Operando spectroscopic investigation of supported metal oxide catalysts by combined time-resolved UV-VIS/Raman/on-line mass spectrometry

Article abstract of DOI:10.1039/b305786k

A new experimental set-up for characterizing supported metal oxide catalysts under working conditions was been developed. The set-up combines product analysis using on-line MS with catalyst characterization by time-resolved operando UV-visible and Raman spectroscopy. The potentials of this approach were illustrated by the propane dehydrogenation over a chromium oxide on alumina catalyst. Different states of the active metal oxide phase can be observed, as well as information on coke formation on the catalyst. These spectroscopic techniques gave valuable complimentary information on the changing catalyst performance during successive propane dehydrogenation operating cycles. They also provide complimentary pieces of information, which agreed well with each other.

Full text of DOI:10.1039/b305786k

View Article Online / Journal Homepage / Table of Contents for this issue
Operando spectroscopic investigation of supported metal oxide
catalysts by combined time-resolved UV-VIS/Raman/on-line
mass spectrometryy
T. A. (Xander) Nijhuis,* Stan J. Tinnemans, Tom Visser and Bert M. Weckhuysen
Department of Inorganic Chemistry and Catalysis, Debye Institute, Utrecht University,
Sorbonnelaan 16, 3508, TB Utrecht, The Netherlands. E-mail: x.nijhuis@chem.uu.nl
Fax: +31-30-2511027; Tel: +31-30-2537763
Received 22nd May 2003, Accepted 29th July 2003
First published as an Advance Article on the web 6th August 2003
A novel experimental set-up for characterizing supported metal oxide catalysts under working conditions has
been developed. The set-up combines product analysis using on-line mass spectrometry (MS) with catalyst
characterization by time-resolved operando UV-VIS and Raman spectroscopy. The potentials of this approach
are illustrated by the propane dehydrogenation over a chromium oxide on alumina catalyst. It is demonstrated
that these spectroscopic techniques give valuable complimentary information on the changing catalyst
performance during successive propane dehydrogenation operating cycles.
olefins are produced using catalytic dehydrogenation pro-
Introduction
cesses. In the propane dehydrogenation the catalyst deactivates
fast. Commonly, a regenerative procedure in air is, therefore,
used for the catalyst once every 20–30 min. After this regenera-
tion, the catalyst can be re-used for the dehydrogenation. This
study provides operando spectroscopic information on the
short- and long-term deactivation process of the catalyst.
The aim of this study is to demonstrate the advantage of the
multi-technique approach in catalytic testing and analysis. It
will be shown that the information collected has good poten-
tial to be used as a tool to more efficiently operate a propane
dehydrogenation process.
Probing events taking place on a supported metal oxide
catalyst under reaction conditions has attracted considerable
attention in recent years.1–3 It requires the development of
time-resolved characterization techniques and the construction
of in-situ spectroscopic-reaction cells, which allow the identi-
fication of active sites, reaction intermediates and/or coke
formation. A number of spectroscopic techniques—such as
UV-VIS, NMR, EPR, IR, Raman and EXAFS—have already
been adapted to study metal oxide catalysts in catalytic action,
each giving information about the catalyst structure.4 It is,
however, very advantageous to combine two or more time-
resolved spectroscopic techniques in one catalytic reactor
set-up because complementary information can be obtained
simultaneously for the same catalyst material under identical
catalytic conditions.
Experimental
The newly developed system (Fig. 1) consists of a 6 mm dia-
meter quartz reactor tube with a special square section in
which the tube wall is made of optical grade quartz windows.
Typically, 300 mg of catalyst is placed in the reactor as a
packed bed supported on a sintered glass grid just below the
optical windows. Catalyst extrudates were milled to a typical
particle size of 50 mm. The bed height is ca. 15 mm. The reactor
is placed vertically in the center of a 10 cm long tubular oven
block; a thermocouple is inserted in the catalyst bed to directly
monitor the catalyst temperature. The metal oven block has on
opposite sides two horizontal 8 mm holes directed at the cata-
lyst bed. One of the holes is used to aim an Avantes FCD-
UV400-G-01CHT UV-VIS reflection probe at the catalyst
bed. This high-temperature probe contains three 400 mm opti-
cal fibers closely packed in a triangle. One fiber is connected
to an Avantes AVS-SD2000 spectrometer and another fiber
is connected to an Avantes halogen–deuterium light source.
