X-ray absorption spectroscopy under reaction conditions: suitability of different reaction cells for combined catalyst characterization and time-resolved studies

Structure–activity correlations of solid catalysts and time-resolved studies generally require that the structure within a solid catalyst is probed simultaneously and at the same location where the catalysis and the structural changes occur.

These requirements lead to a compromise between the spectroscopic arrangement and the optimum design for an in situ reactor cell.

Opportunities and limitations of in situ and time-resolved X-ray absorption spectroscopy (XAS) combined with gas analysis are critically analysed with the help of two different cell designs, an in situ EXAFS cell designed for solids in the form of pressed wafers and a capillary cell where the catalyst is packed similarly to a plug flow reactor.

On the basis of three examples, the reduction of CuO/ZnO, the reduction of PdO/ZrO2 and methane oxidation over PdOx/ZrO2, criteria are developed which allow to judge the appropriate cell design in solid–fluid reactions.

The prerequisites for the design of an in situ cell including the catalyst shape strongly depend on the time resolution required.

Important issues embrace the type of reaction to be investigated (slow vs. fast), the reaction medium, and the porosity of the catalyst material.

Criteria for assessing the role of pore and film diffusion in in situ studies are of paramount importance for a proper experimental design.


Time-resolved spectroscopic studies and the combined investigation of the structure and activity are important in the field of heterogeneous catalysis.1–11

While time-resolved studies can elucidate information on intermediates on the catalyst surface, the combination of structure and activity aims at possible relationships between the active centres and the reactivity of the catalyst, being a first step to rational catalyst design.

Both goals require in situ investigations and the design of appropriate in situ cells for simultaneous structural and catalytic studies.

Ex situ studies have been performed to a great extent but it is generally accepted nowadays2,3,10 that such studies provide structure–activity relationships only in selected cases, i.e. if it has been ruled out in advance that the structure does not change if the catalyst is exposed to ambient atmosphere or reaction conditions.

Even cooling from the reaction temperature to room temperature may change the catalyst structure (cf. example in ref. 12).

In particular, techniques based on X-rays (X-ray diffraction, scattering techniques, X-ray absorption) have been used,4,8,9,13–16 since in situ cells can be constructed that mimic conditions close to the reaction conditions, even in liquid phase or at high pressure.17–19

This can be traced back to the fact that a number of cell window materials exist (Kapton, Al foil, thin diamond, Be and quartz windows), which have sufficient low absorbance of hard X-rays: thus the absorption of the probing beam by reaction gasses or window materials can be held marginal.

For the characterization of heterogeneous catalysts, the EXAFS technique is advantageous,4,20,21 since it can detect element-specifically both amorphous and crystalline structures.

In order to interrelate the structure and activity of the catalysts, the flow properties in the EXAFS cell have to be similar to those usually met in the reactor and for time resolved studies, diffusion (film and pore diffusion) should be faster than the corresponding chemical reaction.

Different designs have been reported, such as a disk of pelletized catalyst powder (also denoted as pressed wafer) placed in an environmental cell (cf. refs. 22 and 23).

Another approach is the use of a powdered catalyst in an in situ cell as, e.g., proposed by Bazin et al24. and Clausen and Topsøe.25

Moreover, Clausen et al26. and Sankar et al27. used a capillary cell for combined XRD–EXAFS studies.

This cell design has recently also been used to obtain both structure and activity data on heterogeneous catalysts.28,29

However, none of the studies compared the advantages and limitations of the different cell designs to provide truly in situ EXAFS studies, both under dynamic and static conditions.

Following the different concepts, we have constructed two kinds of in situ setups (using either an environmental chamber for a plate-like pellet/pressed wafer or a capillary cell for spectroscopic studies), which are shortly described in this paper.

Their characteristics are illustrated using exemplary case studies.

In the first case, the reduction of CuO/ZnO is investigated as an example for time-resolved studies since Cu+ species occur as an intermediate during the reduction of CuO/ZnO which is important for the activation of such catalysts.30–32

In addition, the reduction behaviour of PdO/ZrO2 at room temperature was investigated in dependence of the material used for pressing a mechanically stable catalyst pellet.

In the subsequent chapter, examples based on the PdOx/ZrO2 catalyst are presented during methane combustion at 500 °C.

Both structure and catalytic activity were determined at the same time under reaction conditions.

For comparison, reduction of PdO/ZrO2 by methane and re-oxidation by oxygen are shown.

In situ studies of such catalysts are important, since several mechanisms for the oxidation of methane have been proposed33–38 and the structure of Pd under reaction conditions is still a matter of discussion.35

Based on these examples, the setup of in situ cells in gas–solid, liquid–solid and dense fluids (e.g., supercritical fluids)–solid reactions are discussed with respect to catalyst shape, influence of external film and pore diffusion, catalyst efficiency and time-resolved studies.


