Revealing energy storage mechanism of CsPbBr₃ perovskite for ultra-stable symmetric supercapacitors

Synthesis of the CsPbBr₃ perovskite

CsPbBr₃ nanocrystals were synthesized by a hot injection method, as reported by us previously^[26-28]. Briefly, a precursor solution composed of 203 mg caesium carbonate (Cs₂CO₃), 10 mL 1-octadecene (ODE) (95%, Sigma-Aldrich), and 1.5 mL oleic acid (OA) (90%, Sigma-Aldrich) was heated in a round bottom flask under vacuum at 120 °C for 1 h. The temperature was then increased to 160 °C and the solution was heated for 20 min in a nitrogen atmosphere until the solid was completely dissolved to form a Cs-oleate solution that is kept at 120 °C for further use. 207 mg lead bromide (PbBr₂), 15 mL ODE, 1.5 mL oleic acid, and 1.5 mL oleylamine (OAM) were mixed in another round bottom flask to form the main solution. The main solution was firstly heated at 120 °C for 1 h under vacuum, and then at 160 °C in a nitrogen atmosphere for 20 min until the solid dissolved completely. After that, 2 mL of the precursor (Cs-oleate) solution was injected into the main solution at 160 °C, which was immediately transferred to a round bottom flask in an ice bath to decrease the solution temperature to room temperature. The color of the solution turned light yellow, suggesting the formation of CsPbBr₃ nanocrystals. 15 mL methyl acetate was added to the solution as an antisolvent to promote precipitation of the CsPbBr₃ nanocrystals in the solution. The solution was centrifuged at 12,000 rpm for 10 min. The supernatant was decanted, and the precipitate at the bottom of the centrifuge tube was redispersed in hexane to remove the residual ODE, unreacted reactants, and excessive organic ligand (oleic acid and oleylamine), which can make the CsPbBr₃ more stable. 15 mL methyl acetate was added to facilitate the precipitation of CsPbBr₃ nanocrystals again. After centrifugation at 8,000 rpm for 5 min, the CsPbBr₃ nanocrystals at the bottom of the centrifuge tube were collected and dried under vacuum for 12 h at room temperature to completely remove residual solvent.

Characterization of CsPbBr₃

The morphology and structure of the as-synthesized CsPbBr₃ nanocrystal powder was measured with scanning electron microscopy (SEM, JEOL 7001F) and transmission electron microscopy (TEM, JEOL 2100). The composition of the perovskite was measured by energy dispersive X-Ray spectroscopy (EDS). The voltage used in the SEM and EDS measurements were 5 kV and 15 kV, respectively, to avoid damaging the perovskite material. The optical properties of the CsPbBr₃ nanocrystals were measured by a UV-VIS spectrophotometer (CARY 60 Romulus) and photoluminescence (PL) spectrophotometer (Cary Eclipse Argus) by dispersion of the as-synthesized material in hexane. X-ray diffraction (XRD, CuKα) patterns and X-ray photoelectron spectroscopy (XPS) measurements of thin films of CsPbBr₃ nanocrystals. The films were made by spin coating a solution containing 0.58 mg of the as-synthesised CsPbBr₃ in 10 mL hexane on the surface of a FTO substrate. Inductively coupled plasma mass spectrometry (ICP-MS) was used to quantitatively analyze the composition of the electrolyte solution after the electrochemical cycling test, and the standard deviation and residual standard deviation of the results are shown in Supplementary Table 2.

