Review Article | Published:

# Defects in perovskite-halides and their effects in solar cells

## Abstract

Solar cells based on perovskite-halide light absorbers have a unique set of characteristics that could help alleviate the global dependence on fossil fuels for energy generation. They efficiently convert sunlight into electricity using Earth-abundant raw materials processed from solution at low temperature. Thus, they offer potential for cost reductions compared with or in combination with other photovoltaic technologies. Nevertheless, to fully exploit the potential of perovskite-halides, several important challenges must be overcome. Given the nature of the materials — relatively soft ionic solids — one of these challenges is the understanding and control of their defect structures. Currently, such understanding is limited, restricting the power conversion efficiencies of these solar cells from reaching their thermodynamic limit. This Review describes the state of the art in the understanding of the origin and nature of defects in perovskite-halides and their impact on carrier recombination, charge-transport, band alignment, and electrical instability, and provides a perspective on how to make further progress.

The publication in 2012 of two landmark papers1,2 describing high efficiency solid-state solar cells based on perovskite-halide semiconductors kick-started a new track of photovoltaics research. This family of perovskites merges the highly efficient operational principles of conventional inorganic semiconductors with the low-temperature solution processability of emerging organic and hybrid materials, offering a promising route towards cheap electricity generation using sunlight. The allure of impressive power conversion efficiencies in lab-scale devices, now above 22% (ref. 3), using relatively simple fabrication techniques has moved perovskite-halides from an academic curiosity4,​5,​6 to centre stage.

It is now well established that the optoelectronic properties of perovskite-halides — such as their tunable direct bandgap, high absorption coefficient, low exciton binding energy, and balanced ambipolar carrier transport — meet many of the requirements for a high-efficiency solar energy conversion technology7,​8,​9. However, in spite of significant research effort, many of the finer details of perovskite bhalide device physics remain unclear. Of particular interest is the understanding of the effects of defects. In more established semiconductors, such as Si for example, the understanding and control of defects has been a cornerstone of their successful technological deployment10.

Given the simple processability of perovskite-halides, one could expect a non-negligible level of unintentional defects at temperatures relevant for device operation. Yet, the rate of progress in power conversion efficiencies for these solar cells is unprecedented, suggesting that perovskite-halides have a relatively high tolerance for defect-related losses. However, although impressive, the highest observed power conversion efficiencies still fall short of the thermodynamic limit of 30–33% for bandgaps in the range 1.2–1.6 eV (ref. 11). In addition, much of the debate in the field surrounding the common observations of electrical instability in devices is concerned with the nature of defects in these materials. Defects thus remain one of the interesting material characteristics that underpin limitations in device operation and influence further progress towards reaching the highest possible power conversion efficiencies.

This Review describes what is known about the nature and impact of defects in solar cells based on perovskite-halides, with a focus on traps, recombination mechanisms, electrostatics, and defect conduction, which have an impact in both the bulk material and at the interfaces in devices. Beyond the conceptual aspects in understanding these material and device properties, the experimental techniques that are currently pursued to explore them are critically reviewed. Finally a perspective on future directions for perovskite-halide defect research is provided, much needed at this stage of solar cell development.

## Effects of defects in semiconductors and devices

In this section a brief overview is given of the optoelectronic processes in semiconductors that can be affected by the presence of defects, which influence device performance metrics. To provide context, Box 1 summarizes the crystal structure, basic properties, and processing procedures for perovskite-halides, as well as the most common architectures used for solar cells. Box 2 summarizes defect chemistry in semiconductors, highlighting specific defect species.

Carrier recombination processes. The principle of detailed balance for carrier generation and recombination processes requires that photons must be continuously exchanged between a semiconductor and its environment. Hence, radiative recombination within a semiconductor is a necessary process11. However, photogenerated free carriers in solar cells may also recombine through additional mechanisms for a given excitation density and temperature. Whether a particular mechanism is relevant is dependent on the densities of the background carrier population and the statistics of carrier interactions with defect states in the bandgap.

When a defect state, which is spatially localized, lies energetically within the semiconductor bandgap, there is a likelihood that an approaching electron or hole will become captured or trapped by it. The trapped electron (or hole) is likely to be emitted, or de-trapped, back to the conduction (or valence) band by phonon absorption if the activation energy is sufficiently small. However, if the activation energy is large, then it is more likely that the trapped carrier will annihilate or recombine with an opposite carrier before it can be emitted. This recombination process can be non-radiative and accompanied by the emission of phonons. The rate of this recombination process is determined by Shockley–Read–Hall (SRH) statistics and is considered an important loss mechanism in solar cells12,13. These recombination processes are presented in Fig. 1a–e.

At low excitation densities, the photoexcited population decays through interactions with the more numerous background carriers and defect states with a monomolecular rate. As the photoexcitation density increases, the monomolecular decay channels saturate and bimolecular band-to-band recombination between free electrons and holes increasingly contributes to the overall recombination rate. At much higher densities, three-body interactions become probable and Auger(-like) recombination becomes important. Assuming diffusion can be neglected, the time-dependent carrier population in a material can be described with a rate equation that accounts for these processes given by:

$dndt=−k3n3−k2n2−k1n$

where n is the photoexcited carrier density, and k1, k2, and k3 are the rate constants for monomolecular, bimolecular, and three-body recombination, respectively. These rate constants can be experimentally determined using, for example, time-resolved optical techniques such as photoluminescence decay and pump-probe spectroscopy14,​15,​16.

Non-radiative defect-mediated carrier recombination in particular is of fundamental importance to solar cell performance under open-circuit conditions. Under steady-state solar illumination, electrons are photoexcited from the valence band into the conduction band, splitting the electron and hole quasi-Fermi levels. The difference in quasi-Fermi levels is determined by the charge density at which the recombination rate equals the generation rate. Additional non-radiative recombination processes that have a shorter carrier lifetime than radiative decay thus reduce the steady-state charge density. This reduces the gap between the quasi-Fermi levels, which sets the value of the open-circuit voltage, VOC, for the solar cell. Thus, the external electroluminescence quantum efficiency (EQEEL) of the solar cell is directly related to its VOC under illumination17. Solar cells lose approximately 60 mV in VOC for every order of magnitude reduction in EQEEL.

Charge-transport processes. The absorber layer in a solar cell must be thick enough to maximize light absorption but thin enough to efficiently collect the photogenerated carriers. The collection efficiency is determined by the competition between recombination and charge transport towards the contacts.

