Tables for
Volume I
X-ray absorption spectroscopy and related techniques
Edited by C. T. Chantler, F. Boscherini and B. Bunker

International Tables for Crystallography (2022). Vol. I. Early view chapter

Quick EXAFS studies in catalysis

M. Nachtegaala* and R. Frahmb

aPaul Scherrer Institute, Villigen-PSI, Switzerland, and bDepartment of Physics, University of Wuppertal, Wuppertal, Germany
Correspondence e-mail:

QEXAFS (quick extended X-ray absorption fine structure), or alternatively quick EXAFS, spectroscopy refers to X-ray absorption measurements with a time resolution of a few seconds down to several milliseconds. This is achieved by the fast continuous scanning of a double-crystal monochromator through the Bragg angle and thus a certain energy range. This chapter describes selected applications of state-of-the-art QEXAFS. QEXAFS spectroscopy allows study of the rate of structural change and reaction mechanisms in dynamic systems such as catalysts or during the synthesis of materials. This is illustrated by three examples: identifying the rate-limiting step in the low-temperature reduction of harmful nitrous oxides (as found in diesel-fueled cars) over a copper zeolite, understanding the deactivation mechanism of a nickel-based CO2-methanation catalyst due to intermittent cutoff of renewable hydrogen in the feed, and finally following the growth mechanism and kinetics of iron–organic nanoaggregates under different environmental conditions.

Keywords: QEXAFS; catalysis; synthesis; environment.

1. The need to add time resolution to XAS

X-ray absorption spectroscopy (XAS) has been used since its initial development to monitor the electronic and local geometry of (electro)catalysts (Lytle et al., 1974[link]). Laboratory-based X-ray spectrometers have developed to such an extent that X-ray absorption and emission spectra can now be obtained from concentrated samples. Using such laboratory-based spectrometers, EXAFS spectra can be obtained in less than one hour and up to k = 9 Å−1 (Seidler et al., 2014[link]), even to determine structure–activity relationships in catalysts (Padamati et al., 2017[link]). Fast time-resolved XAS, however, is exclusively available at third-generation synchrotron light sources. Using two optical schemes that evolved in different directions, i.e. energy-dispersive XAFS and QEXAFS, time-resolved EXAFS spectra with good signal to noise and up to k = 12 Å−1 can now be achieved with a temporal resolution of 10 ms (Müller et al., 2016[link]). When faster time scales are needed, the energy range covered is limited to the XANES region. With the general trend of synchrotron facilities upgrading to diffraction-limited light sources, energy-dispersive XAS becomes challenging (Abe et al., 2018[link]), leaving QEXAFS as the method of choice for the study of dynamic systems.

It is the use of QEXAFS that allows the study of dynamic processes, such as the synthesis of materials or (electro)catalytic processes. When EXAFS experiments are coupled with mass spectrometry, gas chromatography or gas-phase infrared spectroscopy (techniques that allow monitoring of the activity or selectivity of the material of interest, for example a catalyst) so-called structure–activity/performance relationships can be determined, where the structure observed with EXAFS reflects the active site (Clausen & Topsøe, 1991[link]). More recently, it has been established that the dominant structure observed with EXAFS (which is a bulk averaging technique) under reaction conditions might not always be representative of the catalytic active site. One way to enhance the sensitivity of XAS to the catalytic active site is to stimulate structural change in the active site by concentration-modulation experiments, in which reactants are periodically added to or removed from the reactor feed. Simultaneously, QEXAFS spectra are collected with sufficient time resolution, which is typically in the subsecond range. In the phase-sensitive detection approach the time-resolved XAS data set is subsequently filtered with the excitation frequency, i.e. the frequency at which reactants are periodically added to or removed from the reactor feed. As a result, only the XAS signal remains, which changes with the excitation frequency, whereas the signal of noise and spectator species (the bulk of the material) is filtered out (see, for example, Ferri et al., 2011[link]; König et al., 2012[link]). This remaining signal is then representative of the structure that changes upon changing the reaction conditions. Another powerful method to extract information from such large time-resolved QEXAFS data sets is the use of multivariate curve resolution (MCR) as a blind-source separation method which does not need a priori knowledge of all of the spectral components that make up the XAS data sets. This method allows the extraction of the `pure' spectra and their corresponding time-dependent concentration profiles (see, for example, Voronov et al., 2014[link]; Rochet et al., 2017[link]).

