Introduction

Generating cost-effective and environmentally benign renewable energy remains a major challenge for scientific development. Solar energy, which is abundant and sustainable, has attracted enormous interest in terms of research and development for many years. The developments in emerging photovoltaic techniques are taking place at a rapid pace. Most such devices involve inexpensive solution processing with high device performance.1, 2, 3 Solar cells (SCs) of these new types differ from traditional SCs in their unique mesoscopic structural features, with large surface areas that dominate their photovoltaic performance.4, 5, 6 For example, a dye-sensitized SC (DSSC) uses a monolayer of light-absorbing sensitizer anchored on nanocrystalline TiO2 to enhance light harvesting in a mesoporous environment in which charge separation also occurs at this interface. The design of a DSSC involves an important feature: the light harvesting and charge transport are spatially decoupled.7 Since the early success of using nanocrystalline TiO2 as a charge collector, DSSCs have been considered a low-cost alternative to semiconductor-based photovoltaic devices.8 In 2011, Yeh, Diau and Grätzel developed an efficient DSSC based on an ortho-substituted porphyrin sensitizer, YD2-oC8, co-sensitized with an organic dye (Y123) with a cobalt electrolyte to attain remarkable power conversion efficiency (PCE=12.3%),9 which opened a new research area on porphyrin-sensitized SCs (PSSCs).10 However, the absorption spectrum of the YD2-oC8 system covers only the visible spectral region, 400–700 nm, and the lack of light-harvesting ability beyond 700 nm (near infrared) limits the PSSC from further enhancement of the device performance. Moreover, the enduring stability of PSSCs might be an issue for future commercialization. Various inorganic light absorbers with a light-harvesting capability that extends into the near-infrared region were hence sought to replace the delicate organic dyes.

Inorganic semiconductor quantum dots (QDs) have been extensively investigated as new photosensitizers or light absorbers to replace conventional organic-type sensitizers in DSSCs because of their excellent properties—tunable energy band gaps via varied and controlled size and shape of the nanocrystals, large optical absorption coefficients, large dipole moments for enhanced charge separation, and multiple exciton generation (MEG).11, 12, 13, 14, 15, 16, 17, 18 For example, the performance of QD-sensitized SCs (QDSCs) has been significantly enhanced for Sb2S3-sensitized TiO2 solar cells using a blended layer of poly[2,6-(4,4-bis(2-ethylhexyl)-4H-cyclopenta[2,1-b;3,4-b′]-dithiophene)-alt-4,7-(2,1,3-benzothiadiazole)] (PCPDTBT) and [6,6]-phenyl C61-butyric acid methyl ester (PCBM) as a hole-transporting material (HTM) to attain η=6.3%.15 Hybrid passivated QDSCs using PbS colloidal QDs (CQDs) as a photosensitizer attained an efficiency of more than 7.0%,13 and a large short-circuit photocurrent density (JSC30 mA cm−2) was reported for a PbS:Hg QDSC.18 In general, using an inorganic light-harvesting material to replace a conventional organic or organometallic sensitizer has the advantage of obtaining superior device durability.

Typical semiconductor QDs have an average diameter of 1–10 nm, which indicates that surface states or midgap trap states can be produced on the surfaces of the nanocrystals during the synthesis of the QDs via layer deposition on the photoanode of the solar cells. For mesoporous TiO2/QD solar cells, nanocrystalline QDs can be adsorbed and deposited on the surface of the TiO2 nanoparticles, but the QD coverage on the surface of the TiO2 nanoparticles may be a problem. As a result, these QDSC devices typically show a small open-circuit voltage (VOC) and a poor fill factor (FF) because of the charge loss at the interface of the TiO2/QD/HTM layers. To increase VOC and FF, we must decrease the charge loss in the QD layer and suppress the charge recombination in the TiO2/QD and TiO2/HTM interfaces. For this purpose, an insulating Al2O3 scaffold, surface passivation of QD films and interfacial engineering of buffer layers with organic and inorganic modifications have been employed to enhance the charge separation and charge transport in solar cells.2, 13, 17

A major advance in the development of novel inorganic sensitizers occurred in 2012: Park and Grätzel reported submicrometer thin-film solid-state solar cells attaining a PCE of 9.7% with a methyl ammonium lead-iodide perovskite sensitizer, CH3NH3PbI3, in a mesoporous TiO2 film (thickness 0.6 μm).12 This perovskite sensitizer has a band gap of 1.5 eV with the energy levels of the conduction band (CB) and the valence band (VB) matching well the CB of TiO2 and the highest occupied molecular orbital of HTM, respectively. At nearly the same time, Snaith and co-workers11 reported a similar perovskite, CH3NH3PbI2Cl, that served as a light absorber for mesoscopic thin-film solid-state solar cells to attain a PCE of 10.9%, for which the mesoporous Al2O3 film served as a scaffold to replace the n-type TiO2 electron-transporting layer. In 2013, Snaith and co-workers reported a significantly enhanced PCE of 12.3% for perovskite CH3NH3PbI3-xClx solar cells with the same device structure based on Al2O3.2 Concurrently, Seok and co-workers3 reported the same superior PCE of 12.3% for another perovskite solar cell, CH3NH3Pb(I1-xBrx)3, with the mesoporous TiO2-based device structure. Also in 2013, Grätzel and co-workers1 reported perovskite (CH3NH3PbI3) TiO2 solar cells with a PCE approaching 15%, which set a new record for all-solid-state hybrid mesoscopic solar cells (HMSCs). The significant discovery of perovskites as novel photovoltaic materials has hence opened a new channel for the development of third-generation solar cells with the advantages of very high efficiency, low cost, ease of processing and considerable durability.

In this review, we summarize the recent investigations into novel inorganic materials, including both metal chalcogenide QDs and halide perovskite nanocrystals, applied as light-harvesting layers for HMSCs. Figures 1a and b are diagrams of the potential levels of various metal chalcogenide QDs and halide perovskites, respectively. Metal chalcogenide sensitizers of three types were introduced for QDSCs: they are classified as cadmium (Cd) chalcogenides, lead (Pb) chalcogenides and antimony (Sb) chalcogenides. The band gap energies (Eg/eV) of the metal chalcogenides decreased systematically as the chalcogenides varied from sulfur (S) through selenium (Se) to tellurium (Te). For the halide perovskite materials, tuning the band gap of CH3NH3PbX3 was achieved by varying the halide constituents (X=Cl, Br or I); the Eg of the system decreased along this direction. This review provides detailed descriptions of the synthesis, characterization, photovoltaic performance and challenges for two inorganic materials, Sb2S3 QDs and perovskite nanocrystals, because of their efficient performance as light-harvesting materials for HMSCs. We also introduce strategies for the improvement of the performances of these QD devices and perovskite materials via surface passivation and interfacial modification with varied device structures.

