Study of Potential‐Induced Degradation in Glass‐Encapsulated Perovskite Solar Cells under Different Stress Conditions

Potential‐induced degradation (PID) is an important reliability issue of photovoltaic modules. For future field applications of perovskite photovoltaic modules, it is important to study the PID behavior under real‐world operating conditions, which has not yet been thoroughly researched. This work presents PID investigation of glass‐encapsulated perovskite solar cells (PSCs) at different stress conditions for an extended duration of 55 h. At room condition (25 °C, 20% relative humidity [RH]) the efficiency is reduced by 59% when −1000 V is applied to the short‐circuited PSCs, whereas under elevated stress condition (60 °C, 60% RH) the device efficiency suffers severe degradation of >90%. The PID effects are analyzed with several characterization methods, revealing that Na+ ion migration from the front glass pane toward the perovskite layer causes the degradation. Application of a reverse voltage bias right after PID results in very poor recovery under elevated condition compared to the recovery of devices under room condition. It is proposed that PID and its recovery rate depend on the external stress conditions and the reverse bias strategy is not sufficient for the recovery. Possible mitigation strategies which can open a new avenue for further research on PID in PSCs are also presented.


Introduction
Organic-inorganic halide perovskites (OIHPs) have shown remarkable progress in the last 10 years toward achieving highly efficient perovskite solar cells (PSCs). Currently, the record efficiencies of single-junction PSCs and perovskite/Si tandem solar cells are 25.7% and 31.3%, respectively. [1] At the photovoltaic (PV) module level, single-junction PSCs have demonstrated over 17.9% of power conversion efficiency (PCE) owing to their excellent electrical and optical properties. [2] However, PSCs have been found to be sensitive to moisture, light, and thermal stress, making them unstable over time. [3][4][5][6] Extensive research is being carried out on improving the long-term stability of PSCs, not only under solar simulators inside the laboratory but also under harsh outdoor environments. [7][8][9][10][11] Nevertheless, several extrinsic and intrinsic issues arising from the environment and the biasing electric field during outdoor operations affect the real-time device operation. Among a number of degradation mechanisms, potential-induced degradation (PID) is considered as one of the major reliability issues in commercial solar cells, which may lead to catastrophic failure of the device performance in a short time. [12,13] PID is a well-known phenomenon in crystalline Si PV modules, but the study of PID effects in perovskite solar cells is still at its early stage.
Some initial investigations on PID were performed for perovskite solar cells [19][20][21] and perovskite/Si tandem minimodules. [22] A summary of the existing works in the literature is listed in Table 1. Initially, Carolus et al. applied high voltage stress of 1000 V along with a thermal stress of 60°C and relative humidity (RH) of <60% for 18 h and observed that the PV efficiency drops by 95%. [19] Under the same test condition, a comparative approach was conducted to study the influence of PID on ni-p-and p-i-n-type perovskite solar cells. [20] The efficiency of the devices with n-i-p architecture reduced to 34.7% of their initial efficiency, and only 28% of the initial efficiency was retained in case of p-i-n-type devices. From these works it is clear that encapsulated perovskite solar cells are susceptible to PID, and the degradation has been shown to be predominantly due to the oversaturation of the perovskite layer with Na þ ions from the glass pane which, in turn, alters the perovskite bulk properties. [19] Brecl et al. also fabricated perovskite minimodules wrapped in a conductive foil which were exposed to À1000 V voltage stress for 168 h at room condition (RC). [21] More recently, Xu et al. [22] verified the degradation of single-device perovskite/ silicon tandem modules subjected to À1000 V at 60°C and <20% RH for 24 h. Around 50% efficiency reduction was reported owing to the accumulation of Na þ ions on the perovskite and DOI: 10.1002/solr.202300100 Potential-induced degradation (PID) is an important reliability issue of photovoltaic modules. For future field applications of perovskite photovoltaic modules, it is important to study the PID behavior under real-world operating conditions, which has not yet been thoroughly researched. This work presents PID investigation of glass-encapsulated perovskite solar cells (PSCs) at different stress conditions for an extended duration of 55 h. At room condition (25°C, 20% relative humidity [RH]) the efficiency is reduced by 59% when À1000 V is applied to the short-circuited PSCs, whereas under elevated stress condition (60°C, 60% RH) the device efficiency suffers severe degradation of >90%. The PID effects are analyzed with several characterization methods, revealing that Na þ ion migration from the front glass pane toward the perovskite layer causes the degradation. Application of a reverse voltage bias right after PID results in very poor recovery under elevated condition compared to the recovery of devices under room condition. It is proposed that PID and its recovery rate depend on the external stress conditions and the reverse bias strategy is not sufficient for the recovery. Possible mitigation strategies which can open a new avenue for further research on PID in PSCs are also presented. silicon cells. However, it is important to note that all the above works on perovskite related PID studies were performed under moderate temperature and humidity condition, and the duration of the PID stress in these works was not sufficiently long to arrive at concrete conclusions on the effects of PID. [19,20] In the previous extensive PID studies on Si PV modules, both temperature and humidity were found to play an important role in the extent of PID, [23][24][25][26][27][28] and neglecting the influence of high temperature and high humidity on PID might lead to misrepresentation of the PID sensitivity of perovskite-related devices in real-world conditions. There is therefore still a large gap in the perovskiterelated PID research under accelerated testing, such as elevated stress condition (ESC). Furthermore, the recovery of PID was successfully demonstrated for silicon PV modules by applying a reverse polarity bias to the affected cells. [29,30] A similar strategy was adapted by Brecl et al. [21] for perovskite solar cells and Xu [22] for perovskite/Si tandem cells. When the PID-affected perovskite devices were exposed to a reverse polarity bias of þ1000 V, it was revealed that the penetration of Na þ ions is reversible to some extent and that the applied positive bias drives Na þ ions away from the perovskite layer. However, it is unclear whether the recovery process can fully or partially recuperate the performance of the encapsulated perovskite solar cells after PID under ESC.
In this work, we demonstrate the effects of PID on glassencapsulated perovskite solar cells under two different conditions: RC (25°C and 20% RH) and ESC (60°C and 60% RH). In particular, we investigate the behavior of encapsulated perovskite solar cells with negatively biased PID (nPID). Light and dark current-voltage (I-V ), external quantum efficiency (EQE), steady-state photoluminescence (PL), photoluminescence quantum yield (PLQY), and X-ray diffraction (XRD) characterizations are carried out to understand the underlying mechanisms. Furthermore, recovery behaviors are also investigated for devices with nPID by applying a reverse (positive) bias under both stress conditions.

