Effects of Light and Electron Beam Irradiation on Halide Perovskites

Jan 25, 2016 - These perovskites do not contain any rare element, the amount of toxic lead .... images, which were taken at the first (pristine surfac...
1 downloads 0 Views 6MB Size
Download PDF
Article pubs.acs.org/accounts

Effects of Light and Electron Beam Irradiation on Halide Perovskites and Their Solar Cells Published as part of the Accounts of Chemical Research special issue “Lead Halide Perovskites for Solar Energy Conversion”. Nir Klein-Kedem, David Cahen, and Gary Hodes* Department of Materials and Interfaces, Weizmann Institute of Science, Rehovot 76100, Israel CONSPECTUS: Hybrid alkylammonium lead halide perovskite solar cells have, in a very few years of research, exceeded a light-to-electricity conversion efficiency of 20%, not far behind crystalline silicon cells. These perovskites do not contain any rare element, the amount of toxic lead used is very small, and the cells can be made with a low energy input. They therefore already conform to two of the three requirements for viable, commercial solar cells−efficient and cheap. The potential deal-breaker is their long-term stability. While reasonable short-term (hours) and even medium term (months) stability has been demonstrated, there is concern whether they will be stable for the two decades or more expected from commercial cells in view of the intrinsically unstable nature of these materials. In particular, they have a tendency to be sensitive to various types of irradiation, including sunlight, under certain conditions. This Account focuses on the effect of irradiation on the hybrid (and to a small degree, all-inorganic) lead halide perovskites and their solar cells. It is split up into two main sections. First, we look at the effect of electron beams on the materials. This is important, since such beams are used for characterization of both the perovskites themselves and cells made from them (electron microscopy for morphological and compositional characterization; electron beam-induced current to study cell operation mechanism; cathodoluminescence for charge carrier recombination studies). Since the perovskites are sensitive to electron beam irradiation, it is important to minimize beam damage to draw valid conclusions from such measurements. The second section treats the effect of visible and solar UV irradiation on the perovskites and their cells. As we show, there are many such effects. However, those affecting the perovskite directly need not necessarily always be detrimental to the cells, while those affecting the solar cells, which are composed of several other phases as well as the perovskite light absorber, are not always due to the perovskite itself. While we cannot yet say whether perovskite solar cells will or will not be stable over the long-term, the information in this Account should be a useful source to help achieve this goal.

■

at an appreciable rate at ∼200 °C, and is lost at a comparable rate from MAPbI3 at only slightly higher temperature (∼225 °C).1 This translates into limited temperature stability for the perovskite semiconductors containing these organic species, with some dependence on the specific alkyl group, with MA and FA being by far the two most common ones. FAPbI3 was reported to be substantially more thermally stable than MAPbI3 with little visible change after 60 min at 150 °C in the former, compared to complete change to yellow PbI2 of the latter in the same time.2 A related property is the very low formation energy of the perovskite from PbX2 and AX (∼0.1 eV for MAPbI3 from theoretical calculations.3 We are not aware of experimental measurements of this parameter although a low value could be inferred from its low decomposition temperature). Additionally, it is generally accepted that ion movement in the perovskite occurs in these cells,4−6 even if there is still a question of just what is/are the moving species, in part because of lack of direct (e.g., isotope tracer) evidence. All these factors lead to the question: Will these semiconductors be sufficient stable for

INTRODUCTION Lead halide perovskite photovoltaic cells have not only set records for increased photovoltaic performance with time, but also catalyzed a quest for similar semiconductors that can be prepared by low energy methods, yet have excellent charge transport and optical properties. Most studies into these photovoltaic cells involve so-called hybrid perovskite semiconductors, where “hybrid” refers to the combination of organic and inorganic constituents. The organic part here is an alkyl ammonium cation (commonly CH3NH3+, abbreviated to MA+). This cation is the A in ABX3 perovskites, where B is normally Pb (can also be Sn or Ge) and X is a halogen. The A cation must be large enough to form the 3D perovskite structure; however, if too large, a 2D (or mixed 2D−3D) structure forms instead. Thus, the choice of suitable cations is very limited (at least, at present): to date, only MA+, FA+ (formamidinium), and, to some extent, Cs+ fulfill the size demands for a fully 3D perovskite structure for the most commonly investigated material, MAPbI3. The alkyl ammonium species have limited temperature stability and readily dissociate and evaporate. For example, thermogravimetric analysis shows that MAI starts evaporating © XXXX American Chemical Society

Received: October 16, 2015

A

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

Figure 1. Sample degradation caused by e-beam. Cell structure: FTO glass/dense TIO2/MAPbI3(Cl); “(Cl)” shows PbCl2 was used in the preparation, although little Cl remains in the product)/spiro/gold. Sequential EBIC scans of the same region showing the decrease in EBIC contrast due to beam damage. (Left) Top, SE image; remaining panels, sequential EBIC scans. (Right) Line scans, corresponding to panels 2−8. Note that the “missing” second EBIC scan was used for the SE image. E-beam current (1 nm beam spot) 110pA; beam voltage, 1.5 kV. Adapted with permission from ref 9. Copyright [2014] [Nature Publishing Group].

