eg occupancy as an effective descriptor for the catalytic activity of perovskite oxide-based peroxidase mimics

A peroxidase catalyzes the oxidation of a substrate with a peroxide. The search for peroxidase-like and other enzyme-like nanomaterials (called nanozymes) mainly relies on trial-and-error strategies, due to the lack of predictive descriptors. To fill this gap, here we investigate the occupancy of eg orbitals as a possible descriptor for the peroxidase-like activity of transition metal oxide (including perovskite oxide) nanozymes. Both experimental measurements and density functional theory calculations reveal a volcano relationship between the eg occupancy and nanozymes’ activity, with the highest peroxidase-like activities corresponding to eg occupancies of ~1.2. LaNiO3-δ, optimized based on the eg occupancy, exhibits an activity one to two orders of magnitude higher than that of other representative peroxidase-like nanozymes. This study shows that the eg occupancy is a predictive descriptor to guide the design of peroxidase-like nanozymes; in addition, it provides detailed insight into the catalytic mechanism of peroxidase-like nanozymes.

Many nanomaterials, including systems based on various transition metal oxides (TMOs), have been explored as possible peroxidase mimics 14 . For example, Yan and colleagues 1,2,19 discovered the unexpected peroxidase-like activity of iron oxide nanoparticles, which were then applied to Ebola detection and tumor immunostaining. We have recently developed Ni oxide-based peroxidase mimics for glucose detection in serum 20 . However, these peroxidase-like nanozymes are generally developed using trial-and-error strategies 14 . The prevalence of empirical approaches is due to the lack of predictive descriptorsstructural characteristics of the nanomaterials that can be used as proxies for their peroxidase-like activities. This lack of predictive descriptors significantly hampers the identification of more active nanozymes.
In this study, we aim to identify a predictive descriptor for TMO-based peroxidase mimics. We reason that the e g occupancy (i.e., the d-electron population of the e g (σ*) antibonding orbitals associated with the transition metal sites) may control the peroxidase-like activity of perovskite TMOs because of the central role of oxygen species in these biomimetic catalytic reactions. We choose ABO 3 -type perovskite TMOs with BO 6 octahedral subunits (where A is a rare earth or alkaline-earth metal and B is a transition metal) as a model system due not only to their low cost and ease of preparation, but, more importantly, also to their diverse and controllable structural and catalytic properties ( Fig. 1a) 24 , which may facilitate the tuning of the e g occupancy by adjusting the ABO 3 composition. We show that the peroxidaselike activity of ABO 3 -type perovskite TMOs is primarily governed by their e g occupancy. In particular, we identify a volcano relationship between the e g occupancy and the specific catalytic activity of perovskite TMO-based peroxidase mimics: namely, perovskite TMOs with an e g occupancy of~1.2 and 0 (or 2) exhibit the highest and the lowest peroxidase-like activity, respectively. These conclusions are further rationalized by density functional theory (DFT) calculations. The identified descriptor successfully predicts the peroxidase-like activity of binary TMOs with octahedral coordination geometries.

Results
Identification of nanozyme activity descriptor. The perovskite samples were prepared by a sol-gel method followed by annealing at the desired temperatures. The as-prepared perovskites were fully characterized by scanning electron microscope (SEM), transmission electron microscope (TEM), powder X-ray diffraction (PXRD), inductively coupled plasma-optical emission spectroscopy (ICP-OES) and Brunauer-Emmett-Teller (BET) surface  area measurements (see Supplementary Figs 1-13 and Supplementary Tables 1-2 for details). The amount of oxygen vacancies  in the perovskites was quantified by iodometric titrations (Supplementary Table 3). To identify a suitable descriptor for the peroxidase-like activity of perovskite TMOs, we initially examined La 1-x Sr x FeO 3-δ compositions (x = 0-1), because the e g occupancy of Fe in this series of perovskites could be gradually tuned by substituting La 3+ with Sr 2+ cation ( Fig. 1b and Supplementary Table 4). Such a substitution would shift the oxidation state of Fe from + 3 in LaFeO 3 to + 3.31 in SrFeO 3-δ , resulting in the corresponding change of the e g occupancy of Fe from 2 to 1.69 (note: as there are no fractional electrons occupying these orbitals, the e g occupancies presented in the current study are averages between integer occupations). A representative TEM image of LaFeO 3 (Fig. 1c) reveals the typical irregular morphology of perovskites with nanoscale features. The formation of phase-pure La 1-x Sr x FeO 3-δ (x = 0, 0.5, and 1) perovskite structures was confirmed by matching their PXRD data to the standard pattern of LaFeO 3 (JCPDS card number 75-0541) (Fig. 1d).