Results and Discussion
In this work, single-junction single-cell glass-encapsulated perovskite solar cells (henceforth referred to as "devices") with an active area of 1.2 cm 2 are used to study the effect of potential-induced degradation. As shown in Figure 1, the perovskite solar cell has a p-i-n configuration with a CsFAMA-based triple-cation perovskite absorber layer, an indium tin oxide (ITO) transparent front electrode, a self-assembled monolayer (SAM) as the hole transport layer, a C 60 /BCP stack as the electron transport layer, and a silver (Ag) layer as the rear electrode. The devices are encapsulated in a controlled nitrogen environment with a cover glass and epoxy resin as encapsulant. Details of the materials and fabrication procedure are provided in the Supporting Information. All the PID experiments in this work are performed using the standard foil method as described in IEC 62804-1. [20,31] To clearly understand the effect of PID under different stress conditions, the devices are divided into two groups, i.e., devices tested under RC and devices tested under ESC, each consisting of 16 devices. Initially, the devices in each group are tested without application of any high voltage stress, to elucidate the effects of RC and ESC on the performance of the devices. These devices are termed RC and ESC, respectively. For RC the climate chamber is maintained at 25°C and 20% RH, while for ESC it is maintained at 60°C and 60% RH. Note that the fluctuation of temperature and the RH in the climate chamber is controlled within AE1°C and AE2% RH, respectively. To understand the impact of PID, the devices are exposed to negatively biased high voltage stress (À1000 V) for 55 h under RC and ESC. These devices are termed as nPID_RC and nPID_ESC, respectively. Later, the PID-affected devices are subsequently subjected to reverse (positive) bias for 90 h under RC (devices are termed as ((nPID) þ _RC)) and ESC (devices are termed as ((nPID) þ _ESC)) to study the respective recovery behaviors. The performance of all these devices is compared with the reference devices named as "ref," which were stored in a N 2 environment all the time. The detailed nomenclature of the devices used in this work is summarized in Table 2. All the samples listed in Table 2 are encapsulated immediately after fabrication.

