Evolution of Pd catalyst structure and activity during catalytic oxidation of methane and ethane

Article abstract of DOI:10.1016/j.apcata.2014.07.028

Methane and ethane oxidation in alkane-rich mixtures over Pd powder in temperature range of 300-450 °C has been studied using combined mass-spectrometry and thermogravimetry. Fresh Pd powder was found to have very low stationary initial catalytic activity. The activity gradually rose with time on stream reaching a certain value where oscillations of the reaction rate began. The oscillation phases with relatively high catalytic activity were accompanied by a periodic accumulation/removal of carbon in Pd bulk, whereas during the low activity phases a periodic accumulation/removal of oxygen in Pd surface layer took place. The correlation between Pd oxidation state and catalytic activity was drawn. The metallic Pd had the lowest catalytic activity in alkane oxidation at high partial pressure of O₂ and while the highest catalytic activity was observed at low partial pressure and high conversion of O₂. Microscopic study of Pd samples revealed that the catalyst evolution caused changes in Pd surface morphology resulting in formation of sponge-like structure. In turn, such morphological changes were found to increase duration and intensity of the high activity phase.

Full text of DOI:10.1016/j.apcata.2014.07.028

Applied Catalysis A: General 485 (2014) 1–9
Contents lists available at ScienceDirect
Applied Catalysis A: General
journal homepage: www.elsevier.com/locate/apcata
Evolution of Pd catalyst structure and activity during catalytic
oxidation of methane and ethane
a,∗
a
a
b
a
V.Yu. Bychkov , Yu.P. Tyulenin , A.Ya. Gorenberg , S. Sokolov , V.N. Korchak
a
Semenov Institute of Chemical Physics, Russian Academy of Sciences, Kosygina str. 4, GSP-1, 119991 Moscow, Russia
Leibniz Institute for Catalysis, Albert-Einstein-Strasse 29a, 18059 Rostock, Germany
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 16 May 2014
Received in revised form 9 July 2014
Accepted 18 July 2014
Available online 30 July 2014
Methane and ethane oxidation in alkane-rich mixtures over Pd powder in temperature range of
◦
3
00–450 C has been studied using combined mass-spectrometry and thermogravimetry. Fresh Pd pow-
der was found to have very low stationary initial catalytic activity. The activity gradually rose with time
on stream reaching a certain value where oscillations of the reaction rate began. The oscillation phases
with relatively high catalytic activity were accompanied by a periodic accumulation/removal of carbon
in Pd bulk, whereas during the low activity phases a periodic accumulation/removal of oxygen in Pd sur-
face layer took place. The correlation between Pd oxidation state and catalytic activity was drawn. The
metallic Pd had the lowest catalytic activity in alkane oxidation at high partial pressure of O2 and while
the highest catalytic activity was observed at low partial pressure and high conversion of O2. Microscopic
study of Pd samples revealed that the catalyst evolution caused changes in Pd surface morphology result-
ing in formation of sponge-like structure. In turn, such morphological changes were found to increase
duration and intensity of the high activity phase.
Keywords:
Methane oxidation
Ethane oxidation
Palladium catalysts
Self-oscillations
©
2014 Elsevier B.V. All rights reserved.
1
. Introduction
to Pd/PdO redox conversions, but also goes through carbonized Pd
phase [8]. The latter is formed during catalytic activity oscillations
taking place in methane-rich environment.
Methane oxidation over Pd-based catalysts has been extensively
studied over the past 25 years due to its attractive potential for
catalytic combustion applications and more than one hundred of
relevant publications can be found in this excellent review [1]. Most
of the studies were performed in methane-lean mixtures since low
methane concentrations are interesting for practical applications.
In particular, it was established that the catalytic activity of Pd
varies significantly with the oxidation state of the metal which is
manifested in the methane conversion rate temperature hysterysis
Another important observation made was a systematic increase
of an average catalytic activity of Pd with time on stream. It is
known that the catalysts in hydrocarbon oxidation rarely remain
unaltered usually changing both their structure and activity while
on stream. Both increase and decrease in catalytic activity with time
on stream are possible, including Pd catalysts in alkane oxidation
[1]. We have studied self-oscillating modes of lower alkane oxida-
tion over Pd and reported that the oscillation parameters depended
on Pd prehistory and vary with time [8,9]. It is known that inves-
tigation of the oscillating catalytic processes can give additional
information about the reaction mechanisms which is difficult or
sometimes impossible to obtain by studying only stationary cat-
alytic processes [10]. Furthermore, evolution of the oscillations
themselves should be viewed in relation to the structural changes
occurring with the catalyst and certain dependencies of the catalyst
properties on its structure can be deduced.
[
2,3]. It was demonstrated by the in situ thermogravimetric analysis
that the changes in methane oxidation rate are concomitant with
the alterations in the catalyst weight originating in Pd/PdO redox
conversions. There are different opinions as to which Pd phase is
the most catalytically active. Some authors regard metallic Pd as
the most active [4,5] while others hold Pd oxide as more active
than metallic Pd [6,7]. In our study on methane oxidation over Pd in
methane-rich atmosphere, we demonstrated that the phase trans-
formations of Pd during the course of the reaction are not limited
König et al. were the first group who observed the oscillatory
behavior during methane oxidation over a thick Pd-film catalyst
and found large time-dependent variations in the reaction rate
[
11,12]. By means of ellipsometry they demonstrated that the for-
∗ Corresponding author. Tel.: +7 495 939 75 46.
E-mail addresses: bychkov@chph.ras.ru, bychkov@polymer.chph.ras.ru
V.Yu. Bychkov).
mation of porous PdO film greatly increased an effective surface
area of the catalyst and thus led to a large increase in the CH4
(
http://dx.doi.org/10.1016/j.apcata.2014.07.028
926-860X/© 2014 Elsevier B.V. All rights reserved.
0

2
V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
Table 1
The conditions of the appearance of the oscillatory behavior.
