Operando tracking of oxidation-state changes by coupling electrochemistry with time-resolved X-ray absorption spectroscopy demonstrated for water oxidation by a cobalt-based catalyst film

Transition metal oxides are promising electrocatalysts for water oxidation, i.e., the oxygen evolution reaction (OER), which is critical in electrochemical production of non-fossil fuels. The involvement of oxidation state changes of the metal in OER electrocatalysis is increasingly recognized in the literature. Tracing these oxidation states under operation conditions could provide relevant information for performance optimization and development of durable catalysts, but further methodical developments are needed. Here, we propose a strategy to use single-energy X-ray absorption spectroscopy for monitoring metal oxidation-state changes during OER operation with millisecond time resolution. The procedure to obtain time-resolved oxidation state values, using two calibration curves, is explained in detail. We demonstrate the significance of this approach as well as possible sources of data misinterpretation. We conclude that the combination of X-ray absorption spectroscopy with electrochemical techniques allows us to investigate the kinetics of redox transitions and to distinguish the catalytic current from the redox current. Tracking of the oxidation state changes of Co ions in electrodeposited oxide films during cyclic voltammetry in neutral pH electrolyte serves as a proof of principle. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00216-021-03515-0.

Various methods have been used to describe the X-ray edge position by a single energy value, the edge energy. Edge energies are usually determined as the half-height of the normalized Xray edge-rise (energy where the fluorescence intensity of the normalized edge is 0.5), or as the energy of the main inflection point of the X-ray edge spectrum (energy of the steepest fluorescence rise) [1,2]. Both methods have the advantage that are intuitively clear and mathematically straightforward to implement. However, the determined edge energy position is often affected by edge shape changes related to structural changes of the material and are not necessarily coupled to oxidation state changes. The alternative option used here is the 'integral method' that derives the X-ray edge position, via suitable integration procedure of the entire edge rise, as detailed elsewhere [3]. The advantage of the 'integral method' is that the resulting edge energy value: (i) is largely insensitive to edge-shape changes (as they may arise due to changes in the coordination environment), (ii) reflects the mixing ratio well for mixed-valent compounds, and (iii) is less sensitive to noise contributions (because the integration effectively involves averaging over a large number of data points) such that it can be determined reliably, without potentially critical data smoothing protocols, also for X-ray data with comparably high noise contributions [3].

Supplementary Note 2. Energy calibration
The energy of the X-rays at most synchrotron beamlines is calculated from the position of the monochromator, often with high relative precision, but typically with significant uncertainties regarding the absolute energy scale. The absolute energy scale may depend on experimental factors like positioning of mirrors and monochromator elements, slit widths and heat load to elements of the beamline optics. Thus, in typical X-ray absorption experiments, an energy calibration procedure is performed, ideally based on the simultaneous measurement of the sample of interest and of a stable reference compound. The reference compound must have a known position of the absorption edge and it is often a metal foil placed behind the sample.
(For metal foils, edge position have been determined on an absolute energy scale and are available as tabulated values [4]). However, this procedure introduces some constraints in the experimental set-up. In the case of a large electrochemical cell which is not transparent for Xrays, like the one used in this work, the measurement of a reference compound placed behind the sample is technically impossible (see Figure S1). Therefore, an alternative procedure for energy calibration was developed, which is based on the alignment of the pre-edge rise. The pre-edge is a feature of the XANES spectrum of cobalt ions, whose position, unlike the position of the main absorption edge, is not strongly influenced by the oxidation state of the sample [5]. The spectra of reference compounds, used to build the calibration curve, are also aligned following the same rationale. Since pre-edges are aligned, what we are practically using for calibration is the distance between pre-edge rise and edge rise.
The alignment here was facilitated by simulating the derivative of the pre-edge feature with Gaussian functions (see Supplementary Information, Figure S4), the center of the Gaussian function, corresponding to the middle of the pre-edge rise was set to 7708 eV. Since the preedge is not perfectly well described by a Gaussian function, the range of the fit slightly affects the final alignment. The variation in oxidation state due to different fit-ranges used for energy calibration is illustrated in the Supplementary Information ( Figure  Other approaches are feasible, for example, the fine-tuning of the energy axis, aiming at optimal alignment of the pre-edge rise of the measured spectrum and a reference spectrum. For X-ray data of sufficiently high quality, optimal alignment can be judged by mere visual inspection at high precision. We emphasize that, whatever method is used, the pre-edge alignment requires extreme care, because its precision is a major determinant of the precision of the final oxidation-state estimate.

