Solution-based methods

The simplest method for depositing perovskites is precipitation from solution: the precursor metal halide and organic halide are dissolved, and evaporation of the solvent by heating or spincoating yields perovskite crystals. Common solvents include Lewis bases such as dimethylformamide (DMF), dimethylsulfoxide (DMSO), and gamma-butyrolactone. Little is known about the species present in solution, although solvents such as DMSO and DMF are known to coordinate to Pb²⁺ halide salts,^[Reference Wharf, Gramstad, Makhija and Onyszchuk^26–Reference Persson, Lyczko, Lundberg, Eriksson and Płaczek^28] and Pb²⁺ halide complexes are common in aqueous solutions of excess halide.^[Reference Haight and Peterson^29, Reference Clever and Johnston^30] The morphology of the resulting crystals and films produced from solution depends critically on the choice of solvent(s), the mode of their removal, and also on the substrate's morphology and surface chemistry.^[Reference Mitzi^31–Reference Im, Jang, Pellet, Grätzel and Park^36] For example, using DMSO in a mixture of solvents produces an intermediate phase after spincoating that then crystallizes into the perovskite upon annealing.^[Reference Jeon, Noh, Kim, Yang, Ryu and Il Seok^33] Forming continuous, pinhole-free films has proven to be challenging and remains a chief concern in fabricating solar cells of high efficiency.^[Reference Sum and Mathews^14] Addition of a hydrohalic acid to the precursor solution has offered some improvement in the film morphology, but the underlying reason remains unclear.^[Reference Heo, Song and Im^34, Reference Eperon, Stranks, Menelaou, Johnston, Herz and Snaith^37]

The single-step solution process has also been the method of choice for creating controllable mixtures of perovskites with higher band gaps than CH₃NH₃PbI₃ such as I–Br perovskites^[Reference Noh, Im, Heo, Mandal and Il Seok^38–Reference Sadhanala, Deschler, Thomas, Dutton, Goedel, Hanusch, Lai, Steiner, Bein, Docampo, Cahen and Friend^40] and Br–Cl perovskites.^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39, Reference Kitazawa, Watanabe and Nakamura^41] In these cases, the global stoichiometry of the resulting films appears to follow the stoichiometry of the precursors in solution, although chemical homogeneity on the nanoscale has not yet been confirmed.

In order to gain better control over the deposition of CH₃NH₃PbI₃, a two-step solution process was developed and has in general produced solar cells with higher efficiencies than the single-step process,^[Reference Burschka, Pellet, Moon, Humphry-Baker, Gao, Nazeeruddin and Grätzel^42] which could be due to the film's different carrier type and doping concentration (discussed in Section 5.1). In this approach, either PbI₂ or PbCl₂ is dissolved in DMF, spincoated onto a substrate, and then dried into a film. This film is then either dipped into an alcoholic solution of excess CH₃NH₃I alone or a CH₃NH₃I–CH₃NH₃Cl mixture^[Reference Burschka, Pellet, Moon, Humphry-Baker, Gao, Nazeeruddin and Grätzel^42, Reference Docampo, Hanusch, Stranks, Döblinger, Feckl, Ehrensperger, Minar, Johnston, Snaith and Bein^43] or spincoated repeatedly with such a solution to convert the lead-halide film into the pervoskite.^[Reference Kutes, Ye, Zhou, Pang, Huey and Padture^44] When such a process is applied to PbCl₂, CH₃NH₃PbI₃ is formed, presumably because according to the theory of hard and soft acids and bases^[Reference Pearson^45, Reference Pearson^46] the softer, more polarizable I⁻ has much greater affinity for Pb²⁺ than the much harder Cl⁻.^[Reference Shimoni-livny, Glusker and Bock^47] As this approach requires diffusion of the CH₃NH₃⁺ cations into the lead-halide matrix, it is possible that there is a thickness limitation on the films formed by this approach or higher temperatures or longer times are required to produce thicker films.^[Reference Burschka, Pellet, Moon, Humphry-Baker, Gao, Nazeeruddin and Grätzel^42]

Vapor-phase methods

Vapor-phase methods have also been used to produce hybrid perovskites, and their morphology and crystal structure differ from perovskites produced from solution. Co-evaporation of PbCl₂ and CH₃NH₃I in a vacuum deposition chamber yields highly uniform CH₃NH₃PbI₃ films with excellent photovoltaic characteristics.^[Reference Liu, Johnston and Snaith^4] In situ XRD^[Reference Pistor, Borchert, Fra, Csuk and Scheer^48] has been used to track the phase of the perovskite formed in a similar reaction and indicates that a large miscibility gap exists between CH₃NH₃PbI₃ and CH₃NH₃PbCl₃. The evaporation rates of the precursors determine the composition of the resulting perovskite, and the CH₃NH₃PbI₃ produced was confirmed to be cubic rather than the usual tetragonal phase that exists at room temperature.

Two-step processes involving vapor-phase reactions, similar to those performed in solution, have also been demonstrated. In the vapor-assisted solution process, PbI₂ is deposited from solution and then converted to the perovskite by exposure to CH₃NH₃I vapor.^[Reference Chen, Zhou, Hong, Luo, Duan, Wang, Liu, Li and Yang^49] Platelets of lead halides have also been synthesized directly from the evaporation of lead-halide powders and then converted to perovskite platelets by exposure to methylammonium halide vapor.^[Reference Ha, Liu, Zhang, Giovanni, Sum and Xiong^50] This two-step, vapor-phase method yields pure iodide, bromide, and chloride perovskites, as well as some mixtures. Interestingly, an I–Cl mixture was produced that exhibited blue-shifted photoluminescence (PL) (~750 nm) as compared with CH₃NH₃PbI₃ (~760 nm), suggesting that the miscibility gap between the two materials might be relaxed in these platelets, whose thicknesses range between 60 and 260 nm.

Role of Cl⁻ in the deposition of CH₃NH₃PbI₃

The earliest perovskite solar cells were produced from CH₃NH₃PbI₃, but it was later discovered that incorporating Cl⁻ during the crystallization process dramatically improved their photovoltaic performance.^[Reference Lee, Teuscher, Miyasaka, Murakami and Snaith^51] Subsequent studies confirmed the superior properties of CH₃NH₃PbI_3−xCl_x: its optical absorption is stronger near the band gap,^[Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza^52–Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca^54] it exhibits longer PL lifetimes,^[Reference Wehrenfennig, Eperon, Johnston, Snaith and Herz^55] and it yields nearly equal electron and hole diffusion lengths of ~1 μm,^[Reference Stranks, Eperon, Grancini, Menelaou, Alcocer, Leijtens, Herz, Petrozza and Snaith^56] as compared with the ~100–300 nm measured for typical CH₃NH₃PbI₃.^[Reference Xing, Mathews, Sun, Lim, Lam, Grätzel, Mhaisalkar and Sum^57]

At first it was proposed that incorporation of Cl⁻ into the film to form CH₃NH₃PbI_3−xCl_x was responsible for this enhancement; however, mounting evidence indicates that Cl⁻ might not be present in the bulk of the CH₃NH₃PbI₃ films synthesized with Cl⁻. From XRD, CH₃NH₃PbI_3−xCl_x has identical structure to CH₃NH₃PbI₃. In films deposited on planar substrates, no Cl⁻ has been detected via energy-dispersive x-ray spectroscopy (EDS), electron-energy loss spectroscopy or x-ray photoelectron spectroscopy (XPS).^[Reference Docampo, Hanusch, Stranks, Döblinger, Feckl, Ehrensperger, Minar, Johnston, Snaith and Bein^43, Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza^52, Reference Zhao and Zhu^58] The detection limit of the EDS measurement was reported as 0.1 wt%, corresponding to 2 at.%,^[Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza^52] and similar detection limits are typical for the other techniques; consequently, none of these techniques rules out incorporation at doping levels, which is still sufficient to play a key role in the passivation of defects. Examples of important trace elements in other photovoltaic technologies include chloride within cadmium telluride (CdTe) and sodium within CuIn_xGa_1−xSe₂ (CIGS).^[Reference Romeo, Terheggen, Abou-Ras, Bätzner, Haug, Kälin, Rudmann and Tiwari^59] In these cases, the impurities are found segregated near the surfaces and grain boundaries, so a similar process of passivation might be at work in the perovskites. Additionally, a recent study using depth-sensitive, angle-resolved XPS on ultrathin films deposited on compact TiO₂ suggests that the Cl⁻ is localized to the TiO₂–perovskite interface,^[Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca^54] which could explain why Cl⁻ was detected via EDS in the perovskite deposited on mesoporous Al₂O₃.^[Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza^52] If it is indeed localized to the interface, then the high surface area of meso-structured templates can enhance the signal above the limit of detection.