Charge transport is also affected by defects. Free carriers in semiconductors accelerate through the crystal under an electric field until they interact with scatterers, such as phonons, defects, impurities, and so on, that alter their acceleration vectors18. The charge carrier mobility, μ, is therefore inversely proportional to the carrier effective mass and proportional to the scattering lifetime. The most typical effect of defects on charge transport in crystalline semiconductors is ionized defect scattering: the Coulomb interaction between a carrier and a defect deflects the carrier, as illustrated in Fig. 1f (ref. 19). As the temperature increases, the average thermal velocity of carriers also increases, reducing the effectiveness of this mechanism because faster carriers experience weaker deflection in the Coulomb potential of a charged defect. It is thus usually only limiting at low temperatures. Neutral defects, although less studied, may also contribute to scattering20.

If the defect states are energetically deep within the bandgap, a proportion of carriers become trapped, reducing the overall conductivity and altering the Coulombic or neutral scattering rates of the remaining free carriers. For example, this has been suggested to be relevant for the formation of potential barriers at grain boundaries21. The extent to which trapping affects charge transport depends on the density of traps and device operating conditions. When the photocarrier density is much higher than the trap density, the traps are likely to be fully populated and the impact on charge transport may be negligible. When the photocarrier density is much lower that the trap density, most carriers become trapped. In this extreme limit of highly disordered materials, charge transport becomes dominated by hopping-like or multiple trapping and release processes22. The resulting effective mobility is thermally activated but with limited magnitude. For intermediate trap densities, charge transport in devices can become dependent on the applied bias where the steady-state carrier density is low near short-circuit conditions, but increases towards open-circuit conditions.

Band alignment. The electric-field distribution within an operating solar cell is one of the defining characteristics of its operational mechanism, determining which regions are dominated by carrier drift or diffusion under given biasing conditions. In ideal circumstances, it is determined by the energies of the band edges and intentional carrier doping levels in the device layers23. However, unintentional defects in the bulk or at device interfaces can affect the electric field inside the device.

In a general sense, perovskite-halide solar cells are designed with a structure reminiscent of a p–i–n (or n–i–p) heterojunction23. An intrinsic perovskite absorber layer is desired, which is sandwiched between electron- and hole-selective contacts. The band alignment at thermodynamic equilibrium results in a linearly varying potential gradient in the intrinsic absorber, as shown in Fig. 1g. Under illumination, this built-in field drives photogenerated free carriers towards the selective contacts for efficient collection at short-circuit conditions. Forward biasing the cell opposes the built-in field. The open-circuit condition is obtained when the applied bias matches the illumination-induced quasi-Fermi level splitting.

If the absorber contains unintentional defects with associated electronic states that are shallow with respect to the band edges, then the defects can ionize at room temperature serving as donor or acceptor states that dope the semiconductor. One possible result of unintentional doping of the absorber layer is the formation of a p–n heterojunction, as shown in Fig. 1h. Photogenerated minority carriers in the absorber depletion region drift across the junction for efficient collection. Minority carriers generated in the quasi-neutral region in the absorber can diffuse into the depletion region efficiently as long as their diffusion length is much longer than the quasi-neutral region itself. Applying a forward-bias reduces the junction barrier height, increasing majority carrier diffusion in opposition to the collection current, thus increasing the recombination rate until it matches the generation rate at the open-circuit condition.

One further consequence of defects emerges when they comprise energetically deep-lying states within the bandgap. A high density of deep defects can pin the Fermi level in the bulk of the intrinsic layer or at the interfaces, limiting control over the electric field distribution through doping23. Fermi-level pinning in the bulk of the intrinsic layer in particular can impose a limitation on VOC where the filling of sub-bandgap states under illumination limits the quasi-Fermi level splitting. A possible band diagram that can emerge due to deep defects is shown in Fig. 1i.

## Defects in perovskite-halides

The most widely studied defects in perovskite-halides are the native point defects in methylammonium lead triiodide (CH3NH3PbI3) — the archetypal perovskite-halide for photovoltaics. CH3NH3PbI3 has 12 native point defects: the vacancies VMA, VPb, and VI; the interstitials MAi, Pbi, and Ii; and anti-site occupations MAPb, MAI, PbMA, PbI, IMA, and IPb (refs 24,​25,​26,​27,​28,​29,​30,​31), where MA is methylammonium (CH3NH3). Theoretical approaches to studying these defects use calculations of their formation energies in a host crystal of the semiconductor in equilibrium with the pure constituents, for example, CH3NH3PbI3 in equilibrium with I2, Pb, and methylammonium24,​25,​26,​27,​28,​29,​30,​31. In these calculations, the requirement for equilibrium of the pure constituents with the host crystal is perhaps unlikely to be a realistic processing scenario given the typical techniques used for growing perovskite-halides32. Nevertheless, such calculations serve to offer qualitative trends in defect formation and guidance for desirable processing conditions. In all studies on perovskite-halides thus far, the calculated quantities are estimated using density functional theory (DFT) calculations. Among those studies, there is variability in the details of the implementation of DFT as well as the processing conditions that are assumed in the calculations. However, some general trends have still emerged.

Across these studies there is a common message that the point defects in CH3NH3PbI3 that would contribute deep levels in the bandgap have high formation energies, and the more important point defects with lower formation energies should be shallow states24,​25,​26,​27,​28,​29,​30. This would suggest that point defects should not contribute a high density of non-radiative recombination centres, whose energies usually lie deep within the bandgap. However, there are some exceptions. The predictions of which defects could contribute deep levels are varied, but include IPb, IMA, Pbi, PbI, VI, and PbMA (refs 25,26,29,31). Among these, there are at least some growth conditions where the formation energies of PbI, VI, and IMA may be low enough to contribute a significant density of recombination centres26,29,31. For example, it has been predicted that under iodine-rich conditions, the formation energy of the PbI anti-site is low and its most stable charge state makes a transition at a Fermi level that sits deep within the bandgap, as shown in Fig. 2 (ref. 31). The interactions of atoms around these defects, in the formation of covalent bonds for example, are predicted to create additional states that also lie deeper in the bandgap26.

The shallow point defects could cause unintentional doping at room temperature. The acceptor defects are VMA, VPb, Ii, MAPb, IMA, and IPb. The donor defects are VI, MAi, Pbi, PbMA, MAI, and PbI (ref. 25). The shallow defects with low formation energies under some growth conditions are VPb, VMA, MAPb, and Ii, which are acceptors, and MAi, VI, and MAI, which are donors25,29. Both shallow donors and acceptors can form with low formation energies, allowing CH3NH3PbI3 to be intrinsically doped from p-type to n-type when carefully controlling the growth conditions. Schottky disorder (cation and anion vacancies form in equal numbers) may dominate the defect formation in perovskite-halides under stoichiometric growth conditions27. In this case, unintentional doping is minimized as each vacancy type contributes an opposite charge carrier, naturally compensating one another, even if the defect formation energies are low. This may be a plausible explanation for the low doping densities often observed in Hall effect measurements, in the range 109–1014 cm−3 (refs 33,​34,​35,​36,​37).