Reactor type
Reaction
Alkane
concentration (%)
Oxygen
concentration
Reactor temperature
( C)
Feedrate of the reactant
mixture (ml/min)
Amount of Pd
powder (mg)
◦
Tubular reactor
Tubular reactor
TGA cell
CH4 + O2
C2H6 + O2
CH4 + O2
C2H6 + O2
40.8–81.5
8.8–16.9
40.8–81.5
8.8–16.9
7.3–14.6
3.8–7.3
7.3–14.6
3.8–7.3
340–420
320–370
400–410
325–340
20–40
20–40
20–30
20–30
10–20
10–30
10–20
10–30
TGA cell
conversion rate. They also have found that under certain condi-
tions the film composition could spontaneously oscillate between
metal-rich and oxide-rich states. However, the authors could moni-
tor only 30 clearly defined oscillatory. The period of the oscillations
decreased with time and the surface became increasingly rough so
that it was no longer possible to obtain reliable ellipsometric data.
Zhang et al. also reported that the oscillations during methane oxi-
dation over palladium wire and foil were not stable. Results of the
ex situ SEM surface analysis revealed a rough and highly porous
surface indicating that the reaction caused a reconstruction of the
metal surface [13]. Oscillatory behavior was also observed over
supported Pd catalysts [14–17].
to the mass-spectrometer. The flow rate of the gas mixture was
maintained by the flow controller built into the SETSYS instrument.
The instrument is described in detail elsewhere [19].
In both catalytic and TGA experiments the gas flow was ana-
lyzed directly after exposure to the catalyst. Atomic mass units
(AMU) of 2 (H ), 15 (CH ), 18 (H O), 28 (CO, CO ), 30 (C H ), 32
2
4
2
2
2
6
(O ), and 44 (CO ) were detected. CO concentration in the outflow
2
2
must have been very low under all experimental conditions. Our
attempts to account for CO using AMU 28 value minus the value
from the standard fragmentation pattern of CO2 (∼10% of the AMU
44 signal intensity) were unsuccessful. GS analysis revealed a very
low mean concentration of CO (around two order of magnitude
During the recent studies of C –C hydrocarbon oxidation over
lower that that of CO ) under conditions of our experiments [8].
1
4
2
Pd catalysts, we encountered poor reproducibility of the oscillation
parameters [9]. It was found that the parameters depend signifi-
cantly on a state of the catalyst surface and a history of the catalyst
treatment. A complex pretreatment procedure including heating
The feed flow rate varied between 10 and 40 ml/min (see Table 1).
Micrographs of the spent catalysts were obtained on a high res-
olution scanning electron microscope (SEM) JSM-7001F (JEOL) at
5 kV.
◦
of the catalyst to 400–450 C followed by a controlled temperature
decrease was established in order to initiate any oscillatory regime
3. Results
of C –C oxidation. Moreover, it was found that no oscillations with
1
4
large amplitudes were possible in oxidation of Cn hydrocarbon after
3
.1. Oscillatory behavior during methane and ethane oxidation in
the catalyst was used in oscillatory oxidation of C
methane and so on).
(ethane after
n−1
the tubular reactor
The aim of this study was (i) to investigate in detail emergence
and evolution of the oscillations of the rate of catalytic methane
and ethane oxidation over Pd in the corresponding alkane rich
atmosphere, and (ii) to clarify what state of Pd catalyst initiates
and sustains such oscillations. A correlation between oscillatory
behavior and catalyst morphology was derived.
Experiments in the tubular reactor were carried out to find the
conditions of the oscillatory behavior during methane and ethane
oxidation over Pd powder. Oscillations during methane oxidation
were observed for feed rates of 20–40 ml/min at temperatures
◦
3
40–420 C for methane-rich reaction mixtures. In ethane oxida-
tion such behavior was obtained at lower temperatures in the range
◦
of 320–370 C and for more diluted reactant mixtures. The condi-
2
. Experimental
tions under which the oscillatory behavior in both systems was
observed are presented in Table 1.
The catalyst was prepared by mechanical grinding of Pd foil and
The reactant mixture comprising 40.8% CH4 and 7.3% O2 in He
was introduced in the reactor containing 9 mg of fresh Pd catalyst
and the reactor temperature was increased stepwise. Temperature
rise was accompanied by increase in the CO2 concentration in the
selecting the 50–100 ␮m fraction from the grit. The catalyst was
used either pure or mixed with quartz particles of similar size. Cat-
alytic experiments were carried out in a tubular quartz flow reactor
◦
(
i.d. 5 mm), operated at atmospheric pressure under shallow bed
products flow and after ca. 35 min at 375 C, a slow development of
conditions. Catalyst loadings were in the 10–30 mg range. The tem-
perature was controlled by a thermocouple on the outer side of
the reactor since chromel–alumel thermocouple wires can gener-
ate the reaction rate oscillations during methane oxidation [18].
Methane- or ethane-rich mixtures were pre-mixed for all catalytc
experiments.
Thermogravimetric analysis (TGA) was performed on Setaram
SETSYS EVOLUTION 16/18 instrument coupled with Pfeiffer
OmniStar GSD 301 quadrupole mass-spectrometer. The operation
unit of the thermogravimetric set-up consists of a vertical flow
tubular furnace with an inner alumina tube. The inner diameter
of the alumina tube is 20 mm and the length of the controlled
temperature zone is ∼30 mm. The standard TGA and DSC thermo-
gravimetric rods could not be used because they contain platinum,
which is active in alkane oxidation. The catalyst samples were
placed in a quartz cup suspended in the center of a heated zone.
Pre-mixed reactants were fed through the top of tube. The products
were sampled below the catalyst via a stainless capillary connected
oscillations was detected (see Fig. 1a). In the course of this induction
period, the catalyst activity and CO2 concentration slowly grew,
at the same time the O2 concentration decreased to a quarter of
its initial value. In case of a longer contact time (27 mg Pd, same
◦
flow), the oscillations started at a lower temperature of 365 C.
