Introduction

Since the pioneering discovery of Fe3O4 nanoparticles as peroxidase (POD) memetics in 20071, nanozymes have garnered significant attention over the past decade2,3,4,5,6,7,8,9. Among the various research areas within nanozyme studies, investigating catalytic mechanisms is essential for the rational design and the application of nanozymes. Currently, single-atom nanozymes are widely selected for mechanism studies due to their well-defined and tunable structures10,11. Nevertheless, elucidating the mechanisms of traditional nanozymes with complex components and structures remains both challenging and crucial12,13,14,15,16,17,18.

Since our group first reported the peroxidase (POD) and catalase (CAT)-like activities of Prussian blue nanozymes (PBNZ), these nanozymes have found wide application across various fields19,20,21,22,23. However, consensus on their catalytic mechanism remains elusive. For instance, our initial work speculated that PBNZ function as POD and CAT mimetics through a mechanism involving the conversion among Prussian blue (PB, KFe(III)[Fe(II)(CN)6]), Berlin green (BG, KFe(III)3[Fe(III)(CN)6]2[Fe(II)(CN)6]) and Prussian yellow (PY, Fe(III)[Fe(III)(CN)6]), driven by matching redox potentials24. Dong’s group highlighted the importance of N-coordinated Fe units of PBNZ for POD-like activity and proposed a catalytic mechanism similar to that of natural peroxidases, involving high-valent Fe (Fe (IV)=O) intermediate25. In contrast, Arkady’s group suggested that PBNZ first react with a reductive substrate to form Prussian white (PW) and argued that the Fe (IV)=O-mediated mechanism is not applicable to PBNZ26. Additionally, several studies have attributed the POD-like activity of PBNZ to the generation of hydroxyl radical (·OH) and singlet oxygen (1O2)27,28,29,30,31,32. Despite this complexity, the overall electron flow direction of POD-like catalysis is from a reductive substrate (e.g., 2,2’-Azinobis (3-ethylbenzothiazoline −6-sulfonic Acid Ammonium Salt) (ABTS)) to H2O2. Two distinct electron transfer pathways could simultaneously exist: as a semiconductor33, PB could undergo a valence band mediated pathway (VBP) where PB initially donates an electron to H2O2 (process 1) and then accepts another electron from ABTS (process 2) (Fig. 1a). Alternatively, the catalytic reaction could proceed through a conduction band mediated pathway (CBP), where PB or their pre-oxidized state (BG, PY) first receive an electron from ABTS (process 1), followed by electron transfer to H2O2 (process 2) (Fig. 1b). From this perspective, current explanations for the catalytic mechanism of PBNZ can be attributed to these two catalytic pathways and a more explicit understanding of the mechanism is urgently needed.

Fig. 1: Two possible electron transfer pathways during the POD-like catalysis of PBNZ.
figure 1

a Electron transfer pathway I (VBP: valence band mediated pathway). b Electron transfer pathway II (CBP: conduction band mediated pathway).

Full size image

Our previous work demonstrated the depletable POD-like activity of Fe3O4 nanozymes through long-term catalysis experiments34. Long-term catalysis is advantageous for accessing the life span of nanozymes and amplifying minor physicochemical changes that may be undetectable during the short-term catalysis. We believe this strategy is also beneficial for examining the catalytic sustainability of PBNZ, detecting subtle variations during the catalytic process, and elucidating their potential mechanisms.

In this work, we report the self-increasing catalytic activity of PBNZ observed during prolonged catalysis. The previously overlooked pre-oxidation of PBNZ during catalysis is responsible for the simultaneous facilitation of both the VBP and CBP, resulting in a dual-path electron transfer mechanism mediating the catalytic process of PBNZ.

Results and discussion

Characterization of PBNZ prepared by double injection method

PBNZ used for long-term catalysis in this work was prepared using our previously reported double injection method. By adjusting the volume of the Fe3+-citric acid solution added initially (the x value) and the reaction time, we could easily control the size and crystallinity of the PBNZ35. Dynamic light scattering (DLS), transmission electron microscopy (TEM), and ultraviolet-visible (UV-vis) spectroscopy confirmed the successful synthesis of PBNZ with sizes of 81 nm and 43 nm, exhibiting low crystallinity and good dispersity (Supplementary Figs. 1 and 2).