Preparation of a CsPbBr₃ electrode and a symmetric supercapacitor

A CsPbBr₃ electrode was made by spin coating the dispersion of CsPbBr₃ nanocrystals in hexane (5.8 mg CsPbBr₃ nanocrystals in 10 mL hexane) on a cleaned FTO substrate. The FTO glass substrate was washed with Decon 90 solvent under sonication for 20 min, followed by a mixed solvent of acetone, ethanol, and isopropanol at a ratio of 1:1:1 under sonication for another 20 min. After this, the FTO substrate was treated under UV light in an ozone atmosphere for 30 min to remove organic residuals and increase the surface wettability. The process for spin coating was performed at 1,000 rpm s^-1 for 10 s with an acceleration rate of 100 rpm s^-1 with a spin coater (CY-SPC8-HAR, Zhengzhou CY Scientific instrument Co., Ltd). After this, the spinning speed was reduced at a rate of 100 rpm s^-1 for 10 s. The substrate with the sample was dried in a vacuum oven for 12 h. This process was repeated several times until a uniform film with a thickness of around 700 nm was obtained. The excess CsPbBr₃ was wiped off with a cotton swab moistened with DMF, leaving the film with a size of 1.0 × 2.0 cm² [Supplementary Figure 1]. The mass loading (around 1 mg) of the activated material was determined by weighing the difference between the unmodified FTO and the FTO with the CsPbBr₃ film. A symmetric CsPbBr₃ supercapacitor was assembled by immersing the two CsPbBr₃ electrodes with similar mass loading of activated materials in an electrolyte solution composed of TBAPF₆ at different concentrations in DCM. In comparison, we also tested the stability of the perovskite electrode in a 1 M KOH aqueous solution.

Electrochemical characterization

The electrochemical properties of the CsPbBr₃ perovskite material were measured using a three-electrode configuration composed of a CsPbBr₃-based working electrode, a saturated calomel reference electrode, and a platinum plate counter electrode. The electrolyte solution was composed of TBAPF₆ in DCM solvent with different concentrations. The Cyclic voltammetry (CV) at different scan rates (2 mV s^-1, 5 mV s^-1, 10 mV s^-1, 20 mV s^-1, and 50 mV s^-1) in a potential window of 0.0 to 1.0 V, galvanostatic charge-discharge (GCD) at different current densities (40-100 mA g^-1) and impedance spectrum of the perovskite electrode were recorded with an electrochemical workstation (BioLogic) in air. The electrochemical impedance spectroscopy (EIS) measurements were performed over a frequency range from 0.1 Hz to 10 kHz. The EIS data were fitted by an equivalent circuit via Zview software.

The specific mass capacitance of the electrode was calculated according to the GCD plots specifically by the equation of $$C_{m}=\frac{i \times t}{\Delta V}$$ where i is the current density (A g^-1), t is the discharge time (s) and ΔV is the potential window. The specific areal capacitance is given by $$C_{a}=\frac{C_{m} \times m}{a}$$, where a is the area of the active electrode material in contact with the electrolyte. The specific areal capacitance is based on the area of the electrode active material rather than the normalized areal capacitance. The energy density (E) and power density (P) were calculated using the expressions $$ E=\frac{0.5 \times C_{m} \times V \times 1,000}{3,600} = \frac{C_{m} \times V}{7.2} $$ and $$ P=\frac{E \times 3,600}{t} $$, respectively^[13].

Effect of electrolyte solution on the stability of CsPbBr₃ perovskite

KOH aqueous solution is one of the most used electrolytes in supercapacitors. However, an aqueous electrolyte is not suitable for metal halide perovskite electrode materials because exposure to water can cause material degradation and ultimately destroy the CsPbBr₃ structure^[29]. To confirm this, we investigated the stability of the CsPbBr₃ electrode in a 1 M KOH aqueous solution. We found that the yellow color of the perovskite film quickly disappears after contact with the 1 M KOH electrolyte within 3 s [Supplementary Figure 2A and B]. These results confirm that CsPbBr₃ is highly unstable in KOH aqueous solution. We also noticed that in a similar report investigating CsPbBr₃ for supercapacitor applications by Thakur et al. that Ni foam was used as the substrate for CsPbBr₃ materials which reported a specific capacitance of 121 F g^-1[19]. However, both our experimental results [Supplementary Figure 2C] and previous literature^[20,21,30] have shown that Ni metal is not a suitable substrate to assess electrode materials in supercapacitors applications as it can cause water splitting when the voltage is above 0.4 V (vs. saturated calomel reference electrode), resulting in inaccuracy or over-estimation of the electrochemical properties of the electrode.