## Carrier recombination

Recent measurements have confirmed trends in earlier theoretical work that the room-temperature exciton binding energy for CH3NH3PbI3 is in the range of a few to tens of millielectron volts38,​39,​40,​41. For this value, the probability that a photoexcitation results in unbound electron–hole pairs is approximately one under solar illumination and geminate excitonic decay channels can be neglected. The most relevant recombination mechanisms are thus likely to be restricted to defect mediated, band to band, and Auger(-like) depending on the photoexcitation density.

Several studies42,43 have taken advantage of absorption and emission reciprocity relationships in complete solar cells to show that the predominant contribution to the deficit in VOC from its ideal value under solar illumination is a low EQEEL value. This indicates that most carriers recombine non-radiatively. These relationships have been recently explored in mixed-cation, mixed-halide perovskites, concluding that the EQEEL is limited by SRH recombination in devices with state-of-the-art power conversion efficiency44.

Under photoexcitation densities comparable to working solar cell conditions, photoluminescence quantum yields in bare thin-films of CH3NH3PbI3 usually do not exceed 1% (ref. 14) and the dominant decay channel is monomolecular in nature14,​15,​16. Non bradiative decay is thus significant in the perovskite films themselves and is not only an emergent property in complete devices42,43. These results show that defects in the perovskite are active and have an important contribution, contrary to the general theoretically predicted properties of point defects24,25. An example plot of the temperature dependence of the monomolecular decay constant, k1, is shown in Fig. 3a (ref. 16). The dependence is consistent with a proportion of the recombination being mediated by thermally ionized defects. The competition of this non bradiative monomolecular decay with band bto bband recombination determines the radiative efficiency. For significant gains in radiative efficiency under solar illumination, and hence improvements in VOC, k1 needs to be reduced by orders of magnitude, as highlighted in Fig. 3b (ref. 45). The density of states within the bandgap that contribute to this monomolecular decay component have been measured to be in the range of 1016−1017 cm−3 using optical techniques14,15,46. However, we note that photoluminescence lifetimes are variable even within the same film, implying defects are inhomogeneously spatially distributed47.

A common misconception in the field is that the long-lived decay times observed in time-resolved photoluminescence studies are direct measurements of carrier recombination lifetimes. However, given that only the emissive lifetime is observed but most recombination is non-emissive when the quantum yield is low, the interpretation is more nuanced. If the non-radiative component is indeed trap-mediated, most carriers must become trapped and can not recombine in a band-to-band transition, so the photoluminescence dynamics are in fact dominated by trapping kinetics. For example, the emissive and free-carrier populations exhibit different dynamics in time-resolved microwave conductivity and photoluminescence48. Thus, a longer-lived decay could be observed compared with the case of purely bimolecular radiative decay, but this is not the recombination lifetime of the initial photoexcited population: longer photoluminescence lifetimes do not directly imply improved device performance or film quality; the quantum yield is also fundamentally important. To allow a consistent interpretation of results, the excitation density dependence of the quantum yield and of the decay lifetime, which allows the determination of the decay regime (that is, trap-limited versus bimolecular), should always be reported.

## Charge transport

The reduced effective mass for carriers in CH3NH3PbI3 has recently been measured by optical magneto-absorption to be 0.104me, where me is the mass of an electron39. Some closely related formamidinium-based and bromine-based analogues have values in the range 0.09–0.117me, which increase roughly proportionally with the bandgap49. These values are comparable to, for example, electrons in GaAs, thus implying that the real mobility in the perovskite could be relatively high. However, the other ingredients of mobility, namely the scattering and trapping lifetimes, are crucially important.

The importance of scattering processes can be inferred from the temperature dependence of the carrier mobility. Using pump-probe terahertz spectroscopy and microwave conductivity, several groups have studied the temperature dependence of the carrier mobility in CH3NH3PbI3 (refs 16,50,51). Typically, in the tetragonal phase, the mobility decreases with increasing temperature closely proportional to T−3/2, as shown in Fig. 4. This is consistent with acoustic phonon scattering and a negligible contribution from defects or other processes.

The reported values for the room-temperature carrier mobility in CH3NH3PbI3 are usually on the order of tens of cm2 Vs−1 (refs 16,33,34,50,51), with both electrons and holes exhibiting similar values52. It remains unclear why there is a discrepancy between measurements when transport appears to be only limited by phonon scattering, an intrinsic property of the crystal, even in the case of polycrystalline films. Nevertheless, given values in this range and recombination rates determined in thin films, the carrier diffusion lengths within this material are on the order of micrometres45. This is several times the optical absorption depth enabling efficient light-harvesting and current collection simultaneously, which explains the impressive performance of the material in solar cells.

## Band alignment and interfaces

Several of the material properties that are required for understanding the band diagrams of perovskite-halide solar cells have been measured for some processing conditions. Taking CH3NH3PbI3 as an example, the bandgap has been estimated with optical methods to be 1.6 eV (refs 14,53). Observations at the surfaces of perovskite-halides using X-ray and ultraviolet photoelectron spectroscopy suggest a valence band edge at 5.4 eV (refs 1,54,​55,​56,​57). The intrinsic doping density has been measured by the Hall effect and usually falls in the range 109–1014 cm−3 (refs 33,​34,​35,​36,​37). Both p- and n- type doping has been observed. The lower end of the doping range is expected of a high-quality or compensated intrinsic semiconductor and is seen in single crystals. These values are qualitatively in agreement with the expected behaviour and formation energies of native point defects. The density of states within the bandgap for polycrystalline films has been estimated to be in the range 1016–1019 cm−3 using capacitance and admittance spectroscopy on complete devices, which covers values derived from optical measurements14,15,46. Several peaks in the energetic distributions of states with activation energies of 0.167, 0.24, 0.3, 0.35, 0.45, and 0.66 eV have been observed, generally over a broad background37,58,59. An example distribution is shown in Fig. 5a. Results from different laboratories do not report agreement in the observed activation energies and there is ambiguity in whether these are hole or electron states. This point is discussed further in the Outlook section. The energetic distribution below the conduction band and above the valence band has only been measured unambiguously in single crystals using temperature-dependent space-charge limited currents (SCLC). The density of states was measured to be in the range of 1010–1011 cm−3 distributed within 0.25 eV of the band edges, as shown in Fig. 5b (ref. 36). The static or low-frequency relative dielectric constant has been measured to be 60.2 using millimetre-wave spectroscopy of pressed pellets assuming Debye relaxation60; 70 using charge-extraction by linearly increasing voltage in solar cells61; and 1,000 by impedance spectroscopy in thin-film or pressed pellet devices62,63. The latter anomalously high value is most likely a result of the slow response of mobile defects, as discussed further in the next section, that does not determine the dielectric behaviour under static conditions. However, the importance of the space charge caused by mobile charged defects, particularly in transient phenomena, is a topic of ongoing debate64,65.