We then focused on the surficial composition of the as-prepared PBNZ. The X-ray photoelectron spectroscopy (XPS) spectra of the 81 nm and 43 nm PBNZ are shown in Fig. 2a, b, and Fig. 2d, e respectively. The Fe2p3/2 peak of PBNZ can generally be deconvoluted into four peaks, similar to our previous work36, which are assigned to C-coordinated Fe (II) at 708.6 eV, N-coordinated Fe (II) at 710.0 eV, N-coordinated Fe (III) at 712.8 eV and Fe2+ satellite peak at 715.1 eV. Given that the PBNZ were prepared with an equivalent molar dosage of Fe3+ and [Fe(II)(CN)6]4-, the presence of N-coordinated Fe (II) and low amount of Fe (III) on the surface of the PBNZ needs further clarification. Firstly, a similar Fe2p spectrum was discovered in the PBNZ prepared by Fe2+ and [Fe(III)(CN)6]3- (Supplementary Fig. 3d). This suggests that the analogous Fe2p spectra originate from the rapid valence state redistribution through inter-metal charge transfer among C-coordinated Fe and N-coordinated Fe once the PBNZ are formed. Additionally, the Fe:N ratio of 81 nm and 43 nm PBNZ detected by EDS and XPS analyses were lower than their theoretical formula (KFe(III)[Fe(II)(CN)6] (Fe:N = 1:3) or Fe(III)4[Fe(II)(CN)6]3 (Fe:N = 1:2.6) (Fig. 2a, c, d, f)). This indicates a deficiency of Fe3+ in the bulk phase and surface of the as-prepared PBNZ due to the incomplete reaction of Fe3+ with [Fe(CN)6]4- during the ultra-fast synthesis. The insufficient use of Fe3+ was further demonstrated by analyzing the components of the filtrate solution of the unpurified PBNZ (Supplementary Fig. 4). In summary, the relatively higher usage of [Fe(CN)6]4- during the actual synthetic process led to the observed surficial coverage of excess C-coordinated Fe.

Fig. 2: Surface composition and catalytic activity of prepared PBNZ.
figure 2

a XPS survey, b Fe2p spectrum, and c EDS spectrum of 81 nm PBNZ. d XPS survey, e Fe2p spectrum, and f EDS spectrum of 43 nm PBNZ. POD-like activity measurement using g ABTS and h TMB as substrates (n = 3 independent experiments). i CAT-like activity measurement (n = 3 independent experiments). Error bars represent standard deviation (SD) from three independent measurements and all data are presented as mean values ± SD.

Full size image

The specific catalytic activity (aPOD for POD-like activity, aCAT for CAT-like activity) of 81 nm and 43 nm PBNZ was calculated. As expected, the 43 nm PBNZ outperformed the 81 nm PBNZ in the catalytic oxidation of ABTS (Fig. 2g), 3,3’,5,5’-tetramethylbenzidine (TMB) (Fig. 2h), and the self-decomposition of H2O2 (Fig. 2i). This is reflected in the higher aPOD and aCAT values, demonstrating superior POD and CAT-like activity due to their larger specific surface area. Additionally, the high catalytic stability of the prepared PBNZ was confirmed (Supplementary Fig. 5).

Fig. 3: Characterization of the PBNZ recycled from POD-like long-term catalysis.
figure 3

a Illustration of the POD-like cyclic catalysis process. Relative POD-like activity of the b 81 nm and c 43 nm PBNZ recycled from cyclic catalysis (n = 3 independent experiments). Fitting XPS Fe2p spectrum of the d 81 nm and e 43 nm PBNZ recycled from cyclic catalysis. Fitting XPS Fe2p spectrum of the f 81 nm and g 43 nm PBNZ recycled from the incubation with 0.12% H2O2. Relative POD-like activity of the PBNZ recycled from the incubation with 0.12% H2O2 using h ABTS and i TMB as chromogenic substrate (n = 3 independent experiments). Error bars represent SD from three independent measurements and all data are presented as mean values ± SD.

Full size image

POD-like long-term catalysis of PBNZ

ABTS was chosen as the chromogenic substrate, and consecutive oxidation of ABTS catalyzed by PBNZ was observed over 120 h (Supplementary Fig. 6). Initially, ABTS undergoes a single electron transfer process to form green-colored ABTS+ radicals, followed by a further electron transfer to generate colorless ABTS2+37. This result indicates that the prepared PBNZ can maintain catalytic activity for up to 120 h, providing valuable insight for the parameter settings of subsequent cyclic catalysis experiments. As outlined in Fig. 3a, the cyclic catalysis was conducted in three rounds, with each round lasting 24 h. Remarkably, the POD-like activity of both 81 nm (Fig. 3b) and 43 nm PBNZ (Fig. 3c) increased with each successive round of catalysis. This continuous enhancement is notably different from the depletable POD-like activity observed in the cyclic catalysis of Fe3O4 nanozymes in previous study34.

