Halide perovskites with ABX₃ formula (A=monovalent cation, B=divalent cation, X=halogen) form corner sharing (B–X) octahedra and cubo-octahedra (A–X). The recently studied methylammonium lead (or tin) halide perovskite materials for photovoltaics were originally discovered in 1978.^1,2  Colorless methylammonium lead chloride was observed to turn orange and black upon replacing the chloride anion with bromide and iodide, respectively, which is indicative of a decrease in the band gap energy. Despite the recent use of methylammonium lead iodide as light absorber in photovoltaic solar cells, little attention was initially paid to such a possibility, probably because the material has other interesting electrical properties (including superconductivity), which was discovered for layered organic–inorganic perovskite materials in 1994.³  In 2009, Miyasaka et al. were the first to report that CH₃NH₃PbX₃ (X=Br and I) perovskite materials showed light harvesting properties when they were used as sensitizers in a liquid electrolyte based dye-sensitized solar cell.⁴  In this cell, the bromide perovskite was deposited on nanocrystalline TiO₂ and displayed a power conversion efficiency (PCE) of 3.1%, with its iodide analogue showing a slightly higher PCE of 3.8%. However, this work was not cited for two years, probably due to problems in reproducing the recipe. In 2011, Park et al.⁵  reported a higher PCE of 6.5% using CH₃NH₃PbI₃ and optimizing the coating solution concentration, TiO₂ film thickness and electrolyte formulation. As can be seen in Figure 1.1, CH₃NH₃PbI₃ nano-dots deposited sparsely on a nanocrystalline TiO₂ surface show a one order of magnitude higher absorption coefficient than the ruthenium-based N719 dye fully covered nanocrystalline TiO₂ surface. Although this work demonstrated the potential of organic lead halide perovskites as photovoltaic materials, CH₃NH₃PbX₃ could not be considered as a promising sensitizer in liquid based dye-sensitized solar cells because of its rapid dissolution in polar electrolyte solutions (Figure 1.2).

In 2012, Park et al.⁶  solved the instability problem of perovskites in polar liquid electrolytes by replacing the liquid with a solid hole-transporting material, leading to the first demonstration of a long-term stable all-solid-state mesoscopic perovskite solar cell with an efficiency of 9.7%. Figure 1.3a shows cross-sectional SEM images of a solid-state perovskite solar cell employing a nano-dot perovskite deposited on a TiO₂ surface. Since the perovskite nano-dots are isolated from each other, electron injection from the perovskite nano-dots to TiO₂ occurs even without encapsulation. This structure is quite stable up to 500 h, as can be seen in Figure 1.3b. This is due to the fact that hydrophobic spiro-OMeTAD protects the perovskite from moisture as it is coating completely the perovskite nano-dots.

Following these three important pioneering works,^4–6  there has been a huge surge of interest in perovskite solar cells. Based on Web of Science data, around 500 peer-reviewed research papers on perovskite solar cells were published in 2014, and the rate of publications continues to increase. As can be seen in Figure 1.4, the PCE of perovskite solar cells has increased rapidly as a consequence of improvements on the perovskite film quality and other structural modifications. As a result, a certified PCE of 20.1% was reported in 2015.⁷ 