A Halon disk was used for white reference measurements for
the spectrophotometer. The UV-VIS spectrometer was used
with an integration time of 28 ms and 40 spectra were averaged
for signal noise reduction. The resulting time resolution
was, therefore, 1 s. The measuring range of the spectrophoto-
meter is 200–870 nm. However, effectively the system only
provided a usable signal to noise ratio above 250 nm, partly
In this work, we present a novel experimental set-up, which
combines two operando time-resolved spectroscopic techni-
ques, namely UV-VIS and Raman, directly coupled to a regu-
lar fixed-bed catalytic reactor via optical fiber technology. Fast
product analysis is performed by on-line mass spectrometry.
UV-VIS together with Raman spectroscopy is a very useful
combination of in-situ techniques. The first technique provides
electronic information on the catalyst system and the latter
vibrational information. Both techniques can be conveniently
applied to a catalytic reactor system, since they both are
non-contact measurements.
The potential of this experimental set-up has been explored
by studying the dehydrogenation of propane to propene over
supported chromium oxide, vanadium oxide and molybdenum
oxide catalysts. As an example, in this paper we will discuss
the results obtained for a 13 wt% Cr/Al2O3 alkane dehydro-
genation catalyst. This catalyst is similar to the one used in
industrial processes. The dehydrogenation of propane is an
industrially very important reaction with a relatively short
catalyst lifetime.5 Per year, close to 10 million metric tons of
y Presented at the International Congress on Operando Spectroscopy,
Lunteren, The Netherlands, March 2–6, 2003.
DOI: 10.1039/b305786k
Phys. Chem. Chem. Phys., 2003, 5, 4361–4365
This journal is # The Owner Societies 2003
4361

View Article Online
from room temperature to a catalyst temperature of 823 K in a
gas stream of 3 ml minꢁ1 of O2 in 12 ml minꢁ1 of He. After
heating, the reactor was operated isothermally in cycles of
alternating dehydrogenation for ca. 50 min with 2 ml propane
in 20 ml minꢁ1 of He and regenerating for ca. 50 min with 3 ml
minꢁ1 of O2 in 12 ml minꢁ1 of He. During the first 120 s of
each cycle UV-VIS measurements were performed with a 1 s
interval to track the rapid changes on the catalyst, after this
time measurements were carried out at a 1 min interval.
Raman measurements were performed at 5 min intervals. Dur-
ing the entire experiment the product gas composition was
monitored using the MS. The temperature of the catalyst
was monitored using a small thermocouple in the middle of
the bed.
Results and discussion
Fig. 2 shows the measured catalytic and spectroscopic data of
the 13 wt% Cr/Al2O3 catalyst. The catalytic activity data are
presented in Fig. 2a for three dehydrogenation–regeneration
cycles. It can be seen that catalyst performance does not
change from one cycle to the next one, but changes signifi-
cantly within one dehydrogenation cycle. The stable perfor-
mance of the catalyst over the different cycles is in agreement
with expectations, since a commercial catalyst is used in indus-
try for ca. two years before it is replaced. In that case a catalyst
will have undergone well over 10 000 cycles. Within one cycle,
it can be seen that the conversion to propene increases gradu-
ally with increasing time-on-stream and reaches a maximum of
55% after 28 min. Longer reaction times result in a gradual
decrease in propene formation.
Fig. 2b shows the fast changes taking place in the UV-VIS
spectra in the first 6 s of the dehydrogenation reaction. The
strong absorption in the 300–400 nm range is from the alumina
support. There is an isobestic point present at 565 nm, indica-
tive of the presence of at least two distinct chromium oxide
species. It can be seen that the Cr6+ CT bands [6] located at
380 and 450 nm decrease in intensity with increasing reaction
time. At the same time, the Cr3+ d–d transition at 630 nm
increases in intensity.6 This decrease in the absorption at 380
and 450 nm and the increase of the band at 630 nm are in
agreement with the reduction of CrO3 to Cr2O3 when the cat-
alyst is switched from an oxidizing to a reducing environment.
Furthermore, a shoulder at about 700–750 nm becomes visible.