On the contrary, a decrease in contact time (5 mg Pd) resulted in
◦
higher oscillations initiation temperature of 385 C. Nevertheless,
the degree of O conversion before the start of the oscillations were
2
approximately the same in all three runs, namely 76.4, 76.5, and
◦
74.6% at 365, 375, and 385 C, respectively.
Fig. 1b–d shows a gradual change in the oscillation waveform
over ∼2 h on stream. The H O MS signal is not shown, because it
2
was distorted due to water re-adsorption on quartz wool. The first
3 oscillations shown in Fig. 1b begin with a small negative O2 peak
and positive CO peak which then followed by a large and extended
2
positive O2 peak and negative CO2 peak. With time on stream,
an average CO2 concentration increased and that of O2 decreased
while the waves of both changed their shape (Fig. 1c and d).

V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
3
a 5.E-07
b 5.E-07
1.E-08
4
3
2
1
.E-07
.E-07
.E-07
.E-07
4
3
2
1
.E-07
.E-07
.E-07
.E-07
0
.E+00
0.E+00
0
.E+00
8
4
85
86
87
0
20
40
60
80
100
120
140
Time, min
Time, min
5.E-07
2.5E-08
2.0E-08
1.5E-08
1.0E-08
5.0E-09
0.0E+00
c
5.E-07
1.5E-08
d
4
3
2
1
.E-07
.E-07
.E-07
.E-07
4
3
2
1
.E-07
.E-07
.E-07
.E-07
1
5
.0E-08
.0E-09
0
.E+00
0
.E+00
0.0E+00
113
179
180
181
182
183
184
1
10
111
112
Time, min
Time, min
Fig. 1. The variation of O2, CO2 (+1.5 × 10−7) and H2 MS signals during stepwise heating of Pd powder (10 mg Pd powder mixed with 50 mg of quartz) up to 375 C in a
◦
mixture 40.8% CH4–7.3% O2–He (20 ml/min) in the tubular reactor: the whole experiment (a) and details of the evolution of the oscillatory behavior (b–d).
Negative O2 peak first seen in Fig. 1b became broader and had a
3.2. TGA study of the oscillatory behavior of methane and ethane
oxidation
lower minimum indicating high O conversion in a–b range (resid-
2
ual O2 content is ∼3% from the initial concentration) with a steep
drop of O concentration at the beginning and a rise at the end of
The conditions for oscillatory behavior established in the tubu-
lar reactor (see Table 1) were reproduced in the TGA experiments;
all experiments were performed in the isothermal mode. Fig. 3a–c
shows a development of the oscillatory behavior during methane
oxidation over Pd powder catalyst. The catalyst was heated up to
2
the range. An average duration of this phase increased with time
on stream while the oscillations themselves occurred less regularly.
Fig. 1c and d show that the fall of O2 concentration in a–b range is
accompanied by a small dip in CO2 concentration (a negative CO2
peak) and H2 evolution. The a–b range can be called a “high activ-
ity” phase of the oscillation cycle. The b–a range characterized by a
lower catalytic activity is located between the high activity phases.
The “low activity” phase starts with an abrupt drop in methane oxi-
dation rate (CO signal fall and O signal rise) followed by a gradual
◦
400 C in He flow (20 ml/min). At this temperature the reactant mix-
ture comprising 85% CH4 and 15% O2 was introduced at a rate of
10 ml/min through an additional inlet (10 min in Fig. 3). Simulta-
neously, the feed rate of He was decreased to 10 ml/min to obtain
the conditions necessary for the oscillatory behavior (20 ml/min of
2
2
increase in the reaction rate up to a sharp start of the high activity
phase. It is clear from Fig. 1b–d that the duration of the low activity
oscillation phase slightly varies as oscillations progress.
reactant mixture comprising 42.5% CH and 7.5% O in He). As seen
in Fig. 3a, the catalytic activity early into the experiment was very
4 2
low. After the induction period of 80 min, CO2 and H O concentra-
2
The oscillatory behavior during ethane oxidation over the Pd
catalyst was very similar. The oscillations of catalytic activity dur-
ing ethane oxidation started at 365 C in more diluted compared to
tions began to rise, while O concentration began to decrease. After
2
∼260 min, when the oxygen concentration decreased to 45% of its
◦
initial value, oscillations of CO , H O, O and CH4 concentrations
2
2
2
methane reactant mixture (16.9% C H6 and 7.3% O2 in He). Fig. 2a
appeared. With time on stream, the amplitude and the period of
oscillations slowly increased. It was found that the induction period
could not be eliminated or shortened by applying a prolonged (sev-
2
demonstrates a slow development of the oscillations at this tem-
perature. The initial reaction rate was very low, but gradually
increased with the CO2 and O2 concentrations changing in oppo-
site directions. After a 40 min induction period the oscillations
appeared. The trends in variation of the period and the waveform
of the oscillations for ethane and methane oxidation were similar.
Fig. 2a–c shows that the period of oscillations increased with time
and the amplitude of H2 peaks grew. Similar to methane oxida-
tion, at loner time on stream a share of the high activity phase in
a given oscillation period became higher which can be clearly seen
by comparing O2 signal dips in Fig. 2b and c.
eral hours) treatment of fresh Pd catalyst in He or O –He flow at
2
◦
400 C. As shown in Fig. 3a, the MS signals are interrupted at 320
and 510 min, at which points the reactant mixture was pumped out,
the catalyst cooled to room temperature and left overnight. Sur-
prisingly, when exposed to the oscillatory conditions on the next
day, the catalyst immediately restored its last detected activity and
continued oscillating. Moreover, the treatment of the catalyst at
◦
400 C in a flow of oxygen once in oscillatory mode did not quench
its activity and did not affect oscillations.