Figure 1
figure 1

Energy levels vs vacuum of various (a) metal-chalcogenide quantum dots (QDs) and (b) organometal-halide perovskites as prospective light absorbers for hybrid mesoscopic solar cells. The particle sizes are shown in parentheses with the related references shown as superscripts for TiO2 (bulk),18 CdS (5 nm),88 CdSe (3 nm),33 CdTe (3.8 nm),33 PbS (2.4 nm),18 PbSe (4.5 nm),48 Sb2S3 (5 nm),86 Sb2Se3 (5 nm),99 CH3CH2NH3PbI3 (1.8 nm),102 CH3NH3PbBr3 (7 nm),106 CH3NH3PbI2Br (10 nm),107 CH3NH3PbI3 (2.5 nm)12 and CsSnI3 (bulk)112; HTM represents spiro-OMeTAD.12

Cadmium chalcogenide QDSCs

Cadmium-chalcogenide QD-based solar cells have been intensively studied over the past two decades to improve device performance and stability in ambient conditions. The photoactive sensitizers of these solar cells are composed of cadmium sulfide (CdS),19, 20, 21, 22, 23, 24, 25 cadmium selenide (CdSe),26, 27, 28, 29, 30, 31, 32, 33 cadmium telluride (CdTe)34, 35, 36, 37, 38 and their alloy nanocrystals.39, 40, 41, 42, 43, 44, 45, 46 CdS has an optical band gap of 2.25 eV, and thus it can only absorb light up to 550 nm; CdSe, with a band gap of 1.7 eV, can absorb light below 720 nm; CdTe has an energy gap of 1.45 eV and an optical absorption edge at 860 nm. The CB edges of CdSe and some CdTe nanocrystals were located below that of the TiO2 film (Figure 2), which resulted in poor and limited electron injection from the QD to the TiO2, and substantial charge recombination occurred at the TiO2/QD/HTM interfaces.34, 39 Moreover, the use of iodide and polysulfide electrolytes led to the degradation of these QD nanocrystals and a gradual deterioration of the photovoltaic performances.20, 27, 28, 34

Figure 2
figure 2

(a) Relative band edges of TiO2, CdS and CdSe in bulk and (b) the proposed band edge structure for a TiO2/CdS/CdSe electrode after redistribution of the electrons between the CdS and CdSe interface (described as alignment of the Fermi level of bulk materials) (adapted from Lee and Lo,39 copyright 2009, Wiley-VCH Verlag GmbH & Co.).

To find a solution for the above problem, co-sensitization of CdS/CdSe QDs on a TiO2 film was reported.39 The co-sensitized CdS/CdSe system showed improved light absorption, incident photon-to-current conversion efficiency (IPCE), and PCE over individual CdS or CdSe systems because more CdSe QDs were loaded onto the TiO2 film. The rate of electron injection from CdSe to TiO2 was significantly improved via the CdS layer and adjustment of the Fermi level, as shown in Figure 2.39 ZnS passivation served to protect the QD sensitizers from the polysulfide electrolyte and to inhibit the charge recombination at the TiO2/electrolyte interface. For devices made of CdSexTe1−x alloy nanocrystals, the range of optical absorption could be extended to 900 nm, which reflects the band gap tuning effect of the CdSe–CdTe alloys, especially to enhance light harvesting in the range of 750–900 nm. Figure 3 shows the procedure used to synthesize the CdSe–CdTe QD nanocrystals by means of the layer-by-layer deposition approach, for which the optical band gaps of the alloy were systematically varied from 1.38 to 1.73 eV.40 Using core–shell nanocrystals, recombination losses at the TiO2/CdTe/CdSe interface can be effectively diminished through a cascade potential approach that yields a superior charge separation, as shown in Figure 4. The most striking feature of this CdTe/CdSe core–shell system, with a particle size 10 nm, is its smaller band gap for light absorption toward the near-infrared region (1000 nm).41 Instead of polysulfide electrolytes, quasi-solid-state polysulfide gel electrolytes, polymer electrolytes and solid-state HTM have been developed and examined for this type of QDSC, but the device efficiencies were reported to be less than 5%.20, 27, 28, 42, 43

Figure 3
figure 3

Schematic outline of the layer-by-layer fabrication process for CdSexTe1-x nanocrystal solar cells. (a) CdSe and CdTe nanocrystals dispersed in pyridine/1-propanol are mixed at the desired ratio. (b) A thin film of CdSex:CdTe1-x is deposited by spin coating onto the ITO substrate. (c) The thin film is subjected to treatment with CdCl2 followed by thermal annealing to promote crystal growth. This process is repeated as necessary to yield the desired device composition and active layer thickness. (d) Schematic image of a completed device, which includes a layer of nanocrystalline ZnO and an evaporated aluminum top contact. (e) Absorption spectra of CdSex:CdTe1-x solutions (adapted from MacDonald et al.,40 copyright 2012, American Chemical Society).

Figure 4
figure 4

Schematic representation of a system consisting of a CdTe/CdSe type-II core/shell QD adsorbed on nanocrystalline TiO2 and coated with a ZnS shell. An assumed band diagram of the system is also shown. Arrows show the charge directions after exciton generation (adapted from Itzhakov et al.,41 copyright 2013, American Chemical Society).

The solid-state solar cells based on CdS, CdSe and CdTe QDs have shown PCEs less than 6% because of their large band gaps, slow electron injection rates and substantial charge recombination at the TiO2/QD/HTM interfaces. Thin-film solar cells made of CdSexTe1−x alloys (x=0.9 or 0.8) exhibited a PCE (7.1%) smaller than that of a CdTe-only device (PCE=7.3% at x=0) because JSC of the former is smaller than that of the latter, indicating that the electron transport and charge collection through the alloy layers encountered some difficulties in the devices. In summary, the ideal cadmium chalcogenide QDSC devices should have a large recombination resistance, a small chemical capacitance and a small series resistance due to a decreased rate of recombination and fewer trap sites at the TiO2/QD interface. Furthermore, the device performance of the CdTe/CdSe QDSCs can be further improved by increasing the QD loading on the TiO2 film to enhance the efficiency of light harvesting.

