Control over perovskite formation mechanisms

Advances in perovskite solar cells have been driven by a continuously increasing understanding of the crystallization processes of these hybrid materials. ^{Reference Jeon, Noh, Kim, Yang, Ryu and Seok2,Reference Nie, Tsai, Asadpour, Blancon, Neukirch, Gupta, Crochet, Chhowalla, Tretiak, Alam, Wang and Mohite4,Reference Im, Jang, Pellet, Grätzel and Park31,Reference Zhao and Zhu32} Despite progress in controlling the crystallinity and composition of perovskite thin films, however, the assembly of the precursor salts into high-quality compounds is only phenomenologically understood at present. This is due to the difficulty in monitoring the growth of the perovskite in solution, which can be investigated only a posteriori. From this perspective, a unique advantage of dual-source vapor deposition over all other deposition methods is the possibility to monitor material growth in real-time.

Pistor et al. studied the crystallization process of hybrid perovskite by recording x-ray diffraction patterns while co-evaporating the MAI and PbCl₂ precursors. ^{Reference Pistor, Borchert, Fränzel, Csuk and Scheer33} By steadily increasing the deposition rate of the metal precursor (i.e., continuously modifying the MAI/PbCl₂ stoichiometry), it was possible to follow crystal formation as a function of time. In particular, for the set of deposition parameters used, three main phases were identified ( Figure 2 a). For low PbCl₂ fluxes, black perovskite films with high iodine content and the expected bandgap energy of about 1.6 eV can be obtained. At higher PbCl₂ fluxes, a transition to a different, chloride-rich phase was identified. The results indicate a narrow miscibility gap for methylammonium lead iodide and chloride phases, consistent with the low chloride content identified in solution processed CH₃NH₃PbI_3−x Cl _x thin films. ^{Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca34,Reference Dar, Arora, Gao, Ahmad, Grätzel and Nazeeruddin35}

Figure 2.

(a) Representation of the x-ray diffraction intensity measured during CH₃NH₃PbI_3−x Cl _x formation. For reference, the temperature ramp of the PbCl₂ crucible ( $T_{{\rm{PbCl}}_{\rm{2}}}$ ) and the deposited film thickness are depicted below. As the temperature increases above 370°C, (1) the first perovskite phase is detected. (2) This structure evolves into a second phase, and (3) then fully converts to pure PbCl₂ at temperatures above 450°C. Reprinted with permission from Reference 33. © 2014 American Chemical Society. (b) X-ray photoelectron spectroscopy Cl 2p core level spectra with increasing PbCl₂ thickness deposited on a methylammonium iodide (MAI) (15.0 nm) thin film. No Cl signal can be measured during the initial deposition (0.2 to 0.5 nm), suggesting that the mismatch between the halide ions hinders the incorporation of Cl^– into the forming perovskite. This is confirmed by observing that (c) the trend of the Cl/Pb atomic ratio for different thicknesses of PbCl₂ deposited onto MAI is under stoichiometric, while the correct stoichiometry is obtained when depositing onto a bare ITO surface (symbols are experimental data; the dotted line is the theoretical value). Adapted with permission from Reference 36. Note: MA, methylammonium; ITO, indium tin oxide.

Using x-ray photoelectron spectroscopy, Ng et al. studied the interaction between PbCl₂ and MAI and the subsequent formation of the mixed CH₃NH₃PbI_3−x Cl _x perovskite. ^{Reference Ng, Chan, Lo, Guan and Lee36} PbCl₂ was evaporated on top of a thin, pre-formed MAI film, and the core levels for both Cl and Pb were studied as a function of the PbCl₂ thickness. As illustrated in Figure 2b, a negligible amount of Cl could be observed close to the MAI/PbCl₂ interface. For comparison, the Cl/Pb atomic ratio for PbCl₂ deposited on bare indium tin oxide (ITO) was also studied, revealing the expected stoichiometry even for very thin (<2 nm) film thicknesses (points on dotted line, Figure 2c). It was suggested that due to the differences in ionic radius of Cl^– and I^– ions, Cl leaves the Pb atom when PbCl₂ encounters the MAI molecules. Again, this result agrees with the negligible amount of chloride observed in CH₃NH₃PbI_3−x Cl _x employed in efficient solar cells. ^{Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca34,Reference Dar, Arora, Gao, Ahmad, Grätzel and Nazeeruddin35}

