Self-regeneration of supported transition metals by a high entropy-driven principle

The sintering of Supported Transition Metal Catalysts (STMCs) is a core issue during high temperature catalysis. Perovskite oxides as host matrix for STMCs are proven to be sintering-resistance, leading to a family of self-regenerative materials. However, none other design principles for self-regenerative catalysts were put forward since 2002, which cannot satisfy diverse catalytic processes. Herein, inspired by the principle of high entropy-stabilized structure, a concept whether entropy driving force could promote the self-regeneration process is proposed. To verify it, a high entropy cubic Zr0.5(NiFeCuMnCo)0.5Ox is constructed as a host model, and interestingly in situ reversible exsolution-dissolution of supported metallic species are observed in multi redox cycles. Notably, in situ exsolved transition metals from high entropy Zr0.5(NiFeCuMnCo)0.5Ox support, whose entropic contribution (TΔSconfig = T⋆12.7 J mol−1 K−1) is predominant in ∆G, affording ultrahigh thermal stability in long-term CO2 hydrogenation (400 °C, >500 h). Current theory may inspire more STWCs with excellent sintering-resistance performance.

S upported metal catalysts are of great importance in both industrial and fundamental catalysis 1 . The methods for generating active sites on metal oxide carriers include chemical or physical deposition techniques (e.g., impregnation method 2 , deposition-precipitation method 3 , co-precipitation process 4 , atomic layer deposition 5 , spraying 6 and so on 7 . Needless to say, those methods have significantly promoted the synthetic chemistry of supported metal catalysts. For example, the control of metal nanocrystal size by modified deposition method reveals the attractive role of corner atoms, which offers a theoretical guidance for catalyst design 8 . Nevertheless, the deactivation of supported metal catalysts, especially those metal nanoparticles (NPs) with low tamman temperatures (e.g., Cu 9 , Ag 10 , Au 11 ) and high surface energies 12 , is often inevitable by sintering 13 under long-term exposure to the high temperature or redox environments. For example, Au/TiO 2 catalysts afforded an ultralow temperature for CO oxidation 14 , while particle growth leading to the activity loss often occurred during high temperature operation 15 .
Early in this century, a self-regenerative strategy by in situ exsolution of active species from host matrix was presented by Nishihatas' group 16 , which can greatly prolong the lifetime of three-way catalysts. The emergence of perovskite-based (ABO 3 ) self-regenerative catalysts exploits a useful approach to optimize catalyst activity and stability [16][17][18][19][20][21] . To our knowledge, the control of nonstoichiometry 18 (A-site defect interactions), judicious choice of B-site composition (e.g., Pd 17 , Pt 22 , Ni 23 , CoFe alloy 24 ) and the surface modification (polished or pretreated perovskite) are frequently used to tailor B-site metals exsolution. This type of catalytic materials can not only in situ generate active sites, but also accomplish reintegration of those species back to parent perovskites during oxidation-reduction cycles [16][17][18][19][20][21] . Meanwhile, introducing high-valence cations (Nb 25 , Ti 26 , Ga 27 , etc.) into ABO 3 structure can improve the structural stability of the parent perovskites, which contributes to the reintegration process during catalysis 19 . This process effectively suppresses the sintering or agglomeration of the supported metals 16 . However, these types of host matrix are limited to perovskite crystal structures and the state-of-art self-generation systems are perovskite-based catalysts [16][17][18][19] (Fig. 1a), which cannot meet the broad requirements by different catalytic processes. To date, the proposal of a selfregenerative principle for the design of sintering-resistance catalysts is still desired.