Supplementary Note 3. X-ray radiation damage
A possible source of misleading results is radiation damage in form of X-ray photoreduction of high-valent metal sites, which is often observed when exposing a sample to strong X-rays for prolonged time periods. This problem has been extensively discussed elsewhere [6]. In this study, to avoid possible radiation damage to the sample, an out-of-focus geometry (with 11 x 2 mm beam spot size) and a low X-ray beam intensity (about 10 11 photons/s [7]) were used.
Furthermore, an additional control experiment was performed, to rule out the presence of radiation damage: the incoming X-ray beam energy was fixed in the middle of the absorption edge (Eexc = 7722 eV) and the X-ray fluorescence was monitored over the course of several minutes ( Figure S6). During this time, the X-ray beam was switched off and on again. The fluorescence signal slightly decreases over time, which can be due either to photoreduction or sample dissolution. However, the rate of the decrease is independent on whether the X-ray beam is off or on; no indications for sharp changes in the film behavior exists, suggesting that photoreduction is not involved. The same experiment was repeated with the energy set well above the absorption edge (Eexc = 8400 eV) and a similar result was obtained. At this excitation energy, the X-ray fluorescence is not sensitive to the metal oxidation state but only to the amount of fluorescing material in the film, suggesting that the decrease in fluorescence is assignable to sample dissolution. Accordingly, the presence of extensive radiation damage, meaning photoreduction of the metal centers caused by the X-ray beam, can be excluded in our case.

Supplementary Note 4. Accuracy of the conversion procedure
The analytical procedure proposed for the conversion of the X-ray fluorescence signal into an oxidation state requires a rather complex data analysis. In this paragraph, we analyze the potential inaccuracy that is added by the different steps of the procedure and estimate a confidence interval for the results obtained in the proof-of-concept. therefore has a lower accuracy, while the comparison between different experiments, using the same set of reference compounds (relative accuracy), is more reliable.
An estimation of the relative accuracy is provided by a repetition of the experiment (Fig. S8a).  Fig. 1d, and the standard deviation (σ) on each of the fit parameter is obtained. Two alternative calibration lines were calculated using the standard deviation: y = (m + σm) x + q -σq and y = (m -σm) x + q + σq. This two alternative calibration lines were employed for the conversion procedure on an exemplary experiment, obtaining different final values for the Co oxidation state during the experiment (Fig. S8b). The variation in the Co oxidation state caused by the use of an alternative "second calibration line" is 0.03 oxidation state unit, we use this value as an estimate on the error caused by the "second calibration line". It is important to note that the reference compounds used for this study were carefully chosen and fit well on a line, using a different set of reference compounds can lead to higher inaccuracies.
The error made in the alignment of the energy axis is discussed in the Supplementary Note 2.
A slightly alternative approach is used for alignment resulting in a different oxidation state value, the variation amounts to 0.06 oxidation state units. This value is assumed to be the error caused by the alignment of the energy axis.
By summing all the different contributions, we obtain an estimate for the confidence interval on the oxidation state absolute value (absolute accuracy) of 0.15 oxidation state units.      Figure S5). The use of a different fit range causes a variation of 0.06 oxidation state units. Data averaged from 9 CVs are shown after smoothing with a moving average over 15 points. The sample used here is thicker than the one used in Fig. 6a, a comparison between the two is presented in Fig. S8.