Another benefit of time resolution in XAS is that it allows the determination of reaction mechanisms, including the rate-limiting step. The group of S. T. Oyama (Bravo-Suárez et al., 2008[link]) showed that the observation of a specific structure by EXAFS under operating conditions does not necessarily mean that this structure is involved in the catalytic reaction mechanism. This structure, resembling a particular structural site, might well be a spectator species to the catalytic reaction. Reactant cutoff or addition experiments are then combined with QEXAFS measurements, which allow measurement of the rate of structural change. If the rate of structural change determined by XAS corresponds to the rate of the overall reaction, as measured for example using a mass spectrometer, then this structure, which is alternatively referred to as the active site, is involved in the rate-limiting step. When the structure of the rate-limiting step is known, attempts can be made to rationally improve the catalyst by changing this structure in order to lower its energy barrier and speed up the reaction.

2. QEXAFS: state of the art

The QEXAFS method (Frahm, 1988[link], 1989[link]) allows fast transmission XAS measurements for `standard' samples and also enables time-resolved fluorescence XAS measurements (Clark, Steiger et al., 2020[link]) for the study of the element of interest either in a heavy absorbing matrix or at low concentrations.

The time resolution covers seconds to milliseconds. The most advanced current QEXAFS monochromators use torque motors that allow oscillating movements in the Bragg angle at 50 Hz and more (Nonaka et al., 2012[link]; Müller et al., 2015[link], 2016[link]). For 95% of practical applications, however, the range from 1 to 10 Hz is sufficient. Here, the monochromators can be operated continuously for the duration of the experiment. The data quality is similar to or better than that obtained on beamlines that collect EXAFS spectra in step-scan mode: the data quality at high-intensity sources mostly depends on the sample quality. At 10 Hz, i.e. 50 ms per spectrum, wide energy ranges can be covered, spanning for example both Cu and Zn K-edge EXAFS ranges. Meaningful scientific investigations at high speed also require very fast changes to be induced in the sample, which are difficult to obtain in reality. Changes are limited by the speed of gas diffusion, temperature gradients, movement of grains in the sample by changes of gas flow etc. This limits the time resolution to 100–50 ms for QEXAFS experiments on most real samples. Using the full capabilities of the QEXAFS method is a major challenge for the design of in situ sample cells, which cannot be underestimated.

For some experiments, a combination of QEXAFS with X-ray diffraction (XRD) and small-angle X-ray scattering (SAXS) is desirable. Whereas XAFS probes the local atomic short-range order and valence states, XRD can follow the growth of crystallites, which could explain, for example, the observed reduction of catalytic activity. Such a combination of techniques was developed quite early on (Clausen et al., 1998[link]) and is now available at different sites (see, for example, Sekizawa et al., 2017[link]). Fluorescence measurements for the detection of dilute elements have progressed from initial work (Frahm, 1991[link]; Lützenkirchen-Hecht et al., 2001[link]) with ion chambers or PIN photodiodes as detectors to the use of large-area PIPS diodes (Clark, Steiger et al., 2020[link]). Furthermore, due to the availability of dedicated monochromators at high-intensity sources, methods such as element-specific tomography, with a spatial resolution in the 10 µm range, which was first demonstrated on an undulator beamline at the Advanced Photon Source in Argonne, USA (Schroer et al., 2003[link], 2010[link]), can be further developed for investigations that need nanometre resolution. Whereas only a few QEXAFS setups were tested during the first 15 years, the number of beamlines with permanently installed QEXAFS setups has been rapidly expanding over the last 15 years, which will stimulate scientific competition and, with that, further progress.