In order to address the moisture instability of CsPbBr₃, we used a non-polar organic solution composed of TBAPF₆ in low polarity DCM for the electrochemical characterization of CsPbBr₃^[9]. The large size of both the cation and anion in TBAPF₆ benefits the structural stability of CsPbBr₃ in solution as they are too bulky to intercalate into the CsPbBr₃ crystal structure^[9]. In addition, the much lower polarity of DCM (polarity index = 3.1) compared to water (polarity index = 10.2) is expected to hinder the decomposition of the perovskite material. Indeed, we found that the CsPbBr₃ film was almost completely intact after 2,000 cycles of electrochemical charge/discharge based on the fluorescence image of the film under a UV light exposure [Supplementary Figure 2D]. These results gave us confidence that TBAPF₆ in DCM is a compatible electrolyte for studying the electrochemical properties of CsPbBr₃. This is the first time the electrolyte was used in a supercapacitor rather than in other applications.

Electrochemical performance characterization

The CV behaviour of a CsPbBr₃ electrode in a three-electrode configuration in an electrolyte with different concentrations of TBAPF₆ in DCM is shown in Figure 1. Figure 1A compares the CV curves of a bare FTO electrode and CsPbBr₃ coated FTO. The bare FTO electrode showed negligible capacitance and can be regarded as only a current collector. In contrast, a nearly rectangular-shaped CV curve was obtained with the CsPbBr₃ electrode over a wide potential window of 0-0.8 V [vs. standard calomel electrode (SCE)] [Figure 1A]. The complete absence of any redox peaks from a faradaic process indicates that the perovskite material operates through an electric double-layer energy storage mechanism that relies on electrolyte ion adsorption/desorption on the electrode surface [Supplementary Figure 3]^[19,31]. We further investigated the ionic conductivity of the electrolyte as a function of concentration of TBAPF₆ in DCM. The results show that the highest conductivity was obtained with 1.5 M TBAPF₆ [Figure 1B]. However, such a high concentration of TBAPF₆ can result in precipitation of TBAPF₆ on the surface of the electrode due to oversaturation of the salt in DCM. The CV curves of the CsPbBr₃ electrode in a three-electrode system with different concentrations of TBAPF₆ are shown in Figure 1C. The capacitance of the CsPbBr₃ electrode increased significantly when the TBAPF₆ concentration increased from 0.1 M to 1 M. Beyond this, the capacitance of the CsPbBr₃ electrode does not change. The GCD curves [Figure 1D] also confirm the superior properties of the 1 M electrolyte compared to 0.1 M TBAPF₆. The 1 M electrolyte exhibits a much longer discharge time at both high (100 mA g^-1) and low (40 mA g^-1) current densities, which implies a higher specific capacitance. Therefore, 1 M TBAPF₆ electrolyte was used in the following experiments because of its high conductivity and lack of precipitation from solution.

Figure 1. (A) Comparison of CV curves of CsPbBr₃/FTO electrode and bare FTO in 0.1 M TBAPF₆ at a scan rate of 50 mV s^-1. (B) dependence of ionic conductivities of the electrolyte with different concentrations of TBAPF₆ in DCM. (C) CV curves of CsPbBr₃ electrode in electrolyte with different concentrations of TBAPF₆ in DCM at a scan rate of 50 mV s^-1. (D) domparison of GCD curves of CsPbBr₃ electrode in 0.1 M and 1 M TBAPF₆ in DCM at the current density of 0.04 A g^-1 and 0.1 A g^-1, respectively.

