Multi-length scale 5D di ﬀ raction imaging of Ni – Pd/ CeO 2 – ZrO 2 /Al 2 O 3 catalyst during partial oxidation of methane † 11331 of

A 5D di ﬀ raction imaging experiment (with 3D spatial, 1D time/imposed operating conditions and 1D scattering signal) was performed with a Ni – Pd/CeO 2 – ZrO 2 /Al 2 O 3 catalyst. The catalyst was investigated during both activation and partial oxidation of methane (POX). The spatio-temporal resolved di ﬀ raction data allowed us to obtain unprecedented insight into the behaviour and fate of the various metal and metal oxide species and how this is a ﬀ ected by the heterogeneity across catalyst particles. We show ﬁ rstly, how Pd promotion although facilitating Ni reduction, over time leads to formation of unstable Ni – Pd metallic alloy, rendering the impact of Pd beyond the initial reduction less important. Furthermore, in the core of the particles, where the metallic Ni is primarily supported on Al 2 O 3 , poor resistance towards coke deposition was observed. We identi ﬁ ed that this preceded via the formation of an active yet metastable interstitial solid solution of Ni – C and led to the exclusive formation of graphitic carbon, the only polymorph of coke observed. In contrast, at the outermost part of the catalyst particle, where Ni is predominantly supported on CeO 2 – ZrO 2 , the graphite formation was mitigated but sintering of Ni crystallites was more severe. of graphite and Ni size at the particles where CeO 2 ZrO 2 Al O 3 (core)


Introduction
The abundance of methane, the main ingredient of natural gas, coupled with the continuous consumption of crude oil has led to high interest in methane reforming and upgrading processes. 1,2 Methane can be directly converted to light olens (such as ethylene) following both oxidative 3-6 and non-oxidative routes [7][8][9] or converted rst into synthesis gas (a mixture of CO and H 2 ) and then to higher hydrocarbon molecules via the Fischer-Tropsch reaction. 10,11 There are several processes that can be used to prepare synthesis gas with the most well-known ones being: steam reforming, 12-14 dry reforming 15-18 partial oxidation of methane [19][20][21] and autothermal reforming (which is essentially a combination of partial oxidation and steam reforming). [22][23][24] The partial oxidation of methane (POX), a mildly exothermic reaction, is considered to be a promising alternative process to the highly endothermic processes of steam and dry reforming of methane. 25,26 The most well studied catalyst candidate for the POX reaction is Ni/Al 2 O 3 due to its good performance and lower cost, when compared to noble metals such as Pd, Pt or Rh. 27,28 However, the Ni based catalysts are prone to fast deactivation due to multiple reasons which include coke deposition on the metallic Ni active sites and sintering of Ni particles. 28 More importantly, coke deposition has a detrimental effect, as it leads to encapsulation of the catalyst active sites, particles expansion and blocking their pores, or even to physical blockage of the reactor itself. 29 It is generally accepted that there are two mechanisms of coke deposition: 2CO#C þ CO 2 DH 298 K ¼ À172 kJ mol À1 (2) where (1) direct dehydrogenation of methane and (2) disproportionation of carbon monoxide (Boudouard reaction). 30 Different types of coke have been reported to form over the Ni/ Al 2 O 3 catalysts: crystalline graphitic carbon, amorphous carbon, lamentous carbon or even metallic Ni carbides. 29,[31][32][33][34] In order to improve the catalyst performance, a large number of studies focused on investigating the preparation method 35,36 as well as the addition of promoters such as noble metals (e.g. Pd, Pt, Rh) 37,38 and modication to support material by adding the alkali earth and rare earth metal oxides (e.g. MgO, CaO, BaO, CeO 2 , ZrO 2 , La 2 O 3 ). 23,[39][40][41][42][43] The role of the noble metal promoters has been associated to an increased stability towards coke deposition 44 and an increased reducibility of oxidised Ni species to metallic Ni, through a hydrogen spillover mechanism from noble metal clusters to the Ni species. As reported by Mukainakano et al., 45 doping Ni catalysts with Pt or Pd resulted in suppression of hot spot formation (in the autothermal reforming of methane) through the formation of bimetallic Ni-Pd particles and enhanced reducibility of Ni species. As reported in several studies, the promoting effect of noble metals is also related to increased activity of the bimetallic catalysts and improved physico-chemical properties, such as enhanced dispersion of the active metals. 46,47 However, to the authors' knowledge there have been no studies investigating the state or stability of these metal-containing species/alloys under real reaction conditions. Modication of the conventional Al 2 O 3 support with CeO 2 is expected to supress coke formation as the CeO 2 has the ability to store and release oxygen (CeO 2 redox properties). 48 Especially when doped with ZrO 2 , the binary solid solution of CeO 2 -ZrO 2 has been proved to possess a higher thermal stability and improved oxygen storage capacity. 17,[49][50][51][52] All these materials combined in the state-of-the-art Ni-Pd/CeO 2 -ZrO 2 /Al 2 O 3 catalyst make it a promising candidate for the POX reaction. More importantly, the 10 wt% Ni-0.2 wt Pd/10 wt% CeO 2 -ZrO 2 /Al 2 O 3 has been previously reported to show an excellent performance with stoichiometric POX reaction (CH 4 : O 2 ¼ 2), reaching a H 2 yield above 90%. 19,22 This study aims to gain a deeper insight into the evolving solid-state chemistry of this complex catalyst during the partial oxidation of methane through synchrotron tomographic X-ray diffraction imaging. X-ray diffraction computed tomography (XRD-CT) is a powerful characterisation technique which has been lately applied to study a variety of functional materials and devices including operating solid catalysts, fuel cells and batteries. [53][54][55][56][57][58][59][60][61][62][63][64] The radial distribution of the various crystalline catalyst components under operating conditions was investigated with the 3D-XRD-CT technique. In addition, the fast solidstate changes occurring along the catalyst bed were followed with X-ray diffraction mapping. Both approaches led to complementary information that allowed us to rationalise the observed solid-state chemistry changes taking place in the working catalyst at different length-scales. Our ambition was to follow the evolving solid-state chemistry at play in order to understand the salient features of the catalyst and how the active and deactivated states look like.