Design of in situ EXAFS cells

Different in situ cell types that have been reported in literature are schematically gathered in Figs. 1a–c.

The first category of cells is based on a pressed wafer placed in an environmental chamber, which has been widely used for in situ studies.

The gas bypasses the pressed wafer either from one or from two sides.

An alternative design is shown in Fig. 1b.

In this case, the gas stream passes through the catalyst pellet instead of just by-passing it.

Another possibility is the use of a design that is similar to a plug flow reactor.

Since the wall thickness of normal glass reactors is too high, a glass tube with thin windows (0.5–2 mm diameter, 0.01 mm thick windows)—typically a glass capillary—has been proposed for such experiments.26,27

Alternative proposals using powdered catalysts have been made in literature as well.24,25,39

Based on the designs in literature, we have constructed two in situ cells that correspond to Fig. 1a (“pellet cell”) and Fig. 1c (“capillary cell”).

While the first cell is optimal with respect to spectroscopy (homogeneous sample), the second cell should be ideal for on-line catalytic studies (optimal thickness d/length l, no dead volume).

The in situ cell applicable for the investigation of self-supporting pellets has been designed according to Fig. 1a, embedding the cell inside an oven with X-ray transparent windows.

A catalyst pellet (diameter of 13 mm) was pressed (ca.

1 ton) and then loaded into the stainless steel cell.

The thickness of the pellet was ca.

2 mm, and the total X-ray path through the cell ca.

6 mm.

The gas could pass along the pellet from both sides.

From these dimensions, a dead volume of about 0.5 cm3 resulted.

Three mass flow controllers (Brooks, 0–50 ml min−1) were used to control the feed of different pre-mixed gasses through the in situ cell.

The outlet of the reactor cell was connected to a mass spectrometer (Balzers).

The whole in situ setup was mounted on an x,z,θ-table (x = translation, z = height, θ = angle to X-ray axis), directly fixed on an X95-profile (Newport) usually applied at synchrotron beamlines.

The setup according to Fig. 1c was constructed on the basis of a previous successful design26 and some more details are given in Fig. 2.

Focus was put on an in situ cell that allows both XAS studies and the determination of the catalytic performance of the catalysts.

The catalyst was loaded between two glass wool plugs in a glass reactor (quartz capillary, Markröhrchen, Hilgenberg GmbH) of 1–1.5 mm diameter.

The capillary is heated by a hot gas stream (here air, which can be substituted also by nitrogen) with a controlled air flow adjusted by a mass flow controller (Brooks, 0–2000 ml min−1).

The temperature of the heater (Tcontrol) was measured by a thermocouple at the heating element (ring cartridge, SUVAG, Zürich, 300 W/230 V) in order to control a constant temperature ramp.

The actual sample temperature (Tsample) was monitored just below the sample.

The heater was embedded in glass wool and the outer side of the oven could be cooled by water or air in order to prevent the surroundings from warming up (i.e. important for ionization chambers).

The capillary was enclosed in a Kapton cap just above the heater where the hot stream of air passed out of the oven.

Similar to the pellet cell, the outlet of the capillary was connected using gas-tight Swagelok fittings to a mass spectrometer (Balzers).

The whole assembly (in situ cell, heater) was mounted on a small x,z-table (x = translation, z = height) to align the sample in the X-ray beam.

Materials and treatment of the samples

Reduction of 30% CuO/ZnO in 5% H2/He

The 30% CuO/ZnO catalyst was prepared by controlled co-precipitation of Cu(NO3)2 and Zn(NO3)2 (precipitation pH = 9.2) followed by filtration through a 0.45 μm cellulose acetate membrane filter.40

After drying at 90 °C and grinding, 1.0 g of the precursor powder was calcined in an oven with a temperature ramp of 3 °C min−1 and a final temperature of 250 °C for 10 h.

The sample was diluted 1∶3 with boron nitride (Aldrich, hexagonal, 99.9%) during dry grinding and pressed either to a pellet of 13 mm diameter or loaded in a capillary cell (see above, 80–120 μm sieved fraction).

For the reduction, 5% H2/He was used.

Reduction of PdO/ZrO2 in 5% H2/He and re-oxidation in 2.5% O2/He

The PdO/ZrO2 material was prepared from amorphous Pd33Zr67 by oxidation in air at 400 °C.41

The sample was diluted with boron nitride (2∶7, Aldrich, hexagonal, 99.9%) or Al2O3 (2∶7, Engelhard) and pressed to a pellet of 13 mm diameter (after dry grinding).

For experiments in the capillary cell, a sieved fraction of about 80–120 μm particles (diluted ca.

1∶3 with Al2O3) was filled into a 1.5 mm diameter capillary.

The reduction was performed at room temperature in 5% H2/He, monitoring the outlet gas by a mass spectrometer.