Nyquist plots of the CsPbBr₃ electrode in a three-electrode setup employing 1 M TBAPF₆ electrolyte over a frequency range from 0.1 Hz to 100 KHz are shown in Figure 2A and B. A nearly vertical line with a slope of 84° indicates that the CsPbBr₃ nanocrystals possess an electric double-layer capacitor (EDLC) capacitive behaviour, which is determined by the diffusion of electrolyte ions between the electrode surface and the electrolyte [Figure 2A]^[32,33]. Fitting the high-frequency region of the Nyquist plot by using the equivalent circuit shown in the inset of Figure 2A clearly shows the resistance from each part of the system [Figure 2B]. The 1 M TBAPF₆ electrolyte had a charge transfer resistance (Rct) of 0.5 Ω and an internal resistance (R_i) of 146 Ω. The Warburg impendence is 165 Ω and the CPE (constant phase element) is 1 Ssⁿ cm^-2. The series resistance Rs comes from each part of the assembled device and the contacts between the parts, which mainly include the CsPbBr₃ electrode, non-dense coating, the organic electrolyte, the FTO glass and the interfacial resistance resulting from the contact between the electrode and electrolyte^[34,35]. Figure 2C shows the CV curves of the CsPbBr₃ electrode in 1 M TBAPF₆ electrolyte over a potential window of 1 V recorded at different scan rates of 2 mV s^-1, 5 mV s^-1, 10 mV s^-1, 20 mV s^-1, and 50 mV s^-1, respectively. The near rectangle shape of the CV plots is consistent with the typical shape of EDLC behavior. In addition, there are no redox peaks observed in Figure 2C, indicating that the electrochemical process in the CsPbBr₃ nanocrystals does not involve a Faradaic charge transfer process^[13,14]. The GCD plots of the CsPbBr₃ electrode in 1 M TBAPF₆ electrolyte with the current density increasing from 40 to 100 mA g^-1 over a voltage window of 1 V were also measured [Figure 2D]. With the increasing current density, the shape of the curves remains triangular without a discharge plateau, indicating excellent reversible EDLC behaviour. It is noticed that the coulombic efficiency at low current density (40 mA g^-1) is poor, which could be caused by the slow ion absorption/desorption and faster degradation of the perovskite^[36]. Based on the GCD plots, the mass specific capacitance of the CsPbBr₃ electrode was calculated to be 528 mF g^-1 (264 mF cm^-2) at a current density of 100 mA g^-1 [Figure 2D]. These tests were repeated in N₂ atmosphere and gave similar results [Supplementary Figure 4].

Figure 2. Nyquist plot of the CsPbBr₃ electrode in 1 M TBAPF₆ electrolyte in (A) Low and (B) high-frequency region. The insert is the enlarged curve at the high-frequency region and the equivalent electric circuit. (C) the CV curves of CsPbBr₃ electrode in 1 M TBAPF₆ at different scan rates. (D) the charge and discharge curves of CsPbBr₃ electrode in 1 M TBAPF₆ at different current densities. All the tests were performed in a three-electrode system in air.

To further confirm the energy storage mechanism of the perovskite electrode, a relationship based on a power law between the current (i) and the scan rate (ν) was investigated according to log i(V) = blogv + loga^[37], where i is the current, ν is the scan rate. We extracted the b value by fitting the plot of log i vs. log ν [Figure 3A]. The value of b is 0.89, which is close to 1, indicating capacitive behaviour. To confirm the contribution of diffusion in the charge storage process, we fitted the plot of current density vs. potential with i(V) = K₁v + K₂v^1/2[37], where i is the current, V is the voltage and ν is the scan rate [Figure 3B]. The result shows that 99.33% of the capacitance is due to an EDLC process. Furthermore, measurement of the high-resolution XPS for Cs, Pb and Br on the surface of the perovskite electrode before and after the electrochemical test shows no change in the XPS peaks after 2,000 cycles [Supplementary Figure 5]. This indicates that there is no change in the chemical state of the elements in the perovskite. Jiang et al. proved that only small ions like Li⁺ and Na⁺ can intercalate into the perovskite structure to form Pb metal which is called an electrochromic phenomenon^[38]; however, large ions like K⁺ and Cs⁺ exceed the intercalation limit of perovskite^[38], and are hence excluded from entering the crystal structure. Therefore, the TBAPF₆ electrolyte cannot provide the required conditions to form Pb metal which would degrade the structure.