Since the discovery of high entropy alloys in 2004 28 and the pioneering introduction of high-entropy oxides (HEOs: oxide solid solutions contain five or more distinct metal cations) in 2015 29 , entropy-stabilized materials such as oxides 30 , sulfides 31 , phosphide 32 , nitrides 33 , carbides 34 , and borides 35 , are flourishing in the library of advanced materials during past five years. Inspired by the principle of thermodynamics 29,30 (Entropic contribution could predominate the thermodynamic landscape), applying the theory of high entropy materials to trigger the selfregeneration during catalytic redox processes may be a choice. First, based on the configurational entropy 34 (it refers to the number of conformations of a molecule and the number of ways that atoms or molecules pack together in the oxide) formula ( Fig. 1.bi), more disorder of system and higher randomness of structure with a lower Gibbs free energy can contribute to the stability of host structure, especially during high temperature catalysis (ΔG = H-TΔS) 31,36 (Fig. 1bii). From another perspective, the reintegration process can be promoted due to an entropy-added motivation by the oxidative transformation of isolated metals back into parent metal oxides. The theoretical conjecture stimulates us to carry out a number of representative experiments.
In this work, we verify the entropy-driven principle for selfregenerative catalyst by high entropy ZrO 2 . A single-phase cubic Zr 0.5 (NiFeCuMnCo) 0.5 O x , whose entropic contribution (TΔS config = T*12.7 J mol −1 K −1 ) is predominant in ΔG 28,29,31 , was prepared and rationally selected as a model catalyst. Interestingly, reductive H 2 treatment could induce the exsolution of CoFeCuNi alloy NPs on the surface of parent high entropy ZrO 2 NPs. With high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x as the host matrix, the exsolution and reintegration of CoFeCuNi alloy were observed during three redox cycles (H 2 600°C + Air 550°C), an interesting feature that may confine metal species under dynamics catalysis (Fig. 1biii). As a proof of principle, the entropy-driven self-regenerative property endows Zr 0.5 (NiFeCuMnCo) 0.5 O x catalyst good stability during CO 2 hydrogenation reaction. The Zr 0.5 (NiFeCuMnCo) 0.5 O x catalyst functioned well at 400°C during 500 h continuous hydrogenation, and at the same time no phase separation or sintering were found on the reused catalyst. The self-regenerative talent of high entropy material may open an alternative for the design of sintering-resistance catalysts.

Results
The formation of high entropy cubic Zr 0.5 (NiFeCuMnCo) 0.5 O x . Inspired by the principle of the high entropy-stabilized structure, the synthesis of high entropy ZrO 2 catalyst was carefully investigated by a mechanochemical process (Fig. 2). Five common transition metal salts (Ni, Fe, Cu, Mn, Co) were mixed with ZrCl 4 and NaOH, followed by calcination in air and sodium salt removal. Definitely, for transition metal dopants, the matching radius (the rationally selected five metals j r Zr -r x r Zr j < 20% 37 40 , and then a detailed study for constructing high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x was carried out. Firstly, the calcination temperature was adjusted to explore the phase evolution of Zr 0.5 (NiFeCuMnCo) 0.5 O x precursors. As shown in the Fig. 3a, obvious diffraction peaks for the typical cubic ZrO 2 phase started to form at 550°C. Notably, no other impurity peaks were observed at an elevated temperature (e.g., 600°C). This phenomenon indicated the possible formation of a high entropy cubic Zr 0.5 (NiFeCuMnCo) 0.5 O x . Then, the solubility of five metals (Ni, Fe, Cu, Mn, Co) within ZrO 2 solid solution was carefully investigated by adjusting their total ratio (0.3-0.9) (Fig. 3b). The single cubic structure of ZrO 2 was well maintained in Zr 0.7 (NiFeCuMnCo) 0.3 O x and Zr 0.5 (NiFeCuMnCo) 0.5 O x , while impure reflections of other crystal structures were detected with a higher ratio (0.7-0.9) of doped metals (Fig. 3b). The molar ratio (Zr:Ni:Fe:Cu:Mn:Co= 5:1:1:1:1:1) was then fixed, and a relative high configurational entropy (S config = 12.47 J mol −1 K −1 ) could be expected.