Catalyst preparation
The 10 wt% Ni-0.2 wt Pd/10 wt% CeO 2 -ZrO 2 /Al 2 O 3 catalyst was prepared by sequential impregnation. First, the CeO 2 -ZrO 2 / Al 2 O 3 was prepared by the co-impregnation method. The (g + d)-Al 2 O 3 support (a fraction between 250-500 mm) was impregnated by aqueous solution of salts (cerium nitrate Ce(NO) 3 -$6H 2 O and oxychloride of zirconium ZrOCl$8H 2 O) at the required ratio. The CeO 2 -ZrO 2 /Al 2 O 3 was then dried at 120 C for 6 h and calcined in air at 850 C for 6 h with a heating rate of 2 C min À1 . Next, the 10 wt% CeO 2 -ZrO 2 /Al 2 O 3 support was impregnated by aqueous solution of nickel nitrate salt Ni(NO 3 ) 2 $6H 2 O of the appropriate concentration and then the Ni/CeO 2 -ZrO 2 /Al 2 O 3 was dried at 120 C for 6 h and calcined in air at 500 C for 4 h with a heating rate of 2 C min À1 . The Ni/ CeO 2 -ZrO 2 /Al 2 O 3 was subsequently impregnated using an aqueous solution of palladium nitrate Pd(NO 3 ) 2 salt of the appropriate concentration. The catalyst was then dried at 120 C for 6 h and calcined in air at 500 C for 4 h with a heating rate of 2 C min À1 .

Operando diffraction imaging measurements at ID15A, ESRF
95 mg of catalyst sample was placed inside a quartz tube reactor (external diameter 6 mm, internal diameter 5 mm) to form a catalyst bed of 7 mm length, supported by glass wool. Pure gases O 2 , CH 4 and He and mixture of 20% H 2 /He were delivered to the reactor by mass ow controllers (Brooks). The catalyst was heated under an atmosphere of He (30 sccm) up to 800 C with a temperature ramp of 20 C min À1 , using a custom-built furnace. The POX reaction was performed at atmospheric pressure and the outlet gases were analysed by mass spectrometry using an EcoSys portable mass spectrometer. The gas conditions and experimental protocol are presented in the Table 1. The last stage of experiment (POX reaction mixture 2 : 1) was only kept for $60 minutes and only ve full XRD-CT slices were collected during this period. Finally, the sample was cooled down to room temperature under an atmosphere of He (30 sccm) for the subsequent TGA analysis. The photograph of the experimental set up and reactor is presented in Fig. S1 in the ESI. † The XRD-CT measurements were performed at ID15A beamline at the ESRF 65 using a monochromatic beam of 91 keV with a size of 20 mm Â 40 mm (H Â V). Diffraction patterns were collected using a PILATUS3 X 2M CdTe (Dectris) area detector, calibrated with a CeO 2 NIST standard. One XRD-CT scan of the catalytic reactor was performed using 150 translation steps and 125 angular steps, with the translation step of 40 mm, angular step of 1.44 and 10 ms of exposure time. Time of measurement for one XRD-CT scan was $9 min. The 3D-XRD-CT measurements consisted of 10 consecutive XRD-CT slices, collected in the middle of catalyst bed with the step of 80 mm between them. Every 2D diffraction image was converted to 1D powder diffraction pattern using the nDTomo and PyFAI soware package 66,67 with implemented trimmed mean lter (10%) 68 to remove the artefacts due to hot spots of (single) crystalline material. The data integration was performed with fast GPU processing. 69 The reconstructed images were obtained using the ltered back projection algorithm. The collected data were corrected in respect to synchrotron current. The analysis of diffraction data was performed with Topas 5 soware. 70 The XRD mapping measurements (i.e. YZ grid) were collected with 40 mm translation step, 500 mm z step and 50 ms of exposure time. Each XRD map consisted of 150 columns (translation steps across the reactor) and 14 rows (z steps). Time of measurement for one complete XRD map was $5 min.

