Radiation tolerance and self-healing in triple halide perovskite solar cells

Recently, the metal halide perovskites have received considerable attention as a low-cost terrestrial photovoltaic (PV) technology. Remarkably, these systems display several combined unique properties that may allow this technology to be commercially viable as a space PV technology first.^1–6 While space requires demanding performance in environments with radiation and thermal stress, there is a reduction in other stressors and constraints, such as moisture, humidity, 25-year operation, and total system costs, which currently provides a considerable bottleneck to the terrestrial implementation of perovskite solar cells. With the proliferation of satellite installations⁶ and the increasing interest in space exploration from both the Federal and commercial space community, advanced photovoltaic technologies are being sought to provide power for various future space missions.

Among various photovoltaic solar cells, the specific power of perovskite solar cells (PSCs) is predicted to be higher than those of current technologies available for commercial space solar cells.^7–9 PSCs have made considerable progress in the past decade, becoming a strong candidate for space applications due to the improved efficiency and impressive radiation stability of these systems.^3,4,10,11 Recent studies assessing the radiation tolerance of perovskite solar cells have considered the nature of proton interaction with the relatively thin absorber in these devices, establishing high energy particles in the excess of 1000 keV to have comparatively little parasitic effects upon the system.^3,4,12–14 Indeed, terrestrial space testing indicates perovskite solar cells can withstand irradiation fluences in excess of 10¹² p/cm²,¹² while perovskite thin films have withstood fluences in excess of 10¹⁵ p/cm² levels.³ These levels of irradiation are well above the tolerance that can be withstood by conventional III-V space solar cells without encapsulation, which require additional cover glass^15–18 to remain effective.

Previous works have developed testing protocols for perovskite solar cells in space and demonstrated the importance of low energy irradiation and elemental analysis for such missions.⁶ Here, the radiation tolerance of perovskite solar cells is assessed at several energies and fluences of proton irradiation to quantify further the specific role of ionizing and non-ionizing energy loss in the degradation of these systems. Indeed, a significant innovation of this work is the demonstration that perovskite solar cells self-heal after radiation induced degradation in dark cycles equivalent to such orbits in space.

The novelty here is that the experiments were designed such that fluence and the energy of proton irradiations were varied to induce the same density of vacancies in the absorber layer due to both non-ionizing nuclear energy loss (predominant at <300 keV) and electron ionization loss (predominant at >300 keV). While proton irradiation of the solar cells initially inflicts damage to the photovoltaic parameters, self-healing phenomena is mapped for these parameters vs energy and fluence of irradiation leading to qualitative tolerance threshold levels for unencapsulated cells.

In this study, the tolerance boundaries of perovskite solar cells are challenged by high irradiation levels. Irradiation fluences up to 4 × 10¹⁴ p/cm² are performed at a range of energies to assess the effects of damage and vacancy generation in the devices, as well as determine the sources of degradation in these systems. Devices based on transparent front and back surfaces are used to independently assess the perovskite absorber optically while correlating this to the PV performance of the solar cells at various radiation levels.

Figure 1(a) shows the layout of the device (including 6 solar cells on it and the pattern of the contacts) under illumination with high energy protons. A stopping and range of ions in matter (SRIM) simulation of the trajectory of the 75 keV protons incident on the perovskite cell is shown in Fig. 1(b). SRIM simulations mimic how protons interact within the device stack to create high vacancy densities. This plot shows a uniform distribution of defects created throughout the whole perovskite layer.