Dual-source vapor deposition

A milestone in the field of perovskite photovoltaics was the demonstration of the first highly efficient solar cell employing a vacuum-deposited perovskite thin-film absorber. ^{Reference Liu, Johnston and Snaith9} CH₃NH₃PbI_3−x Cl _x thin films were prepared by dual-source vapor deposition of the MAI and PbCl₂ precursors in a high-vacuum chamber. The perovskite layer was deposited onto a compact TiO₂ electron-transport layer (ETL) and covered with a solution-processed hole-transport layer (HTL), 2,2′,7,7′-tetrakis-(N,N-di-p-methoxyphenylamine)9,9′-spirobifluorene (spiro-OMeTAD), essentially maintaining the normal embodiment of solution-processed perovskite solar cells. ^{Reference Kim, Lee, Im, Lee, Moehl, Marchioro, Moon, Humphry-Baker, Yum, Moser, Grätzel and Park37,Reference Lee, Teuscher, Miyasaka, Murakami and Snaith38} These planar heterojunction solar cells showed a PCE of over 15%, with an open-circuit voltage (V _oc) of 1.07 V and a short-circuit current density (J _sc) as high as 21.5 mA/cm^–2. ^{Reference Liu, Johnston and Snaith9}

A similar but inverted device concept (the conductive substrate is the hole-extraction contact), where a dual-source evaporated CH₃NH₃PbI₃ perovskite film was sandwiched between organic charge-transport layers ( Figure 3 a), has also been demonstrated. ^{Reference Malinkiewicz, Yella, Lee, Espallargas, Graetzel, Nazeeruddin and Bolink11} The organic semiconductors were chosen to match the conduction and valence bands of the perovskite absorber (Figure 3b). In this case, a thin layer (10 nm) of poly(N,N′-bis(4-butylphenyl)-N,N′-bis(phenyl)benzidine) (polyTPD) was used as the HTL acting, due to its low electron affinity, also as an electron-blocking layer. In between the transparent electrode and the polyTPD, a thin layer of poly(3,4-ethylenedioxythiophene) poly(styrene sulfonate) (PEDOT:PSS) was used as the hole-injection layer (HIL). On the other contact, a 10-nm-thick [6,6]-phenyl C₆₁-butyric acid methyl ester (PCBM) layer ensured selective extraction of the photogenerated electrons. Such simple devices reached efficiencies of 12%, with high open-circuit voltage and fill factor (FF) (Figure 3c–d). ^{Reference Malinkiewicz, Yella, Lee, Espallargas, Graetzel, Nazeeruddin and Bolink11} Further optimization of this same device architecture, in particular by fine tuning the perovskite layer thickness and the electron injection interface, resulted in solar cells with efficiencies as high as 15%. ^{Reference Malinkiewicz, Roldán-Carmona, Soriano, Bandiello, Camacho, Nazeeruddin and Bolink6}

Figure 3.

(a) Structure of a solar cell employing organic charge-transport layers, (b) schematic of the corresponding flat band energy level diagram, and (c) J–V characteristics of a perovskite solar cell recorded at different irradiance and in the dark. Independent of the light intensity, the device exhibited very similar efficiencies. The high open-circuit potential indicates that there are negligible surface and sub-bandgap states in the perovskite film. (d) Optical absorption (red) and incident photon to current conversion efficiency (IPCE) spectra (blue) obtained for the same device structure. The steep onset in the IPCE spectra corresponds to the band-to-band perovskite absorption. Adapted with permission from Reference 11. Note: ITO, indium tin oxide; PCBM, [6,6]-phenyl C₆₁-butyric acid methyl ester; PEDOT:PSS, poly(ethylenedioxythiophene) poly(styrene sulfonate); polyTPD, poly(N,N′-bis(4-butylphenyl)-N,N′-bis(phenyl)benzidine).

This configuration is particularly advantageous since ETL/HTL can be independently varied, selecting the desired materials from the large library of commercially available organic semiconductors. Lin et al. demonstrated this by studying a series of different hole-transport materials in this simple device configuration. ^{Reference Lin, Armin, Nagiri, Burn and Meredith10} Optimum performance was found when the hole-transporting polymer PCDTBT (poly(N-9′-heptadecanyl-2,7-carbazole-alt-5,5-(4′,7′-di(thien-2-yl)-2′,1′,3′-benzothia- diazole))), with an ionization potential of 5.3 eV, was used. Polander et al. took a similar approach by studying the effect of a series of hole-transporting molecules, using again an “inverted” device configuration. ^{Reference Polander, Pahner, Schwarze, Saalfrank, Koerner and Leo39} The HTL was a double layer consisting of a partially oxidized (doped) layer and an intrinsic layer. A series of organic semiconductors with ionization potentials ranging from 5.0 to 5.6 eV were investigated. The devices employing materials with ionization potentials of up to 5.3 eV yielded improved V _oc, while, beyond this range, the absence of a driving force for hole extraction led to a decrease in both J _sc and V _oc . ^{Reference Polander, Pahner, Schwarze, Saalfrank, Koerner and Leo39} Kim et al. used double-layer hole- and electron-transport contacts with improved energy level alignment. ^{Reference Kim, Kim, Choi, Shim and Kim40} They obtained very high V _oc (1.12 V) by employing a MoO₃/NPB (NPB: N,N′-Di(1-naphthyl)-N,N′-diphenyl-(1,1′-biphenyl)-4,4′-diamine) HTL and a C₆₀/BCP (BCP: bathocuproine) ETL. These fully vacuum-deposited devices showed PCEs up to 13.7%, with remarkable cell-to-cell reproducibility.

