In 1954, the first practical photovoltaic (PV) device based on crystalline silicon was demonstrated at Bell Laboratories.1 After many decades of progress, crystalline silicon technology dominates the global PV market with a 55% and 36% market share for polycrystalline- and monocrystalline-silicon modules in 2014, respectively.2 The remaining 9% of the market was split between a variety of other established and emerging PV technologies, including polycrystalline thin films, amorphous semiconductors, dye-sensitized solar cells (DSSCs), organics, and quantum dot solar cells.3 To gain market share from crystal silicon solar cells, alternative technologies have to provide a desirable combination of high power conversion efficiency (PCE), low manufacturing costs, and excellent stability. Recent research suggests that organo-metal halide perovskites (OMHPs), with methylammonium lead iodide ( or ) being the prototypical example, have the potential to meet these conditions and become competitive in the marketplace. As a result of intensive research efforts across the globe over the past three years,184.108.40.206.220.127.116.11.–13 perovskite-based solar cell PCEs are now comparable to or better than most other PV technologies, and the simple device processing promises lower manufacturing costs, suggesting the potential to challenge the prevailing silicon technology in the foreseeable future.14
The term perovskite refers to the crystal structure of calcium titanate (), which was discovered by the German mineralogist Gustav Rose in 1839 and named in honor of the Russian mineralogist Lev Perovski.15 In the field of optoelectronics, OMHPs are a group of materials with the formula , where is an organic cation ( or ), M is a divalent metal cation ( or ), and X is a monovalent halide anion (, , or ). Figure 1 shows the crystal structure and a single crystal of . In a unit cell of the OHMP structure, eight cations are located at the vertices of a cubic cage, an cation is located at the center of the cube, and the latter species is octahedrally coordinated to six species that sit at the cube’s faces. The OMHP family of materials were studied in the 1990s due to their excellent optoelectronic properties and potential for solution-processed fabrication,17,1819.–20 but the main goal of this early work was to develop new materials for field effect transistors and organic light-emitting diodes.21,22
The first known use of OMHP was as a dye in DSSC, which reported a 3% PCE in 2009.23 However, this OMHP-based solar cell contained a liquid electrolyte and received little attention due to the low efficiency and poor stability. The so-called perovskite fever24 did not fully bloom until a solid-state cell was developed and devices with efficiency were reported in 2012.25,26 Since then, OMHP-based PV device performance has rapidly progressed, and a best efficiency record of was achieved in late 2015.27 The pace of progress has been remarkable and unprecedented in PV history and can likely be attributed to several factors related to inexpensive fabrication costs, ease of processing, and the excellent optoelectronic properties of the materials.7,9,1011.12.–13
As will be described in Sec. 3, high-quality perovskite thin films can be fabricated using a variety of processes including solution-26,28 and vapor-based2930.–31 deposition methods. Many of these methods are compatible with low-cost, large-scale, industrial production techniques, which strengthens the potential for the commercialization of perovskite solar cells. Due to the ease of processing, many research groups from around the world have been attracted to work in the area. This includes groups that have past histories and relevant expertise in DSSC, organic photovoltaics (OPV), and solution processing. Consequently, the learning curve for developing perovskite solar cells has been relatively short, and progress has been very rapid.
In addition to flexibility in processing, OMHP materials possess several outstanding optoelectronic properties that make them ideal choices for PV applications. The 1.55 eV band gap of is nearly ideal for single-junction solar cells exposed to the solar irradiance spectrum, and it can be continuously varied in the range from 1.5 to 2.3 eV by exchanging the organic and halide ions.32,33 The optical absorption coefficient of is higher than other PV materials such as Si, CdTe, (CIGS), and amorphous Si:H, so the absorber thickness can be reduced to , thereby lowering the material costs.34,35 In contrast to organic PV materials, the low exciton binding energy (30 to 50 meV) allows spontaneous exciton dissociation into free charges after light absorption.3637.–38 Moreover, the high electron and hole mobility in the range of 10 to and the long carrier lifetime () result in long diffusion lengths () so that charge carriers can be freely transported across the 300-nm thick perovskite absorber before recombination.3940.41.–42 Finally, because the electronic defects are shallow and relatively benign, the nonradiative recombination rates are low, allowing open-circuit voltages to be achieved.43,44
Although the perovskite solar cells show great potential, there are several challenges that need to be addressed before commercialization will be possible. Perhaps most significantly, OMHPs have not yet demonstrated the long-term stability that is necessary to compete with the 30-year lifetime of commercially available Si and CdTe solar panels. Second, there are questions about the current–voltage () hysteresis during voltage scanning, which could be problematic for large-scale deployment. There are also concerns associated with potential environmental impacts due to the fact that OMHPs contain Pb.