Rietveld refinement was then performed to determine the structural parameters of Zr 0.5 (NiFeCuMnCo) 0.5 O x . The initial structural model was built based on the cubic ZrO 2 (a = b = c = 5.100 Å). The calculated lattice constant of Zr 0.5 (NiFeCuMn-Co) 0.5 O x was a = b = c = 5.097 Å (Fig. 3c, d). The difference in atomic sizes between Zr and transition metals would cause the shrinkage of the lattice of ZrO 2 matrix 41 . However, it should be noted that the abundance of oxygen vacancies in doped ZrO 2 may enlarge the ZrO 2 lattice 42 . Meanwhile, the R-factor (Rp (%)=4.6), weighted profiles R-factor (Rwp (%)=4.7) and chi square (χ 2 = 1.330) of the HE ZrO 2 were quite good (Fig. 3c). The simulated PXRD pattern of the high entropy ZrO 2 was in good agreement with the experimental one. Those XRD results suggest the formation of cubic Zr 0.5 (NiFeCuMnCo) 0. 5 With the option temperature (550°C) and composition in hand, a high specific surface area (84 m 2 g −1 ) was obtained by optimizing the dosage of NaCl additive (Supplementary Table 1. Supplementary Fig. 3). To the best of our knowledge, the high temperature for phase transformation (e.g., 900-1300°C) in entropy-driving force (TΔS) would cause the collapse of pores, leading to HEOs with low surface areas, such as, NiMgCuZn-CoOx by solid-state combustion (2 m 2 g −1 ) 43 , NiMgCuZnCoO x by citric acid-based sol-gel method (28 m 2 g −1 ) 43 , (MgTiZnCu-Fe) 3 O 4 by solid-state combustion (12 m 2 g −1 ) 44 , CoNiCuMgZ-nO x with graphene oxide as a sacrificial template (42 m 2 g −1 ) 45 . The current porosity (84 m 2 g −1 ) may be attributed to the particle breakage by mechanochemistry 46 , and meanwhile a mild crystallization temperature (550°C) could effectively avoid the excessive growth of fine NPs and interstitial porosity was retained 47 .
In order to observe the morphology of the Zr 0.5 (NiFeCuMn-Co) 0.5 O x sample, the high-angle annular dark-field signal (HAADF) and energy dispersive spectrometer (EDS) elemental  (Fig. 4). Zr 0.5 (NiFeCuMnCo) 0.5 O x was composed of small NPs (e.g., 10-30 nm), which aggregated together and resulted in abundant interstitial porosity. This feature was in agreement with the found N 2 sorption isotherm of Zr 0.5 (NiFeCuMnCo) 0.5 O x (S BET = 84 m 2 g −1 ). Moreover, the EDS elemental maps of individual metals showed that the five dopants were homogeneously distributed in the ZrO 2 backbone, suggesting that five metals-doped ZrO 2 solid solution was constructed.   Table 2), which matched well with the theoretical formula. These investigations encouraged us to explore its self-regeneration possibility during redox cycles.
