5D operando tomographic diffraction imaging of a catalyst bed

We report the results from the first 5D tomographic diffraction imaging experiment of a complex Ni–Pd/CeO2–ZrO2/Al2O3 catalyst used for methane reforming. This five-dimensional (three spatial, one scattering and one dimension to denote time/imposed state) approach enabled us to track the chemical evolution of many particles across the catalyst bed and relate these changes to the gas environment that the particles experience. Rietveld analysis of some 2 × 106 diffraction patterns allowed us to extract heterogeneities in the catalyst from the Å to the nm and to the μm scale (3D maps corresponding to unit cell lattice parameters, crystallite sizes and phase distribution maps respectively) under different chemical environments. We are able to capture the evolution of the Ni-containing species and gain a more complete insight into the multiple roles of the CeO2-ZrO2 promoters and the reasons behind the partial deactivation of the catalyst during partial oxidation of methane.

Temperature calibration was performed before all experiments using a thermocouple by measuring the temperature at the catalyst bed. The temperature calibration curve for the experimental setup with the two hot air blowers used at ID31 is shown in Supplementary Figure 3.
The maximum temperature was used during the high temperature XRD-CT measurements (i.e. nominal 1000 °C and actual 800 °C). The temperature offset for the experimental setup with the furnace used at ID15A is minimal (i.e. nominal 825 °C and actual 800 °C).

Supplementary note 1. Multi-dimensional chemical imaging
With the continuous advances in data acquisition strategies and methods/techniques development, it is becoming increasingly difficult to define the dimensions of in situ experiments. As such, the following tables serve as a guide to define the dimensions in multi-dimensional chemical imaging experiments. In Supplementary Table 1, the most common type of static scans (i.e. at ambient conditions/ the sample under study is not changing) are described. The conclusion is that for the type of scan (i.e. same number of spatial dimensions), in comparison to an absorptioncontrast scan a chemical scan contains one extra dimension (i.e. the spectral or scattering dimension). For example, a static 3D-XRD-CT consists a 4D dataset (3D spatial and 1D scattering dimensions).   Series of maps or 2D tomograms " 2D " 4D to nD (n > 4) Series of 3D tomograms " 3D " 5D to nD (n > 5)

Supplementary note 2. Fast XRD-CT and comparison with literature
In 1987, Harding et al. were the first to introduce a new materials characterization technique termed X-ray diffraction computed tomography (XRD-CT). 1 XRD-CT couples traditional powder X-ray diffraction with first generation (i.e. pencil beam approach) computed tomography (CT). It was first demonstrated using a lab diffractometer and more than a decade later, in 1998, Kleuker et al. were the first to implement XRD-CT using synchrotron light for medical imaging applications (soft tissue). 2 The total acquisition time of that XRD-CT scan (using an 80 keV monochromatic pencil beam) was 2 h for 100 × 100 pixels reconstructed images using an area detector (900 × 900 μm 2 beam size using slit systems  5,6,7,8,9,10,11,12,13,14,15,16,17,18,19 In the aforementioned studies, XRD-CT was employed either to study static samples or for ex situ characterization of functional materials (e.g. catalysts and batteries). However, it is generally accepted in literature that functional materials can change under working conditions; there are several review papers emphasizing the need for in situ/ operando techniques to characterize catalytic and electrochemical systems under real working conditions in order to gain a proper understanding of structure-activity relationships. 20,21,22,23,24,25,26,27,28,29,30,31 XRD-CT has recently been exploited as a chemical imaging tool to study several catalytic systems under real process conditions and in all cases new physico-chemical information was revealed due to the spatially-resolved signals obtained from such measurements; information that is/can be lost in bulk measurements. 32,33,34,35,36 However, the temporal resolution of the XRD-CT technique has been always considered to be its main drawback. This is clearly implied in Supplementary Table 4 where the experimental details of most XRD-CT studies conducted in the past decade are presented (i.e. from papers where information available). It can be seen that in most studies, the total acquisition time of an XRD-CT scan is in the order of several hours which is far from ideal for dynamic experiments.
Supplementary  36,37 However, there is also another equally important factor which contributes to the time required to perform an XRD-CT scan and it is has not been investigated/discussed in literature in the past. This is the dead time of the measurement which depends not only on the motors' capabilities (fast motors are essential for fast XRD-CT scans) but also on the chosen data collection strategy. We recently provided a short review on the available data collection strategies and also introduced a new one, termed as interlaced XRD-CT, which allows, post experiment, choice between temporal and spatial resolution. 37 Herein, we present a new data collection strategy which minimizes the dead time during an XRD-CT scan. In fact, the total acquisition time of such an XRD-CT measurement can be calculated by just multiplying the ATPP with the total number of diffraction patterns collected. The breakthrough in this ultra-fast XRD-CT data collection strategy lies on the fact that both the rotation and the translation axis are continuously moving during the acquisition of the tomographic scan.
In our first attempt to perform the fast XRD-CT scan at beamline ID31 of the ESRF, the ATPP was 15 ms (11 ms exposure time and 4 ms readout time). When these fast XRD-CT experiments were performed, the newly purchased Pilatus3 X CdTe 2M detector had not been fully integrated with the beamline hardware. More specifically, there was a limitation associated with writing and transferring data from the detector PC to the beamline PCs (i.e. network limitation). However, even with these limitations, the total acquisition time of each XRD-CT scan was less than 2 min (108 s to collect 7200 images) which is a technical breakthrough when compared to what was feasible in the past (Table 4).

