Jackson S.D., Hargreaves J.S.J. Metal Oxide Catalysis

264 6 Photoelectron Spectroscopy of Catalytic Oxide Materials

catalysts towards butene and butadiene increased slightly, in agreement with the

ﬁ ndings of Jackson and coworkers [109] with 1 bar n - butane dehydrogenation at

873 K over the 3.5% V/alumina catalyst. Furthermore, the same group found that

V

x

O

y

/alumina catalysts containing polyvanadates were more effective at dehydro-

genation than isolated species [97] .

An additional feature of the XPS analysis was the formation of surface carbon

during the reaction. Figure 6.4 B shows that during the reaction the gas - phase peak

shifts to higher binding energy, revealing a surface carbon species which increases

as the reaction time increases. The binding energy of the surface carbon species

( ∼ 284.2 eV) is typical of graphene or aromatic carbon [110] . The formation of a

surface carbon species correlates with the mass spectra; surface carbon appears at

the point of maximum in benzene formation. It then increases as benzene forma-

tion decreases. Hence, it appears that the deposition of carbon on the surface is

linked to the retention of benzene. The percentage of surface carbon did not reach

more than 3% of the total surface (XPS measured at a kinetic energy of 290 eV,

which corresponds to an escape depth of ∼ 1 nm owing to the “ universal curve ”

[111] ), during the sub - millibar in situ reaction.

Links between changes in activity and surface structure can be clearly observed

in Figure 6.5 . The oxidation state of vanadium ﬁ rst decreases before carbon begins

to form on the surface. The ﬁ gure shows that deactivation only becomes more

signiﬁ cant after formation of carbon, and the vanadium continues to be reduced.

Figure 6.5 Structure - activity link between growth of surface

carbon, formation of butadiene (BD) and formal oxidation

state of vanadium (reaction conditions: 0.4 mbar n - butane

and 723 K).

6.3 Case Studies 265

Therefore, it can be postulated that by increasing the number of reduced vanadium

sites (as observed by XPS), the number of adjacent sites available for strong hydro-

carbon chemisorption increase. Thus XPS indicates that multiple reduced sites

may be required for coke formation. Although reduced sites may be active for

dehydrogenation, the beneﬁ t of maintaining mixed oxidation states may be to

separate reduced vanadium sites. Literature suggests that the loading and the type

of site (e.g. isolated VO

4

units) inﬂ uences the reducibility of supported vanadium

catalysts [96, 103] . In situ Auger electron yield XAS at the carbon K - edge (not

shown) have a strong feature at ∼ 285 eV during the dehydrogenation reaction, due

to a π * resonance from unsaturated hydrocarbon(s), which agrees well with the C

1s binding energy. In the mechanism of dehydrogenation, hydrogen is initially

removed to form an alkene; if the reaction is selective it may stop here. In the case

that the catalyst is unselective, the alkene can be dehydrogenated further to form

multiple unsaturation and coke precursors. Studies of coke formation suggest

that the dehydrogenation process is followed ﬁ rst by polymerization, then by

aromatization of surface species. The formation of benzene can be explained

by hydrogenolysis of the secondary – secondary C

−

C bond of butane followed by

oligomerization to benzene.

In addition to our in situ study, catalysts deactivated at 100 mbar pressure were

examined after treatment in an adjoining high - pressure reaction cell ( n - butane,

723 K). The samples were transferred without exposure to air and, as expected, the

concentration of surface carbon was much greater ( ∼ 45%). Reduction of the V/Al

ratio suggests that either particle agglomeration or preferential deposition of

carbon on vanadium occurs, resulting in a reduction of the vanadium signal. It

appears that this process is irreversible, as in situ reactivation measurement in

oxygen (after two activation/reaction cycles) does not show the same composition

as the “ fresh ” sample.