The exploration of self-regeneration process. To verify the exsolution and reintegration behavior, several exploratory experiments were carried out with high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x as a model. The exsolution of metals was determined by the reducibility of transition metal oxides 48 , and it was closely related to co-segregation energy (transition metal accompanying with oxygen vacancies) 49 . The ZrO 2 doped with low valence metals could contribute to produce oxygen vacancies 50 , which promoted the exsolution of metals 51 under reductive atmosphere. Subsequently, the H 2 -TPR spectrum was tested to probe the reduction behavior of  Supplementary Fig. 4c), the reduction peaks at 269°C and 325°C were attributed to the reduction of the CuO to Cu 0 and manganese oxides to MnO, respectively 54 . All three types of nickel oxide (363, 467, 570°C) were reduced to the Ni 0 ( Supplementary Fig. 4d) 55 . For Zr 0.5 Fe 0.5 O x sample (Supplementary Fig. 4e), the TPR profile showed two distinct reduction peaks. The peak centered at 401°C was assigned to the reduction of Fe 2 O 3 to Fe 3 O 4 and the peak at 566°C was ascribed to the reduction of Fe 3 O 4 to Fe 0 56 . For cobalt-doped sample ( Supplementary Fig. 4f), the reduction peaks around 295°C and 408°C were attributed to the transitions from Co 3+ to Co 2+ and Co 2+ to Co 0 57 . Indeed, the addition of CuO could decrease the reduction temperature ( Supplementary Fig. 4a, b, c) 54 . Interestingly, only one strong reduction peak was observed at~279°C for Zr 0.5 (NiFeCuMnCo) 0.5 O x sample ( Supplementary Fig. 5). Then, with CuO as the standard reference, the H 2 consumption of Zr 0.5 (NiFeCuMnCo) 0.5 O 1.1+y sample (1.1 ascribes to unreducible ZrO 2 58 , and MnO 49 ) can reach 3.63 mmol g −1 . Then, the y value is calculated to be 0.36 59 . In addition, the y values of other samples were provided (Supplementary Table 3). Based on this result, the Zr 0.5 (NiFeCuMnCo) 0.5 O 1.46 formula suggests that significant amount of oxygen vacancies were created. Hence, these preliminary studies encourage further research about the exsolution process.
Subsequently, the pristine high entropy Zr 0.5 (NiFeCuMn-Co) 0.5 O x was treated in reduction condition (H 2 ) at appointed temperatures for 2 h. As shown in Fig. 5a, the early stage of phase transformation was not clear at 400°C. When the reduction temperature raised to 500-600°C, the diffraction pattern of the exsolution phase showed a broad peak at 44°C, which could be assigned to the reflection of transition metals (111). At the same time, typical peaks for cubic ZrO 2 were still preserved during H 2 treatment. In sharp contrast, the exsolution diffraction peaks of as-prepared binary Zr 0.5 Cu 0.5 O x were sharp, revealing the excessive growth of Cu particles (average particle size by Reitveld refinement analysis: 37.7 nm) 60 (Fig. 5b, Supplementary Fig. 6). Meanwhile, the exsolution peaks of quaternary Zr 0.5 (NiCu-Co) 0.5 O x presented two different phases (Fig. 5b). The multi redox behavior of Zr 0.5 (NiFeCuMnCo) 0.5 O x was detailedly studied in the following (Fig. 5c). Therefore, high-entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x as a host matrix could somewhat suppress the growth or agglomeration of exsolved metal NPs even under harsh reductive conditions (e.g., 600°C), which stood on the surface of parent high entropy ZrO 2 NPs.
Then, the Zr 0.5 (NiFeCuMnCo) 0.5 O x (treated at 400°C, H 2 ) was investigated by STEM and elemental mapping analysis. Firstly, STEM was employed to provide insights into the exsolved NPs in Zr 0.5 (NiFeCuMnCo) 0.5 O x under reducing atmosphere. As shown in Supplementary Fig. 7, reductive H 2 treatment induced the exsolution of grains growing on the surface of parent high entropy ZrO 2 NPs. Meanwhile, elemental mapping analysis exhibited aggregation of some elements (Supplementary Fig. 7b). To further confirm the exsolved metal species, the high resolution elemental mapping images at the edge of the sample revealed the formation of the CoFeCuNi alloy NPs (Fig. 6). To further study clearly the metal exsolution, the O elemental mapping was added ( Supplementary Fig. 8). A typical particle on the edge was highlighted by yellow cycle. It showed a relatively low O content, close to the background. This result somewhat demonstrated the formation of the metallic CoFeCuNi particle.