Re-oxidation of the samples was performed in 2.5% O2/He to 500 °C with a ramp rate of 5 °C min−1.

Methane combustion over PdOx/ZrO2

In the first experiment the PdO/ZrO2 material—prepared from amorphous Pd33Zr67 by oxidation in air at 400 °C41—was heated in the reaction mixture 1% CH4/4% O2/He to 500 °C with 5 °C min−1.

Both the pellet and plug flow reactor cell were utilized using samples prepared in a similar way as for the reduction in 5% H2/He.

The space velocity was adjusted to the conversion degree of methane over the catalyst in order to study the in situ activation of the catalyst.

Also, the reaction temperature was varied between 500 and 550 °C.

Afterwards, the catalyst was reduced by 5% H2/He at room temperature.

In the next step, the reduced Pd/ZrO2 sample was studied during temperature programmed reaction in 1% CH4/4% O2/He.

XAS experiments

The reduction of CuO/ZnO was performed at the Swiss Norwegian Beamline (SNBL) at the European Synchrotron Facility (ESRF) in Grenoble, France.

The electron energy was 6.0 GeV and the ring current between 150 and 200 mA.

A Si(111) crystal was used as channel-cut monochromator.

Harmonic rejection was performed by a double-bounce gold coated mirror system.

EXAFS data were collected in the transmission mode at room and elevated temperatures.

Three ionization chambers were used for detecting the intensity of the X-rays—of the incoming I0 (filled with 3% Ar and 97 % N2, 17 cm length), of the transmitted It (filled with 50% Ar and 50% N2, 31 cm length) and of the through the reference transmitted Iref (filled with 50% Ar and 50% N2, 31 cm length) X-rays.

The spectra were taken in a step scanning mode around the Cu K-edge (for fast spectra 8.96 – 9.08 keV, 3 min per scan).

Characterization of the beam behind the sample in the capillary and the pellet cell was performed by an X-ray eye of Micro Photonics (Allentown, PA, USA).

Using the X-ray eye, the homogeneity of the samples was checked and the alignment of the in situ cell was optimised.

The experiments with PdO/ZrO2 were performed at SNBL, at ESRF, and at beamline X1 at HASYLAB (DESY, Hamburg).

In particular, for faster acquisition time (20–80 s) the QEXAFS mode at the Hamburger Synchrotronlabor (HASYLAB) at Deutsches Elektronen Synchrotron (DESY in Hamburg, Germany) at beamline X1 was applied using a Si(311) double crystal monochromator.

The typical beam current was 80–120 mA (operating positron energy at 4.5 GeV).

Higher harmonics were effectively removed by detuning of the crystals to 70% of the maximum intensity.

Three ionisation chambers filled with Ar were used to record the intensity of the incident and the transmitted X-rays (in situ reactor cell located between the first and second ionisation chamber, a Pd-reference foil for energy calibration between the second and third ionization chamber).

Under stationary conditions EXAFS spectra were taken around the Pd K-edge in the step scanning mode between 24 000 and 25 800 eV.

Faster scans around the Pd K-edge were recorded using the normal step scanning EXAFS mode (ca.

5 min per scan) or the continuous EXAFS scanning mode between 24 300 and 24 800 eV (80 s per scan, QEXAFS, cf42.).

During reduction with hydrogen faster scans were recorded (24 310–24 490 eV, 20 s per scan).

The raw data were energy-calibrated (Pd K-edge energy of the Pd-foil: 24 350 eV, first inflection point), smoothed, background corrected, and normalized using the WINXAS 2.1 software.43

In order to quantify the relative ratio of Cu2+/Cu+/Cu0 and Pd2+/Pd0, respectively, linear combination analysis of the XANES region (8.97 to 9.03 keV for the Cu K-edge and 24.33–24.45 keV for the Pd K-edge) around the edge was performed, using the spectra of the completely reduced (after hydrogen treatment) and the oxidized (as prepared) materials as model spectra.

Additionally, the XANES spectrum of Cu2O was included in the fit.

Results and discussion

Characterization of the X-ray beam behind the catalyst sample in the in situ cells

Fig. 3 gives a comparison of the X-ray beam behind the plug-flow reactor and the pellet reactor in situ cell.

Figs. 3a and 3d represent the optimal position of the in situ cells.

While the beam is homogeneous in the image of Fig. 3d (pellet cell), several shadows of the particles can be seen in the image of Fig. 3a (capillary cell).

Hence, some inhomogeneity of the sieved catalyst in the capillary cell can be seen with the X-ray eye.

Images 3b and 3c (capillary cell) show what happens if the capillary is not properly aligned: the intensity is higher at the outer region (Fig. 3b) or at the bottom of the capillary (Fig. 3c).

However, no pinhole is present as, e.g., shown in image 3e (pellet cell) where the beam bypasses the catalyst pellet.