The peroxidase-like activity of the perovskite-based nanozymes was assessed by using absorption spectroscopy to monitor the catalytic oxidation of 3,3′,5,5′-tetramethylbenzidine (TMB, a typical peroxidase substrate) with H 2 O 2 in the presence of the nanozymes. The oxidation of TMB generates an oxidized product ( ox TMB) with a characteristic absorption peak at 652 nm. The intensity of this absorption peak (A 652 ) increased with increasing Sr content and the highest absorption was obtained for SrFeO 3-δ (Fig. 1e). The time evolution of the A 652 value (Fig. 1f) shows that SrFeO 3-δ also exhibited the fastest reaction kinetics, demonstrating that the Sr substitution effectively enhanced the peroxidaselike activity of La 1-x Sr x FeO 3-δ . The mass-based peroxidase-like activities of nanozymes were measured by steady-state kinetics assays (see Methods section). To separate the effect of surface area from the intrinsic peroxidase-like activity of perovskites (including the Fe-based perovskites discussed in this section), their specific activity (i.e., the mass activity normalized to the surface area) was also calculated, based on the BET surface areas obtained by nitrogen desorption measurements (Supplementary Figs 3 and 6, and Supplementary Table 1). As shown in Fig. 1g, the specific activity of SrFeO 3-δ was 7.76 and 1.79 times higher than that of LaFeO 3 and La 0.5 Sr 0.5 FeO 3-δ , respectively. The dependence of the specific activity on the Sr content of La 1-x Sr x FeO 3-δ and the e g occupancy of Fe is plotted in Fig. 1h. A substantial improvement in the peroxidase-like activity of La 1-x Sr x FeO 3-δ was observed as the Sr content increased from 0 to 1 and the e g occupancy of Fe decreased from 2 to 1.69.
To study the effect of e g occupancy lower than 1 on the peroxidase-like activity of perovskites, we investigated three Mnbased perovskites with the e g occupancy of Mn varying from 0.68 tõ 0.08 (i.e., e g = 0.68, 0.53, and 0.08 for LaMnO 3-δ , La 0.5 Sr 0.5 MnO 3-δ , and CaMnO 3-δ , respectively). The SEM and TEM images shown in Supplementary Figs 4 and 5, and the PXRD patterns in Supplementary Fig. 7a demonstrate the successful synthesis of the Mn-based perovskites. As shown in Supplementary Fig. 7c, the specific activity of LaMnO 3-δ was 1.42 and 43.29 times higher than that of La 0.5 Sr 0.5 MnO 3-δ and CaMnO 3-δ , respectively. Supplementary Fig. 7e shows the effect of the e g occupancy on the specific peroxidase-like activity of the Mn-based perovskites. Different from the trend observed for the Fe-based perovskites (i.e., the peroxidaselike activity increased as the e g occupancy decreased), the catalytic activity of the Mn-based perovskites decreased as the e g occupancy further decreased from 0.68 to~0.08. Taken together, the above results show a strong but non-monotonic correlation between e g occupancy and peroxidase-like activity of perovskites, suggesting the e g occupancy as a potential activity descriptor.
Evaluation of e g occupancy as nanozyme activity descriptor. To further evaluate the correlation between the e g occupancy and their peroxidase-like activity, overall ten perovskite TMOs covering e g occupancies of 0-2, six as described above and four others (i.e., LaCrO 3 , LaCoO 3-δ , LaNiO 3-δ , and LaMn 0.5 Ni 0.5 O 3 ) were investigated (Supplementary Figs 8-14, Supplementary  Table 6, and Supplementary Notes 1-2). As shown in the summary in Fig. 2a, the perovskite TMOs exhibited significantly different specific peroxidase-like activities. Some of them (such as LaNiO 3-δ ) exhibited a high activity, whereas the activity of others (such as LaCrO 3 ) was negligible. This behavior can be understood by plotting the activities of the ten perovskite TMOs as a function of the corresponding e g occupancy associated with the B cations: a definitive volcano relationship is obtained (Fig. 2b). The massbased peroxidase-like activities (i.e., mass activities) of the perovskite TMOs also show a volcano dependence on the corresponding e g occupancies ( Supplementary Fig. 15), confirming that the catalytic activity of the perovskite TMO-based peroxidase mimics is primarily governed by the e g occupancy of the B cations. In particular, perovskite TMOs with e g occupancy of~1.2 exhibit the highest peroxidase-like activity (Fig. 2).