To further understand current observation, in situ highresolution XPS (Fig. 7a, b) was carried out to investigate the surface state of the Zr 0.5 (NiFeCuMnCo) 0.5 O x during H 2 treatment from RT to 873 K. As shown in Fig. 7c, the peak centered at 778.3 eV was assigned to the characteristic Co 2P 3/2 peak of Co 0 61 . Meanwhile, the sharp peak at 852.7 eV resulted from surface Ni 0 (Fig. 7d) 61 . Since the Binding Energy of Cu + generally overlaps with Cu 0 in the Cu 2 P core level (Fig. 7e, peak at 932.6 eV), the X-ray induced Auger electron spectra in the kinetic energy region of 928-906 eV is provided in Fig. 7f. As shown in Fig. 7f, the kinetic energy at 918.6 eV could be contributed to Cu 0 61 . A signal peak is located at 706.9 eV, which is ascribed to the Fe 2P 3/2 metal (Fig. 7g) 61 . As expected, there is no peak for Mn 0 at 638.8 eV 61 . Meanwhile, MnO 2P 3/2 (at 641.0 eV) peaks show their characteristic satellite peak at 636.9 eV (Fig. 7h) 62 . Indeed, manganic oxide is more difficult to be reduced to Mn 0 63 . Under H 2 treatment, several transition metal ions (Co, Ni, Cu, Fe) seem to in situ migrate and separate from the bulk of high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x , while the original cubic ZrO 2 matrix was well maintained (Fig. 5a). The H 2 -induced formation of exsolved CoNiCuFe NPs is different with the Cantor alloy (FeCrMnNiCo) by melting process 64 and those CoNiCuFe NPs could not be described as high entropy alloys 65 . As illustrated in Supplementary Fig. 5, only one strong reduction peak was observed at~279°C, which could be assigned to the produce of the CoFeCuNi alloy. Based on these results, the exsolution phenomenon of alloys occurred under reductive atmosphere, which was just the first step of self-regenerative process.  To figure out the self-regeneration possibility of high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x , the dissolution process was subsequently explored. The reduced Zr 0.5 (NiFeCuMnCo) 0.5 O x (treated at 600°C, H 2 ) was further exposed to oxidative environment (550°C, air). After this cycle, the obvious XRD peaks for CoFeCuNi alloy phase disappeared (Fig. 5c). The interesting exsolution-reintegration behavior of the high entropy ZrO 2 material was further proved by three redox cycles (Fig. 5c). The oxidized sample after three cycles was observed by STEM images. The elemental mapping images in different scale bars, such as 50 nm, 200 nm, 50 μm, have been taken. As shown in Fig. 8 and Supplementary Fig. 9, all five heterometal species were evenly distributed in ZrO 2 matrix, which was the same as the pristine Zr 0.5 (NiFeCuMnCo) 0.5 O x (Fig. 4). As expected, the entropyadded motivation may drive the reintegration of CoFeCuNi alloy NPs into the parent ZrO 2 matrix in air, which contributed to the reversible self-regenerative process. All those results above suggested the feasibility of the high entropy ZrO 2 as a host matrix for self-regenerative process.