Rietveld renement
The diffraction patterns analysed herein were taken only from tomographic voxels where the catalyst was present (i.e. regions of air, reactor vessel and voids between particles were removed). Each XRD-CT scan consisted of ca. 5000 diffraction patterns (i.e. taken from the catalyst particles region only) that were analysed with Rietveld renement. Regarding the XRD map, the regions of the reactor vessel and air were masked prior to the Rietveld analysis. Also, the signal generated by the reactor vessel was subtracted from the catalyst signal in order to ease the data analysis. Each XRD map consisted of ca. 1300 diffraction patterns that were analysed with Rietveld renement. To decrease the total number of parameters and thus improve the stability of the renements, the parameters related to the Al 2 O 3 support (low symmetry q-Al 2 O 3 corresponding to a monoclinic unit cell) were only rened once at room temperature and at high temperature and their values were kept constant in the input le for batch renement. In this study, the presented results are based on normalised scale factors, separately for the spatially-resolved data from tomographic and for the diffraction maps (YZ). The scale factors for each phase were normalised with respect to the maximum value obtained throughout the entire experiment. All identied phases together with their crystallographic details and ICSD database codes are presented in Table S1 in the ESI. †

Initial catalyst composition
The phase distribution volume results of the 3D-XRD-CT scan performed at room temperature are presented in Fig. 1. Full prole Rietveld analysis of the reconstructed diffraction patterns revealed that the initial catalyst comprised NiO, ZrO 2 , CeO 2 and PdO distributed on a q-Al 2 O 3 support (Fig. S2 in the ESI †). NiO was found in all catalyst particles, with a rather uniform distribution and a few particles were seen to be particularly rich in this phase, predominantly near the core. The distribution of ZrO 2 was mainly observed to be located on the edges of the catalyst particles. The ZrO 2 was not present in every catalyst particle; it was possible to identify particles without any ZrO 2 as well as a few particles rich in this phase. As reported in our previous study, 19 the ZrO 2 phase possessed an egg-shell distribution. Regarding the CeO 2 phase, its distribution varied for each particle. Indeed we observed the whole range of particle distributions, ranging from cores rich in CeO 2, uniform distributions, to some particles containing CeO 2 only at the edges, similar to the distribution of ZrO 2 . 71 The distribution of PdO was challenging to extract due to its low loading (0.2 wt%) and also due to peak overlap with ZrO 2 (Fig. S4 in the ESI †).
The spatial distribution of ZrO 2 and CeO 2 in one of the XRD-CT cross sections (cross section from the middle of 3D volume presented in Fig. 1) collected at room temperature as well as the distribution of their lattice parameter a and crystallite size are presented in Fig. 2 (cross-section taken from the middle of catalyst bed volume). Regarding the distribution of the ZrO 2 phase, regions with hot spots (high intensity) of this material are characterised by a larger crystallite size ($20 nm) and a lower lattice parameter a, meaning that the phase comprised relatively pure ZrO 2 . In addition, as previously shown, two regions of mixed ZrO 2 -CeO 2 phases could be identied, one with a lattice parameter a $ 3.61 A and crystallite size of $12 nm and second region with a greater lattice parameter a being around 3.63 A and 8 nm crystallite size. 19 As the size of the ionic radius of Ce 4+ is $20% greater than the ionic radius of Zr 4+ (0.97 A and 0.78 A respectively 72 ), we propose that substitution of Ce 4+ into the structure of ZrO 2 occurs and leads to an increased lattice parameter. 73 Finally, the distribution of a fourth phase, consisting of a relatively pure CeO 2 , was found to be present in the core of the particles, with a uniform distribution of lattice parameter $5.38 A and crystallite size of $10 nm. Summarising the results of CeO 2 -ZrO 2 distribution, the reported four species are in a good agreement with our previous study on this catalyst. 19 These CeO 2 and ZrO 2 components, although having different symmetry (cubic-uorite type with space group Fm 3m and tetragonal lattice with P4 2 /nmc space group respectively), are known to form solid solutions with multiple congurations. The analysis of such structures which aims to determine the structure-composition relationships is anything but trivial. As reported by Yashima et al. [74][75][76] these mixed phases can be categorised into three different tetragonal phases, according to the ratio between the lattice parameters a/c. In addition, the mixed solution with quantities of CeO 2 above 80 wt% crystallises in a cubic structure (uorite type). 73,77 Although multiple solid solutions with different crystal structures are possible, it has been proposed that CeO 2 -ZrO 2 (and the solid solutions) comprise a crucial component for yielding an optimal POX catalyst due to its thermal stability and its ability to supress coke formation through its enhanced redox properties and improved oxygen storage capacity. 78,79 As reported in the work of Ismagilov et al., 80 increasing the content of Zr in the mixed CeO 2 -ZrO 2 oxide improves the catalytic performance of the resulting catalyst in terms of CH 4 conversion as well as CO and H 2 yield. However, there is an optimal composition with Ce and Zr ratio being 1 : 1 aer which no further improvement is observed, possibly even a slight decrease in CO and H 2 yield.