A Keithley power source 2400 multimeter is used to apply discrete bias values sweeping in forward and reverse directions and simultaneously measuring the resultant current. Current density is calculated by dividing the current by the device area illuminated through an aperture. Solar cells were mounted in a cryostat connected to a Linkam LNP95 cooling system and a T95 temperature controller to perform temperature-dependent measurements from 77 to 300 K. A solar simulator, Newport Oriel Sol2A, was used to provide an AM0 solar spectrum. External quantum efficiency (EQE) measurements are performed with an Oriel Quartz Tungsten Halogen (QTH) lamp as the light source coupled to an Oriel Cornerstone 260 monochromator. This system also uses a filter wheel with long-pass filters to block second-order light, the measurements were performed in the pulsed regime with an optical chopper. Two lenses are used for collimating and focusing the light onto the sample and an Oriel Si detector is used to acquire the reference spectrum. The signal is collected using a Stanford Research Systems SR830 lock-in amplifier.

To perform photoluminescence (PL) experiments, the 442 nm line of a He–Cd laser was used as the coherent light source. Borosilicate lenses and Borofloat 33 flat mirrors are used to direct and focus the light beam onto the sample in a Janis SHI-4-5 closed-cycle cryostat (4.2–300 K). The light-induced photoemission from the sample was then collected in a Princeton Instruments Acton SP2500 spectrometer fitted with an air-cooled Si charge-coupled device (2-dimensional CCD array). The data acquisition software Winspec^TM is used to control and record the PL data. After mounting the cell into the cryostat, the vacuum condition is achieved using a turbomolecular pump lowering the pressure of the cryostat down to ∼5 × 10⁻⁵ Torr after which a helium closed cycle cold head cools the sample to 4.2 K. A heater element connected to a copper cold finger inside the cryostat is controlled by a temperature controller system, Lakeshore Model 335, to vary the temperature of the sample for temperature-dependent measurements.

For sample irradiation, the lower energy (75 keV) proton beams were extracted from a TiH solid cathode of a Source of Negative Ions by Cesium Sputtering (SNICS-II) associated with a 3 MV tandem Pelletron accelerator (NEC-9SDH-2). The higher energy (300 keV and 1 MeV) proton beams were produced using a single-ended pelletron accelerator (NEC-9SDH) with a RF ion source. The momentum analyzed proton beams were electrostatically raster scanned across the sample enabling uniform irradiation. The targets were mounted in a vacuum chamber and pumped down to 7 × 10⁻⁷ mbar.^14,19 Before the ion irradiation experiment, the interactions of the energetic proton beam with the target layers were simulated using the Stopping and Range of Ions in Matter (SRIM) code. The SRIM code has been developed based on the binary collision approximation utilizing Monte Carlo random number generator simulations.²⁰ 

The SRIM simulations were used to trace the trajectories, distribution, and energy loss mechanism of the incident protons in the target layer under investigation and to estimate the amount of damage these particles produced in the target layers. In order to estimate the total damages (vacancies and displaced atoms), “Full cascade” calculations were performed for 100 000 proton ions with each energy (75 keV, 300 keV, and 1 MeV) and scaled up to the desired fluence. The device layer details going from back (irradiated side) toward the front (glass)—Al₂O₃: 50 nm, ρ: 3.95 g/cm³; ITO: 300 nm, ρ:7.20 g/cm³; SnO₂: 20 nm, ρ: 6.95 g/cm³; C60: 30 nm, ρ: 1.65 g/cm³; LiF: 1 nm, ρ:2.64 g/cm³; perovskite (FA_0.8Cs_0.2PbI_2.4Br_0.6): 200 nm, ρ: 4.173 g/cm³; PFN-Br: 1 nm, ρ: 1.5 g/cm³; poly-TDP: 5 nm, ρ: 1.15 g/cm³; ITO: 100 nm, ρ:7.2 g/cm³; glass: 90 nm, ρ:2.53 g/cm³. The glass is nominally 1 mm thick, but was kept to only 90 nm to speed calculations. In the SRIM calculation, the displacement energy is the minimum energy needed to displace an atom to create a vacancy. A displacement energy of 25 eV was considered to estimate the damages for the metal target atoms (Li, Al, Na, In, Sn, Cs, Pb) as well as for the halide atoms (F, Br, I). For H (a much lighter element), the displacement energy was 10 eV, and a displacement energy of 28 eV was used for O, C, and N.^21,22