Efficiency improvements are related to the perovskite solar cell structure and perovskite film quality. Figure 1.5 shows the evolution of device structures. Diverse structures have been developed, ranging from the sensitized structure to the p-i-n junction structure. The first version of the perovskite solar cell was based on the perovskite-sensitized concept (Figure 1.5a), analogous to the solid-state dye-sensitized solar cell.⁸  Since the perovskite is deposited on the TiO₂ surface without necking between the nano-dots, photocurrent generation can be explained by electron injection from the perovskite into TiO₂. However, it has been demonstrated that the perovskite also works without electron-accepting TiO₂, when Al₂O₃ is used as the scaffold instead, and a thin layer of perovskite is formed on the Al₂O₃ surface with perovskite networking.⁹  The super-structured device configuration in Figure 1.5b underlines the importance of electron transport in the perovskite. Since the perovskite has the ability to transport electrons, it can be used as a bifunctional material acting as light harvester and n-type semiconductor. Based upon this bifunctionality, the perovskite can be used to completely fill the pores in the TiO₂ film, instead of the usual hole-transporting spiro-OMeTAD. This pore filling creates a capping layer comprising purely the perovskite, eventually leading to the mesoscopic perovskite solar cell structure shown in Figure 1.5c.¹⁰  The mesoscopic structure has two different perovskite morphologies, with nanocrystals in the mesopores and a bulk thin film in the capping layer. Thus, two different pathways for electron transport can be expected. Since electron and hole diffusion lengths in the perovskite have been found to be almost identical,^11,12  the mesoporous oxide layer is not necessarily required to construct a perovskite solar cell. Figure 1.5d shows a planar heterojunction structure employing a compact thin TiO₂ electron-acceptor layer and a hole-accepting/transporting material with the perovskite layer in between.¹³  Since the perovskite has unique opto-electronic properties, as mentioned previously, it can be applied to many types of solar cell structures. Figure 1.5e illustrates an inverted layout similar to the configuration of an organic photovoltaic device, with the perovskite layer being sandwiched between the transparent electrode with a hole-collecting PEDOT:PSS layer and the metal electrode with an electron-collecting PCBM.^14,15 

The three-dimensional organolead halide CH₃NH₃PbI₃ has been found to display a special feature involving the molecular motion of methylammonium ion.^16,17  The CH₃NH₃⁺ ion can have two types of orientational disorder; one is the orientation of the C–N axis relative to the crystal axes, and the other is related to the rotation around the C–N axis, as can be seen in Figure 1.6.¹⁸  It has been observed that CH₃NH₃PbX₃ undergoes a temperature-dependent phase transition for X=Cl, Br and I due to the molecular motion of CH₃NH₃⁺.¹⁷  For instance, the CH₃NH₃PbI₃ structure changes from orthorhombic to tetragonal at 162.2 K, and from tetragonal to cubic at 327.4 K.¹⁷  According to temperature-dependent XRD analysis and complex permittivity results, dynamic disorder of methylammonium cation occurs in the high temperature phase,¹⁷  resulting in a cubic phase. It is noteworthy that the phase transition is accompanied by a large volume change of the unit cell between the cubic (253.5×10⁶ pm³) and tetragonal (992.6×10⁶ pm³) structures. The molecular motion-induced phase transition is depicted in Figure 1.7.

The molecular motion of the CH₃NH₃⁺ ion can induce or affect not only phase transitions, but also other physical properties such as ferroelectric polarization and ion migration. Ferroelectric polarization of CH₃NH₃PbI₃ has been observed directly by piezoresponse force microscopy (PFM).²⁰  Figure 1.8 shows PFM phases along with the topography observed in the dark and under illumination for 700 nm-sized CH₃NH₃PbI₃ cuboids. PFM phase images were obtained without bias voltage (unpoled) and with bias voltages of +3 V (positive poling) and −3 V (negative poling). In the dark, spontaneous polarization is observed even in the absence of an electric field (unpoled), which is due to the large freedom of rotation of the polar CH₃NH₃⁺ ion.²¹  The poling effect is evident for both positive and negative bias potentials, where dipoles do not rotate 180° completely upon changing from +3 V to −3 V. Such imperfect rotation of dipoles with respect to the applied electric field is probably due to their interaction with defects in the crystals. Light-induced polarization using the PFM measurement system equipped with a white light source (50 000 lux) has shown that the spontaneous (unpoled) polarization is partly attenuated by light irradiation, which indicates that the ferroelectric polarization is screened by photo-generated conduction electrons. However, photo-induced polarization is pronounced in the presence of an electric field for positive poling. Upon changing the poling polarity, a similar tendency is observed as in the dark.