Temperature ramping
A series of XRD maps were collected during the temperature ramp from room temperature to 800 C under He ow. In order to follow the solid-state changes occurring in the catalyst as a function of temperature, all diffraction patterns from each line scan were summed (150 diffraction patterns collected in one line scan); these diffraction patterns are presented in Fig. 3. Only aer heating to 780 C does a new peak (position 2q ¼ 3.45 ) appear in the diffraction patterns. This peak was associated to a new phase, identied as a cubic Pd-Ni alloy and it was found to be formed in the entire catalyst bed (i.e. axially). The formation of this alloy is a consequence of PdO reducing to metallic Pd (thermal reduction under He ow at high temperatures). 81 The rened position of the principal reection (111 reection with d ¼ 2.25 A), shown in Fig. 3, indicated that the major element present in the metallic alloy was Pd. 82 In addition, the phase identication of the diffraction data collected at temperatures above 780 C revealed the formation of NiAl 2 O 4 . Although the structures of Al 2 O 3 , NiO and NiAl 2 O 4 possess overlapping reections (Fig. 3, bottom panel), when using the Rietveld analysis it was possible to identify the presence of the NiAl 2 O 4 phase and the corresponding disappearance of NiO 36 (see Fig. S5 in the ESI †).
Aer reaching 800 C, three XRD-CT datasets were collected at different positions in catalyst bed: reactor inlet, middle of reactor and reactor outlet (Fig. 4). During the temperature ramp NiO was seen to react with the Al 2 O 3 support, forming the new phase NiAl 2 O 4 . As the solid-state reaction between NiO and Al 2 O 3 proceeded during the XRD-CT data collection at 800 C, the apparent gradient between NiO/NiAl 2 O 4 in Fig. 4 is due to a time effect (XRD-CT data at the reactor inlet collected rst) rather than a thermal gradient in the catalyst reactor (see Fig. S6 in the ESI †).
The newly formed phase, Pd-Ni alloy was uniformly distributed in the catalyst bed (i.e. as a function of bed height). Also, its composition (i.e. amount of Ni and Pd in the Pd-Ni alloy) in the catalyst bed remained uniform (as veried by the diffraction patterns collected at 800 C under He presented in Fig. S7 in the ESI †). It is important to note that with the resolution of the reconstructed XRD-CT images in this work (i.e. voxel size 40 Â 40 Â 80 mm 3 ) the distribution of the Pdcontaining species can only be treated as an estimate of the real distribution. As shown with the SEM/EDX measurements ( Fig. S8-S9 in the ESI †) and in our previous work on this catalytic material (with a voxel size of 1 Â 1 Â 1 mm 3 ), 19 the exact distribution of Pd-containing species is in the form of discrete regions around the particles edge.

In situ reduction
For the reduction/activation step, the catalyst was reduced in an H 2 environment (20% H 2 /He) for 2 h. During the rst 10 min of reduction two XRD maps were collected to investigate the temporal changes during gas introduction ( Fig. 5 and 6). 3D-XRD-CT data were collected in between the XRD maps (Fig. 7).
