Thermally induced failure mechanisms in double and triple cations perovskite solar cells Open Access

Hybrid metal-halide perovskite solar cells (PSCs) have attracted immense attention from the photovoltaic community over the past decade.¹ Their combination of low fabrication costs and high solar-to-electrical power conversion efficiencies (PCEs) makes them promising in the emerging photovoltaic (PV) technology.² Starting from initial PCEs of 3.8% in 2009, the PCEs of PSCs have increased to levels that exceeded 25% in 2021.^3,4 Thus, PSCs now have PCEs that compete with those of crystalline silicon solar cells.⁵ 

Much of the interest in PSCs is due to the exciting combinations of optoelectronic properties of their perovskite absorber layers. These include high extinction coefficients, excellent light absorption, low exciton binding energy, long range charge carrier diffusion length, and fast ambipolar charge transport.^6,7 However, there are still significant concerns about their recombination losses, long term stability, and upscaling to large-area devices.^8,9 There is, therefore, a need for further research to address these issues, as a part of integrated efforts to develop robust and stable PSCs.

Perovskite solar cells are multi-layered structures that degrade when exposed to different environmental conditions. Several studies have shown that perovskite materials/layers are unstable in the presence of environmental factors such as irradiation,¹⁰ moisture,¹¹ oxygen,¹² and elevated temperatures and temperature ranges.^13,14

Thermally induced degradation phenomena can occur in perovskite solar cells during processing and service under exposure to light. During processing, multiple annealing temperature steps¹⁵ can degrade solar cell performance. Thermal stability testing^16,17 can also lead to the early stage decomposition of the PSCs. The thermally induced degradation persists up to higher operating temperatures (60°C and above),¹⁸ at which device degradation can be characterized via initial exponential decay, proceeded by a slow linear loss regime. The situation can be similar to that of a device aged at −10 °C.¹⁴ These regimes are also associated with significant changes in the layer and interfacial microstructures in the active layer and the adjacent hole and electron transporting layers (HTLs and ETLs).^19,20

In an effort to improve the stability and durability of perovskite materials, more careful engineering of new material compositions has been carried out in the past to develop perovskite compositions that are based on combinations of different cations and mixed halides,^17,21 starting with the initial single cation to methylammonium lead iodide (MAPbI₃) to formamidinium lead iodide (FAPbI₃) and a mix of inorganic cesium and organic materials.²² However, these problems can be mitigated through various approaches. For example, the incorporation of bromide (Br) into the X-site to form I–Br mixed halide perovskites has been shown to improve thermal stability in MAPb(I_1−xBr_x)₃ due to a shorter and stronger Pb–Br bond than the Pb–I bond.²³ Furthermore, recent studies have suggested that the introduction of Br into the MAPbI₃ lattice can increase the crystallinity of the perovskite and also vary the potential barrier at grain boundaries to reduce recombination losses in the perovskite.²⁴ Thus, this emphasizes the effect of bromide doping on the phase stability of α-perovskites. Notably, formamidinium lead triiodide (FAPbI₃) perovskites show promise as a photovoltaic material due to their narrower bandgap (Eg) of 1.48 eV, which is closer to the Shockley–Queisser ideal bandgap of 1.40 eV for single-junction solar cells, implying greater PCEs.²⁵ Moreover, the FA organic cation presents a larger ionic radius than MA, and the corresponding FA lead triiodide (FAPbI₃) perovskite is more resistant to heat stress than MAPbI₃.²⁶ However, both cations have exhibited degradation under ambient conditions with different relative humidities.²⁷ During fabrication, the microstructure of the black perovskite phase (FAPbI₃) is thermodynamically stable only below 160°C. This may result in a yellow photo-inactive δ phase below the phase transition temperature.²⁸ To prevent the yellow phase of FAPbI₃ at room temperature, small amounts of methylammonium (MA)²⁹ were used to stabilize the α-phase of FAPbI₃. In other works, Cl was added to improve crystallization processes and moisture during film growth, which improved performance and reproducibility.³⁰ In addition, introducing MA⁺ cations and Br⁻ anions into FAPbI₃ at the same time to obtain (MAPbX₃)_x(FAPbI₃)_1−x results in a synergetic effect that stabilizes the perovskite phase.³¹ In addition, the approach of adding bromide (Br) and methylammonium (MA) has also been used to enhance the intrinsic operational stability of formamidinium lead iodide (FAPbI3-based) PSCs.³² 