The results show that inhomogeneities in the sample may lead to inhomogeneous penetration of the X-ray through the sample.

These inhomogeneities can be minimized using a small sieved fraction.

They lead in principle to a lower μd, and hence a lower sensitivity of the method.

This effect is not so relevant, unless the beam moves during the experiment (or the beam is inhomogeneous in intensity and the particles move).

Also, the use of a position sensitive detector, as applied, e.g., in the DEXAFS technique,44,45 may be difficult if the sieved fraction is not small enough.

For this technique, it is important to use a sample as homogeneous as possible, which is best achieved with finely ground powder pressed as a wafer, as is visible in Fig. 3d.

QEXAFS is not so sensitive to inhomogeneities in the sample since it probes the whole sample at all energies as an average32,45 and was thus exclusively used in these studies.

Reduction of CuO/ZnO in 5% H2/He

The time-resolved reduction of CuO/ZnO in the form of powder or pressed wafers has been subject to a number of studies.30–32

Here, we compare the reduction of a 30% CuO/ZnO sample (diluted in boron nitride) when using it as pressed wafer (pellet cell, Fig. 1a) or as sieved fraction (capillary cell, Fig. 1c).

Fig. 4 shows the results.

On the left, normalised spectra at the Cu K-edge are shown for illustration (note that there are less “intermediate” spectra in Fig. 4a than in 4b).

The right part of the figure shows that in both cases the reduction started at about 175–180 °C.

Accordingly, on-line gas analysis uncovered the consumption of hydrogen and the formation of water (not shown).

Although hydrogen was only consumed to a small extent during reduction (<30%) and the temperature ramp was the same (1 °C min−1) in both cases, the reduction in the capillary cell was faster than that observed with the pellet cell.

Analysis of the XANES region by linear combination of spectra originating from Cu0, Cu+, and Cu2+ showed that in both cases Cu+ is formed as an intermediate (region 8.97–9.03 keV).

The major part of the copper oxide is, however, directly reduced to Cu0.

This is in agreement with recent studies.30–32

The fraction of Cu+ is also slightly different and principal component analysis gave similar results.

Despite the low heating rate, a different behaviour was found which is attributed to the different design of the in situ cells.

The reduction of CuO/ZnO starts in both cases at the same temperature but the reduction time is markedly different.

Since hydrogen was used in excess and in the mass spectrometer a conversion of <30% was observed, the difference can be attributed to mass transfer limitations or different flow conditions in the pellet cell.

A rough estimate of the significance of the various steps of the gas–solid reaction (external film diffusion, pore diffusion, chemical reaction) can be obtained by applying a shrinking core model with constant size (thickness) of the plate-like pellet/particle.46–49

The duration for complete reduction τt can be expressed aswhere τfilm, τpore, and τchem are the times for external mass transfer, pore diffusion and the chemical reaction, respectively, necessary if the corresponding steps control the overall reaction.

Models describing the fluid–solid reaction for different rate limiting steps have been summarized by Levenspiel both for flat plates and spheres.47

Assuming a laminar flow in case of the pellet cell (200 °C, isothermal pellet), the different contributions to the complete reduction time (τt) can be calculated using the expressions developed for the shrinking core model of unchanging size for flat plates:where ρCuO is the molar density of copper in the pellet, d the thickness of the pellet, cH2 the concentration of hydrogen and km the mass transfer coefficient.

The different geometric parameters and fluid properties used for the evaluation are summarized in Table 1.

In order to estimate the mass transfer coefficient km, we consider a laminar flow across a flat plate of length L = 13 mm, which can be described and analytically solved using a boundary layer model50 resulting in an average mass transfer coefficientwhere ReL is the Reynolds number for a plate of length L, Sc the Schmidt number and DH2 the diffusion coefficient of H2 in He (Table 1).

By using eqn. (2), a time for complete reduction of 6.5 s is estimated if external mass transport (film diffusion) is rate-determining.

Thus, diffusion through the gas film cannot be the limiting step during reduction of CuO/ZnO in the pellet cell.

The control by the chemical reaction cannot explain the difference between the two reaction cells either.

Thus, pore diffusion can be assumed to be the limiting step.

Pore diffusion coefficients in porous solids depend on the porosity of the solid and slightly on the temperature.

Assuming an effective diffusion coefficient De between 10−6 and 10−8 m2 s−1 and a pellet thickness of d = 2 mm, reduction times can be calculated according to:47(disk of 2 mm diameter, molar density of copper in the pellet ρCuO = 67 mol m−3, concentration of hydrogen cH2 = 1 mol m−3, compare with Table 1).

For De = 10−6 and 10−8 m2 s−1 reduction times of 33 s and 56 min, respectively, can be calculated and thus τt can strongly depend on the pore diffusion.