3. QEXAFS examples

The first QEXAFS experiments were performed on copper-based methanol catalysts (Clausen et al., 1998[link]). Due to the nature of the scientific questions, the catalysis research community has been the major user of the QEXAFS technique ever since (Grunwaldt & Frenkel, 2009[link]; Nachtegaal et al., 2016[link]). QEXAFS is especially helpful to follow very dynamic catalytic systems, which include gas after-treatment systems for cars and trucks and electro-catalysis, where H2 is generated from water depending on the availability of electricity from intermittent renewable energy sources such as wind and solar power. To obtain insight into the catalytic mechanism under such dynamic conditions, including the rate-limiting step, temperature, voltage-jump or reactant cutoff/addition experiments have been designed and coupled to QEXAFS studies.

The first example selected here is from a gas after-treatment system for diesel-fueled cars, in which catalytic converters are used together with the intermittent feeding of the selective reductant ammonia to convert toxic nitrous oxides such as NO to N2 and H2O. These catalysts need to be stable under hydrothermal conditions up to high temperatures. Simultaneously, as car engines have become more and more fuel-efficient, the catalysts need to be active at low temperature. A catalyst that meets these criteria and has recently found its way into commercial applications is the small pore-size zeolite (SSZ-13) loaded with 1–3 wt% of copper. The conversion rate of such a catalyst as a function of temperature is shown in Fig. 1[link], where it is obvious that full activity is only achieved at higher temperatures. When driving a car under realistic conditions, the catalyst is exposed to varying temperatures and compositions of the feed gas due to the constantly changing driving requirements of the car and due to the intermittent supply of ammonia, which assists the reduction of NO.

[Figure 1]

Figure 1

NO conversion rate of a Cu–SSZ-13 catalyst measured in a test reactor for monolithic catalysts for an optimized NH3 flow (adapted from Marberger et al., 2018[link]).

As can be seen from Fig. 1[link], this catalyst is not very active at low temperatures, for example when the vehicle is cold-started. To understand why this catalyst did not achieve high NO conversion at low temperature, reactant cutoff and addition experiments were designed to simulate the intermittent supply of ammonia to the catalyst as occurs in real car exhaust-gas applications. Simultaneously, the time-resolved change in the catalytic structure was measured with QEXAFS. The resulting spectra after switching off the selective catalytic reduction (SCR) by NH3 cutoff from the initial gas composition (1000 p.p.m. NH3, 1000 p.p.m. NO, 6 vol% O2, 2 vol% H2O made up with N2) are shown in Fig. 2[link]. The re-oxidation of copper(I) to copper(II) is evident.

[Figure 2]

Figure 2

Time-resolved QEXAFS data set collected at the Cu K edge after NH3 cutoff, where the first spectra are coloured black and the last spectra are coloured green (adapted from Marberger et al., 2018[link]).

This large set of time-resolved XAS can be fitted by a linear combination approach, which requires that the individual components that make up the data set are known and reference spectra are available for all components. If this is not the case, the large data set might be analyzed by using a combination of principal component analysis and MCR alternating least-squares methods to determine the number of components present in the data set and reconstructing their spectra to fit the time-dependent speciation of copper. Fig. 3[link] shows the time-resolved copper speciation determined by linear combination fitting together with the most important species determined by a mass spectrometer during an NH3 cutoff experiment. The initial gas composition is the same as that given above. QEXAFS spectra were collected here at 1 Hz, i.e. 500 ms per spectrum.

[Figure 3]

Figure 3

Changes in the catalyst after NH3 cutoff at t = 0 s, measured at 225°C. Top: dynamic copper speciation derived from linear combination fitting of copper(I) (black) and copper(II) (red) solvated in ammonia. Also shown is copper bound to the zeolite structure (brown) and copper(II) nitrate species (blue). Bottom: mass-spectrometer traces of NH3 and NO (adapted from Marberger et al., 2018[link]).

The most striking feature observed in this graph is a decrease in the amount of NO detected at the end of the reactor immediately after switching off the selective reductant ammonia, which suggests that upon removing ammonia from the feed the catalyst is temporarily more active. Simultaneously, a dip in the presence of copper(I) and the temporary presence of copper(II) species is observed in the linearly fitted XAS spectra. These data suggest that in the low-temperature regime the reoxidation of copper(I) to copper(II) is hindered by ammonia, slowing the overall selective catalytic reduction. With this knowledge in hand, subsequent tests on a catalytic test bench showed that reducing the amount of ammonia fed intermittently to the catalyst at low temperature while increasing the amount of ammonia fed at higher temperature could double the NO conversion of this catalyst at low temperature and increase the NO conversion over the complete temperature range (Marberger et al., 2018[link]). QEXAFS was able to uncover this unexpected, complex behaviour at the atomic level in a unique way.

