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

Подождите немного. Документ загружается.

15.4 Chromia 599

or pre - graphitic paracrystals [32] The shoulder at 1550 cm

− 1

is characteristic of

conjugated oleﬁ ns or polyenes.

Conﬁ rmation of this interpretation of the Raman spectrum came from analysis

of yellow oil produced during extended use of a catalyst. The analysis revealed the

presence of pyrene (C

), ﬂ uoranthene (C

), benzanthracene (C

) and

perylene (C

) among other compounds. The alumina support principally pro-

duced perylene in the absence of chromia but with chromia a wider range of

polynuclear aromatics were detected [26] .

Studies using isotopically labeled propane [30] have shown that the route to

the coke deposits during propane dehydrogenation goes through a C

species,

probably CH(a). However, before these species convert to coke there is a period

when they are labile and can be hydrogenated off the surface as methane [26, 30] .

This was shown using isotopically labeled propane [26] and by using a mixed

continuous/pulse reaction system where a chromia/alumina catalyst was reduced

and subjected to a continuous ﬂ ow of propane for 10 min at 873 K followed by two

pulses of propane in helium [20] . With both pulses the H : C ratio was ∼ 3.3 much

higher than the inlet 2.7; also the mass balance was > 100%. However the propene :

propane ratio was what would have been expected from a continuous ﬂ ow experi-

ment after 10 min. Clearly carbonaceous material retained by the catalyst was

being desorbed into the pulse as methane. Therefore under continuous ﬂ ow condi-

tions there is a reservoir of C

fragments on the surface of the catalyst that is con-

tinually in ﬂ ux between removal as methane and consolidation on the surface to

generate higher molecular weight species. A study by Krause and coworkers [33]

showed that if the chromia is not pre - reduced the reduction of chromates by

propane may proceed through the formation of isopropoxide species, which react

Figure 15.2 UV Raman spectrum of coke deposited on 1%

Cr/Al

catalyst used for propane dehydrogenation.

600 15 Alkane Dehydrogenation over Vanadium and Chromium Oxides

further ﬁ rst to adsorbed acetone and then to formates and acetates. They also

found that dehydrogenation of propane resulted in sequential formation of ali-

phatic, unsaturated/aromatic and graphite - like hydrocarbon deposits, which deac-

tivated the catalyst.

The situation with butane is similar but with signiﬁ cant differences. The forma-

tion of coke on the surface is very similar to that found with propane; however,

Edussuriya and coworkers (unpublished results) found that although the majority

of the catalyst deactivation was due to polynuclear aromatics or other forms of

coke strongly adsorbed reaction intermediates were also detected. These species

can be desorbed at room temperature by treating with a low concentration O

Ar mix. Figure 15.3 shows the desorption of butane. The catalyst had been catalyz-

ing the dehydrogenation reaction for 2 h at 873 K when the gas stream was switched

from butane to argon. Online mass spectrometry indicated that all hydrocarbon

components were rapidly swept out of the reactor at 873 K and no other gases apart

from the argon carrier were detected during the cool - down period. However, when

2% O

/Ar was passed over the sample at room temperature, butane, butenes and

butadiene were evolved. This desorption process can be viewed as an oxidative

displacement, resulting in recombination of reaction intermediates (hydrogen and

alkyl, alkenyl and alkadienyl) from the surface of reduced CrO

species. Removal

of these species did not regenerate the catalyst. The polynuclear aromatics were

only removed when the catalyst was subjected to a full regeneration involving

high - temperature oxidation. Only after this treatment did the catalyst regain its

activity.

Figure 15.3 Desorption of butane and butene at room

temperature in a ﬂ ow of 2% O

/Ar from a 6% chromia/

alumina catalyst after butane dehydrogenation at 873 K.