Several hypotheses have been proposed to explain the improved performance of CH₃NH₃PbI_3−xCl_x. Since the film's morphology changes from its Cl-free counterpart, one proposal is that an insoluble lead-chloride compound acts as a nucleation site for the iodide perovskite, enhancing its nucleation rate and density; this change in morphology determines the film's properties.^[Reference Williams, Zuo, Chueh, Liao, Liang and Jen^22, Reference Tidhar, Edri, Weissman, Zohar, Hodes, Cahen, Rybtchinski and Kirmayer^60, Reference Yu, Wang, Xie, Li, Chen and Zhao^61] The location of the Cl⁻ at the perovskite-scaffold interface is consistent with this hypothesis. Other work suggests that growing the film in a CH₃NH₃⁺-rich or I⁻-poor environment changes the defect structure of the film, which improves the performance.^[Reference Yu, Wang, Xie, Li, Chen and Zhao^61, Reference Buin, Pietsch, Xu, Voznyy, Ip, Comin and Sargent^62]

Intrinsic electrical properties

The starting point for characterizing materials electrically is always quantitative determination of properties such as carrier type, concentration, and mobility; however, in the case of perovskites, these properties have exhibited a large range of values often influenced by the method used to prepare the perovskite. Additionally, the lack of smooth and uniform films on which to perform measurements can make the determination of intrinsic electrical properties challenging since conventional techniques often assume the sample exhibits a specific geometry.

Thermoelectric measurements of the Seebeck coefficient, Hall measurements of the conductivity's response to an applied magnetic field, or a thin-film transistor's response to a gating electric field are typical methods used to determine the carrier type. Early resistivity, Seebeck, and Hall measurements on polycrystalline CH₃NH₃PbI₃ indicated n-type conductivity, a carrier concentration of ~10⁹ cm⁻³, and an electron mobility of 66 cm²/V/s.^[Reference Stoumpos, Malliakas and Kanatzidis^24] A more detailed study determined that tuning the stoichiometry of the precursors during solution-phase synthesis can adjust the carrier concentration and even switch the carrier type to p-type when excess CH₃NH₃I is used, which is the case in two-step syntheses; it is proposed that iodide vacancies are responsible for the n-type conductivity, and electron concentrations of ~10¹⁷–10¹⁸ cm⁻³ were measured.^[Reference Wang, Shao, Xie, Lyu, Liu, Gao and Huang^68] From Hall measurements, the electron mobility for n-type films deposited from stoichiometric precursors was determined to be 3.9 cm²/V/s,^[Reference Wang, Shao, Xie, Lyu, Liu, Gao and Huang^68] which is in accord with the 8 cm²/V/s measured using terahertz spectroscopy,^[Reference Wehrenfennig, Eperon, Johnston, Snaith and Herz^55] a technique that usually provides an upper limit because it neglects long-range carrier scattering^[Reference Wang, Shao, Xie, Lyu, Liu, Gao and Huang^68]; however, it is not surprising that the mobility depends strongly on the preparation of the film. In the most extreme case, for instance, CH₃NH₃SnI₃ prepared by a solid-state reaction in a vacuum-sealed tube shows an electron mobility of 2320 cm²/V/s,^[Reference Stoumpos, Malliakas and Kanatzidis^24] which is much larger than the 200 cm²/V/s measured for solution-processed material.^[Reference Takahashi, Obara, Lin, Takahashi, Naito, Inabe, Ishibashi and Terakura^69]

The electron mobility of polycrystalline CH₃NH₃PbI₃ films compares favorably to that of films of other materials used as absorbers in solar cells. It is larger than the thin-film mobility of polymers (10⁻⁷–1 cm²/V/s)^[Reference Venkateshvaran, Nikolka, Sadhanala, Lemaur, Zelazny, Kepa, Hurhangee, Kronemeijer, Pecunia, Nasrallah, Romanov, Broch, McCulloch, Emin, Olivier, Cornil, Beljonne and Sirringhaus^70, Reference You, Dou, Hong, Li and Yang^71] and colloidal quantum dots (10⁻³–1 cm²/V/s),^[Reference Sandeep, Cate, Schins, Savenije, Liu, Law, Kinge, Houtepen and Siebbeles^72] and it is comparable with that of CdTe (10 cm²/V/s),^[Reference Long, Dinca, Schiff, Yu and Theil^73] CIGS and Cu₂ZnSnS₄ (CZTS) (10–10² cm²/V/s),^[Reference Shin, Gunawan, Zhu, Bojarczuk, Chey and Guha^74, Reference Brown, Faifer, Pudov, Anikeev, Bykov, Contreras and Wu^75] and polycrystalline Si (40 cm²/V/s).^[Reference Kamins^76] Perhaps in its monocrystalline form, the intrinsic mobility might be similar to that of monocrystalline Si (10²–10³ cm²/V/s)^[Reference Green^77] or GaAs (>10³ cm²/V/s),^[Reference Blakemore^78] but further studies must be performed on single crystals synthesized using a variety of methods. Even in polycrystalline form, hybrid perovskites’ inexpensive processing and tolerance to defects offer a significant advantage over conventional semiconductors.

As is the case for other polycrystalline semiconductors, electrical properties in hybrid perovskites are likely correlated with the film's morphology. For instance, the dark and light conductivities of CH₃NH₃PbI_3−xCl_x deposited on a planar scaffold or on mesostructured aluminum oxide are quite different.^[Reference Leijtens, Stranks, Eperon, Lindblad, Johansson, Ball, Lee, Snaith and McPherson^79] This difference has been attributed to an increase in the perovskite's Fermi level in the mesostructured scaffold either through more surface iodide vacancies or through electrostatic gating from the aluminum oxide. Also, the role of grain boundaries in conduction through perovskite films has not been thoroughly explored, although passivation of grain boundaries with PbI₂ has been correlated to increased radiative lifetimes.^[Reference Chen, Zhou, Song, Luo, Hong, Duan, Dou, Liu and Yang^80] Inspired by the literature on organic solar cells, solvent annealing has been applied to CH₃NH₃PbI₃ solar cells to increase the grain size of the films to ~1 μm.^[Reference Xiao, Dong, Bi, Shao, Yuan and Huang^35] Such processing results in an increase in the photovoltaic performance and radiative lifetime.

Extrinsic electrical properties of operating solar cells

To investigate the electrical properties of perovskites in devices, techniques such as impedance spectroscopy (IS)^[Reference Dualeh, Moehl, Tétreault, Teuscher, Gao, Nazeeruddin and Grätzel^81, Reference Gonzalez-Pedro, Juarez-Perez, Arsyad, Barea, Fabregat-santiago, Mora-Sero and Bisquert^82] and electron beam-induced current (EBIC)^[Reference Edri, Kirmayer, Mukhopadhyay, Gartsman, Hodes and Cahen^83, Reference Edri, Kirmayer, Henning, Mukhopadhyay, Gartsman, Rosenwaks, Hodes and Cahen^84] have been applied. IS is typically used to identify the frequency dependence of capacitance, to measure charge diffusion lengths and lifetimes, and to investigate carrier trapping and recombination. An equivalent circuit that has been used successfully to analyze DSSCs can be adapted to identify characteristics of perovskite solar cells; however, the model should be taken from a solid-state DSSC, which also uses a hole-transport layer, such as 2,2′,7,7′-tetrakis(N,N-di-p-methoxyphenylamine)-9,9′-spirobifluorene (spiro-OMeTAD), instead of redox ions in a liquid electrolyte.^[Reference Dualeh, Moehl, Tétreault, Teuscher, Gao, Nazeeruddin and Grätzel^81] Capacitance values extracted from IS of perovskite solar cells show evidence of significant charge accumulation in the perovskite in the dark and under illumination,^[Reference Kim, Mora-Sero, Gonzalez-Pedro, Fabregat-Santiago, Juarez-Perez, Park and Bisquert^85] which could arise from trap states within the material. Such an interpretation is consistent with Kelvin probe force microscopy measurements that found long-lived trapped charges within the perovskite after illumination that accumulate because of poor transport through the hole-transport layer.^[Reference Bergmann, Weber, Ramos, Nazeeruddin, Grätzel, Li, Domanski, Lieberwirth, Ahmad and Berger^86] These trapped charges are one possible source of the hysteresis often seen in the operation of perovskite solar cells (discussed below in Section 5.3).