Mixed cation PSCs that are based on (MAPbX₃)_x(FAPbI₃)_1−x perovskite absorbers have been shown to have the highest power conversion efficiencies, with certified PCEs that exceed 25%.^30,33,34 However, under heat stress, the mixed (MAPbX₃)_x(FAPbI₃)_1−x perovskite is also prone to decomposition due to the presence of MAPbX₃. As a result, it is important to develop a perovskite structure that is stable in both low and high temperature ranges.

To improve light stability of PSCs, the α phase can be promoted by adding small inorganic cesium (Cs) cations that assist black phase crystallization.^28,35 Lee et al.²⁷ had found that Cs can stabilize the black phase of FAPbI₃. Furthermore, mixed CsMAFA cation-based devices have been developed with power conversion efficiencies of more than 22%.³⁶ 

It is also important to note that defects at the interfaces and grain boundaries have adverse effects on the performance and stability^37,38 of PSCs. The defects can also act as recombination sites for charge carriers, as well as sites for moisture ingression.³⁹ 

The stability of PSCs is also dependent on the robustness of electron transport layers (ETLs), hole transport layers (HTLs), and metal electrodes. Among the several organic and inorganic materials, spiro-OMeTAD and its composites are mostly used as HTLs in the best performing PSCs. This is due to its favorable energy level alignment and excellent and simple processing conditions that are compatible with most perovskite films.⁴⁰ 

Spiro-OMeTAD (HTL), commonly used in n-i-p PSC structures, can also represent another pathway of degradation at high temperatures.⁴¹ In addition, some of the materials that are used for the processing of HTLs (such as LiTFSI dopants) are very hygroscopic. They can, therefore, damage the perovskite layer at temperatures as low as 65°C.⁴² When PSCs are heated to 75°C for a long time, gold can migrate from the top electrode through the spiro-OMeTAD layer into the perovskite layer.⁴³ Heat-induced degradation has also been observed in PSCs with spiro-MeOTAD HTM layers.⁴⁴ After heating the PSCs to 50°C, the resulting decrease in the short-circuit current (Jsc) was attributed to changes in the transport properties of the spiro-MeOTAD. In addition, the thermally induced degradation of the ETLs had only a small effect on the instability of PSCs since both SnO₂ and TiO₂ have excellent thermal stability.⁴⁵ This suggests that the thermal degradation mechanisms should depend more on the perovskite absorber and HTL layers than on the ETL layer.

In this paper, we present a comparison of the thermally induced degradation of PSCs with double and triple cations. These include two mixed perovskites: (FAPbI₃)_0.97(MAPbBr₃)_0.03 (FAMA-based) and Cs_0.05(FA_0.95MA_0.05)_0.95Pb(I_0.95Br_0.05)₃ (CsFAMA-based) films; these were chosen and integrated into PSCs due their optimized thermal stability under environmental conditions.^15,46 The underlying changes in the layer/interfacial microstructure, current–voltage characteristics, optoelectronic properties, and solar cell stability are elucidated at 40, 50, and 60°C for 3 h. The underlying device failure mechanisms are presented before discussing the implications of the results for the development of double and triple cation PSCs with improved combinations of stability and performance characteristics (PCE, absorbance, and optical losses).