This band can be attributed to the interaction of Cr3+ with
adsorbates.7 The sharp peak in the UV-VIS spectrum at 532
nm is due to interference by light of the Raman laser scattered
at the catalyst particles, while the spike at 656 nm is an artifact
of the UV-VIS photometer system (most intense light emitted
from the deuterium lamp).
Fig. 2c shows the UV-VIS spectrum changing over a longer
time (50 min) in the propane dehydrogenation cycle. It is clear
that no major peak intensity changes can be observed. On the
other hand, the overall absorbance of the sample increases and
this is related to the formation of coke. The Cr3+ d–d transi-
tion band at 630 nm becomes less pronounced over time. This
can also be explained by the coverage of the chromium oxide
by coke.
Fig. 1 Schematic presentation (A) and picture (B) of the reactor
set-up with combined operando UV-VIS/Raman spectroscopy.
caused by the low transmittance of the optical fibers below this
wavelength.
The second hole in the oven is used for focusing a Raman
laser at the catalyst. Raman spectra were obtained by a Kaiser
RXN spectrometer equipped with a 532 nm diode laser. A 5.500
non-contact objective is used for beam focusing and collection
of scattered radiation. 10 spectra were accumulated with a 3 s
exposure time. The resulting total spectral recording time was
180 s. The laser output power of 70 mW did not cause any
changes to the catalyst, since this energy input is negligible
compared to the heat input by the oven to maintain the
550 ꢀC reaction temperature. Furthermore, the high-loaded
chromium on relatively low surface area alumina catalyst is
very stable. Raman spectra measured during a period of over
1 h in which the catalyst was heated to the reaction tempera-
ture in oxygen/helium did not show any observable changes
in the recorded spectra confirming this stability.
The gas leaving the reactor is analyzed using a Pfeiffer
Omnistar quadrupole mass-spectrometer, to determine all mass
components in the gas from 14 to 44 amu. The scanning time
per mass was 50 ms, making the time resolution of the mass
spectrometer 7.5 s. By performing a deconvolution procedure
using the known fragmentation patterns of the reactants and
all possible products, it is possible to convert the mass signals
quantitatively to concentrations. For a more facile quantifica-
tion of the product gas analyses, a Varian CP-4900 micro-GC
system equipped with a Poraplot-Q and Molsieve 5A column
was added to the reactor system. This GC system is capable
of performing analyses at less than 60 s intervals.
Coke formation is evident from the Raman spectra in Fig.
2d. The bands that appear in the region 1200–1650 cmꢁ1 are
attributed to (poly-)aromatic ring stretching C=C vibrations.8
Fluorescence of the catalyst, causing the increasing baseline at
high wavenumbers, decreases during the dehydrogenation
cycle. It is likely that this phenomenon is also due to the coke
formation on the catalyst. The bands at 200, 605, 810 and 1060
cmꢁ1, which can be observed in the Raman spectra, are origi-
nating from the quartz windows of the reactor.
The reactor system has been used to monitor the changes in
a catalytic solid during a number of dehydrogenation–regen-
eration cycles of an industrial-like 13 wt% Cr-on-alumina cat-
alyst for the dehydrogenation of propane. In these catalytic
experiments, at start-up the reactor was heated at 10 K minꢁ1
Another interesting Raman band could be observed at 542
cmꢁ1. This Raman band is assigned to the presence of
4362
Phys. Chem. Chem. Phys., 2003, 5, 4361–4365

View Article Online
Fig. 2 Propane dehydrogenation using a 13 wt% Cr-on-alumina catalyst [1 bar(a), 823 K]. (A) Catalytic activity measurements for three dehy-
drogenation–regeneration cycles; (B) operando UV-VIS spectra showing the rapid changes during the first 6 s of the first dehydrogenation cycle; (C)
operando UV-VIS spectra of the first dehydrogenation cycle as a function of time-on-stream after the first 6 s; and (D) operando Raman spectra of
the first dehydrogenation cycle as a function of time-on-stream.
Cr2O36,9 and gradually decreases with increasing reaction time.
Since Cr3+ is unlikely to be reduced further at the applied reac-
tion conditions, this disappearance is most probably associated
with the formation of coke covering the Cr2O3 surface as it
limits Raman scattering of the underlying layer.
the propane dehydrogenation cycle is caused by the reduction
of the Cr(VI) to Cr(III) by propane. In industry, this loss of pro-
pane is often circumvented by a short catalyst reduction using
hydrogen immediately before switching to propane.