In addition to the capacitance, the carrier diffusion length can also be derived from IS and has been estimated to be about ~1 μm for CH₃NH₃PbI_3−xCl_x.^[Reference Gonzalez-Pedro, Juarez-Perez, Arsyad, Barea, Fabregat-santiago, Mora-Sero and Bisquert^82] This result is similar to values obtained using EBIC (1.5–1.9 μm)^[Reference Edri, Kirmayer, Mukhopadhyay, Gartsman, Hodes and Cahen^83] and PL quenching (~1 μm).^[Reference Stranks, Eperon, Grancini, Menelaou, Alcocer, Leijtens, Herz, Petrozza and Snaith^56] The carrier diffusion length in CH₃NH₃PbI_3−xCl_x is comparable with or longer than that of other polycrystalline semiconductors with direct band gaps used in solar cells, such as CZTS (350–450 nm),^[Reference Shin, Gunawan, Zhu, Bojarczuk, Chey and Guha^74] CIGS (300–900 nm),^[Reference Brown, Faifer, Pudov, Anikeev, Bykov, Contreras and Wu^75] and CdTe thin films (0.4–1.6 μm).^[Reference Tarricone, Romeo, Sberveglier and Mora^87] It is also much longer than the typical excitonic diffusion lengths in other solution-processed technologies such as conducting polymers (5–10 nm),^[Reference Mikhnenko, Azimi, Scharber, Morana, Blom and Loi^88] and films of colloidal quantum dots (100 nm).^[Reference Koleilat, Levina, Shukla, Myrskog, Hinds, Pattantyus-abraham and Sargent^89]

Origin of electrical hysteresis in hybrid perovskites

In the dark and light, CH₃NH₃PbI₃ devices exhibit distinct hysteresis behavior in both their resistivity versus temperature dependence (R–T characteristic) and current versus voltage dependence [I–V characteristic, Fig. 4(a)].^[Reference Stoumpos, Malliakas and Kanatzidis^24] While the hysteresis on the R–T curve has been associated with a structural phase transition in the perovskite, the cause of the hysteresis in the I–V curve at room temperature is still unclear (Fig. 4). Additionally, CH₃NH₃PbI₃ exhibits a very slow transient (~10–100 s) to reach a steady state in current or voltage [Fig. 4(b)].^[Reference Gottesman, Haltzi, Gouda, Tirosh, Bouhadana, Zaban, Mosconi and De Angelis^90–Reference Unger, Hoke, Bailie, Nguyen, Bowring, Heumuller, Christoforo and McGehee^92] This time scale has been verified from the characteristic times measured by transient photovoltage,^[Reference Sanchez, Gonzalez-Pedro, Lee, Park, Kang, Mora-Sero and Bisquert^93] transient PL,^[Reference Sanchez, Gonzalez-Pedro, Lee, Park, Kang, Mora-Sero and Bisquert^93] and impedance spectroscopies.^[Reference Dualeh, Moehl, Tétreault, Teuscher, Gao, Nazeeruddin and Grätzel^81, Reference Kim, Mora-Sero, Gonzalez-Pedro, Fabregat-Santiago, Juarez-Perez, Park and Bisquert^85, Reference Sanchez, Gonzalez-Pedro, Lee, Park, Kang, Mora-Sero and Bisquert^93] Some variation in the time scale results from changing the crystal growth process. For example, the single-step solution deposition gives transient times that are three orders of magnitude slower than the two-step process.^[Reference Sanchez, Gonzalez-Pedro, Lee, Park, Kang, Mora-Sero and Bisquert^93] Several explanations have been proposed for this behavior, including ferroelectricity, accumulation of charges at grain boundaries, filling of interface or surface trap states, and ionic migration [Figs. 4(c) and 4(d)].^[Reference Snaith, Abate, Ball, Eperon, Leijtens, Noel, Stranks, Wang, Wojciechowski and Zhang^91] Distinguishing between these mechanisms is an area of ongoing research.

Figure 4. Hysteresis in the electrical transport in hybrid perovskites. (a) I–V measurement taken at room temperature on a single crystal of CH₃NH₃PbI₃ exhibiting pronounced hysteresis. (b) The I–V characteristic (left) of a perovskite solar cell depends upon the rate at which the voltage is scanned: in this specific case, more hysteresis is observed for the faster scan rate. The photocurrent stabilizes with a long time constant of 7.2 s (right). Schematic hysteretic I–V curve (c) and phenomena proposed for its origin (d). The as-synthesized perovskite exists in a disordered state (1). When a bias is applied (2), the electric field creates a change in the material, as depicted in (d) for the three proposed mechanisms. Upon return of the field back to 0 V (3), the arrangement of dipoles, trapped charges, or ions has been altered from the initial state, so current continues to flow as this configuration relaxes, creating hysteresis. The material returns to its initial state (1′) upon application of a negative bias. The same process with reversed polarity occurs for points 2′, 3′, and 1″. Panel (a) is reproduced with permission from the supporting information from Ref. Reference Stoumpos, Malliakas and Kanatzidis24. Copyright 2013, American Chemical Society. Panel (b) is reproduced with permission from Ref. Reference Unger, Hoke, Bailie, Nguyen, Bowring, Heumuller, Christoforo and McGehee92. Copyright 2014, Royal Society of Chemistry.

Understanding ferroelectricity within a hybrid perovskite requires consideration of the crystal symmetry as well as the incorporated CH₃NH₃⁺ cations. From a structural perspective, the tetragonal phase of CH₃NH₃PbI₃ does allow for a polar ferroelectric distortion, whereas the cubic phases of the chloride and bromide perovskites do not. Unlike conventional oxide perovskites, however, hybrid perovskites include a molecular dipole, CH₃NH₃⁺, which reorients freely on the picosecond timescale at room temperature.^[Reference Swainson, Chi, Her, Cranswick, Stephens, Winkler, Wilson and Milman^18–Reference Knop, Wasylishen, White, Stanley and Michiel^20] The molecule is not spherically symmetric like an atom, but its unimpeded reorientations at room temperature impart a pseudo-spherical symmetry. It is hypothesized that application of an external electric field stabilizes a state in which the dipoles are aligned and the inorganic lattice is distorted; therefore, the long-time constant in the electrical measurements arises from the slow polarization of the inorganic lattice that must occur because of its interaction with the organic cations.^[Reference Gottesman, Haltzi, Gouda, Tirosh, Bouhadana, Zaban, Mosconi and De Angelis^90] It is possible that such an electric-field-induced structural change might be measured by in situ XRD, as has been demonstrated in certain oxide perovskites.^[Reference Noheda, Zhong, Cox, Shirane, Park and Rehrig^94] If this mechanism is correct, the hysteresis should depend on the magnitude of the dipole moment of the organic cation and its coupling to the halide cage via hydrogen bonding; perovskites synthesized from various cations with different dipoles could be tested and compared.^[Reference Frost, Butler, Brivio, Hendon, van Schilfgaarde and Walsh^95]

Some experimental evidence supports the existence of ferroelectricity in CH₃NH₃PbI₃. Ferroelectric domains have been mapped using piezoforce microscopy on solution-processed CH₃NH₃PbI₃.^[Reference Kutes, Ye, Zhou, Pang, Huey and Padture^44] Also, ferroelectricity would explain the large static dielectric constant (ε ₀) of CH₃NH₃PbI₃ of ~1000, although grain boundaries could also be responsible for it.^[Reference Juarez-Perez, Sanchez, Badia, Garcia-Belmonte, Kang, Mora-Sero and Bisquert^96] In typical ferroelectrics, a discontinuity in the dielectric constant marks the ferroelectric to paraelectric phase transition. For hybrid perovskites such a singularity has been observed only at the phase transition at which the CH₃NH₃⁺ cations become ordered,^[Reference Onoda-Yamamuro, Matsuo and Suga^19] which is well below room temperature. These measurements were performed at 20 Hz, however, so it is likely that they were too fast to allow the polarization change, which occurs on the 10–100 s timescale, to be realized, essentially removing this contribution to the dielectric constant. Since the change in dielectric constant at the phase transition is one of the hallmarks of a ferroelectric material, repetition of these measurements at sufficiently low frequencies would do much to clarify this issue.