Etched fluorine-doped tin oxide (FTO)-coated glass substrates (FTO, ∼7 Ω sq⁻¹) were acquired from MSE Supplies (Tucson, AZ, USA). Dispersion diluted SnO₂ colloidal (15% in H₂O colloidal dispersion, Alfa Aesar), tin (IV) chloride pentahydrate (SnCl₄·5H₂O), formamidinium iodide (FAI; 98%), methylammonium chloride (MACl), methylammonium bromide (MABr), cesium iodide (CsI), lead bromide (PbBr₂), lead iodide (PbI₂; 99.9%), chlorobenzene (anhydrous) dimethyl sulfoxide (DMSO) (anhydrous), dimethylformamide (DMF) (anhydrous), acetone, iso-propyl alcohol (IPA), 4-tert-butylpyridine (tBP), acetonitrile, lithium bis(trifluoromethylsulfonyl) imide (Li-TFSI), 2,2′,7,7′-tetrakis(N,N-di-p methoxyphenylamine)-9,9′-spirobifluorene (Spiro-OMeTAD), and anhydrous chlorobenzene were all purchased from Sigma-Aldrich (Natick, MA, USA). Pure gold (99.999%) was purchased from Kurt J. Lesker Company (Lesker PA, USA).

The etched FTO glass substrates were cleaned in successive steps with detergent, deionized water, acetone, and IPA. This was carried out in an ultrasonic bath for durations of 15 min each. Then the cleaned glass substrates were then blow-dried in nitrogen gas, followed by UV–ozone cleaning (Novascan, Main Street Ames, IA, USA) to remove organic residuals prior to the deposition of the layers.

An ETL bilayer-SnO₂ was deposited onto FTO-coated glass. The solution of SnO₂ was prepared by dissolving tin (IV) chloride pentahydrate (0.016 g) in 1 ml of anhydrous iso-propanol. The precursor was then spin-coated onto the cleaned FTO-coated glass at 2000 rpm for 45 s before annealing at 180°C for 45 min. Subsequently, another layer of ETL (nanoparticle SnO₂ layer) was processed by spin-coating a diluted colloidal solution of SnO₂ (3.5 wt. %) at a speed of 4000 rpm for 45 s. This was then annealed at 150°C in the air for 30 min. The surface of the tin dioxide substrates was also activated prior to spin-coating of the perovskite layer with an ozone plasma.

The double cation (FAMA)-based perovskite solutions were obtained according to the preceding protocol.^15,47–50 The PbI₂ precursor solution was obtained by mixing 599.8 mg of PbI₂ with 950 µl DMF and 50 µl DMSO. The organic salt solution contains FAI:MABr:MACl (60 mg:6 mg:6 mg) in 1 ml IPA. After that, the PbI₂ solution was spin-coated onto SnO₂ at 1500 rpm for 30 s and then annealed at 70°C for 1 min. The organic salt solution was spin-coated on PbI₂ layers at a rotation speed of 1300 rpm for 30 s. Then, the film was annealed at 130°C min for 20 min to form the perovskite layer. The MACl in the precursor aids the achievement of high quality of the perovskite film,^30,51 and it is assumed to be negligible in the final perovskite film due to Cl release during thermal annealing.⁵² 

To fabricate the triple cation (CsFAMA)-based perovskite film, the perovskite precursor solution with excess lead was prepared using our previous protocol.⁵³ This was performed by mixing two stock solutions of 0.8M MAPbBr₃ and FAPbI₃ perovskite solutions in a specific volume ratio. The 0.8M FAPbI₃ stock solution was made by dissolving 0.481 g of FAI and 1.420 g of PbI₂ in a 4:1 (v/v) mixture of dimethylformamide (DMF) and dimethyl sulfoxide (DMSO). Furthermore, 0.8M of MAPbBr₃ stock solution was prepared by dissolving 0.313 g of MABr and 1.130 g of PbBr₂ in the DMF:DMSO (4:1 volume ratio) solvent and stirring with excess PbBr₂. Then, 40 μl of the 1.5M CsI stock solution (in DMSO solvent) was combined with 1.5 ml of the previously stated solution. The perovskite solution was then deposited onto the ETL film using the two-step spin coating technique where the first and second steps correspond to 2000 rpm for 10 s and 6000 rpm for 30 s, respectively. Chlorobenzene (100 µl) was dropped onto the spinning substrate 10 s before the completion of the second step. The substrate was then annealed at 100°C for 30 min.