This review focuses on the recent advances that have allowed perovskite PV to improve to efficiencies in the time period of 2013 to 2015. Reviews that discuss early developments and the materials can be found elsewhere.7,9,1011.12.–13 Here, we provide insights on the key factors that govern the device performance, including device architecture, preparation methods, and advanced device engineering. Notable devices over this time frame are highlighted. Additionally, the stability issues and future directions for perovskite PV devices are discussed.
The first OMHPs employed in PV were used as direct replacements for the dye sensitizers in the DSSCs.23,45The typical DSSC structure employs a several-micron thick porous layer that is coated and penetrated with an absorber dye material. The electrode assembly is contacted by a liquid electrolyte containing a redox couple.46 In these devices, is used to collect and transport the electrons, while the electrolyte acts as a hole conductor. The original perovskite solar cells evolved from this same structure, with the OMHP materials acting simply as a dye replacement.25,26 Interest increased when the so-called mesoscopic device structure [Fig. 2(a)] was formed by replacing the liquid electrolyte with a solid-state hole conductor.25,26 This advance created great interest in the PV community and drew in experts from the thin-film PV and OPV communities. As a result, planar device structures in which the OMHP absorber is sandwiched between electron and hole transporting materials (ETM and HTM) were developed. Depending on which transport material is encountered by the light first, these planar structures can be categorized as either the conventional n-i-p [Fig. 2(b)] or the inverted p-i-n [Fig. 2(c)] structures. Recently, a mesoscopic p-i-n structure [Fig. 2(d)] has also been developed.47,48 Due to processing differences, the device architecture determines the choice of charge transport (ETM and HTM) and collection (cathode and anode) materials, the corresponding material preparation methods, and, consequently, the performance of the devices. To date, no perovskite devices with significant efficiency have been constructed on opaque substrates (e.g., Ti foils)49,50 because the conventional deposition technologies for transparent conducting oxides (TCO) may lead to decomposition of the surface of the OMHP.
Conventional n-i-p Structure
The mesoscopic n-i-p structure is the original architecture of the perovskite PV devices and is still widely used to fabricate high-performance devices. The structure [Fig. 2(a)] consists of a TCO cathode [fluorine doped tin oxide (FTO)], a 50- to 70-nm thick compact ETM (typically ), a 150- to 300-nm thick mesoporous metal oxide (mp- or mp-) that is filled with perovskites, followed by an up to 300-nm perovskite capping layer, a 150- to 200-nm thick layer of 2,2’,7,7’-tetrakis(N,N-di-p-methoxyphenylamine)-9,9’-spirobifluorene (spiro-MeOTAD), which is a hole conductor, and 50 to 100 nm of a metal anode (Au or Ag).
In this structure, the mesoscopic layer is thought to enhance charge collection by decreasing the carrier transport distance, preventing direct current leakage between the two selective contacts and increasing photon absorption due to light scattering. Accordingly, the original mesoscopic perovskite devices used a thick () porous layer to efficiently absorb the incident light.25,51 But because the grain growth of the perovskites is confined by the pores in the structure, a significant amount of the material is present in disordered and amorphous phases.52 This leads to relatively low open-circuit voltage () and short-circuit current density ().53 Surprisingly, thinning the mesoporous layer to to 200 nm results in improved device efficiency due to enhanced crystallinity in the perovskite absorber. Additionally, the pore filling fraction and morphology of the perovskites is critically dependent upon the thickness of mp-.54,55 When the porous layer thickness is reduced to , the pore filling fraction is increased and a perovskite capping layer forms on top of the porous structure. Complete pore filling accompanied by formation of a capping layer assures high charge transport rates and high collection efficiencies at the interface. Once the charges are separated, recombination pathways between electrons in the and holes in the HTM are blocked due to the relative positions in energy of the respective conduction and valence bands (vide infra).54Consequently, the meso n-i-p structure is the most popular structure reported in the literature. The previous record efficiency value (20.2%) was measured from a cell formed in the mesoscopic structure that had discrete perovskite nanocrystals embedded in the porous ETM film with an overlaying continuous and dense perovskite capping layer.56
The planar n-i-p structure [Fig. 2(b)] is the natural evolution of the mesoscopic structure. A larger area mesoporous ETM was initially considered critical for high-efficiency perovskite devices because hole extraction at the HTM interfaces is significantly more efficient than electron extraction at the ETM interfaces.57However, by delicately controlling the formation of the perovskite absorber, and the interfaces among the perovskite, carrier transport layers, and electrodes, high efficiencies can now be achieved without a mesoporous layer.58 To date, the best planar n-i-p device showed a 19.3% efficiency after careful optimization of the electron selective indium tin oxide interfaces.58 Although the planar n-i-p perovskite solar cell usually exhibits enhanced and relative to a comparative mesoscopic device processed with the same materials and approach, the planar device usually exhibits more severe hysteresis (see Sec. 6). Thus, the state-of-the-art n-i-p devices usually include a thin () mesoporous buffer layer filled and capped with the perovskite.56