Both the UV-VIS and Raman spectra at the end of the
regenerative cycle are identical to those at the beginning of
the dehydrogenation cycle. This is in good agreement with the
activity data, which shows an identical performance of the
catalyst in subsequent cycles.
The coke formation, as observed by both Raman and UV-
VIS, is the most likely cause of the decrease in catalyst activity
over time. In our experiments we observe a maximum in the
catalytic activity after ca. 28 min. Looking at the overall cata-
lyst performance with an increasing activity for 28 min and
then a slow decrease in activity, this might seem odd consider-
ing the typical industrial cycle time of 30 min. However, an
industrial feed will contain undiluted alkanes, while in our
experiments only a 9% alkane in helium stream was used. Hav-
ing a pure alkane feed will drastically decrease the timescale in
which changes on the catalyst are taking place (oxidation state
and coke formation) compared to our observations. Further-
more, a more frequent regeneration is beneficial since the heat
liberated by the coke combustion is used for heating up the
catalyst for the next endothermal dehydrogenation cycle.
In the oxidation cycles this coke is combusted from the cat-
alyst as can be concluded from the formation of CO2 in the
beginning of the regeneration cycle and from the increase in
the catalyst bed temperature of about 10–15 K. During the cat-
alyst regeneration the UV-VIS spectra return completely to the
initial state of Fig. 2b in less than 15 s. Both coke combustion
and (partial) chromium oxide reoxidation take place simulta-
neously. This rapid change in the observed catalyst state is in
agreement with the fact that the mass spectrometer detects a
short carbon dioxide production immediately after the start
of the oxidation cycle. Also, the increased catalyst temperature
resulting from the heat released by the coke combustion is
observed only for a short time. The smaller CO2 production
peak observed by the mass spectrometer at the beginning of
In Fig. 3, all relevant signals measured using the different
methods are summarized in a single figure for one complete
dehydrogenation–catalyst regeneration cycle. It can be seen
that there is a good agreement for the observed coke formation
between the UV-VIS and Raman measurements. The UV-VIS
signal at 850 nm (indicative of the increasing overall absor-
bance of the sample due to coke formation) increases similarly
fast as the Raman signal at 1584 cmꢁ1 and at a similar rate as
the Raman band at 542 cmꢁ1 disappears due to coke coverage
of Cr2O3 . Furthermore, it can be seen that the decreasing cat-
alyst activity observed using online mass spectrometry activity
data agrees well with the spectroscopically observed coke
build-up on the catalyst. For the combustion of coke in the
catalyst regeneration steps the information obtained using MS,
UV-VIS, Raman and reactor temperature is also in excellent
agreement.
The fact that it is now possible to use fast and easily applic-
able operando spectroscopic techniques using fiber optics
allows the use of it in the chemical industry as a means of
advanced process control in the near future. Being able to
monitor directly the amount of coke on the dehydrogenation
catalyst allows for the use of process control system, which
dynamically adjusts the cycle time for the dehydrogenation–
regeneration phases to optimally run the process. Since the
Raman measurements provide us with direct information on
Phys. Chem. Chem. Phys., 2003, 5, 4361–4365
4363

View Article Online
Fig. 4 UV-VIS spectra of 13 wt% Cr-on-alumina catalyst in its
regenerated state (1.5 bar(a) 20% O2 in He, 823 K). Spectra are com-
pared of the fresh catalyst, the catalyst after 100 h of propane dehydro-
genation cycles and after a deliberate deactivation procedure.
the catalyst deactivation was accelerated by using it for a pro-
pane dehydrogenation at 933 K for 4 hours, followed by a nor-
mal regeneration at 823 K. In Fig. 4, it can clearly be seen that
the absorbance of the active chromia species at 450 nm is sig-
nificantly less, indicating catalyst deactivation by loss of active
chromium sites. Also, the Cr3+ band at 630 nm has shifted a
bit to the right, which is indicative of chromium moving into
the alumina support. The average propene yield of this catalyst
in a propane dehydrogenation cycle at 823 K indeed was only
31%. This change in the UV-VIS spectrum clearly shows this
operando measurement can provide essential information on
the catalyst activity. Since in industrial processes the alkane
dehydrogenation is often carried out with some excess catalyst
to obtain the equilibrium conversion of the reaction, this loss
in catalyst activity cannot be determined immediately from
the conversion of the reactor.