Although carbon XAS is rarely applied in catalysis, several studies have looked

at the carbon K - edge of post - reaction carbonaceous species using electron energy

loss spectroscopy [91, 112] . This can contain valuable information about the

nature of carbon deposited on the catalyst during deactivation. A range of cata-

lysts (1 – 8% V/alumina) were examined after deactivation in 1 bar of n - butane

at 873 K. The main feature for all of the catalysts was a strong π * ( ∼ 285 eV) reso-

nance, which is due to the presence of unsaturated hydrocarbons. At higher

energies in the region associated with σ - related features, the spectra were rela-

tively featureless and suggested the presence of disordered carbon [113] .

Although the higher loading catalysts were similar, the 1% V/alumina catalyst

showed a shoulder on the low - energy side of the main feature which suggested

the formation of a styrene - like compound, in agreement with ﬁ ndings of Wu

and coworkers [99] .

By combining the techniques of XPS and XAS, insights into the electronic

structure of the dehydrogenation catalysts V

x

O

y

/alumina have been observed, and,

by correlation with activity data, a greater understanding of the dehydrogenation

mechanism and subsequent coke formation, was possible.

266 6 Photoelectron Spectroscopy of Catalytic Oxide Materials

6.3.2.2 Oxidative Dehydrogenation of n - Butane

As mentioned previously, the main advantages of oxidative dehydrogenation are

enhanced thermodynamics together with a reduction in coking. Oxidative dehy-

drogenation of alkanes has been investigated by a number of authors [93 – 96, 114,

115] . Oxidative dehydrogenation catalysts are normally based on mixed - metal

oxides, and vanadium oxide - based catalysts are among the most successful of those

considered. Thus, the V

x

O

y

/alumina catalysts used for in situ butane dehydrogena-

tion (Section 6.3.2.1 ) were also examined under the conditions of in situ oxidative

dehydrogenation.

The addition of oxygen (1 – 2 vol%) to an n - butane feed resulted in strong changes

in the vanadium 2p

3/2

spectra and to a lesser extent in the vanadium L

3

edge (not

shown). Figure 6.6 indicates that on addition of oxygen, a signiﬁ cant amount of

V

5+

species were maintained, compared with direct dehydrogenation conditions.

This is not completely unexpected, as in the absence of oxygen in the reaction

feed any oxygen removed from the catalyst during reaction is not replenished.

Unfortunately, keeping V

5+

as the main species was accompanied by an increase

in the formation of oxygenated side - products such as furan, dihydrofuran and

CO

x

, which have been observed during n - butane oxidation and to a lesser

extent in the initial stage of dehydrogenation [83] . However, surface carbon

was not detected and only the gas - phase peak from butane was observed in the C

1s region. This is likely due to the automatic removal of surface carbon species by

the oxygen in the feed (or prevention of irreversible adsorption), thus reducing

Figure 6.6 Comparison of V 2p

3/2

XPS of 3.5% V/alumina at

723 K in (a) 0.5 mbar oxygen, (b) 0.4 mbar 2% O

2

in n - butane,

(c) 0.4 mbar 1% O

2

in n - butane and (d) 0.4 mbar n - butane.

6.3 Case Studies 267

the need for catalyst regeneration; however, this is at the expense of reduced

selectivity.

Lopez - Nieto and coworkers [93, 95] proposed that tetrahedral V

5+

is the active

and selective site for oxidative dehydrogenation of short chain alkanes (C

2

−

C

4

)

and isolated sites prevent the occurrence of consecutive reactions, that is, to total

oxidation products. Others suggested that neither terminal V

=

O nor bridging

V

−

O

−

V bonds are essential to the catalytic activity of alumina supported vana-

dium oxide catalysts, but rather that V

−

O

−

Al bonds play a crucial role [96] . On

the other hand, Iglesia and coworkers [116] measured the extent of reduction of

active VO

x

surface on γ - Al

2

O

3

during oxidative dehydrogenation ( ODH ) of propane

by in situ UV - Vis spectroscopy, and observed a positive correlation with the reac-

tion rate. Transient observations during changes in C

3

H

8

and O

2

concentrations

clearly indicated that only a fraction of the prevalent reduced centers ( ∼ 30 – 40%)

are responsible for the catalytic turnover. Blasco and coworkers [117] identiﬁ ed V