To study the electrostatic properties of complete solar cells based on CH3NH3PbI3, cross-sectional mapping using electron-beam-induced current (EBIC)66 and scanning Kelvin probe force microscopy (SKPM)67,68 has been demonstrated. Rather than directly probing the electric field, EBIC maps the variation of current collection across the device from which the underlying electric field may be qualitatively described. The electric field distribution66 measured with EBIC was found to be roughly commensurate with a p–i–n structured solar cell but without full depletion of the absorber, leading to a regions of low collection efficiency, as shown in Fig. 6a. SKPM maps the surface potential difference between a scanning probe tip and the sample, giving an indication of the potential variation across the device. The potential distributions measured by SKPM have been found to be commensurate with either p–i–n or p–n junction solar cells, even in very similar device architectures67,​68,​69. In cases where p–n junctions were observed, the perovskite layer was p-type and formed a p–n junction at the perovskite–TiO2 interface, as shown in Fig. 6b. Using independent measurements of the dielectric constants and doping in the n-type TiO2 layer, complementary modelling of the capacitance–voltage characteristics of devices has indicated that the perovskite layer is p-type with a carrier density of 1016 cm−3 assuming an ideal two-sided p–n junction. It should be noted that this carrier density is far in excess of most measurements based on the Hall effect, highlighting the sensitivity of the defect structure to the nature of substrate on which the film is formed. The variation in results from these mapping studies also highlights the fact that the operational mechanism for solar cells based on perovskite-halides is not well defined; it is dependent on the level of unintentional doping in the perovskite layer.

Among the few studies that have attempted to understand interfaces in perovskite solar cells, thermal admittance spectroscopy (TAS)70 is a relatively popular experimental technique that can probe the impact of carrier-selective contact materials on the density of trap states. One of the interesting interfaces studied with TAS is that of CH3NH3PbI3 with [6,6]-phenyl-C61-butyric acid methyl ester (PCBM). Experimental results suggest that a specific interaction with iodine species at this interface reduces the density of states within the bandgap of the perovskite30,71,72, leading to improved device performance37. These devices exhibit reduced current–voltage hysteresis and efficient charge extraction with respect to alternative electron-selective layers in planar devices73. Careful control of the crystallinity and energetic disorder at this interface also correlates with higher open-circuit voltages72.

## Defect mobility

One of the most intriguing topics that has emerged in studying perovskite-halide solar cells is the observation of current–voltage hysteresis when characterizing devices74. This has raised substantial discussion about measurement accuracy in the reporting of high efficiencies and the long-term stability of the power output of perovskite-halide solar cells. Two hypothetical phenomena have dominated the discussion of its origin: ferroelectricity and defect mobility.

The ferroelectric effect is a property of some polar crystals whose dipoles can be permanently aligned by an external electric field. In a solar cell, this crystal polarization could serve either to enhance or screen the built-in potential, allowing two distinct electric-field distributions with different current–voltage characteristics in the same device. It has been proposed that ferroelectricity emerging through ion displacement33,75,76 or dipole orientation of the organic cation77 may be relevant in CH3NH3PbI3. However, piezoelectric force microscopy has been used to show that microscopic ferroelectric domains can only persist for short timescales78. Other conventional measurements of the ferroelectric effect have returned negative results79. In addition, studies on CsPbI3 in its cubic phase conclude that both conventional ferroelectricity and cation alignment are not necessary for hysteresis in devices80.

The alternative hypothesis, that mobile defects are playing a key role, is gathering support. Indeed, other perovskites related to CH3NH3PbI3 have long been suspected to be halide-ion conductors81. A material of this type is usually termed a mixed ionic–electronic conductor and can find applications in energy storage devices82, for example. For photovoltaics, however, this could lead to recombination processes and charge distributions being both space- and time-dependent under operating conditions. In the worst case, this could manifest as electrical instability of the power output or an undesirable transient response under variable illumination and load.

The observation that a single, nominally symmetric device based on CH3NH3PbI3 could be polarized to induce photovoltaic characteristics in two power-generating quadrants of operation was first explained as a consequence of mobile ions that could generate a potential gradient in the device after poling83. Further evidence that this perovskite behaves as a mixed electronic and ionic conductor was obtained from pressed pellets of perovskite powder that were sandwiched between various electrode materials that selectively interact electronically or with various ionic species, as shown in Fig. 7 (ref. 63). The findings suggested that the most important mobile species were point defects, namely the halogen interstitial or vacancy. Similarly, temperature-dependent current transients combined with theoretical calculations suggested that the iodine vacancy is the principal contributing defect, with an activation energy for hopping of 0.6 eV (ref. 28). Other theoretical calculations suggest that both the iodine vacancy and interstitial should have a similar activation energy to transport that is significantly lower than other native point defects84.

Under atmospheric operation, current–voltage hysteresis becomes more severe. This might be due to specific interactions of the perovskite with atmospheric moisture creating mobile species in addition to native point defects85,86. Other preliminary experimental results have suggested that defect states may mediate these interactions with environmental agents, thus playing an additional role in instability87. These results highlight a need for robust encapsulation in commercial applications.

Mobile defects have also been observed to mediate degradation related to light exposure, particularly in mixed-halide perovskites. These perovskites have received special attention because their bandgap can be optimized88 for multi-junction solar cells when coupled with crystalline Si or other thin-film solar cells89. However, it has been demonstrated that emissive sub-bandgap states are formed in mixed iodide–bromide perovskites by the action of white-light illumination90. This is related to photo-induced halide segregation into iodide-rich minority and bromide-enriched majority domains; the iodide-rich domains acting as recombination centres following photoexcitation of the bromide-enriched domains. By comparing the optical properties of nanocrystals (not necessarily quantum confined) containing low defect densities with thin films, there appears to be a correlation between the susceptibility to degradation and the crystal quality, that is, crystals with fewer defects and higher photoluminescence efficiency are less susceptible to photodegradation91,92. Even in the case of CH3NH3PbI3, this phenomenon has been recently observed to modulate the density of non-radiative defect states93. Improving the crystal quality may thus provide benefits to not only the optoelectronic properties of devices, but also their long-term stability.

## Outlook

Although the theoretical predictions of the properties of point defects have proven informative, experimental verification of these results remains paramount. Identifying defects and understanding their properties is a formidable challenge and requires a range of complementary techniques, but studies on more conventional materials can provide hints at what further experiments can be done. For example, both positron annihilation spectroscopy94 and electron paramagnetic resonance95 have been used to selectively study different defect species in semiconductors, such as vacancies and defects with unpaired charges, respectively. In addition, among a host of electrical characterization techniques that are applicable for studying the electronic properties of defects, potentially useful methods for complete devices are the variants of deep-level transient spectroscopy70. These approaches can allow the determination of densities, emission rates, activation energies, and trapped species, and capture cross-sections of deep defects that may be active in transport and recombination processes. These are the properties of defects that are required for rigorously modelling recombination with SRH statistics12,13. This would open up a deeper explanation for the photoluminescence properties of films and the electroluminescence properties of devices, ultimately linking defect structures to the fill factors and open-circuit voltages of solar cells. Using these techniques in combination with processing strategies designed to control the density of specific point defects, as well as commonly used material characterization methods that measure stoichiometry and doping density — such as energy-dispersive X-ray spectroscopy and Hall effect measurements, respectively — a richer picture of phenomena arising from specific defects can emerge.