15.5

Vanadia

Although vanadia catalysts have not been used commercially there is a signiﬁ cant

literature concerning the dehydrogenation reaction over supported vanadia

systems. The structure of supported vanadia catalysts under redox reaction condi-

tions is directly related to the catalytic performance. Vanadia catalysts are usually

reduced to some extent during redox reactions, and the reduced vanadia species

have been proposed as the active sites [20, 34, 35] . Therefore, information on the

valence state and molecular structure of the reduced vanadia catalysts is of great

interest. A number of techniques have been applied to investigate the reduction

of supported vanadia catalysts, such as TPR [36 – 38] , X - ray photoelectron spectros-

copy (XPS) [35] , electron spin resonance ( ESR ) [39] , UV - Vis DRS [40 – 45] , X - ray

absorption ﬁ ne structure spectroscopy ( XAFS ) [46] and Raman spectroscopy [37,

47 – 50] . Most of these techniques give information only on the oxidation state of

vanadia species. Although Raman spectroscopy is a powerful tool for character-

ization of the molecular structure of supported vanadia [34, 42, 51] , it has been

very difﬁ cult to detect reduced supported vanadia species with conventional

(visible) Raman measurements [37, 47 – 52] . A widely accepted explanation for this

phenomenon is that the Raman cross - section of reduced vanadium oxide species

is very small or near zero [53] . Thus, it remains challenging to use Raman spec-

troscopy for obtaining information on the molecular state of reduced vanadia

species.

It is notable that most Raman studies of supported VO

catalysts were carried

out using a single excitation wavelength in the visible region (488, 514 or 532 nm)

[34, 46, 47, 54] . However, several recent investigations on supported transition

metal oxides [55 – 58] , including vanadium oxides [55, 56] under ambient condi-

tions, using both UV and visible wavelength Raman excitations suggest that more

complete and sometimes new structural information of supported metal oxides

can be achieved by using multiple excitation wavelengths. The reason lies in the

strong electronic absorptions in the UV and visible wavelength regions exhibited

by most transition metal oxides, which make it possible to measure resonance -

enhanced Raman spectra. Under circumstances where supported VO

species are

present in a distribution of cluster sizes or coordination geometries, it is likely that

these species also possess a corresponding distribution of electronic absorption

wavelengths. Excitation of Raman spectra within the absorption region will produce

resonance - enhanced spectra from the subset of VO

species with absorptions at

the excitation wavelength. By measuring the Raman spectra at several wavelengths,

more information can be obtained about the various VO

species in the distribu-

tion. Moreover, when UV excitation is employed, even Raman spectra from sup-

ported VO

at low loadings ( < 1 wt%) on oxides having strong ﬂ uorescence are

possible because of the avoidance of ﬂ uorescence and enhanced sensitivity [59,

60] . In addition, the decreased self - absorption effects in the UV region indicated

by in situ UV - Vis DRS studies of reduced VO

and CrO

[40 – 44, 61, 62] suggests

15.5 Vanadia 601

602 15 Alkane Dehydrogenation over Vanadium and Chromium Oxides

that UV Raman spectroscopy may be capable of detecting reduced supported metal

oxides.

Studies on alumina supported catalysts with a range of vanadia weight loadings

revealed [63, 64] that isolated VO

species dominated at surface densities below

1 V/nm

; polyvanadates coexisted with monovanadates at surface densities between

1.2 – 4.4 V/nm

; and V

formed at a surface density higher than 4.4 V/nm

. The

vanadia densities of the catalysts were 1.1 V/nm

for the sample with the 1%

loading, 3.7 V/nm

for the 3.5% loading and 10.4 V/nm

for the 8% loading. Only

the XRD pattern for the catalyst with the 8% loading exhibited additional charac-

teristic peaks of crystalline V

at 2 θ values 26.3 and 34.6 [65, 66] . However,

analysis by

V NMR revealed V

on the 3.5% vanadium as well as on the 8%

(Gladden and McGregor, unpublished results).