Supplementary discussion
Phase identification results of the fresh catalyst Prior to the Rietveld analysis of the reconstructed data (4D matrix in the case of 3D-XRD-CT data), the reconstructed XRD-CT images were processed using the images corresponding to the high intensity peak/peaks of Al2O3 (ca. Q = 4.5 Å -1 ). More specifically, masks (binary images) were created after performing image segmentation of the Al2O3 images by thresholding, in order to mask out the SiO2 capillary (reactor vessel) and the empty space between the catalyst particles. The Rietveld analysis of the acquired XRD-CT data was then performed, after cropping appropriately the reconstructed data, on a line-by-line basis using the TOPAS software (all the powder diffraction patterns present in each row of each XRD-CT dataset were processed simultaneously). 47 These preprocessing steps were performed in order to minimize the required time for the Rietveld analysis of the XRD-CT data (i.e. number of diffraction patterns to be processed). Apart from refining the background and the scale factors of all crystalline phases, the crystallite size of each phase was calculated too. Finally, the unit cell parameters of CeO2 and ZrO2 were added to the refinement as a crude inspection of the reconstructed data prior to the Rietveld analysis revealed that there were significant variations in the unit cell parameters of CeO2 and ZrO2 over the sample (spatial variations). That was not the case for the PdO and the Al2O3 unit cell parameters which were not refined. From a materials chemistry perspective, there are not expected to be significant variations in the unit cell parameters of the Al2O3 it being the support of the catalyst. Additionally, refining the low symmetry Al2O3 unit cell parameters (lattice parameters a, b, c and β angle) does not only significantly increase the required computational time to perform the Rietveld analysis but can also lead to a less stable refinement (higher number of parameters being refined simultaneously).
Regarding the PdO, only the high intensity diffraction peak (ca. Q = 2.35 Å -1 corresponding to (002) and (011) reflections) is observed in some regions of the sample and as it is shown in Supplementary   Figure 5 there are other crystalline phases generating diffraction peaks in that Q region (i.e. CeO2, ZrO2 and Al2O3). Of course, it is also expected that there should not be significant variations in the PdO unit cell parameters as the 3D-XRD-CT measurement was performed at room temperature. As a result, it was considered prudent to not refine the lattice parameters of PdO and Al2O3.
X-ray micro CT and sub-minute XRD-CT measurements at ID15A, ESRF X-ray micro-CT measurements were performed at beamline ID15A of the ESRF using a 42 keV monochromatic X-ray beam. Radiographs were recorded with an X-ray imaging camera (CCD) and the pixel resolution was 3.18 μm. Each micro-CT scan consisted of 1900 projections (radiographs) covering an angular range of 0 -190 ° (i.e. angular step size of 0.1 °). Flat-field and dark current images were also collected prior to the micro-CT measurements and were used to normalize the acquired radiographs before the tomographic reconstruction. The tomographic data were reconstructed using the filtered back projection algorithm.
Ultra-fast XRD-CT measurements were also performed at beamline station ID15A of the ESRF phase is seen to follow an egg-shell distribution. The CeO2 phase is also seen to be predominantly present near the surface of the catalyst particles but there are also numerous particles where it is present near the core of the particles too. The NiO phase is present in all catalyst particles but it can also be seen that not in the same amount. Finally, the PdO phase is seen to follow a similar distribution to the ZrO2 phase but it can also be clearly seen that there are region of very high concentration of this phase (i.e. hotspots). This indicates that the Pd species are not well distributed over the catalyst particles. High resolution XRD-CT measurement at ID15A, ESRF A single catalyst particle was investigated with high resolution XRD-CT at beamline station ID15A of the ESRF using a 50 keV monochromatic X-ray beam focused to have a spot size of 1 μm x 1 μm with a Kb mirror system. 2D powder diffraction patterns were collected using the Pilatus3 X CdTe

Supplementary note 5. Different chemical species of CexZryO2
In order to verify the existence of the four distinct crystalline CexZr1-xO2 species, three summed diffraction patterns were exported from selected regions of interest. As it is shown in Supplementary Figure 10, three masks were created based on the results from the Rietveld analysis results presented in Figure 1 in the main paper and these masks correspond to a Ce rich CexZr1-xO2 region (x >> 0), a Zr rich CexZr1-xO2 region (x << 1) and another region where both phases are present (0 < x < 1). The three masks were applied to the reconstructed data volume (one at a time) and in each case the summed diffraction pattern was exported. These three diffraction patterns are plotted at the right side of Supplementary Figure 10