If CH₃NH₃PbI₃ is ferroelectric, knowledge gained from the extensive study of oxide ferroelectrics can inform the interpretation of effects seen in perovskites. For example, it has been proposed that ferroelectric domains could be responsible for the excellent performance of perovskite photovoltaics by aiding in charge separation.^[Reference Frost, Butler, Brivio, Hendon, van Schilfgaarde and Walsh^95] A similar mechanism has been demonstrated experimentally in BiFeO₃^[Reference Yang, Seidel, Byrnes, Shafer, Yang, Rossell, Yu, Chu, Scott, Ager, Martin and Ramesh^97] and was later explained in detail.^[Reference Seidel, Fu, Yang, Alarcón-Lladó, Wu, Ramesh and Ager^98]

In addition to ferroelectricity, charge accumulation at grain boundaries, ionic migration, and trapping of charges at surfaces or interfaces, have also been proposed as possible sources of the hysteresis.^[Reference Snaith, Abate, Ball, Eperon, Leijtens, Noel, Stranks, Wang, Wojciechowski and Zhang^91] The reported large static dielectric constant is consistent with effects from grain boundaries,^[Reference Juarez-Perez, Sanchez, Badia, Garcia-Belmonte, Kang, Mora-Sero and Bisquert^96, Reference Lunkenheimer, Bobnar, Pronin, Ritus, Volkov and Loidl^99] and measuring the dielectric response of single crystals with varying contact geometries to avoid artifacts from trapped charges at electrode interfaces^[Reference Lunkenheimer, Bobnar, Pronin, Ritus, Volkov and Loidl^99, Reference Pontes, Leite, Longo, Varela, Araujo and Eiras^100] would provide insight into the role of grain boundaries. Also, no hysteresis was detected in solvent-annealed solar cells with grain sizes large enough to permit pseudo-single crystal transport in the through-plane direction.^[Reference Xiao, Dong, Bi, Shao, Yuan and Huang^35] Such an observation cannot rule out a ferroelectric mechanism, however, because the switching of ferroelectric domains is known, at least in oxide perovskites, to be highly sensitive to grain size and morphology.^[Reference Shaw, Trolier-McKinstry and McIntyre^101] In the case of ionic migration, a first step would be to determine the contribution of ionic conduction by comparing frequency-dependent AC and DC resistivity measurements, which would suggest whether either halide anions or CH₃NH₃⁺ cations are mobile and available to screen the electric field. Related inorganic perovskites, CsPbCl₃ and CsPbBr₃ have been determined to be halide-ion conductors, as are the precursor halides PbCl₂ and PbBr₂,^[Reference Mizusaki, Arai and Fueki^102] which have sometimes been found by XRD to contaminate the perovskite films. Cycling of a large bias (±2.5 V) applied to perovskite solar cells has been shown to produce switchable photocurrent and photovoltage, possibly explained by the reversible migration of vacancies that leads to a bias-induced p–i–n doping structure.^[Reference Xiao, Yuan, Shao, Wang, Dong, Bi, Sharma, Gruverman and Huang^103] Hysteresis in perovskites has also been shown to be sensitive to the contacting scheme employed for the perovskite film, which suggests that it might be related at least in part to interfacial trap states or to the filling and emptying of bulk trap states.^[Reference Snaith, Abate, Ball, Eperon, Leijtens, Noel, Stranks, Wang, Wojciechowski and Zhang^91]

Optical constants

Although tuning the band gap of perovskites is well documented, a more detailed understanding of these materials’ optical properties awaits further research. For instance, dielectric constants in the ultraviolet, visible, and near-IR spectral regions are critical to understanding the optical response of perovskites and also to calculating their absorption and emission properties when incorporated into optoelectronic devices. Progress in this area has been hindered by the difficulty of producing continuous films of sufficient smoothness^[Reference Sum and Mathews^14] to avoid measurement artifacts from spectroscopic measurements of reflectance, transmittance, and ellipsometry. Recent work using ellipsometry to characterize films optimized to reduce roughness, and which accounts for remaining roughness via modeling, has produced the most reliable optical constants of CH₃NH₃PbI₃,^[Reference Stuckelberger, Niesen, Filipic, Moon, Yum, Topic and Ballif^104] although they still remain to be independently confirmed. These latest values differ considerably from previous preliminary measurements on CH₃NH₃PbI₃ films,^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39, Reference Lin, Armin, Nagiri, Burn and Meredith^105] which is not surprising since earlier measurements did not account for the film's morphology. Quantitative absorption coefficients have been determined from the absorption of CH₃NH₃PbI₃ films on quartz^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39] and glass,^[Reference Sutherland, Hoogland, Adachi, Kanjanaboos, Wong, McDowell, Xu, Voznyy, Ning, Houtepen and Sargent^25] yielding values of ~10⁴ cm⁻¹ near the band edge, but no detailed data of the film's morphology has been provided and no corrections for the surface's inhomogeneity have been applied; consequently, the reported values are only preliminary, but they are consistent with the absorption coefficients calculated based on the optical constants of CH₃NH₃PbI₃.^[Reference Stuckelberger, Niesen, Filipic, Moon, Yum, Topic and Ballif^104] Additionally, the absorption spectrum for CH₃NH₃PbI₃ when it is deposited within a mesoscopic template differs from that of perovskite deposited on planar substrates, which has been attributed to changes in the crystallite morphology that affect the optical transitions [Fig. 5(a)].^[Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza^52] Band structure calculations indicate that this change in morphology affects the dipole screening of the excitonic transition.^[Reference Even, Pedesseau, Jancu and Katan^64] Such differences further complicate accurate determination of the absorption coefficient and other optical parameters of perovskite films, so control over the materials properties of perovskites is of the greatest importance for characterization and applications.

Figure 5. Absorption and PL spectra (a) and resonant Raman spectra (b) of CH₃NH₃PbI₃ deposited on flat substrates and in mesoporous substrates. Clearly the optical properties of the material depend sensitively on the material's structure, which is determined in part by the substrate on which it is deposited. Solid lines in (b) are the sums of the individual peaks (dotted lines) used to fit the spectra. The figure is reproduced with permission from Ref. Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza52. Copyright 2014, American Chemical Society.