4+

species in post - reaction (ODH of ethane and n - butane) catalysts, the contribution

of which increased with increased conversion of reactants. Additionally, the

highest formaldehyde signal was observed in temperature programmed desorp-

tion experiment from adsorbed methanol when the vanadyls were partially

removed from the V

2

O

5

surface [118] . Furthermore no formaldehyde formed from

perfectly vanadyl - terminated surface. Our results are in line with these latter ﬁ nd-

ings, and suggest that careful control of the surface oxidation state by addition of

oxidizing agents can be useful in stabilizing the active catalytic site as well as in

preventing coke deposition. Steps taken to reduce coke formation include control

of feed gas composition (for example inclusion of hydrogen, steam, carbon dioxide

or oxygen to prevent chemisorption of hydrocarbons [119] ) and addition of pro-

moters to modify the surface properties [119 – 123] . However, addition of oxidizing

gases should be minimized to prevent reduction in selectivity through formation

of side - products.

6.3.3

Selective Oxidation

Selective oxidation of hydrocarbons is of key importance in functionalization of

hydrocarbon molecules. It is always a multi - step process with consecutive abstrac-

tion of hydrogen and addition of oxygen atoms. The difﬁ culty of this reaction is,

undoubtedly, that the process should go through these many steps, but also should

stop at the desired product. Such requirements can be met by complex

mixed - metal oxides, and the XPS characterization of two selected examples is

brieﬂ y reviewed here.

6.3.3.1 n - Butane Selective Oxidation to Maleic Anhydride Over VPO

Maleic anhydride ( MA ) is an important intermediate in the production of chemi-

cals such as unsaturated polyester resins, lube oil additives, maleic copolymers

and others [124] . Currently, MA is predominantly produced by the oxidation of

butane, and all industrial catalysts for this reaction are based on vanadium phos-

268 6 Photoelectron Spectroscopy of Catalytic Oxide Materials

phorus oxide s ( VPO ). Different types of technology were developed (ﬁ xed, ﬂ uid-

ized or transported bed reactors), and an important aspect of the following

discussion is that the catalyst is able to operate in both oxidizing and reducing

environments.

Several review papers have been published on VPO materials, covering

various aspects, from preparation and characterization to the catalytic properties,

with focus on the reaction mechanism [125 – 129] . In brief, VPO as a complex

catalytic system can comprise many distinct and known phases ( α - , β - , γ - , δ -

VOPO

4

, (VO)

2

P

2

O

7

, VO(PO

3

)

2

, VPO

4

, etc.) but also less well characterized amor-

phous material. It is prepared by different (e.g. aqueous, alcoholic) routes with

the intermediate formation of VO(HPO

4

) · 0.5H

2

O, the hemihydrate, which is

the precursor of the active catalyst. Depending on the P : V ratio in the precipi-

tation and the conditions of the heat treatment of the precursor material, the

prepared “ non - equilibrated ” catalyst can comprise different phases with various

degrees of crystallinity and defect structure. However, after long - term operation

the catalysts prepared in the conventional VPD route exhibited high crystallinity

and the only phase identiﬁ ed by XRD was (VO)

2

P

2

O

7

(VPP) [128, 130 – 134] .

This led several research groups to conclude that (VO)

2

P

2

O

7

(i.e. V

4+

) is the

active material and that its (100) crystal plane (which is supposed to be the

most exposed) contains the active sites. Different preparation routes, however,

resulted in materials with non - uniform bulk structure and different Brunauer –

Emmett – Teller surface area, but comparable intrinsic activity [129, 135] . This

led Hutchings to conclude that the surfaces exposed on these different materials

must all be the same, even though their bulk structures are completely differ-

ent. Furthermore, the necessity of crystallinity becomes dubious considering

that the amorphous vanadium phosphate catalyst prepared by the same group

(Hutchings and coworkers) [136] in supercritical CO

2

conditions was more

active (MA formation rate per surface unit) than the comparable crystalline

VPO catalysts, although with worse selectivity. Previously, combined XRD,

31

P

NMR and Raman studies [137 – 139] indicated that XRD alone is not sufﬁ cient

for identifying the presence of minority phases, and that not only V

4+

but some

V

5+

species exist in almost all materials. Many studies [137, 140 – 144] provided

indirect evidence that vanadium in both oxidation states is needed for the whole

catalytic process. Furthermore, Coulston and coworkers [145] argued about the

central role of V

5+

, as its reduction showed a good quantitative correlation with

the MA formation on supported VPO. Thus there is no clear agreement in the

literature about the active sites of the reaction. Even the reaction mechanism,

the participation of lattice or various types of surface oxygen species (

O

2

−

,

O

2

2−

,

O

−

) and the role of phosphorus enrichment remain under discussion [126, 128,

146 – 150] .