During the reduction, both NiO and NiAl 2 O 4 were observed to reduce to metallic Ni. As reported in our previous study, 19 the reduction of NiO and NiAl 2 O 4 to metallic Ni did not depend on the position in the catalyst bed and it was only related to the time-on-stream. As the XRD map was collected starting from the bottom to the top of reactor (reactor inlet to reactor outlet direction), it is clear than the difference in composition was related to time of measurement (Fig. 5). Regarding the CeO 2 , its quantity decreased in the entire catalyst bed and this can possibly be explained by the formation of a CeAlO 3 structure. As previously reported, 40,48,49,83 at high temperatures and under reducing conditions Ce 4+ was found to readily reduce to Ce 3+ and form crystalline CeAlO 3 (110 reection at d ¼ 2.66). In our study it was not possible to identify this phase unambiguously, as there were no new peaks formed during reduction. However, taking into consideration that concomitant CeO 2 disappeared from core of catalyst particles (Fig. S10 in the ESI †) and a slight decrease in Al 2 O 3 intensity was also observed (based on the scale factors, Fig. S11 in the ESI †) this could be attributed to the formation of poorly crystalline/amorphous CeAlO 3 . The possible role of CeAlO 3 phase has been discussed and there are disagreements as to its signicance in the literature. In some reports the CeAlO 3 phase has been suggested to be inactive, 49 whereas in others CeAlO 3 was thought to provide an  improvement in catalyst stability and suppression of coke formation. 40,84 Finally, no changes were observed in the ZrO 2 distribution map during the reduction step (Fig. S11 in the ESI †). Fig. 6 presents the distribution of Pd-Ni lattice parameter and the distribution of Ni crystallite size. The crystallite size of metallic Ni in the catalyst bed showed a gradient from the reactor inlet to the reactor outlet; the crystallite size of Ni is higher for catalyst particles closer to reactor outlet (27 nm and 10 nm for outlet and inlet of reactor respectively) and during the reduction process the crystallite size of metallic Ni gradually increased. At the same time, we observed that the lattice parameter of the Pd-Ni changed as a function of bed height (see Fig. S13 in the ESI †). The bottom of the reactor (reactor inlet) has a higher lattice parameter, meaning that the Pd-Ni contained more Pd in its structure, as the size of Pd atomic radius is $10% larger than Ni (atomic radius of Pd 137 pm and atomic radius of Ni 125 pm. 82 The top of the reactor (reactor outlet) contained less Pd, however the mixed metallic alloy appeared to be retained. During the reduction step the lattice parameter of this mixed phase decreased, meaning that more Ni was incorporated in its structure as the reduction of Ni containing species (mainly NiAl 2 O 4 ) proceed. Regarding the crystallite size of metallic Ni, the observed gradient in the length of catalyst reactor may be a result of temperature gradients induced with the sample movement during the mapping measurements (i.e. in order to probe the top of reactor the sample stage needs to move down and thus the bottom of the reactor is exposed to lower temperatures during the mapping measurements); at the end of the reduction process the crystallite size of Ni species became more uniform with the crystallise size range between 16-25 nm.
The distribution of the various catalyst components during the reduction is shown in Fig. 7 (3D-XDR-CT data collected aer rst two XRD maps and before the third XRD map). The distribution of all Ni-containing phases was seen to be uniform. As concluded from the XRD maps in Fig. 5, the quantity of CeO 2 decreased during the reduction and the new distribution of CeO 2 was mainly found on the edges of catalyst particles, being now very similar to ZrO 2 . Also, the distribution of NiAl 2 O 4 during the reduction step was not uniform; during the data collection the NiAl 2 O 4 phase was found to be more reduced (i.e. more of NiAl 2 O 4 disappeared) in the upper parts of the catalyst bed (closer to reactor outlet) due to the time effect of measurement; the 3D-XRD-CT data were also collected in the order from reactor inlet to reactor outlet.

Partial oxidation of methane
Upon completion of the reduction/activation process, the POX reaction mixture was introduced into the reactor. Three different reaction conditions were tested: rst a reaction mixture with CH 4 : O 2 2 : 1 (POX1), then a reaction mixture with (Right) relative changes of each phase during the reduction step. XRD maps are presented from the inlet of reactor towards the outlet of reactor. The NiO phase is already present in traces when compared to initial composition at room temperature. Corresponding diffraction patterns can be found in the ESI (Fig. S12 †).    Corresponding diffraction patterns can be found in the ESI (Fig. S17 and S22 in the ESI. †). The XRD map collected at the end of POX2 reaction conditions (after 120 minutes) is presented in Fig. S23 in the ESI. † reaction mixture with optimal CH 4 : O 2 ratio intended to evaluate the impact of the changes in the catalyst structure and its performance, induced with the coke deposition.
During each reaction mixture, two XRD maps were collected (during the rst 10 minutes aer the introduction of reactive gases) in order to study the relatively fast solid-state changes taking place aer introducing the reaction mixture to the reactor. Then, 10 cross sections were collected during reaction (what we recently termed a 5D-XRD-CT imaging experiment) and nally, aer 80 min reaction a nal XRD map was collected (two XRD map for the reaction mixture POX2). In the case of the last reaction mixture (POX3), only ve XRD-CT datasets were collected due to experimental time limitations. Fig. 8 shows the distribution maps of CeO 2 , Pd-Ni and graphite, based on normalised scale factors, as well as the distribution of lattice parameter and crystallite size of metallic Ni. The changes of note include the disappearance of the Pd-Ni phase, which is quite noticeable at the end of POX1 reaction and occurs predominantly at the reactor inlet (Fig. S17 in the ESI †). The Pd-Ni peak was seen to continuously shi towards higher scattering angles during the reaction, at the same time becoming broader and of lower intensity. This phenomenon could be explained through the temperature-driven incorporation of Ni in the Pd-Ni structure resulting in the formation of a large spread of solid solutions with different Ni/Pd ratios 85,86 (Fig. S13 in the ESI †). For this reason, at the end of POX1 reaction the main reection of the Pd-Ni phase signicantly decreased.