Excitons

The role of excitons in perovskites has been a topic of considerable debate. Recent studies indicate, however, that no significant population of excitons is present in photovoltaics made from CH₃NH₃PbI₃, whose exciton-binding energy has generally been reported between 20 and 50 meV and is therefore comparable with the thermal energy at room temperature (k _BT = 26 meV).^[Reference D'Innocenzo, Grancini, Alcocer, Kandada, Stranks, Lee, Lanzani, Snaith and Petrozza^53, Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] These values have been obtained by fitting temperature-dependent absorption spectra using the measured^[Reference Tanaka, Takahashi, Ban, Kondo, Uchida and Miura^107] reduced mass of the exciton. Estimating the excitonic radius from the binding energy, however, requires knowledge of the appropriate^[Reference Huang and Lambrecht^65] dielectric constant, which continues to be the subject of debate.^[Reference Lin, Armin, Nagiri, Burn and Meredith^105] Although the population of excitons is small in photovoltaics and remains small even at the higher excitation densities required for stimulated emission,^[Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] the excitonic transition significantly enhances the absorption of hybrid perovskites near the band edge; consequently, the electronic band gap taking this into account is 1.65 eV for CH₃NH₃PbI₃.^[Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] The density of excitons and the associated effects should be greater in perovskites with higher band gaps, such as CH₃NH₃PbBr₃, whose binding energy has been estimated to be 76 meV.^[Reference Tanaka, Takahashi, Ban, Kondo, Uchida and Miura^107]

Photoluminescence

PL has been observed in perovskites of pure and mixed compositions. PL efficiency depends strongly upon pump fluence. At low excitation intensities, trapping of photogenerated charges competes effectively with direct radiative recombination of electrons and holes, reducing luminescence.^[Reference Deschler, Price, Pathak, Klintberg, Jarausch, Higler, Hu, Leijtens, Stranks, Snaith, Atatu, Phillips and Friend^8, Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106, Reference Noel, Abate, Stranks, Parrott, Burlakov, Goriely, Snaith and Al^108, Reference Stranks, Burlakov, Leijtens, Ball, Goriely and Snaith^109] As the excitation increases, these traps are filled and radiative electron–hole recombination dominates. In this regime, studies have confirmed low trap-induced recombination and two-body recombination dynamics over a range of pump intensities.^[Reference Wehrenfennig, Liu, Snaith, Johnston and Herz^110, Reference Manser and Kamat^111] High PL efficiency, ~70% for CH₃NH₃PbI_3−xCl_x,^[Reference Deschler, Price, Pathak, Klintberg, Jarausch, Higler, Hu, Leijtens, Stranks, Snaith, Atatu, Phillips and Friend^8, Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] and ~17–30% for CH₃NH₃PbI₃,^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39, Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] is achieved because amplified spontaneous emission begins at a threshold of only ~10 μJ/cm², which corresponds to an injected carrier density of ~10¹⁸ cm⁻³ or three orders of magnitude greater than the excitation expected under solar illumination.^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39, Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] At higher pumping, the PL efficiency falls as Auger recombination becomes more competitive at the higher carrier densities.^[Reference Sum and Mathews^14, Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39, Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106]

PL lifetime measurements have been reported for CH₃NH₃PbI₃ and CH₃NH₃PbI_3−xCl_x, with longer lifetimes exhibited by the latter. In general, it is difficult to compare lifetimes directly unless they are measured at the same pump fluence, and the wide range of reported lifetimes for each material is likely due to a combination of varying excitation density and synthetic procedures. The PL lifetime of CH₃NH₃PbI₃ has been reported between 3 and 18 ns,^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39, Reference Ha, Liu, Zhang, Giovanni, Sum and Xiong^50, Reference Pellet, Gao, Gregori, Yang, Nazeeruddin, Maier and Grätzel^112] while CH₃NH₃PbI_3−xCl_x exhibits considerably longer lifetimes from 91 to 341 ns at low pump fluences.^[Reference Docampo, Hanusch, Stranks, Döblinger, Feckl, Ehrensperger, Minar, Johnston, Snaith and Bein^43, Reference Stranks, Eperon, Grancini, Menelaou, Alcocer, Leijtens, Herz, Petrozza and Snaith^56, Reference Noel, Abate, Stranks, Parrott, Burlakov, Goriely, Snaith and Al^108, Reference Wehrenfennig, Liu, Snaith, Johnston and Herz^110] Reports that measure both materials under the same excitation conditions indicate an increase from 18 to 91 ns^[Reference Docampo, Hanusch, Stranks, Döblinger, Feckl, Ehrensperger, Minar, Johnston, Snaith and Bein^43] and 10 to 273 ns.^[Reference Stranks, Eperon, Grancini, Menelaou, Alcocer, Leijtens, Herz, Petrozza and Snaith^56] This increase in lifetime correlates well with the factor of ten increase in carrier diffusion length, indicating that lifetime rather than mobility is improved by the addition of Cl⁻; measurements have confirmed that the mobility increases only slightly from 8 to 11.6 cm²/V/s for CH₃NH₃PbI₃ and CH₃NH₃PbI_3−xCl_x, respectively.^[Reference Wehrenfennig, Eperon, Johnston, Snaith and Herz^55] For solution-processed polycrystalline semiconductors, such lifetimes are surprising, especially considering that essentially nothing has been done to reduce recombination at surfaces or grain boundaries. Passivation of grain boundaries with PbI₂^[Reference Chen, Zhou, Song, Luo, Hong, Duan, Dou, Liu and Yang^80] or Lewis bases^[Reference Noel, Abate, Stranks, Parrott, Burlakov, Goriely, Snaith and Al^108] or using solvent annealing^[Reference Xiao, Dong, Bi, Shao, Yuan and Huang^35] to increase grain size have improved the PL lifetime, with lifetimes reaching 2 μs in the case of passivation with Lewis bases. Such long radiative lifetimes in a semiconductor with a direct band gap have been seen in undoped and surface-passivated GaAs films.^[Reference Ahrenkiel^113] In this case, photon recycling causes the PL lifetime to be limited by surface recombination rather than radiative recombination. Considering that the perovskites exhibit clear self-absorption in their PL^[Reference Saba, Cadelano, Marongiu, Chen, Sarritzu, Sestu, Figus, Aresti, Piras, Geddo Lehmann, Cannas, Musinu, Quochi, Mura and Bongiovanni^106] and lasing spectra,^[Reference Xing, Mathews, Lim, Yantara, Liu, Sabba, Grätzel, Mhaisalkar and Sum^39] it would not be surprising for photon recycling to play a role in their excited state dynamics when pathways of non-radiative decay are adequately suppressed.

IR and Raman spectroscopy

Probing the band gap with visible light offers information about the electronic structure of these compounds, but IR spectroscopy can provide a more detailed chemical understanding. At room temperature, CH₃NH₃PbI₃ is tetragonal, while CH₃NH₃PbBr₃ and CH₃NH₃PbCl₃ are cubic. Symmetry of the lattice in principle precludes Raman-active modes for the cubic crystals,^[Reference Maalej, Abid, Kallel, Daoud, Lautié and Romain^114, Reference Calistru, Mihut, Lefrant and Baltog^115] although a weak, broad band has been observed about 66 cm⁻¹ in CH₃NH₃PbCl₃ that is perhaps allowed by disorder within the material.^[Reference Maalej, Abid, Kallel, Daoud, Lautié and Romain^114] For tetragonal CH₃NH₃PbI₃, a recent study comparing the resonant Raman spectrum with density-functional theory (DFT) calculations based on a harmonic approximation has identified modes below approximately 100 cm⁻¹ related to the inorganic octahedron as well as higher energy modes indicative of the disorder of the CH₃NH₃⁺ cations [Fig. 5(b)].^[Reference Grancini, Marras, Prato, Giannini, Quarti, De Angelis, De Bastiani, Eperon, Snaith, Manna and Petrozza^52, Reference Quarti, Grancini, Mosconi, Bruno, Ball, Lee, Snaith, Petrozza and De Angelis^116] Although calculations and comparison with the literature of related compounds have informed the interpretation of the spectrum, further investigation of how these modes shift with structural changes in the crystal can be used to confirm the assignments. The temperature-dependent IR spectra of CH₃NH₃⁺ within all three perovskites were investigated^[Reference Onoda-Yamamuro, Matsuo and Suga^117] and supported the earlier conclusion that the cations remain disordered until phase transitions below room temperature.^[Reference Poglitsch and Weber^21] A detailed description of the IR spectrum below 850 cm⁻¹, where the inorganic bands would be expected to appear and have been documented in CsPbCl₃,^[Reference Hirotsu^118] has not been reported for the hybrid perovskites. Once fundamental assignments of modes have been confirmed, Raman and IR spectroscopy can become tools with which to study structure–function relationships in perovskites. In particular, spatial mapping of Raman modes has the potential to offer a richer understanding of the inhomogeneity of perovskite films with sub-micron spatial resolution.