In traditional semiconductors, unintentional chemical impurities have been considered as an important type of point defect leading to doping and trap levels within the bandgap. Specific processing techniques have been required to minimize impurity concentrations such as ‘gettering’, where additives are used to act as sinks for impurities10. In contrast, the precursor purity used to prepare perovskite-halides has received less attention. Chang et al. have attempted to investigate the purity of the PbI2 as a precursor, but the correlation of purity to morphological variation may mask the chemical influence of impurities96. Zhang et al. have investigated purification protocols for the CH3NH3I precursor, finding that residual hypophosphorous acid may be desirable because it inhibits the formation of I2 in perovskite precursor solutions. This leads to a more favourable stoichiometry during crystallization and a reduction in defect density97. Thus, further work is needed to understand exactly to what extent chemical impurities directly introduce detrimental electronic states, or serve as processing or passivation additives.

The question as to whether the properties of point defects are sufficient to completely describe defect-related phenomena in perovskite-halides also remains open. Higher-dimensional defects have certainly been observed, but their effects in devices remain poorly understood. Dislocations have been observed with atomic-scale scanning tunnelling microscopy at the surface of a freshly cleaved CH3NH3PbBr3 crystal98. Studies based on conductive atomic force microscopy suggest that grain boundaries in polycrystalline CH3NH3PbI3 thin films are active, but are not necessarily significantly detrimental to charge transport99,100. Their precise roles in carrier recombination are not well understood41,47,101, but there are indications that they serve to reduce the carrier diffusion lengths102. Clusters or precipitates of metallic lead54,103,104 have been detected in CH3NH3PbI3. Detection of these precipitates has been shown to correlate with a reduced photoluminescence quantum yield in freshly prepared films, but post-annealing under ambient atmosphere reduces their density104. In contrast, PbI2 precipitates have been correlated with improved device performance44,105. Speculative hypotheses for its function are that it passivates grain boundary defects, or acts as a processing additive increasing the average crystallite size while reducing the density of grain boundaries. In both cases, it may serve to reduce non-radiative recombination at grain boundaries.

The type of the junction in perovskite-halide solar cells has been shown to depend on the self-doping of the perovskite layer. Electrical characterization methods are more commonly used in conventional semiconductors to derive information about the aspects of devices that determine these electrostatic properties. These approaches typically model impedance and admittance spectroscopy, current–voltage, and capacitance–voltage measurements to derive barrier heights, as well as interface and bulk densities of states59,106,107. To reduce the complexity of the interpretation, these techniques are often applied in idealized devices that probe, for example, a single interface rather than a complete solar cell. The basis of these models is knowledge of the doping densities and dielectric constants obtained from complementary techniques. These measurements are typically derived from isolated films, which introduces an inherent uncertainty when used in complete device modelling. This is because the surface energy and roughness of the substrate can change the activity of reactants in the perovskite precursor solution during processing, potentially modifying the formation energy of defects. Thus, the properties of isolated films are not necessarily identical to films formed on top of electron- or hole-selective contact materials. Taking TAS measurements as an example, to be certain that a sub-bandgap state is a trap for either electrons or holes requires prior knowledge that there is only a single majority carrier species throughout the depletion region (this is not true for a p–i–n junction, for example). This cannot easily be determined with a device model because the doping level for the perovskite is not known with certainty when processed in a device configuration. In addition, electrical measurements may be dependent on the device's measurement history through action of charged mobile defects73. Their redistribution under measurement conditions could have an impact on the electric-field distribution in the solar cell. This effect is not well handled by conventional semiconductor modelling and makes such investigations very challenging to interpret. In future studies, time-resolved two-photon photoemission spectroscopy108 could be particularly helpful for detecting the distributions of sub-bandgap surface states. In this measurement, sub-bandgap electron traps are first populated with an optical pump, then the electrons are photoemitted with a second pulse and their kinetic energy is measured. This can allow measurement of energetic distribution of surface states and the corresponding carrier lifetimes within those states.

Following the initial rise of power conversion efficiencies in perovskite solar cells, the understanding of the limiting material properties is catching up. An important area of focus for future effort is the understanding and control of defects, which have an impact on several aspects of device functionality. The important gaps in our current understanding are in the precise identification of bulk and interface defects that are responsible for non-radiative recombination and have an effect on band alignment. Older studies on more conventional materials offer some guidance on how to move forwards. The observations of the high electronic quality of macroscopic single crystals grown in solution demonstrates that further improvements to material quality are not limited by the thermodynamics of defect formation in perovskite-halides. With greater control over the engineering of defect structures in thin films, solar cell power conversion efficiencies can therefore be expected to continue to approach their thermodynamic limit.

## References

1. 1.

et al. Lead iodide perovskite sensitized all-solid-state submicron thin film mesoscopic solar cell with efficiency exceeding 9%. Sci. Rep. 2, 591 (2012). This work was the first study of a solid-state perovskite-sensitized solar cell using a mesoporous TiO2 anode.

2. 2.

, , , & Efficient hybrid solar cells based on meso-superstructured organometal halide perovskites. Science 338, 643–648 (2012). This work was the first study of a solid-state perovskite-halide solar cell without an electrically active mesoporous anode showing that perovskite-halides can also support charge transport.

3. 3.

Best Research-Cell Efficiencies (National Renewable Energy Laboratory, 2016);

4. 4.

CH3NH3PbX3, ein Pb(II)-system mit kubischer Perowskitstruktur / CH3NH3PbX3, a Pb(II)-system with cubic perovskite structure. Z. Naturforsch. B 33, 1443–1445 (1978).

5. 5.

, , , & Conducting layered organic–inorganic halides containing <110>-oriented perovskite sheets. Science 267, 1473–1476 (1995).

6. 6.

, , & Organometal halide perovskites as visible-light sensitizers for photovoltaic cells. J. Am. Chem. Soc. 131, 6050–6051 (2009).

7. 7.

, & The emergence of perovskite solar cells. Nat. Photon. 8, 506–514 (2014).

8. 8.

& Metal-halide perovskites for photovoltaic and light-emitting devices. Nat. Nanotech. 10, 391–402 (2015).

9. 9.

Perovskite solar cells: an emerging photovoltaic technology. Mater. Today 18, 65–72 (2015).

10. 10.

& Defects in semiconductors: some fatal, some vital. Science 281, 945–950 (1998).

