Title Advancements in perovskite solar cells : photophysicsbehind the photovoltaics

Solution-processed organic–inorganic perovskite solar cells are hailed as the recent major breakthrough in low-cost photovoltaics. Power conversion efficiencies approaching those of crystalline Si solar cells (exceeding 15%) have been reported. Remarkably, such phenomenal performances were achieved in a matter of 5 years – up from ∼3.8% back in 2009. Since then, the field has expanded exponentially. In this perspective, we review the basic working mechanisms of perovskite solar cells in relation to their intrinsic properties and fundamental photophysics. The current state-of-the-art and the open questions in this maturing field are also highlighted.


Introduction
Solar power is the world's most abundant energy resource. A year's worth of sunlight contains 1.5 Â 10 18 kW h of energy. By comparison, the known reserves of oil, coal, and gas are 1.75 Â 10 15 kW h, 1.4 Â 10 15 kW h, and 5.5 Â 10 15 kW h, respectively. Thus, a year's worth of sunlight provides more than a hundred times the energy of the world's entire known fossil fuel reserves. Harnessing solar power would yield a never-ending energy supply. 1 The difficulty has always been converting solar energy in an efficient and cost-effective way. Photovoltaic cells are the most promising avenue for directly converting the photons to electricity. For photovoltaic energy to become competitive with fossil fuels and to capture a signicant share of the electricity market, it is necessary to reduce the total cost of solar energy. This can be achieved by either reducing the cost of photovoltaic cells or by increasing their power conversion efficiencies.
The photovoltaic market is currently dominated by crystalline Si solar cells with efficiencies close to 20%. Alternative "third generation" technologies such as organic photovoltaics (OPVs), dye sensitized solar cells (DSCs) and quantum dot solar cells (QDSCs), which are fabricated through solution based processes such as blade coating, screen printing and spraying, promise low cost solar power while allowing the utilization of unconventional substrates. Although the utilization of low temperature processes reduces the energetic costs and the energy payback time, the power conversion efficiencies (PCEs) of these solar cells still lag signicantly behind conventional solar cells. The levelized cost of energy (LCOE), which allows for the comparison of various electricity generation sources, depends critically on the efficiency of the solar cells produced. 2 A more efficient module yields more power per unit area. A signicant fraction of a solar cell cost scales proportional to the installation area, including the cost of the glass, inverter costs and installation costs, among others. A more efficient solar cell allows for a reduction in all the costs associated with installation, while requiring much lower numbers of solar panels to be installed. Thus the PCE is a primary driver of cost for solar cells.
Organic-inorganic halide perovskite solar cells have been the most signicant development in the eld of photovoltaics in the present decade and are the best bet at satisfying the need for high efficiencies while allowing for low cost solution based manufacturing. Since the rst reports of stable solid state solar cells based on the CH 3 NH 3 PbI 3 perovskite in mid-2012, the PCEs of the solar cells have already exceeded 15%, leapfrogging every other solution-processed solar cell technology. The wide range of efficient perovskite solar cell device architectures demonstrated points towards a remarkable semiconducting material with excellent electrical and optical properties. Early pioneering work 3,4 in the area of organic-inorganic halides has clearly shown that this class of materials can behave as low dimensional electronic systems with tunable properties, allowing for the development of newer perovskite solar materials in addition to CH 3 NH 3 PbI 3 .
This review focuses on the recent developments (i.e., up to Feb 2014) in perovskite solar cells as well as their photophysical properties and charge dynamics. We rst review the intrinsic physical and electronic properties of this class of organic-inorganic perovskites, followed by its progress as a photovoltaic material. The review then examines the recent photophysical studies on unraveling the charge dynamics and transport mechanisms in both perovskite thin lms as well as in perovskite solar cells. Due to the rapid pace of research in this area, this review does not aim to be comprehensive but will highlight key studies and ndings. Lastly, we conclude the review with the open questions facing these classes of solar cells and future directions of research.

Intrinsic properties of organicinorganic perovskites (a) Physical structure
Perovskite is the common nomenclature for compounds with the generic chemical formula -AMX 3 . In the cubic unit cell, the A-cation resides at the eight corners of the cube, while the M-cation is located at the body center that is surrounded by 6 X-anions (located at the face centers) in an octahedral [MX 6 ] 4À cluster. Typical inorganic perovskites include CaTiO 3 and SrTiO 3 . Due to the interplay of the charge, spin and structural properties, this family of materials is known to exhibit a plethora of novel and exciting phenomena such as superconductivity, magnetoresistance, ferroelectricity, magnetoelectricity, anti-ferromagnetism, anti-ferroelectricity, etc. 5 In the classes of compounds being discussed here, the A cations are organic (typically CH 3 NH 3 + , C 2 H 5 NH 3 + , HC(NH 2 ) 2 + ), the metal cations (M) are typically divalent metal ions such as Pb 2+ , Sn 2+ , Eu 2+ , Cu 2+ etc., while the X anions are halides (Cl À , Br À , I À ). CH 3  and probing carrier and quasi-particle dynamics in a broad range of emergent nanoscale and light harvesting systems using femtosecond time-resolved spectroscopy.
Dr Nripan Mathews is an assistant professor at the School of Materials Engineering in Nanyang Technological University. He pursued his PhD at a joint Commissariatà l'énergie atomique (CEA) -Centre national de la recherche scienti-que (CNRS) -Universite de Pierre et Marie Curie (Paris VI University) laboratory in the area of molecular crystals, studying the signatures of optical excitations within them (2008). He was also a visiting scientist at Prof. Michael Grätzel's laboratory atÉcole Polytechnique Fédérale de Lausanne (EPFL), working on a pan-european project on photoelectrochemical hydrogen production. His research focuses on a wide variety of novel materials (metal oxides, organic semiconductors, graphene, carbon nanotubes, suldes, and selenides) and novel morphologies (one dimensional structures such as nanowires and nanotubes, thin lms as well as two dimensional nanosheets) produced through a range of fabrication procedures. He has focussed primarily on the electronic and optical properties of these materials and how they can be adapted for practical applications.
the focus was on the effects of dimensionality on the excitonic, optical and electronic properties. 4,6,7 The optoelectronic properties of layered organic-inorganic perovskites were also extensively studied because of the novel properties exhibited by these crystals which include: high mobilities for thin-lm transistors; 8,9 strong excitonic properties for light emitting diodes; 10,11 large nonlinearities with ultrafast responses; 12,13 and even polariton emission in 2D perovskite-based microcavities. [14][15][16][17][18] In CH 3 NH 3 PbI 3 , each [PbI 6 ] 4À octahedron is connected with six neighbours at the iodideforming a 3-D network ( Fig. 1(a)). The countercation (CH 3 NH 3 + ) is located at the void of the network. For the 2-D case 19 e.g., (CH 3 NH 3 ) 2 PbI 4 , each [PbI 6 ] 4À octahedron is connected with four neighbours at the halideforming a 2-D network layer that is sandwiched between two CH 3 NH 3 + layers (also not shown in Fig. 1  close matching of their calculated bandgaps (where spin-orbit coupling (SOC) effect were not considered) with the experimental data is likely to be fortuitous. These ndings are consistent with the studies by T. Baikie et al. 26 and Y. Wang et al.
(low-temperature orthorhombic phase). 27 Investigation on the SOC effect on the electronic band structure in 3-D perovskites (low-temperature orthorhombic phase) was reported by J. Even et al., 28 where they found that the SOC dramatically reduces the energy gap affecting mainly the conduction band.
Having covered the fundamental properties of the large family of organic-inorganic perovskites, we next focus our attention on CH 3 NH 3 PbI 3 and their photovoltaic applications.