Solar cells: better materials for transport layers and contacts

Current high-efficiency perovskite solar cells can be further optimized using appropriate contacts to extract photogenerated charges efficiently. Present designs typically sandwich the perovskite between a hole-transport layer and electron-transport layer, which are then contacted by a metal or a transparent conducting contact. Important characteristics for the selective layers include: (1) suitable band alignment that permits the flow of one carrier and blocks the other; (2) electrical conductivity and in particular a charge mobility that is comparable with that of the perovskite; (3) a homogeneous and intimate contact with the perovskite, which might require a conformal deposition technique; (4) chemical stability against reactions with the perovskite; and (5) transparency of at least one contact to allow the absorption of light.

To create carrier-selective contacts, the use of metal oxides and metal sulfides is promising because they are stable in air and have charge mobilities higher than those of the organic semiconductors currently being used. In addition, metal oxides and sulfides have been used in most commercial thin-film solar cells and have proven better in most cases for organic photovoltaics (OPVs). Typical electron-transport layers in perovskite solar cells are metal oxides such as TiO₂, ZnO, Al₂O₃, and ZrO₂ or fullerene derivatives such as 6,6-phenyl C₆₁ butyric acid methyl ester (PCBM). Since metal sulfides have electron mobilities that are two to three orders of magnitude higher than those of the standard oxides, they are an attractive alternative. Candidates include CdS, ZnS, Bi₂S₃, MoS₂, and Sb₂S₃.^[Reference Ito, Tanaka, Manabe and Nishino^119] Care must be taken to keep the layers thin because these metal sulfides absorb visible light, with the exception of ZnS.

The mobility of the hole-transport layer also needs to be improved, and metal oxides and sulfides can contribute here as well. Highly efficient perovskite cells currently use organic semiconductors with high work functions, such as spiro-OMeTAD,^[Reference Lee, Teuscher, Miyasaka, Murakami and Snaith^51] poly(3-hexylthiophene) (P3HT),^[Reference Abrusci, Stranks, Docampo, Yip, Jen and Snaith^120] poly(3,4-ethylenedioxythiophene)-polystyrenesulfonic acid (PEDOT:PSS),^[Reference You, Hong, Yang, Chen, Cai, Song, Chen, Lu, Liu, Zhou and Yang^121] or other polymer derivatives such as carbazoles,^[Reference Xu, Sheibani, Liu, Zhang, Tian, Vlachopoulos, Boschloo, Kloo, Hagfeldt and Sun^122] polyfluorenes,^[Reference Zhu, Bai, Lee, Mu, Zhang, Zhang, Wang, Yan, So and Yang^123] pyrenes,^[Reference Jeon, Lee, Noh, Nazeeruddin and Il Seok^124] and triarylamines.^[Reference Ryu, Noh, Jeon, Chan Kim, Yang, Seo and Il Seok^125] All of them have much lower hole mobility (μ_h = 10 ⁻⁶–10⁻² cm²/V/s) than CH₃NH₃PbI₃, except for PEDOT:PSS (μ_h = 1 cm²/V/s); however, PEDOT:PSS is not ideal because of its inability to block electrons, acidity, and tendency to absorb water. The latter two disadvantages are often associated with degradation in OPVs^[Reference De Jong, Van Ijzendoorn and De Voigt^126, Reference Voroshazi, Verreet, Buri, Müller, Di Nuzzo and Heremans^127] and would likely be similarly detrimental to perovskite solar cells. In contrast, MoS₂ can have carrier mobilities in excess of 100 cm²/V/s and is chemically less reactive because of its layered crystal structure.^[Reference Kim, Konar, Hwang, Lee, Lee, Yang, Jung, Kim, Yoo, Choi, Jin, Lee, Jena, Choi and Kim^128] MoS_xSe_1−x has been successful as a hole contact in CIGS solar cells, and its valence band is also expected to align favorably with that of the perovskites. Moreover metal oxides such as NiO^[Reference Zhu, Bai, Zhang, Liu, Long, Wei, Wang, Zhang, Wang, Yan and Yang^129] and MoO_x^[Reference Zhao, Nardes and Zhu^130, Reference Shrotriya, Li, Yao, Chu and Yang^131] have also been used with perovskites, and other metal oxides that have higher valence bands, such as V₂O₅^[Reference Shrotriya, Li, Yao, Chu and Yang^131] and WO₃^[Reference Tao, Ruan, Xie, Kong, Shen, Meng, Liu, Zhang, Dong and Chen^132] are of interest because of their successful application in organic solar cells. The hole-transport layer for CH₃NH₃PbI₃ could also be formed from other halide perovskites such as CsSnI₃ or CH₃NH₃SnI₃ since tin halide perovskites have very high hole mobilities (μ_h = 10²–10³ cm²/V/s).^[Reference Chung, Song, Im, Androulakis, Malliakas, Li, Freeman, Kenney and Kanatzidis^133]

Ideally, the transport layers should be able to make homogeneous contact with the perovskite layer to ensure efficient collection of charges. Optimizing the surface chemistry of the layers and the perovskite could induce a preferential orientation for growth of the film and therefore yield better contact between the perovskite layer and its scaffold.

The transport layer should also be chemically inert. For example, mesoporous TiO₂ exhibits pronounced degradation under ultraviolet light, which creates oxygen vacancies on the TiO₂ surface that are deep traps for photogenerated electrons.^[Reference Leijtens, Eperon, Pathak, Abate, Lee and Snaith^134] Coating TiO₂ with ZrO₂^[Reference Mei, Li, Liu, Ku, Liu, Rong, Xu, Hu, Chen, Yang, Gratzel and Han^135] or Sb₂S₃^[Reference Ito, Tanaka, Manabe and Nishino^119] or replacing it with a non-conductive scaffold such as Al₂O₃^[Reference Tan, Moore, Saliba, Sai, Estroff, Hanrath, Snaith and Wiesner^136] greatly improves the stability of the device. Clearly, engineering the interface between the perovskite and its contact layers is important in improving performance and overcoming device instability.

Lastly, developing new transparent contacts is also important. Fluorine- and indium-doped tin oxide (FTO and ITO), the two most popular transparent conducting oxides currently in use, will likely be replaced with less expensive materials. Possibilities include aluminum-doped ZnO, graphene, thin metal contacts, and networks of carbon nanotubes or metal nanowires.

Testing a variety of interfacial layers and contact materials does pose challenges because perovskites cannot withstand the conditions of many conventional processes used to produce thin films. Perovskites’ decomposition at relatively low temperatures (~200 °C)^[Reference Pistor, Borchert, Fra, Csuk and Scheer^48] precludes many vapor-phase chemical deposition processes, and their dissolution in water and some polar organic solvents limits solution-based processes such as chemical bath deposition, sol–gel chemistry, and spincoating. Additionally, the organic component of the perovskite is likely to be damaged by plasma-assisted processes or oxidizing environments. Currently, the most commonly employed methods for depositing films on perovskites are thermal and electron-beam evaporation and spincoating from non-polar solvents. While deposition of the transport layer beneath the perovskite can be achieved using any method compatible with the underlying electrode, methods for depositing layers on top of the perovskite are restricted.

Multi-junction photovoltaics

The overall cost of electricity generated from a photovoltaic system is determined by the expenses associated with the solar module, the balance-of-system components, and the system's installation and maintenance. Since the cost of the module is small relative to other costs, improving the module's efficiency while only marginally increasing its cost is the best way to reduce the cost of generated energy. The efficiency of a single-junction solar cell is limited by the Shockley–Queisser limit, which includes losses from transmitted below-band-gap photons and from thermalization of hot photogenerated carriers.^[Reference Shockley and Queisser^137] The concept of multi-junction photovoltaics is to improve the efficiency by minimizing these losses. Within a stack of solar cells with decreasing band gaps, photons with higher energies are converted by the high-band-gap solar cells on top, while the photons with lower energy pass through to be converted by the lower-band-gap solar cells. Conventional fabrication of multi-junction photovoltaics is costly because the stacked crystalline absorber layers must be lattice matched and are produced by expensive epitaxial growth processes. Incorporating defect-tolerant materials that give devices with high efficiencies, such as perovskites, into the multi-junction concept has the potential to be much less expensive than current III–V technology and might improve the overall efficiency. This goal can be achieved by stacking perovskite cells on well-established photovoltaic cells, such as those made from crystalline Si (c-Si), GaAs, CIGS, or CdTe. The perovskite cell can be used as the top cell because its band gap can be tuned to transmit sufficient light to the cell beneath.