The second example comes from a different dynamic catalytic system. In the power-to-gas process, power in the form of electricity is used to split water into hydrogen and oxygen. The emerging H2 is then reacted with CO2, recovered for example from the exhaust gas of a cement plant, over a nickel catalyst to form methane. Since the electricity for this process is not always available from renewable sources due to changing wind or solar radiation conditions, the catalysts will experience dropouts in H2 availability and thus be constantly exposed to dynamic conditions. An experiment was designed that mimics these dropout events by periodically interrupting the H2 supply to the reactant feed for 30 s at a time. The time-resolved response of the nickel-based methanation catalyst in the middle of the catalyst was monitored simultaneously with QEXAFS. The time-resolved data set was then linearly fitted with oxidic and metallic references. The results are summarized in Fig. 4[link].

[Figure 4]

Figure 4

Methanation of CO2 using renewable H2 during dynamic operation, switching every 30 s between methanation conditions (H2/CO2 = 4) and CO2. The figure shows the valve signal in the upper part (black), the methane signal of the mass spectrometer in the middle part (green) and the fraction of reduced (blue) and oxidized (red) nickel from linear combination analysis of the XANES spectra. The numbers in circles count the H2 dropouts (reproduced from Mutz et al., 2017[link]).

Fig. 4[link] shows that during the first six cycles in which hydrogen was periodically cut off no effects were observed on the speciation of nickel in the middle of the reactor, i.e. the methanation catalyst stayed nicely reduced. During the seventh cycle partial oxidation of nickel was observed during hydrogen cutoff, which was attributed to the few hundred p.p.m. of oxygen present in the technical CO2 feed. When hydrogen was added back to the feed the catalyst could not be fully reduced back. This oxidation of the nickel catalyst, inherent to the absence of the reducing hydrogen in the feed, deactivates the catalyst, as seen in the mass-spectrometer traces. However, a slight deactivation was also observed during the first cycles. This was attributed to oxidation of the catalyst at the beginning of the reactor bed. Since the XAS spectra were collected in the middle of the catalyst bed, it took a few cycles before the oxidation front moved through to the reactor part where the XAS data were collected. Further detailed evaluation of the full EXAFS spectra showed surface-oxidation/reduction processes, in which the core of the nickel particles remained reduced (Mutz et al., 2017[link]). These results demonstrate how well-designed time-resolved EXAFS measurements can help to improve the efficient operation of a catalyst bed by understanding the processes at the atomic level.

The last example comes from environmental science and is representative of QEXAFS studies that focus on understanding and controlling material synthesis in the liquid phase. In natural water, iron and organic matter aggregates are abundant and control the fate of inorganic pollutants such as heavy metals. Little is known about how their complex structure varies as a function of environmental conditions such as O2 concentration and pH. The organization of this complex structure has a large influence on the availability of metal-binding sites. Vantelon et al. (2019[link]) followed the various iron phases formed, their growth processes and kinetics upon the oxidation/hydrolysis of initial iron(II)–humic acid (HA) complexes with QEXAFS. Such humic substances are organic compounds which are the major organic fraction of soil and constituents of many streams, lakes and ocean water.

The kinetics of iron(II) oxidation in the presence of HA is different under oxic (air) or anoxic (N2) conditions, as shown in Fig. 5[link](a), where component 1 is representative of iron(II) species and components 2 and 3 are representative of iron(III) species. Fig. 5[link](b) shows the advantage of collecting large XANES data sets: by applying MCR analysis to large XAS data sets, three `pure' components could be identified and their evolution over time plotted (Fig. 5[link]a). EXAFS analysis of these three components (Figs. 5[link]c and 5[link]d) allowed the structure of the three phases that were present to be obtained. The starting phase is consistent with iron(II) ions in solution which are partially complexed by HA. This starting phase disappears rapidly as iron is oxidized. In the absence of organic matter (HA), lepidocrocite, an iron oxide–hydroxide mineral, would be formed, which is not the case in the presence of organic matter. In the presence of HA, nanosized ferrihydrite bound to organic matter is formed (component 3). Component 2 is similar in structure to component 3, but the fitted component is rather representative of oligomers of individual iron octahedra bound to HA via carboxylic groups. The ratio of the two phases formed (components 2 and 3) was shown to depend on the pH and has large implications for the capability of these complexes to trap metal and metalloid pollutants, where the ferrihydrite nanoparticles and oligomers of octahedral iron, both bound to organic matter, contain significantly different pollutant absorption sites.