Being relatively surface sensitive, XPS can contribute to an understanding of

this catalytic system, and has already done so. It gives valuable information on the

formal oxidation state and the stoichiometry of the constituents in the sampling

depth of the technique. However, XPS investigations reported in the literature are

6.3 Case Studies 269

not without problems. VPOs are insulators at room temperature and in vacuo

(under conventional XPS conditions); hence the emission of electrons from the

samples leaves the surface charged making the exact binding energy identiﬁ cation

difﬁ cult. Charge compensation using an electron ﬂ ood gun reduces charging

effects, and referencing the binding - energy scale to “ adventitious ” carbon contami-

nation is widely applied. In fact, in the reviewed research papers on VPO - related

materials, the vast majority calibrate the binding - energy scale according to the

observed C 1s energy. A few others use arbitrarily taken P 2p energies. As not the

same C 1s binding energy is used as reference (C 1s: 284.5 – 285 eV), and because

this peak can be relatively broad (because of charging or overlapping components),

the uncertainty of position will be high, and the reported E

B

s of V/P/O core levels,

even in very similar materials, strongly diverge (Table 6.3 ). A further complication

arises if the main VPO material is differentially charged compared to the carbon

taken as reference. This can frequently happen considering the different conduc-

tivity of carbon and the oxide material. Most VPOs have open valence orbitals with

V 3d electrons at ∼ 1 ± 0.2 eV E

B

(Figure 6.7 ), similar to binary vanadium oxides in

non - V

5+

states [63] . Since, except for charging, no signiﬁ cant spectral difference

between the insulating and metallic phases of VO

2

has been observed, a more

appropriate referencing of VPOs can be achieved, circumventing differential

charging, by means of the V 3d transition. In the reviewers ’ opinion, differential

charging is the main reason that most of the reported core level binding energies

in the VPO - related literature are markedly higher than for example, in the V

x

O

y

system. For example, the V 2p energy of binary oxides is found at 517.1 ± 0.1 eV

for V

2

O

5

and 516.0 ± 0.2 eV for VO

2

. On the contrary, most of the E

B

values of V

2p in VPOs corresponding to the V

4+

state are reported in the range 516.8 – 518.2 eV,

which is 1 – 2 eV higher than in the binary VO

2

. This binding - energy offset is

usually conserved in O 1s and P 2p core levels, as well. Fortunately, Garbassi and

coworkers [160] noticed that the binding - energy difference (O 1s – V 2p) can be

related to the oxidation state of vanadium. Later, Coulston [158] proposed a quan-

titative correlation between the average oxidation state of vanadium and the previ-

ously mentioned splitting of the transition centroids, according to the following

equation:

VOsVp

ox

=− −

()

[]

13 82 0 68 1 2 3 2..

(6.7)

This equation has been extensively accepted and applied, and thus comparison

of various VPO materials in different publications has been possible (see Table

6.3 ). Although calculating the mean oxidation state of vanadium is exceptionally

valuable (especially with spectra lacking resolution), it allows determination of the

fraction of different contributions if only two oxidation states are present in the

sample (i.e. it is not useful if V

5+

/ V

4+

/ V

3+

are simultaneously present). Further-

more, with a signiﬁ cantly high E

B

contribution in the O 1s line (from adsorbed

water, OH groups and carbon - related O) the equation will overestimate the degree

of reduction of the formal vanadium valence.

270 6 Photoelectron Spectroscopy of Catalytic Oxide Materials

Table 6.3 XPS binding energies of VPO and related materials according to the literature survey.