Progress in perovskite photovoltaics
Initial studies in the area of perovskite solar cells arose as an evolution of the dye sensitized solar cell 29 architecture. DSCs typically consist of a mesoporous n-type TiO 2 electrode which has been sensitized by a dye and placed in a liquid electrolyte (typically the I À /I 3 À redox couple in such liquid electrolyte solar cells were made by N. G. Park and coworkers in 2011 through a careful optimization of the mesoporous layer thickness, perovskite concentration and surface treatment. Surface modication of TiO 2 with Pb(NO 3 ) 2 prior to deposition of perovskites resulted in an efficiency of 6.54% (J SC ¼ 15.82 mA cm À2 , V OC ¼ 0.706 V and FF ¼ 0.586). Despite the efficiencies achieved in such congurations, the overall instability of the solar cells due to the dissolution of the perovskite in the liquid electrolyte appeared to be a challenge. A breakthrough in both efficiency and stability was achieved in 2012 through utilization of a solid-state hole transporter 2,2 0 ,7,7 0tetrakis(N,N-p-dimethoxy-phenylamino)-9,9 0 -spirobiuorene (spiro-OMeTAD) with CH 3 NH 3 PbI 3 and CH 3 NH 3 PbI 3Àx Cl x as light absorbers (Fig. 3). 32,33 N. G. Park, M. Grätzel and coworkers reported a PCE of 9.7% (J SC ¼ 17.6 mA cm À2 , V OC ¼ 0.888 V and FF ¼ 0.62) for CH 3 NH 3 PbI 3 on 0.6 mm TiO 2 layers. The use of spiro-OMeTAD dramatically improved the device stability compared to liquid junction cells with ex situ long-term stability tests conducted for over 500 h, where the devices are stored in air at room temperature without encapsulation. 32 H. Snaith and coworkers on the other hand utilized a mixed halide system -CH 3 NH 3 PbI 3Àx Cl x on both TiO 2 and Al 2 O 3 mesoporous layers. Incredibly, the highest efficiencies (PCE ¼ 10.9%, J SC ¼ 17.8 mA cm À2 , V OC ¼ 0.98 V and FF ¼ 0.63) were obtained for the mesoporous Al 2 O 3 devices where they act purely as scaffolds and do not take part in the electrical processes. 33 These concurrent reports sparked an explosion of research activities where a variety of device congurations, deposition protocols and material sets have been employed.