To prepare the HTL solution, 72 mg of Spiro-OMeTAD was diluted in 1 ml of chlorobenzene, followed by addition of 30 µl of tBP and 35 µl of Li salt solution (260 mg of Li-TFSI dissolved in 1 ml of acetonitrile) to the spiro-OMeTAD solution. The mixture was then sonicated for 5 min before being spin coated onto the perovskite films at 4000 rpm for 30 s. Finally, a 90 nm thick layer of gold was thermally evaporated onto the Spiro-OMeTAD film using an Edward E306A thermal evaporator (Edward E306A, Easton PA, USA). A shadow mask was used to achieve the desired device area of 0.045 cm².

A Scanning Electron Microscope (SEM) (JEOL JSM-700F, Hollingsworth & Vose, MA, USA) was used to obtain top-view and cross-sectional SEM morphologies of the perovskite films, while the crystallographic characteristics of the perovskite films were determined using an x-ray diffractometer (Malvern PANalytical, Westborough, MA, USA) that was operated under a Cu Kα radiation, with a beta nickel filter at 40 kV and 40 mA. An ultraviolet–visible (UV–vis) spectrometer (Avaspec-2048, AVANTES Starline, BV, USA) was used to characterize the light absorption of perovskite films under both control and operating temperature conditions. Steady-state photoluminescence (PL) measurements were recorded with an excitation wavelength of 450 nm using laser light and a transient fluorescence spectrometer (FluoTime 300, Germany).

The photocurrent density–voltage (J–V) curves of the devices were measured using a Keithley source meter unit (Keithley, 2400 system, Tektronix, Newark, NJ, USA) connected to an Oriel solar simulator (Oriel, Newport Corporation, Irvine, CA, USA). The irradiance of light was calibrated at 100 mW cm⁻² (1 sun condition) using a 91150V silicon reference (MKS/Newport Instruments, Newport Beach, CA, USA). Electrochemical Impedance Spectroscopy (EIS) was then carried out to measure the charge transfer characteristics using a potentiostat (SP-300, BioLogic Instrument, Knoxville, TN, USA), with a bias of 0 V; the process was carried out in the dark state with an AC amplitude of 10 mV in the frequency range between 1 and 10 MHz.

Figures 1(a1-a4) and 1(b1-b4) present typical SEM microstructural images of FAMA and CsFAMA cation perovskites, respectively. Starting from the control [1(a1) and 1(b1)], a smooth compact and pinhole-free perovskite film layer was obtained at room temperature (RT), ∼25°C. The SEM images of the perovskite films that were exposed to operating temperatures between 40 and 60°C are presented in Figs. 1(a2-a4) for FAMA and Figs. 1(b2-b4) for CsFAMA.

The SEM images were analyzed using ImageJ software (Image J, National Institutes of Health, Bethesda, MD, USA) to estimate and compare the grain size values of several perovskite films exposed at RT and operating temperatures for 3 h. The average grain size estimated for FAMA increased from 343 ± 135.4 to 712 ± 157.3 nm, as shown in Figs. 1(a1) and Figs. 1(a4). In the case of CsFAMA perovskite film, a similar trend was also observed, but they are not as pronounced as those of FAMA perovskites. The average grain size increased from 212 ± 9.2 to 269 ± 8.7 nm, and the perovskite films do not show any serious decomposition as the temperature increased from RT (25°C) to 60°C.

Thus, the results show that perovskite grain size increases with increasing operating temperature. In addition, FAMA perovskite film had a rougher surface with pin-holes than CsFAMA perovskite film at ∼50–60°C. The increase in grain size suggests a tensile strain in the film, leading to compressive strain along the grain boundaries. Similar observations have been reported in prior work.^15,47,54 The strain at the grain boundaries can initiate cracking due to thermal mismatch.

