts at a lower temperature but ends at a higher one than for polymolybdates.

In the presence of EDTA. Ni is sulfided at a higher temperature, while the sulfidation

behaviour of Mo is similar to the catah st with NEAfNi - 3.

An indirect effect of the ligands on the sulfidation of Alo may explain why EN also

has the same effect on the FIDS activit}. All the tested catalysts preferentially form

complexes with Ni, i.e., thev hinder direct contact between Ni and the support (8). In the

absence of the ligands. the Ni~" cation in the impregnating solution (pFI 8) is attracted by

the SiO" groups that cover the surface of the support (point of zero charge of SiO: is at

95

Chapter 4

about pH 2). The interaction of Ni with silica changes the surface properties of the support

and, consequently, the way in which Mo interacts with it. This furthers the absorption of

the molybdenum entities on SiO:, as concluded from the sulfidation behaviour of Alo in the

catalyst without ligand (Fig. 6). The thiophene FIDS results (Fig. 2) reveal that the activity

of the catalysts decreases at NTA:Ni ratios higher than 2 and at EN:Ni ratios higher than 4.

This negative influence may be due to the fact that, at the mentioned concentrations, the

ligands start to form complexes with molybdenum, because only one NTA molecule and

three EN molecules can complex Ni. No mention of an Mo-EN complex was. however,

found in the literature. Therefore, the negative effect of high EN concentrations may be the

result of the absorption of excessive amounts of protonated EN on the support and a change

in the surface properties. As a consequence, the more positively charged surface could

more easily absorb Alo species of the same kind that was detected by means of Raman

spectroscopy, as reported in Chapter 3. Such pohanions interacting with the support have

likely a different sulfidation behaviour in comparison with crystalline heptamolybdate

anions that have no connection w ith the support.

EDTA is present in a relativ eh low concentration (EDTA:Ni - 1). At this

concentration Mo does not perceive the presence of the ligand, and is present as polyanion

on the support.

Conclusions

Ehe collected QEXAFS data led to an interpretation of the sulfidation processes of

Mo and Ni on a support. It was shown that the sulfidation of Mo takes place in four

regions. Two intermediate regions were observed during the sulfidation of molybdenum,

the first consisting of mohbdenum oxysulfides and the second of apure Mo(IV)-S product,

which might be similar to amorphous MoSs.

QEXAFS measurements demonstrated that also the sulfidation of Ni is a complex

reaction involving at least one intermediate product, fhe shift of the first shell in the Ni K-

edge Fourier transformed data represents the substitution of the O or N ligands with sulfur

in the first coordination sphere of Ni. Subsequently. Ni forms an intermediate product

whose nature was not clarified }et. Ihe effect of the chelating ligands on the sulfidation

96

Sulfidation ofNiMo/Si02 eatah sts followed by QEXAFS

temperature ofNi is very pronounced. All chelating agents induce a delay in the sulfidation

of Ni. Interaction between Ni and Mo is supposed to take place after the intermediate

region of the sulfidation of Ni. Ehe increase in the IIDS activity, caused by the presence of

the ligands. is therefore ascribed to the sulfidation of Ni at higher temperatures. It was

suggested that there must also be an optimum sulfidation region for Mo, which judging

from the results, is relatively narrow and should be contained in the temperature range from

80 to 300°C, and depends on the Mo species present and on the molybdenum-SiO:

interactions. An excessive broadening of this interval could cause a decrease in catalytic

activity.

Ehe use of Quick EXAFS showed the advantage of following the sulfidation process

in situ; a comparison with classical EXAFS spectra measured at liquid nitrogen temperature

showed that signals observed in QEXAFS spectra are reliable and enable qualitative

analysis in spite of the higher temperatures. Flowever. with such a time resolution, the

diffusion limitations must be considered, what can be made negligible decreasing the

heating rate of the sulfidation and increasing the collecting time for QEXAFS spectra.

Another factor that must be considered when interpreting EXAFS spectra is the overlap of

the different shells.

Literature

1. Scheffer. B.. Arnoldy. P., and Moulijn. E A., ,7 Catal 112, 516 (1988).

2. Payen. E., Kasztelan. S., IToussenba}. S.. Sz}manski. R.. and Grimblot. E. J Phys.

Chem 93,6501 (1989).

3. Topsoe, EF, Clausen. B. S., and Massoth. F. E.. "Catalysis Science and

Technology". Springer Verlag. New York. 69 If (1996).

4. Medici, L., and Prins. R.. J Catal 163. 38 (1996).

5. De Boer, AE. van Dillen. A. E. Koningsberger. D. C. and Geus. J. W., Jpn J. Appl.

Phys 32-2,460(1992).

6. Jong. A. M. D., Muijsers. J. C. Weber. Eh.. van IJzendoorn, L. E, de Beer, V. EI. E.

van Veen. E A. R.. and Niemantsverdriet, J. W., in "Transition Metal Sulphides.

97

Chapter 4

Chemistry and Catalysis", Weber, Th.. Prins, II. and van Santen, R.A. Eds.. Kluwer,

Dordrecht. 207 (1998).

7. Tröger. L.. Synchr. Rad. News 6, 11 (1997).

8. Frahm, R., Rev. Sei. lustrum. 60, 2515 ( 1989).

9. Krolzig, A., Matcrlik, G., Swars, AE» and Zegenhagen. E, Nucl. Instr. and Meth.

219,430(1984).

10. Kampers. F. W. EE, Alaas, T. AT E, van Grondelle, E, Brinkgreve, P., and

Koningsberger. D. C. Rex. Sei. Instrum. 60. 2635 (1989).

11. Vaarkamp, M., Firing. E, Oldman. R. E. Stern, E. A., and Koningsberger, D. C,

Phys. Rev. B 5i). 7872(1994).

12. Ankudinov. A, L.. and Rehr, J. E, Phys. Rev. JB56, Rl712 (1997).

13. Zabinsky, S. !.. Rehr, E E. Ankudinov. A., Albers. R. C, and Ellcr. AT E. J. Phys.

Rev. 5 52,2995(1995).

14. Chiu, N.-S.. Bauer. S. IE. and Johnson. Al. F. E.., J. Catal. 89, 226 (1984).

15. Eouwers, S. P. A.. Crajé. Al. W. E. van der Kraan. A. M., Geantet, C, and Prins, R.,

J. Catal. 144, 579 (1993).

16. Cattanco. R., Shido. T.. and Prins. R,. J. Catal. 185. 199 ( 1999).

17. Crajé, M. W. E, Louwers, S. P. A., de F3eer, V. H. E, Prins, R., and van der Kraan,

A. M., A Phys. Chem. 96, 5445 (1992).

18. Williams. C. C, Ekerdt. E G., Jehng. E. Hardcastie, F. D., Turek, A. M., and

Wachs, I. E.. J. Phys. Chem. 95. 8781 (1991).

19. Mcdici. E., and Prins. R., J. Catal. 163. 28 (1996).

20. Weakley. E. E, Polyhedron E 17 ( 1982).

21. Evans, 11. T. jr., Gatehouse, B. M., and Leverett, P., J. Chem. Soc. Dalton Trans.

1975.505(1975).

22. Rockcnberger. E, Tröger. E., Kornowski, A., Vossmcyer, T., Eychmüller, A.,

Feldhaus. E. and Weller. IE. J. Phys Chem B 10E 2691 (1997).

23. Weber, Eh., Aluijsers. .1. C, van Wolput. E Fl. Al. C, Verhagen. C. P. E, and

Niemantsverdriet. J. W.,,7. Phys. Chem. 100. 14144 (1996).

24. Müller. A., Diemann. E.. Krickemeyer, E.. Walberg, IT.-.k. Bögge, H., and

Armatage, A.. Eur J. Solid State Inorg. Chem 30. 565 (1993).

98

Sulfidation ofNiMo/SiO: catalysts followed In QEXAFS

25. Müller, A., Nolle. W.-O., and Krebs, B.. Angev. Chem 90, 286 (1978).

26. Muijsers, J. C, Weber, Th., van Ilardeveld, R. AT, Zandbergen, H. W., and

Niemantsvcrdriet. J. W..,/ Catal. 157, 698 (1995).

27. Weber. Th.. Muijsers. J. C, and Niemantsv erdriet, J. W., J. Phys. Chem. 99, 9144

(1995).

28. Cramer. S. P., Tiang, K. S., Jacobson, A. E. Chang. C. H., and Chianelli, R. R..

Inorg Chem 23. 1215(1984).

29. Eeliveld. R. G.. van Dillen, A. E, Gcus. J. W.. and Koningsberger, D. C. J. Catal.

171, 115(1997).

30. De Boer, M.. van Dillen, A. E, Koningsberger. D. C, and Geus. J. W.. J. Phys.

Chem 98.7862(1994).

99

Chaplei 4

100

Chapter 5

Influence of chelating ligands on sulfided NiMo

catalysts: an EXAFS study

Introduction

Ehe sulfidic state of supported NiMo (CoAIo) catalysts has been investigated by

means of a large number of techniques, among which the most powerful ones are

Mössbaucr emission spectroscop}' (AIES) [1-3]. X-ray photoelectron spectroscopy (XPS)

[4-61, electron microscopy [7-91. and extended X-ray absorption fine structure (EXAFS)

[10-14]. In this chapter we employ the latter technique in order to study sulfided

NiMo/SiO: catalysts. EXAFS spectroscopy has the great advantages of being a bulk,

element specific technique that can be applied to dispersed and amorphous materials.

Ehercfore, this technique is very suitable for the study of this bimetallic system, where the

loading of the studied elements ranges from 1 to 10 vvt%.

Ehe parameters obtained from the anah sis of EXAFS data allow us to understand the

composition of the coordination spheres around the absorber atom. Nevertheless, the

obtainable information is dependent on the quality of the collected data. Ehe quality of

EXAFS data is influenced b} various factors, the most important ones being the

concentration aucl the edge energy of the studied material. Ehe two elements studied in our

system have very different characteristics, while the loading of Alo is 7.1 wt%, that of Ni is

1.3 wt%. The energy of the Alo A-edge is 20000 eV and that of the Ni A-edge is 8333 eV.

Chapter 5

These parameters induce a significant difference in the data quality for the two metal

sulfides, the quality of the Alo A-edge data being substantially better.

In this chapter the findings are reported of an EXAFS study at the Mo and Ni A-edge

of SKE-supported NiMo catalysts prepared with nitrilotriacetic acid (NTA) and

ethylenediamine tetraacetic acid (EDEA). Ihe structure of these catalysts is compared with

that of catalysts prepared in the absence of ligands. Ehe conclusions drawn in the previous

two chapters are taken into account throughout this chapter in order to understand the

causes of the differences observed in the structure of the sulfided catalysts.


Catalysts preparation and activity tests

Ehe catalysts precursors were prepared b\ incipient wetness impregnation of SiO:

with an aqueous solution containing Ni(NO;):. MoO^ and the chelating ligands as described

in Chapter 3. Ehe oxidic precursors were sulfided for 2 h at 400°C (heating rate 6°C/min)

with a mixture of 10% HiS in IE (Messer Griesheim 3.0). The activity of all catalysts was

tested in the hydrodesulfuri/ation of thiophene at 400°C. The feed (3% thiophene in IE)

was obtained by bubbling IE through a series of four thiophene saturators that were cooled

to 2°C. Ehe product stream was analyzed on line with an 1IP5890 gas Chromatograph. The

sulfidation and thiophene HDS reactions occurred at atmospheric pressure in the apparatus

presented in Fig. 1 of Chapter 3.

EXAFS measurements

The EXAFS spectra reported were collected at the Swiss Norwegian Beam Fine

(SNBE) at the European Synchrotron Radiation Eacilitx (ESRF), Grenoble France. Ehe set

up of the beam line is described in Chapter 3 and ref. [15]. The catalyst precursors were

pressed in self-supporting wafers, mounted in a sealed EXAFS cell [16[ and sulfided in the

chemistry laborator} of the ESRF. The thickness of the samples was chosen to adjust the

total absorption to px - 4 for the Ni A-edge (low Ni concentration) and the edge jump to 1

for the Mo A-edge.

102

Sulfided NiMo/Si02 catalysts

An H2S/H2 (10/90) mixture was flown through the cell at a rate of 60 ml/min. The

unreacted hydrogen sulfide was abated by leading the gas exiting from the cell through two

washing bottles containing a basic FeCE solution. Ehe cell was heated to 400°C with a

heating rate of 6°C and was kept at this temperature for 30 min. After decreasing the

temperature to 40°C, the cell was flushed for 10 min and filled with He to an overpressure

of about 0.5 bar. Ehe cell containing the treated sample was then transported to the beam

line and placed in the hutch. Ehe sample was cooled with liquid nitrogen by means of a

dewar, whose structure allowed the cooling liquid to come in direct contact with the cell.

As far as the Alo A-edgc was concerned, the collection times for each data point in a

scan were 1, 2. 2, 3 and 3 s for the intervals 19454-19954 cV (pre-edge). 19954-20164 eV

(edge region) and the three 6.5-10. 10-15. 15-21 A"1 post-edge regions between 20164 and

21684 cV, respectively. Ehe Ni A-edge spectra were divided into six regions: 7900-8300

eV (pre-edge). 8300-8370 e\T (edge region) and the four post-edge regions (3-6.5, 6.5-10.

10-15, 15-17 A"1) between 8370 and 9433 eV. fhe collection times for the data points in

each scan region were 1, 1.2. 3. 4 and 4 s, respectively. Ehe distance between the post-edge

data points was determined so that the difference in their k values was smaller than

0.05 Ä"1. Five scans were averaged for the Ni A-edge and three for the Mo A-edge.