To investigate the underlying reason for the self-increasing POD-like activity of PBNZ, we carefully characterized the recycled PBNZ. Although the morphology of the collected PBNZ remained unchanged (Supplementary Fig. 7a, b), the hydrodynamic diameter of 81 nm and 43 nm PBNZ increased after three rounds of cyclic catalysis, indicating slight nanoparticle aggregation (Supplementary Fig. 7c, d). Additionally, we observed an increase in zeta potential and a slight decrease in absorbance in the adsorption spectra (Supplementary Fig. 7c–f). These findings suggest an irreversible oxidation process in which the surficial Fe (II) in PBNZ was oxidized to Fe (III) by H2O2 during the catalysis, leading to the decreased surface negative charge and reduced adsorption around 700 nm. XPS results further confirmed this speculation. During the cyclic catalysis, a certain amount of high-spin N-coordinated Fe (II) was oxidized by H2O2. Specifically, the surficial Fe (III) increased from 20.2% to 21.1% for 81 nm PBNZ and from 20.5% to 27.2% for 43 nm PBNZ, respectively (Fig. 3d, e). Although Dong’s group reported undetected oxidation of PBNZ treated with 20 mM H2O2 for 60 min25, our results revealed the actual existence of this irreversible oxidation process during PBNZ catalysis.

Various control experiments for cyclic catalysis were conducted. When PBNZ were dispersed in HAc-NaAc buffer with or without 0.385 mg mL−1 ABTS, slight aggregation and decreased catalytic activity were observed, while their morphology and zeta potential remained nearly unchanged (Supplementary Figs. 8 and 9). In contrast, when PBNZ were incubated with 0.12% H2O2 in HAc-NaAc buffer, there was a more significant decrease in surface negative charge and characteristic adsorption compared with the PBNZ recycled from cyclic catalysis (Supplementary Fig. 10). XPS results directly revealed further oxidation of PBNZ by H2O2, with the amount of surficial Fe (III) increasing to 24.3% and 29.2% for 81 nm (Fig. 3f) and 43 nm (Fig. 3g) PBNZ, respectively. As expected, the POD-like activity of the oxidized PBNZ showed a significant enhancement (Fig. 3h). This growing POD-like activity was also observed when using positively charged TMB as a chromogenic substrate, indicating that the increase in surficial Fe (III), rather than the decrease in surface negative charge decrease, is the main reason for the enhanced catalytic activity (Fig. 3i).

To study the influence of crystallinity on the oxidation of PBNZ, highly crystalline PBNZ (denoted as 90 nm PBNZ) were synthesized with an x value of 1.0 mL and a reaction time of 2 h. These 90 nm PBNZ exhibited similar particle size, surficial valence state, and POD-like activity but differed in crystallinity compared with the 81 nm PBNZ (Supplementary Fig. 11a–f). Notably, the increment in Fe (III) and POD-like activity of 90 nm PBNZ after cyclic catalysis and incubation with 0.12% H2O2 were lower than that of the 81 nm PBNZ (Supplementary Fig. 11g–i). This suggests that low-crystallinity PBNZ are more easily oxidized due to their larger specific surface area and stronger interaction with H2O2, resulting in a higher increase in catalytic activity. Additionally, increased Fe leaching was observed during the cyclic catalysis and H2O2 incubation of PBNZ, indicating that Fe leaching could also reflect the oxidation of PBNZ (Supplementary Fig. 12). Taken together, the irreversible oxidation of PBNZ by H2O2 is responsible for their improved POD-like activity, and low-crystallinity PBNZ are more susceptible to oxidation. Although the addition of the reductive ABTS weakens this oxidation effect during actual catalysis, the increasing amount of surficial Fe (III) during POD-like catalysis remains significant.

CAT-like long-term catalysis of PBNZ

The catalytic behavior of PBNZ shifts from POD to CAT as the pH of the solution increases from acidic to neutral conditions24. In the prolonged CAT-like catalysis, the self-decomposition of H2O2 catalyzed by PBNZ was observed in pure water using a hydrogen peroxide detection kit. Due to the weaker oxidizing ability of H2O2 at neutral environment, the final concentration of H2O2 was increased to 3.6%, compared with 0.12% used in the POD-like experiments. Pure water was chosen over phosphate-buffered saline (PBS) due to the long-term instability of PBNZ in PBS24. The detection kit uses molybdic acid, which chelates with H2O2 to form a yellow-colored product with an absorption peak at 405 nm. As shown in Supplementary Fig. 13, the absorbance at 405 nm of the reaction system continuously decreased, indicating that the PBNZ remained active as CAT mimetics for at least 24 h.