To further understand the effects of thermal stress on the structure and crystallinity, we used an x-ray diffractometer (XRD) to produce x-ray diffraction patterns of the perovskite films at different operating temperatures. Perovskite thin films subjected to thermal stress can undergo two possible degradation mechanisms. First, the perovskite might decompose, resulting in the formation of lead iodide (PbI₂), and second, the phase of the perovskite film might change.⁵⁵ The patterns in Figs. 2(a) and 2(b) give the characteristic diffraction peaks of FAMA and CsFAMA, respectively. The diffraction peak at a 2θ value of ∼14° observed for the perovskite films suggests the presence of a pure black α-perovskite phase.^56,57 From the XRD analysis, we observed that the control films for the two mixed perovskites show strong and sharp diffraction peaks, indicating high crystallinity for the perovskite phase. The crystallinity of the films decreases as the operating temperature increases from RT to 60°C, as shown by the decrease in diffraction intensity of the main peaks at ∼14°, which is indexed to the (110) and the (101) plane in the samples for FAMA and CsFAMA, respectively. Thus, these also confirm the changes in perovskite bulk properties [as shown in 1(a4) and 1(b4)] and upon thermal stress.

We also noticed changes in the intensity and the area related to the PbI₂ in the perovskite films. The decomposition of the perovskites is confirmed by the increase in the peak intensity of PbI₂ along with the increase in the operating temperature over time.⁵⁸ Although the role played by excess lead iodide on the thermal stability of perovskite layers has not been fully understood to date, some studies indicate an influence of lead iodide on long-term stability.^55,59,60 When exposed to other stress factors such as heat or ambient atmosphere, perovskite layers with excess lead iodide degrade faster.^57,59 At 60°C, the PbI₂ is more predominant and pronounced for the FAMA perovskite, as shown in Figs. 2(a) and 2(c), which is consistent with the above-mentioned observation of the decrease in intensity at the (110) plane with an increase in operating temperature. In contrast, CsFAMA shows relatively less formation of lead iodide, but its intensity peaks at ∼14° under the same conditions [Figs. 2(b) and 2(d)]. In addition, more information on the variation of the (110) and (101) peaks is given in Table S1. This suggests that more structural defects are formed in FAMA while the structural defects are mitigated in CsFAMA. Since structural defects are more sensitive to operating temperature, this could be one of the significant reasons for better stability of CsFAMA, as FAMA perovskites appear to undergo more degradation upon thermal stress, leading to an increase in the formation of PbI₂, which is detrimental to the perovskite’s functional characteristics over time.⁵⁴ 

However, since there is a change in the bulk properties as indicated by surface morphology and XRD measurements, we also expect the thermal operation to affect the optical properties of the perovskite films. Therefore, we measured the optical absorbance and photoluminescence (PL) spectra of the perovskite films. The absorption spectra are presented in Figs. 3(a) and 3(b) for FAMA and CsFAMA, respectively, which are exposed to different operating temperatures. Before thermal aging, the control samples for both FAMA and CsFAMA films exhibit good absorbance throughout the visible region. We noticed that the absorption reduced with increasing operating temperature for both perovskite films, indicating a reduction in photo-active material.⁵⁸ This is also consistent with the trends observed in our previous report.⁶¹ 

In the case of FAMA, the absorbance spectrum reduces faster and stronger degradation occurs at 60°C thermal stress for 3 h [see Fig. 3(a)] as compared with the counterpart. The rapid decrease in the absorbance of the films at higher operating temperatures is associated with the initiation of grain boundary cracks/defects, which is driven by the strains along the grain boundaries.¹⁵ Thus, these results are consistent with the formation of pin-hole defects and rough surfaces observed in the SEM images, which suggest to have contributed to the reduction in perovskite intensities observed in the PL intensity as the temperature increases. The tensile strains in the perovskite films have been presented earlier for different temperatures.⁴⁷ 

To further gain insights into the effect of operating temperature on the optical properties of perovskites, steady-state photoluminescence (PL) spectroscopy was performed to assess charge transfer efficiency and quenching properties between the perovskite film and the ETL. Figures 3(c) and 3(d) show PL spectra of the half-cell consisting of FTO/SnO₂/perovskites for the two mixed perovskites. It is important to mention that since the characterization techniques are applied on half-cells, a quenching with FTO/SnO₂ is expected. However, we noticed an increase in PL intensity with no significant red shift in the PL emission for the two mixed perovskites before and after the thermal stress condition.