(a) Device architectures
A common device conguration for CH 3 NH 3 PbI 3 based solar cells consists of inltrating the perovskite within an n-type mesoporous layer ( Fig. 3(a)). The solar cell fabrication process commences with the deposition of a compact TiO 2 hole-blocking layer on top of the uorine doped tin oxide (FTO) substrate. This is typically done through the spray pyrolysis of precursors such as titanium diisopropoxide bis(acetylacetonate) at $450 C. It is important to ensure that the compact layer is pinholefree and uniform to prevent the recombination between carriers from the perovskite layers and FTO. On top of the compact layer, a mesoporous layer of n-type TiO 2 is formed either by screen printing or spincoating a nanoparticle TiO 2 paste followed by annealing to remove the polymeric binders. The thickness and porosity of these layers can be modulated by changing the ller and solvent concentrations in the TiO 2 paste. The perovskite lms are then deposited on top of the n-type mesoporous layer by spincoating it from a solvent such as g-butyrolactone (GBL) or N,N-dimethylformamide (DMF). This is followed by the deposition of a hole transporting material (HTM) such as spiro-OMeTAD with appropriate dopants to improve conductivity. Finally, a metal electrode is deposited on top of the HTM to complete the solar cell.
Although TiO 2 nanoparticles are most commonly used, 32,34-37 there have also been reports of solar cells employing TiO 2 nanosheets, 38 nanorods, 39 nanobers 40 as well as other n-type materials such as ZnO. 41,42 Despite the wide variety of mesoporous architectures being employed, there has been no clear evidence that the efficiencies of the perovskite solar cells can be effectively increased by mesoporous layer modication. The most signicant effect of the mesoporous layer that has been noted is how their thicknesses can affect the power conversion efficiencies. Both in the liquid junction conguration, as well as in its solid state equivalent, it has been demonstrated that lower mesoporous layer thicknesses perform better. 31,32,39 The highest efficiencies in these kinds of congurations have been obtained with a 350 nm thick mesoporous TiO 2 layer inltrated with CH 3 NH 3 PbI 3 . 35 The high efficiencies of these solar cells at relatively low TiO 2 thicknesses (in contrast to $3 mm thickness for solid state DSCs 43 and 10-15 mm thickness for DSCs 44 ) can be traced to the high optical absorption coefficient ($10 5 cm À1 ) of CH 3 NH 3 PbI 3 . 31,45,46 The exact dependence of the mesoporous layer thickness on the PCE is determined by the nature of the perovskite distribution within the TiO 2 layers 47 as well as the perovskite overlayer thickness 34 (which is in turn dependent on the perovskite solution concentration). The demonstration of good efficiencies 39 in rutile TiO 2 nanowires, which have been shown to perform poorly in DSCs (due to poor electron transport), 48 again indicates that the thickness of the mesoporous layer is a more critical factor. An interesting change in this device conguration stems from the work of Etgar and coworkers who have demonstrated that such mesoporous TiO 2 based solar cells do not require a HTM to function. 38,49 The initial study employed anatase TiO 2 nanosheets as the mesoporous layer, 38 onto which the perovskite layers were spuncoated and followed by the evaporation of a gold electrode. The solar cells so fabricated displayed an efficiency of 5.5% (J SC ¼ 16.1 mA cm À2 , V OC ¼ 0.63 V and FF ¼ 0.57). Further optimization of the perovskite layer thickness through consecutive spincoating resulted in an efficiency of 8%. 49 The successful functioning of such devices indicate that CH 3 NH 3 PbI 3 can act as an effective hole transporter. However, the V OC s of such devices are lower than that produced with spiro-OMeTAD indicating that the lack of the electron selection layer results in increased recombination.
Another efficient solar cell conguration that has been chiey employed by Snaith and coworkers is the meso-superstructured solar cell (MSSC). This device conguration employs an insulating mesoporous layer on top of the compact TiO 2 layer as a scaffold to load the perovskite within them. Much of the reports employing this conguration utilize Al 2 O 3 as the mesoporous layer and the mixed perovskite CH 3 NH 3 PbI 3Àx Cl x as the absorber. 33,[50][51][52] However, the high efficiencies observed in this device conguration are not specic to this material combination.
Hagfeldt and coworkers have also demonstrated efficient solar cells (PCE ¼ 10.8%, J SC ¼ 17.3 mA cm À2 , V OC ¼ 1.07 V and FF ¼ 0.59) by employing CH 3 NH 3 PbI 3 deposited on an insulating ZrO 2 mesoporous layer. 36 The lack of an n-type mesoporous layer in this device conguration clearly indicates that efficient electron transport occurs within the perovskite itself ( Fig. 3 (c)). This also indicates that the perovskite within the mesoporous layer is continuous, in contrast to previous studies which had considered them as isolated quantum dots. 30,31 The contrast between both the views could be explained by recent work, 47 which indicated that CH 3 NH 3 PbI 3 exists as two components within the mesoporous TiO 2 , one component with medium range crystalline order (30 atom%) and another with only local structural coherence (70 atom%). The electron transporting nature of the perovskite is further highlighted in the work of Ball et al. 50 By thinning down the Al 2 O 3 scaffold thickness ( Fig. 4(a)), the authors were able to form solar cells which appeared more as thin lm solar cells. The J SC for such solar cells were 16.9 AE 1.9 mA cm À2 for thin alumina scaffold layers (80 nm) indicating that thick perovskite lms could generate sufficient photocurrent. The ability of the perovskites to be employed in such electron transporter-free conguration, as well as to perform efficiently in an HTM-free conguration, 38,49 points towards the ambipolar nature of transport in them. However, it is important to note that the MSSCs do not have the reduced V OC s associated with the HTM-free solar cells primarily due to the presence of the hole selective contacts. The effects of carrier accumulation within the perovskite could also be a possible reason for the enhanced V OC s.
The success of the "pseudo-thin lm" MSSC solar cells has driven explorations into planar thin lm congurations of the perovskite solar cells. Thin lm solar cells allow for a simpler processing route. One approach to realize such thin lm solar cells is to avoid the use of any mesoporous layer and deposit the lms directly on the TiO 2 compact layer. The initial report on MSSCs had included results from a planar conguration (FTO/ compact TiO 2 /CH 3 NH 3 PbI 3Àx Cl x /spiro-OMeTAD/Ag), but with a relatively low efficiency of 1.8%. 33 The difficulties associated with the fabrication of a thin lm solar cell stem from the challenges of depositing a homogeneous pin-hole free perovskite layer through solution based processes. Poor coverage results in poor light absorption as well as shunting paths through the light absorber layer which reduces efficiency. Unlike polymeric lms which are mostly amorphous when deposited by spincoating, perovskites are crystalline. Even mild heating to remove solvent residues can result in dewetting and roughening of the as deposited perovskite lm. Eperon et al. have carefully studied the morphological effects of the underlying substrate, annealing time, temperature and initial perovskite(CH 3 NH 3 PbI 3Àx Cl x ) lm thickness. Through control of the various factors involved, the authors achieved an efficiency of 11.4% (J SC ¼ 20.3 mA cm À2 , V OC ¼ 0.89 V and FF ¼ 0.64). 54 A much simpler approach 55 involved the vapour deposition of the perovskite lms onto the TiO 2 lms which yielded a shortcircuit photocurrent of 21.5 mA cm À2 , an open-circuit voltage of 1.07 V and a ll factor of 0.68, and an efficiency of 15.4%. An interesting vapour-assisted approach to form efficient planar solar cells has also been demonstrated by Yang and coworkers. 56 A 350 nm thick CH 3 NH 3 PbI 3 lm formed on the compact TiO 2 layer yielded an efficiency of 12.1%, despite the relatively thick HTM layer employed in the devices. A similar planar solar cell approach, which allows for low temperature processing, involves the utilization of ZnO as the blocking layer. The rst approach by Kumar et al. employed electrodeposited ZnO on both rigid FTO and exible ITO on top of which CH 3 NH 3 PbI 3 was spuncoated. 41 The second approach utilized spuncoated lms from a solution of 5 nm ZnO nanoparticles suspended in a butanol-chloroform mixture requiring no calcination or sintering. The authors argue that the lack of a constricting mesoporous layer allows for unconstrained CH 3 NH 3 PbI 3 perovskite crystal growth yielding an impressive efficiency of 15.7% on FTO and 10.2% on exible ITO substrates. 57 Another growing area of research involves the use of electron and hole transport layers commonly utilized by the organic photovoltaic community for the implementation of the planar solar cell conguration. The rst work in this direction was performed by Chen and coworkers who employed a device architecture consisting of poly(3,4-ethylenedioxythiophene):poly-(styrenesulfonic acid) (PEDOT:PSS) as the hole transport layer and C 60 derivatives including (6,6)-phenyl C61-butyric acid methyl ester (PCBM) and indene-C 60 bisadduct (ICBA) as the electron transporters in a planar conguration. 58 As with forming perovskites on top of TiO 2 thin lms, the authors faced challenges in forming a uniform CH 3 NH 3 PbI 3 coating on top of the PEDOT:PSS layer. This restricted the total thickness of the CH 3 NH 3 PbI 3 utilized, hence limiting the efficiencies to 3.9%. Sun et al. employed a similar device conguration but succeeded in making a thicker CH 3 NH 3 PbI 3 (110 AE 5 nm) lm through the utilization of a twostep conversion approach. The highest efficiency reported in such a device was 7.41%. 45 Snaith and coworkers 53 screened a wider material set employed in OPVs, including NiO and V 2 O 5 hole transport layers and poly[(9,9-bis(30-(N,N-dimethylamino)propyl)-2,7-uorene)-alt-2,7-(9,9-dioctyluorene)] (PFN) as an electron selective contact with CH 3 NH 3 PbI 3-x Cl x . However, the most efficient device conguration still employed PC 61 BM and PEDOT:PSS as the two previous reports covered above. A $300 nm thick CH 3 NH 3 PbI 3-x Cl x layer ( Fig. 4(b)) yielded an efficiency of 9.8% while exible solar cells were also fabricated on ITO coated glass yielding 6.3%. Crucially, the lower efficiencies on the exible devices were attributed to the smoother surface of the ITO layer which affects the uniformity of the perovskite coating. However, even higher efficiencies have been demonstrated for exible CH 3 NH 3 PbI 3-x Cl x solar cells on ITO by Yang and coworkers (9.2%). 59