EXAFS analysis

Ehe program XDAP (version 2.2.2) was used to analyze and fit the data 117]. The pre¬

edge background was approximated by a modified Victoreen curve and the background was

subtracted using a cubic spline routine, "fhe spectra were normalized by the edge jump. The

k'-weighted and k1 -weighted EXAFS functions were Fourier transformed and fitted in R-

space. The free parameters were interatomic distance, coordination number, Dcbye-Waller

factor and the correction of the edge energy. The errors of the parameters were statistically

estimated using the random errors of the observed data. The goodness of fit was calculated

for every model from the k- and R-space fit range and the number of free parameters as

described in Chapter 4. Reference spectra lor the Alo Ä'-edge were obtained from the

measured spectrum of AloS: whereas for the Ni A-edge they were calculated using Feff7

[18].

103

Chapter 5

Results

Ehe catalysis studied in this chapter were chosen in order to explain which features of

these materials in the sulfided state influence their catalytic performance. A comparison

was made between catahsts showing the lowest (without ligands. NTA:Ni ~ 6.66, see

Fig. 1) and the highest (NEA. FDTA:Ni= 1-2) IIDS rates. In Fig. 2 the quality of the data

at the Mo and Ni A-edges for the catalyst prepared without ligands can be compared. The

0 14

0 12 y^ ^^\

^Aa *i-

"o 0 08 4B

£ 0 06

p

« 0 04

To

0 02 j

0

0 12 3 4 5 6 7

Ligand/Ni molar ratio

FIG. I. IIDS activity profile of SKA-suppotted catahsts prepated with different amounts oEN FA (\) and

FDTA () (molat tatto Ni Mo-0 3 1)

Mo A-edge data show an extremeh good signal to noise ratio, whereas the quality of the Ni

A-edge data is less good. Elus difference is due partly to the higher loading of Mo (7.1

wt%) in comparison to that of Ni (1 3 wt%) and partly to the lower edge energy of the Ni

A-edge. since at this energy the X-ra} absorption, and thus the corresponding noise,

produced by the other elements present in the sample is still relatively high. The higher

number of averaged scans for measurements at the NI A-edge (5 versus 3) allowed only a

limited improvement of the data quality, fhe k-ranges used for the Fourier transformations

were 3-19 and 3-15 Â"1 for the Alo and the Ni A-edge data, respectiv eh.

104

Sulfided NiA to/Si02 catalysts

¥

-16

9 12

kl A"1]

15

11 13 153 5 7 9

k|A-'|

FIG. 2. k'-weighted %(k) data of the catahst prepaied without ligands measiited at the Mo (above) and Ni

(below) Aedges

105

Chapter 5

Mo K-edge

Fig. 3 shows the Mo A-edge k3-weighted Fourier transformed EXAFS spectrum of

the catalyst prepared without ligands. Ehe first two signals, at 2 and 2.8 Â (phase

uncorrected) correspond to the nearest Alo-S and Mo-Mo shells, respectively. Fhe weak

signal at about 3.5 A (phase uncorrected) is produced by the second nearest Mo-S shell,

18 -,-,

-18 '

R[A]

FTG. 3. Fourier tiansformatton of the Mo À'-edge E-weighted 1 XAFS function of fhe catalysts prepared

without ligands

whereas the two signals around 5 and 6 A (phase uncorrected) are the result of a

superposition of the multiple scattering signal arising from the nearest Mo~Alo shell and of

further Alo-S and Alo-A1o shells [9J. During the fitting procedures, significant differences

were observed onh' in the parameters concerning the nearest Mo-Mo shell. Fig. 4 shows a

detailed view of this shell for the inv estigated catal} sis. From this figure one can notice that

the amplitudes of the various signals are different. 1 low ever, it is not possible to understand

106

Sulfided N1M0/S1O2 catalysts

44

H

16

14

12

10

8

6

4

2 3

— no ligands

NTA Ni=1 5

- NTA Ni=6 66

-EDTANI=1

EDTA'Ni=E5

EDTA Ni=2

^"

2 5

R[Â]

2 9 3 1

FTG. 4, Closer view of the Mo-Mo shell in the Mo Aedge Fourier transformed EXAFS functions of the

various sulfided catalvsts

TABEE 1. Parameters obtained from the fit of the Alo ÄAedge k'-weighted EXAFS

spectra of the sulfided NiAlo/Si02 catalysts (AR: 1.00-3.25 A. Ak: 3-19 Ä"1)

Catalyst Shell CN R Aa2 AE° Goodness

[Â] [1<E4A2] [eV] of fit

No ligands Mo-S 5.7 2.42 5.0 2.9 0.75

Mo-Alo 3.6 3.15 6.7 3.8

NTA:Ni-1.5 Alo-S 5.8 2.42 7.2 0.66

Mo-Mo 3.4 3.15 5.7 4.5

NTA:Ni=6.66 Alo-S 5.8 2.42 8.1 3.9 0.72

Mo-AIo j.J 3.15 3.3 5.2

EDTA:Ni=l Alo-S 5.9 2.42 5.4 3.3 0.71

Mo-Alo 3.7 3.15 5.5 4.3

EDTA:NaE5 Alo-S 5.8 2.42 4.8 3.8 0.76

Mo-Mo 3.6 3.15 5.2 4.8

EDTA:Ni=2 Mo-S 5.9 241 6.3 4.2 0.76

Mo-AIo 3.4 3.15 5.1 4.8

107

Chapter 5

the reason for these differences without fitting the data because the intensify of the signals

is a lunction of the coordination number and of the Debye Waller factor. Since the two

parameters are correlated, a fitting procedure is needed to discriminate the contributions of

the two parameters. Table 1 shows the results of the fits for the first two shells.

As far as the first shell is concerned, no significant change can be noticed in the

structural parameters between catalysts prepared with and without chelating ligands. On the

contrary, the parameters of the second shell show some interesting features. Increasing

amounts of NT/A and EDEA tend to lower the coordination number and the Debye Waller

factor of the Mo-AIo shell. Ehe effect of NEA is already observable for a molar ratio

NTAfNi - 1.5. whereas for EDTA a molar ratio of EDTA:Ni -~ 2 is needed to reduce the

Mo~Alo coordination number. From the Alo-Mo coordination number it is possible to

deduce the size of the MoSi crystallites [19], Mo-Mo coordination numbers of 3.2 to 3.7

suggest the presence of particles with a diameter varying approximately from 20 to 30 A,

what corresponds to MoSi slabs composed of40 and 90 Mo atoms, respectively.

Ni K-edge

The Ni A-edge Fourier transformed EXAE S spectra of the catalysts prepared without

ligands, and with different amounts of NTA and EDEA sulfided at 400°C are shown in

Figs. 5 and 6. The first shell was fitted with a Ni-S contribution, whereas the low intensity

and the proximity of the signals at 2.3 and 2.7 A (phase uncorrected) required a careful

checking of the results of the fits. Alan} fitting attempts were needed to study these two

signals.

In previous studies these two signals were fitted with various combinations of shells

[10, 12]. The larger k-range of the data obtained in our experiments made the results

presented here more reliable, however. Ehercfore, we fitted the two shells combining in

pairs the backseattercrs S. Ni and Alo and trying all possible sequences of shells. We

excluded contributions from the support because similar signals were obtained by Louwers

and Prins with carbon-supported NiAio catalysts, for which it can be assumed that the

carbon atoms from the support do not interact with Ni 110].

108

Sulfided XiMo/SiO2 catalysis

R[A|

FIG. 5. Fouitet ttansfotmation of the Xt A-edge L -weishted LXAFS function ol the catalysts piepated

with diffeient amounts of NTA

R[A]

FIG. 6. } outlet üanstotmation ot the Ni À. edge k -weighted LXAFS function of the catalysts piepared

with diftetent amounts ol FDl A

109

( 'haptcr 5

The main criterion adopted to accept a combination of shells was that the fit would

converge for the k1- and ^-weighted spectra. Moreover, the value of the edge energy

correction (AE°) should be smaller than 15 eV. Ewo models satisfied these two restrictions.

Ehe first model consisted of three Ni-S shells, fhe parameters resulting from the fits of the

spectra with these shells are presented in Eable 2.

TABLE 2. Parameters obtained from the fit of the k3-weighted Ni K-edge EXAFS spectra

of the NiMo/SiO: sulfided catal} sts with three Ni-S shells (AR: 0.90-3.15 A, Ak: 3-15 A"1)

Catalyst Shell CN R A<rz AE° Goodness

[Al [103Â2| |eV] of fit

No ligands Ni-S 3.6 2.19 0.55 2.0 0.30

Ni-S 0.7 2.75 -3.75 -3.2

Ni-S 0.6 3.11 -4.08 -7.8

NTA:NE-E5 Ni-S 3.6 2.19 0.39 2.4 0.31

Ni-S 0.7 2.75 -2.84 -0.9

Ni-S 0.5 3.12 -4.83 -8.5

NTA :NEA). 66 Ni-S 3.7 2.19 0.88 1.5 0.34

Ni-S 0.5 2.74 -6.34 -1.7

Ni-S 0.5 3.11 -5.99 -8.5

EDTA:Ni=l Ni-S 3.6 2.20 0.18 1.1 0.63

Ni-S 0.5 2.73 -4.96 3.3

Ni-S 0.3 3.09 -7.23 -0.3

EDTA:Ni=E5 Ni-S 3.7 2.19 0.63 1.9 0.31

Ni-S 0.6 2.76 -4.49 -3.8

Ni-S 0.5 3.11 -5.00 -8.4

Ehe second model was composed of two Ni-S shells and one Ni-Mo shell (Table 3).

The values of the goodness of the fit and AE° are slightly better for the first model. In

addition, the difference in the coordination numbers between the fits of the k1- and the ki-

weighted spectra varies from 0 to 0.15 for the first model and from 0.1 to 0.25 for the

110

Sulfided NiMoJSi02 catalysts

second one. All these factors suggest, therefore, that the model composed of three

subsequent Ni-S contributions has to be preferred.

As far as the differences between the various catalysts are concerned, one must be

very careful when comparing the coordination numbers obtained from the fits of the two

models, because the standard deviations van from 5% for the first shell to 40% for the

third one. It is. therefore, not possible to compare the catalysts on the basis of the

coordination numbers.

TABLE 3. Parameters obtained from the fit of the k""-weighted Ni A-edge EXAFS spectra

of the NiMo/SiO: sulfided catalysts with two Ni-S shells and one Ni-Mo shell (AR: 0.90-

3.15Â. Ak:3-15 A1)

Catalyst Shell CN R Act2 AE° Goodness

[A| [103A2] leVJ of fit

No ligands Ni-S 3.6 2.19 0.55 1.9 0.31

Ni-S 0.4 2.74 -4.85 0.1

Ni-Alo 0.7 3.07 0.17 10.6

NEA:NE-E5 Ni-S 3.6 2.19 0.42 2.3 0.36

Ni-S 0.4 2.73 -4.31 3.0

Ni-Mo 0.6 3.09 -0.34 9.0

NEA:Ni-6.66 Ni-S 3.7 2.19 0.92 1.4 0.42

Ni-S 0.3 i 70 -7.62 2.9

Ni-Alo 0.6 3.07 -0.91 9.1

EDTA:Ni--l Ni-S 3.6 2.19 0.22 1.0 0,68

Ni-S 0.3 2.71 -6.66 7.4

Ni-Alo 0.3 3.05 -3.11 17.5

EDEA:Nh-E5 Ni-S 3.7 2.19 0.63 1.8 0.32

Ni-S 0.4 2.74 -5.63 -0.7

Ni-Mo 0.5 3.07 -1.06 10.2

111

Chapter 5

Figs. 7 to 10 show the results of the fit for the catalyst prepared with EDEA (molar

ratio EDTA:NET .5) usine the two models.

13

k [A-

03

22*

03

R[A]

FTG. 7. Compaiison between measured and calculated spectra k'-wetghted %(k) (above) and Fomier

Iransfotmed (below) spectia of the catalyst containing LDIA with the molar ratio EDTA Ni - 1 5. The fit

was obtained usine three Ni-S contributions

112

SulfidedNiMo/Si02 catalysts

2T

10

— measured

fit I

k[A

measured

fit

R[\]

FIG. 8 Compaiison between measuied and calculated spectia EAveighted xOA (above) and 1 ounei

tiansfotmed (below) spectia of the catahst contamina: EDTA wnh the molat tatio EDTANi- 1 S fhe fit

was obtained usin« three Ni-S conti lbutions

in

ci 5

2'

k[A

_4

22

-0 1

-0 2

-0 3

RjÂ]

HG. 9. Compaiison between measuied and calculated spectia k-weighted %{k) (above) and Foiniei

tiansfoimed (below) spectia of the catalyst containing f D1A with the molai tatto EDTANi = i 5 The fit

was obtained using two Ni-S and one Ni-Mo conti ibuttons

Sulfided N1M0/S1O2 catah sis

22

Ph

R[A]

FTG. 10. Compaiison between measuied and calculated spectia lAweighted y(k) (above) and Fouitei

tiansfoimed (below) spectia oi the catahst containing FDl A with the molar ratio EDTANi = 1 s The fit

was obtained using two Ni-S and one Ni-Mo conti ibutions

IIS

Chapter 5

Discussion

In this section the structure of Mo and Ni in the sulfided catalysts will be discussed

taking into account the findings concerning the catalysts precursors discussed in Chapter 3

and the sulfidation mechanisms treated in Chapter 4. fhe role of chelating ligands through

the whole preparation procedure will be considered,

Molybdenum

Molybdenum is present on SiCE in the oxidic precursors as a mixture of polyanions

such as M07CE4A and molvbdatc units (AIoöaA Ihe chelating ligands NTA and EDEA

form complexes with Mo onh when the molar ratio ligand:Ni is larger than i [15, and

Chapter 3, 20]. Ehe presence of chelating ligands induces a broadening of the sulfidation

interval of Alo, i.e. the sulfidation starts at lower and is completed at higher temperature in

comparison with the catalyst prepared in absence of chelating agents. The consequences of

the different sulfidation behaviour can be noticed in the final sulfided catalysts, whose

structure is discussed here. The data of Table 1 show that the addition of chelating ligands

induces a reduction of the AI0S2 particle size and an increase in the structural order, as

deduced from the values of the coordination number and of the Debye Waller factor,

respectively. It can be noticed (see Table 1) that these effects are observable for ligand:Ni

molar ratios larger than 1. which suggests that as long as only Ni is complexed by the

organic molecules. Alo does not perceive the presence of the chelating ligands, either in the

oxidic state, nor in the sulfided state.