In contrast reections for the metallic Ni phase (111 and 200) showed remarkably little change in height, a gradual shi towards lower scattering angle and only aer 80 min reection broadening (Fig. S18 in the ESI †); we propose that this phenomenon is due to intercalation of carbon into the Ni crystal structure. Although the formation of Ni-C occurs along the entire catalyst bed, it is more pronounced at the reactor inlet where the rst deposits of graphitic carbon were also identied. In this study, we modelled this phenomenon by rening the Ni crystallite size and lattice parameter values. We note that the changes seen in the behaviour of the metallic Ni species maps remarkably well onto the catalytic behaviour (CO/H 2 production) from both the operando study and a separate laboratory experiment (see Fig. S19 †). More specically, the catalyst reached its best performance during the POX1 reaction mixture aer 2 h of time on stream which could suggest that the interstitial solution of Ni-C is part of catalyst active species. It should be noted that we did not observe any reoxidation of Ni species for the duration of the POX experiment.
Aer introducing a methane rich gas stream (reaction mixture POX2), signicant formation of graphitic carbon could be observed in the rst ve minutes of reaction, this time occurring uniformly along the entire length of the catalyst bed. As expected, the deposition of coke was coincident with Fig. 9 Volume rendering of the normalised scale factors data volume (normalised over the maximum value scale factor reading in the entire process) obtained from the Rietveld analysis of the 3D-XRD-CT data collected during the reaction mixture POX1, POX2 and POX3. Plots on the right corresponds to normalise mean scale factors 3D volume during the POX reaction conditions (for normalised mean scale factors during the entire experiment see Fig. S24 †). The normalisation was done with respect to maximum value for the entire experiment. Note that only five XRD-CT slices were collected for POX3 reaction mixture. a decrease in crystallite size of Ni (observed as peak broadening) and an increase in its lattice parameter (observed as peak shi towards lower scattering angles), 87 suggesting that the prerequisite for graphite formation is the intercalation of carbon into the Ni lattice. Although, the presence of Ni-C was found to be crucial for the catalyst activity, its progressing carbonatation during the POX2 reaction mixture led eventually to a catalyst with lower activity. The accumulation of carbon species over Ni catalyst can effectively block the access to active components, through the formation of coke, lamentous carbon and/or metallic carbides. [88][89][90][91] The coke formation can possibly be explained with two different mechanisms: (1) methane dehydrogenation occurring for the high partial pressure of methane, observed in this study mainly at the reactor inlet at the end of POX1 reaction mixture and (2) CO disproportionation to carbon and CO 2 , observed in this study in the higher parts of reactor (towards reactor outlet) as the CO is expected to form through the reaction between CH 4 and CO 2 (formed through the initial combustion of CH 4 at the reactor inlet). 29,30,33 Under the employed reaction conditions in the present study (C/O $ 2) the formation of carbon is predicted and in good agreement with the C-H-O diagram determined from the thermodynamic calculations. 92 It is important to note that when the Ni-Pd/CeO 2 -ZrO 2 /Al 2 O 3 catalyst operates in the carbon-free region (C/O < 2), it shows high stability and no apparent formation of coke deposits. 93,94 The coke formation was seen to slow down at the end of reaction mixture POX2 while its quantity remained high for the duration of the POX3 reaction mixture measurements. The diffraction data show the presence of crystalline graphite till the end of the experiment (Fig. S20 †) and complementary laboratory TGA measurements (see Fig. S21 †) performed on the recovered catalyst sample did not reveal the presence of any other forms of coke.
The phase distribution volumes of all components present during the studied POX conditions (i.e. POX1, POX2 and POX3 reaction mixtures) are shown in Fig. 9 (the overall changes in normalised mean scale factors during the entire experiment are shown in Fig. S24 †). During the POX reaction, the distribution of the mixed CeO 2 -ZrO 2 solution remained unchanged. More importantly, the formation of graphite was found to develop on the particles rich in metallic Ni and supported on Al 2 O 3 , thus suggesting that the carbon was being built up in the catalyst particles, signicantly changing their shape (i.e. particles cracking) and density (i.e. particle expansion) (Fig. S25-S26 in the ESI †). Therefore, the apparent overall decrease in scattering signal intensity of all other catalyst components during the POX2 and POX3 reaction mixture is associated with a decrease in the density of the catalyst particles.