Reference E

B

V2p (3/2) E

B

O1s

DE

B

(O1s - V2p

3/2)

E

B

P2p E

B

reference V

5+

/V

4+

P/V Note

[137] 518 and 516.9 530.5 12.5 and 13.6 131.7 C1s: 284.5 0.21

Co:0.3; Fe:0.14

∼ 2%

VPO

Co - VPO

Fe - VPO

[151] 518 and 516.9 C1s: 284.5 15% V

5+

VPO

[138] 518 and 516.9 C1s: 284.5

V5+: 47 ↓ to 37%

VPOs with different

activation time on

stream

[134] 518.2 and 516.9

518.2 and 516.9

517

531.3 (532.6)

531.3 (532.5)

13.1 and 14.4

14.3

133.6

133.7

C1s: 284.9 50:50?

50:50?

Only V

4+

Calcined sample

Non - equilibrated VPO

Equilibrated VPO

[152] 532.35 135 P2p: 135 Na

4

P

2

O

7

· 10H

2

O

[153] 517.7

( ∼ 532.0) 533.51 ∼ 14.3

135 P2p: 135 (V

4+

) VO(H

2

PO

4

)

2

V3d: ∼ 3 eV!

[154] 518.25 532.62 14.37 135 P2p: 135 (V

4+

) VOHPO

4

* 0.5H

2

O

V3d: ∼ 3.5 eV!

C1s:285 eV diff.

charging

[155] 519.84 533.07 13.23 135 P2p: 135 (V

5+

) VOPO

4

· 2H

2

O

C1s:286.1 eV!

[156] 517.7 – 517.9 532.2 14.3 – 14.5 C1s: 284.6 2.8

2.6

Eq.VPO

Addition of Co(2+)

6.3 Case Studies 271

Reference E

B

V2p (3/2) E

B

O1s

D E

B

(O1s - V2p

3/2)

E

B

P2p E

B

reference V

5+

/V

4+

P/V Note

[140] 517.9 532.3 14.4 134.3 C1s: 284.6 1.8 (up to 7.8) (Ref compounds)

and VPOs

After H

2

or inert:

P/V ↑ ↑ ↑

a)

[157] 517.7 – 517.9 532.2 14.3 – 14.5 C1s: 284.6 2.8

2.6 – 3.5

Un - promoted VPO

Co - promoted VPO

Surface enrichment of

Co

[158]

∼ stoichiometry ∆ E

B

(O1s - V2p 3/2)

Formula

[141] 518

516.9

C1s: 284.5 30% – 70%

50% – 50%

1.6

1.55

VPO

Nb

5+

doped VPO

[142] 517.5 – 517.2

Fit: 518,516.9

∼ 531.4

13.9 – 14.2 C1s: 284.5 47% – 53%

37% – 63%

1.6 VPO (0.1 h TOS)

VPO (132 h TOS)

[159] 517.0

517.1

531.3

531.5

14.3

14.4

134.0

134.1

C1s: 284.6 1.4

2.4

Bulk VPO

VPO/ZrO

2

(H

3

PO

4

)

[160] 516.0 and 517.2 530.2 13.0 and14.2

∼ 133.3 as

overall max.

C1s: 284.6 1.2 - 2.3 Depending on the

nominal P/V, but

surface P enrichment

[161] 517.7/517.6

517.5/517.5

532.2/532.4

531.9/532.1

14.5/14.8

14.4/14.6

133.9/134.1

134/134

C1s: 284.8 1.23/1.54

1.6/1.82

VPO1/VPO2

VPBiO1/VPBiO2

[162] 518 532.3 14.3 133.9 – 134.3 C1s: 284.6 2.5 - 2.9

Refs: 1.2!!

b)

VPO, Mo - VPO

β - VOPO

4

and

(VO)

2

P

2

O

7

a After treatment with H

2

or inert, the P/V ratio strongly increased (up to 7.8).

b Reference VPO phases gave P/V ratio 1.2, while catalysts up to 2.9.