Lead chalcogenide QDSCs

Lead-chalcogenide QD-based solar cells have been intensively examined in recent years because of the small band gaps of the QDs, which allow solar energy harvesting in the near-infrared region. For example, PbS QDs have an energy band gap in the range of 0.9–1.1 eV, and the optical absorption edge can be extended to 1300 nm;13, 18 the band gap energies of PbSe QDs were tuned from 0.7 to 1.7 eV by varying the sizes of the QDs, and the optical absorption edge was further extended to 1500 nm.17, 47, 48 Under light illumination, the shunt resistance of the QDSCs was smaller than in the dark; the rates of charge recombination increased at the interfaces of QD–QD and TiO2–QD because of the increased surface states and trap states. For the PbS CQD films, a high density of midgap trap states (1017 cm−3 eV−1) may deteriorate the device performance (Figure 5), which could be a crucial limiting factor in improving the device efficiency.13 PbSe QD solar cells suffered from significant current leakage and interfacial charge recombination; the film thickness of the photoactive layer was expected to be an important parameter in optimizing the PbSe device performance.48

Figure 5
figure 5

Performance of CQD photovoltaics as a function of passivation. (a) Schematic of the depleted heterojunction CQD device used in this work. Inset: photograph of a typical device (substrate dimensions, 25 mm × 25 mm). (b) Cross-sectional s.e.m. image of the same device. Scale bar, 500 nm. (c) Measured current–voltage characteristics under AM1.5-simulated solar illumination for representative devices employing organic (red), inorganic (green) and hybrid (blue) passivation schemes. Black diamonds denote the JV curve for a hybrid passivated device as measured in an accredited photovoltaic calibration laboratory (Newport Technology and Application Center-PV Lab). Inset: EQE curve of a hybrid passivated device. The integrated current value is also shown. (d) Simulated JV curves of devices with varying midgap trap densities, demonstrating the detrimental effect of traps (adapted from Ip et al.13 copyright 2012, Macmillan Publishers Limited).

For PbS devices treated with CdCl2, the trap states near the middle of the PbS band gap were determined to be 1016 cm−3 eV−1, which is one-fifth of that in organic cross-linked and inorganic-treated PbS films.13 The PbS film treated with CdCl2 showed the greatest hole mobility, μp=4.2 × 10−3 cm2 V−1 s−1, and CdCl2 plays a major role in the passivation of the midgap trap states and charge transport in the valence band. Greater photocurrent and voltage are achieved through decreased recombination and improved charge transport in the PbS device. Using PbS:Hg nanocrystals, the short-circuit photocurrent density of the device attained 30 mA cm−2.18 The CB of the PbS QDs shifted upward with Hg2+ doping, and more rapid electron injection became feasible in the PbS:Hg QDSCs, as shown in Figure 6. The structural reinforcement of the Pb–S bonds and the stability provided by the addition of Hg2+ were two major factors for the decreased charge recombination and enhanced charge transport in the PbS:Hg device. For the hydrazine-treated PbSe QDs, the external quantum efficiency (EQE) and internal quantum efficiency (IQE) were 114% and 130%, respectively, proving the MEG effect in the PbSe device shown in Figure 7.17 PbSe QDs were treated with ethane dithiol and hydrazine to increase the electronic coupling of the QD nanocrystals and decrease the surface states and trap states in the PbSe film. Approximately 1 mA cm−2, or 4% of the total photocurrent, was estimated to derive from the MEG effect in the PbSe device.17

Figure 6
figure 6

Morphology and energetic properties of PbS and PbS:Hg QDs. (a) Deposition of a PbS:Hg QD on a mesoporous TiO2 nanostructure using successive ionic layer adsorption and reaction (SILAR). (b) TEM micrograph of bare TiO2 and QD-coated TiO2 nanoparticles using a cationic solution containing 0, 2, 4, 6 and 8 mmol of HgCl2. (c) Tauc plot calculated using the Kubelka-Munk equation from the reflectance spectra. (d) UPS spectra of PbS:Hg QD-adsorbed TiO2 and extrapolation of the region of small binding energy. (e) Band edge alignment diagram of PbS and PbS:Hg QD (adapted from Lee et al.,18 copyright 2013, Scientific Reports).

Figure 7
figure 7

(a) EQE signals for 18 independent devices made with QD band gaps of 0.71 eV (yellow), 0.72 eV (blue) and 0.73 eV (red) and for a device with an antireflective coating (black). (b) Collected IQE curves vs the ratio of photon energy to band gap, hυ/Eg, for the three QD sizes. The dashed curve is a published fit for colloidal QD, whereas here it has been normalized for intrinsic loss in the cell due to recombination. (c) Maximum IQE values for seven QD sizes. The peak IQE values have been corrected for the intrinsic loss in the solar cell (estimated at 15%). Error bars indicate the propagated uncertainty of 5 to 30 measurements at the given wavelength of both the reference detector and the test solar cell. The solid black curve is the original fit to colloidal QD; the dashed curve is the same as in (b). Hollow triangles and squares represent ultrarapid transient absorption measurements of PbSe QD solutions (adapted from Semonin et al.,17 copyright 2011, American Association for the Advancement of Science).

In the core–shell structure, the PbSe core was covered with a PbS layer (thickness 0.5 nm). PbSe/PbS core–shell QDs have an absorption edge at 960 nm (Eg 1.29 eV).49 The PbS shell covering the PbSe core acted as surface passivation to protect the PbSe core, thus improving the chemical stability of the QD film. The IPCE reached 100% at 420 nm, indicating that the core–shell QDs on the TiO2 surface had efficient light harvesting, enhanced electron injection from the QD to the TiO2 CB, superior charge transport in the TiO2 layer and retarded charge recombination in the QD film. For PbS0.9Se0.1 alloy QDSCs, an EQE above 100% at a photon energy of 2.76 eV (440 nm) was obtained.50 PbS0.9Se0.1 alloy QDs showed a sharp absorption peak at 1076 nm due to strong quantum confinement. PbS0.9Se0.1 alloy QD films were treated with ethane dithiol to increase electronic coupling in the PbS0.9Se0.1 QDs. Greater EQE and JSC values were obtained upon increasing the film thickness to 360 nm because of the increased optical absorption and higher charge generation in the device. Other PbS(Se) QD solar cells have been reported with efforts to improve the charge transport and retard the charge recombination in the devices.51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83

Antimony chalcogenide QDSCs

Synthesis of Sb2S3 QDs on a mesoporous TiO2 film

Antimony sulfide Sb2S3 QDs have been synthesized on a mesoporous TiO2 film by means of the chemical bath deposition (CBD) method.14, 15, 84 Messina et al. reported a CBD procedure to form an Sb2S3 thin film coating on microscope glass slides.84 SbCl3 (1.3 g) was dissolved in acetone (5 ml); this SbCl3 solution was kept at a temperature below 4 °C for several hours. A Na2S2O3 solution (50 ml, 1 M) and deionized water (145 ml) were also kept at a temperature below 4 °C for several hours before the solutions were mixed. The cold SbCl3 and Na2S2O3 solutions were mixed with continuous stirring, and the cold deionized water was added to the mixed solution. This final mixing initially gave a nearly clear solution. The substrates containing the TiO2 films were placed vertically in the CBD bath for 2–3 h in a refrigerator (4–7 °C). The color of the Sb2S3-coated TiO2 film changed from yellow to orange-yellow and finally to brown during the CBD. Thereafter, the TiO2 films coated with Sb2S3 were carefully washed with deionized water and dried at around 23 °C for several hours in air.