The ferroelectric polarization remains unchanged even after removal of the external electric field, as can be seen in Figure 1.9. The PFM phase of the positively poled perovskite was measured at intervals of 30 min under illumination, and highly aligned dipoles were observed for 60 min. It was found that the retention of the light-induced polarization depends on the CH₃NH₃PbI₃ crystal size: larger crystals demonstrated a better retention behavior than smaller ones.²⁰  The observed size-dependent retention behavior is probably related to defects in the perovskite crystals.²² 

It is expected that iodide ions may be able to migrate in the perovskite ionic crystal due to molecular motion. Dynamic CH₃NH₃⁺ motion is likely to lead to the weakening of the chemical interactions with iodide, which can be freed from restrictions in the lattice. Ion migration has been observed in the presence of a bias potential under illumination. Interestingly, the effect could be switched by reversing the bias polarity as shown in Figure 1.10.^23,24  Utilizing the field switchable photovoltaic effect in organolead halide perovskites could offer several advantages. Using a single perovskite layer solar cell without electron and hole transporting layers might be feasible as free charges in the perovskite can be transported effectively by self-doping via poling. If one could fabricate a lateral structure with spacing between the metal electrodes, a TCO (transparent conductive oxide)-free solar cell with an extremely high photovoltage could be realized. The demonstration of electric-field-manipulated doping opens up new applications for perovskites in devices such as both electrically- and optically-readable memristors.

In the first version of a solid-state perovskite solar cell shown previously in Figure 1.3, the perovskite deposited on the TiO₂ surface is in the form of nano-dots that are not interconnected. Therefore, perovskite nano-dots are expected to behave in the same way as a molecular sensitizer, where photoexcited electrons are injected to TiO₂ and holes are injected into spiro-OMeTAD. Femtosecond transient absorption spectra (TAS) of different perovskite assemblies were measured (Figure 1.11) to elucidate the mechanism of charge separation.⁶  First, samples containing the perovskite material deposited on Al₂O₃ and TiO₂ without hole transporting layers were prepared to compare their electron transfer (Figure 1.11a and b). The negative signal at around 483 nm is ascribed to the bleaching of the absorption of the perovskite. The same bleaching is observed for both Al₂O₃ and TiO₂ samples. Since no electron transfer is expected between Al₂O₃ and the perovskite due to their large conduction band mismatch, the small difference in TA spectra observed between Al₂O₃ and TiO₂ indicates that no significant charge injection takes place from the excited state of the perovskite into TiO₂.

After introducing a spiro-OMeTAD hole transport medium (HTM) (Figure 1.11c and d), the bleaching peaks in the 480 nm-region are less pronounced than those for the perovskite without HTM, for both the Al₂O₃ and TiO₂ samples. The positive absorption signal in the 630–700 nm region is also attenuated for the TiO₂ sample and disappears for the Al₂O₃ one. These results suggest a rapid reductive quenching of the excited state of the perovskite by spiro-OMeTAD. From the TAS study, hole separation by the HTM is obvious, however, electron injection is not clear. Nevertheless, the photocurrent observed from the sensitization-type system provides strong evidence of electron injection under short-circuit conditions. Otherwise, charge collection at the FTO substrate could not be explained. Then, the absence of evidence of electron injection to TiO₂ in the TAS results is still questionable. As shown in Figure 1.3, the TiO₂ surface is not fully covered with perovskite, and the uncovered surfaces contacting directly with spiro-OMeTAD are likely to act as recombination sites. However, the solid-state sensitized perovskite solar cells deliver a remarkably high open circuit voltage of 888 mV, despite the bare TiO₂ surfaces, which indicates that photo-excited electrons may accumulate in perovskites without injection under open-circuit conditions.

The charge accumulation properties of CH₃NH₃PbI₃ have been studied using impedance spectroscopy. Figure 1.12 shows the J–V curves for CH₃NH₃PbI₃ perovskite solar cells with three different types of electrodes: FTO/Compact TiO₂ (Flat), FTO/Compact TiO₂/Nanostructured TiO₂ (NS·TiO₂) and FTO/Compact TiO₂/Nanostructured ZrO₂ (NS·ZrO₂), along with the capacitance as a function of the applied voltage.²⁵  All three samples demonstrate photovoltaic behavior with different performances. In Figure 1.12b, the capacitances of TiO₂ and ZrO₂ electrodes in a liquid redox electrolyte based on I⁻/I₃⁻ are compared as a function of the applied bias potential to investigate electrochemical charging effects. This experiment proves that the NS ZrO₂ layer cannot be charged, whereas the NS TiO₂ one is charged with the applied bias (V_app).