There are two basic designs of the multi-junction concept: two-terminal (two stacked cells that are connected electrically in series) and four-terminal (two stacked cells that are electrically independent). The four-terminal design has advantages over the two-terminal design including no requirement of current matching, no need for a tunnel junction or recombination layer, freedom to choose any combination of materials without the need for lattice matching, and simpler prototyping. The primary disadvantage is the parasitic absorption, which is any absorption that cannot lead to photocurrent, that occurs in the four-terminal cell's extra contacts.

The perovskite cells can be mechanically stacked in a four-terminal tandem cell with current established c-Si solar cells to yield higher efficiencies than either of the single cells. For example, adding a semi-transparent perovskite solar cell to low-performing silicon and CIGS solar cells has already been shown to increase the cell's total efficiency.^[Reference Bailie, Christoforo, Mailoa, Bowring, Unger, Nguyen, Burschka, Pellet, Lee, Grätzel, Noufi, Buonassisi, Salleo and McGehee^138] Looking forward, 30%-efficient tandem cells can be achieved for CH₃NH₃PbI₃ (E _{g_optical} = 1.6 eV)/c-Si and CH₃NH₃PbBr₃ (E _{g_optical} = 2.2 eV)/c-Si pairings under one-sun illumination using a minimum 21.0%-efficient CH₃NH₃PbI₃ cell or a 13.2%-efficient CH₃NH₃PbBr₃ cell on top and a 25%-efficient c-Si cell below.^[Reference White, Lal and Catchpole^139] The maximum attainable efficiencies for optimized versions of these tandem devices are estimated at 31.6% and 32.5%, respectively. These estimates of combined efficiency are based on the absorption coefficient, diffusion length, and radiative efficiency of CH₃NH₃PbI₃, which are taken to be 10⁴ cm⁻¹, 100 nm, and 10⁻²%, respectively. An efficiency of over 30% for hybrid perovskite/c-Si tandem devices is realistic, since the maximum efficiency to date for solar cells based on CH₃NH₃PbI₃ is 20.1%^[Reference Green, Emery, Hishikawa, Warta and Dunlop^1] and 10.4% for solar cells made from CH₃NH₃PbBr₃.^[Reference Heo, Song and Im^34] Also, low-temperature processing of the hybrid perovskite cells allows them to be produced on top of other cells without damaging them. Combining solution-processed solar cells such as perovskites with conventional solar cells can increase their efficiency without significantly increasing the cost of production.

The main challenge in fabricating a successful tandem device is minimizing parasitic absorption. The top cell's bottom contact must be semi-transparent so that light reaches the cell underneath. Also, the material used for the top cell itself should have minimal sub-band-gap absorption, for instance from trap states or compositional inhomogeneity. Recent measurements of the perovskites’ sub-band-gap absorption using photothermal deflection spectroscopy indicate that CH₃NH₃PbI₃ performs well in this regard, exhibiting a sharp decrease in absorption below the optical band gap.^[Reference Sadhanala, Deschler, Thomas, Dutton, Goedel, Hanusch, Lai, Steiner, Bein, Docampo, Cahen and Friend^40] Mixtures of iodide and bromide, however, show comparatively greater sub-band-gap absorption that can perhaps be reduced by improved synthetic techniques.

Building-integrated photovoltaics

Integrating semitransparent solar cells into windows is one highly attractive application for perovskite solar cells. Perovskites are one of the few materials that can deliver tunable semitransparency while maintaining high-power-conversion efficiency. Their success is enabled by their unique morphology: micron-sized islands of perovskite with gaps between them. The micron-sized islands are thick enough to absorb most of the visible light, but the gaps are sufficiently large to allow transmission. Controlling the morphology and thickness of the islands along with the sizes of the gaps produces a device with acceptable efficiency (reaching 6.4% so far) and up to 60% transmittance from 400 to 750 nm.^[Reference Eperon, Burlakov, Goriely and Snaith^140] Similar transparency has also been demonstrated using DSSCs^[Reference Kang, Park and Park^141] and OPVs^[Reference Lunt and Bulovic^142]; however, IR-absorbing dyes and polymers exhibit non-uniform absorption of visible light that tends to make these devices appear tinted rather than color-neutral.

A possible target for implementation is to improve the efficiency to 12% while maintaining the existing 60% transmittance. A step toward this goal might come from replacing the gold electrode on hybrid perovskite solar cells with conductive materials that are more transparent, such as graphene or meshes of carbon nanotubes or silver nanowires. Although this approach is technically challenging, it has been shown to be promising in OPVs^[Reference Kang, Xu, Park, Luo and Guo^143, Reference Rowell, Topinka, McGehee, Prall, Dennler, Sariciftci, Hu and Gruner^144] and perovskite tandem solar cells.^[Reference Bailie, Christoforo, Mailoa, Bowring, Unger, Nguyen, Burschka, Pellet, Lee, Grätzel, Noufi, Buonassisi, Salleo and McGehee^138] Additionally, maximizing the photon absorption in the invisible regions of the solar spectrum by engineering the band gap or the optical density of states of the perovskites is also an interesting prospect.

Light emission and detection

The observation of PL in pure and mixed-composition perovskites makes these materials of great interest for optoelectronic applications beyond photovoltaics. Light-emitting diodes,^[Reference Tan, Moghaddam, Lai, Docampo, Higler, Deschler, Price, Sadhanala, Pazos, Credgington, Hanusch, Bein, Snaith and Friend^6, Reference Kim, Cho, Heo, Kim, Myoung, Lee, Im and Lee^7] lasers,^[Reference Deschler, Price, Pathak, Klintberg, Jarausch, Higler, Hu, Leijtens, Stranks, Snaith, Atatu, Phillips and Friend^8–Reference Zhang, Ha, Liu, Sum and Xiong^10] and photodetectors^[Reference Lee, Kwon, Hwang, Ra, Yoo, Ahn, Park and Cho^11, Reference Dou, Yang, You, Hong, Chang, Li and Yang^12] have already been demonstrated, yet the continued success of these applications hinges upon the development of appropriate materials. As in the case of photovoltaics, selection of contacts and transport layers with suitable conductivity and band alignment for the injection of charges is a challenge, and minimizing series resistance from these layers is crucial since light-emitting diodes can operate at current densities that are orders of magnitude larger than the current densities in solar cells. Because of the full tunability of the emission of perovskites, tandem stacked light-emitting diodes, such as those currently produced using organic light-emitting diodes,^[Reference Wang, Zhi and Müllen^145] are an interesting possibility for a white-light source. As an alternative, perhaps a microscopically phase-segregated mixture of the perovskites tuned across the visible spectrum could yield white-light emission, given the right contacting design.

Solar water-splitting

Perovskites also hold promise for the conversion of sunlight directly into chemical fuels, a process called artificial photosynthesis. Perovskite solar cells have already been used to split water into O₂ and H₂,^[Reference Luo, Im, Mayer, Schreier, Nazeeruddin, Park, Tilley, Fan and Gratzel^5] although they were not immersed in the electrolyte. Two perovskite solar cells connected in series were required to achieve the photovoltage necessary to overcome the thermodynamically required minimum voltage of 1.23 V for splitting water and the additional 0.1 and 0.3 V for kinetically driving the H₂ and O₂ evolution reactions.^[Reference Höfle, Schienle, Bernhard, Bruns, Lemmer and Colsmann^146] The maximum solar-to-hydrogen conversion efficiency demonstrated by this system was 12.3%, which is close to the most notable example of water-splitting using GaInP/GaAs tandem cells that achieved 12.4%.^[Reference Hu, Xiang, Haussener, Berger and Lewis^147] Achieving a solar-to-hydrogen conversion efficiency of over 20% is very realistic by using a tandem architecture, for example by combining a perovskite top cell with a c-Si or CIGS bottom cell. The appropriate selection of catalysts is also critically important in order to reduce the overpotential of the chemical reactions. In terms of efficiency and cost-effectiveness, the use of perovskites has already been exceptional. The greatest remaining challenge is how to overcome the instability of perovskites since they show degradation in performance after only hours of operation and are typically soluble in water. Encapsulation of the perovskite in a conductive oxide such as TiO₂ is one strategy toward achieving stability in water since this approach has already been shown to stabilize other semiconductor photoelectrodes.^[Reference Khaselev^148]