The catalytic performance of high entropy ZrO 2 catalyst. Then, the activity and thermal stability of high entropy Zr 0.5 (NiFe-CuMnCo) 0.5 O x were explored in high temperature catalysis (Fig. 9). CO 2 hydrogenation (reverse water gas shift, RWGS), with an enthalpy ΔH of 41.3 kJ mol −1 , is a typical endothermic reaction 66 . From the sight of both kinetics and thermodynamic, high temperature is more conducive to accelerating the reaction rate and moving the equilibrium in the direction of CO generation 67 . In other words, RWGS reaction is usually performed at high temperatures (e.g., 300-500°C), whose catalysts are easy to be sintered 68 . In addition, a redox reaction mechanism of active sites was proposed to explain this process 69 Fig. 10a-d), the Zr 0.5 (NiFeCuMnCo) 0.5 O x afforded better catalytic activity during CO 2 hydrogenation (300-400°C). When the temperature reached 400°C, the CO 2 conversion was~29% by Zr 0.5 (NiFeCuMn-Co) 0.5 O x and the selectivity for CO is over 90% (Fig. 9a,  Supplementary Fig. 10e). Then, the long-term stability of Zr 0.5 (NiFeCuMnCo) 0.5 O x during CO 2 hydrogenation reaction was studied. No obvious loss of catalytic activity and selectivity was found during continuous CO 2 hydrogenation for 500 h (Fig. 9b). In sharp contrast, severe deactivation by the ternary doped Zr 0.5 (MnCu) 0.5 O x catalyst was observed in relatively short reaction time (40 h) (Fig. 9b). Meanwhile, the XRD pattern of spent Zr 0.5 (NiFeCuMnCo) 0.5 O x catalyst showed little tendency towards segregation or sintering, further illustrating the advantage of high entropy (Zr 0.5 (NiFeCuMnCo) 0.5 O x ) as the host structure (Fig. 9c). In addition, the dissolution of metal alloys somewhat underwent during CO 2 atmosphere (600°C), as suggested by the diminished XRD peaks for metallic species (Fig. 9d). Furthermore, control experiments were carried out for comparison. First, the 5%(NiFeCuCo)/ZrO x and 10%(NiFe-CuCo)/ZrO x were synthesized by wet-impregnation method. The long-term stability of supported NiFeCuCo catalyst during CO 2 hydrogenation reaction was studied at the same condition ( Supplementary Fig. 11). Unfortunately, severe deactivation was observed in relatively short reaction time (1 h and 3 h). Compared with the high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x material (>500 h), 5%(NiFeCuCo)/ZrO x and 10%(NiFeCuCo)/ZrO x exhibited poor stability. Based on those results, unique properties of the high entropy structure indeed promoted the thermal stability of doped ZrO 2 catalysts.

Discussion
In summary, a cubic high-entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x was designed and simply prepared. The contribution of high entropy for the sintering-resistance ability of supported transition metals was figured out, while severe particle growth occurred on Zr 0.5 Cu 0.5 O x and Zr 0.5 (CuCoNi) 0.5 O x treated in H 2 . Meanwhile, the reversible exsolution-dissolution of fine CoFeCuNi alloys on high entropy Zr 0.5 (NiFeCuMnCo) 0.5 O x was observed in H 2 /Air treatments. Those results somewhat validate that entropy-driving force may contribute to selfregenerative process and sintering-resistance performance of supported transition metals. During high temperature catalysis (400°C), no obvious loss of catalytic activity and little tendency towards segregation were found during continuous CO 2 hydrogenation for 500 h, once again arguing for the excellent sintering-resistance ability of high entropy catalysts.
The high entropy oxide was invented in 2015 29 and related catalysis has been unveiled in around 2018 70 , and their unexpected talent for heterogeneous catalysis is a young topic 63 , yet full of promise. The idea of considering entropic contribution for stabilized catalyst is just the beginning, and more unique materials for heterogeneous catalysis would be inspired in the near future.
The as-made catalysts (75 or 50 mg) were used for the H 2 -TPR experiments (Micromeritics Autochem II 2920). The material was pretreated in Ar (30 mL min −1 ) at 300°C for 60 min and then cooled to the room temperature. Then, the 5% H 2 in Ar (30 mL min −1 ) was switched on in a heating rate of 10°C min −1 . Finally, the H 2 consumption of the as-made materials was monitored by a TCD detector.
The RWGS reaction worked at atmospheric pressure in a fixed-bed quartz reactor. Firstly, 30 mg catalyst and 100 mg quartz sand were mixed together well. Then, the catalyst with quartz was exposed to a stream of feed gas (H 2 :CO 2 = 3:1, 10 mL min −1 ). When the catalyst reached the appointed temperature, the gaseous products were analyzed by gas chromatograph with a TCD and FID detector.
Detailed procedures are in the Supplementary Methods section.

Data availability
The authors declare data supporting the findings of this study are available within the paper and its Supplementary Information. All data are available from the authors on reasonable request.