Figure 1.12c shows the capacitance for TiO₂ (Flat and NS) and ZrO₂ electrodes with and without perovskite. For the flat sample with perovskite at low V_app, the capacitance is almost invariant while, at high V_app, an increase in the capacitance slope is observed. This exponential enhancement of the capacitance can be attributed to the chemical capacitance of the flat TiO₂ compact layer, since the blank NS samples without perovskite show a similar tendency in capacitance change with V_app. In the case of complete NS devices with the perovskite, higher capacitances are clearly observed for the samples with perovskite than for the blank samples. It can be seen that the capacitances display a similar behavior regardless of the electrode material (TiO₂ or ZrO₂), the NS layer thickness and the illumination conditions. This indicates that the capacitance is originated in the perovskite deposited in the NS layer, showing direct evidence of charge accumulation in the perovskite material.

From the meso-superstructure concept with a scaffold oxide layer whose surface is coated with an extremely thin perovskite layer as reported elsewhere,²⁶  electron transport in CH₃NH₃PbI₃ is proved because Al₂O₃ cannot be charged by injection from the perovskite due to a conduction band mismatch, as can be seen in Figure 1.13. To deposit the perovskite on an Al₂O₃ mesoporous film, one mole of PbCl₂ and 3 moles of CH₃NH₃I are mixed in a polar aprotic solvent, instead of using an equimolar mixture of PbI₂ and CH₃NH₃I. From the stoichiometric point of view, the reaction of one mole of PbCl₂ with three moles of CH₃NH₃I yields CH₃NH₃PbI₃+2CH₃NH₃Cl. Since CH₃NH₃⁺ is a strong acid (because CH₃NH₂ is a weak base), CH₃NH₃⁺ tends to dissociate into CH₃NH₂ and H⁺. The strong acid HCl gives rise to the weak conjugate base Cl⁻. Thus, two moles of CH₃NH₃Cl are expected to form CH₃NH₂, H⁺ and Cl⁻, which are volatile when drying. Therefore, CH₃NH₃PbI₃ is eventually expected as the final product, along with the possibility of the presence of Cl as an impurity. Both the sensitized structure and meso-superstructure are beneficial for long-term stability because the moisture-sensitive perovskite is fully wrapped with the hydrophobic HTM.

In Figure 1.5, we saw diverse structures for perovskite solar cells. Among them, the mesoscopic and planar heterojunction structures have been widely and intensively studied because they show an abnormally high photovoltaic performance. The mesoporous TiO₂ layer is included in the mesoscopic structure by depositing it on a compact thin TiO₂-blocking layer, while the compact TiO₂ (or other oxide) layer alone is used as an electron-accepting layer in the planar heterojunction configuration. Spin-coating of the perovskite solution usually leads to formation of a capping layer on the mesoporous TiO₂ film. Either one-step or two-step sequential coating is available for the preparation of the perovskite layer. It has been found that the CH₃NH₃PbI₃ crystal size is easily controlled by the concentration of CH₃NH₃I in the two-step spin coating procedure.²⁷  The size-controllable two-step procedure is presented schematically in Figure 1.14. First, a solution of PbI₂ in N,N′-dimethyl formamide (DMF) is spin-coated on the mesoporous TiO₂ film, and this is followed by spin-coating of a CH₃NH₃I solution in isopropanol (IPA). Finally, a dark black perovskite film is formed after mild heating. In the second step, the crystal size can be controlled by varying the CH₃NH₃I concentration. As can be seen in the SEM images in Figure 1.14b, a relatively low concentration is found to increase the perovskite crystal size, but smaller sizes are obtained at high concentrations. Perovskite crystals about 700–800 nm in size are obtained from 38 mM [CH₃NH₃I], while the use of a 63 mM solution leads to ca. 100 nm-sized crystals. It is noteworthy that the size distribution becomes broader when bigger crystals grow and also, large crystals are not closely packed on the substrate.

Interestingly, the photovoltaic performance is influenced significantly by the concentration of CH₃NH₃I, which indicates that the perovskite crystal size plays an important role in determining the photovoltaic parameters. Figure 1.15a–d show that the photocurrent density increases as the crystal size increases. The fill factor exceeds 70% when the size is larger than 200 nm, and the open-circuit voltage is maximized for the intermediate size. As a result, an average PCE exceeding 16% is achieved for crystal sizes larger than 200 nm, whereas less than 14% is observed for smaller sizes of around 100 nm. The high photocurrent achieved with large crystal films can be explained by the light harvesting efficiency and photo-CELIV (charge extraction with linearly increasing voltage) data in Figure 1.15e and f, respectively. Overall, the light harvesting efficiency increases with the increasing crystal size, where the highest efficiency at long wavelengths observed for the 800 nm size is due to enhanced internal light scattering in the gaps between crystals. The large amount of extracted charges can also explain the high photocurrent density. Moreover, high hole mobility, along with a high charge extraction ability, can explain the highest voltage for the intermediate size, as confirmed by photo-CELIV in Figure 1.15f.