11. 11.

& Detailed balance limit of efficiency of p–n junction solar cells. J. Appl. Phys. 32, 510–519 (1961).

12. 12.

Electron-hole recombination in germanium. Phys. Rev. 87, 387 (1952).

13. 13.

& Statistics of the recombination of holes and electrons. Phys. Rev. 87, 835–842 (1952).

14. 14.

et al. Correlated electron–hole plasma in organometal perovskites. Nat. Commun. 5, 5049 (2014). This was the first thorough study of the carrier-density dependent recombination dynamics in CH3NH3PbI3 suggesting that defects were important at densities relevant for solar cell operation.

15. 15.

et al. Recombination kinetics in organic–inorganic perovskites: excitons, free charge, and subgap states. Phys. Rev. Appl. 2, 034007 (2014).

16. 16.

, , , & Temperature-dependent charge-carrier dynamics in CH3NH3PbI3 perovskite thin films. Adv. Funct. Mater. 25, 6218–6227 (2015).

17. 17.

Reciprocity relation between photovoltaic quantum efficiency and electroluminescent emission of solar cells. Phys. Rev. B 76, 085303 (2007).

18. 18.

Semiconductor Physics (Springer-Verlag Berlin Heidelberg, 2004).

19. 19.

& Theory of impurity scattering in semiconductors. Phys. Rev. 77, 388–390 (1950).

20. 20.

& Neutral impurity scattering in semiconductors. Phys. Rev. B 11, 5208–5210 (1975).

21. 21.

The electrical properties of polycrystalline silicon films. J. Appl. Phys. 46, 5247–5254 (1975).

22. 22.

& Electronic transport in amorphous silicon films. Phys. Rev. Lett. 25, 509–511 (1970).

23. 23.

The Physics of Solar Cells (Imperial College Press, 2003).

24. 24.

, , & The role of intrinsic defects in methylammonium lead iodide perovskite. J. Phys. Chem. Lett. 5, 1312–1317 (2014).

25. 25.

, & Unusual defect physics in CH3NH3PbI3 perovskite solar cell absorber. Appl. Phys. Lett. 104, 063903 (2014). The first study of the energetics of native point defects in CH3NH3PbI3 leading to the suggestion that highly detrimental defects would be difficult to form in this perovskite.

26. 26.

, , & Strong covalency-induced recombination centers in perovskite solar cell material CH3NH3PbI3. J. Am. Chem. Soc. 136, 14570–14575 (2014).

27. 27.

, , , & Self-regulation mechanism for charged point defects in hybrid halide perovskites. Angew. Chem. Int. Ed. 54, 1791–1794 (2015).

28. 28.

et al. Ionic transport in hybrid lead iodide perovskite solar cells. Nat. Commun. 6, 7497 (2015).

29. 29.

et al. Materials processing routes to trap-free halide perovskites. Nano Lett. 14, 6281–6286 (2014).

30. 30.

et al. Perovskite–fullerene hybrid materials suppress hysteresis in planar diodes. Nat. Commun. 6, 7081 (2015).

31. 31.

, , , & Halide-dependent electronic structure of organolead perovskite materials. Chem. Mater. 27, 4405–4412 (2015).

32. 32.

, , , & Formation of thin films of organic–inorganic perovskites for high-efficiency solar cells. Angew. Chem. Int. Ed. 54, 3240–3248 (2015).

33. 33.

, & Semiconducting tin and lead iodide perovskites with organic cations: phase transitions, high mobilities, and near-infrared photoluminescent properties. Inorg. Chem. 52, 9019–9038 (2013).

34. 34.

et al. Understanding the formation and evolution of interdiffusion grown organolead halide perovskite thin films by thermal annealing. J. Mater. Chem. A 2, 18508–18514 (2014).

35. 35.

et al. Electron–hole diffusion lengths >175 μm in solution-grown CH3NH3PbI3 single crystals. Science 347, 967–970 (2015).

36. 36.

et al. The in-gap electronic state spectrum of methylammonium lead iodide single-crystal perovskites. Adv. Mater. 28, 3406–3410 (2016). This work was the first unambiguous measurement in the sub-gap density-of-states spectrum in a perovskite-halide single crystal.

37. 37.

, , , & Origin and elimination of photocurrent hysteresis by fullerene passivation in CH3NH3PbI3 planar heterojunction solar cells. Nat. Commun. 5, 5784 (2014). This work was the first to show that the interface between the perovskite and PCBM played an important role in suppressing J–V hysteresis induced by mobile defects.

38. 38.

, & Analysis of multivalley and multibandgap absorption and enhancement of free carriers related to exciton screening in hybrid perovskites. J. Phys. Chem. C 118, 11566–11572 (2014).

39. 39.

et al. Direct measurement of the exciton binding energy and effective masses for charge carriers in organic–inorganic tri-halide perovskites. Nat. Phys. 11, 582–587 (2015).

40. 40.

et al. Intrinsic femtosecond charge generation dynamics in single crystal CH3NH3PbI3. Energy Environ. Sci. 8, 3700–3707 (2015).

41. 41.

et al. Role of microstructure in the electron–hole interaction of hybrid lead halide perovskites. Nat. Photon. 9, 695–701 (2015).

42. 42.

et al. Radiative efficiency of lead iodide based perovskite solar cells. Sci. Rep. 4, 6071 (2014). This work was the first study to link non-radiative recombination channels to the open-circuit voltage in perovskite-halide solar cells.

43. 43.

et al. Predicting the open-circuit voltage of CH3NH3PbI3 perovskite solar cells using electroluminescence and photovoltaic quantum efficiency spectra: the role of radiative and non-radiative recombination. Adv. Energy Mater. 5, 1400812 (2015).

44. 44.

et al. Efficient luminescent solar cells based on tailored mixed-cation perovskites. Sci. Adv. 2, e1501170 (2016).

45. 45.

& Hybrid perovskites for photovoltaics: charge-carrier recombination, diffusion, and radiative efficiencies. Acc. Chem. Res. 49, 146–154 (2015).

46. 46.

et al. Low-temperature solution-processed wavelength-tunable perovskites for lasing. Nat. Mater. 13, 476–480 (2014).

47. 47.

et al. Impact of microstructure on local carrier lifetime in perovskite solar cells. Science 348, 683–686 (2015).

48. 48.

et al. Charge carrier lifetimes exceeding 15 μs in methylammonium lead iodide single crystals. J. Phys. Chem. Lett. 7, 923–928 (2016).

49. 49.

et al. Determination of the exciton binding energy and effective masses for methylammonium and formamidinium lead tri-halide perovskite semiconductors. Energy Environ. Sci. 9, 962–970 (2016).