Evaluation of other parameters as potential descriptors. As several other potential descriptors (i.e., oxidation state of transition metal, 3d electron number of B-site ions, O 2p-band center, and B-O covalency) have been studied to predict the electrocatalytic and photocatalytic activities of perovskites, we also investigated the relationship between the peroxidase-like activity h Specific peroxidase-like activity of the Fe-based perovskite TMOs as a function of e g occupancy. Source data are provided as a Source Data file and these parameters. As shown in Supplementary Fig. 16, although the oxidation state of B sites affects the peroxidase-like activity of perovskites, there is no apparent relationship between them. These results indicated that the oxidation state of B sites is not an effective descriptor and cannot provide guidance for the rational design of peroxidase-like nanozymes. We then studied the relationship between the peroxidase-like activity and the 3d electron number of B-site ions. As shown in Supplementary  Fig. 17, an "M-shaped" relationship with the maximum peroxidase-like activities around d 4 and d 7 was obtained. Clearly, although the 3d electron number is indicative, it is not a straightforward descriptor, as two maxima are associated with it. Several recent studies suggested that the O 2p-band center could be a better activity descriptor than the e g occupancy to design catalysts for oxygen reduction reaction and oxygen evolution reaction 33,34,40 ; therefore, we also evaluated it as a potential descriptor for the peroxidase-like activity of perovskites (Supplementary Note 6). As shown in Supplementary Fig. 18, the O 2p-band center was not well correlated with the peroxidase-like activity of perovskites, suggesting that it is not an effective descriptor for the perovskite-based peroxidase mimics. Last, we studied the relationship between the peroxidase-like activity and B-O covalency. The B-O covalency was approximately quantified by the normalized O 1s → B 3d -O 2p absorbance from O K-edge X-ray absorption spectra 21 . As shown in Supplementary Fig. 19, there is no apparent relationship between the B-O covalency and the peroxidase-like activity of the six representative perovskites. Interestingly, for the perovskites with e g occupancy close to 1 (i.e., LaMnO 3-δ , LaCoO 3-δ , and LaNiO 3-δ ), their peroxidase-like activity increases with the increasing of covalency strength of B-O. These results suggested that the B-O covalency may act as a secondary descriptor for peroxidase-like activity when the e g occupancy of B-site is close to 1 (Supplementary Note 3).
In short, in contrast to the e g occupancy, none of the four parameters discussed in this section showed a volcano relationship with the peroxidase-like activity of perovskites. These results further validated that the e g occupancy as an effective activity descriptor to predict the peroxidase-like activity of perovskites.  Table 9 and Supplementary Figs 20-22). We proposed that these perovskites mimicked peroxidases via mechanisms of   Supplementary Fig. 23), suggesting step I does not determine the overall reaction rate. The variations of absorption energies (E ads ) for O (E ads,O ) and OH (E ads,OH ) with respect to e g occupancy are shown in Fig. 3b, c, and that for H 2 O 2 (E ads; H 2 O 2 ) in Supplementary Fig. 24. Volcano-like relationships were found for E ads,O and E ads,OH with e g occupancy (Fig. 3b and c and Supplementary Fig. 24). Further analysis shows that the five perovskites with e g occupancy < 1.  Fig. 3a) is not the only ratedetermining step. Reportedly, when a kinetic profile goes through a maximum as a function of a given parameter, it means that there is a change of the rate-determining step governing the reaction mechanism 32 . To identify all the rate-determining steps and to validate the mechanisms of Fig. 3a, we further calculated the energies for species involved in the proposed reaction pathways ( Supplementary Figs 23-25). Supplementary Fig. 25 plots the energies of species involved in the proposed reaction pathways. It reveals that for the five perovskites with e g occupancy < 1.2, which are all located on the left side of Fig. 2b's volcano-like plots, the rate-determining step should be the oxidation of the substrate (i.e., IIIb and IV of Fig. 3a); for the other five with e g occupancy > 1.2, which are all located on the right side of the volcano-like plots, the rate-determining step should be the O-O bond splitting of the adsorbed H 2 O 2 * (II of Fig. 3a); LaNiO 3 is the maximum point where the rate-determining step changes. Taking these results together, e g occupancy influences the peroxidase-mimicking activities of perovskites by altering the E ads of reaction intermediates and the rate-determining step governing the catalytic reactions. Perovskites with e g occupancy of~1 possess optimal E ads and can facilitate these rate-determining steps efficiently, which further lead to the high peroxidase-like activity.