272 6 Photoelectron Spectroscopy of Catalytic Oxide Materials

Reference E

B

V2p (3/2) E

B

O1s

DE

B

(O1s - V2p

3/2)

E

B

P2p E

B

reference V

5+

/V

4+

P/V Note

[163] 517.3 – 517.1

517.1 – 517.4

531.4 – 531.6

532.6 – 9 (SiO2)

14.1 – 14.5 133.6 – 7

133.7 – 134

C1s: 284.6 VPO bulk

VPO/SBA - 15

[143]

518 and 516

C1s: 284.5 28/72 In

fuel rich (O2/C4:0.6)

[164] 516.8

517

C1s: 284.6 1.5

After r: 1.3

c)

5% VPO/MCM - 41

5% VPO after reaction

[165] 517.5 134.1 C1s: 285 VPO glasses

for cross section

[166] 516.6 531.2 14.6 C1s: 284.8 VOHPO

4

· 0.5H

2

O

[167] 518, 516.9

∼ 532 ∼ 14, 15.1

C1s: 284.5? 82/18 fresh VPO.

[79]

∼ 517.2 ∼ 531.3

14.1

14.4

C1s: 284.5 VPO

UHV effect on V

5+

[149] 517.4 – 517.1 531.6 – 531.2 14.5 – 13.9 133.6 – 133.9 C1s: 285 1.37 – 1.68 (Bi) - VPOs

[150] 517.6 – 517.8 532 – 532.6 133.7 – 134.1 C1s: 285 1.32 – 2.05 (Co) - VPOs

[168]

∼ 517.2

531.2 C1s: 284.6 1.3 – 1.4 (Cs) - VPOs

(Ref. VPO glasses)

[20] 530.3 – 531.7

530.4 – 531.9

132.3 – 133.8

132.8 – 134.5

C1s: 285 Alkali phosphate,

Pyrophosphate

∆ E

B

O - P ∼ const.

c The P/V ratio decreased after reaction to 1.3.

Table 6.3

Continued

6.3 Case Studies 273

In addition, the existence of different oxidation states can be determined by

curve ﬁ tting of the recorded spectrum. In the case of vanadium, however, the situ-

ation is a bit more complex, as ﬁ nal state effects can complicate the spectral

interpretation. As all the lower valence vanadium cations have open valence shells,

multiple splitting, which is typically not resolved in the V 2p region, will signiﬁ -

cantly broaden (sometimes asymmetrically) the lower valence components. There-

fore one has to use known, but widely different, FWHM values to ﬁ t the individual

components. Spectral resolution depends strongly on the instrumentation; hence

adequate ﬁ tting results can be obtained after carrying out XPS experiments on

known reference compounds, and applying the observed (or at least similar) spec-

tral function in the ﬁ tting procedure of VPOs. Unfortunately such data treatment

has almost never been carried out, and only arbitrarily taken reference positions

were ﬁ xed in ﬁ tting VPO spectra and forcing the algorithm to observe different

valences in an unresolved spectral envelope. Considering the strong scattering and

shift of the core level E

B

s (Table 6.3 ), the reliability of such data treatment is highly

questionable.

Another widely neglected fact in the VPO literature when interpreting XPS data

is that information arises from a certain depth. This, using Mg or Al K α excitation,

is as high as ∼ 3 nm, the technique being thus a priori not sensitive to the top

surface layer on which the catalytic reaction proceeds. Furthermore, the sample

can be inhomogeneous in the top few layers, allowing no conclusions to be drawn

as to whether a certain state coexists on the surface or not. This problem can be

circumvented by carrying out XPS investigations with a tunable X - ray source at a

synchrotron as detailed below. Figure 6.8 compares the in situ V 2p core levels of

two, approximately equally active, VPO materials (VPO

P4

, VPO

P9

: prepared in

Hutchings ’ laboratory [169] ), under fuel - lean n - butane oxidation conditions, at two

different excitation energies. The spectra were published in a preliminary form

[100] . Note that referencing the binding - energy scale to the V 3d valence transition

Figure 6.7 Top of the valence band of a VPO catalyst showing

the band gap transition of V 3d.

Jackson S.D., Hargreaves J.S.J. Metal Oxide Catalysis

Подождите немного. Документ загружается.