As the CBD-processed Sb2S3 exhibited an amorphous phase, the Sb2S3-coated TiO2 films were annealed at 300–330 °C for 30 min under Ar (or N2) inert conditions to form a crystalline stibnite structure. Itzhaik et al.85 reported that this annealing method, which forms an Sb2O3 surface passivation layer on the surface of the Sb2S3 sensitizer, improved the device performance, resulting in a PCE of 3.4%. This passivation layer on the Sb2S3 surface can inhibit the electron-hole recombination through its insulating nature (Eg 3.7 eV), but an Sb2O3 layer that was thick enough to act as an energy barrier, disturbing the electron injection from the Sb2S3 sensitizer to the TiO2 film, thus degrading the device performance significantly.

Electronic properties of Sb2S3 sensitized on TiO2 films

Patrick and Giustino86 reported on the structural and electronic properties of Sb2S3-sensitized QDSCs and found an almost perfect interface between the TiO2 film and the QD sensitizer because of the lattice match of the TiO2 substrate with the Sb2S3 layer. As Figure 8 shows, no significant structural defects in the TiO2/Sb2S3 interface were found; the electronic distributions corresponding to the CB of Sb2S3 show a direct interfacial Ti–S coupling between Sb2S3 and TiO2. The CB edge (ECB) of the Sb2S3 sensitizer is located 0.5 eV above the ECB of the TiO2 film (Figure 1a); superior electron injection from the CB of Sb2S3 to the CB of TiO2 would hence be expected. For the TiO2/Sb2S3 solar cells, the theoretical value of VOC can be predicted to be as high as 1.6 V, based on the calculated energy-level diagram; this value indicates that VOC might be increased experimentally by optimizing the other parameters related to hole-transporting materials.

Figure 8
figure 8

(a) Atomistic model of the TiO2/Sb2S3 interface derived in this work. The colored atoms represent the periodic repeat unit; the view is along the TiO2 [010] direction. The lengths of the Ti-S and O-Sb bonds are 2.67 and 2.81 Å, respectively. Inset: Schematic representation of the stibnite-sensitized solar cell of Chang et al., with the TiO2/Sb2S3 interface highlighted. (b) Isodensity plot of the Kohn-Sham LUMO state of Sb2S3 at the TiO2/Sb2S3 interface. The charge density is plotted in a plane through the Ti-S bond. The coupling between the S-3p states of the sensitizer and the Ti-3d states of the substrate, which provides a path for electron injection, is highlighted (adapted fromPatrick and Giustino,86 copyright 2011, WILEY-VCH Verlag GmbH & Co., Weinheim Germany).

Optical absorption and IPCE of the Sb2S3-based solar cells

Crystalline Sb2S3 has an optical absorption coefficient α 105 cm−1 in the visible region. Stibnite Sb2S3 nanocrystals deposited on a mesoporous TiO2 film have a particle size of 5–10 nm.14, 15 The TiO2/Sb2S3 system shows an excellent match of energy levels for electron injection and transport after light harvesting by the Sb2S3 sensitizer. The corresponding TiO2/Sb2S3 solar cell (SC) showed IPCE values between 70 and 90% in the visible region with various hole-transporting materials, such as CuSCN, spiro-OMeTAD and poly(3-hexylthiophene-2,5-diyl) (P3HT).14, 15, 87, 88, 89, 90, 91, 92, 93, 94 The IPCE curves feature an optical absorption onset at 750 nm, which is consistent with the crystalline Sb2S3 band gap of 1.65 eV. Seok and co-workers reported an IPCE loss15 in the wavelength range of 450–650 nm, as P3HT can absorb light in this wavelength region, but the charges generated by P3HT were incompletely transferred to the Au counter electrode (CE) in the TiO2/Sb2S3/P3HT device. This IPCE loss was fully recovered by adding PCBM for additional conducting channels, as shown in Figure 9.

Figure 9
figure 9

(a) EQE spectra: the region marked with blue lines is the EQE difference between the T/S/P-P and T/S/P samples. (b) J-V curves at 100 mW·cm−2 light illumination (photo) and no illumination (dark). (c) Transmission spectra: the inset shows the transmission difference between the T/S/P-P and T/S/P samples. (d) PL spectra. The following abbreviations appear in the figure: T=mp-TiO2, S=Sb2S3, P=P3HT and P-P=P3HT/PCBM (adapted from Chang et al.,15 copyright 2012, American Chemical Society).

Photovoltaic properties of the Sb2S3-based solar cells

Etgar and co-workers fabricated Sb2S3/P3HT solar cells with a 1-decyl phosphonic acid (DPA) surface treatment to improve the device performance.94 Such a device attained a PCE of 3.9%, which was significantly superior to that of a device without this surface treatment (η=3.1%). The enhanced performance was due to the smaller dark current and greater electron lifetime after the DPA surface treatment. For Sb2S3/CuSCN solar cells, oxygen plays an important role during the fabrication of solar cells: oxygen can decrease the series resistance of the device to enhance the overall device performance.87 It was also reported that the formation of the Sb2O3 surface passivation layer helps to promote the device performance.85 For the Sb2S3/spiro-OMeTAD solar cells, JSC increased nonlinearly with increasing light intensity.92 This condition might indicate that the hole diffusion from spiro-OMeTAD to the Au electrode is a bottleneck responsible for the nonlinear behavior of the photocurrents with increasing light intensity. The charge losses at higher light intensities can be decreased by improving the Sb2S3/Spiro-OMeTAD interface and using new HTM with a greater hole mobility. The highest efficiency of Sb2S3-based solar cells was reported to be 6.3% using PCPDTBT as an HTM layer, giving the device a remarkable photovoltaic performance, with JSC=16.0 mA cm−2, VOC=0.595 V and FF=0.655.15 Ito and co-workers89 reported TiO2/Sb2S3/CuSCN solar cells with a TiO2 surface treatment (η=4.1%). The TiO2 devices in which the surface was treated with BaTiO3/MgO showed much greater efficiency compared with untreated TiO2/Sb2S3/CuSCN devices (4.1% vs 2.8%). The TiO2/Sb2S3 interface has been improved by a BaTiO3/MgO surface treatment; passivation of the TiO2 surface was also performed at the TiO2/CuSCN interface.