50. 50.

et al. Thermally activated exciton dissociation and recombination control the carrier dynamics in organometal halide perovskite. J. Phys. Chem. Lett. 5, 2189–2194 (2014). This was the first study to demonstrate that the carrier mobility in the room temperature phase of CH3NH3PbI3 is limited by phonon scattering rather than by defects.

51. 51.

, , , & Improved understanding of the electronic and energetic landscapes of perovskite solar cells: high local charge carrier mobility, reduced recombination, and extremely shallow traps. J. Am. Chem. Soc. 136, 13818–13825 (2014).

52. 52.

et al. Electron–hole diffusion lengths exceeding 1 micrometer in an organometal trihalide perovskite absorber. Science 342, 341–344 (2013).

53. 53.

et al. Organometallic halide perovskites: sharp optical absorption edge and its relation to photovoltaic performance. J. Phys. Chem. Lett. 5, 1035–1039 (2014).

54. 54.

et al. Electronic structure of TiO2/CH3NH3PbI3 perovskite solar cell interfaces. J. Phys. Chem. Lett. 5, 648–653 (2014).

55. 55.

et al. Interface energetics in organo-metal halide perovskite-based photovoltaic cells. Energy Environ. Sci. 7, 1377–1381 (2014).

56. 56.

et al. Chemical and electronic structure characterization of lead halide perovskites and stability behavior under different exposures — a photoelectron spectroscopy investigation. Chem. Mater. 27, 1720–1731 (2015).

57. 57.

et al. Electronic level alignment in inverted organometal perovskite solar cells. Adv. Mater. Interfaces 2, 1400532 (2015).

58. 58.

et al. Defect density and dielectric constant in perovskite solar cells. Appl. Phys. Lett. 105, 153502 (2014).

59. 59.

et al. The identification and characterization of defect states in hybrid organic–inorganic perovskite photovoltaics. Phys. Chem. Chem. Phys. 17, 112–116 (2014).

60. 60.

& Dynamic disorder in methylammoniumtrihalogenoplumbates (II) observed by millimeter-wave spectroscopy. J. Chem. Phys. 87, 6373 (1987).

61. 61.

et al. Electro-optics of perovskite solar cells. Nat. Photon. 9, 106–112 (2014).

62. 62.

et al. Photoinduced giant dielectric constant in lead halide perovskite solar cells. J. Phys. Chem. Lett. 5, 2390–2394 (2014).

63. 63.

, , , & The significance of ion conduction in a hybrid organic–inorganic lead-iodide-based perovskite photosensitizer. Angew. Chem. Int. Ed. 54, 7905–7910 (2015). The first study that was able to separate the contributions of electronic and ionic species to conduction in CH3NH3PbI3 and show that halogen point defects were the dominant species in ionic conduction.

64. 64.

, & Modeling anomalous hysteresis in perovskite solar cells. J. Phys. Chem. Lett. 6, 3808–3814 (2015).

65. 65.

et al. Can slow-moving ions explain hysteresis in the current–voltage curves of perovskite solar cells? Energy Environ. Sci. 9, 1476–1485 (2016).

66. 66.

et al. Elucidating the charge carrier separation and working mechanism of CH3NH3PbI3−xClx perovskite solar cells. Nat. Commun. 5, 3461 (2014). This work was the first attempt to experimentally map the type of junction in a perovskite-halide solar cell.

67. 67.

et al. Real-space observation of unbalanced charge distribution inside a perovskite-sensitized solar cell. Nat. Commun. 5, 5001 (2014).

68. 68.

et al. Carrier separation and transport in perovskite solar cells studied by nanometre-scale profiling of electrical potential. Nat. Commun. 6, 8397 (2015).

69. 69.

, , , & Electrical field profile and doping in planar lead halide perovskite solar cells. Appl. Phys. Lett. 105, 133902 (2014).

70. 70.

, & Advanced Characterization Techniques for Thin Film Solar Cells (Wiley, 2011).

71. 71.

et al. Ion migration and the role of preconditioning cycles in the stabilization of the J–V characteristics of inverted hybrid perovskite solar cells. Adv. Energy Mater. 6, 1501453 (2015).

72. 72.

, & Correlation of energy disorder and open-circuit voltage in hybrid perovskite solar cells. Nat. Energy 1, 15001 (2016).

73. 73.

et al. 17.6% stabilized efficiency in low-temperature processed planar perovskite solar cells. Energy Environ. Sci. 8, 2365–2370 (2015).

74. 74.

et al. Anomalous hysteresis in perovskite solar cells. J. Phys. Chem. Lett. 5, 1511–1515 (2014).

75. 75.

et al. Synthesis and crystal chemistry of the hybrid perovskite (CH3NH3)PbI3 for solid-state sensitised solar cell applications. J. Mater. Chem. A 1, 5628–5641 (2013).

76. 76.

, , , & Complete structure and cation orientation in the perovskite photovoltaic methylammonium lead iodide between 100 and 352 K. Chem. Commun. 51, 4180–4183 (2015).

77. 77.

, & Molecular ferroelectric contributions to anomalous hysteresis in hybrid perovskite solar cells. APL Mater. 2, 081506 (2014).

78. 78.

et al. Polarization switching and light-enhanced piezoelectricity in lead halide perovskites. J. Phys. Chem. Lett. 6, 1408–1413 (2015).

79. 79.

, , , & Non-ferroelectric nature of the conductance hysteresis in CH3NH3PbI3 perovskite-based photovoltaic devices. Appl. Phys. Lett. 106, 173502 (2015).

80. 80.

et al. Inorganic caesium lead iodide perovskite solar cells. J. Mater. Chem. A 3, 19688–19695 (2015).

81. 81.

, & Ionic conduction of the perovskite-type halides. Solid State Ionics 11, 203–211 (1983).

82. 82.

Factors governing oxygen reduction in solid oxide fuel cell cathodes. Chem. Rev. 104, 4791–4843 (2004).

83. 83.

et al. Giant switchable photovoltaic effect in organometal trihalide perovskite devices. Nat. Mater. 14, 193–198 (2015).

84. 84.

, , & Defect migration in methylammonium lead iodide and its role in perovskite solar cell operation. Energy Environ. Sci. 8, 2118–2127 (2015).

85. 85.

et al. Mapping electric field-induced switchable poling and structural degradation in hybrid lead halide perovskite thin films. Adv. Energy Mater. 5, 1500962 (2015).

86. 86.

et al. Photovoltaic switching mechanism in lateral structure hybrid perovskite solar cells. Adv. Energy Mater. 5, 1500615 (2015).

87. 87.

et al. Mechanistic insights into perovskite photoluminescence enhancement: light curing with oxygen can boost yield thousandfold. Phys. Chem. Chem. Phys. 17, 24978–24987 (2015).

88. 88.

, , , & II Chemical management for colorful, efficient, and stable inorganic-organic hybrid nanostructured solar cells. Nano Lett. 13, 1764–1769 (2013).