In comparison, using a sieved fraction of 100 μm in a capillary microreactor under continuous flow conditions the terms for both film diffusion and pore diffusion are less significant as the following calculation shows.

Applying the expressions for spherical particles (constant particle size), the contribution from film diffusion in a plug flow reactor can be calculated as:where Rp is the radius of the spherical particle.

Mass transfer coefficients (km) for fluids flowing around spheres in packed-bed reactors have been widely estimated for different cases (e.g., refs. 50–52).

For Reynolds numbers in the range 0.1 < Rep < 50 applied in this study (Table 1), a suitable relation is reported in ref. 50 and 52:

The calculation of the mass transfer coefficient from the Reynolds number and the Schmidt number is given in Table 1.

Calculation of the reduction time, assuming that film diffusion is rate-limiting (eqn. (5)), results in a τfilm of 1.7 ms.

Assuming pore diffusion instead of film diffusion being the rate limiting step, the corresponding reduction times for round-shaped particles can be calculated according to:

Also, these reduction times are much smaller with 0.02–2 s for a diffusion coefficient De of 10−6–10−8 m2 s−1, respectively.

As long as the CuO/ZnO reduction takes more than 1 s at a De > 10−7 m2 s−1, the rate-limitation results from the chemical reaction.

Hence, the differences between the two in situ cells is expected to be due to pore diffusion effects.

Note that here we assumed different pore diffusion coefficients for hydrogen.

The same accounts for the gaseous product (water formation) and it has recently been shown that the reduction of CuO/ZnO is strongly affected by the presence of water.53

Hence, also the pore diffusion of water could be limiting.

In addition, we can conclude from these estimations that for gas–solid reactions in the subsecond region,4,32,54,55 intermediates may not only be a consequence of the reaction itself but also be indirectly affected by film and/or pore diffusion.

This is, in particular, the case if the flow properties are not as good as in a plug flow reactor or the particle diameter is too high.

Moreover, differences can be more significant if larger molecules and, in particular, liquids are used (smaller diffusion coefficient).

Both in situ cells can give similar results during temperature programmed reaction if the ramp rate is small, the flow properties are good and the pore diffusion of the molecules is sufficiently fast.

The latter is also strongly dependent on the material used to dilute the catalyst sample as shown in the next part.

Comparison of differently prepared PdO/ZrO2 samples during reduction

Reduction of an oxidized PdO/ZrO2 catalyst by hydrogen turned out to lead to a more active methane combustion catalyst.41,56

Reduction occurs very fast at room temperature.37

Fig. 5 shows the reduction of PdO/ZrO2 (capillary cell, diluted 1∶3 with Al2O3) in 5% H2/He using the QEXAFS (20 s per scan) mode.

On-line gas analysis showed that the response time inside the reactor is <5 s, while the reaction occurs over a time interval of 2 min.

The XANES region (24.33–24.45 keV) of all spectra could be reconstructed from the spectrum of the reduced and the oxidized form of the sample.

The fraction α of oxidized or metallic palladium is shown in Fig. 5b.

Note that some palladium hydride formed as well, since EXAFS analysis indicated the formation of some Pd hydride due to an increased Pd–Pd distance to 2.80 Å.

The analysis in Fig. 5b reveals that the reduction occurs via an induction and acceleration period.

Fig. 6 shows the comparison of two different pellets in the pellet cell.

Using a pressed wafer of PdO/ZrO2 diluted once with BN (thickness d = 1.5 mm) and once with Al2O3 (thickness d = 2 mm), shows a striking difference in the reduction behaviour.

The same reduction conditions were used in both cases and also the response time of the pellet cell was only a few seconds.

Nevertheless, both the start of the PdO/ZrO2 reduction, as well as the reduction rate, are negatively affected in the case of the dilution with boron nitride.

This shows that not only the catalyst shape (sieved catalyst powder vs. pellet) but also the material plays an important role.

BET measurements of the reduced pellets revealed that the surface area and the pore structure of the BN- and Al2O3-pellet are significantly different.

While the specific pore volume of the used Al2O3 pellet is 0.327 cm3 g−1 (the BET surface area is ca.

133 m2 g−1), the pore volume of the BN pellet is only 0.036 cm3 g−1 and the BET surface area 2 m2 g−1.

Hence, using BN for pressing the pellets led to pellets with significantly lower porosity, and these results suggest the use of high porosity materials.

The oxidation state of palladium during catalytic combustion of methane over PdOx/ZrO2

Several mechanisms concerning the oxidation of methane over Pd/ZrO2 catalysts have been proposed.35,37

Interestingly, Pd/ZrO2 (pre-reduced in hydrogen) performs better than PdO/ZrO2 in this reaction.56

Hence, the question arises on the role which metallic Pd plays in the combustion of methane.

Once again, results over PdOx/ZrO2 in a pellet and a capillary cell are considered.