Figure 2(a) shows the schematic structure of the solar cell. The perovskite solar cells investigated in this study are constructed utilizing a spin-coated FA_0.8Cs_0.2Pb_1.02I_2.4Br_0.6Cl_0.02 (∼1.7 eV band gap, thickness = 200 nm) as an absorber layer. The absorber layer is integrated upon an ITO front contact, a poly-TPD hole transport layer, with a PFN-Br (poly[(9,9‐bis(3′-((N,N-dimethyl)-N-ethylammonium)-propyl)-2,7-fluorene)-alt-2,7-(9,9-dioctylfluorene)]dibromide) interface layer to improve carrier extraction and interface passivation. The back of the cell architecture consists of a SnO₂/C₆₀ electron transport layer separated from the back ITO contact by a thin LiF interfacial layer. A 50 nm multilayer TiO₂/Al₂O₃ nanolaminate deposited by ALD used as an encapsulant layer completes the device structure.^1,23–25 The double transparent contact configuration enables this study to choose the desired side for illumination and particle irradiation. Here, proton irradiation enters through the back [in Fig. 1(a)] Al₂O₃-based nanolaminate.

The perovskite absorber used here is thinner than the optimal value (for a standard solar cell), allowing for some transmission of the light. This way the effect of the induced defects due to various energetic ion irradiations can be studied through absorption/transmission measurements, in addition to providing a more systematic device analysis. The PCE of the solar cells is 11% for AM 1.5G spectrum. The PCE is low due to the non-optimized thickness of the absorber layer and the higher than optimum band gap (1.7 eV) of the perovskite materials, which is designed to be the top junction for a tandem solar cell.

Figure 2(b) shows the light and dark J-V results taken both in forward and reverse directions illuminated through the glass substrate. The forward and reverse curves are identical, and the absence of hysteresis reflects the quality of the solar cells and the homogeneity of the perovskite lattice.^26,27 In Fig. 2(b), the extracted J_SC is ∼12 mA/cm² reflecting the reduced thickness of the absorber layer and the transparent back contacts (which do not allow for current generation in a second pass of light, as per normal). The inset in Fig. 2(b) shows the perovskite solar cell device that contains 6 pixels on each substrate. The illuminated area of each pixel is defined via apertures positioned on the pixels with an active area of 0.061 cm².

Four samples were sent for proton beam irradiation to the ion beam laboratory at the University of North Texas (UNT) to evaluate their radiation tolerance. Three of these samples were proton irradiated with the following energies and fluences: (1) 75 keV and 10¹³ p/cm² (a flux of 3.3 × 10¹⁰ p/cm²-s); (2) 300 keV and 10¹⁴ p/cm² (a flux of 7.8 × 10¹⁰ p/cm²-s); and (3) 1 MeV and 4 × 10¹⁴ p/cm² (a flux of 7.8 × 10¹⁰ p/cm²-s). The fourth sample was used as a traveler device to monitor the changes in the cell’s performance due to transit to and from the characterization (University of Oklahoma) and irradiation facilities (UNT) located about 150 miles apart. The reason for choosing higher fluences for higher energy protons is to induce the same number of vacancies due to predominant nuclear interactions in the absorber layer as low energy protons create higher densities of vacancies in the absorber layer. By doing so, the effects due to ionizing energy loss (IEL) and elastic nuclear (non-ionizing energy loss: NIEL) mechanisms can be more effectively deconvolved.