then either are collected at the two electrodes, giving a current that is measured (the e-beam-induced current) or recombine. There are two fundamental differences between EBIC and normal photovoltaic behavior: each electron in the e-beam excites many electron−hole pairs and the higher spatial resolution of the e-beam compared to a light beam allows mapping the charge collection across the cell thickness (typically micrometer or, for perovskite cells, submicrometer size). It is this latter property that allowed us to demonstrate that MAPbI3 cells with spiro-OMeTAD as hole transport medium (HTM) behave as p−i−n devices and, by measuring electron and hole diffusion lengths, explain differences between MAPbI3 made with and without PbCl2 as precursor, as well as showing that HTM-free cells behave as a heterojunction.9 We also showed that MAPbBr3 cells with HTM behaved, as HTMfree iodide cells, like a p−n heterojunction (probably due to the much higher doping of the bromide than the iodide perovskite), as well as having charge diffusion lengths that increase markedly upon illumination with monochromatic (406 nm) light, compared to what we measured under only e-beam exposure.10 In this Account, we concentrate on the effects of the e-beam on the perovskite cells. The energy of an e-beam is much higher than that of a solar photon (keVs vs few eV). Because highenergy electrons will create many electron−hole pairs, this increases their ability to damage the perovskite, known to be relatively thermally unstable. We have noted9 that the e-beam (1.5 keV) damages the (MAPbI3) cell over time, with clear deterioration of the EBIC intensity after just a few scans (Figure 1). Therefore, and based on our experience using EBIC for perovskites, we developed a procedure to minimize this

long-term use (for photovoltaic or other applications)? At the same time and from a more positive perspective, these factors, and ion movement in particular, could potentially be beneficial in that, if ions can move back and forward in such a way so as to end up (in the operating photovoltaic cell) in the most beneficial position, this could result in self-healing of “defects” (translated loosely here as ions in any positions/electronic states that do not give the highest efficiency). CuInSe2-based photovoltaic cells are an example of such beneficial ion movement7 and CdS/CdTe cells may also blessed with this property. In this Account, we focus on issues related to changes due to irradiation, by light or electron beams, that occur in perovskite semiconductors and photovoltaic cells made from them, during the actual characterization measurement of material/cell properties or during cell operation. While we do not treat solar cell stability directly, this issue does arise in relation to light-induced effects, in particular in the last two sections.

■

ELECTRON BEAM EFFECTS ON PEROVSKITES AND PEROVSKITE CELLS We used electron beam-induced current (EBIC) to learn about the mechanism of perovskite cell action.8−10 In EBIC, the electron beam (e-beam) in a scanning electron microscope is used to excite electron−hole pairs in a complete photovoltaic cell, either in planar configuration (scanning the e-beam over the cell surface) or, as used by us to study the perovskite cell’s mechanism of operation, in cross-section configuration, scanning the e-beam across the cell cross-section (obtained by cleaving a cell). The e-beam-excited electron−hole pairs B

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

Figure 2. Secondary electron and EBIC cross-section images of a cleaved FTO/dense TiO2/MAPbI3(Cl)/spiro-MeOTAD/Au cell (top) and corresponding line profiles taken from 12 sequential scans of the same area (bottom). The EBIC signal starts to decay, after increasing during the first 5−6 scans. After that signal decay occurs, increasing with increasing dose, which may suggest a quasi-neutral region close to the middle of the film. Beam current, ∼5pA (∼2.5 nm beam size); beam voltage, 3 kV.

12 (right). Initially, with the increase of dose up to 5−6 scans, a higher EBIC signal is seen over the entire film thickness. One possible reason for this EBIC signal increase is passivation of surface states, created by cleaving the sample (resulting in breaking bonds) and possibly also air exposure prior to introduction into the scanning electron microscope. The e-beam can then affect these states in two ways. One is (dis)charging of these surface states; increase in EBIC signal with beam dose is well-known for Si-based devices, where (dis)charging of surface states by beam electrons results in surface passivation and apparent increase in collection efficiency.11 The other is thermal annealing of the surface states by the e-beam. Even in the absence of surface states, thermal annealing may initially improve performance, in analogy to improvement due to light soaking of perovskite cells. Planar EBIC scans of uncleaved samples may help to clarify the issue. The effect of damage is seen at doses starting from scan 6. The signal decay exhibits two main characteristics: reduced overall signal and change in profile shape. In scans 6−12, a clear valley develops in the EBIC profile with dose. The position of the valley close to the center of the film may suggest lower drift/diffusion lengths for both carrier types. By normalizing the profile curves and taking the low point of the valley in scans 5−12 the dose effect on the EBIC signal decay can be obtained. This decay should be indicative of the sensitivity of the material to beam damage. Such analysis is presented in Figure 3; a monoexponential decay of 1.5 × 1017 e/cm2 is obtained, an order of magnitude higher than the electron dose; thus, the imaging parameters that were used are suitable for this material.