To gain further insight into the evolution of graphite and Ni crystallite size, a binary mask was created for each dataset based on the ZrO 2 distribution maps presented in Fig. 10A and applied to the graphite and Ni crystallite size maps. The average values for each dataset were calculated and the results are plotted in Fig. 10B. It can be clearly seen that the graphite forms and grows predominantly at the regions of the catalyst particles where CeO 2 -ZrO 2 is absent. However, it can also be seen that although CeO 2 -ZrO 2 can provide enhanced resistance to graphite it cannot prevent its growth (Fig. S25-S26 †).
The distribution of the Ni crystallite size during the reduction process and during the reaction mixture POX1 were comparable (Fig. 10). Generally, the core of the catalyst particles was seen to contain larger Ni crystallites when compared to the edges which is also evident from Fig. 10B. This provides evidence of a direct correlation between the presence of the CeO 2 -ZrO 2 promoter and the dispersion of Ni crystallites. 48,95 As the reaction mixture POX2 was introduced to the reactor, the Ni crystallite size generally decreased in the catalyst particles (with simultaneous increase in lattice parameter), which suggests the intercalation of carbon into the Ni structure. It can also be seen that the slope between POX1 and POX2 is more negative for the areas in the catalyst particles where CeO 2 -ZrO 2 is absent. This is consistent with the graphite results as it implies that the incorporation of the interstitial carbon in the Ni structure is taking place at a higher extent in the Ni supported on Al 2 O 3. However, a reversed egg-shell distribution for the Ni crystallite size is observed when comparing the results obtained from POX1 and POX3 reaction conditions. This implies that the metal-support interaction between Ni and CeO 2 -ZrO 2 is not sufficiently strong and indeed weaker compared to the Ni-Al 2 O 3 interaction and as a result the Ni crystallites gradually sinter under POX reaction conditions. The strong interaction between the metallic species and the support material was shown to be crucial in preventing the Ni species from sintering as well as improving the resistance to coke formation and growth, since larger crystallites are more prone to carbon deposits. 18,96,97  These results are further supported by the histograms presented in Fig. 11; these histograms are derived from the images that yielded the average values shown in Fig. 10. Specically, it can be seen that the shape of the Ni crystallite size distribution is similar for both the Al 2 O 3 and the CeO 2 -ZrO 2 areas but there is a right hand asymmetry towards larger values for the former. Importantly, in POX2, the Ni crystallite size distribution corresponding to the Al 2 O 3 areas is becoming wider and is shied signicantly towards lower values which, as previously discussed, is attributed to the interstitial carbon formation. Under POX3, this distribution is becoming narrower and shied to higher values implying the partial disappearance of this species. On the other hand, it is evident that the Ni crystallites are beginning to sinter signicantly in the CeO 2 -ZrO 2 areas; the wider and shied distribution towards larger values implies a weaker metal-support interaction.
The growth of graphitic coke had a detrimental effect on physical structure of catalyst particles as it led to their expansion, formation of cracks in their structure and even their breakage ( Fig. S27-S28 †). In addition, the formation of coke signicantly decreased the catalyst performance veried by operando mass spectrometry and laboratory quantitative measurements (Fig. S19 †). The catalyst showed an excellent performance with stoichiometric POX1 reaction mixture (2 : 1), reaching a H 2 yield above 90% and CH 4 conversion of 84%. Switching to POX2 reaction mixture resulted in a decrease of both H 2 yield and CH 4 conversion, as during these conditions there was not enough O 2 to convert the rich in CH 4 inlet stream resulting in signicant catalyst coking. Aer introduction of POX3 reaction mixture, the H 2 yield was lower when compared to POX1 reaction mixture (70% according to quantitative analysis) and the decrease in catalyst performance was due to the multiple physico-chemical solid-state chemistry changes and mechanical changes occurring during the POX2 reaction mixture (i.e. coke deposition and Ni crystallites sintering on the catalyst surface).

Summary and conclusions
In this work we presented the results from an operando 5D chemical imaging experiment during the reduction/activation and POX reaction employing a Ni-Pd/CeO 2 -ZrO 2 /Al 2 O 3 catalyst. Specically, we investigated in 3D the real-time evolution of the solid-state chemistry of this complex catalytic system under POX operating conditions. We aimed to gain a better understanding of the role of each catalyst component by looking at the changes in their physico-chemical properties upon placing the catalyst under both stoichiometric and CH 4 -rich reaction conditions. This multi-length scale chemical imaging experiment performed with a combination of 3D XRD-CT and XRD mapping allowed us to track the evolving solid-state chemistry taking place in the catalyst both at the particle and at the reactor level, investigating the changes in the range of mm to nm. We were able to observe signicant chemical gradients present both at the single catalyst particle level (i.e. intraparticle radial gradients) and along the reactor (i.e. reactor axial gradients). More specically we showed: (1) At operating temperatures (i.e. 800 C), the Pd-containing species are present in the form of a solid solution of Pd-Ni metallic alloys. The formation of such alloys under the ow of He clearly illustrates the promoting role of Pd, as the reduction of the Ni species occurred in an inert gas atmosphere. During the reduction stage, the quantity of metallic Ni incorporated in the Pd-Ni alloy was increased due to the reduction of the Nicontaining species (i.e. mainly NiAl 2 O 4 ). The further incorporation of Ni into the Pd-Ni resulted in the spread of the alloy mass population, leading to gradual decrease in the intensity of Pd-Ni diffraction peak, until it was no longer possible to observe it. The resolution of the reconstructed XRD-CT images (voxel size of 40 mm Â 40 mm Â 40 mm) was not high enough to allow us to extract any further information regarding the Pd-Ni metallic alloy during this experiment. However, our results suggest that the metallic Pd-Ni alloy is not stable under reaction conditions and one can argue that the signicant alloying of Pd with Ni may hinder the promoting role of Pd in this catalyst.