Under Room Condition
Initially, the average of the absolute efficiencies of all the devices measured immediately after encapsulation was 19.6%, while the average efficiency of the reference devices stored in the N 2 environment after 55 h was 20.1%, as shown in Figure S1, S2, Supporting Information. The normalized box plots of the I-V  The PID experiment is conducted using the foil method where À1000 V high-voltage stress is applied between the short-circuited perovskite solar cell and the aluminum foil on the front glass.
www.advancedsciencenews.com www.solar-rrl.com measurements under RC are shown in Figure 2a-d. When the devices were biased at À1000 V for 55 h (nPID_RC devices), we observed that the normalized efficiencies reduced dramatically by 59% compared to the reference devices as shown in Figure 2a. The V OC decreased by 13%, and the reduced efficiency was primarily from the J SC (decreased by 27%) and FF (decreased by 36%). The dark I-V characteristics for these devices are shown in Figure 2e. The dark I-V curves of negatively biased PID devices have a steeper slope, indicating a lower shunt resistance and increased leakage current compared to the reference devices (Table S1, Supporting Information). For the devices with negatively biased PID, %40% reduction in the EQE was observed compared to the reference devices, as shown in Figure 2f. This reduction in the EQE agrees well with the decrease in J SC for negatively biased PID devices, as can be observed in Figure 2d. Strategies to recover the devices with negatively biased PID ((nPID) þ _RC) were also investigated in our work. We performed experiments by applying the reverse bias (þ1000 V) to the affected devices under RC for 90 h. From the normalized efficiency plots in Figure 2a, it is found that the performance of the devices after recovery is enhanced from 37% to 58% of the reference device efficiency. It was observed earlier that the J SC and FF are significantly reduced for devices with negatively biased PID (Figure 2b,d). With the recovery via reverse (positive) biasing, the J SC and FF of the devices are improved from 69% and 63% of the reference devices to 83% and 77%, respectively (Figure 2b,d), indicating a partial recovery.

Under Elevated Stress Condition
We also performed similar experiments under ESC to understand if the PID behavior is dependent on the operating conditions. Initially, the average of the absolute efficiencies of the devices kept under ESC for 55 h without the application of external voltage stress was observed to be 18.5% ( Figure S3, S4, Supporting Information). Hence, without the high voltage stress, the percentage loss in normalized PCE due to only the ESC is %7% (Figure 3a). When the devices were subjected to high negative bias under ESC (nPID_ESC), there was a catastrophic effect and more than 90% of efficiency was lost with respect to the reference devices in 55 h (Figure 3a). Comparing with the percentage loss of (normalized) efficiency in the sister devices under RC (nPID_RC), which was 59% in 55 h (Figure 2a), the results clearly show that there exists a domino effect of high voltage stress and high temperature on the PID of perovskite solar cells. These devices show much stronger PID compared to devices stressed under the RC. It is therefore of utmost importance to investigate the underlying mechanism of PID at ESC for future field applications. The dark I-V curves of the devices under ESC are shown in Figure 3e. The leakage current in devices with negatively biased PID under ESC is as high as 58.0 mA cm À2 (Table S2, Supporting Information). Furthermore, the EQE reduces by %60% for devices with negatively biased PID compared to the reference devices, as shown in Figure 3f.
We also perform recovery to the devices with negatively biased PID under ESC by applying þ1000 V for 90 h. From Figure 3a, it can be observed that the efficiency of the (nPID) þ _ESC devices could be recovered by only %10%. The results of the I-V measurements show that applying a reverse bias to the devices with negatively biased PID results only in partial recovery in RC, while the recovery under the ESC is minimal.

PL Analysis
To understand the underlying mechanism for the observed PID behaviors of our devices, we performed PLQY spectroscopy measurements (Figure 4). For devices under RC, except for the devices with negatively biased PID, the PL peak intensity of all other devices is centered around the wavelength of 736.7 AE 0.2 nm. For the devices with negatively biased PID, a redshift of approximately 8 nm is observed along with a strong quench in the PL peak intensity by more than 50%, as shown in Figure 4a. The redshift is an indication of the formation of iodine-rich regions in the perovskite layer, while the quenching suggests that there is nonradiative recombination in the devices. The corresponding PL spectra under ESC are shown in Figure 4b. For devices under ESC without high voltage stress, the redshift of the PL peak is only 1 nm from the reference device, whereas the PL peak redshifted significantly by 22 nm and quenched by %62.5% for devices with negatively biased PID. The findings from the PL spectra suggest that the negative high voltage stress tends to result in more iodine-rich regions and hence more nonradiative recombination sites. Such crystal deformation is further aggravated under ESC. These PL results are also in excellent agreement with the analysis made from the I-V characteristics. A similar redshift is also observed in the normalized steady-state PL spectra ( Figure S5, Supporting Information) of the corresponding devices.
The PLQY spectra of Figure 4a also provide insights into the recovery mechanism under reverse (positive) bias application. Under RC, the PL peak position of the devices with negatively biased PID blueshifted from 745.0 to 740.0 nm after the recovery process and the PL intensity was also partially enhanced. As the peak position is still redshifted compared to the peak position of the reference devices (736.4 nm), it indicates partial recovery. Under ESC, although there is a slight blueshift in the PL peak intensity (from 757.6 to 750.0 nm) for devices with negatively biased PID after recovery, we observe that the final PL peak position is still clearly different to the PL peak position of the reference devices (736.0 nm). Furthermore, in this case the recovery of the PL intensity is much weaker compared to the RC case, indicating a very poor recovery.