As a basis, the re-oxidation of a (reduced) Pd/ZrO2 sample in 2.5% O2/He and reduction in 1% CH4/He are discussed.

Re-oxidation of Pd/ZrO2 in oxygen and reduction of PdO/ZrO2 in methane

Figs. 7 and 8 show the oxidation of Pd/ZrO2 in 2.5% O2/He and the reduction of PdO/ZrO2 in 1% CH4/He, respectively.

While reduction of PdO/ZrO2 in hydrogen already occurs at room temperature (Figs. 5 and 6), the re-oxidation only partly occurs at room temperature.

Interestingly, this oxidation has not been observed by TPO on Pd/ZrO2.57

Some shift in the edge position was observed as well, indicating the decomposition of the Pd hydride phase (also supported by EXAFS analysis).

Above 320 °C, most of the palladium is re-oxidized and XANES/EXAFS analysis reveals no detectable metallic palladium after oxidation at 500 °C, supported by thermal analysis (not shown).

The reduction by methane also occurs only at higher temperatures (>275 °C) and results in a completely reduced catalyst.

No reduction took place up to 275 °C.

Above this temperature, a very sudden reduction within 1 scan (50 s per scan) was observed.

It seems that once methane can be activated on the PdO/ZrO2 catalyst a fast reaction occurs.

PdO was rapidly reduced once metallic particles started to be formed, which seem to catalyse the activation of methane.

This is in agreement with temperature programmed reduction and Raman studies,57 where it was also found that a certain temperature is needed to activate methane.

Once metallic palladium is formed, dissociative adsorption of methane is enhanced (on the metallic Pd particles) and further reduction occurs.

The fact that an autocatalytic reduction occurs at 275 °C is also supported by the fact that dissociative adsorption of methane is reported to occur already at 200 °C58.

Reaction in 1% CH4/4% O2/He

The oxidation behaviour of Pd/ZrO2 during temperature programmed reaction in 1% CH4/4% O2/He (Fig. 9) uncovers in the low temperature region (<350 °C) a similar behaviour as in 2.5% O2/He.

The methane combustion (according to on-line gas analysis) sets in at 300–320 °C.

Likewise, in the reduction of the CuO/ZnO, both the plug-flow reactor cell (based on the capillary) and the cell using a pelletized sample (disk-shape) were compared.

In both setups, methane combustion was more effective over Pd/ZrO2 (pre-reduced) than PdO/ZrO2 (oxidized form by direct oxidation of amorphous PdZr alloy).

During the temperature programmed reaction the conversion of methane over Pd/ZrO2 reached already its maximum well below 500 °C, whereas only partial conversion of methane was observed over the PdO/ZrO2.

As Table 2 summarizes, the conversion over the Pd/ZrO2 pellet only reached about 80%, while total conversion was observed over Pd/ZrO2 in the capillary cell.

This is traced back to the dead volume of the pellet cell, allowing some of the methane to pass through the cell without contact with the catalyst (by-passing).

The observed conversions and rates are summarized in Table 2.

In both cases, it was observed that PdO/ZrO2 was not reduced to an extent measurable by EXAFS/XANES, while starting from Pd/ZrO2 evidenced that Pd was not completely re-oxidized or possessed a different structure (Fig. 10), both if either the pellet or the glass capillary cell was used.

The “whiteline” in the XANES spectra was slightly different and EXAFS analysis indicated a different number of coordination shells in the PdO/ZrO2 and the Pd/ZrO2 catalyst after methane combustion at 500 °C.

While the PdO/ZrO2 sample had mainly contributions at 2.02 Å (Pd–O) and at 3.02 Å (Pd–Pd in oxide), a significantly higher contribution at 2.76 Å (Pd–Pd shell) was found in the Pd/ZrO2 catalyst (cf. more details and similar results in ref. 37).

Hence, the findings on the structure of the catalyst are similar in both in situ studies, though the catalytic activity is different but with the same tendency in the two cells.

Nevertheless, the significantly lower methane oxidation rate over the catalyst pellet compared to the corresponding particles in the capillary reactor requires an explanation.

Comparison of the effectiveness of the catalyst in a plug flow and a pellet in situ cell

One reason for the lower catalytic activity of disk-shaped catalyst samples compared to the sieved catalyst powder is by-passing of the gas due to the dead volume in the pellet reactor cell.

This would, however, still lead to a representative result for both the catalytic activity and the active species if the whole catalyst contributes to the catalytic activity.

In order to verify this hypothesis for methane combustion, the influence of pore diffusion in both kinds of cells is considered.

The pore diffusion resistance can dramatically influence the effectiveness of heterogeneous catalysts.

The phenomenon has been intensively investigated in kinetics and reaction engineering studies in heterogeneous catalysis.47,59,60

This concept can be directly transferred to estimate the influence of the catalyst shapes (porosity, characteristic length) in the two cell designs used in the present study.