Figure 15.4 a shows simple pulse - acquire

V MAS NMR spectra for catalysts

with vanadium loadings of 1, 3.5 and 8 wt%. This simple 1 - D spectroscopy identi-

ﬁ es a

V resonance only in the 3.5 and 8 wt% catalysts. The spectra are dominated

by spinning sidebands, with the isotropic chemical shift (denoted by * ) at − 612 ppm

being indicative of V

in a V

environment. The broad background upon which

these peaks are superimposed in the case of the 8 wt% sample is characteristic of

in more amorphous environments. The spectra clearly suggest that the popula-

tion of V

- like species in the 8 wt% catalyst is larger than in the 3.5 wt% catalyst,

consistent with the Raman data. Alongside characterizing the

V environment,

NMR spectroscopy of the

Al species is also of interest. Figure 15.5 a shows a

simple pulse - acquire

Al spectrum of an as - prepared 8 wt% catalyst. Two broad

resonances are observed; the peak at ∼ 6 ppm is characteristic of sixfold, octahe-

drally coordinated Al, while Al in four - fold and ﬁ ve - fold coordinated sites contrib-

ute to the broader resonance in the range 25 – 70 ppm. Interpretation of the broader

Figure 15.4 51 - V MAS NMR spectra. (a) As - prepared

catalysts, from top 8, 3.5 and 1 V; (b) 8 V after regeneration at

873 K. * indicates an isotropic resonance. In spectrum (b) the

peaks to negative intensity correspond to Al atoms in the

support.

resonance is made signiﬁ cantly easier – although the NMR technique itself is more

challenging – by using multiple quantum NMR spectroscopy. A triple - quantum

spectrum from the 8 wt% catalyst is shown in Figure 15.5 b. The projection of such

spectra in the F1 dimension yields the isotropic peaks, while the anisotropic line-

shapes (corresponding to a standard MAS spectrum) are displayed by the projec-

tion in the F2 dimension. The resolution of the isotropic resonance provided by

the triple - quantum MAS experiment unambiguously conﬁ rms the presence of

three distinct

Al environments, assigned to 4 - , 5 - and 6 - coordinate Al.

Studies reveal that reduction takes place over a broad range of temperatures and

the onset of reduction follows the weight loading with 8% < 3.5% < 1% but reduc-

tion is complete for each sample by ∼ 873 K. This difference in onset of reduction

may be attributed to different reducibilities of vanadia species, monovanadates,

polyvanadates and V

coexisting on the catalyst ’ s surface. The weight loss at the

reduction step increases with vanadia loading: when these weights are translated

into oxygen loss we see that there is a 1 : 1 relationship between vanadium atoms

and oxygen atom loss (Table 15.1 ). The same trend in mass loss is observed during

Figure 15.5

Al NMR spectra. (a)

Al MAS NMR spectrum of

as - prepared 8%V/alumina catalyst; (b)

Al 3Q MAS NMR

spectrum of as - prepared 8%V/alumina; (c)

Al 3Q - MAS NMR

spectrum of 8%V/alumina after regeneration at 873 K.

15.5 Vanadia 603

604 15 Alkane Dehydrogenation over Vanadium and Chromium Oxides

Tapered Element Oscillating Microbalance ( TEOM ) studies of catalyst reduction

(Gladden and McGregor, unpublished results). These studies also yield an average

1 : 1 relationship between vanadium atoms and oxygen atom loss.

This would suggest a change of 2 in the vanadium oxidation state, so if the

vanadium were in a +5 oxidation state before reduction it would be converted to

+3 after reduction. This was conﬁ rmed by in situ UV - Vis DRS. Before reduction

the catalysts were in a V

oxidation state with typical charge transfer bands at

267 nm for a 1% loading, 290 nm for a 3.5% loading and 357 nm for an 8% loading.

This change in position of the charge transfer band is indicative of the increasing

polymeric nature of the vanadia [67] . Over the course of the reduction the spectrum

changes to give spectra typical for a V

state with d – d transitions around 575 nm

and 625 nm for 3.5 %V and 8 %V. However, the spectrum of the 1% sample is

more indicative of V

, with a single band at 585 nm [31] .

V NMR spectroscopy

revealed a total loss in signal after reduction, indicating the presence of V(III) or

V(IV) species (Gladden and McGregor, unpublished results).

Using oxygen chemisorption at 293 K and 873 K the average oxidation state of

the 1% VO

/alumina was calculated as 3.8, whereas the average oxidation state for

the 3.5% VO

/alumina was calculated as 2.6 [31] . Both of these ﬁ gures are in excel-

lent agreement with the conclusions from the UV - Vis DRS.