Similar to the prolonged POD-like catalysis, significant physicochemical changes were observed in the PBNZ recycled from the long-term CAT-like reactions. Aggregation was evident, as indicated by the increased hydrodynamic diameter of the PBNZ (Fig. 4a). The surface negative charge of the PBNZ decreased rapidly (Fig. 4b), and the color of the recycled PBNZ solution turned slightly green, particularly for the 43 nm PBNZ (Supplementary Fig. 14). This color change was accompanied by a decreased adsorption intensity around 700 nm and an increase in 300–500 nm range (Fig. 4c). Additionally, the surface of the 81 nm became slightly rough (Fig. 4d), indicating mild structural disruption similar to the deformation of PB or its analogs after H2O2 exposure as reported in previous38,39. XPS spectra revealed the disappearance of N-coordinated Fe (II) and a decrease of C-coordinated Fe (II). In contrast, the amount of N-coordinated Fe (III) increased and a new peak representing C-coordinated Fe (III) at 710.9 eV appeared (Fig. 4f, g). Additionally, oxygenated groups such as Fe-OH and Fe=O prominently emerged (Supplementary Fig. 15). Increased Fe leaching was also observed (Supplementary Fig. 16). These results collectively revealed the oxidative effects of H2O2 on PBNZ during CAT-like catalysis. Consequently, the CAT-like activity of the recycled PBNZ was notably enhanced (Fig. 4h). In conclusion, even in a neutral environment, the irreversible oxidation effect during CAT-like long-term catalysis is evident when H2O2 dosage is sufficiently high, contributing to the improvement of the CAT-like activity of the recycled PBNZ.

Fig. 4: Characterization of the PBNZ recycled from CAT-like long-term catalysis.
figure 4

a Hydrodynamic diameter (n = 3 independent experiments), b zeta potential (n = 3 independent experiments), c UV-vis spectra, d, e TEM images, f, g fitting XPS Fe2p spectrum, and h relative CAT-like activity of the recycled PBNZ (n = 3 independent experiments). TEM images were collected three times with similar results. Error bars represent SD from three independent measurements and all data are presented as mean values ± SD.

Full size image

Catalytic mechanism of PBNZ as POD and CAT mimetics

The active intermediates Fe-OH and Fe=O are considered crucial for the catalytic activity of PBNZ. To identify these intermediates, methyl phenyl sulfoxide (PMSO) was used as a probe. PMSO can be specifically oxidized by Fe-OH and Fe=O, with the latter process producing PMSO2 which exhibits light adsorption at 215 nm. Thus, the formation of Fe=O during catalysis can be quantified by measuring the transformation efficiency (η) of PMSO2 from PMSO40. As shown in Fig. 5a, both 81 nm and 43 nm PBNZ systems exhibited high η values at pH 3.6, indicating significant Fe=O formation. In contrast, the 90 nm PBNZ system displayed a lower η value (57.0 ± 2.7%) (Supplementary Fig. 17), highlighting the difficulty of oxidizing more crystalline 90 nm PBNZ. The PMSO2 production and η value decreased rapidly in pure water (pH 7.0), partly due to the reduced oxidizing ability of H2O2 at neutral pH, which directly lowers the production of Fe=O24. Additionally, the competitive CAT-like reaction at neutral pH consumes Fe=O to produce O2 (Supplementary Fig. 18). The incomplete transformation of PMSO2 also indicated the presence of Fe-OH during PBNZ catalysis, further confirmed by electron spin resonance spectrometer (EPR) (Supplementary Fig. 19a). The contribution of 1O2 to PBNZ-mediated catalysis was excluded as no signal was detected by EPR (Supplementary Fig. 19b). In summary, both Fe-OH and Fe=O intermediates are involved in the catalysis of PBNZ, as demonstrated through various experimental analyses.

Fig. 5: Dual-path electron transfer mechanism of PBNZ.
figure 5

a PMSO and PMSO2 detection by HPLC (n = 3 independent experiments, error bars represent SD from three independent measurements and data are presented as mean values ± SD). b Energy levels of catalytic substrates and conduction/valence band of PBNZ (VB valence band, CB conduction band). c Mechanism and energy profiles (energy in eV) mediated by VBP. d Mechanism and energy profiles (energy in eV) mediated by CBP.