(b) Deposition processes
Much of the initial work on CH 3 NH 3 PbI 3 and CH 3 NH 3 PbI 3Àx Cl x solar cells utilized spincoating to deposit the light absorbers from a single precursor solution. CH 3 NH 3 PbI 3 lms were formed by dissolving stoichiometric quantities of CH 3 NH 3 I and PbI 2 in polar solvents such as GBL or DMF. 32,34 CH 3 NH 3 PbI 3Àx Cl x is typically formed from a solution in DMF where the PbCl 2 and CH 3 NH 3 I are in a molar ratio of 1 : 3. 33,50 With appropriate optimization of the precursor concentration and spincoating conditions, the perovskite can be deposited within the pores of the mesoporous layers or be used to form compact layers for planar solar cells. As described previously, the formation of a uniform perovskite layer for planar devices through spincoating requires the optimization of multiple parameters including postdeposition processes. 45,54,58 Due to wettability differences, the perovskite would need to be separately optimized for each underlying layer. When spincoating on mesoporous layers, conditions similar to the deposition of HTMs within the mesoporous TiO 2 layers of a solid state DSC could be expected to occur. 60 In such a case, the excess solution on top of the lm can act as a reservoir during the spincoating process. The amount of inltration within the mesoporous layer would depend critically on the solution concentration, spincoating speed and the solvent utilized. The tendency of the perovskite lms to crystallise, could lead to rough surface morphologies 34 which could introduce shunts into the solar cells.
A signicant development for solution based deposition has been the application of the sequential deposition process (originally developed by Mitzi and coworkers 61 ) for the fabrication of perovskite solar cells by Grätzel and coworkers. 35 The process consists of rst spincoating PbI 2 on the TiO 2 layer from a solution under appropriate conditions (solution concentration, spincoating speed) to enable inltration within the mesoporous layer (Fig. 5(a)). Subsequently, the yellow coloured substrates are dipped in a CH 3 NH 3 I solution in 2-propanol solvent. During the dipping, the yellow PbI 2 converts to form the dark brown CH 3 NH 3 PbI 3 in a few seconds (Fig. 5(b)). It is important to note that the conversion time can vary between the different PbI 2 deposition conditions. A 20 minute conversion time has been reported for perovskite solar cells, 40 while the initial work by Mitzi and coworkers indicated that the conversion required 1-3 h. 61 It is likely that the PbI 2 layer deposited on top of mesoporous substrates has increased roughness that allows the conversion reaction to proceed faster. 35 Due to the volume expansion 56 ($75%) occurring due to the conversion of PbI 2 into CH 3 NH 3 PbI 3 , it can be expected that the mesoporous layer would be better inltrated through the sequential deposition process. The top view of the sequentially deposited samples reveals a highly crystalline lm with complete coverage (Fig. 5(c)).
Apart from the widely utilised solution based deposition processes, vapour deposition has also been employed to form perovskite solar cells. Snaith and coworkers 55 demonstrated efficient planar solar cells (15.4%) of CH 3 NH 3 PbI 3Àx Cl x formed by dual source evaporation of PbCl 2 and CH 3 NH 3 I (Fig. 5(e)). The evaporated lm was sandwiched between a compact TiO 2 layer and a spuncoated spiro-OMeTAD layer acting as electron and hole transporters, respectively. The vapour-deposited lms are extremely uniform, with crystalline features on the length scale of hundreds of nanometres. Another example of vapour deposition was demonstrated by Bolink and coworkers 62 who deposited thin lms of CH 3 NH 3 PbI 3 from the dual source evaporation of PbI 2 and CH 3 NH 3 I. The authors employed organic electron and hole transport layers such as PCBM and PEDOT:PSS to form solar cells with 12.04% efficiency. Both the dual source evaporation examples described above required careful optimization to yield the desired perovskite layers and efficient solar cells. An interesting approach which employs both solution based deposition and vapour phase transformation has been reported by Chen et al. 56 In the vapourassisted solution process (VASP), PbI 2 was rst deposited from solution onto a compact TiO 2 substrate. Subsequently, the lms were exposed to a vapour of CH 3 NH 3 I at 150 C in N 2 for 2 h (Fig. 5(d)). The slow rate of conversion resulted in CH 3 NH 3 PbI 3 lms which exhibited micron sized grains with very low surface roughness of $20 nm. Solar cells made from such lms exhibited an efficiency of 12.1%.
In conjunction with these exciting device-centric advancements, fundamental studies into the photoexcited species and their photogeneration and recombination dynamics in perovskites also began in earnest. The next section traces these early studies to the latest ndings of the fundamental photophysical mechanisms in this system.

Photophysical mechanisms in CH 3 NH 3 PbI 3 thin films
The early photophysical studies in this class of organic-inorganic perovskites are mainly centered on the excitonic properties of layered perovskites such as the optical non-linearity around the excitonic resonances 63 and the exciton-exciton interactions. 64 Coherent transient spectroscopy such as four wave mixing has been used to investigate the exciton and biexciton dynamics 65 and temperature dependent TRPL spectroscopy for the exciton recombination dynamics 66 in these 2D perovskites. Depending on the exciton binding energy E b , the fundamental excited species following photoexcitation could exist as bound electron-hole pairs or as free carriers. Hence, E b have also been extensively investigated for CH 3 NH 3 PbI 3 , the photovoltaic material of choice. Table 1 shows the E b of selected 3-D perovskites with low-dimension perovskites included for comparison. These E b are estimated using optical absorption 7 and magneto-absorption spectra 67,68 as well as from temperature-dependent photoluminescence (PL) intensities. 45,69 The excitons in CH 3 NH 3 PbI 3 are expected to be of the more delocalized Wannier-type with exciton Bohr radius, r B $ 30Å. 7,70 Larger binding energies for CH 3 NH 3 PbBr 3 and the mixed halide CH 3 NH 3 PbI 3Àx Cl x system indicate a more tightly bound nature of the excitons resulting from the halogen substitution. In contrast, for 2-D layered perovskites, the excitons are the Frenkel type and their large binding energies are ascribed to dielectric modulation between the organic and inorganic layers and the two dimensionality of the inorganic structure. 71 With decreasing dimensionality to 1-D and 0-D, the exciton binding energies increase in accordance with the quantum connement effects. 70 Absorption of photons creates electron-hole pairs in perovskite. Following carrier thermalization, these carriers could either continue to exist as free carriers or form excitons depending on the exciton binding energy. Presently, it is still unclear whether these species at room temperature exist as excitons or free chargeswhich gave rise to the 3-D perovskite's exceptional properties of long diffusion lengths. This uncertainty stems from the low exciton binding energies E b ranging from 19 meV to 50 meV, which are comparable to the room temperature thermal energies of k B T $ 25 meV. Given the uncertainties and the assumptions involved with the estimation/extraction of the values for E b based on the various methods employed (optical absorption, 73,74 magneto-absorption 68 and temperature dependent PL 75 ), it is reasonable from Table 1 that the E b for CH 3 NH 3 PbI 3 is comparable to the thermal energies of k B T $25 meV at room temperature. As an illustration, even for an E b $19 meV, the faction of the excitons with an energy greater than E b at $300 K can be calculated from statistical physics to yield: $57% of the photo-generated excitons dissociating spontaneously and $43% remaining as excitons.
It is an open question on how the co-existence of free carriers and excitons, whose dynamic populations could vary over their lifetimes, will affect the carrier dynamics in perovskite solar cells.
To the best of our knowledge, there have been no reported studies on the charge dynamics in 3D perovskites until the advent of the 9.7% perovskite solar cells in 2012. 32 Despite the rapid progress in organic-inorganic perovskite solar cells, the fundamental photophysical processes driving the high performance of these devices is still severely lacking. The bulk of the research efforts to date are predominantly focused on device development, with limited studies on charge carrier dynamics in the CH 3 NH 3 PbI 3 perovskite materials. 32,33,46,[76][77][78] Nevertheless, efforts into applying ultrafast optical spectroscopy (UOS) techniques to investigate the structure-function relationships in perovskite solar cells are on the rise. UOS techniques are powerful probes of carrier dynamics and charge transfer mechanisms in materials. A clear understanding of the charge generation and transport mechanisms in perovskite solar cells will provide valuable feedback to guide the materials design and device engineering. Here, we rst examine the intrinsic charge dynamics in the bare perovskite thin lms. The charge transfer processes in a typical perovskite solar cell will be examined in the next section.