In FAMA [Fig. 3(c)], it is seen that the PL intensity rapidly increases as compared to CsFAMA [Fig. 3(d)] with increasing operating temperature after 3 h. This suggests a poor transfer of electrons from the perovskite to the SnO₂ layer due to the high surface defects at the interface (Fig. 4), thus allowing recombination of the charges.⁶² 

After identifying material deterioration from a thin-film perspective, we fabricated the device for the mixed halide perovskites with FAMA and CsFAMA cation compositions to further investigate the effect of operating temperature on the optoelectronic properties of the PSCs. The architecture of the PSCs is presented in Fig. 5(a), while the current–voltage characteristics of the devices are presented in Fig. 5(b), with the device parameters [Fig. 5(b) inset], for control FAMA and CsFAMA based devices. The results show that the FAMA device had a device efficiency of 17.43% and the CsFAMA device had an efficiency of 18.40%.

In an effort to understand and compare the effects of operational thermal stress on device performance, we monitored the operational stability at 40, 50, and 60°C for 3 h. The current density–voltage (J-V) curves in each condition were plotted after 3 h set intervals for the operations. Figure 5(c) depicts the J–V curves of FAMA-PSCs for the control device as well as at 60°C operating temperature. One can observe a significant change in J–V curves of the device aged at 60°C. The control device has an open-circuit voltage (V_OC) of 0.96 V and a short-circuit current density (Jsc) of 24.50 mA/cm². After exposure to 60°C for 3 h, we recorded a reduction in V_OC and J_SC of 0.85 V and 8.20 mA/cm², respectively, showing a rapid degradation of the PSCs. In contrast, as shown in Fig. 5(d), the control for the CsFAMA-PSCs gave a Voc and Jsc of 1.05 V and 22.08 mA/cm², respectively. As the temperature of PSCs increases to 60°C for 3 h, we notice a reduction in short-circuit current density of 14.56 mA/cm² and open-circuit voltage of 0.93 V. Figure S1 presents more detailed information on the J–V curves for all investigated operating temperatures. Although PCE loss is expected with a reduction in J_SC and V_OC as the operating temperature increases, we observed a higher degradation rate for the FAMA device compared to the CsFAMA one. These reports suggest that more structural defects are formed in FAMA layers, which are more severe at 60°C operation temperature. A similar trend was observed in the earlier reports.⁵⁴ 

To better understand the evolution of device characteristics with an increase in the operating temperature effect, the data of device parameters were analyzed for both FAMA and CsFAMA PSCs. Figs. 6(a) and 6(b) and Figs. S2(a–d) summarize the Jsc, Voc, FF, and PCE of the FAMA and CsFAMA based PSCs. The PSC devices operated between RT (25°C) and 60°C temperatures. In Fig. 6(a), we observe that the FAMA-based PSCs showed fast degradation of averaged photovoltaic parameters at a decreased Voc of (0.96, 0.94, 0.87, and 0.85) V, Jsc of (24.5, 21.34, 18.37, and 8.2) mA/cm², FF of (74.12%, 72.01%, 49.87%, and 33.54%), and PCE of (17.43%, 14.49%, 7.97% and 2.97%), which were undertaken for an operating temperature of (25, 40, 50, and 60)°C, respectively. On the other hand, the CsFAMA-based PSCs [Fig. 6(b)] exhibited much slower degradation of averaged photovoltaic parameters at a decreased Voc of (1.05, 0.99, 0.96, and 0.93) V, Jsc of (22.08, 21.83, 17.81, and 14.56) mA/cm², and FF of (79.4%, 67.9%, 54.7%, and 46.2%), resulting in a decreased PCE of (18.4%, 14.67%, 9.28%, and 7.06%) at an operating temperature of (25, 40, 50, and 60)°C, respectively.