Previous studies have shown that proton energies in the window of 50–1000 keV have a strong interaction by stopping within the perovskite solar cells, while the protons with energies higher than 1 MeV penetrate through the cell with much fewer collisions with the target nuclei,^6,14 although these energy values depend on the structure and thickness of layers that high-energy protons encounter. Here, to induce defects in the system and study the damage mechanisms specifically, high radiation fluences (above 10¹² p/cm²) are chosen with varied energies to determine the role and effects of nuclear displacement and/or electron ionization in PSCs under conditions that induce defects through these two different mechanisms, while inducing the same total defect concentration in the absorber layer. The total number of vacancies in the absorber layer for the 75 keV, 300 keV, and 1 MeV irradiated cells as determined by full damage cascades simulations using SRIM are 1.95 × 10¹², 1.66 × 10¹², and 1.82 × 10¹², respectively (see Fig. 3). This results in equivalent defect concentrations despite their generation via different mechanisms (NEIL and IEL).

The 75 keV protons lose their energy before reaching the end of the cell, stopping in the absorber [Fig. 3(a)], while the high energy protons have a larger projected range and pass through the solar cell depositing a predominant amount of their energy close to the glass substrate while inducing IEL and local heating throughout their trajectory. For all the energies simulated, the ITO layers are the most affected by irradiation forming high numbers of In and O vacancies. Within the perovskite material itself, in the case of the 75 keV irradiation H atoms constitute the highest number of vacancies followed by the halide-iodine atoms. At higher energy, for both 300 keV and 1 MeV, iodine (I) vacancies constitute the largest vacancies generated, followed by lead (Pb) [see Figs. 3(d)–3(f)].

The prevalence of H vacancies under 75 keV irradiation reveals that direct collisions (NIEL) with the perovskite induces significant damage to the organic molecules. The high density of H vacancies at lower energies has been seen previously,⁶ this is attributed to the abundance of hydrogen (H) in these lattices, its low displacement energy, and the fact that H has a mass nearly identical to that of a proton. In Ref. 6, the systematic increase in the energy of protons was also observed to increase Pb–I vacancies as the protons increasingly interact with the metal halide framework—a hypothesis supported by the analysis here (Fig. 3). This behavior is attributed to the fact that higher energy protons are much more energetic than the H atoms and therefore have a reduced probability of interaction with lower mass atoms in the absorber layer.⁶ 

Based on SRIM results [Figs. 3(d)–3(f)], it is expected that the 300 keV and 1 MeV irradiated cells would be quite similar. On close inspection, the effect of irradiation upon the sample indicates different mechanisms are in play with increasing energy, which might be expected since the effects of NIEL and IEL change as the energy is increased. [Stopping and Range of Ions in Matter (SRIM) calculated electronic energy loss (electronic ionization) (a) and nuclear (recoil) energy loss (b) for 75 keV, 300 keV, and 1 MeV protons are presented in supplementary material (1).]

In order to diagnose any changes in the samples upon irradiation, spectroscopic measurements on the traveler and irradiated samples (Fig. 4) were also performed. These data are presented at 4.2 K to increase the resolution of the well-defined excitonic complex, as well as the increased potential to observe defect complexes in the absorption prior to thermal broadening and exciton ionization (temperature-dependent absorption and PL graphs of all the cells are presented in Fig. SI-2). Figure 4(a) compares the normalized photoluminescence of the cells at 4 K, while (b) shows the normalized absorption of the cells at the same temperature (4 K). Both PL and absorption spectra show a slight red shift upon irradiation (∼7 meV). This red shift indicates that the displaced atoms have affected the band gap of the material slightly, either via local relaxation of strain or subtle decomposition of the perovskite.

Figures 4(a) and 4(b) show that the red shift is slightly larger for the solar cell irradiated at 300 keV as compared to the solar cells irradiated at both higher and lower fluences; the red shift for the lower 75 keV exposure and higher 1 MeV irradiated cells are equivalent. The larger red shift for the 300 keV irradiation can stem from following causes: (1) for the higher energy irradiation, the SRIM simulations of the trajectory of the protons inside the solar cells indicates more collision effects in 300 keV due to the more lateral scattering of the particles as compared to the 1 MeV case (see Fig. 5). Compared to the lower energy irradiation, the fluence of 300 keV irradiation is ten times higher than the 75 keV case.