From the results of size-dependent photovoltaic performance, we wondered whether or not a further increase in size (achieved by further lowering the CH₃NH₃I concentration) would improve the photovoltaic performance. Sizes over 1 µm were obtained from a 32 mM CH₃NH₃I solution, as confirmed by the SEM images in Figure 1.16. However, contrary to our expectation, the corresponding PCE decreased to about 11.3%, due to a relatively low photocurrent density of 17 mA cm⁻² and open circuit voltage of 0.92 V. This is ∼10% lower than the voltages exceeding 1.1 V observed for the intermediate size of ∼200 nm. Taking into account the fact that V_oc=1.06 V for the 800 nm size crystals, one can conclude that an increase in crystal size over 200 nm is not beneficial to voltage gain. According to a micro photoluminescence (PL) study,²⁸  strong PL was observed from intermediate sized crystals (e.g. 200 nm), whereas large crystals of 800 nm and over 1 µm did not exhibit bright PL. The weak PL from large crystals suggests that non-radiative recombination is important, which is probably due to the presence of many defects and/or grain boundaries in large crystals. Such non-radiative recombination seems to be the responsible for low open-circuit voltages.

As illustrated in Figure 1.5, the mesoscopic structure has a pure perovskite layer in the form of a thin capping layer (around 100–200 nm) that is directly contacting with the HTM. Thus, holes generated in the perovskite layer travel according to the random walk model²⁹  and finally separate at the HTM/perovskite interface. One-dimensional (1D) nanostructured perovskites may meet the requirements for reducing the length of hole transport and improving charge separation. 1D nanowire perovskites can be fabricated by a two-step coating procedure, where the addition of a small amount of aprotic solvent to an isopropanol solution of CH₃NH₃I (MAI) is found to lead to nanowire formation.³⁰  Figure 1.17 shows the effect of adding aprotic solvents such as DMF to the isopropanol solution of CH₃NH₃I. Compared to the 3D nanocuboids formed in the absence of DMF, 1D nanowires are grown in the presence of DMF. 50 µL DMF in 5 mL isopropanol is the optimal formulation for the growth of long perovskite nanowires with diameters ranging from ∼30 to ∼200 nm. In terms of the mechanism involved in the anisotropic growth of MAPbI₃, the locally dissolved PbI₂, caused by the small amount of DMF present during the second spinning step, serves as the preferential site for reaction with MAI and growth of 1D structures, like in a liquid catalyst cluster mode.³⁰  This underlying mechanism suggests that the solubility of PbI₂ in the aprotic solvent can affect the nanowire formation. The effect of adding different aprotic solvents, such as dimethyl sulfoxide (DMSO) and gamma butyrolactone (GBL) to the MAI solution has been investigated. DMSO is able to produce nanowires but not perfectly, whereas GBL cannot induce nanowire formation due to the low solubility of PbI₂ in it. Nanowire growth correlates well with the order of PbI₂ solubility, DMF>DMSO≫GBL, and thus it can be said that the liquid catalyst cluster model is likely to be involved in nanowire growth.

As expected, the nanowire morphology was found to be better for hole separation at the HTM/perovskite interface, as can be seen in the time-resolved PL spectra in Figure 1.18. Approximately 79% of the holes are injected into the HTM from the perovskite nanowires with a lifetime of about 700 ps, whereas hole extraction in 3D bulk perovskites is not so efficient (ca. 68% of holes are injected into the HTM with a lifetime of 2.9 ns). This more efficient hole extraction by the nanowire perovskite is related to the increased surface area, which ensures a better contact with the HTM. Electron extraction of the nanowire perovskite deposited on a TiO₂ layer is not as efficient as for the bulk 3D perovskite, which is probably due to poor contact with the TiO₂ layer.