The decreased PCEs in both FAMA and CsFAMA-PSCs are related to Jsc and FF. As the temperature increased, the Jsc of the device contributed more severely to the decreased device efficiency, which is stronger in FAMA- PSCs. A similar trend was observed in the earlier reports.⁵³ We attribute the reduction in FF and J_SC at elevated temperature to the exciton trapping that occurs at the interfacial defects between the ETL/perovskite/HTL layers. The occurrence of interfacial delamination (Fig. 4) prevents the splitting of excitons and the collection of charge carriers.⁶³ 

To establish the real cause of the observed PCE and J_SC degradation in the PSC device, we probed the device’s charge transport and recombination process relative to the operating temperature by Electrochemical Impedance Spectroscopy (EIS). EIS has been extensively used to characterize different photovoltaic devices to provide useful information on the device working mechanism.⁶⁴ Figures 7(a) and 7(b) show the EIS spectra as a function of temperature for both PSC devices, represented in Nyquist. All the impedance measurements were carried out in the dark. In the Nyquist plots, the arc closest to the origin is associated with the higher frequency regime and is generally associated with the charge transport resistance.⁶⁵ The other low frequency regime is attributed to the impedance arising from charge accumulation at the interface or Warburg ion diffusion within the layered PSC structure.⁶⁶ 

We see an increase in the charge transport resistance with increasing device temperature [Figs. 7(a) and 7(b)], indicating weaker charge transfer but a decrease in recombination resistance. In FAMA-PSCs [Fig. 7(a)], the values of the series resistance (R_s) and charge transport resistance (R_ct) were obtained. R_s shows a slight change from 19.72 to 29.82 Ω for the control and 60°C operated device, respectively. A significant difference is found in R_ct, increasing from 4.92 to 14.01 kΩ. This suggested the recombination influence of the ionic transport or the polarization effect at the interface between the active layer and the adjacent layer.⁶⁷ The increment in R_s indicates internal deterioration of the layers and interfaces, which in this case contributes to the lowering of the fill factor (FF). These results support the fast reductions observed in all the solar cell parameters at higher operating temperatures. In the case of CsFAMA devices [Fig. 7(b)], R_s was determined to be 20.06 and 22.62 Ω for the control and the 60°C operation of PSC devices, while the R_ct values were increased from 8.58 to 12.01 kΩ, respectively. Thus, the increase in Rs and R_ct values was lower for the CsFAMA-based PSCs than for the FAMA-based PSC.

To further understand the degradation trend, we carried out the External Quantum Efficiency (EQE) measurements of the PSCs and compared the effects of thermal stress degradation for the devices. The EQE mainly depends on the efficiency of exciton diffusion, charge transfer, and charge collection efficiency at a given wavelength region.⁶⁸ Figures 7(c) and 7(d) compare the EQE’s evolution under the same thermal stress conditions of exposure for FAMA and CsFAMA PSCs. As can be seen for both types of devices, there is a decrease in EQE with increasing operating temperature, which is consistent with the observation of J_SC loss from J–V characterization. In FAMA devices, we observe a fast decrease in the EQE [Fig. 7(c)], with the maximum EQE value dropping from 92% for the control to 30% at 60°C thermal stress. By exploring the advantage of the possibility offered by EQE spectral information, we can gain insight into the behavior of the different regions of the cells. Considering the structures of devices that correspond to the ETL/perovskite/HTL interface, since the front/back of these devices is thermally stressed at different operating temperatures, it is reasonable to assume that the degradation should start from these layers. The faster EQE decay in the long-wavelength region indicates lower charge transfer/collection efficiency and shorter carrier diffusion length bulk of the photoactive layer and even in the interfacial region between the hole transport layer (HTL) and the perovskite, which is associated with the thermally enhanced charge trapping and recombination.⁶⁸ We also notice a rapid drop in the low range region for the FAMA-based PSCs at 60°C. It is assumed that the Spiro-OMeTAD has been adversely affected with some increase in defect.^48,69 Therefore, this suggests that the charge transfer properties between the SnO₂/perovskite/Spiro-OMeTAD interface becomes poor as the device operating temperature increases. In Fig. 7(d), we noticed a reduction in EQE at 60°C to almost 52% of its control for CsFAMA-based PSCs compared to the equivalent FAMA-based PSC. However, this suggested a less defect in the grain boundaries since it is in this area where moisture ingress is mostly accessible (see Fig. 1) before it extends to the interfaces. Similar patterns of degradation have been previously reported.⁷⁰ Therefore, the reduction in EQE is consistent with the optical loss effects and increase in charge carrier recombination as the temperature increases from RT to 60°C.⁷¹ 