Ehe IIDS activity profile of the investigated catahsts (Fig. I) shows a decrease for

NTA:Ni ratios larger than 2. fhe explanation of this decrease in activity must be connected

with the observed changes in the structure of the M0S2 particles, since no change In the

structure of Ni is expected and observed for large ligand concentrations. The fact that a

lower activity is observed for smaller AI0S2 crystallites (Table 2) is in apparent contrast

with the idea that for smaller particles a higher number of active sites should be present. A

reduction of the particle dimensions should raise the fraction of Alo atoms at the edges of

the M0S2 crystallites. According to the C0AI0S model, these atoms are responsible for the

116


formation of the active sites [1. 21]. If it is assumed that this model is correct, the fact that

we find a lower activitv when the particles are smaller, would mean that cither the

arrangement of the Ni atoms on the edges of the MoS2 crystallites is hindered by some

factor, or that the ratio of Alo atoms exposed at the MoS: edges and Ni atoms becomes

unfavourable for the catahtic activity.

The Mo-Mo coordination number of the catalyst prepared in the absence of ligands is

3 6. According to Shido and Prins [19], a coordination number of 3.6 corresponds to a

M0S2 particle diameter of about 30 A. A crystallite with such a dimension is composed

approximately of 90 Mo atoms, as sketched in Fig. 11. One third of them are situated at the

edges. For the catalyst prepared with the highest amount of NFA (NTA:Ni -= 6.66) the

Mo-Mo coordination number is 3.2, which indicates that the AI0S2 particles have a

diameter of ca. 20 Â. and half of the atoms are situated at the edges. In our catalyst the

31 A

C-A^ -V V^

1

1 1 !

1 1 ' i

1'

î

-^^ %,• *S,v. V,\ i 1 f

la A k A >.

y^-

v.-* N>^ ^~-* x^.

1 ! ! ! Ï

- 'A. ^ -

^ ^^- N v^ X-

\ ] 1 \

f A A A XyA, l~A ^

y ^"" v-1 Ï 111 1

lit'

~"~Y~Y y"AXv v X,- "s~- X,- l^

t ! 1 t 1

ANv v\^ .. \ v-*, - "w -

v.

IS ! 1 !

1^

II!1!S s

"V\ y Va- 1 1 ] 1

1 1 i 1

j | | 1

' 1 \

S ^

X - A,- N/K

1 ! !

' 1

f y '""

-A,-A- A.

I9Ä

l l 1 ' l

v -yA/laAAa

ii

A

Fig. 11. M0S2 paiticles with diffeient diameteis.

Ni:Mo ratio is 0.3:1, which means that there are slightly more than 3 Mo atoms for each Ni

atom. We assume that with the help of the ligands all Nt becomes situated on the M0S2

edges. It is clear from these approximate calculations that for the larger MoS: particles

(diameter of ca. 30 A) the number of edge Alo atoms and that of Ni atoms present in the

catalysts are very similar. On the contrary, m the smaller particles the number of edge Mo

117

Chapter 5

atoms is significantly larger than the number of Ni atoms. The increase of the fraction of

free Mo atoms on the MoS: edges could be the reason for the decrease in activity observed

for higher N FA concentrations.

The explanation of this effect could be twofold: either the Ni atoms arranged on the

AI0S2 edges are too far apart to carry out the hydrodesulfurisation reaction properly, or the

presence of MoS: vacancies has a negative effect on the active sites. It has been proposed

that the promoting of Ni or Co is due to an increase of the electron density on Mo, which

would lead to more anion vacancies and consequently to a higher activity [22]. In the

presence of a high number of Mo atoms exposed at the edges, the electron donating effect

of Ni would be distributed over a larger edge area and the promoting effect would be

spread.

To understand the changes in the structural features of the sulfided catalysts, the

sulfidation mechanism of Alo has to be considered. We showed that before M0S2 is formed,

Mo is present on the support as a MoSs-like material [23, and Chapter 4]. This material is

transformed to M0S2 and the aggregation to larger crystallites takes place from the end of

the third intermediate sulfidation region (AloSMike material) until 400°C. As observed by

fitting the QEXAFS data, no significant change in the Alo-Alo signal was noticed after

reaching a sulfidation temperature of 400°C, i.e.. even though the temperature wras kept at

400°C for 30 min. small changes were observed only in the first few minutes. This suggests

that the size of the MoS: particles is dependent on the length of the temperature interval

ranging from the formation of the first MoS: units to 400°C. In the case of the catalyst

prepared without ligands. AloS: starts to form at 295°C (see Fig. 6 in Chapter 4), whereas

in the catalyst containing NTA with the molar ratio NTA:Ni =" 6.66, the first MoS: units are

detected at 360°C. Ehese observations are understandable when considering that Alo is

present in the catalyst prepared without ligands as polyanions, where groups of

molybdenum atoms are already collected in units of 7-8 atoms, whereas, when the amount

of NEA is large enough to bind also all Alo (at least for NfA:Ni = 4.3), Mo is present on

the support as isolated [MoOnNTA)p complexes. Therefore it is more likely to obtain

smaller particles when employing larger amounts of ligands.

Another phenomenon could explain why differences are observed in the structure of

Mo in the sulfided catalysts. The conclusions drawn from the QEXAFS results suggested

lis

Sulfided NiMo/Si02 catalysis

that the sulfidations of Ni and of Mo are mutually influenced. The reason for this effect is

still unclear but the fact that Ni is sulfided at higher temperatures prevents the formation of

segregated clusters of nickel sulfide and should increase the number of interactions between

Ni and M0S2. The presence of a higher number of Ni-MoS: interactions could be the reason

for a smaller M0S2 particle size. Ni could prevent further enlargement of the particles

because of its structure (see Eable 2 and 3). which is different from the trigonal prismatic

geometry of Mo in AI0S2. A similar hypothesis was proposed from 'FEM results by Ledoux

el al who suggested that the role of Co in C0AI0/AI2O3 catalysts is to stabilize small M0S2

patches and to avoid sintering of M0S2, thus keeping the M0S2 dispersion high [24]. Other

authors arrived at the same conclusion for unsupported CoAlo and NiMo compounds [25,

261.

Nickel

Nickel is present as silicate in the catalyst precursors prepared in the absence of

ligands. whereas the formation of complexes with the chelating ligands isolates the metal

ion from the support. The difference in structure in the oxidic precursors leads also to

marked differences during the sulfidation process. Increasing amounts of NTA. EDTA and

EN delay the end of the sulfidation of Ni. Ehe goal of the EXAFS investigation presented

here was to explain the consequences of the different sulfidation behaviour on the structure

of the final sulfided catalysts. Indeed, the EXAFS technique gave useful information. We

calculated that Ni is surrounded by four sulfur ligands at a distance of 2.2 A, which either

corresponds to a tetrahedral or to a square planar geometry.

The nickel-sulfur system is rather complex and the combination of the two elements

can give rise to a large number of compounds. The most common minerals composed of

these two elements are hcazlewoodite (NES:), millerite (NTS), polydymite (NES4) and

vacsite (NTS:) [27]. These four minerals are present in nature and are established mineral

species. Other binary nickel sulfides have been reported, such as NißSs, NES6, NESg [28],

and Ni|7Si8 |29J. The crystallographic data about NES:. NTS, NE,St, NiS2 and Ni^Sis can

be found in the literature. Among all these compounds, the only one in which Ni has a

tetrahedral coordination and a Ni-S distance of 2.21 A is Ni3Si. This material, however.

119

Chapter 5

decomposes at 200°C to NiS when treated with FES and hydrogen. In principle it is,

therefore, impossible that it is present at our sulfidation conditions (400°C). EXAFS

measurements are carried out at liquid nitrogen temperature, so that the presence of M3S4

is, in principle, possible. It is. however, very unlikely, since it means that it should be

reformed when the sample is cooled after the sulfidation. However, fits of the QEXAFS

spectra measured at 400°C gave coordination numbers around 3.5 and a Ni-S distance of

2.19 Ä. demonstrating that the observed phase is the same as the one observed in the

EXAFS spectra measured at liquid nitrogen temperature. Ehe other inorganic Ni-S

compound with a tetrahedral structure is heazlewoodite (NES:), in which a Ni-S distance is

reported at 2.28 A and a Ni-Ni distance at 2.51 A. which is not observed in our spectra.

A survey of organometallic compounds gave us a more complete picture of the

possible structures of Ni in a sulfur environment. In general, the Ni-S distance in

compounds where Ni has a tetrahedral coordination ranges from 2.25 to 2.35 A [30],

whereas it amounts to 2.14-2.20 A in case of a square planar structure [31, 32]. Ehe

distance obtained by EXAFS (2.19 A) thus suggests a square planar geometry. In

accordance with this suggestion. Ni is sterically protected by the S ligands in case of a

tetrahedral structure and the absorption of a substrate during catalytic reaction would be

hindered. On the contrary, the square planar geometry leaves the surface free for interaction

with substrate molecules.

'fhe nature of the second and third shells might clarify the structure of the

investigated material. Louwers and Prins suggested that these two shells arc due to Ni-Mo

and Ni-Ni contributions [10]. Ehe data range that they used for the Fourier transformation

was 3-11 A"1. Fig. 12 shows how different the spectrum of the same sample can be when

using different k-ranges. As a consequence, also the interpretation of the data can change

dramatically. Thanks to the more powerful X-ray source and long collecting times, we

could collect data that were usable in the range 3-15 A"1. We tried to fit the data with the

shell combinations proposed by formers and Prins but the fits did not satisfy the adopted

criteria: the fits converged for the IE-weighted spectra but never for the k'-weighted data.

We obtained two possible structures of Ni in the sulfided state. The first model, consisting

of three Ni-S shells docs not show any Ni-Mo proximity. If this model were valid, it would

mean that the EXAFS results do not allow to say anything about the position of the Ni

120

Sulfided NiA4o/Si02 catalysts

R [A]

FIG. 12. Comparison between Fourier transformed data obtained using different k-range.

atoms on the MoS: edges in these catalysts. Ihe second model, in which Ni is surrounded

by two Ni-S shells at 2.2 and 2.75 A and a Ni-Alo contribution at 3.07 A, would be in

agreement with the proposed structure of the active phase, in which Ni decorates the edges

of the MoS: crystallites [33].

Ehe inability to discern which is the correct model and the inaccuracy of the

coordination number of the second and third shell, show some limitations in the use of

EXAFS. Nevertheless, both models have many similarities. In both models Ni is

surrounded by 3.6-3.7 S atoms at 2.19 A and about 0.4-0.7 S atoms at 2.75 A. They differ

in the composition of the third shell, being S for the first and Alo for the second model. 'Fhe

parameters of the first Ni-S shell suggest a square planar geometry. As far as the Ni-S shell

at 2.75 A is concerned, a similar Co-S distance was found by Crajé et al. in Co/C sulfided

at 400°C [ 34]. Their explanation for this observation was a shrinkage of the second nearest

Co-S distance belonging to CoçSg. which usually is 3.47 A. Such a reduction seems,

however, excessive.

121

Chapter 5

No other inorganic Ni-S compounds reported in the literature show a similar Ni-S

distance. Ehe only compound which has a Ni-S distance around 2.75 A is trimeric

bis(dithiobenzoato)-nickel(H). whose structure is depicted in Fig. 13 [35].

v_ T~ A v

Fig. 13. Structure of trimeric bis(dithiobenzoato)-nickel(ll) (from [35])

In this compound every monomer is composed of a Ni atom positioned in the centre

of four sulfur atoms of two dithiobenzoate anions with a Ni-S distance of 2.22 A. The two

external Ni are arranged in a square pyramidal geometry, whereas the central one has either

a octahedral geometry or a square planar structure with significant Jahn Teller distorsions.

The three units are interconnected by means of Ni-S bridges having a bond length of 2.78

and 3.11 A (Fig. 14). These two latter distances are different because the external Ni atoms

are slightly displaced towards the central monomer. Ihe resulting Ni-S distance is, thus,

reduced to 2.78 A.

Because of the strong resemblance between the distances observed in our data and in

this compound, we simulated the spectrum of the trimer by means of the Feff code. For the

simulation an overall Debye Waller factor of 0.004 A" was used. The phenyl rings were

excluded from the simulation, only the carbon atoms as indicated in Fig. 14 were retained.

In Fig. 15 the x(k)lA data of the simulation are plotted together with the data of the catalyst

containing N'EA with the molar ratio NFA:Ni=- 1.5. A horizontal shift can be observed

between the two curves, which is more marked around 13 A"1. Ehis shift is likely due to the

122


3,107

\ 2.776

Fig. 14. Distances in trimeric bis(dithiobenzoato)-ntckel(lI).

fact that the distances of the first and second shell in our compounds are shorter in

comparison to the corresponding distances in the trimer. A shorter distance produces a

longer period of the EXAFS oscillations as can be understood from equation (12) in

Chapter 2. Apart for this discrepancy, the two functions are practically identical. Fig. 16

shows the Fourier transformed spectra of the same compounds. Also in this plot a shift of

the imaginary part can be noticed. A comparison between the k - and the IE-weighted

Fourier transforms for the two spectra showed that the second and third shells behaved in

the same way for the two weighting factors. Vaarkamp showed that it is possible to

distinguish between Alo and S backseattercrs comparing the k - and the k3- weighted

spectra [361. Since Mo is a heavier backscatterer. the signal belonging to a Ni-Mo shell

should be amplified by a larger extent than a Ni-S signal, when increasing the power of the

123

Chapter 5

10

22

NTA:Ni=E5

Ni-trimer

k IÄ-1]

Fig. 15. Comparison between the Ni A-edge %(k) E data of the measured NiMoNTA/Si02 catalyst (molar

ratio NTA:NI-1.5) and the simulated spectrum of trimeric bis(Dithiobenzoato)-nickeI(II).