Inverted p-i-n Structure
When the deposition order is changed and the HTM layer is deposited first, the device is fabricated in the p-i-n structure [Fig. 2(c)]. In this case, the p-i-n type perovskite device is built on a 50- to 80-nm p-type conducting polymer such as poly(3,4-ethylenedioxythiophene) poly(styrene-sulfonate) (PEDOT:PSS), which is deposited on ITO-coated substrates. After depositing a 300-nm intrinsic perovskite thin film, the device is completed with a 10- to 60-nm organic hole-blocking layer [6,6]-phenyl -butyric acid methyl ester (PCBM) and a metal cathode (Al or Au). Early device design utilized a perovskite and fullerene () donor–acceptor pair, which is typical in OPV.59 In fact, the commonality in structure has allowed OPV researchers to easily move into the field of perovskites. As the field has advanced, the organic acceptor has been omitted in favor of an ETM layer, leaving the planar perovskite absorber sandwiched between two opposite organic charge transporting materials.60 Recently, the efficiency of the planar p-i-n device has improved significantly due to the use of more advanced material preparation methods, such as a multicycle solution coating process, and a best efficiency of 18.9% was achieved.61
Further development of the p-i-n device structure has expanded the selective contact options from organic to inorganic materials. For example, NiO and layers have recently been used for the hole and electron selective contacts, respectively, which makes the perovskite device distinct from its organic conterpart.62,63 Inorganic charge extraction layers (NiMgLiO and ) have been used to fabricate large-area (), high-efficiency (15%) perovskite cells, representing a potentially important step in the path toward commercialization.62 The use of oxide HTMs also allows for construction of the mesoscopic p-i-n device structure [Fig. 2(d)], in which NiO/mp- or c-NiO/mp-NiO are used as the HTM.47,48 The best mesoscopic p-i-n device with a nanostructured NiO film demonstrated a 17.3% efficiency.64
The device performance of most thin-film solar cells is mainly determined by the film quality of the absorber. High-quality perovskite films with appropriate morphology, uniformity, phase purity, and crystallinity are essential for high-performance PV devices. To meet these quality criteria, well-controlled crystallization and engineering of the composition and interface properties of perovskite films are required. Critical issues include the deposition approach, precursor composition, processing condition, and additive control, all of which can greatly affect the crystallization and quality of the perovskite films. Focusing first on the deposition approach, the preparation processes can be categorized as follows: single-step solution deposition,26 two-step solution deposition,28 two-step vapor-assisted deposition,30 and thermal vapor deposition.29
Single-Step Solution Deposition
Single-step solution deposition [Fig. 3(a)] is commonly used for perovskite thin film preparation due to ease of processing and low fabrication cost. Generally, organic halides [methylammonium iodide (MAI)] and lead halides (, , Br, or Cl) are dissolved in gamma-butyrolactone (GBL), dimethylformamide (DMF), or dimethyl sulfoxide (DMSO) to prepare the precursor solution. The perovskite films can be prepared by spin-coating of the precursor solution followed by a postdeposition heating at 100 to 150°C. Since the perovskite tolerates composition variation,65 high-efficiency devices can be fabricated through a wide range of MAI to precursor ratios from MAI-poor ()66 to MAI-rich ().58 However, it is critical to choose appropriate processing temperatures and times based on differing precursor compositions to achieve the desired crystallinity, phase, and morphology of the perovskite films.55,65,67 In addition to the choice of precursor composition and processing temperature, the environment (oxygen and humidity levels), substrate material, and deposition parameters must also be controlled. The first solid-state device prepared using the single-step solution process produced a perovskite device that exhibited 9.7% efficiency.68 After developing advanced engineering techniques (discussed in Sec. 4), a best efficiency of 19.7% has been achieved with single-step solution deposition.69
In addition to spin-coating, other solution-based deposition methods, including spray,70 doctor-blade,71 inkjet printing,72 and slot-die printing,73 have also been employed to fabricate perovskite PV devices. These techniques demonstrate the potential for large-scale roll-to-roll manufacture of perovskite solar cells. However, the efficiency of devices prepared by these methods is still lower than that of spin-coated devices due to the difficulties associated with controlling the film morphology and compositional uniformity at present.