Full size image

We focused on identifying the electron transfer pathway during the POD and CAT-like catalysis of PBNZ. Similar to our previous work predicting superoxide-dismutase activity through energy level principle41, the electron transfer pathway of PBNZ could also be initially estimated from the perspective of energy level. As a semiconductor, the N-coordinated Fe and C-coordinated Fe in PB contribute to their conduction band and valence band, respectively33. The optical band gaps of the 81 nm and 43 nm PBNZ were obtained from their adsorption spectra (Supplementary Fig. 20). Their energy levels relative to the standard hydrogen electrode (SHE) were then calculated and compared with those of ABTS and H2O2 (Supplementary Table 1). The conduction band energy level (ECB) of PBNZ lies between the energy levels of H2O2/H2O and ABTS/ABTSox, while their valence band energy level (EVB) is below the energy level of H2O2/H2O. Thus, the CBP mediated by the conduction band of PBNZ is estimated to be thermodynamically preferred during the POD and CAT-like catalysis compared with the VBP. The catalysis begins with the electron injection from ABTS or H2O2 into the PBNZ conduction band, forming ABTSox or O2. Subsequently, electron transfer from the PBNZ conduction band to H2O2 generates H2O (Fig. 5b). Furthermore, the oxidation effect of H2O2 narrowed the band gap (Eg) of PBNZ with an increase of EVB (Supplementary Fig. 20, Fig. 5b). Notably, the oxidation primarily occurs on the surface of the particles, whereas the optical method measured the Eg in the bulk phase. Thus, although the calculated EVB of the oxidized PBNZ remains below the H2O2/H2O energy level, the actual increase in the surficial EVB is likely higher and may elevate the valence band energy level above H2O2/H2O, facilitating the occurrence of VBP.

Density functional theory (DFT) calculations were conducted to elucidate the atomistic-level mechanism underlying the POD-like activity of PBNZ and the role of oxidation in enhancing their activity. Based on the experimental results, Fe-OH and Fe=O are likely the key intermediates in the catalysis. Therefore, three PBNZ structures were evaluated: pristine PBNZ (A1 of Fig. 5c), PBNZ modified with -OH (B1), and PBNZ with =O (C1 and D1). For any given PBNZ structure, the catalysis may be initiated by reacting with H2O2 (via the VBP) or ABTS (via the CBP). The calculations revealed that the pristine PBNZ (A1) exhibited minimal POD-like activity. Instead, it could be easily oxidized by H2O2 to form a thermally stable intermediate (A3), which had a deep potential energy well (−3.25 eV) and was difficult to be reduced back to A1 by ABTS (Fig. 5c). Further calculations indicated that the two-OH structure A3 could dehydrate to form a one-O structure (D1), with an energy change of 0.7 eV. As more OH groups are present on PBNZ, the dehydration becomes more energetically favorable. The dehydration of three-OH structure B3 to one-OH-and-one-O structure C2 was energetically favorable with an energy change of −0.11 eV. Thus, the concentration of Fe=O would increase with the extent of oxidation. These findings align with experimental observations showing that unoxidized PBNZ had lower POD-like activity and that Fe-OH and Fe=O groups were prevalent in the PBNZ structures due to irreversible oxidation.

In contrast, the oxidized PBNZ structure B1 had a shallow potential energy well and should have considerable POD-like activity. As illustrated in Fig. 5c, the one-OH structure B1, representing the moderately oxidized PBNZ, could catalyze the reaction between H2O2 and ABTS via the VBP. Initially, an H2O2 molecule dissociated on B1 to form B3, resulting in an energy decrease of −0.76 eV. B3 was then reduced back to B1 by two ABTS molecules, completing the catalytic cycle. The two-O structure C1, representing the heavily oxidized PBNZ, could catalyze the POD-like reaction via the CPB. C1 was first reduced by two ABTS to form one-O structure C3, with an energy decrease of −0.25 eV. C3 was then oxidized back to C1 by H2O2 (Fig. 5d). One-O structure D1, however, could not undergo the POD-like catalysis via the CBP due to the high energy of the intermediate D3 (2.6 eV). Similarly, the two-OH structure E1 as shown in Supplementary Fig. 23 could catalyze the reaction via the VBP, the three-OH structure B3 could catalyze the reaction via the CBP (B3-B4-B1-B2-B3), and the one-O structure could catalyze the reaction via the VBP (C3-C4-C1-C2-C3). Notably, the Fe=O is present in both the PBNZ and horseradish peroxidase (HRP)-mediated POD-like processes, involving two sequential one-electron transfers from the chromogenic substrates, which to some extent indicates similarity. However, the specific transformation of Fe=O during their catalysis differs (PBNZ: Fe=O → Fe-OH → Fe; HRP:.Fe=O+ → Fe=O → Fe42). Moreover, since the CAT-like process can be viewed as a special POD-like process where ABTS is replaced by reductive H2O2 under neutral pH, the above conclusions are also applicable to explain the enhancement of CAT-like activity in oxidized PBNZ. Taken together, unoxidized PBNZ (e.g., A1) exhibits little catalytic activity. However, once dispersed in the catalytic solution and irreversibly pre-oxidized by H2O2, moderately and heavily oxidized PBNZ show significant POD and CAT-like activity via the VBP and CBP, respectively.