According to Thiele61 and Aris,62 for a plate-like pellet and a first order reaction the effectiveness factor η can be described in terms of the Thiele modulus ψ47

For a plug flow reactor, the equations are correspondingly given (see Table 3), and for a rough estimation the same relation as given in eqn. (8) can be used with a characteristic length: Lc = dp/6.

For estimating the effectiveness factor in the methane oxidation reaction over a cylindrical platelet of thickness d = 2 mm and over spherical particles of dp = 80 μm both the reaction rate and the diffusion coefficient were taken from a recent work on methane combustion over Pd/Al2O3 catalysts by Groppi et al.63

The calculations summarized in Table 3, show that the effectiveness factor of the plate-like pellet is only 0.03%, which means that due to pore resistance only a thin layer on each side is contributing to the reaction.

For an efficiency of 50% over a Pd/ZrO2 catalyst pellet, a very thin platelet of d ≤ 0.06 mm is required.

Fig. 11 shows the estimated concentration gradient in the pellet, it dropped by 90% already at 70 μm.

If this drop in reactant concentration can be tolerated in the middle of the pellet, a Thiele modulus of about ψ = 3 is required, corresponding in the case of a flat plate to a thickness of d = 160 μm.

For the capillary cell the effectiveness factor is much larger but still only 88% of the catalyst are effectively contributing to the reaction (due to the fast reaction rate).

The concentration gradient of the reactants inside a catalyst particle is flat, as Fig. 11 shows.

Hence, the structure monitored by the bulk technique EXAFS is particularly representative if the catalyst is used as powder.

For pellet-like shapes a more surface-sensitive technique (e.g., electron yield XAS, grazing incidence XAS) would be required.


The experiments in this study have shown the importance of appropriate cell designs for in situ X-ray absorption spectroscopy and related operando spectroscopic techniques.

The EXAFS technique offers a number of opportunities in the field of time-resolved studies and for combining structural analysis with catalytic activity measurements.

Here, the reduction of CuO/ZnO and PdO/ZrO2 was uncovered in a time-resolved manner and the structure of PdOx/ZrO2 was monitored during methane combustion.

The reduction of CuO occurs via Cu+ species, as also supported in .refs. 30–32

However, it seems that the reduction is dependent on the cell design/sample preparation which may also influence the occurrence of Cu+ species.

In case of methane oxidation over PdOx/ZrO2 catalysts, the pre-reduced catalyst was more active under reaction conditions, in accordance with literature, and some of the palladium remained in reduced state at 500 °C under reaction conditions.

Reduction of PdO/ZrO2 to Pd/ZrO2 by hydrogen occurred within 5 min at room temperature, while re-oxidation was only to a small extent observed up to 250 °C.

Above this temperature, the palladium constituent was continuously and completely oxidized.

These studies support the conclusions in literature that reduced palladium is one of the active species in methane combustion.37,57,64

Studies with an X-ray eye uncovered that the X-ray intensity behind a sample pressed as pellet was more homogeneous than behind a catalyst sample used in a sieved powder fraction.

Thus, a pellet will lead to a better signal/noise ratio (due to a higher μd), in particular, if the position of the beam varies slightly.

The use of pellets also allows to irradiate a larger area resulting in a better signal/noise ratio.

A further limitation of the capillary cell used in this study is that the temperature is measured below the catalyst sample and not in the catalyst bed.

Temperature measurement in the catalyst sample may become of paramount importance for highly exothermic or endothermic reactions, particularly at high conversion.

Moreover, the catalyst composition may change in such a plug flow reactor, so that local probing of the catalyst composition over the catalyst bed would be interesting as we recently proposed.65

Effects due to pore diffusion resistance can be encountered both in time-resolved studies and in catalytic studies, in particular in the case of pressed wafers.

Eqns. (2), (4), (5) and (7) show that time-resolved changes on heterogeneous catalysts depend on the gas–solid mass transfer coefficient and the effective diffusion of the reactants/products in the solid.

Effects caused by pore diffusion limitations can be eliminated or at least mitigated by the use of macroporous materials and/or small particles or very thin pellets.

Our study has shown two extreme cases: the pore structure/volume of BN and Al2O3 are very different.

This shows that high porosity materials, such as Al2O3 used in this study, are preferable.

Since boron nitride is a favoured material in X-ray absorption spectroscopy due to the light elements, the use of a high porosity form would be of interest for spectroscopic purposes—here hexagonal BN with low porosity was used.

Nevertheless, in slow gas–solid reactions the pore diffusion is relatively fast compared to the chemical reaction rate so that using a disk may be preferred due to the better signal/noise ratio in spectroscopy.

We suggest therefore a comparison of the measured reaction time τt with the time τfilm expected if film resistance is rate-limiting or with the time τpore if pore diffusion is limiting, using eqns. (2)–(7), respectively.