13

22

-6 5

— NTA:Ni=E5

" " "

Ni-trimer

Fig. 16. Ni K-edge Fourier transformed spectra of the cataly st containing N FA with the NfA:Ni molar

ratio 6.66 and the simulated spectrum of trimeric bis(Dithiobenzoato)-nickel(TT).

124


weighting factor. The amplitudes of the second and third shells for the simulated spectrum

and the spectra of our catalysts were similar not only when using a k'-weighting but also

when using a lA-weighting, Therefore, both signals of the catalyst spectra consist of Ni-S

contributions. If one of the signals had been a Ni-Alo contribution there would have been a

larger amplification passing from the k1- to the IE-weighted spectrum. This consideration

excludes the second model that we proposed which involved the presence of Mo in the

third shell.

The main difference between the two spectra of Fig. 16 consists in the signal at 3.4 A

(phase uncorrected), which has a relatively large intensity for the simulated spectrum,

whereas it is very weak for the spectrum of the catalyst. Ehis signal is due to the Ni-Ni

contribution at 3.75 A. which is the distance between two Ni atoms belonging to two

different monomers. The fit of the signal at 3.3 A (phase uncorrected) in the spectra of the

catalysts was converging only in the k'-weighted spectra, but not when using ^-weighting.

We attribute this inability to the weak intensity of the signal. The weak amplitude of the

Ni-Ni signal suggests that not all Ni in the catalyst is present as a small cluster. However,

the explanation of this feature could be twofold. One interpretation is that part of it is

present as isolated units. In fact, we noticed that the addition of isolated square planar Ni

complexes gave a marked decrease in the amplitude of the Ni-Ni signal in the simulated

spectrum. Another interpretation is that Ni forms larger sulfided clusters, in which more

than one Ni-Ni distance is present. The superposition of the signal of Ni-Ni shells at

different distances could lead to the weakening and eventually to the disappearance of the

signal in the EXAFS spectra. A possible structure of the clusters could be similar to the

well-known PdCE structure, in which chains of square planar PdCf units are ordered

parallel with a distance of 3.8 A to each other (37). Another difference between the two

spectra of Fig. 16 is the ratio between the Ni-S signals at 2.3 and 2.8 A (phase uncorrected).

In the simulated spectrum of the trimer the first signal has a significantly larger intensity

than the second one. whereas in the catalyst spectrum the two signals have almost the same

125

Chapter 5

intensity. An accurate analysis of the composition of the Ni-S signal at 2.3 A (phase

uncorrected) showed that two contributions overlap and produce the observed peak, one

being the Ni-S contribution at 2.77 A, the other the Ni-C contribution at 2.69 A. A

simulated spectrum in which the Ni-S contribution at 2.78 A was omitted showed that a

good fit can not be obtained without the Ni-S contribution. Proof for the presence of Ni-C

is difficult to obtain because carbon is a weak backscatterer and even the presence of a high

number of carbon neighbours has a minor effect on the intensity of the signal. Thus, the

elimination of the Ni-C contributions from the simulated spectra gave only a slight decrease

of the signal at 2.3 A (phase uncorrected). Changes in the intensity of the signal at 2.3 A

can, therefore, be induced by a change in either the Ni-S or the Ni-C coordination number.

From these observations a new picture is obtained of the structure of Ni in SiO:

catalysts. The correspondence between the structure of trimeric bis(dithiobenzoato)-

nickel(ll) and the data obtained from our EXAFS spectra suggests the presence of small

sulfided Ni(II) clusters in our catalysts. This would explain the simultaneous presence of

two Ni-S distances, at 2.77 and 3.11 A. Aloreover. the weak signal at 3.3 Â (phase

uncorrected) suggests that a Ni-Ni shell is even visible in our spectra.

The question that arises at this point is whether Ni can be present as a thiocarbamate

complex in the sulfided state. We first consider the catalysts prepared with chelating

ligands. It has been proved that Ni in the catalyst precursors is complexed by NTA and

EDTA [15, 20]. Fhe behaviour of the organic ligands under sulfidation conditions has not

been investigated, yet. Flowever. it is likely that the keto groups in the acetate arms of the

ligands may be substituted by FES and that thiocarboxylic groups are formed, which can

replace the other ligands around Ni and form dithiocarbamate-Ni complexes as proposed in

the mechanism in Fig. 1 8.

In the catalyst prepared without ligands. the presence of carboxylic groups can be

excluded since no organic compound was used during the preparation of this catalyst.

Therefore, the formation of metallorganic complexes in the catalyst prepared without

ligands can not be considered. It is more likely that larger clusters are formed of the kind

proposed in Fig. 17. Nevertheless, no prediction can be made about the size of the cluslers.

126

Sulfided NiAlo/Si02 catalysts

K

S—H

H.XS 0

,0H2

N- -CH;

M

H.

HO-

,©

,0G

N-

N

-CH2

S H

-CHo

M

S. S"

—-CH?

HS JS

© \f°H2o-rY

N- -CHc

HH.

NE

.©

,0H

-CH,

Fig. 18. Proposed mechanism for the formation of the dithiocarbamate group from the carboxylic groups

of the chelating ligands.

According to this interpretation the role of the chelating ligands is to avoid formation

of larger Nl-S clusters and favour the separation of the metal sulfide in smaller units.

The EXAFS data did not allow us to investigate the presence and structure of

metallorganic complexes on the catalysts. Ehe only sign that could indicate the presence of

0 at 2.7 A from Ni is the increase of the intensity of the signal at 2.3 A (not phase-

corrected) for the spectrum of the catalyst containing NEA with the molar ratio

NEA:Ni = 6.66 (Fig. 5). In the same spectrum it is possible to observe that the signal at

3.3 A is significantly lower in comparison to the spectra of the other catalysts. As far as the

relationship between Ni and AloS: is concerned, it is impossible to tell from our data

whether an interaction is present between the Ni complexes and the M0S2 crystallites.

127

Chapter 5

Conclusions

The use of EXAFS for the study of supported sulfided NiAlo catalysts demonstrated

that this technique is a powerful tool for the study of this amorphous and dispersed catalytic

system. We could show that molybdenum does not perceive any significant effect from the

use of chelating ligands during the preparation of the catalyst precursors when low amounts

of ligands are employed. From this we deduced that the beneficial effect of chelating

ligands is due to their effect on Ni, since higher ligand amounts influence the structure of

Mo and induce a decrease in the FIDS catalytic performances. A change in size and

structural order of the MoS: crystallites was detected for higher NEA and EDEA

concentrations. It would have been extremely difficult to prove differences in particle size

of less than 10 A on the same catalytic system by means of other techniques. From these

observations it is possible to say that the lower HDS activity rate observed for high ligands

concentrations must be ascribed to the changes in the structure of Mo in the sulfided

catalysts. No Mo-support interaction was detected in the studied catalysts.

Even though the quality of the Ni A-edge data was lower than the Mo A-edge data, it

was possible to propose a structure for Ni in the sulfided catalysts. Ehe relatively short

distance of the nearest Ni-S shell (2.19 A) suggests that Ni is present in a square planar

geometry rather than in tetrahedral coordination. Moreover, from the Ni-S distances (2.77

and 3.11 A) and the signal at 3.3 A (phase uncorrected), that we attributed to a Ni-Ni shell,

it was possible to propose that Ni forms a lay ered structure.

The mechanism of sulfidation of the ligands was discussed, 'fhe presence of the

chelating ligands during the sulfidation could enable the formation of a thiocarbamate

complex composed of a small number of Ni atoms. Therefore, the role of the ligands is to

improve the dispersion of Ni by forming complexes that stay apart from the larger nickel

sulfide clusters.

To improve the signal to noise ratio of the Ni A-edge EXAFS data, the Ni loading

should be raised proportionally with the Alo loading trying to stay below the monolayer

coverage of the support. Ehe support could be changed to carbon, on which the fraction of

Ni present as active phase is relatively high in comparison with other supports. In this way

the danger of having a mixture of different Ni compounds on the support would decrease

128

Sulfided NiA4o/Si02 catalysts

significantly. Moreover, measuring EXAFS in fluorescence mode, instead of transmission

mode, could improve considerably the quality of the data because the noise produced by the

other elements present in the samples would be excluded and the Ni signal could be

isolated, fhe scarce availability of appropriate detection systems makes, however, the

collection of EXAFS data in fluorescence mode still impracticable.

Literature

1. Fopsoe, FE, Clausen. B. S.. Candia, R., Wivel. C, and Morup, S,, J. Catal. 68, 433

(1981).

2. van der Kraan, A. AT. Crajé, M. W. E, Gerkema, E.. Ramselaar, W. L. T. AE, and

de Beer. V. El. L.Appl Catal. 39. L7 ( 1988).

3. Crajé. AE W. E, de Beer. A. FF E. and van der Kraan. A. M.. Appl. Calai 70. L7

(1991).

4. Weber. Eh., Muijsers. J, C. and Niemantsverdriet, E W., J. Phys. Chem 99, 9144

(1995).

5. Muijsers. E C. Weber, fh.. van Ilardeveld. R. AI., Zandbergen, EI. W7., and

Niemantsverdriet. EW..J Catal 157.698(1995).

6. Weber. Th.. Aluijsers. E C, van Wolput. J. FI. M. C, Vcrhagen. C. P. E, and

Niemantsverdriet.E W.,,7 Phys Chem 100. 14144(1996).

7. Eouwers, S. P. A.. Crajé. Al. W. E. van der Kraan. A. AE, Geantet. C, and Prins, R.,

,7 Catal 144, 579(1993).

8. Stockmann. R. M., Zandbergen. FI. W.. van Eangeveld, A. D., and Moulijn. J. A..

J.Mol Catal. 102, 147(1995).

9. Calais. C, Matsubayashi. N., Geantet, C., Yoshinuira. A7., Shimada, FF, Nishijima.

A.. Lacroix, M., and Breysse. AE. ,7 Catal. 174. 130 (1998).

10. Eouwers. S. P. A., and Prins. R.. J Calai 133. 94 (1992).

11. de Boer. M., van Dillen. A. E. Koningsberger. D. C. and Gcus, J. W., J. Phys

Chem 98,7862(1994).

12. Aledici, L.. and Prins. R.. J Catal 163. 38(1996).

129

Chapter 5

13. Topsoe, IT., Clausen, B. S., and Massoth. F. E., "Catalysis Science and

Eechnology", Springer Verlag, New York, 175 (1996).

14. Eeliveld, R. G., van Dillen. A. E, Geus. J. W.. and Koningsberger, D. C,,/ Catal.

165.184(1997).

15. Cattaneo, R.. Shido. T.. and Prins, R.. d Catal. 185, 199 (1999).

16. Kampers. F. W. FF. Alaas, E. AE E. van Grondellc, E, Brinkgreve, P., and

Koningsberger, D. C, Rev. Sei. Instrum. 60, 2635 (1989).

17. Vaarkamp. AE. Dring, !.. Oldman, R. E. Stern, E. A., and Koningsberger, D. C,

Phys. Rev. «50,7872(1994).

18. Zabinsky. S. E, Rehr. J. E. Ankudinov. A., Albers. R. C. and Eller, AT E. Phys.

Rev. S 52. 2995 (1995).

19. Shido, T.. and Prins, R.. J. Phys. Chem. B 102. 8426 (1998).

20. Aledici. L., and Prins. R.. ,7 Catal. 163. 28 (1996).

21. Ratnasamy. P., and Sivasanker, S., Catal Rev.-Sei Eng. 22. 401 (1980).

22. Harris. S., and Chianelli. R. R., J. Catal. 98. 17 (1986).

23. Cattaneo. R.. Weber. Th.. Shido, T.. and Prins. IE, .7. Catal. 191, 225 (2000).

24. Eedoux. Al. E. Maire. G.. ITantzer. S., and Alichaux. O.. in "Proc. 9th Int. Congr.

Catal.". 'fernan, AE and Philips, M.E Eds.. Chem. Institute of Canada, Ottawa, 74

(1988).

25. Pratt. K. C. and Sanders, J. V., in "Proc. 7th Int. Conf. Catal.". Seiyama, T. and

Tanabe, K. Eds., Kodansha'Elsevier. fokyo/Amsterdam, 1420 (1981).

26. Vrinat, M. L., and Alourgues, E. d„ Appl Catal. 5, 43 (1983).

27. Kullerud, G.. and Yund, R. A.. J. Petrology 3, 126 (1962).

28. Honig. J. AE. and Spalek. E. Chem. Mater. 10. 2910 ( 1998).

29. Collin. G., Chavant. C. and Comes. R.. Acta Crystallogr. Sect. B: Struct. Sei. 39,

289(1983).

30. Rheingold. A. L.. Beall. K. S.. Riggs. P. E. and Groh. S. E., Acta Crystallogr., Seel.

C 49, 542(1993).

31. Coucouvanis, D.. Patil, P. R.. Kanatzidis. AE G.. Dctering. B., and Baenziger, N. C.

fnorg. Chem. 24,24(1985).

130


32. Kushch, N., Faulmann, C, Cassoux. P.. \ralade. L.. Alalfant, E, Legros, J.-P.,

Bowlas, C. Errami. A.. Kobayashi, A., and Kobayashi. IE, Mol. Cryst. Liq. Cryst.

Sei. Techno!., Sect. A 284, 247 (1996).

33. Candi a, R., Sorenson, O., Villadscn. E. Eopsoe. N. Y., Clausen, B. S.. and Topsce.

H.,.7. Phys. Chem 95. 123 (1991).

34. Crajé. AI. W. E. Eouwers. S. P. A., de Beer, V. H. E, Prins, R., and van der Kraan,

A. AE,,/. Phys. Chem. 96, 5445 (1992).

35. Bonamico, M., Dessy, G.. Fares, AE. and Scaramuzza, L., ,/. Chem. Soe. Dalton

Trans. .2250(1975).