Bulk re - oxidation data in conjunction with thermogravimetric analysis ( TGA )

and TEOM results indicated that the ease of replacing oxygen that had been

removed during reduction was 8% > 3.5% > 1% [31] , (Gladden and McGregor,

unpublished results). This difference was also seen in the UV - Vis DRS and in

TEOM data. TEOM data provide information on the quantity of oxygen taken up

by the catalysts and on their rate of re - oxidation. The rate of catalyst re - oxidation

was seen to be greater for the 8 wt% catalyst than for materials of lower vanadium

loading (Gladden and McGregor, unpublished results). Once reduced, the 3.5 %V

sample did not regain its original state even after high temperature oxidation;

however, the reduced 8 %V sample did convert to a species that was similar to the

original oxidation state.

Supported VO

catalysts show good catalytic performance in both oxidation and

reduction reactions; for example, they are among the most active and selective

simple metal oxides for dehydrogenation and oxidative dehydrogenation of

alkanes. The possibility of using oxide systems other than chromia is being

Table 15.1 Extent of oxygen removal during reduction of vanadia/alumina catalysts.

Catalyst (V/nm

) V atoms in 1 g of

catalyst ( × 10

)

O atoms removed from

1 g of catalyst ( × 10

)

V : O ratio

1.1 1.18 1.13 1.0 : 1

3.7 4.13 4.14 1.0 : 1

10.4 9.45 7.90 1.2 : 1

explored in an effort to make the alkane – alkene conversion process more selec-

tive and stable. Supported VO

catalysts have been extensively studied for the oxi-

dative dehydrogenation of light alkanes [40, 54, 68 – 72] . They have also been

explored in the dehydrogenation of light alkanes including propane and butane

[35, 39, 73, 74] and have shown promise to be both active and selective for the

dehydrogenation reactions. However deactivation of VO

catalysts by coke species

is also observed with time - on - stream in butane dehydrogenation reactions. Owing

to the limited number of studies of supported VO

for alkane dehydrogenation,

the mechanism of deactivation/coke deposition and also the structure of sup-

ported VO

species under dehydrogenation conditions are unclear. Conventional

visible Raman spectroscopy usually cannot provide meaningful spectra for a cata-

lytic system that produces carbonaceous deposits because of strong ﬂ uorescence

interference. These two issues were addressed in an investigation of butane dehy-

drogenation reaction on V/ θ - Al

catalysts with various surface VO

densities

(0.03 – 14.2 V/nm

) via in situ UV Raman spectroscopy [75] .

V/ θ - Al

catalysts with surface VO

density from 0.03 to 14.2 V/nm

were tested

for butane dehydrogenation under low partial pressures of butane at 873 K, and

the activity and selectivity data on these different catalysts were compared [75] . In

essence, the initial dehydrogenation activity increases as a function of surface VO

density. However, the activity decreases signiﬁ cantly with prolonged reaction time

for samples with VO

density higher than 1.2 V/nm

, apparently due to the forma-

tion of surface coke deposits and blocking of surface sites during the reaction.

Interestingly, the dehydrogenation activity remains nearly unchanged on the cata-

lysts with surface VO

density no higher than 1.2 V/nm

. However, at higher

butane pressures, although the behavior of the 1.2 V/nm

sample is unchanged,

catalysts with higher V/nm

densities show enhanced activity (Table 15.2 ).

In the in situ UV Raman study [75] , V/ θ - Al

samples (0.03 V, 1.2 V, 4.4 V and

14.2 V) that contain surface VO

species with different structures were treated in

dilute butane ﬂ ow at different temperatures and examined by Raman spectra

measurements at each temperature. Taking the 1.2 V sample as an example, the

Raman spectra collected during butane dehydrogenation at different temperatures

are shown in Figure 15.6 .

Table 15.2 Activities and turnover frequencies for butane dehydrogenation

at 1 bar pressure and 873 K with time - on - stream.