Challenges in Sb2S3-based solar cells

For Sb2S3-based solar cells, theoretical simulations by Bisquert and co-workers88 have shown an optimized efficiency of 8.5% upon decreasing the series and hole-transport resistances and increasing the charge-collection efficiency. Because the electron-hole recombination occurs mostly between the electrons of the CB in the TiO2 layer and the holes in the HTM layer, the hole-transport properties of the HTM become important to improve the device efficiency of solar cells of this type.88, 93 The challenge is hence to find new HTMs with great hole mobility and stability.

Haque and co-workers recorded nanosecond transient absorption spectra (TAS) of thin film samples with a TiO2/Sb2S3/spiro-OMeTAD configuration.93 As Figure 10 shows, the rate of charge recombination at the TiO2/Sb2S3 interface decreased when more successive ionic layer adsorption and reaction (SILAR) cycles were involved, making the optical onsets of the Sb2S3 nanocrystals systematically red-shifted. This effect might decrease the rate of electron injection from the Sb2S3 sensitizer to the TiO2 film and thus decrease the device efficiency, particularly in the wavelength region beyond 750 nm. Hole transfer from the Sb2S3 sensitizer to spiro-OMeTAD was much less sensitive to the spectral shift of the Sb2S3 absorption onset, indicating that the hole transport to spiro-OMeTAD might initiate charge separation at the TiO2/Sb2S3/spiro-OMeTAD interface and thus improve the device performance. The same group investigated Sb2S3/P3HT films using the TAS technique.95 After excitation at 567 nm, the P3HT+ polaron band at 950 nm was monitored to investigate the kinetics of hole transport and charge recombination in Sb2S3/P3HT films. According to the TAS results, the P3HT+ polaron band was strongly dependent on the hole transfer from Sb2S3 to P3HT, indicating that this interfacial hole transport is a key parameter for improving the device performance.

Figure 10
figure 10

(a) Decay kinetics of the transient absorption of Sb2S3-localized holes in metal-oxide/Sb2S3/films probed at 800 nm. (b) Transient decay kinetics of spiro-OMeTAD+ absorption in TiO2/Sb2S3/spiro-OMeTAD (main figure) and ZrO2/Sb2S3/spiro-OMeTAD (inset) films probed at 1600 nm. In both (a) and (b), λpump is varied to probe increasingly red-absorbing nanocrystals while ensuring similar ground state absorbance (2 SILAR cycles (black trace)—450 nm excitation, 3 cycles (red)—510 nm, 4 cycles (blue)—570 nm, 5 cycles (pink)—650 nm). Values of ΔOD are scaled to the number of photons absorbed at λpump. Laser excitation energy density=26 μJ cm−2 at 450 nm. (c) Transient absorption spectra of ZrO2/Sb2S3 (blue trace) and ZrO2/Sb2S3/spiro-OMeTAD (red trace) recorded 1 μs after photoexcitation at 510 nm (laser excitation energy density=40 μJ cm−2). Absorption spectrum of spiro-OMeTAD in chlorobenzene chemically oxidized with N(PhBr)3SbCl6 (black dashed trace) (adapted from OMahony et al.,93 copyright 2012, Royal Society of Chemistry).

Another approach for improving the device efficiency of QDSCs is to extend the optical absorption range of the QD to 1000 nm to improve the light-harvesting performance of the device. The Sb2S3 nanocrystals can absorb light up to 750 nm with an optical band gap of 1.65 eV. To further extend the optical absorption into the near-infrared region, we must consider semiconductors with smaller band gaps, Eg<1.3 eV, such as Sb2Se3.

Sb2Se3-sensitized solar cells

Antimony chalcogenide semiconductor Sb2Se3 has an optical band gap of 1.0–1.2 eV and shows a strong optical absorption, with a coefficient α of 105 cm−1 in the visible region.84, 96, 97, 98 Theoretical predictions for JSC as a function of the band gap Eg give the upper limits as 22 and 43 mA cm−2 for Sb2S3 (Eg=1.7 eV) and Sb2Se3 (Eg=1.13 eV) sensitizers, respectively.96 For the Sb2Se3-based devices, the challenges involve finding an effective synthetic method for the Sb2Se3 nanocrystals deposited on the mesoporous film and finding a suitable HTM to match the energy levels.

Haque and co-workers reported TAS results for TiO2/Sb2Se3/spiro-OMeTAD devices.99 Sb2Se3 nanocrystals were deposited on the surface of TiO2 (or ZrO2 as a control experiment) with the SILAR method using SbCl3 dissolved in acetone and selenide dissolved in ethanol under a N2 atmosphere. The optical absorption of the Sb2Se3 nanocrystals increased and red-shifted with increasing SILAR cycles because of the increased surface coverage (QD loading) and size of the nanoparticles. Figure 11 shows the TA profiles and the energy level diagrams for the Sb2Se3 system with TiO2 and ZrO2 films; the yield of long-lived charge-separated states in the TiO2 films was triple that in the ZrO2 films because of the more rapid charge recombination of the latter. With increasing annealing temperature, the yield of charge separation was decreased, and the recombination lifetime also decreased because of the increased aggregation of the Sb2Se3 nanoparticles and the oxidation of the Sb2Se3 surface. A passivation layer of Inx(OH)ySz between TiO2 and Sb2Se3 can diminish the interfacial charge recombination and thus increase the decay coefficients of the transient signals by 10 times. For the TiO2/Sb2Se3/spiro-OMeTAD devices, the spiro-OMeTAD+ charge yield was increased from 45 to 80%, and the lifetime of the separated charges was 50 ms because of an effective match of the energy levels, a greater Sb2Se3 coverage on the TiO2 film and the decreased interfacial charge recombination. For the ZrO2/Sb2Se3/spiro-OMeTAD devices, small transient signals were observed due to an electron transfer from Sb2Se3 to the spiro-OMeTAD. In this device, the transient signals in the ZrO2 films decayed much more rapidly than those in the TiO2 film. The rapid decays observed in the ZrO2 films are related to the recombination of the localized electrons inside the Sb2Se3 layer and the holes in the spiro-OMeTAD. The TAS results thus indicate that Sb2Se3 might be an effective material for light harvesting in a highly efficient solid-state solar cell.

Figure 11
figure 11

(a) Transient kinetics of spiro-OMeTAD/Sb2Se3/TiO2 and spiro-OMeTAD/Sb2Se3/ZrO2 films recorded at 1600 nm. (b, c) The energy diagrams of the systems include the recombination paths monitored with transient optical measurements (solid arrows) and previous processes (dashed arrows). All measurements were taken under N2, exciting all samples at 450 nm (fluency 6.9 μJ cm−2). The energy level for spiro-OMeTAD was reported elsewhere (adapted from Guijjaro et al.,99 copyright 2012, American Chemical Society).