XRD Analysis
The degradation of perovskite solar cells under PID stress is observed using XRD characterization and the tests are performed without the presence of encapsulation glass www.advancedsciencenews.com www.solar-rrl.com ( Figure S6, Supporting Information). Initially, three important observations can be made from the XRD plots for devices under RC (marked by star, diamond, and spade symbols in Figure 5a): 1) all the devices under RC exhibit a typical perovskite peak at 14.2°, but the intensity of this peak is weakest for the device with negatively biased PID; 2) there is a clear diffraction peak at 12.54°o nly for the devices with negatively biased PID, indicating the presence of PbI 2 which is a result of halide segregation; [32] 3) for the devices with negatively biased PID, there is another unique peak at 38.2°, which corresponds to the presence of AgI in the perovskite layer. [33,34] This undesired presence of AgI is a clear indication of ion migration into the perovskite layer. Similar results are observed for devices under ESC, as shown in Figure 5b. First, the intensity of the intended perovskite peak at 14.2°is weakest for the device with negatively biased PID. Moreover, as it is well known that elevated temperatures accelerate the formation of excess PbI 2 , [35,36] the devices under ESC without any voltage bias exhibit a tiny PbI 2 peak at 12.54°. This undesired peak is significantly enhanced in the devices with negatively biased PID under ESC and it clearly dominates the perovskite peak (14.2°). Finally, the AgI peak is also present in the devices under nPID_ESC, confirming the significant degradation.
After the application of þ1000 V of reverse bias for both room and ESC, it can be observed that the intensity of the intended perovskite peak at 14.2°is partially recovered. However, the undesired AgI and PbI 2 peaks are still present in these devices. After the reverse biasing, for the devices under room www.advancedsciencenews.com www.solar-rrl.com temperature stress, the intensity of the PbI 2 peak is reduced whereas the AgI peak remains almost unchanged. This confirms that the devices exposed to reverse bias are only partially recovered from the poor performance after the PID stress. On the other hand, recovery of the devices with negatively biased PID under ESC is limited by the presence of the strong PbI 2 peak arising from the perovskite decomposition, while the corresponding peak is much weaker for the case under RC. In addition, the AgI peak is still evident in the devices under reverse bias. Hence, the negative high-voltage bias has a detrimental effect to the operation of our perovskite solar cells and results in catastrophic failure in the device performance especially under the ESC.