General applicability of the e g occupancy. To test whether e g occupancy could also predict the activity of non-perovskites TMOs with the same metal-oxygen octahedral coordination geometry as the perovskites described above, we investigated the peroxidase-like activity of five binary metal oxide nanoparticles (Supplementary Figs 26-28, Supplementary Table 7, and Supplementary Note 4). First, to demonstrate the predictive power of the descriptor, CoO and Mn 2 O 3-δ nanoparticles with unit e g occupancy were tested, as their peroxidase-like activities are unknown. If the e g occupancy descriptor was also applicable to the binary metal oxides, these nanoparticles would be expected to exhibit high peroxidase-like activities. As shown in Supplementary Figs 29, 30a and Fig. 4a, both nanoparticles exhibited excellent activities, in agreement with the prediction based on the e g occupancy descriptor. By contrast, the measured peroxidaselike activities of MnO 2 (e g = 0), Fe 2 O 3 (e g = 2), and NiO (e g = 2) nanoparticles were nearly negligible (Supplementary Fig. 29 and Fig. 4a), again in agreement with the prediction. These results clearly demonstrate that the peroxidase-like activity of binary metal oxides with octahedral coordination geometry is similarly associated with the e g occupancy, with a similar volcano dependence to that obtained for the perovskite TMO-based peroxidase mimics ( Supplementary Fig. 30b and Fig. 4b).
Comparison with other peroxidase mimics. Among the 16 TMOs studied in this work (Figs 2, 4 and Supplementary Fig. 10), LaNiO 3-δ was identified as the most active peroxidase mimic, in terms of both specific and mass activities (Supplementary Note 5). Over the last decade, dozens of nanomaterials have been proposed as peroxidase mimics 41 . A comparison between the nanozymes developed in this work and those reported in the literature may be useful for searching for new nanozymes. However, a direct comparison between data produced by different studies is difficult, because the applied protocols or even the specific test conditions, such as temperature and H 2 O 2 concentration, could significantly influence the peroxidase-like activity of the nanomaterials. To allow for a reliable and rigorous comparison, we synthesized several peroxidase mimics reported in previous studies (Fig. 5 Fig. 33 and Fig. 5c) shows that the mass activity of LaNiO 3-δ is 28.9 and 13.6 times higher than that of the Fe 3 O 4 nanoparticles and Cu(OH) 2 supercages, respectively. Moreover, Fig. 5d shows that the specific activity of LaNiO 3-δ was 91.4 and 49.0 times higher than that of the Fe 3 O 4 nanoparticles and Cu (OH) 2 supercages, respectively, because of the smaller surface area of the LaNiO 3-δ nanoparticles prepared by the sol-gel method. Other representative nanozymes (such as CeO 2 , CuO, single-walled carbon nanotubes, and graphene oxide (GO-COOH)) were also investigated. The results in Fig. 5c,d confirm the superior performance of LaNiO 3-δ , in terms of both specific and mass activity, further demonstrating the power of the e g occupancy descriptor for identifying nanozymes of particularly high activity.

Discussion
Using experimental measurements and DFT calculations, we have identified the e g occupancy as a predictive and effective descriptor for the peroxidase-like activity of TMO (including perovskite TMO) nanomaterials. The catalytic activity of peroxidase-like nanozymes with metal-oxygen octahedral coordination geometry shows a volcano dependence on the e g occupancy. Namely, nanozymes with e g occupancy of~1.2 had the highest catalytic activity, whereas e g occupancies of 0 or 2 corresponded to negligible activities. The systematic comparison of more than 20 representative peroxidase-like nanozymes revealed that LaNiO 3-δ had the highest catalytic activity. Besides supporting an approach to the design of highly active peroxidase mimics based on the e g occupancy, the present study also provided deep insight into the catalytic mechanism of the peroxidase-like activity of the nanozymes. Taking into account the adaptable structures and catalytic activities of TMO-based nanozymes, the current study has prompted us to further explore the application of the e g occupancy descriptor to predict the enzyme-like activities of other metal oxides.