Time - on - stream

(min)

Rate ( µ mole g

− 1

) Turnover frequency (s

− 1

)

1.1 V/nm

3.7 V/nm

10.4 V/nm

1.1 V/nm

3.7 V/nm

10.4 V/nm

15 1.1 20.3 5.8 0.051 0.269 0.037

30 1.0 5.8 3.7 0.047 0.077 0.023

45 1.0 4.7 2.9 0.044 0.063 0.018

60 1.0 4.2 2.5 0.044 0.056 0.016

105 0.9 3.5 0.041 0.046 –

15.5 Vanadia 605

606 15 Alkane Dehydrogenation over Vanadium and Chromium Oxides

At temperatures below 673 K, a weak Raman band at 1620 cm

− 1

, assigned to C

stretching in polyalkenes, is observed together with the two bands at 1021 and

915 cm

− 1

due to the V

O and V

–

Al modes of surface VO

species, respectively.

After butane dehydrogenation at 673 and 773 K, an intense band at 1601 cm

− 1

due

to polyaromatic hydrocarbons develops. Simultaneously, bands at 1500, 1438,

1183, 1004 and 845 cm

− 1

are also observed. Meanwhile, Raman bands at 1021 and

915 cm

− 1

from surface VO

species are no longer observable, owing to a combina-

tion of the reduction of VO

species [43, 45, 47, 65, 71] and strong optical absorp-

tion by surface coke species. The band at 1500 cm

− 1

is usually assigned to conjugated

polyalkenes or cyclopentadienyl species [76, 77] . The band at 1438 cm

− 1

is due to

the bending mode of CH

/CH

or C

–

H in aromatic rings. C

–

H bending in aro-

matics usually appears near 1180 cm

− 1

. The two sharp bands at 1004 and 845 cm

− 1

are rarely reported for coke species, and thus one might ascribe them to surface

species. However, the absence of an isotope shift on an

O - exchanged V/ θ -

sample conﬁ rms that these bands are due to surface coke deposits. As the

dehydrogenation temperature is increased to 873 K, the intensity of the Raman

bands below 1500 cm

− 1

decreases signiﬁ cantly. Raman spectra obtained after the

reaction of butane with pre - reduced 1.2 %V (not shown) are very similar to those

on pre - oxidized 1.2 %V. This suggests that the initial valence state of surface VO

does not affect the nature of coke species and most likely the VO

species is at

least partially reduced under butane dehydrogenation conditions. This conclusion

can also be inferred from the similar activity/selectivity results from butane dehy-

drogenation tests on either oxidized or pre - reduced 1.2 %V and also the similar

UV - Visible spectra from V/ θ - Al

treated in either hydrogen or butane.

Figure 15.6 UV Raman spectra of butane dehydrogenation

over oxidized 1.2 V at different temperatures.

The set of Raman bands near 850, 1002, 1183, 1437, 1501 and 1603 cm

− 1

is also

observed to grow simultaneously on all other V/ θ - Al

samples when treated in

butane at high temperatures. To aid the assignment of these bands, the adsorption

and reaction of different C

oleﬁ ns were studied on the V/ θ - Al

samples via

Raman spectroscopy. It turned out that they all form similar coke species, indicat-

ing that they are precursors to coke species in butane dehydrogenation on V/ θ -

catalysts. It is interesting to ﬁ nd out that 1,3 - butadiene adsorption at room

temperature results in similar spectra to that from butane dehydrogenation on

1.2 V at 673 K. Considering the low temperature (298 K), a possible chemical

change for 1,3 - butadiene is its cyclization to form styrene on the surface. A com-

parison of the spectrum from styrene with that after butane dehydrogenation at

673 K on 1.2 %V (Figure 15.7 ) reveals a very similar set of bands. Also included is

the spectrum from a mixture of 1.2 %V with styrene heated at 393 K (Figure

15.7 c).