Figure 3. (A) Demonstration of the Trasatti method. (B) the simulation calculation of CV curve CsPbBr₃ electrode in three-electrode system in 0.1 M TBAPF₆ at a scan rate of 50 mV s^-1 (Dunn method).

Symmetrical supercapacitors (SCs) composed of two identical CsPbBr₃ electrodes with similar active material loading were fabricated to assess the energy and power density of the device as well as cycling stability. To demonstrate the importance of the TBAPF₆ concentration on the device performance, we compared supercapacitors using 0.1 M and 1 M TBAPF₆ in DCM electrolytes. The SC using 0.1 M TBAPF₆ is named SS-0.1, while the one using 1 M TBAPF₆ in DCM is named SS-1, respectively. The CV curves of SS-1 [Figure 4A] at different scan rates over a potential window of 0-1.0 V are even closer to a rectangle than the CV curves in a three-electrode system, further confirming the EDLC behaviour of the CsPbBr₃ nanocrystals^[10,13]. The shape of the CV is well maintained at a high scan rate of 50 mV s^-1, indicating good rate performance. Figure 4B shows the charge and discharge curves of the SS-1 at different current densities over a current density range from 40 to 100 mA g^-1. It was found that the shape of the curves remained triangular with no noticeable voltage drop upon an increase in the current density. The calculated specific capacitances of SS-1 and SS-0.1 are 119.8 mF g^-1 (60.88 mF cm^-2) and 30.5 mF g^-1 (17.75 mF cm^-2), respectively, at a current density of 100 mA g^-1. Figure 4C compares the specific capacitances of SS-0.1 and SS-1 at different current densities, where it is worth noting that the specific capacitance of SS-1 is nearly 4 times higher than SS-0.1. Both devices demonstrate more than 90% capacitance retention with an increase of current density from 40 mA g^-1 to 100 mA g^-1, which indicates the excellent rate performance of the perovskite.

Figure 4. (A) The CV curves of SS-1 under different scan rates. (B) the charge and discharge curves SS-1 at different current densities. (C) the specific capacitance of SS-0.1 and SS-1 at different current densities. (D) cycling stability of SS-0.1 and SS-1 after 2,000 and 10,000 cycles at a current density of 100 mA g^-1. (E) energy density and power density of SS-0.1 and SS-1 at different current densities. (F) relationship between energy density and power density of SS-0.1 and SS-1 (Ragone plot).

Figure 4D shows the cycle stability of SS-0.1 and SS-1. It was found that the specific capacitance of SS-0.1 dropped to 78.86% of its initial value after 2,000 cycles at a current density of 100 mA g^-1 [Figure 4D]. In contrast, SS-1 showed superior cycle stability with negligible change in the specific capacitance in the first 2,000 cycles. Even after 10,000 cycles, a retention rate of 90.43% was obtained. The energy density and power density of SS-0.1 and SS-1 are shown in Figure 4E, where SS-0.1 and SS-1 show the same power density which increases from 10 mW kg^-1 to 25 mW kg^-1 upon an increase in the current density. On the other hand, the energy density of SS-1 is four times higher than that of SS-0.1 at the same applied current density. The small decreases in energy density from 9.4 mWh kg^-1 to 8.5 mWh kg^-1 for SS-0.1 and from 35.0 mWh kg^-1 to 33.3 mWh kg^-1 for SS-1 were observed with an increase in discharge current density from 40 mA to 100 mA g^-1. Ragone plots of SS-0.1 and SS-1 in Figure 4F are almost two horizontal lines, and the increase in power density does not weaken the energy density significantly, which is highly desirable for the practical application of this material in a supercapacitor.