With increasing temperature, three main effects are observed in the PL spectrum of the solar cell: (1) the emission broadens and (2) the red shift observed in the irradiated solar cells fades gradually as temperature is increased up to RT. At room temperature, the irradiated solar cells all have identical band gaps (see supplementary material, Fig. SI-3). Finally, (3) the band gap energy shifts toward higher energies consistent with typical perovskite temperature dependence. These effects are likely due to ionization of defects and/or the convolutions of such states with temperature but the exact origin requires further spectroscopic study.

The integrated intensity reduces systematically with the increase in the fluence of the proton irradiation as shown in supplementary material, Fig. SI-4. That all three irradiated samples have weaker intensities at low temperatures is the most compelling spectroscopic evidence of the short term deleterious effects from the irradiation. Based solely on the room temperature spectroscopic measurements (PL and absorption), the integrity of the perovskite film appears to be preserved.

However, the loss in PV performance is seen more clearly in electrical characterization measurements most significantly J_sc and FF [see Figs. 5(b) and 5(c), black symbols], which is associated with short term (initial measurements black symbols) vacancies and diminishment of carrier lifetime/diffusion length due to the NIEL and IEL from the irradiation. One cannot rule out the possible degradation of transporting layers and/or interfaces within the device.

Figure 5 shows the remaining factor (RF, final value/initial value) of the PV parameters of the solar cells with increasing proton energy/fluence (the J-V data are given in SI-5). The post-irradiation data consist of two sets of measurements: (1) J-V parameters immediately after receiving the samples upon irradiation (black squares), and (2) the same parameters after two months following irradiation (red triangles). As it is seen in Fig. 5, the PV shows degradation upon receipt from the irradiation facility (black symbols), and these effects increase with increasing energy and fluence. Considering the results shown in Figs. 3 and 5, it is evident that immediately after irradiation the J-V data show more degradation for the higher energy/fluence proton irradiated solar cells, which SRIM analysis indicates results in significant I- and Pb-type vacancies.

Although the total number of vacancies is higher for the 75 keV irradiated cell, the majority of these vacancies are due to H displacement. Since the organic molecules do not have a dramatic effect upon the band gap or optical properties of the perovskite absorber,²⁸ this solar cell consequently shows the lowest (little) degradation in its PV performance. The J_SC of the solar cell exposed to 300 keV irradiation is the lowest (9.6 mA/cm²) of the devices assessed under different irradiation energies, while (perhaps surprisingly) its fill factor (RF ∼ 0.95) is better than that of the 75 keV cell (RF ∼ 0.92).

These properties can once again speak to the origin of the defects produced under proton irradiation and the subtle details related to the effects of the NIEL and IEL processes, in addition to the ability of perovskites to self-heal with increasing energy and local heating. This self-healing has been observed by several authors,^14,29,30 resulting in improved performance in these systems under typical space testing protocols.^31,32 This has also led to the development of space-specific protocols for perovskites with regard to space power applications.⁶ 

Notably, Fig. 5 also shows that these systems tend to self-heal and return to their original levels of performance when simply stored under ambient conditions in the dark. This is evident in Fig. 5 for each of the PV parameters shown after two months of storage (solid red triangles). These data show that the solar cells return to nearly their original performance levels in all cases, except those devices irradiated at 1 MeV: 4 × 10¹⁴ p/cm², which completely lost their performance and were irreversibly damaged. This indicates that despite their apparent superior radiation tolerance, there remains a radiation threshold for which irreversible damage results. This suggests perovskite solar cells still require the development of radiation-hard encapsulation systems for long and/or hostile missions in space.³³ 