36. Vaarkamp, AE. Cat. Today 39, 271 ( 1998).

37. Wells. A. F.. Zeit. Kristall. 100, 189 (1938).

131

Chaptei 5

132

Chapter 6

Influence of chelating ligands on the HDN and HDS

behaviour of y-Al203-supported NiMo catalysts

Introduction

After the detailed study of the effects of chelating ligands on SiO:-supported catalysts

presented in the previous three chapters, we concentrate in this chapter on alumina-

supported catalysts, fhis support is the one mainly used In industrial applications because

of the stronger interactions between the metal sulfides and the support, which induces a

higher stability of the catalysts at elevated temperatures [1, 2].

In this chapter first the catalytic performances in hydrodcnitrogenation (HDN) and

hydrodesulfurisation (IIDS) are compared. Ehen the QEXAFS results measured during the

sulfidation of the catalysts are presented, that allow to study the sulfidation process of Mo

and Ni in the presence and absence of chelating ligands. To complete the characterisation

part, the FXAES data of the sulfided catalysts arc discussed in order to understand the

structural features that influence the activity of the studied materials.

HDN reaction network

Hydrodcnitrogenation (HDN) is an important step in hydrotreating. Several authors

have studied the activity and structure of the catalysts [3] as well as the mechanisms and the

kinetics of the E1DN reaction [4. 5]. We have choosen toluidine as a model compound

because in its network the most important reactions occur which take place In a HDN

network, folmdine (TOT) can be hydrogenated to methy Icyclohexylamine (MCHA) or can

Chapter 6

react to toluene (T). Finder our conditions. E can not react further, as was shown in a study

of its selectivity in our laboratory [6. 7]. AICHA then reacts via elimination to

methylcyclohexene (MCHE) that can react further via hydrogénation of the double bond to

the final product methyleyclohexane (MCEI). AICHA can also react directly to MCEI via

nncleophilic substitution of the amine group by a sulfhydril group |8]. The reaction network

is shown in Fig. 1.

Fig. 1. O-tohndine HDN reaction network

IIDS reaction mechanism

'fhe field of hydrodesulfurization has been the subject of numerous reviews |9-11 ].

Thiophene (TP) is often used to test the HDS catalytic activity of catalysts [12, 13]. Its

network (Fig. 2) is relatively uncomplicated. Vt elevated IE pressures the major reaction

path in the IIDS of thiophene is via hydrogénation of thiophene to tetrahydrothiophene

[14]. fhis intermediate can react to butadiene through two successive ß-EI eliminations. At

high EE pressure, butadiene reacts quickly further to 1-butène, 2-butene (eis and trans) and

eventually to butane.

Ehe HDS mechanism of thiophene at one atmosphere pressure is still under debate

after several decades of investigation. Based on the presence of butadiene (BDE) and the

absence of fetrahydrothiophene (EHE) m the reaction products obtained in atmospheric

pressure studies, several authors have proposed that the path through BDEi (a. b pathways

shown in Fig.l) is the major route [15. 16]. A direct pathway (pathway f shown in Fig. 2)

for the hydrodesulfurization of IP directly to butène (BE) has been suggested on the basis

of the absence of Ulf [17]. Ehe reaction intermediates are supposedly retained on the

134

IIDN-HDS ofNiAdo'AfO^ catalysts

surface sites during reaction. Another study with tetrahydrothiophene over MoS: at low

pressure found substantial quantities of thiophene, as well as butadiene and suggested two

reactions paths (e. b and c. a. b) for HDS of this compound [18]. It has been revealed that

hydrogenated S-intermediates are present, suggesting parallel paths [19] or even that

prehydrogenation may be necessary before C-S bond cleavage occurs [20].

THE

B

BDE

Fig.2. Network of the HDS of thiophene

Experimental

Catalysts preparation

The catalysts used in this work contained about 7 wt% Mo and 2.5 wt% Ni and were

prepared by pore volume coimprcgnation of y-AKE (CONDEA. pore volume: 0.5 cmVg,

specific area: 100 nr'g) with an aqueous solution (pll 9.5) of (TNFH)6A1o70:a4H:0

(Aldrich) and Ni(NOY'6H:0 (Aldrich) with 25% ammonia in the presence or absence of

the chelating ligands nitrilotriacetic acid (NTA) and ethylenediamine tetraacetic acid

(EDTA). The support was dried at 120°C for 12 h prior to impregnation. The impregnated

powder was dried in air at ambient temperature for 4 h and dried in an oven at 120°C for 15

h. Calcination was carried out only for the catalyst prepared without ligands at 500°C for

4h. For all other catalysts calcination was not carried out in order to avoid the

decomposition of the complexes in the cataly st precursors. Ehe investigated catalysts are

listed in Eable 1.

135

Chapter 6

TABEE 1. List of catalysts used in this work

Catalyst Loading (wt%) Ligand :Ni

Ni Mo (molar ratio)

NiMo calcined 6.8 2.6 0.0

NiMo not calcined 6.9 2.5 0.0

NiAloNFA 6.0 2.2 0.5

NiMoNTA 5.8 2.2 1.0

NiAloNFA 6.0 2.2 1.5

NiMoEDFA 6.0 2.2 1.0

HDN activity tests

70 mg of the oxidic precursor were diluted with 8 g SiC to achieve plug-flow

conditions in the continuous flow fixed bed reactor. The catalyst was sulfided in situ with a

mixture of 10 % IES in H: at 400°C and 1.0 AlPa for 2 h. After sulfidation, the pressure

was increased to 5.0 MPa. and the liciuid reactant was fed to the reactor by means of a high

pressure syringe pump (ISCO 500D). All reactions were performed at 370°C.

Dimethyldisulfidc (E)A1E)S) was added to the liciuid feed to generate FES in the reaction

stream. Toluidine was used as model compound to study the HDN reaction and

cyclohexene was added to the feed to study the hydrogénation reaction. The composition of

the gas phase reactant for the catalytic tests was 7 kPa of toluidine, 4 kPa of cyclohexene,

4800 kPa of H2. 134 kPa of octane. 20 kPa heptane (reference) and 17.5 kPa of H2S.

"fhe reaction products were analysed by on-line gas chromatography with a Variait

3800 GC instrument equipped with a 30 m DB-5 fused silica capillary column (J & W

Scientific, 0.32 mm Ed.. 0.25 pm film thickness), a flame ionisation detector (FID) and a

pulsed flame photometric detector (PFPD). Space time was defined as x = wc / niCCd. where

wc denotes the catalyst weight and ntccj the total molar flow fed to the reactor. Space time

(t) was changed by varying the liquid and gaseous reactant flow rates, while their relative

ratios remained constant.

136

HDN-HDS ofNiAfo'AI-,0, catalysts

HDS activity tests

100 mg of the oxidic precursor mixed with 1 g of SiC were sulfided for 2 h at 400°C

(heating rale. (ACTnin) with a mixture of 10% H:S in ET: (Messer Griesheim) that flowed

through the reactor from the beginning of the heating process. The activity of all the

catalysts was tested in the hydrodesulfurization of thiophene at 370°C. The feed (3%

thiophene in TE) was obtained by bubbling IT: through a series of four thiophene saturators

that were cooled to 2°C. Ehe product stream was analyzed online with an HP5890 gas

Chromatograph. The sulfidation of the oxidic precursors and the thiophene HDS reactions

occurred at atmospheric pressure in the apparatus described in Fig. 1 of Chapter 3.

X4FS measurements

The Quick EXAFS measurements were carried out at the XI (RÖA10 II) beam line at

EIASYEAB (Flamburg. Germany), whose set-up is described in Chapter 4. Si (311) and the

Si (111) crystals were used in the monochromator for the Mo and Ni A'-edges, respectively.

Ehe accumulation time was about 0.2 sAtep at the Alo A-edge and about 0.4 s/step at the Ni

Ä'-edge. Ehe k-ranges used for the analysis of the data were 3-17 A"1 for the Mo and 3-12

A"1 for the Ni A-edge. Ehe EXAFS spectra were collected at the Swiss Norwegian Beam

Line at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France. Ehe

beam line is described in Chapter 3. "Fhe k-range used for the analysis of the data was 3-

19.5 A"1 for the Mo A-edge and 3-16 A"1 for the Ni A-edge.

For both kinds of measurements the catalyst samples were pressed into self-

supporting wafers and mounted in an in situ EXAFS cell [21]. Ehe thickness of the samples

was chosen to adjust the edge jump to 1 for the Alo A-edge and the total absorption to

px - 4 for the Ni A-edge (lower Ni concentration).

For the QEXAFS measurements first two spectra of the fresh samples in an He

atmosphere were collected, 'fhe samples were then sulfided m situ during data collection. A

stream of 10% H:S/1E flowed through the cell while it was being heated to 400°C. Ehe

heating rate was 3°C/min.

For the classical EXAFS measurements, the samples were cooled to room

temperature after they had been sulfided at 400°C for 30 min. Once at room temperature,

137

Chapter 6

the ILS still present in the cell was replaced by He by flushing for 10 min. Ehe cell was

then cooled to liquid nitrogen temperature prior to the EXAFS measurement.

EXAFS analysis

Ehe prognim XDAP (version 2.2.2) was used to analyze and fit the data [22] as described in

Chapters 3 and 4. Apart from AloS: (references for Alo-S, Alo-Mo), reference spectra were

calculated using the Feff7 code [23. 24].

138

HDN-HDS of NtA4oC\I20, catalysts

Results

Catalytic performance

HDN activity tests

Ehe HDN of o-toluidine (IOL) shows that toluene (T) and methylevelohcxylamine

(MCHA) are primary products. Fhe selectiv ity of toluene as a function of conversion is

constant, which proves that f is produced in parallel to the other products [6]; the path from

EOE to E corresponds for all catalysts tested to about 5-10% of the total conversion. The

produced AlCITA reacts so quickly that its concentration can not be detected, or it is

100 — — — — — — —

1

80I i i |

['

.2 60 -

i—| I I'

o I l i

>

o 40 ;o

20

oLJ _l—L _L_L I—I 1—1__ I—1_Calcined Dried NTA Ni=0 5 NTA Ni=1 NTA Ni=1 5 EDTANi=15

Fig. 3. Activities of'NiMoAEO- catalysts in the HDN of o-toluidine

observed in trace amounts only when the HDN of EOE alone is carried out [8].

Fig. 3 summarizes the results of the IIDN of fOL with the different catalysts

investigated in this work. No significant difference is observed in the HDN activity

between the calcined and the dried catalysts prepared in the absence of chelating ligands.

Ehis comparison is important because calcined catalysts arc usually employed in industrial

applications, fhe addition of NEA induces an increase in catalytic activity and a maximum

139

Chapter 6

is obtained for the catalyst with the molar ratio N fA:Ni = 1, which is about 55% more

active than the catalyst prepared without ligands. For higher and lower NTA concentrations

lower activities are obtained. Ehe addition of EDEA does not have the same effect as NTA.

Only a slight increase in EIDN activity is observed for the catalyst containing EDTA with

the molar concentration EDTA:Ni = 1.

Fig. 4 shows the conversion of CITE with the different catalysts. In this case the

behaviour of the catalysts is completely different in comparison to the FOL reaction.

Increasing the NTA loading provokes an increase in the conversion of CHE. Ehe calcined

catalyst shows a higher activity than the dried catalyst prepared without ligands. The dried

catalyst prepared without chelating ligands shows a slightly higher activity than the catalyst

containing the smallest NTA amount.

100 ,-—---

80

> |

o 40 [

20 !

I

0 _L__J ^l 1 I 1 I 1 L__J _,—I 1_

Calcined Dried NTANi=0 5 NTANiA NTA Ni=1 5 EDTA Ni=1

Fig. 4. Conversion of cyclohevene on AECAsiipported "NiMo catahsts prepared with and without chelating

ligands.

77775* activity tests

'Fhe I IDS catalytic performance was tested in the hydrodesulfurisation of thiophene at

atmospheric pressure. Ehe results of the tests are shown in Fig, 5. 'Ehe activity of the

calcined catalyst is significantly higher than that of the dried catalyst prepared in the

absence of ligands. Ehe effect of the addition of NEA is dependent on the used amount of

the chelating ligand. For the molar ratios NlA:Ni - 0.5 and 1 a decrease in activitv is

140

HDN-IIDS ofNiMo/AM, catalysts

observed, whereas for the molar ratio NEANT =-1.5 the activity becomes comparable to

that of the calcined cataly st.

0 12 >

0.1

w" i 1

| 0 08

o) I I!

I 0 06 | I I

g

? o 04 ;

0 02

0J—L_ I 1 1—I J—L J—I—,—I 1

calcined dried NTANi=0 5 NTANi=1 NTANi=15 EDTANi=1

Fig. 5. fhtophene HDS acttv it\ täte fot NiMoAECE catah sts ptepai ed in the presence and absence of

chelating ligands

Although no important catalytic improv ement is observed when chelating agents are

employed, it is interesting to compare Fig. 5 with the data obtained for the

hydrodenitrogenation of! OL (Fig. 3) and the hvdrogenation of CHE (Fig. 4). The catalyst

showing the lowest HDS activity is the most active in the HDN reaction. Similarly, it can

be noticed that the catalysts showing the highest FIDS activity are the least active in the

IIDN of FOL. A comparison between the HDS and the hvdrogenation reaction shows that

the two plots (Figs. 4 and 5) have a similar profile. Ihe catalysts containing NTA with the

molar ratio NTA:Ni = 0.5 and 1 arc the least active ones, while the calcined and those with

the molar ratios NTA:Ni = 1.5 and FDTANi - 1 are the most active.

The various reactions investigated in this work follow different pathways. From the

data presented in Figs. 3 to 5 it is possible to see that a change in the structure of the

catalyst enhances some reactions and hinders others. From the QEXAFS and classical

EXAFS data we wanted to learn what features in the sulfidation behaviour and in the

141

Chapter 6

structure of the final catalysts induce the observed differences in catalytic performance for

the various reactions.