3D-XRD-CT at high temperature under He flow
The summed diffraction patterns of the 3D-XRD-CT data (30 XRD-CT scans) collected at 800 °C under He flow are shown at the right side of Supplementary Figure 11. For comparison reasons, the summed diffraction patterns of the 3D-XRD-CT data (30 XRD-CT scans) collected at room temperature are also presented at the left side of Supplementary Figure 11. It can be clearly seen that the intensity of main NiO diffraction peak (reflection (002)) has significantly decreased while the diffraction peak at ca. Q = 2.55 -2.65 Å -1 has not. Since there is a NiO diffraction peak in this Q region (reflection (111)), one would expect that the intensity of this peak would decrease too.
However, this is not the case and this phenomenon was attributed to the formation and growth of the undesired NiAl2O4 phase (ICSD: 9554). 50,51 It has been previously reported in literature that crystalline NiAl2O4 can be observed in Ni/Al2O3 catalysts calcined at temperatures above 600 °C. 52 It is no wonder then that the NiAl2O4 phase is seen to be present in the 10 wt. % Ni -0.2 wt. Pd/ 10 wt. Scale bar corresponds to 1 mm.

In situ XRD-CT measurements during catalyst activation at ID15A, ESRF
In order to validate our assumption that the observed chemical gradient along the catalyst bed during the reduction process is a purely temporal phenomenon (i.e. for the Ni-containing phases), we performed a second diffraction experiment focusing on the behaviour of the catalyst during activation. The protocol for the catalyst pre-treatment was the same as in the 5D

Mass spectrometry data during the in situ POX XRD-CT measurements
The mass spectrometry data acquired during the POX experiment are presented in Supplementary Figure 18, where the signals from specific masses of interest are shown and serve to prove that the catalyst was captured in its active state during the POX reaction. It can be clearly seen that upon switching to POX reaction conditions (i.e. region 1 in Supplementary Figure 18), the signal from masses 2 and 28 increase which correspond to fractions from the POX reaction products (i.e. H2 and CO respectively). It can also be seen that the signal from masses 15, 17 and 44 (corresponding to CH4, H2O and CO2 respectively) also increase and are present for the duration of the POX experiment. This means not only that CH4 is not fully consumed (which is expected from the CH4/O2 molar ratio used) but that other side reactions take place too resulting in the formation of H2O and CO2. It can be clearly seen that the signal from all masses is stable for the duration of the POX experiment and no apparent deactivation of the catalyst was observed. However, as it is shown in  is indeed detrimental to the long-term stability of this catalyst and is probably not removed when the CH4:O2 ratio is switched back to 2:1. Further studies could focus on optimizing a catalyst regeneration step but it should be mentioned that introducing an O2 rich stream (in order to remove the carbonic species present in the catalyst) can also be harmful to the catalyst (e.g. by leading to sintering of the Ni particles as shown in Figure 9 in the main paper or by creating local hotspots due to the very exothermic nature of the reaction).
Scanning electron microscopy and elemental analysis This is further supported by the SEM/EDS images presented in Supplementary Figure 22 where it can be seen that the Zr species are predominantly present near the surface of the catalyst particles. This result is in direct agreement with the results obtained from the Rietveld analysis of the XRD-CT data that show the presence of a Zr rich CexZr1-xO2 phase following an egg-shell distribution.
In contrast to the Zr species, Ce species are seen to be present even near the core of the catalyst particles. This result is also in direct agreement with the XRD-CT results. However, it should be emphasized that the Rietveld analysis of the XRD-CT showed that the Ce rich CexZr1-xO2 phase present near the core of the catalyst particles corresponds to larger crystallites compared to the surface species. It also allowed us to discriminate between two different CexZr1-xO2 phases present near the surface of the catalyst particles (i.e. a Ce rich and a Zr rich one). This detailed physicochemical information could have not be obtained with any other means.
In both Supplementary Figures 21 and 22, it can be seen that the Ni-containing species are present everywhere in the catalyst particles; a result which is also in full agreement with the results derived from the XRD-CT data. Finally, it can be seen that the signal of the Pd-containing species is higher near the surface of the catalyst particles. However, it can also be seen that there are regions of high concentration of this phase (i.e. hotspots). These results are also in direct agreement with the PdO (and Pd after reduction/activation) phase distribution maps and indicating that there is room for improvement regarding the catalyst design. The Ni species are also seen to be present everywhere in the catalyst particles; a result which is also in agreement with the XRD-CT data obtained under POX reaction conditions. The Pd signal was very weak in both scans so it is not easy to reach any definite conclusions. On the other hand, the SEM image presented in Supplementary Figure 24 shows there are minor cracks near the catalyst surface. However, this is not the case for the catalyst particle shown in Supplementary   Figure 23. This phenomenon is probably associated with the position of the catalyst particles in the reactor. As this is an ex situ measurement of random particles taken from the spent catalyst, it is impossible to provide this information. This is another example showing why in situ and indeed operando characterization is essential in order to understand the structure-function relationships of a working catalyst and the reasons leading to their deactivation.