To further describe the mechanisms of degradation, an SEM was used to characterize cross-sectional views of the layers and their interfaces. We compared the cross-sectional SEM images of FAMA and CsFAMA devices [Figs. 4(a1-a4) and 4(b1-b4)] under the same operating conditions at different temperatures from 40 to 60°C for 3 h. We observed uniform microstructures with no evidence of microvoids or cracks at the interface between the layers for both FAMA and CsFAMA control devices. However, as temperatures increased from 40 to 50°C in FAMA, evidence of multiple micro-voids was initiated and observed in the perovskite layer as well as the interface between the Spiro-OMeTAD (HTL)/perovskite and SnO₂ (ETL)/perovskite layers [Figs. 4(a2 and a3)]. The micro-voids merged to form bigger cracks as the size of the grain increases [Figs. 1(a1-a4)]. Interfacial cracks were observed both along and across the interface of the film [Fig. 4(a4)] at 60°C. This shows that the FAMA device undergoes fast degradation in ways that lead to the delamination of the perovskite.⁶⁹ Therefore, the occurrence of interfacial cracks is attributed to the thermal stress that is induced due to the thermal expansion mismatch between the layers in the PSCs. Such mismatch stresses can also be amplified by the crack-tip fields associated with the presence of interfacial cracks.^53,72,73 Therefore, the delamination of the perovskite films can limit charge transfer between the perovskite and the SnO₂ layer, resulting in an increase in resistance and recombination channels, which mainly led to decreased J_SC and PCEs during operation [Figs. 5(c), 5(d), and 7(a)–7(d)].

In the case of CsFAMA at an operating temperature of 50/60°C, the dynamic features of the void/crack were also observed and continue only along the interface SnO₂/perovskite at 60°C thermal stresses. This phenomenon is ascribed to the different degradation processes of CsFAMA compared to the FAMA-PSC. Unlike the FAMA-PSCs at 60°C, the CsFAMA PSCs were more stable to show relatively less PSC decomposition.

Degradation and failure mechanisms have been studied in perovskite solar cells with FAMA and CsFAMA cations perovskite active layers. We show that both types of PSCs exhibit significantly different PCEs over their range of operating temperatures. The PCEs of FAMA-PSCs undergo severe degradation down to 2.72%, while those of CsFAMA-PSCs reduce to a PCE of 7.2% after thermal stress at 60°C. In both cases, device degradation is initiated via the formation of microvoids and interfacial deterioration that results ultimately in interfacial crack formation and bulk decomposition across the interfaces of the SnO₂/perovskite/Spiro-OmeTAD layers for FAMA devices at elevated temperature. However, the CsFAMA devices deteriorate with the formation of voids, and the incidence of microvoids later increases at the SnO₂/perovskite interface after thermal stress at 60°C. These changes are associated with increases in the series resistance and the charge transfer resistance values, along with rapid drops in EQE. Furthermore, FAMA devices develop much wider interfacial defect layers than CsFAMA devices. Thus, these findings pave the way for understanding the thermally induced degradation of PSCs and suggest the enhancement of thermal stability. The obtained SEM results show that PSC degradation can be dramatically suppressed by engineering the interface and encapsulating their films as a blocking layer, which will allow them to improve their operation lifetimes at higher temperatures.

The supplementary material contains the J–V characteristics of PSCs at different operating temperatures for 3 h and variation in peak intensity, peak FWHM, peak area, and peak ratio of (110) and (101) peaks from the XRD plot.

This research was supported by the grants from the Pan African Material Institute (PAMI) of the African Centers of Excellence Program (Grant No. P126974) and Worcester Polytechnic Institute. The authors acknowledge the African University of Science and Technology (AUST) for financial support.

The authors have no conflicts to disclose.

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