deterioration. In all images, used in our recent studies, microscope calibration and focusing is done on a (sacrificial) region of the sample, different from that used for the EBIC scan. Then, the focus is moved to a fresh spot and the EBIC image taken upon the first exposure. The image is taken using pixel averaging, and the pixel integration time kept to a minimum, while avoiding a loss in signal-to-noise ratio. Because conclusions drawn from EM imaging the same spot repeatedly are highly suspect, best practice is to determine trends from statistics gathered from many spots imaged only once. Before loading the cleaved samples into the scanning electron microscope vacuum chamber, we measured directly their photovoltaic response to confirm they function properly. Also, due to improvements in instrumentation over time, we could reduce the e-beam current, from 110pA used initially to ∼5pA nowadays, thus further reducing beam damage. This reduction in beam damage reveals a trend, hidden in the earlier studies, namely an initial increase in EBIC signal with ebeam irradiation. In a series of images presented below for an FTO/TiO2/MAPbI3(Cl)/spiro-MeOTAD/Au device (Figure 2), sequential and repetitive imaging of a selected spot is presented. For all images, the beam current was 4−5 pA and pixel integration time was 60 μs, which corresponds to a dose of 1.5−4.5 × 1016 e/cm2 per imaging scan. Results of 12 sequential, repetitive scans are presented. A secondary electron (SE) image of the device cross-section is presented (top, left) beside the EBIC images, which were taken at the first (pristine surface), fifth and twelfth scans. The EBIC signal is presented as recorded, so that differences in brightness represent differences in EBIC signal. Line profiles from all scans are presented below the images for scans 1−6 (left) and scans 5− C

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

characteristic of a p−n junction,10 can be used to measure the material’s electron diffusion length. In addition to an overall decay in EBIC signal with increased beam exposure, the diffusion length decreases from 190 ± 5 to 125 ± 5 nm. In both the iodide and bromide case, built-in field-assisted charge separation and collection accounts for the slower signal decay. In the quasi-neutral regions, where collection is diffusionlimited, the signal is more affected by damage than in regions with built-in field, as is evident in the bromide cell. This may suggest that the valley in the EBIC profile in the iodide cell at scans 6−12 highlights an area of zero field, in agreement with the model proposed in ref 9. This relatively narrow quasineutral region (in contrast to the much wider regions in ref 9) in Figure 2 that develops for longer-exposure times may result from beam damage in previous scans. The beam-induced changes in the EBIC signals seem to be irreversible with various combinations of time (up to 1 h), bias (up to ±1.5 V), and illumination at photon flux energies equivalent to >Eg ones in the solar spectrum, for MAPbBr3 and for MAPbI3, with 405 and 560 nm laser beams, respectively. We can make an order of magnitude calculation of the absorbed power of the e-beam used by us today, taking into account pixel size and charge generation volume inside the sample. Assuming all the absorbed power is converted to heat, we find a value of ∼109 W cm−3. This compares with the equivalent power absorbed from sunlight (absorption depth of ∼1 μm) of 103 W cm−3. In terms of energy/pixel, the difference will still be ∼200×, using an ∼0.2 ms/pixel dwell time of the ebeam with our scanning conditions, which gives a qualitative idea of the heating and possible damage to the perovskite in the vacuum of the scanning electron microscope. Our earlier experiments used more than 10× higher beam current (but half the beam voltage), which increases the total power, but also increases the charge generation volume, which acts in the opposite direction. While heating is not the only factor that could lead to cell degradation by the e-beam, it is the most obvious one. We found recently that the all-inorganic Cs bromide perovskite is much more stable to the scanning electron microscope e-beam than the MA one,12 which fits its much greater thermal stability12−14 (e.g., we normally anneal CsPbBr3 at 250 °C), which is probably related to lower volatility of CsBr than of MABr. In ref 12, we showed that while the EBIC signal from MAPbBr3-based devices decays fast under exposure to the e-beam, this is not so for CsPbBr3-based devices. To highlight the better stability of the Cs-based perovskite, we first imaged pristine films of both MAPbBr3 and CsPbBr3, grown using the same procedure.12 Both films were then exposed for 2 min to >103 higher beam current than that for imaging and again imaged at the same spot. The results (Figure 5) show that the MAPbBr3 morphology changed dramatically. The formerly crystalline material now looks amorphous without clear grain boundaries. We note that film decomposition cannot be excluded without further analysis. For CsPbBr3 little, if any high beam current effect is seen on film morphology. Even though the high beam current may have affected the electronic properties, these results show CsPbBr3 to be much more stable toward an e-beam than MAPbBr3. Recently, cathodoluminescence (CL), which involves e-beam excitation of a sample, usually in a scanning electron microsope, and measuring the resulting luminescence spectrum, was reported from various perovskite samples with emphasis on sample changes caused by the e-beam irradiation.15 In general, two permanent changes were observed: defect formation (seen

Figure 3. EBIC signal decay as a function of e-beam dose. The value of the signal is taken from the minimum of the valley in scans 5−12, Figure 2. A monoexponential decay fit was used to estimate the signal decay rate.