(a) Long electron-hole diffusion lengths and hot hole cooling dynamics
In tandem with the developments of increasing perovskite solar cell efficiencies, indications of ambipolar charge transport in perovskites became apparent when efficient perovskite-based devices in a broad range of device architectures are reported. The perovskite material functions well as an absorber in a conguration used by Kim et al. 32 and Heo et al. 34 that sandwiches the thin perovskite layer between a mesoporous TiO 2 photoanode and a HTM layer (spiro-OMeTAD). Lee et al. 33 demonstrated that the perovskite material can also work effectively as both an absorber and an electron transporter by fabricating solar cells with an insulating Al 2 O 3 scaffold instead of the TiO 2 photoanode. Surprisingly, Etgar et al. 38 fabricated devices with appreciable performance in a conguration without the HTM layerindicating that the perovskite material can also work as an absorber and a hole transporter.
These reports provide us with a compelling case to devise quenching experiments utilizing femtosecond transient optical spectroscopy (i.e., TAS and time-resolved PL (TRPL) spectroscopy) of CH 3 NH 3 PbI 3 heterojunctions with selective electron or hole extraction to decouple the electron and hole dynamics in this material. 46 Our ndings revealed clear evidence of balanced and long-range electron-hole diffusion lengths of at least 100 nm in solution processed CH 3 NH 3 PbI 3 (Fig. 6). Concurrently, using the same PL quenching approach, H. J. Snaith and coworkers 77 also performed diffusion length measurements on CH 3 NH 3 PbI 3 and the mixed halide CH 3 NH 3 PbI 3Àx Cl x . Their ndings of the electron-hole diffusion lengths in CH 3 NH 3 PbI 3 concur with ours. Amazingly, the mixed halide CH 3 NH 3 PbI 3Àx Cl x possesses diffusion lengths one order longer (i.e., >1 mm) than  79 determined the effective masses of both the electron and hole in CH 3 NH 3 PbI 3 to be small, (i.e., m e * ¼ 0.23m 0 and m h * ¼ 0.29), thus providing further validation of their longrange ambipolar charge transport property. 46,77 In this same work, we also examined the early time relaxation dynamics in the CH 3 NH 3 PbI 3 system. It is important to note that the carrier dynamics in the perovskite system are strongly pump uence dependent due to their large optical absorption coefficients and long charge diffusion lengths. Multi-particle Auger (third order) recombination processes becomes dominant for pump uence >2.6 mJ cm À2 . In fact, we recently discovered that at a pump uence >12 mJ cm À2 , amplied spontaneous emission (ASE) prevails and even outcompetes the Auger processes. 80 Therefore, careful control of the pump uence in ultrafast optical spectroscopy of perovskites is absolutely essential for uncovering their intrinsic photophysical properties. Femtosecond TAS measurements with selective 400 nm and 600 nm pump excitation (uence < 1.3 mJ cm À2 ) and a white-light continuum (WLC) probe uncovered a slow 0.4 ps hot hole cooling process from a deeper valence band level (VB2) to the valence bandedge (VB1)see Fig. 7. With careful tailoring of the HTM's energy levels, one could efficiently extract these hot hole energies before they cool down to VB1. Potentially, this could be utilized in perovskite solar cells to exceed the theoretical Shockley-Queisser limit. 81 Further investigations into this area should be conducted.

(b) Origins of the long electron-hole diffusion lengths
The mechanism of the long electron-hole diffusion lengths in CH 3 NH 3 PbI 3Àx Cl x and CH 3 NH 3 PbI 3 and their charge carrier mobilities were elucidated by L. M. Herz in collaboration with H. J. Snaith and coworkers 78 using transient THz spectroscopy and TRPL spectroscopy. Their ndings show that both the monomolecular (rst order, i.e., from geminate recombination of excitons and/or from trap-or impurity-assisted recombination) and bimolecular (second order) charge carrier recombination rates are extremely low, with the latter defying the Langevin limit by at least four orders of magnitude. However, the Auger (third order) recombination rates were found to be high $10 À29 cm 6 s À1comparable to those of strongly conned colloidal quantum dots. 82 Comparatively, Auger recombination in highly-doped bulk Si wafers is $2 orders smaller. 83 The lower bound values of the charge carrier mobilities for  View Article Online CH 3 NH 3 PbI 3Àx Cl x and CH 3 NH 3 PbI 3 were determined to be 11.6 cm 2 V À1 s À1 and $8 cm 2 V À1 s À1 , respectively, which are extremely high for the solution-processed perovskites. Comparatively, these values are >20 times larger than that of mesoporous TiO 2 and several orders larger than those of typical p-conjugated molecular semiconductors. The origins of the long electron-hole diffusion lengths stem from the novel combination of low charge carrier recombination rates and high charge carrier mobilities in these perovskites.