Perovskite-based mesoscopic solar cells

Perovskite originally referred to a mineral containing CaTiO3, named after Russian mineralogist Lev Perovski, and the term was later extended to encompass a class of compounds with a crystal structure of the same type as CaTiO3. Figure 12 shows the unit cell of a three-dimensional crystal structure for perovskite compounds with the general chemical formula ABX3. Instead of oxide perovskite species (X=O), halide perovskite compounds (X=Cl, Br or I) were found to feature excellent light-harvesting and electron-conducting properties and are perfectly suitable for use as prospective photovoltaic materials.1, 2, 3

Figure 12
figure 12

Schematic representation of the organometal-halide perovskite structures ABX3 with possible species A, B and X as indicated.

Perovskite-sensitized solar cells with liquid electrolytes

In 2009, Miyasaka and co-workers100 reported the first perovskite-sensitized TiO2 solar cell using liquid electrolytes based on iodide and bromide. According to their approach, lead bromide perovskite (CH3NH3PbBr3) was deposited on the TiO2 film with a spin-coating procedure with the precursor solution containing CH3NH3Br and PbBr2. The coated TiO2 film showed nanocrystalline particles approximately 2–3 nm in size on the surface of the TiO2 nanoparticles. For these TiO2/CH3NH3PbBr3 solar cells, the IPCE action spectrum showed a maximum value of 65% at 400 nm, but it fell to zero beyond 550 nm. The corresponding device with JSC=5.57 mA cm−2, VOC=0.96 V and FF=0.59 gave a PCE of 3.1% under one-sun illumination.100 For these TiO2/CH3NH3PbI3 solar cells, the IPCE spectrum showed effective optical absorption up to 800 nm with a maximum value of 45% at 500 nm. As a result, the PCE (3.8%) was larger for the iodide perovskite device than for the bromide perovskite device because JSC (11.0 mA cm−2) was much greater for the former than for the latter, even though the VOC of the former device was only 0.61 V.

Later in 2011, Park and co-workers101 reported on the improved efficiency of TiO2 solar cells sensitized with lead iodide perovskite (CH3NH3PbI3) and an electrolyte based on iodide. In their approach, Pb(NO3)2 was used for modification of the surface of a mesoporous TiO2 film before coating the perovskite QD, such that Pb(NO3)2 acted as a blocking layer in the solar cells. After the TiO2 surface modification, the device showed an improved PCE of 6.5% with JSC=15.8 mA cm−2, VOC=0.70 V and FF=0.58; the improved VOC was due to the retarded charge recombination effected by the surface treatment. The authors also mentioned the stability of the perovskite nanocrystals in the iodide electrolyte, whereby the perovskite QD gradually dissolved into the iodide electrolyte after being irradiated for 10 min.

The organic part of the perovskite species is modifiable through a similar synthetic approach. For example, a CH3CH2NH3PbI3 sensitizer can be deposited on a mesoporous TiO2 film via spin coating with an equimolar mixture of CH3CH2NH3I and PbI2 in a γ-butyrolactone solution.102 The average diameter of the CH3CH2NH3PbI3 nanocrystals was approximately 1.8 nm; the optical band gap of this species increased to 2.2 eV, as estimated from the diffuse reflectance spectra. For the TiO2/CH3CH2NH3PbI3 solar cells, the IPCE spectrum attained a maximum value of 60% at 600 nm, giving a PCE of 2.4% in the iodide-based electrolyte. The poor performance of the device was due to the large optical band gap of the perovskite sensitizer of this type, which limited light harvesting in the visible region.

Synthesis of lead-iodide perovskites on metal-oxide films

Perovskite CH3NH3PbI3 nanocrystals were formed on mesoporous TiO2 films via spin coating with a solution containing CH3NH3I and PbI2 in an equimolar ratio.12, 101 The concentrations of the spin-coating solution were varied from 10 to 40% by mass in γ-butyrolactone. The nanocrystalline (CH3NH3)PbI3 layer formed on the surface of the TiO2 film after drying at temperatures from 40 to 160 °C for 30 min, typically at 100 °C for 30 min. With an increasing concentration of the coating solution, the color of the perovskite (CH3NH3)PbI3 on the TiO2 film varied from yellow (10% by mass) to black (40% by mass). These perovskite nanocrystalline materials have an average diameter of approximately 2.5 nm, showing QD behavior on the surface of the TiO2 nanoparticles, according to the TEM images shown in Figure 13.101

Figure 13
figure 13

TEM micrographs of (a) a wide view of (CH3NH3)PbI3 deposited on TiO2, (b) a magnified image of (CH3NH3)PbI3 deposited on TiO2, (c) pure (CH3NH3)PbI3 and (d) bare TiO2. Arrows in (b) indicate quantum dot perovskite (CH3NH3)PbI3 with a hemispherical shape. Inset in (c) shows a selected-area electron-diffraction pattern of pure (CH3NH3)PbI3. For the (a) and (b) TEM images, 30.18 and 40.26% (by mass) perovskite precursor solutions were used, respectively (adapted from Im et al.,101 copyright 2011, Royal Society of Chemistry).

Instead of the mesoporous TiO2 film, an insulating Al2O3 layer served as a scaffold for coating the perovskite halide (CH3NH3PbI2Cl) as a light harvester on the surface of the Al2O3 film. The perovskite nanocrystalline film was prepared from a precursor solution containing CH3NH3I and PbCl2 in a molar ratio of 3:1 in anhydrous N,N-dimethylformamide.11 After spin-coating and drying, the halide perovskite absorber on the Al2O3 film had an I:Cl ratio of approximately 2:1, according to the energy-dispersive X-ray analysis.

Electronic properties of lead iodide perovskites on metal-oxide films

Park and co-workers reported an alignment of the energy levels in a lead iodide perovskite (CH3NH3PbI3) QDSC. The CH3NH3PbI3 nanocrystals served as the light absorber, and spiro-OMeTAD was used as the hole-transporting layer. This perovskite nanomaterial adsorbed on a TiO2 film had a direct optical transition with an energy gap Eg of 1.5 eV.12 From the energy-level alignment shown in Figure 1, we expect excellent properties of the device for electron injection and charge separation; that is, electrons can feasibly inject from the perovskite species into the CB of a TiO2 film, and holes can transfer from the sensitizer perovskite to the spiro-OMeTAD HTM.