Understanding the cycle stability of the CsPbBr₃ electrode

It is clear that the performance of SS-1 is significantly better than SS-0.1, in particular, the cycle stability. Therefore, further research was undertaken to investigate the underlying reasons. After the electrochemical cycling test, characterization of the electrode material in SS-0.1 by EDS shows the formation of larger crystals that are different from the pristine CsPbBr₃ crystals on the surface of the CsPbBr₃ electrode [Figure 5]. EDS mapping [Figure 5A-F] shows that F and P elements are concentrated on the large crystals while Br and Pb elements are distributed around the large crystals while Cs exists across the whole area. Therefore, this infers that the large crystals are mainly composed of three elements, Cs, P, and F. The EDS line scan also confirms these findings [Supplementary Figure 6A and B]. Supplementary Figure 6B shows the position of the EDS line scan on the large crystal. The concentration of F and P is higher in the region of the large crystals than in their surroundings, while the opposite trend is found for Br and Pb elements. The distribution of Cs is largely uniform. The atomic ratio of P and F is very close to 1:6, proving the existence of the PF₆^- functional group [Supplementary Figure 6C]. Therefore, we suspect that the large crystal is probably a secondary phase, such as CsPF₆ formed on the electrode during the electrochemical charge-discharge process. The XRD pattern in Figure 6A confirms that, indeed, besides CsPbBr₃, CsPF₆ also exists in the electrode. This is consistent with previous reports where CsPF₆ was formed in CsPbBr₃ in the presence of a TBAPF₆ electrolyte^[9]. On the other hand, Samu et al. postulated that the phase transformation of CsPbBr₃ is triggered by the reduction of Pb²⁺ at -1.4 V vs. Ag/AgCl reference electrode^[9]. However, the reduction potential of Pb²⁺ is -0.13 V vs. standard hydrogen electrode, and -0.37 V vs. saturated calomel electrode)^[39-41], respectively. In addition, our results show that the transformation from CsPbBr₃ to CsPF₆ can occur even at a low voltage of 0.4 V rather than -1.4 V^[6]. Supplementary Figure 7A and B show the SEM of the CsPbBr₃ electrode in SS-0.1 after 40 CV cycles at a scan rate of 5 mV s^-1 over a potential window of 0.4 V. The large cubic crystal should be CsPF₆, and the surrounding tiny crystals are CsPbBr₃[Supplementary Figure 7A].Supplementary Figure 7B shows the distribution of CsPF₆ on CsPbBr₃ electrode. Therefore, we propose CsPF₆ is formed by an ion exchange reaction, as shown in Equation (1).

Figure 5. EDS mapping of CsPbBr₃ electrodes in SS-0.1 after the electrochemical test. (A) the layered image of mapping. Mapping result of different elements (B) F; (C) P; (D) Cs; (E) Br; (F) Pb.

Figure 6. (A) The XRD scan of SS-0.1 and SS-1 after the cycles of charge/discharge. (B) the ICP-MS results of SS-0.1 and SS-1 after 2,000 cycles of charge/discharge. (C) XPS of CsPbBr₃ electrode in SS-1 after 2,000 cycles of charge and discharge without etching process. (D) TEM of TBAPF₆ covered CsPbBr₃ crystal obtained from CsPbBr₃ electrode after 2,000 cycles of charge and discharge.