Styrene is known to polymerize easily upon heating. Comparison of spectrum

c to that reported for polystyrene [78] conﬁ rms the identity of spectrum c: the set

of Raman bands near 850, 1002, 1183, 1437, 1501 and 1603 cm

− 1

is due to polysty-

rene. Thus, it appears that polystyrene is a key intermediate in the coke formation

process during butane dehydrogenation on the 1.2 V catalyst. Further Raman

spectroscopy study of styrene on V/ θ - Al

samples heated to 873 K showed that

the characteristic Raman bands of polystyrene disappear, owing to its decomposi-

tion. This is consistent with catalytic studies of polystyrene degradation over solid

acid catalysts [79] . An interesting new observation is that although the Raman

Figure 15.7 Comparison of Raman spectra from (a) styrene,

(b) butane dehydrogenation over 1.2 V at 673 K and (c) a

mixture of 1.2 V and styrene heated at 393 K.

15.5 Vanadia 607

608 15 Alkane Dehydrogenation over Vanadium and Chromium Oxides

spectrum seems to indicate that the catalyst surface is free of coke species after

polystyrene decomposition, the follow - up TPO (Temperature Programmed Oxida-

tion) shows the evolution of a considerable amount of CO

. This suggests that the

absence of typical Raman bands (1000 – 1650 cm

− 1

range) due to coke species does

not necessarily mean the catalyst is coke free. Investigation of the catalytic surface

with multi - wavelength Raman spectroscopy should be helpful in determining the

chemical nature of this surface carbon. Microanalysis of a used 1.2 V catalyst for

carbon and hydrogen gave a C : H value of 0.9 as per polystyrene monomer in

agreement with the Raman spectra [31] .

One important general characteristic of coke is its topology, which can be

assessed by the intensity ratio of the band at around 1600 cm

− 1

(G band) to the

band at around 1400 cm

− 1

(D band) [76, 80] . UV Raman spectra from a series of

polyaromatic compounds [76, 81] show that the intensity in the spectral range

1600 – 1650 cm

− 1

is signiﬁ cantly higher than that in the region 1300 to 1450 cm

− 1

for coke species with a 2 - D, sheet - like topology. By contrast, the intensity is more

nearly equal in these two spectral regions for coke species with chain - like topolo-

gies. The topology of coke species formed from butane dehydrogenation on the

various V/ θ - Al

catalysts at 873 K is compared by plotting the intensity ratio of

to I

as a function of surface VO

density. It is shown that the coke species are

more 2 - D, sheet - like from butane dehydrogenation on V/ θ - Al

with high surface

density ( > 1.2 V/nm

) while more 1 - D - like on V/ θ - Al

with lower surface

density ( ≤ 1.2 V/nm

). The 2 - D coke species can, presumably, reorganize into

pre - graphitic entities that have been thought to be the kind of coke that causes the

deactivation of dehydrogenation catalysts [33, 82, 83] . A study of coke deposited

on a 1 wt% V/Al

catalyst, in which the deactivation was studied as a function

of time - on - stream ( TOS ) using

C CP MAS ( cross - polarization magic angle spin-

ning ) NMR reveals a signiﬁ cant increase in the aromatic peak at 130 ppm up to

3 h, then a gradual decrease in this peak [84] . The initial increase in the peak

intensity was explained by an initial coke build - up, whereas the decrease in inten-

sity was due to formation of very slowly relaxing, NMR “ invisible ” , highly polyaro-

matic coke, which formed under prolonged exposure to the reaction conditions.

This interpretation is consistent with the constant decrease in T

of adsorbed

pentane observed, which also suggests an increasing aromaticity of coke with

increasing TOS. Further, a monotonic decrease in pentane self - diffusion coefﬁ -

cient as a function of TOS was observed; these data are consistent with restriction

in molecular motion resulting from pore blockage due to coke deposition as deac-

tivation proceeds. However, in another study it was found that the majority of the

catalyst deactivation in the early stages of reaction was not due to polynuclear

aromatics or other forms of coke but to strongly adsorbed reaction intermediates

[31] . These species can be desorbed at room temperature by treating with a low

concentration O

in Ar mix. Figure 15.8 shows desorption of butane from a series

of VO

/alumina catalysts.

The catalyst had been running under butane for 2 h at 873 K when the gas stream

was switched from butane to argon. Online mass spectrometry indicated that all

hydrocarbon components were rapidly swept out of the reactor at 873 K and no