Snaith and co-workers11 reported on a perovskite (CH3NH3PbI2Cl) absorber on the surface of an insulating Al2O3 film. This insulating Al2O3 has a wide band gap Eg of 7–9 eV and acts purely as a mesoporous scaffold for the perovskite (CH3NH3PbI2Cl) to be deposited. In this case, the photo-induced electrons could not inject into the Al2O3 film because of the insulating nature of the metal-oxide film, but the holes generated in the perovskite could propagate into the HTM layer, as shown in Figure 14. After charge separation between the CH3NH3PbI2Cl absorber and the spiro-OMeTAD HTM, the photo-induced electrons were transported directly from the perovskite on the surface of the Al2O3 to the FTO electrode. The perovskite species thus acts as a bi-functional material to perform both light-absorbing and electron-transport functions. The authors also found that the rates of electron transport (diffusion) in the Al2O3-based devices were 10 times those in TiO2-based devices, according to the data for the transient photocurrent decay shown in Figure 14. These data demonstrate the superior photovoltaic performance of the perovskite SCs using a mesoporous insulating metal-oxide layer rather than a conventional mesoporous TiO2 film.

Figure 14
figure 14

(a) Illustration of charge transfer and transport in a perovskite-sensitized TiO2 solar cell (left) and a non-injecting Al2O3-based perovskite solar cell (right). (b) Photo-induced absorbance (PIA) spectra of mesoporous TiO2 (black open circles) and Al2O3 (red crosses) films coated with perovskite with (solid lines) and without (dashed lines) spiro-OMeTAD hole transporter. λex=496.5 nm, repetition rate 23 Hz. (c) Charge transport lifetime determined with a small-perturbation transient photocurrent decay of perovskite-sensitized TiO2 (circles with a black line) and Al2O3 cells (red crosses with a line). Inset shows normalized photocurrent transients for Al2O3 (red trace with crosses every seventh point) and TiO2 (black trace with circles every seventh point) cells, set to generate 5 mA·cm−2 photocurrent from a background light bias (adapted from Lee et al.,11 copyright 2012, American Association for the Advancement of Science).

Photovoltaic characteristics of perovskite solar cells

Perovskite CH3NH3PbI3 nanocrystals have large coefficients of optical absorption in the visible region, ranging from 104 to 105 cm−1.12 These nanocrystals also have excellent matching of energy levels with those of TiO2. Interfacial bonding of perovskites with TiO2 and spiro-OMeTAD was effective in ensuring uniform penetration and complete filling of the pores. Figure 15 shows the photovoltaic performance of the corresponding device; the IPCE values attained a maximum higher than 60% at 450 nm and were maintained greater than 50% up to 750 nm. The IPCE characteristics indicate that the CH3NH3PbI3 nanocrystals deposited on the mesoporous TiO2 film feature excellent light-harvesting properties, bestowing an excellent quantum efficiency to the device in the visible region. The perovskite device of 0.6-μm film thickness showed a high photocurrent density (JSC=17.6 mA cm−2) and VOC (=0.88 V), a satisfactory FF (=0.62) and excellent overall device performance (η=9.7%) under one-sun irradiation.12 With increased thickness of the TiO2 film, both VOC and FF of the device decreased because of the increased dark current and resistance to electron transport. The results obtained from the TAS measurements indicate that the holes were completely transferred from the perovskite sensitizer to the spiro-OMeTAD HTM after charge separation in the sensitizer; the normalized photocurrent density was linearly proportional to the light intensity (Figure 15), without charge loss.12

Figure 15
figure 15

Photovoltaic characteristics of a (CH3NH3)PbI3 perovskite-sensitized solar cell. (a) Photocurrent density as a function of forward bias voltage. (b) IPCE as a function of incident wavelength. (c) Short-circuit photocurrent density as a function of light intensity (adapted from Kim et al.,12 copyright 2012, Scientific Reports).

Mixed-halide perovskite (CH3NH3PbI2Cl) nanocrystals have an optical absorption onset (λ800 nm) similar to that of the iodide perovskite (CH3NH3PbI3) mentioned previously. Figure 16 presents a comparison of the photovoltaic performance of the CH3NH3PbI3 SCs using either the TiO2 or the Al2O3 film as the metal-oxide layer.11 For the Al2O3-based device, the IPCE curve attains a maximum value greater than 80% at 400 nm, and the values are maintained at greater than 60% up to 700 nm; for the TiO2-based device, the IPCE values are slightly smaller than those of the Al2O3-based device, with similar JSC for both devices.11 The VOC of the Al2O3 device was, however, much greater than that of the TiO2 device because the loss of injection potential was much less for the former than for the latter. As a result, VOC significantly increased (200 mV) for the devices from the TiO2 film to the insulating Al2O3 film, and the best device efficiency of 10.9% was found for the Al2O3 device under one-sun conditions.11

Figure 16
figure 16

(a) IPCE action spectrum of an Al2O3-based and perovskite-sensitized TiO2 solar cell, with a device structure of FTO/compact TiO2/mesoporous Al2O3 (red trace with crosses) or mesoporous TiO2 (black trace with circles)/CH3NH3PbI2Cl/Spiro-OMeTAD/Ag. (b) Current-voltage characteristics under illumination (simulated AM1.5G 100 mW cm−2) for Al2O3-based cells, with one cell exhibiting great efficiency (red solid trace with crosses) and one exhibiting greater than VOC 1.1 V (red dashed line with crosses); a perovskite TiO2-sensitized solar cell (black trace with circles); and a planar-junction diode with a structure FTO/compact TiO2/CH3NH3PbI2Cl/Spiro-OMeTAD/Ag (purple trace with squares) (adapted from Lee et al.,11 copyright 2012, American Association for the Advancement of Science).

In 2013, Snaith and co-workers2 reported highly efficient perovskite CH3NH3Pb(I,Cl)3 solar cells, with the efficiency of the best device approaching 12.3% based on a mesoporous Al2O3 film that was 0.4 μm thick. For the Al2O3 films with a thickness less than 0.4 μm, the perovskite solar cell had an efficiency of 9.1% with a capping layer of thickness of 0.3–0.4 μm. For the Al2O3 films with a thickness greater than 0.4 μm, the perovskite solar cells had, in contrast, an efficiency near 12% without a perovskite capping layer. Seok and co-workers reported on perovskite CH3NH3Pb(I,Br)3 solar cells with a maximum efficiency of 12.3%.3 In contrast with the approach of Snaith, using the Al2O3 film as a scaffold for perovskites, the best device performance of Seok’s method was achieved using TiO2 films of thickness 0.6 μm. For the mesoporous TiO2-based devices, the efficiency was optimized by using an alloy comprising both CH3NH3PbI3 and CH3NH3PbBr3 crystals, resulting in increased device efficiency and superior device stability. The mixing ratios of the solutions containing perovskites CH3NH3Pb(I1−xBrx)3 were reported for x varying in the range of 0.06–0.20. Figure 17 presents the stability data of the devices, indicating that the devices containing less Br (x=0.06 or zero) had a higher efficiency than the others in the initial stage of the test, whereas those containing more Br (x=0.20 or 0.29) exhibited superior durability under ambient conditions because of their compact nanocrystalline structure.