4
V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
a 1.5E-06
a
CO
2
1
.0E-07
9
.3
1
.0E-06
9.2
8.0E-08
9
.1
9
6
4
.0E-08
.0E-08
O
2
5
.0E-07
8
.9
0
.0E+00
8.8
2.0E-08
4
0
60
80
100
120
140
160
180
0
100
200
300
400
500
600
Time, min
Time, min
b 9
.095
8.E-08
.E-08
b
2.0E-06
1.5E-08
H
2
7
1
1
5
.5E-06
.0E-06
.0E-07
1
5
.0E-08
.0E-09
9.075
9.055
6.E-08
5.E-08
4.E-08
CO
2
2
O
3
.E-08
9
.035
2.E-08
425
0
.0E+00
0.0E+00
3
85
390
395
400
405
410
415
420
6
0
65
70
75
80
Time, min
Time, min
c
9.55
9.E-08
c
2.5E-06
1
.5E-08
2
1
1
5
.0E-06
.5E-06
.0E-06
.0E-07
H
2
9
9
9
.45
.35
.25
7.E-08
5.E-08
3.E-08
1.0E-08
CO
2
5
.0E-09
O
2
0
.0E+00
0.0E+00
6
05
610
615
620
625
630
635
9
5
97
99
101
103
Time, min
Time, min
Fig. 3. The variation of O2, CO2, H2O MS signals and the catalyst weight variation
Fig. 2. The variation of O2, CO2 (+7 × 10− ) and H2 MS signals after heating of Pd
7
during the reaction of Pd powder (19.4 mg) with a mixture 42.5% CH4–7.5% O2–He
◦
powder (10 mg Pd powder mixed with 50 mg of quartz) up to 350 C in a mixture
6.9% C2H6-7.3% O2–He (20 ml/min) in the tubular: the whole experiment (a) and
details of the evolution of the oscillatory behavior (b, c).
◦
−8
−9
(
(
3
20 ml/min) in TGA cell at 400 C: (a) the whole experiment (CO2, +2 × 10 ) and
1
b, c) details of the evolution of the oscillatory behavior ((b) H2O, +5 × 10 , O2,
−8
−8
−8
−8
× 10 , CO2, 3.5 × 10 , (c) O2, 2.5 × 10 , CO2, 2.5 × 10 ).
The dynamics of the catalyst oscillating activity are shown in
particular, during the weight increase an additional H O evolution
2
Fig. 3b and c as MS signals of H O, CO2 and O2 and weight of the
and drop in CO2 concentration (negative peak) are observed. Dur-
ing the following weight decrease, one can see a rise of CO2 signal
2
sample. It is apparent from the plots that the period and the ampli-
tude of the oscillations increase with time and the waveform varies
in a complex manner. During the first 20 min in Fig. 3b, emerging
weight oscillations had low amplitude which however gradually
increased. The maxima of O2 signal oscillations correspond to the
accompanied by a decrease of H O signal.
2
These effects become more pronounced further into the exper-
iment. Fig. 3c shows final oscillations with a period of ca. 11 min.
Here it is clearly seen that the weight increase at point a is accom-
minima of the weight oscillations; CO and H O signals are in phase
panied by the rise of H O signal and the fall of CO2 signal. The
2
2
2
with each other and with the weight. Such phasing of the signals
indicates that the catalyst reaches its maximum weight at the peak
of catalytic activity. After 20 min on stream, weight oscillations
exhibited a steep increase in amplitude marked by progressively
growing sharp peaks. It is seen in Fig. 3b that the weight increase
at point a corresponds to the fall of O2 signal, while the weight
decrease at point b corresponds to the rise of O2 signal. Also, in
dominance of H O over CO2 signal signifies accumulation of car-
2
bon on the catalyst, whereas a reversed trend corresponds to its
removal. This result is in a good agreement with our previous study
[8] where we also demonstrated that such unusual dynamic behav-
ior is the result of carbon accumulation-removal on the Pd catalyst.
The current experiments have shown that with time the amount
of carbon participating in oscillatory cycles increases leading to
the interval a–b H O signal does no longer replicate CO2 signal. In
variation in the waveforms of CO2 and H O signals.
2
2

V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
5
2
5
0.015
0.01
a 7.E-07
tg
b
-
-
-
-
-
5.9
6
5
4
3
2
1
.E-07
.E-07
.E-07
.E-07
.E-07
.E-07
20
6
2
Н О
1
5
6.1
6.2
6.3
СО
2
10
5
c
a
0
0
.005
О
2
-6.4
-6.5
0
0
5
10
15
0
50
100
150
200
250
Osciilation cycle number
Time, min
Fig. 5. The dependence of the duration of (a) a–b (high activity phase) and (b) b–a
low activity phase) stages on the oscillation cycle number for the experiment shown
(
in Fig. 4. (c) Variation of weight fall at point a due to Pd oxide reduction.
b
7
5
3
1
.E-07
.E-07
.E-07
.E-07
the high activity phase significantly increased with time, while the
length of the low activity phase remained nearly constant.
-
6.03
3.3. SEM study of the sample microstructure during the evolution
of the oscillatory behavior
-6.08
In order to study morphological changes that could have
occurred with Pd catalyst during the induction period and the oscil-
lations, the catalytic tests identical to that illustrated in Fig. 3a were
quenched at stages A, B, and C and the samples were investigated
by SEM ex situ. Fig. 6a and e shows SEM images of the sample A
-6.13
2
0
25
30
35
40
45
50
55
60
Time, min
◦
that was quenched at 400 C immediately after exposure to the
reactant mixture, which displayed very low activity and no oscil-
lations. The sample morphology is typical for metallic Pd featuring
grains with some grooves resulting from the mechanical processing
of Pd foil. The SEM images in Fig. 6b and f demonstrate the cata-
lyst B just after beginning of oscillations. At low magnification the
samples B and A are practically indistinguishable, however at high
magnification some structural defects on the surface of sample B
can be noticed which we relate to the emerging oscillations. The
images in Fig. 6c and g show the catalyst C after 10 h on stream
during which the oscillations of methane oxidation were lasting.