89. 89.

et al. Monolithic perovskite/silicon-heterojunction tandem solar cells processed at low temperature. Energy Environ. Sci. 9, 81–88 (2016).

90. 90.

et al. Reversible photo-induced trap formation in mixed-halide hybrid perovskites for photovoltaics. Chem. Sci. 6, 613–617 (2015).

91. 91.

et al. High photoluminescence quantum yield in band gap tunable bromide containing mixed halide perovskites. Nano Lett. 16, 800–806 (2016).

92. 92.

et al. Solution synthesis approach to colloidal cesium lead halide perovskite nanoplatelets with monolayer-level thickness control. J. Am. Chem. Soc. 138, 1010–1016 (2016).

93. 93.

et al. Photo-induced halide redistribution in organic-inorganic perovskite films. Nat. Commun. 7, 11683 (2016).

94. 94.

& Defect identification in semiconductors with positron annihilation: experiment and theory. Rev. Mod. Phys. 85, 1583–1631 (2013).

95. 95.

& Defects in irradiated silicon: electron paramagnetic resonance and electron-nuclear double resonance of the Si–E center. Phys. Rev. 134, A1359–A1377 (1964).

96. 96.

et al. Boosting the performance of planar heterojunction perovskite solar cell by controlling the precursor purity of perovskite materials. J. Mater. Chem. A 4, 887–893 (2016).

97. 97.

et al. Enhanced optoelectronic quality of perovskite thin films with hypophosphorous acid for planar heterojunction solar cells. Nat. Commun. 6, 10030 (2015).

98. 98.

et al. Real-space imaging of the atomic structure of organic–inorganic perovskite. J. Am. Chem. Soc. 137, 16049–16054 (2015).

99. 99.

et al. Benefit of grain boundaries in organic–inorganic halide planar perovskite solar cells. J. Phys. Chem. Lett. 6, 875–880 (2015).

100. 100.

et al. Microscopic investigation of grain boundaries in organolead halide perovskite solar cells. ACS Appl. Mater. Interfaces 7, 28518–28523 (2015).

101. 101.

et al. CH3NH3PbI3 perovskite single crystals: surface photophysics and their interaction with the environment. Chem. Sci. 6, 7305–7310 (2015).

102. 102.

et al. Solvent annealing of perovskite-induced crystal growth for photovoltaic-device efficiency enhancement. Adv. Mater. 26, 6503–6509 (2014).

103. 103.

& Charge trapping in photovoltaically active perovskites and related halogenoplumbate compounds. J. Phys. Chem. Lett. 5, 1066–1071 (2014).

104. 104.

et al. Observation and mediation of the presence of metallic lead in organic-inorganic perovskite films. ACS Appl. Mater. Interfaces 7, 13440–13444 (2015).

105. 105.

et al. Controllable self-induced passivation of hybrid lead iodide perovskites toward high performance solar cells. Nano Lett. 14, 4158–4163 (2014).

106. 106.

, & Bulk and metastable defects in CuIn1−xGaxSe2 thin films using drive-level capacitance profiling. J. Appl. Phys. 95, 1000–1010 (2004).

107. 107.

Electronic Properties of Semiconductor Interfaces (Springer-Verlag Berlin Heidelberg, 2004).

108. 108.

Electron dynamics at surfaces. Surf. Sci. Rep. 21, 275–325 (1995).

109. 109.

et al. Formability of ABX3 (X = F, Cl, Br, I) halide perovskites. Acta Crystallogr. B 64, 702–707 (2008).

110. 110.

Mineral Commodity Summaries 2016 (US Geological Survey, 2016);

111. 111.

et al. Cesium-containing triple cation perovskite solar cells: improved stability, reproducibility and high efficiency. Energy Environ. Sci. 9, 1989–1997 (2016).

112. 112.

, , & Low-temperature processed meso-superstructured to thin-film perovskite solar cells. Energy Environ. Sci. 6, 1739–1743 (2013).

113. 113.

, , , & Efficient organometal trihalide perovskite planar-heterojunction solar cells on flexible polymer substrates. Nat. Commun. 4, 2761 (2013).

114. 114.

et al. Optical properties and limiting photocurrent of thin-film perovskite solar cells. Energy Environ. Sci. 8, 602–609 (2015).

115. 115.

et al. Economic assessment of solar electricity production from organic-based photovoltaic modules in a domestic environment. Energy Environ. Sci. 4, 3741–3753 (2011).

116. 116.

et al. First-principles calculations for point defects in solids. Rev. Mod. Phys. 86, 253–305 (2014).

## Acknowledgements

A.P. and J.M.B. thank the European Union Seventh Framework Programme (FP7/2007-2013) for funding under grant agreement no. 604032 of the MESO project.

## Affiliations

1. ### Center for Nanoscience and Technology, Italian Institute of Technology, via Giovanni Pascoli 70/3, Milano 20133, Italy.

• James M. Ball
•  & Annamaria Petrozza

## Authors

### Competing interests

The authors declare no competing financial interests.

## Corresponding author

Correspondence to Annamaria Petrozza.

### DOI

https://doi.org/10.1038/nenergy.2016.149

• ### Dimensional tailoring of hybrid perovskites for photovoltaics

• Giulia Grancini

Nature Reviews Materials (2019)

• ### Gas-solid reaction based over one-micrometer thick stable perovskite films for efficient solar cells and modules

• Zonghao Liu
• , Longbin Qiu
• , Emilio J. Juarez-Perez
• , Zafer Hawash
• , Taehoon Kim
• , Yan Jiang
• , Zhifang Wu
• , Sonia R. Raga
• , Luis K. Ono
• , Shengzhong Liu
•  & Yabing Qi

Nature Communications (2018)

• ### Accelerated discovery of stable lead-free hybrid organic-inorganic perovskites via machine learning

• Shuaihua Lu
• , Qionghua Zhou
• , Yixin Ouyang
• , Yilv Guo
• , Qiang Li
•  & Jinlan Wang

Nature Communications (2018)

• ### Defects engineering for high-performance perovskite solar cells

• Feng Wang
• , Sai Bai
• , Wolfgang Tress
• , Anders Hagfeldt
•  & Feng Gao

npj Flexible Electronics (2018)

• ### Dipolar cations confer defect tolerance in wide-bandgap metal halide perovskites

• Hairen Tan
• , Fanglin Che
• , Mingyang Wei
• , Yicheng Zhao
• , Makhsud I. Saidaminov
• , Petar Todorović
• , Danny Broberg
• , Grant Walters
• , Furui Tan
• , Taotao Zhuang
• , Bin Sun
• , Zhiqin Liang
• , Haifeng Yuan
• , Eduard Fron
• , Junghwan Kim
• , Zhenyu Yang
• , Oleksandr Voznyy
• , Mark Asta
•  & Edward H. Sargent

Nature Communications (2018)