(2) During the POX reaction, the growth of crystalline graphite was observed with XRD mapping; rst deposits were identied at the reactor inlet aer 80 minutes under the initial reaction mixture with CH 4 : O 2 ratio of 2 : 1. This was coincident with the broadening and shi of Ni diffraction peaks towards lower scattering angles, which suggested that the presence of graphite was preceded by the formation of an interstitial solid solution of Ni-C, considered as the intermediate species formed in the methane activation step. In fact, the presence of such species in the structure of catalyst at its highest performance can only suggest that Ni-C is part of the catalyst active species. Aer switching to reaction mixture with CH 4 : O 2 ratio of 4 : 1 (reducing conditions) the formation graphite was observed instantaneously, while the activity of catalyst was gradually decreasing with time on stream. The accumulation of carbon species due to their insufficient removal (both through gas and bulk oxygen reaction) suggests that the catalyst active species are susceptible to changes in the gas phase composition. It is important to note that both the analysis of the diffraction data and post-reaction thermogravimetric measurements do not suggest the presence of any other major carbonic species, such as amorphous carbon.
(3) The coke deposition during the POX reaction, especially when the CH 4 rich stream was applied, occurred predominantly in regions where the mixed binary solution of CeO 2 -ZrO 2 was absent. With the diffraction imaging technique, we showed that across one catalyst particle it is possible to differentiate between two regions: (1) core of catalyst particles where the Ni is supported over Al 2 O 3 and (2) shell of catalyst particles, where the Ni is supported on CeO 2 -ZrO 2 /Al 2 O 3 . The rst region is characterised with the metallic Ni strongly interacting with Al 2 O 3 support, which provides a good resistance towards the Ni particles sintering. However, the Al 2 O 3 support, with both Lewis and Brønsted acidity, promotes coke deposition in the POX reaction. The second region, which contains the CeO 2 -ZrO 2 support, although it exhibits enhanced resistance to coke formation due to the oxygen storage properties of CeO 2 -ZrO 2 , it does not prevent the Ni crystallites from sintering under reaction conditions, which is most likely caused by a weaker metal-support interaction. The characterisation of catalysts with such design, where the support materials exhibit different properties (inactive and acidic Al 2 O 3 vs. redox CeO 2 -ZrO 2 ), should be performed with spatially-resolved techniques as the conventional bulk measurements provide only average information and do not take into account the intraparticle heterogeneities.
(4) During the catalyst activation/reduction step, the CeO 2 species were stabilised only in the regions where the ZrO 2 was present (i.e. catalyst particle shell) in the form of a binary CeO 2 -ZrO 2 solution. The CeO 2 species present initially in the particle core were seen to disappear during the reduction, most likely interacting with the Al 2 O 3 support and forming the CeAlO 3 phase. Although it has been proposed that the CeAlO 3 phase may be effective in carbon removal, through the chemical cycle of Ce 3+ /Ce 4+ 40,98 our results suggest that only the interaction of CeO 2 with ZrO 2 yields a support that can successfully prevent coke deposition whereas the interaction between CeO 2 and Al 2 O 3 decreases the redox properties of the CeO 2 oxygen storage material. This observation also showed that more attention should be paid to the preparation method, aiming towards improved distribution of the ZrO 2 in the catalyst core, which would help to stabilise the CeO 2 and mitigate the coke deposition on the metallic Ni.
With the 5D diffraction imaging technique we were able to investigate an inhomogeneous by design Ni/Al 2 O 3 based catalyst. The analysis of the spatially-resolved diffraction patterns provided new insight into the structural changes induced by the operating conditions, not possible to obtain with conventional bulk measurements. The current composition yields a catalyst with high initial performance and we believe that its stability can be further improved by dispersing more uniformly the Pdcontaining species and enhancing the interplay between the two distinctive particle regions (i.e. core and shell regions).

Data availability
All data are available from the corresponding authors on reasonable request.

Conflicts of interest
There are no conicts to declare.