The content of each crystal phase in the electrode at different cycling stages was then analyzed quantitatively based on XRD measurements [Figure 6A] using TOPAS software, as shown in Supplementary Table 3. Before the test, CsPbBr₃ electrodes contain 89% cubic CsPbBr₃ and 11% orthorhombic CsPbBr₃. Upon electrochemical cycling, the amount of CsPF₆ in the perovskite electrode of SS-0.1 increased with the increase of cycle numbers and reached 24.48% after 2,000 cycles of charge/discharge. In contrast, the contents of CsPF₆ in the CsPbBr₃ electrode of SS-1 is only 4.01% even after 10,000 cycles, which is much lower than the amount of CsPF₆ in SS-0.1. Therefore, a high concentration of TBAPF₆ can effectively hinder the dissolution of the perovskite in the electrolyte, preventing the formation of CsPF₆. To confirm this hypothesis, we measured the content of Cs and Pb in the electrolyte of the SS-0.1 and SS-1 devices after 2,000 cycles by ICP-MS. The results [Figure 6B] show that the concentrations of Cs and Pb ions in the electrolyte of SS-0.1 are 4.4 ppm (mg L^-1) and 8.7 ppm (mg L^-1), respectively, while their concentrations in the electrolyte of SS-1 are significantly lower with values of 0.083 and 0.91 ppm (mg L^-1), respectively. This confirms the importance of having a high concentration of TBAPF₆ in the electrolyte to stabilize the perovskite electrode materials. This could be related to suppressed decomposition or reorganization of CsPbBr₃ surrounded by a high concentration of large electrolyte ions, because of the formation of the nano-scaled film on the surface of the CsPbBr₃ electrode, thereby increasing the overall cycling stability of the supercapacitor. The XPS wide scan without etching for the CsPbBr₃ electrode in SS-1 after 2,000 cycles of charge and discharge in Figure 6C indicates that the elements present in the film on the CsPbBr₃ electrode surface only include C, N, O, P, and F rather than Cs, Pb, and Br. For further investigation, the components on the electrode in SS-1 were redispersed in hexane for TEM measurements. TEM images in Figure 6D visually show that the CsPbBr₃ nanocrystals still keep a cubic shape and were wrapped in an amorphous substance, which could be TBAPF₆. Further investigations carried out by FTIR [Supplementary Figure 8] proved the existence of the PF₆^- functional group on the electrode. The absence of TBAPF₆’s peaks in the XRD pattern [Figure 6A] also indicates that the TBAPF₆ is amorphous, which is consistent with the TEM results. Therefore, the dramatically increased stability resulted from the coverage of amorphous TBAPF₆ on the CsPbBr₃ electrode, which only happens at high concentrations. Additionally, although the TBAPF₆ layer (solid electrolyte interphase) covered on the surface of the electrode prevents the perovskite from decomposing, it also reduces the coulombic efficiency because it hinders the ion transfer from electrolyte to the surface of CsPbBr₃ for adsorption. Moreover, side reactions such as the formation of CsPF₆ could reduce the coulombic efficiency to some extent, especially at low current densities. Supplementary Figure 9 shows the cycle stability of SS-1 at a current density of 40 mA/g and Cs ions concentration in the electrolyte after 2,000 cycles. The retention rate of SS-1 at a current density of 40 mA/g after 2,000 cycles is 89.77% [Supplementary Figure 9A]. We can see that the concentration of Cs and Pb under low current density is much higher than the one under high current density [Supplementary Figure 9B].

Specifically, it is known that the chemical bonding strength between the Cs⁺cation in the perovskite crystal structure with the [PbBr_x]^- the framework is weak, and the energy barrier for migration of Cs⁺ in the perovskite crystal structure is 0.3-0.7 eV^[42-44]. Experimental evidence has shown that Cs⁺ can drift away from the perovskite crystal structure under certain electrical fields which are high enough to overcome the energy barrier for Cs⁺ migration. Under low current density, it takes a longer time for the electrode to reach the maximum voltage compared to the situation under higher current density. Specifically, according to the GCD plots, it takes 45 s for the charging plot to reach the maximum at 40 mA/g, whereas the charging time is much shorter (55 s) under 100 mA/g. The much higher content of Cs⁺ ions in the electrolyte at 40 mA/g than that at 100 mA/g indicates that Cs⁺ ions have sufficient time to drift out of the perovskite crystal structure when the current density is low (40 mA/g), which eventually leads to the collapse of the perovskite structure because Cs⁺ plays a role in stabilizing the crystalline lattice in perovskite^[45-47]. The high concentration of Cs⁺ will also improve the formation of CsPF₆ (side reaction). In comparison, the faster charging/discharge at the larger current density means Cs⁺ ions might not have sufficient time to leave the perovskite.