Nanorod oxides can be used alternatively to induce pseudo-one-dimensional perovskites. As nanorod oxides, TiO₂ and ZnO are the best candidates when considering matching the band positions with that of the perovskite. Since ZnO is better than TiO₂ in terms of electron transport kinetics^31,32  and fabrication process, ZnO nanorods are expected to be good candidates from the viewpoint of electron transport and modulation of the perovskite morphology. Figure 1.19 shows solution-processed ZnO nanorods on an FTO substrate and the corresponding perovskite solar cell. To grow ZnO nanorods, the FTO should be well covered with a ZnO seed layer. An ethanol solution of zinc acetate dihydrate (Zn(CH₃COO)₂·2H₂O) is used to prepare the seed layer. In this case, the coverage and thickness of the seed layer are important because the final morphology is influenced by these factors. The aqueous solution for growing ZnO nanorods is composed of equimolar zinc nitrate hexahydrate (Zn(NO₃)₂·6H₂O) and hexamethylenetetramine. The diameter of the ZnO nanorod can be controlled by varying the solution concentration. The length of the ZnO nanorods was found to depend on reaction time.

The photovoltaic performance of the ZnO nanorod perovskite solar cell is presented in Figure 1.20. The current density of 20.08 mA cm⁻² observed under solar simulator illumination is fully consistent with the integrated value of 20.03 mA cm⁻² calculated from external quantum efficiency (EQE) data. The relatively low PCE of 11.13% is mainly due to a relatively low fill factor of 0.56 and voltage of 0.99 V. It is interesting to see that the EQE spectral shape is almost squared, as shown in Figure 1.20b, which is related to the perovskite being embedded in the ZnO nanorod film with a negligible capping layer.

As mentioned previously, the coverage of the seed layer on the FTO surface is important for the subsequent growth of ZnO nanorods. The nature of the seed layer can be influenced by the coating solution. The effect of the seed layer on the growth of ZnO nanorods and the resulting photovoltaic performance has been investigated.³⁴  Three different seed layers were prepared using a clear zinc acetate solution (designated as ‘solution’ in Figure 1.21), a ZnO colloidal solution (designated as ‘colloidal’) and ZnO powder dispersed in ethanol (designated as ‘powder’). Coverage of the seed layer is much better when using the colloidal solution compared to the coverage obtained from the other two solutions, as shown in Figure 1.21a–c. The difference in coverage is found to influence significantly the ZnO nanorod morphology. The seed layer formed from the colloidal precursor results in vertically grown ZnO nanorods, whereas some or most of the nanorods are tilted when grown on a seed layer formed from the solution or powder precursor (Figure 1.21d–f). Focused ion beam (FIB) assisted cross-sectional SEM images for the full cell structure reveal that the ZnO nanorods are well attached, while some of the ZnO nanorods are detached from the substrate for both the solution- and powder-based seed layers, resulting in some of the FTO surfaces being exposed directly to the perovskite layer (Figure 1.21g–i). The tilted nanorods are expected to be easily detached due to the reduction in anchoring surface. It is also noted that some of the detached ZnO nanorods are directly contacted with the HTM, acting potentially as sites for recombination.

The colloidal-based ZnO nanorod cell shows a higher voltage and fill factor compared to the solution- and powder-based ones (Figure 1.22a). The values of series resistance (R_s) and shunt resistance (R_SH) estimated from the current–voltage curves explain the differences in the fill factor. It is interesting to note that a similar short-circuit current density and EQE (also known as incident photon to current efficiency, IPCE) are observed (Figure 1.22b), which indicates that the photocurrent is hardly altered by the ZnO nanorod morphology. The significant difference in voltage is related to the recombination resistance (R_rec), associated with the interface and morphology of the ZnO nanorods. At low bias potential, R_rec (which is related to the interface or bulk of the perovskite) increases in the order of colloidal >solution>powder (Figure 1.22c). A higher recombination resistance leads to a higher V_oc. At high bias potential, R_rec is probably related to the ZnO nanorod/perovskite interface or perovskite/HTM interface. It decreases in the order of colloidal>solution>powder, which suggests that suppression of recombination at the ZnO nanorod/perovskite interface is greatest for the colloidal-based seed layer. The resistance of the selective contacts can be obtained from the high frequency resistance (R_hf) in Figure 1.22d. In the bias potential range between 0.2 and 0.7 V, the higher R_hf observed for the colloidal-based ZnO nanorod cell compared to those of the solution- and powder-based cells is due to the full coverage on FTO, which is also responsible for the high voltage.