One of the advantages of dual-source vapor deposition is the possibility of preparing highly crystalline perovskite films of arbitrary thickness. Momblona et al. prepared inverted solar cells, with the layout PEDOT:PSS/polyTPD/CH₃NH₃PbI₃/PCBM/Au, employing perovskite layer thicknesses ranging from 200 to 900 nm. ^{Reference Momblona, Malinkiewicz, Roldán-Carmona, Soriano, Gil-Escrig, Bandiello, Scheepers, Edri and Bolink41} By increasing layer thickness, most of the incident light with energy above the CH₃NH₃PbI₃ bandgap is absorbed ( Figure 4 a). J _sc increased with increasing perovskite layer thickness, first quickly up to 300 nm and then more slowly for devices with thicker active layers. However, due to a decreasing FF, PCE is reduced for the thicker devices. Partial oxidation of the HTL in cells with a 900 nm perovskite layer led to recovery of the FF and efficiency (Figure 4b). ^{Reference Momblona, Malinkiewicz, Roldán-Carmona, Soriano, Gil-Escrig, Bandiello, Scheepers, Edri and Bolink41} This shows that with non-limiting organic layers, cell performance is independent of perovskite thickness and demonstrates very large charge diffusion length for electrons and holes in these materials.

Figure 4.

(a) Absorption spectra for perovskite films with increasing thickness. (b) Photocurrent density (J) versus voltage under 1 sun illumination for a device using non-doped hole-transport layers with a thin (285 nm) and a thick (900 nm) perovskite layer and for a device with a 900-nm-thick perovskite layer but employing a doped hole-transport layer. One advantage of vacuum deposition is the layer absorption (thickness) can be arbitrarily controlled. For thick perovskite layers, however, the conductivity of the charge-transport layers must be adjusted to account for the larger carrier density. Reprinted with permission from Reference 41. © 2014 AIP Publishing. (c) External quantum efficiency (EQE) spectra of devices with different perovskite layer thicknesses. In the cavity region, where the perovskite has lower absorption, interference causes the EQE to be highly thickness-dependent. (d) EQE and internal quantum efficiency (IQE) spectra of a CH₃NH₃PbI₃ device. The IQE is spectrally flat and close to 100% within the error of measurement (±5%). The inset shows the ratio of EQE measured without bias and with a −1 V reverse bias showing that carrier extraction is essentially lossless. Reproduced with permission from Reference 10. © 2015 Nature Publishing Group. Note: polyTPD, poly(N,N′-bis(4-butylphenyl)-N,N′-bis(phenyl)benzidine).

In hybrid perovskite devices, the optimum photoactive layer thickness has been experimentally found to lie between 300 and 400 nm. Lin et al. accurately measured the refractive index (n) and extinction coefficient (k) for a CH₃NH₃PbI₃ layer and hence were able to model the optical-field distribution in a complete device. ^{Reference Lin, Armin, Nagiri, Burn and Meredith10} They predicted two different regimes of behavior: (1) for wavelength λ < 500 nm, no optical field reaches the back electrode as a result of the high-absorption coefficient of the perovskite; (2) for λ > 500 nm, the perovskite is weakly absorbing, and the optical field is dominated by the cavity interference. They consolidated the model by measuring the external quantum efficiency (EQE) of solar cells as a function of the CH₃NH₃PbI₃ layer thickness (Figure 4c). For λ < 500 nm, there is minimal influence of the film thickness, due to the almost complete light absorption by the perovskite cell, while for λ > 500 nm, the EQE is strongly thickness dependent, as predicted by the cavity modeling. By using organic charge-transport layers and a 370-nm-thick perovskite film, record vapor-deposited solar cells, with a PCE of 16.5%, were demonstrated. ^{Reference Lin, Armin, Nagiri, Burn and Meredith10} Interestingly, the correspondent internal quantum efficiency (IQE) is spectrally flat and close to 100% between 450 nm and 750 nm (Figure 4d), confirming the near-unity charge generation and collection efficiency of these perovskite solar cells.