The “oscillating” sample C underwent the most severe alteration
yielding a complex spongy morphology. In addition, the sample D
Fig. 4. The variation of O2, CO2 (+2 × 10−7), H2O (+3 × 10 ) MS signals and the
−7
catalyst weight variation during the reaction of Pd powder (26.0 mg) with a mixture
.8% C2H6–3.8% O2–He (20 ml/min) in TGA cell at 340 C: the whole experiment
a) and details of the evolution of the oscillatory behavior (b). An inset shows a
◦
8
(
comparison of H2O and CO2 MS signals in a–b range.
Similar trends were observed during ethane oxidation. Fig. 4a
and b shows the development of the oscillatory behavior during
ethane oxidation started over the fresh catalyst at 340 C. Similar
◦
to methane oxidation, the activity of Pd catalyst slowly increased
during 24 min before the oscillations appeared. Again, the weight
oscillations began with low amplitude, while the oscillations of CO2,
H O and O signals proceeded with sufficiently greater amplitudes.
With time on stream, the period of oscillations slowly increased,
while the oxygen concentration in the minima of the oscillations
decreased. The system spent progressively more time in the high
activity phase of the oscillatory cycle where accumulation of car-
bon in the catalyst occurred. Fig. 4b demonstrates development of
the wave shapes especially evident in case of weight oscillations.
It can be seen that the cycles began with a slow rise followed by
a rapid fall (point a) of the catalyst weight. After these alterations,
the weight rapidly reached a maximum and dropped to the level
◦
was produced by heating of the “oscillating” sample C up to 1000 C
2
2
in 5% O –He flow. The micrographs of this sample are presented in
2
Fig. 6d and h. After the high-temperature treatment, the surface
became much smoother and the sponge morphology disappeared.
The faces corresponding to low-index crystal planes of Pd could be
identified. The catalytic activity of the sample D was very similar
to that seen on the “fresh” sample A. Upon exposure to the 42.5%
◦
CH /7.5% O /50% He gas mixture at 400 C, the sample D demon-
4
2
strated a very low activity, which gradually increased following the
trend seen in Fig. 3a.
Fig. 7 displays SEM images collected on Pd samples used in
the ethane oxidation experiments stopped at the same A, B and C
points as above. Although in this case Pd powder exhibited sustain-
able oscillations at lower temperature and in more diluted reactant
mixture (see Table 1), the surface morphology of the catalyst just
before the beginning of the oscillations (A) was very similar to
that obtained at the beginning of methane oxidation and shown
in Fig. 6e. The main difference between the effect of methane and
ethane oxidation on the catalyst surface is evident from the com-
parison of the “oscillating” catalysts. SEM images of Pd powders
removed from the reactors after prolonged oscillations (point C)
reveal that the sponge evolved in methane oxidation (Fig. 6g) had
more open structure while that from ethane oxidation (Fig. 7c)
appeared to be more sintered.
(
point b) at which the next cycle started. As in case of methane oxi-
dation, based on the MS signal profiles H O evolves faster than CO
during the weight increase stage (see an inset in Fig. 4b). The max-
imum of the CO2 signal occurs at the end of the interval a–b when
the weight loss is well underway and the H O signal is also on the
2
2
2
downward trend. This shift of the CO2 maximum in respect to H O
2
as well as the weight gain–loss in the interval a–b are caused by the
carbon accumulation-removal. With time on stream, the value of
the weight maxima gradually increased indicating that more and
more carbon participated in the oscillatory process.
An oscillatory cycle consists of the phases of high and low
activity defined in Fig. 4b as a–b and b–a respectively. Fig. 5 demon-
strates a change in the phase duration with growing number of
oscillatory cycles for the first 16. One can see that the duration of

6
V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
◦
Fig. 6. SEM images of samples A (a, e), B (b, f), C (c, g), and D (d, h) quenched during the oscillatory methane oxidation at 400 C with low (a–d) and high (e–h) magnification.

V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
7
4. Discussion
In our earlier reports we have revealed a multi-stage mechanism
of the oscillations during low alkane oxidation over Pd catalysts
powders and foils) [8,9]. The mechanism includes the following
periodic successive stages:
(
(1) Drastic decrease of Pd weight due to PdO reduction to metallic
Pd.
(
2) Intense weight gain due to Pd carbonization occurring during
the high activity phase of the oscillation cycle with high degree
of O2 conversion.
(
3) Weight drop and CO evolution as oxidation of the accumulated
2
carbon proceeds.
(
4) Decrease in O2 conversion, blocking of metallic Pd by
chemisorbed oxygen, minimal catalytic activity and at this
point the beginning of the low activity phase of the oscillation
cycle.
(
5) Gradual Pd oxidation to PdO with increase in the catalyst
weight, simultaneous increase of the catalytic activity to a cer-
tain level.
(6) Return to stage (1).
Accumulation of carbon during the high activity stage was con-
firmed by XRD and TPO of Pd powders recovered from the catalytic
tests quenched at this stage. Formation of PdO was proven by TPD
testing of quenched Pd sample. Mathematical modeling strongly
suggested that the observed oscillatory behavior can be rational-
ized by the mechanism outlaid above [20].
The findings made in this work are in agreement with the
described mechanism and help to extend our knowledge about
the nature of the oscillatory behavior of Pd catalyst during oxi-
dation of methane and ethane and correlating such behavior
with the catalyst evolution. It is shown that fresh Pd powder
demonstrates only low catalytic activity that rises with time on
stream in the alkane–oxygen mixture at constant temperature.
The results of SEM investigation reveal that the activation of Pd
powder was caused by the development of surface morphology
and the formation of numerous surface defects and sponge struc-
tures. This intensive surface development cannot proceed in the
inert or oxidative media and needs the reactive mixture con-
taining high concentration of a reductive agent, such as methane
or ethane. An oscillating mode of alkane oxidation causes much
more intense surface development than a stationary mode of the
reaction.