(c) Summary of photophysical processes in pristine lms
A generalized scheme of the dynamic interplay of the various photophysical processes and loss mechanisms in bare perovskite thin lms following photoexcitation is shown in Fig. 8. Absorption of photons results in the generation of electronhole pairs that evolve towards the formation of highly delocalized Wannier excitons aer thermalization. A fraction of which would dissociate spontaneously back into free carriers. The excitons and free carriers coexist and their dynamic populations continue to vary over their lifetimes. Geminate recombination of the excitons, or the recombination involving an electron and a hole generated from the quenching of a single exciton is inefficient. Likewise, trap-assisted recombination, another monomolecular process, is also suppressed in these CH 3 NH 3 PbX 3 perovskites. At stronger light intensities, nongeminate recombination originating from the recombination of two free charges (i.e., bimolecular in nature) is extremely lowdefying the traditional Langevin limit by at least four orders of magnitude. Auger recombination (involving a three particle process) on the other hand is dominant in CH 3 NH 3 PbX 3 perovskites and surprisingly, ASE even occurs at higher pump excitationsout-competing the Auger processes in CH 3 NH 3 PbI 3 . 80 Eventually, in these bare perovskite lms without a HTM and in the absence of any charge extraction, the photoexcited species (excitons and free carriers) undergo radiative (luminescence) or non-radiative processes within the perovskite. It is also important to note that under solar light intensities (low intensity excitation), Auger recombination or ASE would be strongly suppressed. With the non-radiative pathways (geminate recombination, trap-assisted recombination and Auger recombination) weak or inactive under solar light intensities, it is therefore understandable that these perovskites make excellent photovoltaic materials.

Charge transfer mechanisms in perovskite solar cells
As described previously, various congurations of perovskite solar cells have been explored. These include the congurations where the perovskite is interfaced with mesoporous TiO 2 , mesoporous Al 2 O 3 (MSSC) as well as other organic electron and hole transport layers. Ultrafast measurements have clearly indicated that CH 3 NH 3 PbI 3 and CH 3 NH 3 PbI 3Àx Cl x , both have long and balanced electron-hole transport lengths. 46,77 It has also been clearly shown that injection into electron acceptors such as PCBM and hole transporters such as PEDOT:PSS and spiro-OMeTAD is efficient. 46,53,59,77 Thus, planar congurations of perovskite solar cells can be expected to function as thus: under photoexcitation, a mixture of weakly bound excitons and direct electron and hole generation occurs. Due to the crystalline nature of the perovskite as well as low trap densities, the recombination within the perovskites is limited. The long-lived nature of electrons and holes allows them to be collected by electron and hole acceptor layers, before making their way out of the solar cells as photocurrent. The high open circuit voltages noted in such planar congurations also point to a scenario where energetic costs associated with exciton splitting is not prevalent. Light intensity dependent measurements on a PEDOT:PSS/CH 3 NH 3 PbI 3Àx Cl x /PCBM solar cell indicate that the main recombination mechanism is free carrier recombination. 59 The meso-superstructured solar cells 33,36,50,51 which utilize an insulating mesoporous layer as a scaffold for the perovskite can be expected to work in a similar manner. Here the primary charge separation interface is at the perovskite/hole transporting layer (spiro-OMeTAD) into which Fig. 8 A schematic of the photophysical processes and loss mechanisms in perovskites following photoexcitation. Efficient (or strong) pathways and suppressed (or weak) pathways are denoted by the black and grey lines, respectively. Monomolecular recombination is charge carrier density independent, while bimolecular and Auger recombination are charge carrier density dependent processes that would typically be present under high intensity photoexcitation. In fact, under even higher photoexcitation densities, amplified spontaneous emission (ASE) will out-compete Auger recombination. However, under solar light intensities (low intensity excitation), these latter processes will be strongly suppressed. hole injection occurs efficiently. The good electron transport properties of the perovskite layer ensure that electrons are collected through the compact TiO 2 layer below the insulating mesoporous layer. Thus, a connected perovskite layer within the mesoporous layer is critical. The exact working principle behind solar cells fabricated by depositing the perovskite on a mesoporous TiO 2 layer is still not clearly established. The pertinent questions when considering this architecture are: (a) is there injection of electrons from the perovskite into mesoporous TiO 2 ; (b) is the more efficient path for electron collection through the mesoporous TiO 2 or within the perovskite itself? (c) Which pathway results in higher extraction efficiency in the perovskite solar cell? N. G. Park in collaboration with M. Grätzel and coworkers 32 published the earliest study on the dynamics of the charge separation processes in CH 3 NH 3 PbI 3 /TiO 2 solar cells probed using femtosecond transient absorption spectroscopy (fs-TAS). However, clear evidence of efficient electron injection into TiO 2 could not be observed due to the overlapping signals from the stimulated emission from the perovskite. Although the question on whether perovskite can inject electrons into the mesoporous TiO 2 has more or less been answered by photoinduced absorption spectroscopy (PIA) from the evidence of a broad absorption feature in the region $1.1 mm attributed to free electrons in the titania from CH 3 NH 3 PbI 3Àx Cl x , 33,85 clear evidence of efficient injection remains elusive. Most recently, J.-E. Moser in collaboration with M. Grätzel and coworkers unravelled the mechanism of the charge transfer processes in perovskite solar cells ( Fig. 9(a)) and presented clear evidence of not only efficient electron injection from photoexcited CH 3 NH 3 PbI 3 into TiO 2 , but also efficient hole injection from photoexcited CH 3 NH 3 PbI 3 into the HTM (spiro-OMeTAD) occurring simultaneously over comparable ultrafast timescales (#3 ps). 84 They had overcome the challenges of a spectral overlap encountered in the earlier study 32 by probing in the infrared (1.4 mm) which allows direct observation of only the carrier's population decay within the perovskite itself ( Fig. 9(b)).
The second and third questions on whether electron collection is more efficient through the mesoporous TiO 2 or within the perovskite itself; and which pathway results in higher extraction efficiency in perovskite solar cells remains open. H. J. Snaith in collaboration with T. N. Murakami and T. Miyasaka and coworkers 33 showed using transient photocurrent measurements that the charge collection in the insulating Al 2 O 3 -based devices is faster than the TiO 2 -based devicesindicating that the perovskite material itself is more efficient in transporting the negative charge than mesoporous TiO 2 . However, recent ndings by Marchioro et al. using transient absorption spectroscopy showed that the amount of long-lived charges in the TiO 2 /CH 3 NH 3 PbI 3 /HTM samples is higher than that in the Al 2 O 3 /CH 3 NH 3 PbI 3 /HTM samples ( Fig. 9(b))indicating a more efficient charge separation in the former. Furthermore, charge recombination with oxidized HTM species was also found to be slower on TiO 2 lms compared to Al 2 O 3 lms (Fig. 9(c)). The efficiency of charge extraction in a perovskite solar cell also depends on the ratio between charge recombination and charge separation rates. Nevertheless, Marchioro et al.'s ndings showed that it is advantageous to use TiO 2 as the electron acceptor and transporter, together with a HTM in perovskite solar cellsproviding only partial answers to question 3.
Impedance spectroscopy measurements have also been applied to perovskite solar cells in a bid to understand the electrical processes occurring in them. Bisquert and coworkers analysed perovskite solar cells on TiO 2 and ZrO 2 and proposed that carrier accumulation occurs in the perovskite layers. 86 This indicates that the working principle of the perovskite solar cells is different from that of a pure DSSC where instantaneous injection from the dye into the TiO 2 occurs and no charge accumulation is observable in the light absorber. Dualeh et al. in their impedance spectroscopy analysis have argued that the observation of the mesoporous TiO 2 chemical capacitance in perovskite solar cells indicates that the electron transport is channeled through the mesoporous TiO 2 . 87 The capacitance of TiO 2 in perovskite solar cells has also been observed by Abrusci et al. in their differential capacitance measurements when employing a C 60 -self assembled monolayer on top of the TiO 2 surface. 85 Zhu and coworkers also point to a similar working mechanism in solid state DSCs and perovskite solar cells dominated by electron transport within the mesoporous TiO 2 layer itself, from intensity modulated photocurrent/photovoltage spectroscopy (IMPS/IMVS) measurements. 88 However Bisquert and coworkers have reported a similar impedance spectral shape in planar CH 3 NH 3 PbI 3Àx Cl x and nanostructured CH 3 NH 3 PbI 3 solar cells. Since there is an absence of the TiO 2 mesoporous layer in one of the devices, they attribute the features in the impedance spectra to charge transport and recombination processes occurring within the perovskite layer solely. The authors have also noted that more complicated impedance spectral patterns can be observed in other cases which necessitates the development of a complete model.
The unclear view of the exact working principles of perovskite-mesoporous TiO 2 solar cells could stem from the variations in the solar cell structures studied and their ensuing interfacial charge transfer dynamics. The impact of the degree of coverage of the mesoporous TiO 2 by the perovskite is unclear. The presence or absence of a perovskite overlayer on the mesoporous TiO 2 will modulate the recombination at the perovskite/TiO 2 /HTM interfaces. Additionally the distribution of the perovskite within the TiO 2 pores may also play a dening role in what mechanism of electron transport dominates. For example, the charge transport time in CH 3 NH 3 PbI 3 formed from the sequential deposition process was found to be 3 times longer than that obtained from the single precursor spincoating process. 36 From spectroscopic evidence, it is clear that charge injection from perovskite into mesoporous TiO 2 is possible; 33,84,85 however, the exact ratio between electrons being transported through the mesoporous TiO 2 and that transported through the perovskite is unknown. Possibly, spectroscopic measurements in working solar cells at different applied voltages may unravel this question.