In summary, the catalytic mechanism of PBNZ has been elucidated through the experimental and theoretical analyses on the PBNZ recycled from long-term catalysis. Our results demonstrate the unique advantage of PBNZ, which exhibits self-enhancing POD and CAT-like activities, compared with natural enzymes and iron-based nanozymes that deplete over time. The pre-oxidation effect of H2O2 leads to an increase in the valence state of surface Fe, accompanied by the formation of oxygenated groups, which is crucial for enhancing both VBP and CBP during catalysis. The proposed dual-path electron transfer mechanism thoroughly explains the catalytic behavior of PBNZ, ensuring a long service life.

Methods

Chemicals

All chemicals were of analytical grade and were used without further purification. Iron (III) chloride hexahydrate (FeCl3·6H2O), Citric acid monohydrate (C6H8O7·H2O), Sodium acetate (C2H3NaO2), Acetic acid (CH3COOH, 98.0%), Hydrogen peroxide (H2O2, 30%), Dimethyl sulfoxide (DMSO), Acetonitrile were all obtained from Sinopharm Chemical Reagent Co., Ltd. Potassium hexacyanoferrate (II) trihydrate (K4[Fe(CN)6]·3H2O) were purchased from Shanghai Lingfeng Chemical reagent Co., Ltd. TMB and ABTS were bought from Aladdin (Shanghai, China). PMSO and PMSO2 were purchased from Shanghai Yuanye Bio-Technology Co., Ltd. 2,2,6,6-tetramethyl-4-piperidine (TMP) and 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) were bought from Sigma Aldrich. Hydrogen peroxide detection kit were obtained from Jiangsu KeyGEN BioTECH Co., Ltd. Deionized water was used in all experiments.

Characterization

Particle size, morphology, crystallinity, and element component analysis of PBNZ were conducted using TEM (JEM-2100, JEOL, Japan) equipped with EDS and Digital Micrograph software (v3.2). Hydrodynamic diameter and zeta potential were measured by DLS (Nano-ZS90, Malvern, England) coupled with Malvern Zetasizer software (v7.12). UV-vis absorption spectra were pictured on a UV-VIS-NIR spectrophotometer (UV-3600, Shimadzu, Japan) coupled with UV Probe software (v2.42). Fe element concentration in filtrate solution was determined by an inductively coupled plasma spectrometer (ICP)-mass spectrum (MS) device (ICP-MS 7800, Agilent, USA) coupled with ICP Expert II software. The valence state of elements on the PBNZ surface was measured by XPS (Escalab 250Xi, Thermo, USA) coupled with Thermo Advantage software (v5.967).

Synthesis of PBNZ by double injection method

PBNZ with different sizes and crystallinity were prepared by our previously reported double injection method35. In a general procedure, 1 mmol FeCl3 and 3 mmol C6H8O7 were added into 20 mL deionized water under magnetic stirring for 30 min to form Fe3+-citric acid solution. 1 mmol K4[Fe(CN)6] was added into 20 mL deionized water under magnetic stirring for 10 min. Prior to the double injection, x mL Fe3+-citric acid was added into the reaction flask containing 60 mL deionized water under vigorous stirring at 60 °C. Subsequently, (20 - x) mL Fe3+-citric acid and 20 mL K4[Fe(CN)6] solution were simultaneously injected into the flask at a flow rate of 40 mL h−1. The reaction solution was continuously stirred with a certain reaction time after the injection. The x value was set as 1.0 mL and 20.0 mL to obtain 81 nm and 43 nm PBNZ with low crystallinity, respectively. The reaction time was set as 25 min and 2 h for 81 nm and 43 nm PBNZ, respectively.