Stability and mechanisms of degradation

By far, the chief concern raised in the development of perovskites into practical technology is their long-term stability. At present, moisture and heat have been identified as triggers for instability in CH₃NH₃PbI₃. Since hybrid perovskites are ionic solids, they are unstable in the presence of many polar solvents, most notably water. Films of CH₃NH₃PbI₃ degrade when exposed to high humidity >50%,^[Reference Noh, Im, Heo, Mandal and Il Seok^38] while such degradation was not observed for CH₃NH₃Br₃; consequently, the device performance of CH₃NH₃PbI₃ solar cells decreased while that of CH₃NH₃Pb(I_1−xBr_x)₃ with x = 0.20 and 0.29 remained relatively stable over a period of 20 days of storage [Fig. 7(a)]. Although the exact degradation mechanism is not known, it is postulated that water accepts the proton from the CH₃NH₃⁺, yielding the dissolved hydrohalic acid and volatile methylamine.^[Reference Frost, Butler, Brivio, Hendon, van Schilfgaarde and Walsh^95] The insoluble lead halide remains, as is consistent with experimental observations.

Figure 7. (a) Comparison of the stability of solar cells made from mixed perovskites CH₃NH₃Pb(I_1−xBr_x)₃. Devices were stored in air without encapsulation and tested periodically. On the fourth day the devices were exposed to 55% humidity to evaluate the effect of water on their stability. (b) Stability of a solar cell made from CH₃NH₃PbI_3−xCl_x on an alumina scaffold under continuous illumination of 76.5 mW/cm² (~3/4 one-sun intensity) at 40 °C. The device was encapsulated in a nitrogen environment. (c) Stability over 42 days (1008 h) of a perovskite solar cell made from mixed organic cations CH₃NH₃PbI₃ and 5-ammoniumvaleric acid. The unsealed device was continuously illuminated under one-sun, A.M. 1.5 conditions in air, although the carbon back contact is expected to provide some encapsulation. Panel (a) is reproduced with permission from Ref. Reference Noh, Im, Heo, Mandal and Il Seok38, copyright 2013, American Chemical Society. Panel (b) is reproduced with permission from Ref. Reference Leijtens, Eperon, Pathak, Abate, Lee and Snaith134. Copyright 2013, Nature Publishing Group. Panel (c) is adapted with permission from the supplementary materials from Ref. Reference Mei, Li, Liu, Ku, Liu, Rong, Xu, Hu, Chen, Yang, Gratzel and Han135. Copyright 2013, American Association for the Advancement of Science.

Additionally, hybrid perovskites decompose when heated to relatively low temperatures. While the standard annealing step at 100–150 °C improves crystallinity and preferential orientation, annealing at 200 °C even in vacuum, where no reactive oxygen or water is present, thermally decomposes the films: the PbI₂ remains on the surface, while the methylammonium halide volatilizes.^[Reference Pistor, Borchert, Fra, Csuk and Scheer^48] In situ XRD measurements taken during annealing of CH₃NH₃PbI_3−xCl_x films suggest that decomposition to PbI₂ can occur after annealing 50 min at 100 °C in a nitrogen atmosphere and after only 10 min in air.^[Reference Tan, Moore, Saliba, Sai, Estroff, Hanrath, Snaith and Wiesner^136]

Finally, an open question remains if the operation of the photovoltaic device inherently degrades the perovskite or another material in the stack even when isolated from the environment. Encapsulated TiO₂ appears to generate deep electron traps under UV illumination, which degrade performance. More stable operation has been achieved using Al₂O₃ as the perovskite's scaffold, although the efficiency stabilized to about half of its initial value with most of the loss arising in the V _oc [Fig. 7(b)].^[Reference Leijtens, Eperon, Pathak, Abate, Lee and Snaith^134] Improvements in stability have also been reported in devices that do not use a hole-transport layer. By adding 5-ammoniumvaleric acid to the perovskite, using a combined scaffold made of TiO₂ and ZrO₂, and removing the hole-transport layer, stable operation at 10.5% efficiency was maintained for over 1000 h [Fig. 7(c)].^[Reference Mei, Li, Liu, Ku, Liu, Rong, Xu, Hu, Chen, Yang, Gratzel and Han^135] Clearly as more materials are investigated for their contacting properties, it will become essential that they be simultaneously evaluated for their stability. Much work still remains in evaluating the stability of hybrid perovskites and their devices, discerning the mechanisms of instability, and developing new materials to overcome these challenges.

Achieving high V _oc in perovskites with larger band gaps

CH₃NH₃PbBr₃ has an estimated band gap (E _g) of at least 2.2 eV, making it an attractive option for producing the top cell in a tandem solar cell. The highest V _oc produced from this material, however, is only 1.51 V,^[Reference Heo, Song and Im^34] which yields a loss-in-potential of 0.7 V. In contrast, CH₃NH₃PbI₃ loses only 0.5 V between its band gap (E _g = 1.6 eV) and qV _oc, demonstrating that the iodide perovskite is much closer than the bromide perovskite to achieving its maximum efficiency. In comparison, GaInP (E _g ~ 1.9 eV), which is used for the top cell in high-performance tandem devices, has a loss-in-potential of ~0.5.^[Reference Takamoto, Ikeda, Kurita and Ohmori^150]

The origin of the higher loss-in-potential of CH₃NH₃PbBr₃ remains unclear. One possibility is reduced shunt resistance resulting from large pinholes in the CH₃NH₃PbBr₃ film, since some increase in the V _oc has been seen for denser films.^[Reference Heo, Song and Im^34] Another possibility is a larger density of charge traps or sub-band-gap states in the grain boundaries of CH₃NH₃PbBr₃ than in CH₃NH₃PbI₃. The poorer performance could also result from contacts with improper band alignment or inadequate interfacial passivation because the most recent record V _oc for this material was obtained simply by tuning the hole-transport layer.^[Reference Heo, Song and Im^34, Reference Ryu, Noh, Jeon, Chan Kim, Yang, Seo and Il Seok^125]

Lead-free perovskites

The ultimate goal of materials used in optoelectronic applications should be based on the “triple-E” terms: energy efficient, economically inexpensive, and environmentally benign. The Pb-based halide perovskites have exhibited exceptional performance and will likely satisfy the first two criteria within the coming years; however, the presence of Pb is undesirable because of its toxicity. For a simple substitution, candidates should have a 2⁺ oxidation state to take the B site within the perovskite ABX ₃. These elements include those in the same group with Pb, such as Ge and Sn, but unfortunately their stability as B²⁺ is relatively low as compared with their stability in the B⁴⁺ state. CH₃NH₃SnI₃ has been demonstrated to give a device efficiency as high as 6.4%, but it is much less stable against oxidation than CH₃NH₃PbI₃.^[Reference Hao, Stoumpos, Cao, Chang and Kanatzidis^151, Reference Noel, Stranks, Abate, Wehrenfennig, Guarnera, Haghighirad, Sadhanala, Eperon, Pathak, Johnston, Petrozza, Herz and Snaith^152] Other options can be found in the first row of the transition metals, such as Mn²⁺, Ni²⁺, Cu²⁺, and Zn²⁺, because their 2⁺ state is relatively stable as compared with their higher oxidation states, if any exist. In the case of copper, however, the pronounced Jahn–Teller distortion is likely to affect the band gap and crystal symmetry of the perovskite. Work on two-dimensional (2D) perovskites (A₂BX ₄) has demonstrated the possibility of using transition metals in hybrid perovskites,^[Reference Long, Wei and Willett^154] which can then be adapted to 3D perovskites.^[Reference Mitzi^153, Reference Long, Wei and Willett^154] In addition to non-toxicity, the criteria for selecting suitable replacements should also include the earth abundance^[Reference Haxel, Hedrick and Orris^155] and criticality, which accounts for the potential for a disruption in supply and how well such a disruption can be accommodated by the industry.^{[156, 157]} The transition metals above are abundant non-critical materials.