An intrinsic feature of dual-source vapor deposition is the low-temperature processing of the perovskite, especially in the case of using stoichiometric conditions where no post annealing is needed. ^{Reference Malinkiewicz, Yella, Lee, Espallargas, Graetzel, Nazeeruddin and Bolink11} Roldán-Carmona et al. demonstrated flexible perovskite solar cells with a PCE of 7% by employing a 50-μm-thick polyethylene terephthalate (PET) foil coated with a transparent electrode as the substrate. ^{Reference Roldán-Carmona, Malinkiewicz, Soriano, Mínguez Espallargas, Garcia, Reinecke, Kroyer, Dar, Nazeeruddin and Bolink42} The electrode consisted of AZO (30 nm)/Ag (9 nm)/AZO (30 nm), where AZO is Al-doped ZnO at 2 wt% Al. Solar cells were built using a 260-nm-thick CH₃NH₃PbI₃ active layer sandwiched in between organic transport layers. Since the hybrid perovskite is a polycrystalline material, cracks or other imperfections may occur upon bending it. The authors studied the effect of repeated bending of the solar cells and apart from a slight initial decrease of the FF, did not observe any reduction in key performance indicators. Notably, the PCE decreased only about 0.1%, indicating that, despite their polycrystalline nature, hybrid perovskites have the potential of being applied in future flexible electronics applications.

Sequential vacuum deposition

As described previously, sequential vacuum deposition is being increasingly adopted as an alternative effective method for obtaining highly crystalline hybrid perovskites. Several variations of this process have been reported so far. Liu et al. used the thermal evaporation of PbI₂ films with subsequent dipping in MAI solutions as a reproducible method for preparing perovskite thin films with precise control over final film thickness. ^{Reference Liu, Gangishetty and Kelly18} At the optimum thickness of 330 nm, ITO/ZnO/CH₃NH₃PbI₃/P3HT/Ag devices were found to have PCEs of up to 11.3%, which is remarkable for devices using poly(3-hexylthiophene) (P3HT) as the HTL. A similar modified process, where the MAI is deposited by hybrid CVD, also demonstrated devices with PCE as high as 11.8%. ^{Reference Leyden, Ono, Raga, Kato, Wang and Qi19}

Further variation of this method is the full-vacuum sequential sublimation of the inorganic and organic precursor. ^{Reference Chen, Kang, Hsiao, Yang, Chiang and Lin7,Reference Hu, Wang, Zhou, Zhang, Lv, Pang, Chen, Liu, Padture and Cui20} In this case, the sublimed PbCl₂ converted to perovskite in situ upon MAI deposition, resulting in uniform thin films with large-scale homogeneous crystalline structures at an optimized substrate temperature. The corresponding solar cells, employing an organic double ETL (C₆₀/bathophenanthroline), did not make use of an HTL but relied solely on a thin PEDOT:PSS HIL. This is possible due to the ambipolar character of CH₃NH₃PbI_3−x Cl _x and the presence of a high built-in voltage in the device, resulting in efficient carrier extraction even without selective contacts. ^{Reference Mei, Li, Liu, Ku, Liu, Rong, Xu, Hu, Chen, Yang, Grätzel and Han43,Reference Laban and Etgar44} The HTL free configuration delivered remarkable performance, with J _sc = 20.89 mA/cm^–2, FF = 72.2%, and PCE as high as 15.4%.

In a similar fashion, perovskite can be obtained through sequential deposition of alternating thin PbX₂/MAI multilayers, where the MAI reaction can take place in solution ^{Reference Sutherland, Hoogland, Adachi, Kanjanaboos, Wong, McDowell, Xu, Voznyy, Ning, Houtepen and Sargent22} or in the vapor phase. ^{Reference Chen, Chen and Dai8} This strategy allows for precise control over the thickness and thus optical absorption of the CH₃NH₃PbI_3−x Cl _x films ( Figure 5 a). PCE of solar cells with structure ITO/PEDOT:PSS/CH₃NH₃PbI_3−x Cl _x /PCBM/Ca-Al shows an almost linear dependence on the number of deposited PbCl₂/MAI layers (Figure 5b), suggesting highly uniform growth of well-defined crystalline films. For seven deposition cycles, the PCE exceeded 15% with high FF and J _sc, and the J–V characteristics are essentially free of hysteresis (Figure 5c–d), which points again to a well-crystalized, defect-free active layer.

Figure 5.

(a) Optical absorbance spectra of layer-by-layer growth CH₃NH₃PbI_3−x Cl _x perovskite films with different number of layers, and (b) corresponding solar-cell power-conversion efficiency (PCE). (c) J–V characteristics under illumination (red) and in the dark (blue) for a layer-by-layer perovskite device with corresponding (d) forward and reverse bias scan. Adapted with permission from Reference 8. Note: V _oc, open-circuit voltage; J _sc, short-circuit current density; FF, fill factor.