The same experiment carried out for the MAPbBr3 cell shows no initial EBIC signal increase but rather a decrease from the first scan (Figure 4). If, as hypothesized above for the iodide

Figure 4. Secondary electron and EBIC images of cross sections of a planar MAPbBr3-based device (top) and corresponding line profiles taken from six sequential scans in the same area (bottom). The EBIC signal decays already from scan 1. The electron diffusion length Ln decreases from 190 to 125 nm after four scans and then saturates. Beam conditions as in Figure 2.

cell, cleaving/humidity introduces surface states, this implies that no such states are formed in the bromide cell under similar conditions. We checked this by using an increased acceleration voltage that produces free carriers farther away from the surface. In the limit of resolution loss due to the higher interaction volume, no difference was found as carriers are produced deeper in the material, suggesting that for the cleaved bromide-based cell, surface states do not play a role. In the bromide cell, the exponential decay of the signal from the peak and away from the electron selective layer, a D

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

the ions ensure strong inelastic interaction. Damage includes ion beam sputtering, amorphization of a few nanometers of each side of the sample, and heat-induced damage. Importantly, the absence of the ions used (mostly Ga) in the FIB-treated sample cannot tell if damage was done. Finally, besides stressing that e-beam methods such as SEM and CL require great care to ensure that the intended, rather than a beam-damaged sample, is measured, the results of other forms of analysis in which high energy is carried into the material, either focused onto a small area or not, such as STEM, FIB, secondary ion mass spectroscopy (which may also be selective toward some elements), Rutherford backscattering spectroscopy, and so forth, should be considered with extreme caution.

■

EFFECT OF ILLUMINATION ON PHASE SEGREGATION IN MIXED HALIDE PEROVSKITES In principle, MAPbI3/MAPbBr3 and MAPbBr3/MAPbCl3 can form solid solutions over the entire composition range (but see below) while MAPbI3/MAPbCl3 are essentially immiscible. This means that mixed I/Br or Br/Cl can be used to tailor the band gaps between the two end members. At present, there is little interest in increasing the band gap of the bromide (2.3 eV) (although such materials might be of interest for applications, such as UV or high energy detectors). However, the mixed Br/I perovskite is of particular interest for tandem cells. It is recognized that it may be feasible to couple perovskite cells with commercial Si cells, using the perovskite cell as the high band gap one in a two-junction tandem. This combination could increase the cell efficiency considerably compared to single junction Si cell efficiency. For this purpose, while the ca. 1.6 eV band gap of MAPbI3 is already useful, a higher band gap (∼ 1.8 eV) is better. This can be obtained by forming an MAPb(Br,I)3 solid solution. Interestingly, while there are many reports on this mixed system, none has shown a higher VOC than good cells with with pure iodide perovskite. A recent study showed that the I/Br solid solution, while forming a band gap between the two end members, is not stable under illumination.16 For compositions where the Br is greater than 20% (of the total I + Br), an additional photoluminescence (PL) peak at ∼1.68 eV appears and increases under the PL excitation illumination, eventually dominating the PL spectrum. Also the position of this peak was independent of the starting composition (for >20% Br; for lower Br concentrations, it also appeared but needed more light soaking). Signs of this feature at ∼1.7 eV have been seen previously for the mixed I/Br perovskite but were not associated with illumination. Thus, Sadhanala et al. found, in both absorption and photoluminescence spectra, that they occurred at intermediate compositions (notably for [Br] > 20% of the total halide concentration) but disappeared after aging.17 We also observed this PL peak in our work with mixed Br/I lead perovskites (60% Br, 40% I).18 Importantly, Hoke et al. also found that the change on light soaking was reversible if the samples were kept in the dark for some minutes and that this reversibility was maintained over many on−off illumination cycles.16 The PL changes on illumination were also paralleled by appearance of a shoulder in the light absorption spectra at ∼1.7 eV. X-ray diffraction (XRD) showed that the original sharp peaks of the samples that were not light-soaked split into two peaks upon illumination (Figure 6). All these results were consistent with formation of a minority phase (with a crystal size, estimated from XRD, of

Figure 5. Secondary electron image of MAPbBr3 (top) and CsPbBr3 (bottom), taken with 3 keV beam energy, 5pA beam current, and 60 μs pixel integration time. “After” images (RIGHT) are after exposing the image area to a 8 nA beam with the same energy and pixel integration time for both materials.