Catalysts characterisation

Molybdenum

Mo A'-edge Quick EXAFS. In Fig. 6 the QEXAFS spectra measured during the

sulfidation of the calcined NiMo/AEO^ catalyst are plotted against the sulfidation

temperature. In the spectrum of the fresh catalyst one recognizes the Mo-0 contributions

w

o

CD

B

a>>-i

CD

O

R IÂ]

Fig. 6. Mo A-edge QtXAFS spectra collected dining the sulfidation ofthe calcined NiMo catalyst

142

HDN-HDS ofNiMo/AfO, catalysis

between 0.8 and 2 À (phase uncorrected) and a Mo-Mo shell at 3 A (phase uncorrected).

fhis latter shell suggests the presence of poly moly bdates in the catalyst precursor. The

nature of the signal at 2.2 A (phase uncorrected) is not clear.

Mo starts to perceive the effects of the sulfidation in the spectrum measured at 50°C,

in which the amplitude of the Mo-0 signal has decreased in comparison to the spectrum of

the fresh catalyst. In the spectrum measured around 105°C a Alo-S signal starts to grow at a

distance of 2 A (phase uncorrected) from the absorber atom. Around the same temperature

22

2h

400

380

350

330

in

Spu

O

285

260

230

0r-t-

3A3CD

205

180 0>

1 55

145

o

n

1 25

95

70

50

fresh

2

R[A]

Fig. 7. Mo /v-edge QtXAFS spectia collected dining the sulfidation of theNiMoNTA/AECh catalyst with

fhe molar tatio N fA Ni-1

143

Chapter 6

another signal is visible in the spectra at 2.5 A (phase uncorrected). Ehe same signal was

observed in SiOo-supported NiMo catalysts and was attributed to a Mo-Mo shell present in

an intermediate product of the sulfidation of molybdenum [25, and Chapter 4]. 'Ehe Mo-S

and the Mo-AIo signals grow with increasing sulfidation temperature, but in the spectrum

recorded around 255°C the Alo-Mo shell starts to shift towards larger distance. At this

temperature MoS: starts to form and two Mo-Mo signals overlap: one belonging to the

AloSrlikc material, which is an intermediate product of the sulfidation, the other produced

by the Mo-Mo shell of MoS: at 3.16 A. It is interesting to see that the Mo-0 signal is still

clearly present in the spectra collected at 400°C.

The sulfidation of Alo in the NiAloNTA/ALOi catalyst with the molar ratio

NTA:Ni -- 1 is represented in Fig. 7. The QEXAFS spectra show that the sulfidation of Mo

In this catalyst is very similar to that in the calcined catalyst. Nevertheless, for the spectra

collected between 205 and 260°C a decrease can be observed in the Mo-S signal at 2 A

(phase uncorrected). After fitting the QEXAFS spectra with a Mo-S and a Mo-Mo shell we

could see that in the mentioned temperature range the Mo-S coordination number first

decreases and then increases acain.

100 150 200 250 300 350 400

r ^q

Fig. 8. Mo-S cooidmation number obtained ttom the fit of the QEXAFS spectia as a function of the

sulfidation tempetature tot the alumina suppotted catahsts

CT)1

o

HDN-HDS of NiMo/Ah03 catalysts

A similar behaviour was observed for all dried catalysts. Fig. 8 shows the values

obtained by fitting the QFXAFS spectra for the Alo-S coordination number as a function of

the sulfidation temperature for the various catalysts. A maximum is present for every

catalyst, after which a decrease is observed, that corresponds to the decrease of the

amplitude of the Mo-S signal noticed in the QEXAFS spectra. The maximum is reached at

different temperatures for the various catalysts. Ihe catalyst for which the maximum is

reached at the lowest temperature (215°C) is the one prepared using NTA with the molar

ratio NTA:Ni - 1. Ehe other maxima are observed at 240°C for the catalyst containing

NEA with the molar ratio N fA:Ni = 1.5.at 255°C for the one prepared without ligands

and at 250-260°C for the catalyst prepared with EDEA. A further clarifying plot is shown

in Fig. 0, where the Alo-S and Alo-Mo distances are plotted as a function of the sulfidation

temperature. From this figure it is visible that, in the same temperature regions where the

maxima for the Mo-S coordination number were observed, a shift of the Mo-S shell (from

2.49 to 2.42 A) and of the Mo-Mo shell (from 2.79 to 3.19 A) are detected. The shift of the

Alo-S shell is minor but significant, whereas the change in the Mo-Mo distance is very

3.3

3.2

3.1

3

< 2.9 -

Sp 2.8

DO

aû 2.7

2.6

2.5

2.4

23

Mo-S distance

%âuA^)AÂj;i

No liganda NTA:Ni=1

+ NTA:Ni=1.5

o EDTA:Ni=1

x Calcined

â Ay ®>AAfHQA^y_AffiA^

100 150 200 250 300 350 400

Sulfidation temperature [°C|

Fig. 9. Mo-Mo and Mo-S distances obtained horn the QCNAFS spectia as a function of the sulfidation

temperature

145

Chapter 6

marked. Because of some experimental problems the spectra of the sulfidation of the

catalyst containing no ligands were not measured in the region between 270 and 350°C.

Nevertheless, it is visible that the sequence observed in Fig. 8 is kept here. In the catalyst

containing NEA with the molar ratio NTA:Ni = 1 the shift of the distances takes place at

the lowest temperature, whereas the catalyst prepared with EDTA is the last to undergo this

step.

In fact, the distances 2.42 Â for the Alo-S shell and 3.19 A for the Mo-Mo shell are

very close to the corresponding distances in AloS:. which are 2.42 and 3.16 A, respectively.

We could, therefore, find a precise method to detect when AloS: starts to be present in our

catalyst during the sulfidation process. The next step consists in understanding what

consequences this shift of the temperature of formation of MoS: has on the structure of the

final catalysts. Classical EXAFS data will allow us to investigate the structural features of

the sulfided catalysts.

As far as the beginning of the sulfidation is concerned no clear difference could be

noticed between the spectra of the different catalysts, fhe data presented in Chapter 4

evidenced that in the sulfidation of Alo in silica-supported NiAlo catalysts two intermediate

products were detected, consisting of an oxysulfide and of a M0S3 like material, which was

recognized from the signal at 2.5 A. On alumina, on the contrary, the MoSrlike material

starts to be formed at lower temperatures and there is no clear evidence of the presence of

the oxysulfide intermediate. However, its presence can not be excluded.

Mo JÇ-edgeEXAFS. For three catalysts, classical FXAFS spectra were collected

after 30 min sulfidation at 400°C. fhe quality of the data can be viewed in Fig. 10, where

the k -weighted %(k) data of the dried catalyst containing no ligand are plotted. The Fourier

transformed data are collected in I ig. 11, in which a comparison of the various spectra

evidences a clear difference in the amplitude of the Alo-Mo shells at 2.9 A (not phase

corrected). Ehe results of the fits, presented in "fable 2. revealed that the observed

difference is mainly due to dissimilar Debye Waller factors for the Mo-Mo shells. In the

catalyst prepared without ligands. the Debye Waller factor has the highest value, suggesting

a higher static disorder and a lower crystallinity of the AloS: particles. The presence of

NTA, on the contrary, seems to improve the order of the MoS: slabs. For the catalyst

146

IIDN-HDS ofNiMo/Al203 catalysis

containing NTA with the molar ratio NTA:Ni -1 the Debye Waller factor of the Mo-Mo

shell has the lowest value fhcre is a clear correlation between the value of the Debye

Waller factor and the sequence of the temperature of formation of M0S2 (see Fig. 8). The

lower is the formation temperature of MoS:. the higher is the static order in the final MoS:

22

-16

k | A"1]

Fig. 10. Mo A-edge k'-weighted fXAFS function of the dtied N1M0 catalyst containing no ligands

sulfided at 400°C

yj^

~no ligand

NTA-Ni=1

NTANi=1 5

R [A[

Fig. IE Mo A'-cdgc Foutiet tiansformed EXAFS spectra of the sulfided catalysts prepared in the presence

and absence ofNTA

147

Chapter 6

particles. The same trend was observed for the SKE-supported catalysts presented in

Chapter 4 and 5 but the effect on the Debye Waller factor was not so marked as here.

As far as the first shell is concerned, it is noteworthy that the Mo-S coordination

number of the catalyst made with NTA with the molar ratio NTA:Ni = 1 amounts to 6.0,

whereas for the other catalysts, and in general (see Chapter 5). values below 6 are obtained.

This observation suggests that all the edges of the MoS: crystallites are saturated with S.

Table 2. Parameters obtained from the fit of the spectra of the sulphided NiMo/AbCE

catalysts.

Catalyst Shell CN R A(T AE° Goodness

[À] [1(F4A2] [cV] of fit

No ligands Alo-S 5.8 2.42 5.0 3.1 0.79

Mo-Mo 3.6 3.16 8.2 3.7

N fA:Nr-l Alo-S 6.0 2.42 5.0 3.8 0.79

Alo-Mo 3.6 3.16 4.1 4.7

NEA:Ni=E5 Alo-S 5.9 2.42 4.6 4.1 0.60

Mo-Mo 3.7 3.16 5.4 4.8

Nickel

Ni A-edge QEXAFS. Ehe sulfidation of Ni in the catalysts was investigated by

means of QEXAFS. Fig. 12 shows the example of the sulfidation of Ni in the

NiMoNTA/AEO, catalyst (\TA:Ni - I) as followed by QEXAFS. Ehe spectrum of the

catalyst precursor (fresh) shows two main signals. Ehe first, at 1.8 A (phase uncorrected), is

produced by the Ni-0 shells, whereas the second one. at 2.2 A (phase uncorrected) is due to

the presence of the carbon atoms belonging to NEA around Ni [13. and Chapter 3]. A

careful observation of the first signal allows us to detect when the oxygen atoms around Ni

are replaced by sulfur. The Ni-S distance Is about 0.2-0.3 A larger than the Ni-0 distance.

148

1IDN-HDS ofNiMo/AfOi catalysts

22

400

390

365t/3

A!!->

330Pr-t-

o

2900r+

o

p

250pr-1-

C

205 CD

r~,

0

o

160 1 J

135

100

60

fresh

Fig. 12. Ni A-edge QFXA1 S spectra collected during the sulfidation of the NtMoNTA/AEO catalyst with

the molar tatio M 1A NiA

A more precise picture about the sulfidation of Ni in the different catalysts can be

achieved by monitoring the XANES spectra. As can be seen in Fig. 13 the shape of the

spectra changes during the sulfidation process. By means of a linear fitting it is possible to

estimate the fraction of sulfided Ni in the catalyst. Ehe spectra during sulfidation are

considered a linear combination of the spectrum of the fresh catalyst and ofthat collected

149

Chapter 6

after sulfidation at 400°C. The spectrum of the dried catalyst containing no ligands was

used as reference for the sulfided state for the calcined catalyst. The results of this

procedure are plotted in Tig 14. As far as the dried catalysts are concerned it is visible that

the presence ofNTA shifts the sulfidation of Ni at lower temperatures. No difference can

be noticed between the two NTA containing catalysts. On the contrary, EDTA delays the

sulfidation of Ni. Ihe behaviour of the calcined catalyst is different. In this catalyst even

8350 8400

X-ray energy [eV|

Fig. 13. Ni K-edge XANES spectia collected dining the sulfidation ot the dtied NiMo catahst containing

no ligands

150

8300

HDN-HDS ofN,Mo/Al203 catalysts

though the initial sulfidation rate is fast, the fraction of sulfided Ni after sulfidation at

400°C for 30 min is only about 70%. This is due to the fact that under the calcination

condition part of Ni was incorporated in the support. This fraction of nickel was not

sulfided. fhe EXAFS signal is therefore, an average of the sulfided Ni present on the

surface of the support, and of the unsulfided Ni segregated in y-AEOv

0 100 200 300 400

Sulfidation temperature [°C]

Fig. 14 Degtee of sulfidation ofNi on the alumina-suppoited catahsts as obtained from the XANES spectia

Ni Ä-edge EXAFS. Ihe k1-\\eighted x(k) data of the catalyst containing no

ligands plotted in Tig. 15 allow to have an idea about the quality of the Ni X-edge data

obtained from our measurements. Ehe data were Fourier transformed using the k-range 3-

16 A"1. Ehe spectra of the Fourier transformed data are plotted in Fig. 16. They are

composed of three main signals. According to the interpretation proposed in Chapter 5 all

three signals are produced by Ni-S contributions. A comparison between the k1- and k'-

wcighted spectra with a Ni-S and a Ni-Alo signal simulated with the Feff code clearly

ISj

Chaptei 6

22"

k[\']

Fig. IS. Ni A-cdge 12-weighted FXAFS function of the di tied catalyst containing no ligands

11

55

no ligand

NTA:Ni=1

NTANi=1.5

22><;

>—i

(A

-5 5

-11

0 £2S2^

R[Ä]

Flg. 16. Ni A-edge Fouiiet tiansfoimed EXAf S spectia of the sulfided catahsts ptepaied m the piesence and

absence ofNIA

H2

HDN-HDS of NiMo/AhOi catalysts

Table 3. Structural parameters obtained from the fits of the Ni A-edge EXAFS

spectra of the alumina supported catalysts prepared in the presence and absence ofNTA

Catalyst Shell CN R Ac/ AE° Goodness

[Â] [io3Â2i [eV] of fit

No ligands Ni-S 3.4 2.20 1.24 1.4 0.17

Ni-S 1.2 2.74 2.08 -0.6

Ni-S 0,7 3.06 -0.19f\ /"¥

NTA:NEd Ni-S 3.5 2.19 1.05 2.4 0.20

Ni-S 1.2 2.74 3.21 -0.1

Ni-S 0.6 3.09 -3.53 -4.9

NTA:NiM.5 Ni-S 3.4 2.20 1.01 1.1 0.17

Ni-S 1.6 2.74 4.13 -0.1

Ni-S 0.9 3.05 0.24 0.1

confirmed that also the last two signals are Ni-S contributions, fhe results of the fits arc

given in Eable 3. fhe coordination number of the first shell has a relatively low value for

all catalysts, which suggests that some Ni could have a coordination number of the first

shell lower than 4.