These effects become even more relevant in liquid/solid reactions.

Recently, we observed in the reduction of PdO/Al2O3 by benzyl alcohol that the reduction of palladium occurred within less than 2 min using PdO/Al2O3 as powder, while the reduction over a PdO/Al2O3 pellet took more than 45 min.

Assuming a diffusion coefficient of De = 1 × 10−9 m2 s−1 of the alcohol in a fluid and a typical viscosity of 0.33 × 10−3 m−1 kg s−1, it already turns out with eqns. (2) and (5) that the external film resistance (fluid–solid) in liquid phase is much higher than in gas phase.

In case of film diffusion controls the reduction, some example calculations are given in Table 4 for reduction of a catalyst pellet (platelet) and a catalyst powder (spherical particles) of 5% Pd/Al2O3 in toluene.

Also, pore diffusion resistance is supposed to be considerably higher than in gas phase.

Hence, combination of on-line catalytic monitoring combined with EXAFS analysis has to be performed on finely sieved catalysts, as recently reported in refs. 66 and 67.

Another possibility is the use of shell-impregnated catalysts.

We applied this approach in the field of supercritical (or dense) fluids, since sufficiently large particles are required for proper flow conditions.68

A recent ATR-IR study69 has proven that the supercritical fluid inside the pores has a similar density to the corresponding liquid and it can thus be assumed to have a similar effective diffusion coefficient as in liquid carbon dioxide.

The fluid–solid mass transfer coefficient is, however, much higher than in liquid phase.

Another important consequence of the presented influence of film and pore resistance is the appropriate design of cells for time-resolved studies on heterogeneous catalysts.

Both the QEXAFS and DEXAFS techniques are used with the aim to identify possible intermediates and/or reveal the mechanism of reduction/oxidation.

Considering the reduction of a catalyst by hydrogen on a pressed wafer, it turns out according to eqn. (2) that studies on the second scale will even be influenced by the gas film diffusion using the in situ cell sketched in Fig. 1c.

The calculations using a plug flow reactor with catalyst particles of mean particle size of 100 μm show that in this case the response time is ca.

1 ms.

Hence, 1 ms seems to be the lower reasonable limit of time-resolved studies in a fluid/solid reaction.

Not only during time-resolved studies but also in case of reactions, such as methane oxidation, pore resistance may influence the spectroscopic results, in particular if the diffusion of reactants inside the pores is slow compared to the chemical reaction.

In this case, only a fraction of the catalyst may be involved in the reaction.

Though methane oxidation is a rapid reaction, such limitations by pore/film diffusion have also been observed in other studies.70

A possible solution may be the use of electron yield XAS.71

However, special care has to be taken concerning the influence of the reaction mixture on the detection technique.

Another possibility is once again the use of particles with small characteristic length Lc.

It is suggested to calculate the Thiele modulus ψ

If ψ < 0.5, the effectiveness factor is higher than 90%.

For ψ > 3 the reactant concentration drops to less than 10% inside the particle.

Thus sufficiently thin pellets or particles (resulting in a small characteristic length Lc) will lead to η > 0.9.

In the case when the reaction rate kr cannot be estimated, the Weisz modulus can be used instead as well known from mass transfer literature (cf. ref. 59).

As pointed out above, also shell-impregnated particles can be used for the studies.

In this way, operando studies become feasible not only for gas/solid but also for liquid/solid or dense fluid/solid reactions.


In summary, the different examples show that, as in kinetic studies, special care has to be taken to account for possible film and pore diffusion limitations during time-resolved or operando spectroscopy studies.

Limitations due to film and pore diffusion may occur due to flow conditions, catalyst shape, the chosen reactor design or material to dilute the catalyst to an optimal absorption length μd.

Low porosity materials, such as hexagonal boron nitride, are not ideal for this purpose, even though samples can be easily pelletized and its low absorption coefficient of hard X-rays makes it a preferable material for spectroscopy.

The use of a high porosity material is desired.

Both the cell design and the catalyst system need to be adapted to the reaction of interest (kind of fluid, reaction rate regime, diffusion coefficient, pore size distribution, etc.).

In order to find a balance between optimal design for spectroscopy and catalysis we suggest to estimate the importance of film diffusion (τfilm), pore diffusion (τpore) and/or the effectiveness factors (Thiele modulus), as discussed in this work.

These criteria are important for describing and estimating the reaction conditions during operando spectroscopy and time-resolved studies, not only in X-ray absorption spectroscopy but in situ spectroscopy in general such as XRD, Raman or IR spectroscopy (transmission mode).

The discussed criteria will make both in situ spectroscopic and time-resolved studies more powerful for elucidating the working mechanism of heterogeneous catalysts, as exemplified here for two well-known catalyst systems (Cu/ZnO for methanol synthesis and PdOx/ZrO2 for methane combustion).