as CL peak broadening and quenching) and phase transformations caused by e-beam heating and observed as peak shifts. Two perovskites, FAPbI3 and CsPbI3 in the metastable black perovskite form prepared by heating to 350 °C, were studied. While the MA and FA perovskites showed similar CL stability to the e-beam, the Cs perovskite was found to be more stable, in agreement with our results. Furthermore, for CsPbI3 phase conversion to the stable nonperovskite yellow form was initiated at the grain boundaries. Finally, single crystals of MAPbI3 were found to be more stable than films; this was attributed to a lower defect density in the former. While clearly e-beam damage of the perovskites in SEM (scanning electron microscopy) imaging must be taken into account and carefully minimized, it is worth comparing SEM with TEM (transmission electron microscopy), commonly used to investigate materials and interfaces (although little used for the perovskites, possibly because of expected or observed beam damage). Typical beam energy in TEM is 50−300 keV with typical currents of ∼1 nA. These high energies and currents, compared to SEM, do not, however, necessarily mean even higher beam damage. TEM samples must be thin enough to allow beam electrons to pass through. At such small thickness and high beam energy, most of the electrons interact elastically with the sample. However, inelastic interaction, which may damage the sample, should not be overlooked. Elemental analysis such as EDS and electron energy loss spectroscopy (EELS) rely on inelastic scattering. In these methods, the energy deposition rate/area may be extensive, especially when scanning TEM is used for high resolution of the elemental analysis. The cross section for inelastic scattering will be high for heavy elements like Pb and I. In short, imaging and elemental analysis may prove to cause less damage to the perovskites if done with TEM than with SEM. However, unless shown otherwise, use of these methods to study this material system entails beam damage analysis. Another issue with TEM imaging is sample preparation, especially using focused ion-beam (FIB) microscopy and milling. There, the relatively low energy and heavy nature of E

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

decreased by only 8% compared to the initial value after 63h (the fractional drop was a few times higher compared to the maximum JSC that was measured after some light soaking). The open circuit voltage (VOC) showed a greater drop −180 mV after this time. Importantly, these losses were similar to those measured under one sun for the same time. This suggests that the cells (and the perovskite itself) are not very sensitive to illumination when indirect effects are removed. It was noted in ref 19 that this stability was only found for the above cell configuration; cells with other HTMs, with mp-Al2O3 or planar cells were all much less stable, even in the dark, suggesting that the cause of instability in these cells was not due to the perovskite itself but to either changes in other phases or in the interface between the perovskite and other phases. Another study investigated the perovskite materials (i.e., not cells) using a 100× concentrated solar spectrum.20 The samples were encapsulated and temperature controlled by a thermoelectric cooler. Both MAPbI3 and MAPbBr3 were studied. The MAPbI3 was found to be stable to 100 suns at ∼25 °C, but suffered slow photobleaching at ∼50 °C while the MAPbBr3 was found to be stable even at the higher temperature. Two hypotheses were put forward for the higher photostability of the bromide: the stronger bonds between Br and Pb and/or H (the latter from the MA) and different crystal structure of the two halide perovskites. A simple comparison for these two studies (for MAPbI3) suggests a much greater stability for the perovskite in the former (which covered a much longer time frame) than for the latter. The higher light intensities in the latter do not seem enough to explain this difference. A likely reason is the presence of UV in the latter, but not in the former (see next section). We also briefly note reported structural changes in encapsulated MAPbI3 films due to illumination (LED white light) deduced from slow changes in Raman and PL spectra which were fully reversible in the dark.21 Because of the very slow process, these changes were attributed to photoinduced structural changes in the perovskite, specifically illuminationinduced rearrangement of the Pb−I structural unit and its effect on the MA cation alignment/rotation in the perovskite crystals. In a previous study it was proposed that illumination weakens the H-bonding between the MA+ and the Pb−I framework, enhancing the rotational freedom of the MA+.22

Figure 6. XRD pattern of the 200 peak of MAPb(Br0.6I0.4)3 before (black) and after (red) white light soaking for 5 min at ∼50 mW cm−2. XRD patterns of a MAPb(Br0.2I0.8)3 film (dashed green) and a MAPb(Br0.7I0.3)3 film (dashed brown) are included for comparison. Reproduced with permission from ref 16. Copyright [2015] [Royal Society of Chemistry].