Two-Step Solution Deposition
The two-step solution deposition approach to preparing OMHPs was first introduced by Mitzi et al. in 1998.74Following this pioneering work, Gratzel et al. developed a sequential deposition method [Fig. 3(b)] to prepare perovskite solar cells, which has resulted in efficiencies .28 In a typical two-step solution procedure, a seed layer is spin-coated and then converted to by dipping the substrate into an MAI/isopropanol solution.28 Spin-coating has also been used to introduce MAI molecules into the network.68 Compared with the single-step solution process, the two-step sequential deposition process results in more uniform and dense perovskite films.75 The process can be well controlled and, consequently, has been extensively used to fabricate high-efficiency devices.28,56,76,77
The two-step solution method provides a reproducible way to fabricate high-quality perovskite thin films. Through varying the MAI solution concentration, the perovskite grain size can be controlled.68 However, one of the drawbacks of the two-step solution deposition method is the trade-off between perovskite grain size and surface smoothness. Films with large perovskite grains typically exhibit poor surface coverage, which can limit the performance of devices. The other issue with this method is incomplete perovskite conversion. The conversion from to rapidly occurs as the film is dipped into the solution because the layered structure of heavy metal halide is prone to interaction with small molecules.78 Thus, a dense perovskite capping layer usually forms on the surface of and hinders the MAI diffusion to the underlying layer, leading to incomplete perovskite conversion. These issues have been overcome by some new techniques that have been developed recently (Sec. 4), and now the champion cell efficiency using the two-step solution method has been improved to 20.2%.56
Vapor-Assisted Solution Deposition
In one modification to the two-step solution deposition method, MAI is introduced through a vapor deposition technique rather than through solution processing [Fig. 3(c)].30 This deposition method allows better control of morphology and grain size via gas–solid crystallization and effectively avoids film delamination that can occur during liquid–solid interaction. The perovskite films prepared by this method exhibit uniform surface coverage, large grain size, and full conversion. However, the use of this method is limited because the gas–solid reaction typically required tens of hours for the full conversion, and devices prepared by this method have exhibited only 10 to 12% efficiency.30,79
Thermal Vapor Deposition
Vapor phase deposition is widely used for fabricating high-quality semiconductor thin films with uniform thickness and composition. The thermal vapor deposition of OMHP thin films was first demonstrated by Mitzi et al. in 1999.18 After modifying the technique for dual-source thermal evaporation [Fig. 3(d)], Snaith et al. prepared the first planar heterojunction perovskite solar cell with an efficiency that exceeded 15%.29 Similar vapor-based deposition techniques, such as sequential layer-by-layer vacuum sublimation31and chemical vapor deposition,80 have also been developed.
The perovskite films prepared by thermal vapor deposition are extremely uniform and pinhole-free. Compared with the incomplete surface coverage that can be found for perovskite films prepared by solution processing, vapor-deposited perovskite layers can conformally coat and PEDOT:PSS layers.29,60,81 However, both the precursor sources and the products have low thermal stability, so the vapor deposition requires precise control over temperatures during deposition. Thus, only a few research groups have reported high-efficiency devices prepared by this method.29,31,60,81,82
Advanced Device Engineering
In early 2013, the state-of-the-art of perovskite solar cells prepared by various deposition techniques had demonstrated device efficiencies in the range of 12 to 15%.2829.–30 Since then, the efficiency has improved to 18 to 20%, mainly due to the advancements of several device engineering strategies.10,83,84 These engineering strategies, which focused on controlling the precursor solution, processing condition, perovskite composition, and interface properties, lead to smooth and pinhole-free perovskite thin films consisting of large grains with good crystallinity. The combination of these advanced engineering methods has improved the optoelectronic properties of the perovskite films and, consequently, the device performance as well.