In literature one can find other examples of Pd activation dur-
ing alkane oxidation and explanations of these effects. In the
review [1] three possible reasons of the activation of supported
Pd catalysts are mentioned: (1) poisoning of fresh catalyst with
chlorine that originates from the catalyst precursors; (2) crystal-
lization of initially amorphous PdO; (3) metal-support interaction
of Pd. Obviously, the reasons (1) and (3) are irrelevant in our
case. The reason (2) can be also excluded because it was demon-
◦
strated that treating Pd powder at 400 C in O2 that could have
led to PdO formation and crystallization did not, however, cause
Pd activation. In [21] it was shown that Pd particles can be acti-
vated at high temperature where recrystallization takes place.
This effect was explained by increasing with recrystallization the
fraction of low-index Pd faces that have higher catalytic activ-
ity. But this effect cannot also explain Pd activation in our case.
◦
SEM images of Pd sample D heated to 1000 C (Fig. 3d and h)
evidence that the recrystallization of metallic Pd and the forma-
tion of low-index Pd faces occurs. But the catalytic activity of
this sample was at the same low as that of the fresh Pd pow-
der. Authors of [22] observed a reversible increase of the metal
surface area after treatment of Pd foils in excess of methane at
Fig. 7. High magnification SEM images of samples quenched during the oscillatory
ethane oxidation at 340 C: (a) initial sample, (b) beginning of the oscillations, (c)
several hours of the oscillatory behavior.
◦

8
V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
Table 2
5
98 K in a batch reactor. A batch reactor is not suited for studying
Maximal observed carbon and oxygen accumulation during the oscillatory alkane
oxidation.
oscillations in catalytic activity, but we suppose that the effect
reported in [22] corresponds to the evolution of Pd powder which
in our case happened at the beginning of the oscillatory cycle.
The appearance of self-oscillations during catalytic oxidation of
methane or ethane is closely connected to the evolution of Pd mor-
phology and oxidation state accompanied by changes in its catalytic
activity. The oscillations are absent on fresh Pd powder with the
low catalytic activity and appear when it reaches a certain level.
The results obtained for methane oxidation in the tubular reactor
demonstrate that the decrease of Pd loading drives up the temper-
ature of the oscillation start, but the reaction rate (O2 conversion)
at this point were practically the same: 23.5–25.5% of the initial
O2 concentration remained. The fact that during the oscillations
observed in TGA set-up the O2 content decreased to only ∼37% is
apparently caused by mass-transfer limitations and zero contact
time for a certain portion of feed inevitable in other than flow-
through reactor scheme. The structure of oscillation cycle observed
here suggests the following explanation of this effect. In the cur-
rent and earlier [8,9] studies, it was shown that during oscillations,
a periodic swing between oxidative or reductive conditions takes
place despite the fact that the initial gas mixtures were alkane-rich.
Alkane type
CH4
400
6.9
0.3
620
21
C2H6
Reactor temperature (◦C)
Maximal carbon content (mg C/g Pd)
Maximal oxygen content (mg O/g Pd)
340
4.6
0.7
220
22
2
Maximal rate of carbon accumulation (mg C/cm × min)
2
Maximal rate of oxygen accumulation (mg C/cm × min)
to PdC0.13 which can give a weight gain of 14.7 mg C/g Pd. Values of
6.9 and 4.6 mg C/g Pd measured during the oscillating oxidation of
correspondingly methane and ethane do not reach this level prob-
ably because of the slow carbon diffusion into Pd grains. An
assumption that the measured values are the results of carbon dif-
fusion into a 50 ␮m thick metal layer (50–100 ␮m fraction of Pd
particles) allows to estimate the depths of carbon diffusion as 23.5
and 15.5 ␮m respectively. It is clear from Table 2 that the rate of
carbon accumulation during methane oxidation was much higher
than the one for ethane oxidation. This difference is reflected in
the fact that the intervals with anti-phase evolution of H O and
2
CO2 are clearly seen during methane oxidation, but these signals
are only partially asynchronized during ethane oxidation. Possibly,
this difference in the carbon accumulation rates originates from the
difference in the temperatures at which methane and ethane oxida-
Under the oxidative conditions, a significant amount of O remains
2
in the gas phase (low activity b–a phase in Figs. 1c and 3c), while
under the reductive conditions O content falls sharply from the ini-
2
◦
tial content (high activity a–b phase in Figs. 1c and 3c). It is known
tion started (400 and 340 C, respectively) because carbon diffusion
[
1] that Pd is much more reactive to O2 than to CH . Apparently,
obviously must accelerate with temperature.
4
high O2 concentration in gas phase can prevent transformation of
Pd into the reduced state possessing high catalytic activity. On other
hand, if catalytic activity becomes too high due to temperature
increase or prolonged Pd surface development, the oxidative condi-
tions become hardly accessible, the self-oscillations disappear, and
a steady state reaction with almost total O2 consumption prevails.
Pd development leads not only to the appearance of the self-
oscillating mode of alkane oxidation, but also to a significant
modification of the oscillation waveform. In paper [8] we com-
pared periodic variations of Pd weight in the course of methane
oxidation with the variations of chemical composition of Pd cata-
lyst. On the basis of the aforementioned multi-stage mechanism of
the oscillations and the results obtained here, it was concluded that
the weight increase at the beginning of the high activity a–b phase
of the oscillation was caused by carbon accumulation and disso-
lution in Pd grains. The following weight decrease was the result
of the removal of the accumulated carbon via its oxidation to CO2.