To interpret the results concerning the second and third shell the structure proposed in

the previous chapter should be considered. In the trimer bis(dithiobenzoato) Ni(II) complex

reported by Bonamico et at the coordination number of the second and third Ni-S shell

should amount to 0.66 because there are two Ni-S bridges with a ditance 2.78 A and two

with a distance 3.11 A (Chapter 5. Fig. 14) [26). In the present data of the alumina-

supported catalysts, on the contrary, the coordination number of the Ni-S shell at 2.74 A is

1.2 for the first two cataly sts and 1.6 for the last one.

In our alumina-supported catalysts the Ni loading is relatively high (2.5%) in

comparison to the A1o loading (7°o). fhe resulting NEAlo molar ratio is about 0.6. The

higher Ni loading was chosen in order to improve the quality of the XAFS data. On the

other hand, at this high concentration only a fraction of Ni is present in the catalyst as

153

Chapter 6

active phase. The remaining nickel is disturbing the EXAFS signal because what is visible

in the spectra is an average of all Ni species present in the catalyst. According to the data

presented in the previous chapter, in which the Ni loading was lower (1.3%), the active

species should consist of small Ni clusters, fhe inactive phase could consist of larger

clusters in which Ni has a similar geometry. Ehe presence of this compound is suggested by

the coordination number of the first Ni-S shell close to 4 and would explain the high

coordination number obtained for the second and third Ni-S shell.

Ehe third Ni-S shell belongs to the Ni atoms sandwiched between two other

monomers. Assuming that the clusters can be composed of more than three units, the

coordination number of the last Ni-S shell could increase and reach 2 for extremely large

clusters. In our catalysts we obtained values ranging from 0.7 to 0.9 suggesting that the

clusters arc composed of a relatively small number of units. In the spectrum of the catalyst

containing NEA with the molar ratio NTA:Ni = 1 the third shell is expanded towards larger

distances. This feature is explained by the large Debye Waller factor of the shell. The Ni-S

bond length distribution must be relatively large.

The presence of small Ni clusters would involve the detection of a Ni-Ni shell at

around 3.3 Â (phase uncorrected) by means ol'EXAFS. A weak signal at this distance can

be observed for the catalyst containing NEA with the molar ratio NEA:Ni - 1 but not for

the other two materials. The absence of a Ni-Ni shell can, however, be explained by a large

bond length distribution which induces the disappearance of the signal in the EXAFS

spectra.

154

HDN-1IDS of NiAh/ihO, catalysts

Discussion

In this section we will first consider the mechanism of sulfidation of Alo and the way

it is affected by the ligands. Ehen we will concentrate on the sulfidation of Ni. In both cases

we will explain how the structural features influence the catalytic performance.

Molybdenum

fhe use of QEXAFS enabled us to distinguish three regions in the sulfidation

mechanism of Mo: the oxidic state, an intermediate MoSs-like material and the final MoS:.

Ehis sequence differs from the one observed for SiO:-supported catalysts discussed in

Cahpter 4. On silica two intermediate regions were detected, a Mo-oxysulfide and the

MoSs-like product. On alumina the oxysulfide intermediate can not be clearly

distinguished. Because of the high pH of the impregnation solution (9.5) Alo is present on

the AEOj-supported catalyst precursors mainly as AI0O42" units, even though a fraction of it

is present as polyanions as the signal at 3 A suggests. On the contrary, polymolybdates are

the main species on SiO: [E3]. The smaller molybdate molecules are easier to sulfide

because no Mo-O-Alo bridging bonds have to be cleaved. Therefore, the passage from

MoOf" to the MoS^-like material proceeds in a much faster way and can not be detected by

QEXAFS.

In addition, the behaviour of the Mo-S and Mo-Mo shells evidences the passage from

the intermediate product to AloS:. Fhis feature is very important because it allows us to

detect precisely when AloS: starts to he formed and the temperature of formation is not

assigned only by an optical analysis but also by monitoring the distance of the two shells.

Ehe comparison between the formation of AloS: in the various catalysts showed that the

presence of N fA and ETEfA has opposite effects on the sulfidation of Mo (Figs. 8 and 9).

In catalysts prepared with NFA AloS: is formed at lower temperatures, whereas EDTA

retards the formation of AloS:. At the concentrations NEA:Ni = 1 and EDTAfiNi = 1 Mo is

not complexed by the ligands. because they preferentially form complexes with Ni.

Therefore, the shift in the formation temperature of MoS: must be rather due to an

interaction of Alo with Ni or with the support, fhe sequence for the sulfidation of Ni is, in

fact, the same as for the formation of AloS: as can be seen in Fig. 14. The kind of

155

Chapter 6

interactions affecting the sulfidation of Mo are difficult to predict and the data at our

disposal do not allow us to understand the chemical nature of this mutual influence.

Nevertheless, now we want to discuss which consequences a shift of the formation

temperature of Mo has on the final MoS: slabs, fhe answer to this point can be found in the

Mo A-edge EXAFS data. Ehe results of the fits showed that the catalyst in which AloS: is

formed at the lowest temperatures (NTA:NI - 1) is the one with the highest Alo-S

coordination number and the lowest Mo-Mo Debye Waller factor. The high coordination

number indicates a complete saturation of the MoS: edges, whereas the second parameter

points to a high order of the AloS: slabs. A large regularity of the Mo-Mo distances and a

narrow particle size distribution can induce a decrease in the Debye Waller factor.

A comparison of the XAFS data with the results of the IIDN reaction (Fig. 3) shows a

clear correlation between the order of the AloS: particles and the HDN catalytic

performance. The catalyst NiAloNTA/AfO, with the molar ratio NEA:Ni ~ 1 is the most

active. The inverse correlation is valid for the cyclohexene hydrogénation. The explanation

oE these activity profiles must be searched in the mechanism of the HDN reaction (Fig. 1).

The rate-determining step of the HDN of o-toluidine is the hvdrogenation of the phenyl ring

which leads to the formation of AICHA. A1CEIA then reacts quickly to MCIIE and MCH.

EEydtogenation of olefins (CHE) over sulfided Alo catalysts takes place relatively easily,

occurring already at 1 atni pressure of by drogen. Ehe reactivity generally decreases with

increasing olefin chain length and number of substituent groups adjacent to the double bond

[27 [. Contrary to olefins, hvdrogenation of aromatics requires high pressures of hydrogen to

effect the saturation. Ehis is partially due to the low reactivity of the aromatic structure

induced by resonance stabilization of the conjugated system. Ehe highest conversion of

EOE is obtained with the catalyst which is least active in the hvdrogenation of CHE. Ehe

same trend is shown by most of the tested catalysts. Ehis observation indicates that

different sites are Involved in the two reactions. Catalysts active for olefin hydrogénation

are not always capable of catalyzing aromatic hvdrogenation [28]. Moreover, benzene

(aromatic) and cyclohexene (olefin) hvdrogenation occur at different sites on a sulfided

catalyst [29]. ITowever. the dissimilar activities for the two reactions may be due to

different degrees of adsorption for the two compounds. The CHE can form a cr-bond with

Mo and Ni, while the aromatic ring is more likelv to be rt-bonded to the metal sulfides and.

156

HDN-jHDS ofNtAlo/Al203 catalysts

thus, less strongly adsorbed. TOL should adsorb parallel to the edge plane of MoS:, in

p6-adsorption mode. Considering the size of the aromatic ring, such adsorption needs more

than one Mo site [30]. A more crystalline structure of the active phase should favour the

planar adsorption and thus the hydrogénation of toluidine. The EXAFS results show that

the catalyst with the highest crystallinity is the one containing NTA with the molar ratio

NTA:Ni = 1 followed by the other NTA cataly sts. fhe dried catalyst containing no ligands

shows the most amorphous structure and, at the same time, the lowest HDN activity. 'Fhe

adsorption of the aromatic ring is enhanced by the regularity of the edges of MoS:. On the

contrary, the irregular Mo-Mo distances in the catalyst prepared without ligands make the

adsorption of the ring more unlikely.

This consideration does not affect the HDS of thiophene, because it is assumed that

thiophene adsorbs on the catalyst by means of the S atom present in the heterocyclic ring.

Therefore, as for the hydrogénation of cyclohexene. the regularity of the catalyst surface

does not play such an important role.

Nickel

From the Ni A'-edge QEXAFS measurements we could see that the sulfidation of Ni

in the various catalysts proceeds at different rates. In the catalyst containing NEA, Ni is

sulfided at much lower temperatures than in the other catalysts. From the QEXAFS spectra

of the fresh catalysts it is clearly visible that in the catalyst containing NTA and EDEA Ni

is complexed by the ligands because the Ni-C shell is detected at around 2.4 A (phase

uncorrected). In the catalyst prepared without ligands, on the contrary. Ni is believed to

interact with the support. Ehe difference in the sulfidation behaviour between the NTA- and

the EDEA- containing catalysts can be explained with the higher stability of the EDTA

complexes. ITowever. the sulfidation mechanism could be the same for both kinds of

catalysts, fhe most probable reaction pathway is a substitution of the keto group of the

acetate arms of the ligands by l ES. The resulting S" group would substitute for the acetate

oxygen group of the ligand coordinated by Ni. which would also be replaced by HAS. In

this way a thioacetate would be formed which could complex Ni and form a complex in

which Ni is situated in a plane formed by four S atoms as proposed in Chapter 5, either in a

square pyramidal or in a octahedral geometry, fhis reaction can take place only in the

157

Chapter 6

presence of the chelating ligands, and it is likely enhanced in catalysts where the sulfidation

of Ni occurs at higher temperatures. When Ni is sulfided at lower temperatures FES can

replace the ligand without reacting with it. For this reason we think that in the EDTA

containing catalyst the fraction of Ni which forms the thioacetate complex is higher than in

the other catalyst. The formation of this complex has the main result of achieving a higher

dispersion of Ni on the support. In the catalyst containing no ligands Ni is supposed to be

present in larger clusters. The fact that no Ni-Ni EXAFS signal can be detected by EXAFS

could be due to the fact that Ni-Ni shells with different distances are present, whose signals

interfere and cancel each other.

We believe that the formation of small Ni clusters has a higher importance in the

HDS reaction of thiophene and in the hydrogénation reaction of cyclohexene than in the

HDN of o-toluidine. The fact that a difference in IIDS activity is observed between the

catalysts containing NIA with the molar ratios NTA:Ni = 1 and 1.5 suggests that not only

the sulfidation temperature of Ni is an important factor for the formation of the thioacctate-

Ni complex. An increase of the amount of ligand in the catalyst precursor can also have a

beneficial influence on it. However, a too large amount of ligands would start to interact

with Mo and therefore have a negative effect on the HDS performance as the data of the

previous chapters showed.

Conclusions

The data presented in this chapter allowed us to extend the knowledge about the

effect of chelating ligands obtained from the study of SiO:-supported NiMo catalysts to

y-AEOrsupported catalysts. Aloreover. we compared the effect of the chelating agents on

two hydrotreating reactions, the hydrodenitrogenation and the hydrodesulfurisation.

Ehe IIDN catalytic performance is strongly enhanced by the use of NfA. Ehe effect

of this ligand is believed to be a decrease of the temperature of formation of MoS: during

the sulfidation process, which enables the development of more regular MoS: slabs, onto

which the hydrogénation of o-toluidine is enhanced.

On the contrary, we observed that, for the tested ligand concentrations, the

improvement in activity obtained in the HDS of thiophene is limited. Nevertheless, also for

158

HDN-HDS o/AWoAAA catalysts

this reaction a trend was observed in the performance of the catalysts. The improvement in

HDS activity was ascribed to the presence of NTA and EDEA. Ehe active phase for the

FIDS reaction is believed to be composed of a thioacetate complex, which enables the

formation of small Ni clusters. Its presence allows a better dispersion of Ni in the final

catalysts. Ehe high Ni loading did not allow us to detect this phase clearly by means of Ni

A-edge EXAFS.

'fhe different trend in the HDN and HDS reactions was explained with the fact that

for the HDN reaction the adsorption and hydrogénation of o-toluidine on MoS: is the rate

determining step, and this step is enhanced by the presence of more regular MoS:

crystallites. On the contrary, for the HDS of thiophene the presence of a higher Ni

dispersion plays a more important role.

Literature

1. van Veen, J. A. FE, Gerkema, E.. van der Kraan, A. M., and Knocster, A.,./

Chem Soc, Chem Commun. 1684(1987).

2. Thompson. Al. S., European Patent Application 0.181.035 (1986).

3. Clausen, B. S.. Eopsoe, H.. Canctia, R... Villadsen. E. Ecngeler, B., Als-

Niclsen. E. and Christenscn. F.. ,7 Pins Chem. 85, 3868 (1981).

4. Perot. G.. Catal Today 10, 447 (1991).

5. Satterficld. C. NA. and Yang. S. IE. hid Eng. Chem. Process Des Dev. 23,

11 (1984).

6. Rota, F.. and Prins. IE, Stud Surf Sei Catal 127, 319 (1999).

7. Rota. F.. and Prins. R., Topic m Catal 11/12, 327 (2000).

8. Rota, F.. and Prins. R.,./. Mol Catal (to be published).

9. Prins. R.. de Beer. V. EI. E. and Somorjai. G. A.. Catal Rev.-Sci. Eng. 31. 1

(1989).

10. Chianelli. R„ Daage. AE. and Eedoux. M. E. Adv. Catal. 40, 177 (1994).

11. Eopsoe, IE. Clausen. B. S.. and Alassoth, F. E., "Catalysis Science and

Technology". Springer Verlag, New York. (1996).

12. Startsev. A. N.. Catal, Rev -Sei Eng 37, 353 (1995).

159

Cattaneo, R., Shido, T., and Prins. R.. J. Catal. 185, 199 (1999).