∼50 nm) containing ∼20% Br and a majority phase (depending on total Br content) with a somewhat reduced Br content compared to the original films and a crystal size estimated of 33 nm. An Arrhenius analysis of the rate of growth of the 1.68 eV peak as a function of temperature gave an activation energy of 0.27 eV, similar to measured activation energies for halide migration in various ternary halide perovskites. This supports the mechanism of the phase segregation by illumination as being due to ion (probably halide) migration; of course, it is clear that, for such phase segregation to occur, ions have to move at least the tens of nm of the composite crystal sizes. Hoke et al. suggest a possible mechanism for this phase segregation in which holes are stabilized by the formation of Irich domains (with a higher-lying EV edge) and this stabilization provides a driving enthalpy for halide segregation upon illumination. When the trapped holes recombine with electrons (upon stopping illumination), entropy and lattice strain (which was shown, by XRD, to occur to a greater extent in the minority I-rich phase than in the rest of the composite) might cause the return to a single phase, well-mixed material. A related hypothesis was that the smaller band gap of the minority I-rich phase and the resulting lower exciton energy might drive the segregation. It is clear that such light-induced changes will be, in most cases, highly undesirable for photovoltaic cells using such materials. At the same time, the reversibility of the effect is a specific example of self-healing so that, if only the light-induced changes be sufficiently slowed down (∼5 orders of magnitude slower would be required), this could be beneficial for the cells.

■

HOW DO THE OTHER CELL CONSTITUENTS AFFECT THE PEROVSKITE? Up to now, we have dealt with the effect of radiation on the perovskite phase directly. In this final section, we discuss the effect of radiation on adjacent phases that subsequently affect the cell stability. Leijtens et al.23 demonstrated how the TiO2 electron transport medium (ETM) present in the majority of perovskite cells is responsible to a major degree for cell instability and emphasizes the importance of considering the whole cell structure and not just the perovskite itself when analyzing changes in cell output over time. This instability was caused by UV radiation absorbed by the TiO2, and the cells were much more stable (although still degraded appreciably) when the UV component was removed with a filter. They concluded that the UV radiation absorbed by the TiO2 created deep electron traps through desorption of adsorbed oxygen by reaction of photogenerated holes with electrons at the site of the adsorbed oxygen, and that the deeply trapped electrons were immobile and eventually recombined with holes in the HTM. Oxygen

■

CELL STABILITY UNDER (STRONG) ILLUMINATION One question for which there is not yet a decisive answer is this: Does illumination by visible light directly affect the single halide perovskites and/or the solar cells that are based on them? Several studies addressed this question. In one study, only visible light at high intensity (up to 50 suns) was used.19 The rationale for this was that, since there is no IR irradiation, heating of the cells is relatively low (not more than 80 °C at light intensities above 50 suns). Therefore, indirect effects due to cell heating are minimized, since the high light intensities allow testing over a shorter time than under one sun illumination. Using sealed mesoporous (mp) TiO2/MAPbI3/ spiro cells, they found that the short circuit current (JSC) F

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research therefore passivates these deep traps, explaining the initially unexpected observation in this study that encapsulated cells were less stable than unencapsulated cells when illuminated by UV-containing light. They further showed that in the absence of the mp-TiO2 (using instead the much higher band gap mpAl2O3 that does not absorb the UV in sunlight), the resulting cells are much more stable in the presence of UV-containing simulated sunlight, although still showing a slow initial decrease in output before eventually stabilizing. Physical separation of the perovskite from the TiO2, by very thin Sb2S3,24 Al2O3 or Al2O3/PCBM,25 was shown to improve UV stability. Leijtens et al. suggest several ways to get around this UV/ TiO2 problem26 (the method used in refs 24 and 25, preventing direct contact between the perovskite and TiO2, is one example). However, they also refer to a study on nonencapsulated cells with a thick, porous carbon back electrode with direct contact between the TiO2 and the perovskite where good cell stability (over 7 days) was obtained in outdoor tests, that is, including UV.27 Leijtens et al. speculated that the porous carbon, a good adsorber of oxygen, might supply oxygen from the atmosphere to the TiO2 and passivate the traps at the TiO2 interface with the perovskite.26 While an interesting phenomenon, it is hard to see how this method can be used for long-term stability of encapsulated cells unless an oxygenporous, water-blocking encapsulant (silicone rubber?) can be found.

David Cahen (BSc Chemistry & Physics, Hebrew University, HU; Ph.D., Northwestern University, postdoctoral studies in photosynthesis at the HU and WIS) is professor at the WIS, where he studies materials for solar cells and (bio)electronics. Gary Hodes (B.Sc. and Ph.D. in Chemistry from Queen’s University, Belfast) is an emeritus professor at the WIS, where he studies solar cells.

■

ACKNOWLEDGMENTS We acknowledge the support in carrying out our part of the work described here of the Weizmann Institute’s Alternative sustainable Energy Research Initiative, the Israel Ministry of Science, and the Israel National Nano-Initiative.