On the contrary, weight increase during the low activity b–a phase
was connected with Pd oxidation. Such weight increase due to Pd
oxidation is clearly observed in b–a intervals of the first oscillation
It is seen from Table 2 that the maximal amounts and rates of
oxygen accumulation in Pd powder observed during the oscillatory
methane and ethane oxidation were much lower than the corre-
sponding values for carbon. Nevertheless, it is well known that O is
2
easily chemisorbed on Pd surface forming structures whose compo-
sition depends on temperature and O pressure [1]. The conditions
2
of our experiments allow existence of amorphous or crystalline
palladium oxide PdO. Assuming formation of PdO on the surface
of 50 ␮m thick Pd particles, the thickness of PdO film would be
at most 0.1 and 0.24 ␮m during the rate oscillations of methane
and ethane oxidation respectively. Since the measured rates of oxy-
gen accumulation were similar for methane and ethane oxidation
(Table 2), the greater thickness of PdO layer formed on Pd during
ethane oxidation is the result of the longer low activity phase in
this reaction. XRD analysis of the samples quenched after the oscil-
latory methane oxidation did not reveal crystalline PdO [8] which
leads us to conclude that the formed layer contained amorphous
Pd oxide [25].
The difference in oxygen and carbon reactivity and diffusivity in
respect to Pd results in modifications of oscillation waveform with
time on stream. Carbon diffused into the bulk of Pd crystallites does
not have time to diffuse back completely during the carbon removal
stage at the end of the high activity phase. Therefore, a certain part
of the accumulated carbon is still being oxidized and removed dur-
ing the following low activity phase. During this phase Pd oxide is
formed, and the weight gain due to Pd oxidation overlaps with the
weight loss caused by carbon removal (see, e.g. b–a phase in Fig. 3c).
In the case of oscillatory ethane oxidation, because of relatively low
rate of carbon accumulation/removal (carbon diffusion), a signifi-
cant part of accumulated carbon cannot be removed which results
in a gradual increase of the catalyst weight as seen in Fig. 4a. In con-
trast to carbon, accumulation/removal of oxygen remains almost
unchanged through the oscillation cycles. This can be concluded
from the weight drop at the boundary a being practically constant
through the cycles (curve c in Fig. 5). The weight drop at point a
(see Fig. 4b) corresponds to oxygen removal from PdO film during
its reduction to Pd. Constant weight drop during PdO reduction is
indicative of constant degree of Pd oxidation during the preceding
b–a phases.
ꢀ
ꢀ
cycles of methane (b –a in Fig. 3b) and ethane (Fig. 4b) oxidation
where this oscillation phase constitutes a greater part of the oscil-
lation cycle. In both cases, the minima in the weight curves, i.e.
the minima of Pd oxidation state correspond to the minima of cat-
alytic activity (CO and H O evolution), while gradual Pd oxidation
2
2
(
weight increase) leads to a rise of catalytic activity. Higher rates of
methane/ethane oxidation result in falling concentration of O2 in
gas phase which in turn leads to an abrupt PdO reduction (weight
loss at point a in Fig. 4b) and beginning of the high activity phase.
The high activity a–b phase starts with the fast weight gain due to
carbon accumulation and finishes with the weight loss because of
the removal of the accumulated carbon as CO . In all experiments
2
carried out in this study, the duration of the high activity phase
increases with time on stream (see Figs. 1–4).
In Table 2 one can find maximal rates and values of carbon and
oxygen accumulation measured during the oscillatory oxidation of
methane and ethane over Pd powder. It was shown in [23,24] and in
our report [8] that under the applied experimental conditions, car-
bon dissolves in Pd crystallites up to a certain level corresponding

V.Yu. Bychkov et al. / Applied Catalysis A: General 485 (2014) 1–9
9
Data presented in Figs. 4a and 5 demonstrate that the dura-
tion of the low activity phase does not change significantly with
time on stream, but the duration of the high activity phase strongly
increases with the number of cycles. The latter effect evidently
owes to growing catalytic activity of Pd resulting from the metal
surface development as evidenced by SEM. Under the conditions of
methane-rich feed, increase in Pd catalytic activity will deplete the
gas phase of O2 thus establishing reductive conditions. In the self-
oscillating mode, it causes an extension of the high activity phase
up to disappearance of the oscillating mode.
Effect of the “deep” carbon remaining in Pd bulk through all
phases of the oscillations on the catalytic activity is not clear. Pos-
sibly, presence of such carbon in Pd also stabilizes a high activity
state of carbonized Pd as seen in Fig. 4a. But the fact that grad-
ual buildup of “deep” carbon with number of cycles during ethane
oxidation practically did not influence the rate of oxygen uptake
during the low activity phases and the duration of these phases
allows to assume that Pd carbonization under the surface has weak
or negligible influence on the surface reactions.
and H O concentration is a characteristic difference between the
2
low and high activity phases respectively. It was established that
the catalytic activity oscillations were accompanied by periodic
changes in the catalyst weight. The weight variations during the
high activity phase of the oscillations were caused by the accumu-
lation/removal of carbon whereas during the low activity phase
such variations were a result of oxygen accumulation/removal.
Deposited carbon diffuses into a bulk of Pd crystallites, whereas
oxygen remains on the surface forming Pd oxide layer. It was found
that evolution of the Pd surface morphology increased duration and
intensity of the high activity phase, but practically did not affect the
low activity phase.
In situ TGA measurements reveled that both metallic and
oxidized Pd were catalytically active, but on the former alkane
conversion occurred predominantly via carbon accumulation-
removal, while on the latter direct catalytic oxidation took
place.
Acknowledgments
The relation between Pd catalytic activity and its oxidation state
is discussed for years and publications with different opinions can
be found in review [1]. Most researchers agree that in methane-
lean feeds PdO is more active in oxidation than metallic Pd in the
This work was supported by the Russian Foundation for Basic
Research (grant N 12-03-00282-a). This work was produced using
equipment of MIPT Center of Collective Use (CCU) with the financial
support from the Ministry of Education and Science of the Russian
Federation.