[Figure 5]

Figure 5

(a) Concentration of iron components (Cpt) extracted by MCR analysis for experiments performed in air (Exp-Air) and in N2 (Exp-N2). (b) Reconstructed XANES spectra of the pure components derived from MCR analysis. (c, d) The corresponding EXAFS spectra and Fourier transforms, respectively, plotted with the fit results shown as dotted lines. Ferrihydrite (Fh) and lepidocrocite (Lp) spectra are shown as references (adapted from Vantelon et al., 2019[link]).

This study showed how QEXAFS data in combination with chemometric tools provided an unprecedented insight into how the structure of Fe–HA complexes evolves as a function of pH and O2.

4. Conclusions

During the last decade, permanent and dedicated QEXAFS setups for in situ and operando research have been installed at an increasing number of synchrotron-radiation facilities. This has led to the development of easy-to-use software packages for analyzing large quantities of XAS spectra (see, for example, Clark, Imbao et al., 2020[link]) and thus to QEXAFS becoming a method that is used by a growing number of user groups from different scientific communities. QEXAFS is well versed in tracking dynamic reactions and has been used in catalysis research since its first development. More recently, QEXAFS has become a valuable tool for understanding dynamic processes in various scientific communities, including catalysis, environmental chemistry, material synthesis and energy storage and conversion technologies. This is based on easier access to QEXAFS and methods and tools that can extract a maximum of information from very large data sets. As shown by the examples in this chapter, the understanding of catalytic processes by QEXAFS was essential to rationally optimize these dynamic catalytic processes. Such QEXAFS studies will prove to be useful to understand dynamic reactions in different scientific communities in the decades to come, thanks to the development of QEXAFS as an easy-to-use tool at many synchrotrons worldwide.