Upon closer inspection, it can be seen that initial irradiation degradation of solar cell primarily impacts J_sc and efficiency with minimal effect on the V_OC – Fig. 5. This suggests that the perovskite lattice or structure is not prohibitively affected by irradiation (large decomposition, segregation, and/or changes to the band gap) but that the vacancies induced by proton exposure result in inhibited carrier extraction and/or transport presumably via the formation of shallow traps or defect centers that localize carriers reducing their diffusion length and mobility. While the carrier diffusion length for the perovskites has been suggested to be as much as 1 µm,⁴ much larger than the typical thickness of the absorber layer perovskite solar cells (here, 200 nm) defects—even shallow complexes—can result in significant localization and recombination losses. Here, since V_oc is relatively unaffected by irradiation (perhaps inducing some pinning of the fermi-level at the interface)—these devices would appear to be dominated by radiative rather than non-radiative processes that are less prohibitive than defects typically induced in conventional (for example) III-V systems used in space. This further demonstrates the novelty of perovskites and their potential for space applications.³⁴ 

The unique ability for perovskites to self-heal and recover their performance over time is a consequence of the soft and dynamic nature of the perovskite lattice; when defects are formed, they are both shallow and unstable, allowing the perovskite to return to its original state with minimal degradation.^35,36 This is consistent with the data presented in Figs. 4 and 5 suggesting that the effects of defects induced in space environments may be limited and minimized (possibly removed) as the temperature increases or as the solar cell recovers during dark periods in their orbit in space. Albeit, provided the threshold for irreversible degradation discussed above is not exceeded. However, this threshold (1 MeV, 10¹⁴ p/cm²) is equivalent to 1 year of exposure at Jupiter⁶ for an unencapsulated solar cell (i.e., much more than experienced in most Earth orbits over a reasonable satellite lifetime). This could be circumvented by novel encapsulation systems currently being developed for the perovskites.

In conclusion, several triple halide perovskite solar cells have been exposed to high radiation dosages under space relevant conditions, including a range of proton energies (75 keV to 1 MeV) and high irradiation levels (>10¹⁴ p/cm²). While proton irradiation is shown to lead to a reduction in the performance of the solar cell immediately after irradiation, the performance of the solar cells also recovers ∼2 months after radiation exposure in dark ambient conditions—the system self-heals. While this work confirms the high radiation tolerance of perovskites, a threshold for irreversible damage is observed for 1 MeV protons at fluences of 4 × 10¹⁴ p/cm², which represents exposure for ∼40 000 years on the ISS orbit, or ∼40 years at Jupiter (on the JUNO orbit).

Included in the supplementary material are descriptions of the chemicals and fabrication steps to process the solar cell devices. Also included is SRIM analysis showing the relative IEL and NIEL contributions at various proton energies and the density of vacancies induced via these processes. Additional experimental data, including temperature dependent photoluminescence and absorption spectra of the perovskite absorber, are also presented, in addition to current density–voltage data under 1-sun illumination before and after irradiation, and after self-healing in dark ambient conditions

The authors would like to acknowledge Axel Palmstrom for the ALD of the TiO₂/Al₂O₃ nanolaminate layer at NREL. The OU group acknowledges funding support from the National Aeronautics and Space Administration (NASA) under Agreement No. 80NSSC19M0140 issued through NASA Oklahoma EPSCoR and the Center for Quantum Research and Technology (CQRT) at the University of Oklahoma. U.N.T. acknowledges partial support from NSF (Grant No. HBCU-EiR-2101181).

Giles Eperon works for Swift Solar. That is disclosed on the supplementary material.

Hadi Afshari: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Writing – review & editing (equal). Sergio A. Chacon: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal). Shashi Sourabh: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal). Todd A. Byers: Data curation (equal); Investigation (equal). Vincent R. Whiteside: Conceptualization (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Rose Crawford: Data curation (equal); Formal analysis (equal). Bibhudutta Rout: Conceptualization (equal); Data curation (equal); Investigation (equal); Supervision (equal); Writing – review & editing (equal). Giles E. Eperon: Conceptualization (equal); Formal analysis (equal); Resources (equal). Ian R. Sellers: Conceptualization (equal); Methodology (equal); Project administration (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.