◦
temperature range where PdO is stable (<800 C). Low catalytic
activity of reduced Pd can be caused by its very high reactivity to
O2 which can adsorb on active Pd sites and block them. Our results
demonstrate that even in the methane-rich feed, catalytic activity
References
(
O2 conversion, CO2 and H O yield) increases with Pd oxidation
2
[
[
1] D. Ciuparu, M.R. Lyubovsky, E. Altman, L.D. Pfefferle, A. Datye, Catal. Rev. 44
2002) 593–649.
2] R.J. Farrauto, M.C. Hobson, T. Kennelly, E.M. Waterman, Appl. Catal. A: Gen. 81
(1992) 227–237.
state as seen in the low activity phase of the cycle where O2 partial
pressure is rather high during (b–a in Fig. 3b and c). But the maxi-
mal catalytic activity is observed on the reduced and carbonized Pd
during the high activity a–b phase. The weight increase after point
a in Fig. 3c indicates that after the beginning of Pd carbonization,
the metal is saturated by carbon and remains in this state for sev-
eral minutes, and sometimes tens of minutes. The most of carbon is
dissolved in Pd grains and do not act as intermediate species in the
catalytic reaction taking place on the surface. We hypothesize that
such carbonization of Pd prevents blocking the metal surface by
oxygen and concomitant deactivation. Based on in situ TGA exper-
iments, we hypothesize that there are two states of metallic Pd
with a drastically different catalytic activity in alkane oxidation:
carbonized metallic Pd with high catalytic activity and metallic
Pd with chemisorbed oxygen with very low catalytic activity. In
agreement with literature [1], Pd oxide also demonstrated rather
high catalytic activity that increased with the oxidation state. But
under the conditions of our experiments, the activity of Pd oxide
was lower than that of carbonized metallic Pd.
(
[3] R.J. Farrauto, J.K. Lampert, M.C. Hobson, E.M. Waterman, Appl. Catal. B: Environ.
(1995) 263–270.
[
[
[6] J.G. McCarty, Catal. Today 26 (1995) 283.
[
[
67 (2009) 181–187.
[9] V.Yu. Bychkov, Yu.P. Tyulenin, M.M. Slinko, V.N. Korchak, Catal. Lett. 141 (2011)
02–607.
10] M.M. Slinko, N.J. Jaeger, Stud. Surf. Sci. Catal. 86 (1994) 269.
11] D. König, W.H. Weber, B.D. Poindexter, J.R. Mcbride, G.W. Graham, K. Otto, Catal.
Lett. 29 (1994) 329–338.
6
4] R.F. Hicks, M.L. Young, R.G. Lee, H. Qi, J. Catal. 122 (1990) 280.
5] M. Lyubovsky, L. Pfefferle, Catal. Today 47 (1999) 29.
7] R. Burch, F.J. Urbano, Appl. Catal. A: Gen. 124 (1995) 121.
8] V.Yu. Bychkov, Yu.P. Tyulenin, M.M. Slinko, D.P. Shashkin, V.N. Korchak, J. Catal.
2
6
[
[
[
[
12] G.W. Graham, D. König, B.D. Poindexter, J.T. Remillard, W.H. Weber, Top. Catal.
8
(1999) 35–43.
13] X.L. Zhang, C.S.-M. Lee, D.M.P. Mingos, D.O. Hayward, Appl. Catal. A: Gen. 240
(2003) 183–197.
14] Y. Deng, T.G. Nevell, Faraday Discuss. 105 (1996) 33–46.
15] Y. Deng, T.G. Nevell, J. Mol. Catal. A 142 (1999) 51–60.
16] B. Kimmerle, A. Baiker, J.-D. Grunwaldtz, Phys. Chem. Chem. Phys. 12 (2010)
2288–2291.
[
[
[
[
[
17] J. Stötzel, B. Kimmerle, M. Nachtegaal, J.-D. Grunwaldtz, J. Phys. Chem. C 116
(
2012) 599–609.
18] X.L. Zhang, D.O. Hayward, D.M.P. Mingos, Catal. Lett. 72 (2001)
47–152.
[19] V.Yu. Bychkov, Yu.P. Tyulenin, V.N. Korchak, E.L. Aptekar, Appl. Catal. A: Gen.
04 (2006) 21–29.
20] N.V. Peskov, M.M. Slinko, V.Yu. Bychkov, V.N. Korchak, Chem. Eng. Sci. 84 (2012)
84–694.
[21] M. Lyubovsky, L. Pfefferle, A. Datye, J. Bravo, T. Nelson, J. Catal. 187 (1999)
75–284.
22] R.S. Monteiro, D. Zemlyanov, J.M. Storey, F.H. Ribeiro, J. Catal. 201 (2001) 37–45.
[23] S.B. Ziemecki, G.A. Jones, D.G. Swartzfger, R.L. Harlow, J. Faber Jr., J. Am. Chem.
Soc. 107 (1985) 4547–4548.
24] J.A. McCaulley, J. Phys. Chem. 97 (1993) 10372–10379.
25] H. Gabasch, K. Hayek, B. Klo1tzer, W. Unterberger, E. Kleimenov, D. Teschner,
S. Zafeiratos, M. Ha1vecker, A. Knop-Gericke, R. Schlögl, B. Aszalos-Kiss, D.
Zemlyanov, J. Phys. Chem. C 111 (2007) 7957–7962.
5
. Conclusions
1
Methane and ethane oxidation in alkane-rich mixtures over Pd
3
powder was studied. CO2 and H O were the main reaction prod-
2
[
ucts. The reaction was accompanied by a significant evolution of
the catalyst surface and a change of the reaction regime from sta-
tionary to oscillating. The oscillations began when the reaction rate
increased and O2 content in gas phase dropped to a certain level.
The oscillating regime resulted in a dramatic modification of Pd
surface morphology compared to the stationary one.
6
2
[
[
[
One oscillation cycle can be divided into two phases with high
and low catalytic activity. In-phase and antiphase change of CO2