Schulz. H.. Schon, M., and Rahman. N. AE. Stud. Surf. Sei Calai. 27, 201

(1986).

Hargreaves. A. E.. and Ross. J. R. H.. J. Catal. 56, 363 (1979).

McCarthy. K. F.. and Schrader. G. L.. ,7 Catal 117, 246 (1987).

Startev, A. N.. Burmistrov. Y. A., and Yermakov, Y. h., Appl. Catal. 45, 191

(1988).

Blake. M. R., Eyre, M., Moyes. R. B., and Wells, P. B., Bull. Soc. Belg. 90,

1293(1981).

Daly. F. P..,/ Catal. 51, 221 (1981).

Geneste. P.. Amblard, P.. Bonnet. AT, and Graffin, P., J. Catal 61, 1015

(1980).

Kampers. F. W. H., Maas, E. AE E. van Grondelle, E, Brinkgreve, P., and

Koningsberger, D. C, Rev Sei Instrum 60. 2635 (1989).

Vaarkamp. M., Dring, F, Oldman. R. E. Stern, E. A., and Koningsberger, D.

C.Phys Rev Z? 50, 7872 (1994).

Zabinsky, S. F. Rchr, J. E, Ankudinov. A.. Albers. R. C, and Eller, M. E. ,7

Phys Rev B 52, 2995 (1995).

Ankudinov. A. E.. and Rehr. E E. Phys Rev. B56, R1712 (1997).

Cattaneo, R.. Weber. Th., Shido. f.. and Prins, R., J. Catal. 191. 225 (2000).

Bonamico, M.. Dessy. G., Fares. V.. and Scaramuzza, L.. ,7 Chem. Soc.

Dalton Tram .2250(1975).

Uchytil.E. and Jakubickova. E..J Catal 64. 143 (1980).

Krishnamurthy. S.. Panvelker. S., and Shah. Y. X.AlChEJ 27. 994 (1981).

Voorhoeve. R. E FF. and Stuiver. J. C. AE.,/ Catal 23, 228 (1971).

Drew, AE G. B., EAjmondson, S, E, Forsyth. G, A., Flobson, R. .T., and

Mitchell. P. C. IF. Catal Today 2. 663 (1988).

Chapter 7

Summary & Conclusions

Ehe aim of this dissertation was a structural characterisation of supported NiMo

hydrotreating catalysts during the whole preparation, in the absence and presence of

chelating ligands. In this final chapter, we want to give a brief overview of the obtained

results, starting from the structure of the catalyst precursors, considering then the

sulfidation process and the final catalysts.

The oxidic state

From the combined EXAFS. Raman. EV-Vis study of the Siü:-supported catalyst

precursors we can conclude that in the absence of ligands Ni interacts strongly with the

support, 'fhe chelating ligands. on the contrary, avoid the interaction between the NE'

ions

and the support and induce a dispersion of the Ni-chelate complex over the support.

As far as Mo is concerned, we could conclude that in the absence of chelating agents

Ato is predominantly present on the support as polyanion units. Ehese polymolybdates are

present even when low ligand concentrations are used. In fact, the first amounts of ligands

bind preferentially Ni and Alo does not perceive any effect of them. Only when all Ni is

complexed by the organic molecules. Alo starts to form complexes with the chelating

agents. However, the formation of Mo-chelates was observed only when using NTA and

Chapter 7

EDTA. With these two ligands polymolybdate anions were no longer present on the

support.

The sulfidation

With this background it is possible to understand the behaviour of Ni and Alo during

the sulfidation process. Ihe QEXAFS studies of Si02- and AEOrsupported catalysts

clearly showed that the temperature of sulfidation ofNi in the presence of chelating ligands

is different from that in the catalyst prepared without ligands. For silica-supported catalysts

the presence of NEA and EDEA induced a higher sulfidation temperature of Ni. EN also

has a delaying effect on the sulfidation of Ni. although this influence is less detectable by

means of QEXAFS. On alumina-supported catalysts NFA and EDEA have opposite effects

on the sulfidation of Ni. In comparison to the catalyst prepared without ligands, Ni is

sulfided at lower temperatures in NLA-containing catalysts, while in the presence of EDTA

it is sulfided at higher temperatures. Hie sulfidation temperature of Ni in catalysts

containing the same ligand amounts is similar on both supports, 'fhis suggests that in the

absence of ligands NI is sulfided at higher temperatures on AEO^ than on SKA.

Combining these considerations with IIDS activity data, it becomes clear that the

catalytic activity increases with increasing sulfidation temperature of Ni. Nevertheless, this

is not always valid. On AEOi the NTA-containing catalysts had an activity comparable to

and even higher than the dried catalyst containing no ligands. even though Ni was sulfided

at lower temperatures in the NTA-based catalysts. These considerations lead to the

conclusion that the presence of chelating ligands has a beneficial effect on the HDS activity

not only because of a delay in the sulfidation temperature of Ni. We think that the

retardation of the sulfidation of Ni is only a sign that indicates that the mechanism of

sulfidation of Ni changes when using chelating agents. Probably the ligands are directly

involved in the sulfidation mechanism of Ni. A study of the mechanism of sulfidation of

the ligands could clarify this question. A'T-NMR investigation of the system would be an

interesting way to shed light on this process. We think that the chelating property of the

ligands is essential for the formation of a more active structure of Ni. In fact, the formation

of thiocarbamate complexes my be aided by the fact that uncoordinated chelating arms can

162

Summary and Conclusions

be sulfided more easily than the carboxylic arms coordinated to NE Ehe formed

thiocarboxylic groups could then replace the other oxidic ligands coordinated to Ni because

of the high affinity between Ni and S.

By monitoring the sulfidation of the catalysts by QEXAFS. it has become clear that

also Mo perceives some effects from the presence of the chelating ligands during the

sulfidation. The sulfidation interval of Alo becomes wider with increasing amounts of NEA

and EN, i.e. the sulfidation starts at lower and ends at higher temperatures. Differently from

the other ligands, a relatively small amount of EDfA is needed to strongly broaden the

sulfidation range of Alo. However, these observations do not mean that the changes in the

sulfidation of Mo are due to a direct interaction between Mo and the chelating ligands. It is

much more likely that these changes are due to the formation of the Ni-chelates in the

catalyst precursors. As we saw in the previous section the ligands protect Ni from any

interaction with the surrounding environment. Fhis effect has clear consequences also on

Alo. Ehe sulfidation intervals of Mo show-n in Fig. 6 of Chapter 4 suggest that the more Ni

is complexed by the ligands. the more the sulfidation of Mo resembles that of the catalyst

containing only Mo (Alo'SiO:). The presence of the ligands enables an independent

sulfidation of Alo and Ni. This means that, in the absence of ligands, the sulfidation of Mo

is affected by the presence of Ni. We can not say with certainty whether this influence

arises from a direct contact between the two metal ions or from the interactions ofNi or Mo

with the support. However, the explanation could be that the removal of Ni from the

support enhances stronger Mo-support interactions. This would explain why the sulfidation

of Mo takes such a large temperature range in the cataly st containing no Ni and in those

containing the chelating agents. Assuming that in these cases Alo interacts with the support,

the fraction of Mo exposed to the external atmosphere can be sulfided at lower

temperatures, whereas the part of Mo interacting with the support is only sulfided at

elevated temperatures.

In fact, there should be a correlation between the width of the sulfidation range of Mo

and the degree of removal of Ni from the support. The degree of detachment of Ni from the

support should be proportional to the number of coordination sites occupied by the ligands.

For this reason an EDTA:Ni molar ratio of only 1 is needed to make the sulfidation range

of Mo as wide as the Alo/SiO: cataly st. ED fA has six ligand coordination positions, so that

163

Chapter 7

only one molecule is needed to occupy all 6 coordination sites of Ni~A On the contrary, the

ENENi ratio must be larger than 3 because FN has only 2 binding amino groups so that the

chelate [Ni(ENL,]~ must be formed to avoid contact between Ni and the support. The

sulfidation range of Mo becomes wider with increasing NTA concentration, which suggests

that more than one NIA molecule is needed to completely detach Ni from SKA.

The sulfidic state

Hie relatively high quality of the Ni A-edge EXAFS data allowed us to shed some

more light on the structure of Ni in the sulfided catalysts. We could determine that the

second and third shells visible in the Ni A-edge spectra are two Ni-S shells and not Ni-Ni or

Ni-Mo shells as was erroneously interpreted in the past. Ehis observation has a great impact

on the clarification of the structure of Ni. It shows that the postulated proximity between Ni

and Mo in the sulfidic state can not be detected by FXAFS. The correspondence of the fits

with the structure of a trimeric Ni complex reported in the literature suggests that Ni is

present in the final cataly its in a layered structure.

The chelating ligands are supposed to favour the formation of smaller units and,

therefore, to increase the dispersion of Ni. However, the absence of a Ni-Ni shell from the

EXAFS spectra does not allow us to confirm this hypothesis.

We could observe that the structure of Alo in the final sulfided catalysts is influenced

by the sulfidation process, and. therefore, by the presence of the chelating ligands. Ehe

static order of the AloS: particles increases with increasing ligand concentrations. Ehis is

valid as long as only Ni is complexed by the chelating agents In the catalyst precursors.

Apparently, the presence of stronger Mo-support interactions in the catalyst precursor

favours the formation of more crystalline AloS: particles. However, for larger N'FA and

EDTA concentration the si/e of the MoS: cry stallites decreases. This is due to the fact that

at high ligand concentrations also Alo is complexed by NfA and EDEA in the catalyst

precursors, which induces a higher dispersion of Alo in the final catalysts. Ehe increase in

static order of MoS? has no apparent effect on the HDS activity but is extremely beneficial

in HDN. The decrease in particle size, on the contrary, has a negative influence on the FIDS

catalytic activity. The reason for this is still unexplained.

164


Remarks & suggestions

The variation of the ligand concentrations allowed us to obtain a detailed picture of

Ni and Mo during the whole preparation procedure. We tried to concentrate especially on

the study of Ni. The understanding of its structure has been generally considered somehow

of secondary importance because Ni is considered the promoter in this catalytic system.

Moreover, the low concentration of Ni in hydrotreating catalysts makes the understanding

of its structure even more difficult. Our results show, however, that more attention should

be devoted to the structure of Ni. Concerning the question whether Ni or Mo is the actual

catalyst, we have no clear answer. The presence of chelating ligands primary affects Ni and

has onlv an indirect influence on Mo. Therefore, we think that, indeed. Ni could be the

actual catalyst.

The QEXAFS technique was applied successfully to the study of hydrotreating

catalysts. Wc showed that this method can deliver a large amount of information and is

much more powerful than the temperature programmed sulfidation techniques used in the

past. Its applicability has been confirmed also in the study ofNiW/AfO-, catalysts in recent

experiments of our group (Ph. D. projects M. Sun. A. van der Vlies).

We think that the exact sulfidation mechanisms of Ni in the presence and absence of

chelating ligands could clarify the formation of the active sites in hydrotreating catalysts,

'fo this purpose fluorescence EXAFS should be used to characterize the final catalysts and

catalysts sulfided at intermediate temperatures. However, good data can be obtained from

fluorescence EXAFS only if an appropriate multi-element detector is used. Only a limited

number of such detectors is obtainable in the synchrotron facilities and they are partially

still under development. '"T-NA1R could he a successful tool to understand the mechanism

of sulfidation of the chelating ligands.

Thiocarboxylic ligands or thiocarbamate-Ni complexes should be tested for the

synthesis of hydrotreating catalysts. Catalysts could be prepared in which a trimeric Ni

complex similar to the one presented in Chapter 5, is deposited on the surface of sulfided

supported AloS:.

165

Chapter 7

166


Acknowledgements

This work would not have been possible without the help of many people that I

would like to sincerely thank.

In particular 1 am grateful to Prof. Prins for all his advises and his art of transforming

a rough manuscript into interesting text.

I am thankful to Prof. Wokaun for agreeing to be the co-examiner.

I am also in debt with Dr. Takafumi Shido. he introduced me to EXAFS spectroscopy

in a very kind and subtle way.

In spite of the hard work I had a really funny time with all the colleagues that shared

with me day- and night-shifts in Grenoble and Hamburg. Fhanks to all of you.

In addition I want to thank all the people that helped us solving many problems:

Mr. Urs Krebs of the workshop of the technical chemistry laboratory at ETFl for his nice

ideas about the EXAFS cells; Wouter van Beek. Elermann Emerich of the Swiss Norwegian

beam line at ESRF and Mathias Hermann of the RÖA10 II station at HASYLAB for their

technical help and their kindness; Larc Tröger of EIASYEAB for his patience and his

enthusiasm.

1 am grateful to Christoph Stinner and to Patrizia Fabrizioli for the Raman and

EW-Vis measurements. 1 appreciated a lot the scientific collaboration of Dr. Thomas Weber

and Fabio Rota.

1 considered myself very lucky to have met all the nice people that have worked in

the group of Prof. Prins in the past 4 years, fhe atmosphere in the group made the work

much more pleasant.

Thanks to the friends with whom I have spent v ery funny lunches and evenings.

Thanks to Alonica and to my family for their proximity.

Thanks to mv bike.

167

Chapter 7

\

168


Curriculum Vitae

Riccardo Cattaneo

Date of birth: 25 November 1971

Nationality: Italian

Education

1977- 1982:

1982- 1986:

1986-1990:

1990- 1995:

1995

1996- 2000

Primary school, Porlezza (CO) - Italy

Secondary school, Lugano - TI

High school in Lugano - TI: mattira type B

ETFI Zürich, Diploma in Chemical Engineering

Diploma work in Birmigham (EJK) under the supervision of

Prof. Bourne in the field of surface science

Ph. D. Thesis in the Taboratorv of Technical Chemistry at ETH

Zürich in the group of Prof. Prins

169

why do chelating ligands improve the activity of nimo hydrotreating

Documents