Future directions and open questions in the field
In the wake of the rapidly expanding eld, several pressing open questions remain. They have been largely swept aside in the initial surge for high efficiencies. As the perovskite photovoltaic eld matures, these gaps in our understanding need to be progressively tackled and lled: (a) Interfacial charge transfer dynamics For any given material system, factors such as device processing and the ensuing morphology and coverage can strongly affect the carrier dynamics and the device performance. Systematic studies into understanding the morphological effects arising from the single precursor vs. sequential deposition solutionprocessed approach or thermal evaporation approach on the charge dynamics in pristine and heterojunction perovskite lms are urgently needed. Given that the 100 nm e-h diffusion lengths estimated from the single step processed CH 3 NH 3 PbI 3 are minimum values, 46 further work is urgently needed to carefully examine the e-h diffusion lengths in sequential deposited lms and evaporated lms to establish their upper limits. Over the next phase of development in this eld, one can look forward to an upsurge of systematic transient spectroscopy studies of various pristine/heterojunction congurations over the entire charge generation to charge extraction timescale, in particular: (i) the dynamic interplay and interactions between the excitons and the free charge population and their effects on the charge separation, recombination and transport; (ii) the interfacial charge dynamics and the mechanism of the charge injection at the perovskite/mesoporous TiO 2 ; perovskite/semiconductor nanostructure; and perovskite/graphene interfaces; and (iii) the exact role of higher lying states (if any) for charge separation. Time-resolved optical pump-terahertz probe spectroscopy would be extremely useful to monitor the early time excitons and free-carrier dynamics and populations in the pristine lms and heterojunctions. These studies could be extended with transient microwave photoconductance measurements over longer time scales for correlation with device properties. Double excitation techniques like the pump-push probe spectroscopy can be used to study the higher lying electronic states and their roles in the charge transfer. Field modulation techniques like eld assisted pump-probe spectroscopy could also lead to new insights into the interplay of the exciton dissociation and the free carrier dynamics on the charge transfer under device conditions. With the concerted efforts of various research groups, a complete picture and detailed model of the charge transfer processes of perovskite solar cells will be unravelled. Understanding the structure-function relationships in perovskite solar cells through UOS holds the key to the development of optimal solar cells with efficiencies that could surpass the 20% target.
(b) CH 3 NH 3 PbI 3 vs. CH 3 NH 3 PbI 3Àx Cl x Highly efficient solar cells have been prepared from CH 3 NH 3 PbI 3 and CH 3 NH 3 PbI 3Àx Cl x . The precursor solution of the latter compound consists of 3 parts of CH 3 NH 3 I for 1 part of PbCl 2 . Although initial studies had termed the mixed composition as CH 3 NH 3 PbI 2 Cl, the crystallographical and optical absorption properties are identical to that of CH 3 NH 3 PbI 3 . 33 Stranks et al. have shown that 77 the diffusion lengths measured in CH 3 NH 3 PbI 3Àx Cl x are signicantly longer than that in CH 3 NH 3 PbI 3 , although the device performance of both compositions are comparable. In addition, while the charge carrier mobilities in both CH 3 NH 3 PbI 3Àx Cl x and CH 3 NH 3 PbI 3 are similar, the bimolecular recombination rates in the mixed halide system are approximately one order lower, suggesting that electronic structure modulation by Cl has a role in reducing the spatial overlap of the electron and holes. 78 Colella et al. have also proposed that Cl within CH 3 NH 3 PbI 3 can act as a dopant, improving the transport properties. 89 XPS measurements have revealed that the nal composition of the CH 3 NH 3 PbI 3Àx Cl x contains a signicantly low amount of Cl (Cl/Cl + I ¼ 2.2%) within them. 59 XRD analysis of evaporated CH 3 NH 3 PbI 3Àx Cl x lms 55 also revealed impurity peaks of PbI 2 , although the precursors were PbCl 2 and CH 3 NH 3 I. This points to the Cl source (either PbCl 2 or CH 3 NH 3 Cl) playing a role in the lm formation. The development of a preferred orientation in CH 3 NH 3 PbI 3Àx Cl x lms 33 in contrast to CH 3 NH 3 PbI 3 lms 35 again indicates that the Cl plays a role in lm formation. Studies on CH 3 NH 3 PbI 3 and CH 3 NH 3 PbI 3Àx Cl x lms with similar lm properties (morphological and crystallographical) are required to reveal the electronic role of Cl within the perovskite.