In the case of mesoscopic perovskite solar cells comprising mesoporous TiO₂ films filled with perovskite, a contentious issue is whether or not photo-generated electrons are injected from the perovskite into TiO₂. To answer this question, other n-type oxides having similar conduction band position are required for comparison with the anatase form of TiO₂. The rutile form of TiO₂ was selected initially because its conduction band position is similar to that of anatase. In order to exclude unwanted factors, the rutile TiO₂ film must be carefully controlled in order to have almost the same pore size and layer thickness. Mesoporous rutile TiO₂ films with a pore size of about 40 nm have been prepared in order to compare them with anatase films with pore size of about 43 nm.³⁵  Figure 1.23a shows that the rutile-based cells reproducibly exhibit slightly higher photocurrent densities but lower voltages than anatase-based cells. Regarding the different photovoltaic performance, a different extent of electron injection can be assumed. The electron diffusion coefficient and the time constant for electron recombination are useful tools to investigate electron transport and life time. Compared to the anatase-based perovskite solar cell, the rutile-based one exhibits an electron diffusion coefficient that is almost one order of magnitude lower, and a time constant for electron recombination that is ten times higher. The difference in diffusion coefficient and recombination kinetics are indicative of electron injection, otherwise the electron diffusion coefficients in the two cases would be expected to be similar regardless of the crystal phase, since electron transport occurs only in the perovskite layer. The slower electron transport kinetic in the rutile-based perovskite solar cell is related to the poor interconnectivity associated with long rice-shaped rutile particles. The higher time constant for recombination indicates slower recombination kinetics. The much slower recombination in the rutile-based perovskite solar cell is a consequence of more injected electrons in rutile TiO₂, as recombination is accelerated in the case of ineffective charge separation at the oxide/perovskite interface. Moreover, the lower V_oc for the rutile-based device is also related to the extent of electron injection. When considering that the conduction band edge of TiO₂ is lower than that of the perovskite, an increase in the amount of the injected electrons will lower the Fermi level of the perovskite, and thereby decrease the voltage. According to transient photocurrent–voltage measurements along with the observed current–voltage characteristics, electron injection from the perovskite to the TiO₂ layer occurs partially and, evidently, the rutile layer accepts relatively more photo-excited electrons than the anatase layer. One can conclude that a dual pathway for electron transport seems to be present in the mesoscopic structure. The extent of electron injection is schematically compared between anatse and rutile in Figure 1.23.

CH₃NH₃PbI₃ is a key component in current perovskite solar cells. However, the material needs to be engineered further to achieve a higher absorbance and lower band gap. The valence band of CH₃NH₃PbI₃ is related to the filled p orbitals of the X⁻ ions, and the conduction band is associated with the empty p orbitals of Pb²⁺ ions.³⁶  It is also mentioned in ref. 35 that the organic cation cannot influence the band position directly, but it can indirectly affect the band gap. Thus, it is expected that band gap changes without significant tuning of the band position will occur upon replacing methylammonium with other organic cations. Replacement of methylammonium by formamidinium (HC(NH₂)₂) results in a decrease in the band gap from 1.55 eV to 1.5 eV.³⁷  The corresponding red shift of the onset wavelength by around 30–40 nm is observed in the UV-Vis absorbance spectra going from the CH₃NH₃PbI₃ perovskite to the HC(NH₂)₂PbI₃ one (Figure 1.24), without altering the absorption coefficient.