Methods
Synthesis of perovskite TMOs. The perovskite TMOs were synthesized via a solgel method 43 . Briefly, the respective metal nitrate salts in appropriate stoichiometric ratios (3 mmol in total) and citric acid (12 mmol) were dissolved in 100 mL of H 2 O, followed by the addition of 1.5 mL of ethylene glycol. The resulting transparent solutions were treated at 90°C under stirring to condense them into gel, which were then decomposed at 180°C for 5 h to form the solid precursors. The latter were decomposed at 400°C for 2 h to remove the organic components and obtain foam precursors, which were further annealed at 700°C (850°C in the case of CaMnO 3-δ ) for 5 h with a ramp rate of 5°C min −1 , to obtain the final perovskite TMOs.
Structure characterization. PXRD data were collected at room temperature using a Rigaku Ultima diffractometer using Cu Kα radiation. The diffractometer was operated at 40 kV and 40 mA, with a scan rate of 5°min −1 and a step size of 0.02°. TEM images were recorded on a JEOL JEM-2100 or FEI Tecnai F20 microscope at an acceleration voltage of 200 kV. SEM measurements were performed on a Hitachi S-4800 microscope operated at 5 kV. UV-visible absorption spectra were collected using a spectrophotometer (TU-1900, Beijing Purkinje General Instrument Co. Ltd, China). Nitrogen adsorption-desorption isotherms were measured at 77 K using a Quantachrome Autosorb-IQ-2C-TCD-VP analyzer and were used to calculate the surface areas of the nanozymes with the BET method. The temperature-dependent magnetization was measured on a MPMS SQUID magnetometer (MPMS-3, Quantum Design) with a magnetic field of H = 1 kOe under field-cooling procedures. O K-edge X-ray absorption spectroscopy (XAS) measurements were performed at the beamline BL12B-a (CMD) in Hefei Synchrotron Radiation Facility, National Synchrotron Radiation Laboratory.  buffer solution (pH 4.5) was used as the reaction buffer and 10 μg mL −1 of nanozymes were used for their kinetics assays. The kinetics data were obtained by varying the concentration of H 2 O 2 while keeping the TMB's concentration constant (Supplementary Table 8). The kinetics constants (i.e., v max and K m ) were calculated by fitting the reaction velocity values and the substrate concentrations to the Michaelis-Menten equation as follows: where v is the initial reaction velocity and v max is maximal reaction velocity. v max is obtained under saturating substrate conditions. [S] is the substrate concentration. K m , the Michaelis constant, equals to the concentration of substrate when the initial reaction velocity reaches half of its maximal reaction rate. As for TMOs with negligible activity (i.e., LaCrO 3 , LaFeO 3 , CaMnO 3-δ , NiO, MnO 2 , and Mn 3 O 4 ), we assumed the initial reaction velocity in the presence of 10 μg mL −1 of nanozymes, 1 mM TMB, and 100 mM H 2 O 2 as the v max , because the kinetics measurements for them were difficult and not reliable. The mass activities of the nanozymes were defined as follows: The specific activities of the nanozymes were calculated from Eqs (3) and (4): Normalized BET area ¼ BET area of nanozyme BET area of LaNiO 3Àδ ð4Þ DFT calculations. The bulk structure of each defect-free perovskite was modeled using the A 8 B 8 O 24 unit cell, which was sufficiently large to consider all possible G-type antiferromagnetic (G-AFM), A-type antiferromagnetic (A-AFM), and paramagnetic (PM) magnetic orderings previously reported for perovskites ( Supplementary Fig. 20). Geometrically relaxed ground-state bulk structures were then used to build the (001) slabs, each of which contained six layers: three AO and three BO 2 ( Supplementary Fig. 21). For geometry optimizations of bulks, their space group symmetries (Supplementary Table 9) were used to constrain the geometries. For geometry optimizations using slab models, atoms in the bottom two layers (i.e., one AO and one BO 2 layer) were frozen and those in the above layers were allowed to move; lattice parameters were frozen for calculations with slab models. The generalized gradient approximation with the Perdew-Burke-Ernzerhof functional 44 was used for all geometry optimizations and energy calculations, in a planewave basis set with an energy cut-off of 500 eV and Gaussian smearing of 0.05 eV. The Hubbard U correction, where U is defined as U eff , was applied for B metals of perovskites to treat the strong on-site Coulomb interaction of their localized d electrons (Supplementary  Table 10) [45][46][47] . For calculations of bulks and slabs, the (3 × 3 × 3) and (3 × 3 × 1) Monkhorst−Pack 48 meshes were used for the k-point samplings, respectively. The convergence thresholds for the electronic structure and forces were set to be 10 −5 eV and 0.02 eV Å −1 , respectively. All calculations were performed using the VASP code 49 . More details of the computations can be found in Supplementary Note 6.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.