Discussion of PID Mechanism and PID Recovery
Our results provide an in-depth insight about the impact of different stress conditions on PID of perovskite solar cells and its recovery behavior. They indicate that PID of perovskite solar cells is dependent on the operating conditions. The results reveal that  www.advancedsciencenews.com www.solar-rrl.com application of a high negative voltage bias under ESC results in more severe PID compared to RC. In both cases, the underlying PID mechanism indicates that application of a high negative voltage between the short-circuited solar cell and the Al foil on the module's front glass surface leads to migration of Na þ ions from the soda-lime glass pane into the perovskite solar cell as shown in Figure 6. We have performed energy-dispersive X-ray (EDX) characterization on reference devices under RC, devices with negatively biased PID and devices after recovery, as shown in Figure 7. The scanning electron microscope (SEM) images of the perovskite solar cell stack are shown in Figure 7a. In case of reference devices, Na þ ions are not spotted even in the ITO layer as observed from Figure 7b. However, after application of high negative voltage bias, high concentration of Na þ ions appears in the perovskite layer as evident from Figure 7c. These results confirm the movement of Na þ ions toward the perovskite layer for devices with negatively biased PID. The perovskite solar cells do not experience any electric field as they are shorted, but the electric field between the Al foil on the glass pane and the ITO electrode causes the movement of Na þ ions. [19] These Na þ ions from the glass move into the perovskite layer under high negative voltage bias and interact with the negatively charged iodine (I À ) ions in the perovskite layer, which may lead to the formation of a weak Na þ I À bond as depicted in Figure 6. The rate of transport of Na þ ions is further enhanced when the perovskite solar cell is operated under ESC, resulting in an altered perovskite structure as shown in the PLQY and XRD measurements (Figure 4 and 5). On the other hand, at the rear surface of the solar cell, the incursion of Ag þ ions from the rear Ag contact into the perovskite layer and the formation of a strong AgI bond is clearly evident under both room and ESCs. Subsequently, when the devices with negatively biased PID are treated by applying a reverse (positive) bias for recovery at RC, the blueshift of the PL peak and increase of the PL intensity upon recovery confirms the reduction of the defect densities and an improvement of the structural quality of the perovskite layer. The XRD results also confirm the recovery of the perovskite peak and the suppression of the PbI 2 peak. The EDX results in Figure 7d also indicate that the Na þ ions are retracted away from the perovskite layer by the applied reverse bias and the reduction in the concentration of Na þ ions in the perovskite layer is clearly visible in devices after recovery. Although most of the Na þ ions are driven back to the glass pane by the reverse bias, the amount of AgI seems to remain unchanged, indicating only a partial recovery. It is believed that the iodine (I À ) from the perovskite layer and the silver (Ag þ ) from the rear electrode react to form AgI, which is a more stable form as compared to bonds such as NaI. As a result, the AgI peak cannot be reduced, and the devices experience only partial recovery under RC.
On the other hand, the insignificant yet noticeable redshift in the PL peak and the presence of PbI 2 in the XRD results for the samples kept in ESC without any voltage bias also reiterate the fact that the perovskite layer is affected by higher temperature. [3][4][5][6] Application of negatively biased PID further worsens the performance of the devices and it is observed that the PL peak has an additional redshift, resulting in excess ion migration compared to the devices without any high voltage stress. Also, in this case the PbI 2 peak in the XRD plot is significantly intensified compared to the intended perovskite peak, indicating perovskite decomposition. After application of a reverse bias, the presence of a strong PbI 2 characteristic peak depicts that the devices with negatively biased PID under ESC can barely be restored by only applying a reverse bias. In addition, the AgI peak could not be removed or reduced even after recovery through Figure 6. Mechanism of degradation in the investigated PSCs under negatively biased PID. The presence of the strong electric field due to the applied potential difference forces the Na þ ions in the glass to drift into the ITO layer. As a result of the increasing Na þ concentration in the ITO, some of these ions then diffuse through the SAM layer into the perovskite layer. The depth of penetration into the perovskite layer depends on the potential difference, temperature, and thickness of the ITO and SAM layers. Note that the other elements of the perovskite layer (Cs, FA, MA) are not shown in the figure for the sake of simplicity. www.advancedsciencenews.com www.solar-rrl.com reverse biasing. Although many of the Na þ ions might migrate back into the front glass pane, it is unlikely to at least partially recover the devices via reverse biasing under ESC.
The findings of this work show that the application of a reverse bias alone is not sufficient for complete recovery of the PID of glass-encapsulated perovskite solar cells, especially under ESC. The alternative device-level solutions to PID may include (but are not limited to) the addition of a blocking or passivation layer at the front surface of the perovskite layer and/or the identification of an inert material that can occupy any of the A, B, X sites of the perovskite structure to inhibit the penetration of Na þ ions. In addition to the prevailing studies on the stability of perovskite solar cells under ESC it is important to also address the PID issue, especially under real-world operating conditions such as high device temperatures.

Conclusion
This work reports the first detailed investigation of the effects of PID in glass-encapsulated perovskite solar cells under both RC and ESC. After 55 h of stress testing, the devices with negatively biased PID suffered significant and rapid degradation under ESC retaining only %10% of the initial PV efficiency, while the devices under RC show better tolerance to PID retaining %41% of the initial efficiency. Under RC the degradation can be recovered partially by applying a reverse bias (þ1000 V), whereas the recovery is very small under ESC. Dark I-V, PL, PLQY, XRD, SEM, and EDX characterizations were performed to study the underlying degradation mechanisms of PID and its recovery under RC and ESC. Our results confirm that the ESC plays an important role for the magnitude of PID, and that reversing the polarity alone is not a sufficient condition to recover the PID. We also proposed alternative strategies for PID mitigation. The underlying degradation mechanisms revealed here will also guide future work on mitigation of PID in perovskite solar cells, which is a prerequisite for the successful commercialization of this novel PV technology.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.