■

(1) Song, Z.; Watthage, S. C.; Phillips, A. B.; Tompkins, B. L.; Ellingson, R. J.; Heben, M. J. Impact of Processing Temperature and Composition on the Formation of Methylammonium Lead Iodide Perovskites. Chem. Mater. 2015, 27, 4612−4619. (2) Eperon, G. E.; Stranks, S. D.; Menelaou, C.; Johnston, M. B.; Herz, L. M.; Snaith, H. J. Formamidinium lead trihalide: a broadly tunable perovskite for efficient planar heterojunction solar cells. Energy Environ. Sci. 2014, 7, 982−988. (3) Buin, A.; Pietsch, P.; Xu, J.; Voznyy, O.; Ip, A. H.; Comin, R.; Sargent, E. H. Materials Processing Routes to Trap-Free Halide Perovskites. Nano Lett. 2014, 14, 6281−6286. (4) Unger, E. L.; Hoke, E. T.; Bailie, C. D.; Nguyen, W. H.; Bowring, A. R.; Heumüller, T.; Christoforo, M. G.; McGehee, M. D. Hysteresis and transient behavior in current-voltage measurements of hybridperovskite absorber solar cells. Energy Environ. Sci. 2014, 7, 3690− 3698. (5) Tress, W.; Marinova, N.; Moehl, T.; Zakeeruddin, S. M.; Nazeeruddin, M. K. Grätzel. Understanding the rate-dependent J-V hysteresis, slow time component, and aging in CH3 NH 3 PbI 3 perovskite solar cells: the role of a compensated electric field. Energy Environ. Sci. 2015, 8, 995−1004. (6) Eames, C.; Frost, J. M.; Barnes, P. R. F.; O’Regan, B. C.; Walsh, A.; Islam, M. S. Ionic transport in hybrid lead iodide perovskite solar cells. Nat. Commun. 2015, 6, 7497. (7) Guillemoles, J. F.; Rau, U.; Schock, H.-W.; Kronik, L.; Cahen, D. Cu(In,Ga)Se2 solar cells: Device stability based on chemical flexibility. Adv. Mater. 1999, 11, 957−961. (8) Edri, E.; Kirmayer, S.; Henning, A.; Mukhopadhyay, S.; Gartsman, K.; Rosenwaks, Y.; Hodes, G.; Cahen, D. Why Methylammonium tri-IODIDE perovskite-based solar cells require a mesoporous electron transporting scaffold (but not necessarily a hole conductor). Nano Lett. 2014, 14, 1000−1004. (9) Edri, E.; Kirmayer, S.; Mukhopadhyay, S.; Gartsman, K.; Hodes, G.; Cahen, D. Elucidating the Charge Carrier Separation Mechanism in CH3NH3PbI3‑xClx Perovskite Solar Cells Provides Insight in How the Cells Work. Nat. Commun. 2014, 5, 3461. (10) Kedem, N.; Brenner, T. M.; Kulbak, M.; Schaefer, N.; Levcenko, S.; Levine, I.; Abou-Ras, D.; Hodes, G.; Cahen, D. Light-Induced Increase of Electron Diffusion Length in a p-n Junction Type CH3NH3PbBr3 Perovskite Solar Cell. J. Phys. Chem. Lett. 2015, 6, 2469−2476. (11) Holt, D. B., Joy, D. C., Eds, Techniques of Physics: 12. SEM Microcharacterization of Semiconductors; Academic Press, San Diego, 1989. (12) Kulbak, M.; Gupta, S.; Kedem, N.; Levine, I.; Bendikov, T.; Hodes, G.; Cahen, D. Cesium Enhances Long-Term Stability of Lead Bromide Perovskite-Based Solar Cells. J. Phys. Chem. Lett. 2016, 7, 167−172. (13) Kulbak, M.; Cahen, D.; Hodes, G. How Important Is the Organic Part of Lead Halide Perovskite Photovoltaic Cells? Efficient CsPbBr3 Cells. J. Phys. Chem. Lett. 2015, 6, 2452−2456.

■

CONCLUSIONS AND PROGNOSIS The relatively low thermal stability of the halide perovskite semiconductors requires greater care than normal when characterizing them, especially when using high intensity and/or high energy particle or radiation beams. We used our work with electron beams to show the need for care for electron microscopy in any form due to the changes caused by electron beams. It is important to be aware of the possibility of such changes in the perovskite and to quantify and minimize them as far as possible. This need for care also extends to solar photons (UV in particular but also visible), which have been shown to affect (not necessarily always in a negative sense) the perovskites in some cases. Moreover, for photovoltaic cells, it is important to consider the whole cell and not just the perovskite absorber in isolation as there are a number of pathways, some unrelated to the perovskite itself, that can lead to cell instability. In this respect, when considering the literature on perovskite cell stability, what is important is not that there may be many experiments showing poor or only moderate stability, but there will be some that show long-term (years or equivalent via accelerated tests) stability under realistic conditions. Such experiments are critical for the future of these cells.