First citationAbe, H., Aquilanti, G., Boada, R., Bunker, B., Glatzel, P., Nachtegaal, M. & Pascarelli, S. (2018). J. Synchrotron Rad. 25, 972–980.Google Scholar
First citationBravo-Suárez, J. J., Bando, K. K., Lu, J., Haruta, M., Fujitani, T. & Oyama, T. (2008). J. Phys. Chem. C, 112, 1115–1123.Google Scholar
First citationClark, A. H., Imbao, J., Frahm, R. & Nachtegaal, M. (2020). J. Synchrotron Rad. 27, 551–557.Google Scholar
First citationClark, A. H., Steiger, P., Bornmann, B., Hitz, S., Frahm, R., Ferri, D. & Nachtegaal, M. (2020). J. Synchrotron Rad. 27, 681–688.Google Scholar
First citationClausen, B. S. & Topsøe, H. (1991). Catal. Today, 9, 189–196.Google Scholar
First citationClausen, B. S., Topsøe, H. & Frahm, R. (1998). Adv. Catal. 42, 315–344.Google Scholar
First citationFerri, D., Newton, M. A. & Nachtegaal, M. (2011). Top. Catal. 54, 1070–1078.Google Scholar
First citationFrahm, R. (1988). Nucl. Instrum. Methods Phys. Res. A, 270, 578–581.Google Scholar
First citationFrahm, R. (1989). Rev. Sci. Instrum. 60, 2515–2518.Google Scholar
First citationFrahm, R. (1991). X-ray Absorption Fine Structure, edited by S. S. Hasnain, p. 731. Chichester: Ellis Horwood.Google Scholar
First citationGrunwaldt, J.-D. & Frenkel, A. I. (2009). Synchrotron Radiat. News 22(1), 2–4.Google Scholar
First citationKönig, C. F. J., van Bokhoven, J. A., Schildhauer, T. J. & Nachtegaal, M. (2012). J. Phys. Chem. C, 116, 19857–19866.Google Scholar
First citationLützenkirchen-Hecht, D., Grundmann, S. & Frahm, R. (2001). J. Synchrotron Rad. 8, 6–9.Google Scholar
First citationLytle, F. W., Sayers, D. E. & Moore, E. B. Jr (1974). Appl. Phys. Lett. 24, 45–47.Google Scholar
First citationMarberger, A., Petrov, A. W., Steiger, P., Elsener, M., Kröcher, O., Nachtegaal, M. & Ferri, D. (2018). Nat. Catal. 1, 221–227.Google Scholar
First citationMüller, O., Lützenkirchen-Hecht, D. & Frahm, R. (2015). Rev. Sci. Instrum. 86, 093905.Google Scholar
First citationMüller, O., Nachtegaal, M., Just, J., Lützenkirchen-Hecht, D. & Frahm, R. (2016). J. Synchrotron Rad. 23, 260–266.Google Scholar
First citationMutz, B., Gänzler, A., Nachtegaal, M., Müller, O., Frahm, R., Kleist, W. & Grunwaldt, J. (2017). Catalysts, 7, 279.Google Scholar
First citationNachtegaal, M., Müller, O., König, C. & Frahm, R. (2016). X-ray Absorption and X-ray Emission Spectroscopy: Theory and Applications, edited by J. A. van Bokhoven & C. Lamberti, pp. 155–183. Chichester: John Wiley & Sons.Google Scholar
First citationNonaka, T., Dohmae, K., Araki, T., Hayashi, Y., Hirose, Y., Uruga, T., Yamazaki, H., Mochizuki, T., Tanida, H. & Goto, S. (2012). Rev. Sci. Instrum. 83, 083112.Google Scholar
First citationPadamati, S. K., Angelone, D., Draksharapu, A., Primi, G., Martin, D. J., Tromp, M., Swart, M. & Browne, W. R. (2017). J. Am. Chem. Soc. 139, 8718–8724.Google Scholar
First citationRochet, A., Baubet, B., Moizan, V., Devers, E., Hugon, A., Pichon, C., Payen, E. & Briois, V. (2017). J. Phys. Chem. C, 121, 18544–18556.Google Scholar
First citationSchroer, C. G., Boye, P., Feldkamp, J., Patommel, J., Schropp, A., Samberg, D., Stephan, S., Burghammer, M., Schöder, S., Riekel, C., Lengeler, B., Falkenberg, G., Wellenreuther, G., Kuhlmann, M., Frahm, R., Lützenkirchen-Hecht, D. & Schroeder, W. H. (2010). Acta Phys. Pol. A, 117, 357–368.Google Scholar
First citationSchroer, C. G., Kuhlmann, M., Günzler, T. F., Lengeler, B., Richwin, M., Griesebock, B., Lützenkirchen-Hecht, D., Frahm, R., Ziegler, E., Mashayekhi, A., Haeffner, D. R., Grunwaldt, J. & Baiker, A. (2003). Appl. Phys. Lett. 82, 3360–3362.Google Scholar
First citationSeidler, G. T., Mortensen, D. R., Remesnik, A. J., Pacold, J. I., Ball, N. A., Barry, N., Styczinski, M. & Hoidn, R. (2014). Rev. Sci. Instrum. 85, 113906.Google Scholar
First citationSekizawa, O. T., Uruga, K., Higashi, T., Kaneko, Y., Yoshida, T., Sakata, Y. & Iwasawa, Y. (2017). ACS Sustain. Chem. Eng. 5, 3631–3636.Google Scholar
First citationVantelon, D., Davranche, M., Marsac, R., La Fontaine, C., Guénet, H., Jestin, J., Campaore, G., Beauvois, A. & Briois, V. (2019). Environ. Sci. Nano, 6, 2641–2651.Google Scholar
First citationVoronov, A., Urakawa, A., van Beek, W., Tsakoumis, N. E., Emerich, H. & Rønning, M. (2014). Anal. Chim. Acta, 840, 20–27.Google Scholar

to end of page
to top of page