Fig. 3 Overview of the most relevant observed changes in measured
signals during a complete dehydrogenation–regeneration cycle of pro-
pane over a 13 wt% Cr-on-alumina catalyst. (A) Observed propene
fraction in the gas phase; (B) observed carbon dioxide fraction in the
gas phase; (C) catalyst bed temperature; (D) ratio between the mea-
sured UV-VIS absorbances of the two oxidation states of chromium,
Cr(III) at 630 nm and Cr(VI) at 450 nm; (E) UV-VIS absorbance at
850 nm, corresponding to the overall increase in absorbance due to
coke formation on the catalyst; (F) Raman signal at 1584 cmꢁ1 Raman
shift corresponding to the major coke component formed; (G) Raman
signal at 542 cmꢁ1 corresponding to crystalline Cr2O3 .
Conclusions
A novel operando spectroscopic reactor system has been devel-
oped. The results presented in this paper demonstrate the value
of combining fast time-resolved operando spectroscopic mea-
surements using UV-VIS and Raman with quantitative fast
product analysis using mass spectrometry. Different states of
the active metal oxide phase can be observed, as well as infor-
mation on coke formation on the catalyst. It has been shown
that the different techniques provide complimentary pieces of
information, which are in excellent agreement with each other.
We anticipate that this new set-up will become a very useful
tool for operando catalyst characterization.
the amount of coke on the catalyst, this technique is most sui-
table for that. Operando UV-VIS measurements can be used to
monitor the long-term deactivation of the catalyst. Puurunen
and Weckhuysen10 have reported that the active chromium
oxide species in the dehydrogenation is most likely the chro-
mium oxide, which undergoes a redox cycle in the dehydro-
genation–regeneration cycles. Permanent deactivation of the
catalyst is linked to a decrease in the amount of this redox type
chromium oxide. Therefore, the change in the relative peak
intensities of the 450 and 630 nm bands (Fig. 3) on switching
from air to propane can be used to determine when the catalyst
needs to be replaced.
Acknowledgements
The capability of operando UV-VIS measurements to deter-
mine long-term deactivation was verified by performing a pro-
pane dehydrogenation experiment in which the catalyst was
purposefully deactivated. First propane was dehydrogenated
for 100 hours at the usual reaction temperature of 823 K
and a pressure of 1.5 bar(a). In Fig. 4 it can be seen that the
UV-VIS spectrum of the catalyst in the oxidated state does
not change significantly from the initial state compared to that
after 100 hours of dehydrogenation–regeneration cycles. In
each of these cycles the average propene yield was 63% con-
firming the catalyst stability observed with UV-VIS. After this
B.M.W. acknowledges financial support by the Dutch
National Science Foundation (NWO-CW–van der Leeuw
and VICI grants) and NRSC-Catalysis.
References
1
2
B. M. Weckhuysen, Chem. Commun., 2002, 97–110.
A. Bruckner, Chem. Commun., 2001, 2122–2123.
¨
4364
Phys. Chem. Chem. Phys., 2003, 5, 4361–4365

View Article Online
´
M. O. Guerrero-Perez and M. A. Ban˜ares, Chem. Commun., 2002,
3
4
5
6
7
D. Scarano, G. Spoto, S. Bordiga, L. Carnelli, G. Ricchiardi and
A. Zecchina, Langmuir, 1994, 10, 3094–3104.
Y. Tan Chua and P. C. Stair, J. Catal., 2003, 213, 39–46.
M. A. Vuurman, F. D. Hardcastle and I. E. Wachs, J. Mol.
Catal., 1993, 84, 193–205.
1292–1293.
In-situ spectroscopy in heterogeneous catalysis, ed. J. F. Haw,
Wiley-VCH, Weinheim, 2002.
M. M. Bhasin, J. H. McCain, B. V. Vora, T. Imai and P. R.
Pujado, Appl. Catal., A, 2001, 221, 391–419.
B. M. Weckhuysen, I. E. Wachs and R. A. Schoonheydt, Chem.
Rev., 1996, 96, 3327–3349.
8
9
10 R. L. Puurunen and B. M. Weckhuysen, J. Catal., 2002, 210,
418–430.
Phys. Chem. Chem. Phys., 2003, 5, 4361–4365
4365