A perovskite solar cell based on two-step deposited HC(NH₂)₂PbI₃ in a mesoscopic structure demonstrated a PCE of 16% in 2014.³⁷  The mesoporous TiO₂ layer has been found to play an important role in the photovoltaic performance of formamidinium perovskite cells, and the influence of the mesoporous TiO₂ film on charge collection kinetics has been investigated.³⁷  Figure 1.25a displays cross-sectional SEM images showing the mesoporous TiO₂ layer and the perovskite capping layer, showing that the pores are completely filled with perovskite. The layer structures are illustrated schematically in Figure 1.25b. The kinetics of charge collection is influenced significantly by the presence or absence of the mesoporous TiO₂ film, as well as by the thickness of the TiO₂ film (Figure 1.25c). The time-dependent photocurrent signal shown in Figure 1.25c confirms that the signal amplitude increases with the increasing TiO₂ film thickness, along with fast collection kinetics. This underlines the fact that the mesoporous TiO₂ film plays an important role in facilitating photocurrent collection. Moreover, the fill factor is improved after the introduction of mesoporous TiO₂ and the increase of its thickness. The inclusion of mesoporous TiO₂ results in a decrease in the series resistance and an increase in the shunt resistance, which is responsible for the improved fill factor.

In order to investigate the electron injection behavior at the TiO₂/HC(NH₂)₂PbI₃ interface, a nano-helical TiO₂ layer was introduced. Helical TiO₂ arrays with different pitch (p) and radius (r), helix-1 (p/2=118 nm, r=42 nm), helix-2 (p/2=353 nm, r=88 nm) and helix-3 (p/2=468 nm, r=122 nm), were grown on FTO glass by oblique-angle electron beam evaporation, as confirmed by the SEM images in Figure 1.26. It is essential to rotate the substrate at controlled time intervals and high substrate-tilt angles. HC(NH₂)₂PbI₃ perovskite was deposited in the helical TiO₂ films by the two-step technique. The dependence of the photovoltaic performance and electron transport behavior on the helix morphology was investigated.³⁸ 

The cell fabricated using helix-1 showed a higher short-circuit current density, while the cell using helix-3 exhibited a slightly higher open-circuit voltage (Figure 1.27). The higher photocurrent from the helix-1 cell is related to the enhanced light scattering efficiency and higher absorbed photon-to-current conversion efficiency. The time-dependent photocurrent response also depends on the helical TiO₂ nanostructure used. At 4 Hz and 10 Hz, the helix-1 cell shows the fastest photocurrent response, while the helix-2 cell shows the slowest response, which implies that charge collection is less efficient for the helix-2 cell compared to the others. At an increased chopping rate of 50 Hz, the photocurrent response is less intense, and a fast component can be identified. This fast component is probably due to charge transport through the perovskite layer itself, given the three orders of magnitude higher diffusion coefficient of electrons in the perovskite compared to that in TiO₂. The slower component seems to be related to charge transport in helical TiO₂. The amplitude of the fast component increases in order helix-2<helix-3<helix-1, which is in reverse order of the surface area. With a larger surface area, a larger fraction of electrons generated in the perovskite appear to be injected into the TiO₂ network, so that the fraction of the electrons transported through the perovskite layer decreases. This result again proves partial electron injection from the HC(NH₂)₂PbI₃ perovskite to TiO₂, and a strong probability of injection in the vicinity of the interface.

The emergence and progress of perovskite solar cells have been described briefly. Thanks to the important pioneering works on perovskite-sensitized solar cells reported in 2009, 2011 and 2012, a surge of reports on perovskite solar cells has followed. There is no doubt that these pioneering works should be appreciated. Perovskite materials can be applied to any kind of junction structure, although mesoscopic solar cells containing mesoporous or nanostructured oxide layers have been the most intensively reviewed in this chapter. The basic fundamentals of CH₃NH₃PbI₃ have also been considered, and it is clear that careful attention should be paid to the molecular motion of the organic ion because of the underlying correlation between molecular motion and photovoltaic properties. Mesoporous TiO₂ was found to play an important role in electron transport and the resistance components in the devices. Electron injection was evident at the junction between nanocrystalline TiO₂ and the perovskite, which could be isolated by dual function fit of the data obtained from transient photocurrent measurements. In order to achieve even higher efficiencies with mesoscopic perovskite solar cells, defect-free high quality perovskite grains without non-radiative recombination are essential.

This work was supported by National Research Foundation of Korea (NRF) grants funded by the Ministry of Science, ICT & Future Planning (MSIP) of Korea under contracts No. NRF-2012M3A6A7054861 (Nano Material Technology Development Program) and NRF-2012M3A6A7054861 (Global Frontier R&D Program on Center for Multiscale Energy System).