(c) Hole transporting layers
The hole transporting material of choice for perovskite solar cells is spiro-oMeTAD, which has been well studied due to its popularity in solid-state DSCs. However, the high commercial prices of spiro-oMeTAD (due to synthesis complexity) as well as its tendency to be uncontrollably doped by O 2 , necessitate the development of alternatives. The wide variety of hole transporting organic layers originally developed for organic thin lm transistors as well as light emitting devices could possibly be applied in conjunction with the perovskite system. An early study on alternatives to spiro-oMeTAD for perovskite solar cells had indicated that thiophene based systems may not be suitable. 34 Arylamine based hole transporters such as poly-triarylamine, 34 N,N-di-p-methoxyphenylamine-substituted pyrene derivatives 90 and swivel cruciform thiophene-based molecules 91 have shown performances between 11 and 12%. An interesting development 92 has been the utilisation of the inorganic CuI hole transport layer with CH 3 NH 3 PbI 3 . Although they performed worse than spiro-oMeTAD solar cells due to high recombination, the work does indicate the possibility of applying other inorganic hole transporters such as CuSCN to the perovskite system. An equally intriguing avenue is the application of nanocarbon (e.g. graphite, carbon nanotubes, graphene/polymer composites) based hole transporters. As has been pointed out by Johansson and coworkers, the efficiencies can be primarily determined by recombination at the perovskite-HTM interface. 76 The low thicknesses of HTM layers ($100 nm) in well optimised solar cells as well as the utilisation of dopants point to very low losses within the HTM itself, indicating the crucial nature of the perovskite-HTM interaction.

(d) Newer perovskite compositions
Much of the high efficiency solar cells have utilised CH 3 NH 3 PbI 3 as the light absorber. Although this composition currently yields the most efficient solar cells, newer perovskite compositions which have tuneable bandgaps would be of interest to tandem solar cell congurations or in power applications that require high voltages. Seok and coworkers have elegantly shown how partial substitution of I with Br can yield colourful solar cells with varying photocurrent onset. One benet of the partial substitution seems to be improved stability. 37 Similarly, Hodes and Cohen have demonstrated CH 3 NH 3 PbBr 3Àx Cl x solar cells with open circuit voltages as high as 1.5 V. 93 Another key effort required is the reduction of the bandgap of the perovskite solar cells for increased spectral response and therefore improved efficiencies. CH 3 NH 3 PbI 3 based solar cells do not efficiently harvest photons close to its optical absorption onset (600-780 nm), resulting in photocurrents not approaching the theoretical maximum. The development of formamidinium (HC(NH 2 ) 2 + ) lead perovskites 94-97 with a lower bandgap (1.48 eV) that allows for high photocurrents is thus a promising development. These high photocurrents allowed Eperon et al. to demonstrate planar heterojunction solar cells with power conversion efficiencies of up to 14.2%. 94 The most pressing demand for newer photoactive perovskites is driven by the need to replace Pb. Hodes has estimated that a production capacity of 1000 GW per year from CH 3 NH 3 PbI 3 solar cells requires less than 10 000 tons of leadmuch lower than the 4 million tons per year of lead used for lead-acid batteries currently. 98 Thus, although the total amount of lead that is required is low, the risk of leaching Pb into the environment needs to be managed. Sn 2+ may serve as a replacement to Pb 2+ , but its tendency to be easily oxidised is a drawback. 99 Computational predictions play a critical role in Pb replacement efforts, due to the large variety of halide perovskites possible.

(e) Towards printability and scalability
The rapid growth in the efficiencies of perovskite solar cells within a short time period has been catalogued previously. 98 101 The device architecture consisted of a screen printed TiO 2 layer, a ZrO 2 layer and a carbon black/graphite composite layer onto which CH 3 NH 3 PbI 3 was dropcast. Such devices displayed an efficiency of 6.64%. A critical barrier in the implementation of a printable perovskite solar cell is the deposition of the electron selective compact layer (typically TiO 2 formed by spray pyrolysis at high temperature). Planar perovskite solar cells employing organic electron and hole transport layers do not suffer from this limitation, but still display lower efficiencies than mesoporous layer based solar cells. 45,53,58,59,62 Approaches to avoid the high temperature processes during the formation of the compact layer includes the utilization of electrodeposited ZnO 41 or a spuncoated ZnO nanoparticle layer. 57 A similar nanoparticle based approach has been demonstrated by the utilization of a spuncoated graphene ake/ TiO 2 nanoparticle layer as the compact layeryielding efficiencies of 15.6%. 102 Very recently, a new low temperature compact layer TiO 2 deposition recipe has yielded impressive efficiencies of up to 15.9% for a solar cell fabricated at temperatures less than 150 C. 103 Much of the high efficiency reports on perovskite solar cells have been demonstrated on cell areas much less than 1 cm 2 ( Table 2). Malinkiewicz et al. who reported an efficiency of 12.04% (cell area ¼ 0.09 cm 2 ) fabricated an $1 cm 2 solar cell using the same process which yielded an efficiency of 8.27%. 62 The primary difference in the photovoltaic parameters was the FF which reduced from 0.67 to 0.52 when going to the larger area solar cells. The reason for the reduction in the FF is unclear, but could arise from series resistance in the electrode itself. The rst perovskite based modules have also been reported (up to 16.8 cm 2 module area) by Matteocci et al. who employed both P3HT and spiro-OMeTAD as hole transporters yielding efficiencies of 5.1%. 104 Such large scale demonstrations of perovskite solar cells are necessary for perovskite solar cells to develop into a technology for widespread deployment.

(f) Stability studies
The 3-D methylammonium trihalogenoplumbates (II) crystals (i.e., CH 3 NH 3 PbCl 3 , CH 3 NH 3 PbBr 3 and CH 3 NH 3 PbI 3 ) are known to undergo 1 st order phase transitions. 105,106 In particular, CH 3 NH 3 PbI 3 undergoes a tetragonal I to cubic phase transition at $327 K (or $54 C), which is close to the device operating temperatures under direct sunlight. This cubic-tetragonal transition results in the methylammonium ions exhibiting a disordered character while the PbX 6 octahedron exhibits a displacive character. 106,107 Systematic studies of such effects on the carrier dynamics and on the device properties under operation are warranted. Since the phase transitions involve a volume change, it is pertinent to investigate whether inltrating the perovskite within mesoporous layers can improve temperature stability. Reports on long term device performance tests are few, with 500 h stabilities under constant illumination being reported by Grätzel and coworkers. 35 Snaith and coworkers have reported performance data for 1000 h in MSSCs and have pointed towards UV induced changes in TiO 2 based solar cells due to desorption of surface-adsorbed oxygen. 108 Systematic studies on the degradation mechanisms (including the hole transport layers) are required to manage the lifetimes of these solar cells. Newer perovskite compositions which do not undergo phase transitions at device operational temperatures should be pursued as well.
In summary, the eld of perovskite solar cells has rapidly grown to become the most efficient solution processed photovoltaics, leapfrogging other 3 rd generation photovoltaic technologies. As this eld matures, rapid improvements in the power conversion efficiencies will become harder to come by. Thus, a deeper understanding of the fundamental working mechanisms become increasingly important especially if efficiencies exceeding 20% need to be achieved. Concurrently, technological developments in the area of scalable manufacturing techniques, toxicity and stability need to be effectively addressed.