■

REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest. Biographies Nir Klein-Kedem (B.Sc. Materials Engineering at Ben Gurion University of the Negev; Ph.D. Weizmann Institute of Science, WIS) is a postdoc at the WIS specializing in EBIC characterization of perovskite semiconductors. G

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research (14) Eperon, G. E.; Paternò, G. M.; Sutton, R. J.; Zampetti, A.; Haghighirad, A. A.; Cacialli, F.; Snaith, H. J. Inorganic caesium lead iodide perovskite solar cells. J. Mater. Chem. A 2015, 3, 19688−19695. (15) Xiao, C.; Li, Z.; Guthrey, H.; Moseley, J.; Yang, Y.; Wozny, S.; Moutinho, H.; To, B.; Berry, J. J.; Gorman, B. P.; Yan, Y.; Zhu, K.; AlJassim, M. M. Mechanisms of Electron-Beam-Induced Damage in Perovskite Thin Films Revealed by Cathodoluminescence Spectroscopy. J. Phys. Chem. C 2015, 119, 26904−26911. (16) Hoke, E. T.; Slotcavage, D. J.; Dohner, E. R.; Bowring, A. R.; Karunadasa, H. I.; McGehee, M. D. Reversible photo-induced trap formation in mixed-halide hybrid perovskites for photovoltaics. Chem. Sci. 2015, 6, 613−617. (17) Sadhanala, A.; Deschler, F.; Thomas, T. H.; Dutton, S. E.; Goedel, K. C.; Hanusch, F. C.; Lai, M. L.; Steiner, U.; Bein, T.; Pablo Docampo, P.; Cahen, D.; Friend, R. H. Preparation of Single Phase Films of CH3NH3Pb(I1‑xBrx)3 with Sharp Optical Band Edges. J. Phys. Chem. Lett. 2014, 5, 2501−2505. (18) Saponaro, R. Tuning the bandgap of perovskite semiconductors for high open-circuit voltage solar cells. MSc Thesis, University of Ferrara, 2014. (19) Law, C.; Miseikis, L.; Dimitrov, S.; Shakya-Tuladhar, P.; Li, X.; Barnes, P. R. F.; Durrant, J.; O’Regan, B. Performance and Stability of Lead Perovskite/TiO2, Polymer/PCBM, and Dye Sensitized Solar Cells at Light Intensities up to 70 Suns. Adv. Mater. 2014, 26, 6268− 6273. (20) Misra, R. K.; Aharon, S.; Li, B.; Mogilyansky, D.; Visoly-Fisher, I.; Etgar, L.; Katz, E. A. Temperature- and Component-Dependent Degradation of Perovskite Photovoltaic Materials under Concentrated Sunlight. J. Phys. Chem. Lett. 2015, 6, 326−330. (21) Gottesman, R.; Gouda, L.; Kalanoor, B. S.; Haltzi, E.; Tirosh, S.; Rosh-Hodesh, E.; Tischler, Y.; Zaban, A.; Quarti, C.; Mosconi, E.; De Angelis, F. Photoinduced Reversible Structural Transformations in Free-Standing CH3NH3PbI3 Perovskite Films. J. Phys. Chem. Lett. 2015, 6, 2332−2338. (22) Gottesman, R.; Haltzi, E.; Gouda, L.; Tirosh, S.; Bouhadana, Y.; Mosconi, E.; De Angelis, F.; Zaban, A. Extremely Slow Photoconductivity Response of CH3NH3PbI3 Perovskites Suggesting Structural Changes under Working Conditions. J. Phys. Chem. Lett. 2014, 5, 2662−2669. (23) Leijtens, T.; Eperon, G. E.; Pathak, S.; Abate, A.; Lee, M. M.; Snaith, H. J. Overcoming ultraviolet light instability of sensitized TiO2 with meso-superstructured organometal tri-halide perovskite solar cells. Nat. Commun. 2013, 4, 2885. (24) Ito, S.; Tanaka, S.; Manabe, K.; Nishino, H. Effects of Surface Blocking Layer of Sb2S3 on Nanocrystalline TiO2 for CH3NH3PbI3 Perovskite Solar Cells. J. Phys. Chem. C 2014, 118, 16995−17000. (25) Sutherland, B. R.; Johnston, A. K.; Ip, A. H.; Xu, J.; Adinolfi, V.; Kanjanaboos, P.; Sargent, E. H. Sensitive, Fast, and Stable Perovskite Photodetectors Exploiting Interface Engineering. ACS Photonics 2015, 2, 1117−1123. (26) Leijtens, T.; Eperon, G. E.; Noel, N. K.; Habisreutinger, S. N.; Petrozza, A.; Snaith, H. J. Stability of Metal Halide Perovskite Solar Cells. Adv. Ener. Mater. 2015, 5, 1500963. (27) Li, X.; Tschumi, M.; Han, H.; Babkair, S. S.; Alzubaydi, R. A.; Ansari, A. A.; Habib, S. S.; Nazeeruddin, M. K.; Zakeeruddin, S. M.; Grätzel, M. Outdoor Performance and Stability under Elevated Temperatures and Long-Term Light Soaking of Triple-Layer Mesoporous Perovskite Photovoltaics. Energy Technol. 2015, 3, 551− 555.

H

DOI: 10.1021/acs.accounts.5b00469 Acc. Chem. Res. XXXX, XXX, XXX−XXX