Measurement of the specific catalytic activity of PBNZ as POD mimetics

Typically, 200 μL HAc-NaAc buffer (0.2 M, pH = 3.6), 20 μL PBNZ, 10 μL 10 mg mL−1 chromogenic substrate (ABTS, TMB) and 30 μL 1% H2O2 were sequentially added into the reaction system at room temperature. The absorbance change of the reaction system which was monitored in 60 s by a Microplate Reader (Model 680, BIO-RAD, USA) (coupled with Tecan i-control software (v1.6.19.2)) displayed the POD-like activity of PBNZ. When ABTS was selected as substrate, the concentration of PBNZ was set as 60 μg mL−1, absorbance change of solution was observed at 415 nm; for TMB, the concentration of PBNZ was diluted to 6 μg mL−1, absorbance change of the solution was observed at 650 nm.

The aPOD of PBNZ as POD mimetics was defined following Eq. (1) based on our previous work36:

$${a}_{{{{{{\rm{POD}}}}}}}=[V(\varDelta A/\varDelta t)]/({\varepsilon {lm}}_{{{{{{\rm{Fe}}}}}}})$$
(1)

where aPOD is the specific catalytic activity of PBNZ (U mg−1); V is the total volume of the reaction solution (μL); ΔA/Δt is the initial change rate of reaction solution absorbance after correcting with reagent blank rate (min−1); ε is the molar absorption coefficient of chromogenic substrate derivative (ABTS: 36000 mol−1 L cm−1; TMB: 39000 mol−1 L cm−1); l is the optical path of the cuvette (cm); mFe is the total Fe element mass contained in added PBNZ (mg).

Measurement of the specific catalytic activity of PBNZ as CAT mimetics

The CAT-like activity of PBNZ was measured in pure water. Typically, 5 mL pure water and 50 μL 80 μg mL−1 PBNZ were mixed in the measurement system at room temperature. 167 μL 30% H2O2 was added into the reaction system and the concentration change of dissolved oxygen (DO) in 5 min which was monitored by a dissolved oxygen meter (JPSJ-605F, Lei ci, China), showing the CAT-like activity of PBNZ. Measurement of reagent blank rate: all the procedures are the same as described above except that PBNZ was replaced by deionized water.

The aCAT of PBNZ as CAT mimetics was defined following Eq. (2):

$${a}_{{{{{{\rm{CAT}}}}}}}=[2V(\varDelta D/\varDelta t)]/{m}_{{{{{{\rm{Fe}}}}}}}$$
(2)

where aCAT is the specific catalytic activity of PBNZ (U mg−1); V is the total volume of the reaction solution (mL); ΔD/Δt is the initial change rate of the dissolved oxygen in the reaction solution after correcting with reagent blank rate (mmol L−1 min−1); mFe is the total Fe element mass contained in added PBNZ (mg).

Long-term catalysis of PBNZ

For POD-like activity, 3.2 L HAc-NaAc buffer, 320 mL 60 μg mL−1 PBNZ, 160 mL 10 mg mL−1 ABTS and 480 mL 1% H2O2 solution were mixed in the reaction system (final concentration of reagents: PBNZ: 4.62 μg mL−1; ABTS: 0.385 mg mL−1; H2O2: 0.12%). The catalytic reaction lasted for 24 h at room temperature under magnetic stirring. After the reaction, PBNZ was separated and washed with deionized water several times by ultrafiltration following with the reuse of the as-collected PBNZ for a new round of catalysis. Ultimately, three rounds of the above cyclic catalysis were conducted. PBNZ from different cyclic days were recycled for further characterization.

For CAT-like activity, 2 L pure water, 200 mL 60 μg mL−1 PBNZ, and 300 mL 30% H2O2 solution were mixed in the reaction system (final concentration of reagents: PBNZ: 4.80 μg mL−1; H2O2: 3.60%). The catalytic reaction lasted for 24 h at room temperature under magnetic stirring. After the reaction, PBNZ was separated and washed with deionized water several times by ultrafiltration. PBNZ were recycled for further characterization.

Free radical detection by EPR

For the detection of·OH, DMPO was used as the spin-trapping agent to trap ·OH. After the addition of PBNZ, H2O2 and DMPO into the HAc-NaAc buffer, solutions were monitored by EPR (EMX PLUS, Bruker, Germany) coupled with Bruker Xenon software (v1.2). The concentration of PBNZ, H2O2 and DMPO were 4.62 μg mL−1, 0.12% and 50 mM, respectively.

For the detection of 1O2, DMPO was substituted by TMP as the spin-trapping agent. The concentration of reagents is the same as ·OH detection experiment.