Single-step spin-coating is the simplest method for preparing perovskite thin films; however, it is difficult to achieve a homogeneous composition and uniform thickness over large areas. The reason for this is that single-step solution deposition using DMF and GBL solvents often results in the formation of needle- and spherical-shaped colloidal intermediates.28,85 To improve the surface morphology of spin-coated perovskite films, several precursor solution additives have been employed to suppress the formation of deleterious intermediates.
DMSO is one of the best and widely used additives.77,86 The precursor solution with added DMSO forms a uniform and flat MAI--DMSO intermediate film when spin-coated. After a thermal treatment, the intermediate film is converted into a uniform perovskite film through a solid-state reaction. Several other additives, such as ,87 HI,88 ,89 ,90 ,91 1,8-diiodooctane,92 aminovaleric acid,93 and phosphonic acid ammonium,94 have also been used to improve the crystallinity and morphological uniformity of perovskite films.
The formation of uniform perovskite film by incorporating additives is the result of decoupling the nucleation and grain growth processes. For precursor solutions without additives, these two processes occur simultaneously. Since grain growth favors large-size nuclei (the free energy of volume expansion eclipses that of interface formation), the unbalanced growth rate leads to the formation of large perovskite grains with a significant number of voids between grains. The introduction of additives retards the crystallization kinetics of perovskite formation and results in a uniform intermediate phase film during deposition. A thermal treatment provides the energy for conversion to the perovskite phase and promotes crystal growth to form pinhole-free films.
Additive incorporation was introduced to the two-step methods after its success in single-step deposition. The precursor solution for can be mixed with DMSO,95 ,96 and low concentrations of MAI76 to improve the surface coverage of the final perovskite film. As with single-step deposition, the introduction of additives results in an intermediate state that retards the rapid reaction between MAI and and effectively avoids the formation of a dense perovskite capping layer on the surface of the layer that hinders further conversion.
In addition to modifying the precursor solution, improved device performance has been achieved by adapting the deposition and postdeposition processes. While slowing the crystal growth kinetics has resulted in higher-quality films, the same results have been obtained by speeding the nucleation kinetics. Hot casting, in which crystallization of the perovskite film occurs immediately after a hot precursor solution is loaded onto the substrate at an elevated temperature, has been used to obtain pinhole-free perovskite films with millimeter-scale grains.97 Using this approach, the island-shaped grains rapidly integrate into a dense perovskite film with millimeter-size grains following Volmer-Weber growth.98 Devices with efficiency of were fabricated using this technique.97
Another demonstration of process engineering for fabricating extremely uniform and dense perovskite films is adding an antisolvent that does not dissolve perovskite films (e.g., toluene) during the last few seconds of the spin process.77 The introduction of toluene rapidly extracts DMF from the precursor solution, which results in a rapid precipitation of perovskite before significant growth of the perovskite grains. Thus, a dense, small-grain perovskite film can form uniformly across the entire substrate surface. In addition to toluene, other antisolvents, such as diethyl ether,69 chlorobenzene, benzene, and xylene, are also effective in forming highly uniform perovskite films.99 Since the grain growth kinetics are suppressed during the deposition, this process needs an optimized thermal annealing to achieve both smooth morphology and large grain size.
The postdeposition grain growth process can also be engineered to achieve a uniform and high-quality perovskite film. Although thermal annealing helps increase grain size and improve crystallinity, it may cause decomposition of the perovskite phase and reduce surface coverage.55,65 Solvent annealing with DMF leads to recrystallization and regrowth of perovskite grains, resulting in improved crystallinity and electronic properties and enhanced device efficiency (15.6%).100 Annealing with pyridine or MAI vapor has demonstrated enhanced luminescence and carrier lifetimes, indicating the formation of high-quality absorber material and the potential for high-efficiency devices.101,102
The advanced solvent and process engineering techniques both aim to decouple the nucleation and growth processes so that the perovskite film formation can be precisely controlled. By applying one or a combination of these techniques, high-quality perovskite films with smooth morphology and large grains were prepared (Fig. 4) and devices with efficiencies of 15 to 19% were fabricated. Details on these devices are summarized in Table 1.