High valent Fe detection by HPLC

3 mL HAc-NaAc buffer or deionized water, 300 μL 60 μg mL−1 PBNZ, 150 μL 10 mM PMSO and 450 μL 1% H2O2 were mixed to evaluate the producibility of Fe in high valence when PBNZ were mixed with H2O2. After 24 h incubation, PBNZ were removed by ultrafiltration. Afterward, PMSO and PMSO2 in the filtrate were examined by HPLC (HPLC e2695, Waters, USA) coupled with Empower 3 software according to the reported article40. Briefly, solutions were separated by Waters C18 column (250 × 4.6 mm, 5 μm particle size, XBridge). The mobile phase consisted of 30% acetonitrile and 70% water with 0.1% acetic acid at a flow rate of 1.0 mL min−1. PMSO and PMSO2 were detected at wavelengths of 230 nm and 215 nm, respectively. The transformation efficiency η was defined following Eq. (3):

$$\eta={\varDelta C}_{{{{{{\rm{PMSO}}}}}}_{2}}/{\varDelta C}_{{{{{\rm{PMSO}}}}}}$$
(3)

where ΔCPMSO and \({\varDelta C}_{{{{{{\rm{PMSO}}}}}}_{2}}\) are the consumption amount of PMSO and formation amount of PMSO2, respectively.

Energy level calculation of PBNZ versus SHE

ECB and EVB of PBNZ versus SHE were calculated using the following empirical formula (Eqs. (4) and (5))43:

$${E}_{{{{{{\rm{CB}}}}}}}={\rm X}-4.5{-}0.5{E}_{{{{{{\rm{g}}}}}}}$$
(4)
$${E}_{{{{{{\rm{VB}}}}}}}={\rm X}-4.5+0.5{E}_{{{{{{\rm{g}}}}}}}$$
(5)

where ECB and EVB are the conduction band energy level and valence band energy level of PBNZ respectively (eV). Χ represents the electronegativity of PBNZ (Fe4[Fe(CN)6]3) (Eq. (6)):

$${\rm X}={[{\chi ({{{{{\rm{Fe}}}}}})}^{7}{\chi ({{{{{\rm{C}}}}}})}^{18}{\chi ({{{{{\rm{N}}}}}})}^{18}]}^{1/43}$$
(6)

where χ which reflects the electronegativity of atoms for Fe, C and N are 4.03, 6.26 and 7.30, respectively.

Eg was calculated according to Eq. (7):

$${(\alpha {hv})}^{n}=B({hv}-{E}_{{{{{{\rm{g}}}}}}})$$
(7)

in which Eg, α, hν and B are optical band gap (eV), absorption coefficient, photon energy (J) and proportionality constant, respectively. n represents the transition categories of semiconductor (n = 1/2 for indirect transition; n = 2 for direct transition). Given that PB is considered as indirect transition semiconductor44 and α is linear with absorbance (A), variation of (Ahv)0.5 versus hv were plotted. The linear range of these plots was extrapolated to the x-axis (y = 0) to obtain the value of Eg.

Structural models

According to the experimental cell parameters of PBNZ, an ideal structural model of PBNZ was established. The molecular formula of the model was K4Fe8(CN)24 and it had F43m symmetry. Firstly, the cell parameters were optimized and the magnetic moments of Fe atoms were calculated. The results showed that the N- and C- coordinated Fe were attributed to Fe3+ and Fe2+, respectively. Then, the PBNZ (001) slab was cut from the cell structure for the calculation of the subsequent catalytic reaction. In order to save computing costs, the bottom four atomic layers were fixed during the optimized process. More details of the calculations can be found in the supplementary computational details (Supplementary Figs. 21 and 22).

Computational methods

The spin polarized density functional theory calculations were applied to investigate reaction process. The geometry optimization of all structures was performed using the projector augmented wave (PAW) method45 implemented in the Vienna Ab initio Simulation Package (VASP, version 5.4.4)46. The cutoff energy of plane-wave was set to 400 eV. The Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional with the generalized gradient approach (GGA) was applied. The Ueff value was set to 6.0 because the band gap calculated using the Ueff value of 6.0 was close to the experimental band gap47. All slab structures have a vacuum value of 15 Å in z direction to avoid interaction between adjacent cells. The electronic and geometry optimization convergence criteria were set to 10−5 eV and 0.02 eV Å−1, respectively. Monkhorst-Pack k point mesh was set to 5 × 5 × 5 when optimizing the cell parameters of K4Fe8(CN)24. However, Monkhorst-Pack k point mesh was set to 3 × 3 × 1 for calculating other slab structures. All small molecules, such as H2O2, H2O, ·OH, ABTS and ABTS-H were placed in a cube box of 15 × 15 × 15 Å, optimizing geometry by using k point of 1 × 1 × 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.