Figure 17
figure 17

Variation of the power conversion efficiency of heterojunction solar cells based on CH3NH3Pb(I1-xBrx)3 (x=0, 0.06, 0.20, 0.29) with time, stored in air at around 23 °C without encapsulation. The humidity was maintained at 35%; the cells were exposed to 55% humidity on the fourth day to investigate the variation in performance under high humidity (adapted from Noh et al.3 copyright 2013, American Chemical Society).

Seok and co-workers also fabricated perovskite CH3NH3PbI3/poly-triarylamine (PTAA) solar cells with an efficiency of 12%.103 The device using PTAA as an HTM layer showed much better performance than those using other polymeric hole conductors such as P3HT (η6.7%), PCPDTBT (η5.3%) and PCDTBT (η4.2%). Park and co-workers104 demonstrated perovskite CH3NH3PbI3 solar cells based on rutile TiO2 nanorods with an efficiency of 9.4%. Park’s results indicate that the photovoltaic performance was strongly dependent on the length of the TiO2 nanorods because of the efficiency of charge generation rather than because of the kinetics of charge recombination. Hagfeldt and co-workers105 reported perovskite CH3NH3PbI3/spiro-OMeTAD solar cells with an efficiency of 8.5%. The spiro-OMeTAD device showed a much improved performance compared with devices using P3HT (η4.5%) and 4-(diethylamino)-benzaldehyde diphenylhydrazone (DEH) (η1.6%) because of the greater electron lifetime of the former. Other perovskite-based HMSCs have been studied, with various TiO2 nanostructures and HTMs.106, 107, 108, 109 Hagfeldt and co-workers110 reported ZrO2/CH3NH3PbI3/spiro-OMeTAD solar cells with an efficiency of 10.8%. Perovskite CH3NH3PbI3/fullerene heterojunction hybrid solar cells have also been fabricated.111

Chung et al.112 reported the use of perovskite CsSnI3 as an HTM in solid-state DSSCs. Perovskite CsSnI3 is a direct band gap p-type semiconductor with an optical absorption onset at 1.3 eV (Figure 1b), and it has a large hole mobility (μh=585 cm2V−1 s−1) near 23 °C; the latter value is 100–1000 times that of a polymer-type HTM. As a result, the device consisting of CsSnI2.95F0.05/N719 dye/TiO2 attained a PCE of 8.5% with an enhanced IPCE in the red visible region.112 Chen et al.113 reported Schottky solar cells based on perovskite CsSnI3 thin films; these films were prepared by coating multiple layers of SnCl2 and CsI on a glass substrate followed by annealing at 175 °C on a hotplate for 1 min under ambient conditions. The CsSnI3 film was formed during annealing from the SnCl2/CsI stack. The JV characteristic curve showed a small PCE (0.88%) for an SC of this type. The perovskite mesoscopic solar cell without an HTM layer was reported to possess a PCE of 5.5%.114

Challenges in perovskite-based solar cells

The synthesis of perovskite CH3NH3PbI3 nanocrystals is sensitive to ambient humidity. According to results reported elsewhere,101 the perovskites gradually dissolved into a liquid-type electrolyte after irradiation for 10 min. As a result, anhydrous starting materials should be used in a dry box during synthesis. Another issue is the thickness and porosity of the HTM layer (such as spiro-OMeTAD) coated on the surface of the perovskite. For example, the contact resistance and series resistance at the interface of the CH3NH3PbI3/spiro-OMeTAD/cathode must be decreased by optimizing the thickness of the spiro-OMeTAD layer. A thinner spiro-OMeTAD layer is required to decrease the series resistance and hole-transport resistance through the spiro-OMeTAD to the cathode. As the conductivity of a typical halide perovskite is on the order of 10−3 S cm−1, the spiro-OMeTAD layer should be thick enough to prevent an electrical short circuit between the perovskite layer and the counter electrode. Greater FF and PCE of the devices can be achieved by employing new types of HTMs with greater mobility and by introducing a thin capping layer on the perovskite sensitizer. Nevertheless, the best-performing perovskite solar cell was reported in 2013 with a structure of TiO2/perovskite/spiro-OMeTAD/Au, which exhibited a remarkable PCE of 15.0% (JSC=20.0 mA cm−2, VOC=0.993 V, FF=0.73) under an illumination of 96.4 mW cm−2.1 It is worth noting that when the device was encapsulated under argon, it could maintain greater than 80% of its initial PCE after 500 h.1

Conclusions

Highly efficient dye-sensitized nanocrystalline TiO2 solar cells have been developed since 19918; the most promising devices are based on ruthenium or porphyrin dyes with efficiencies of power conversion exceeding 12%,9, 10 but these traditional DSSCs have durability problems due to (i) instability of the organic sensitizers and (ii) leakage of the devices containing typical liquid-type electrolytes. To solve these critical problems for DSSCs, various inorganic light harvesters have been intensively developed, in combination with hole-conducting materials of organic or polymer type. Therefore, herein we have reviewed inorganic nanomaterials of two types—metal chalcogenides and halide perovskites—as prospective photosensitizers or light absorbers for hybrid mesoscopic solar cells. A summary of the photovoltaic performances of these inorganic photosensitizers in liquid-type electrolytes appears in Table 1. The power conversion efficiencies of these devices are randomly distributed from 2 to 6%, and their performances are much poorer than those of the liquid-type DSSCs, according to the comparison shown in Figure 18. Although a large photocurrent was reported for a mercury-doped PbS-sensitized SC (JSC=30 mA cm−2),18 its poor photovoltage is a typical problem for QD-based solar cells. All-solid-state mesoscopic solar cells have been rapidly developing and have shown astonishing progress in their device performance during the period 2012–2013. A summary of the photovoltaic performances of all-solid-state quantum-dot-sensitized solar cells and lead halide perovskite-based solar cells appears in Table 2; their progress trends are demonstrated in Figure 19. The first all-solid-state DSSC was reported in 1998 with a small PCE (0.74%); it had a ruthenium complex dye and an organic HTM (spiro-OMeTAD) as a solid-state electrolyte.116 The device efficiency of an all-solid-state DSSC quickly rose to 4.0% in 2005 (Z907)117 and to 5.0% in 2010 (C106).118 In 2011, two metal-free organic dyes, C220119 and Y123,120 showed promising performances for all-solid-state devices, attaining efficiencies of 6.1 and 7.2%, respectively. Since 2010, metal chalcogenide QDs have played an important role as suitable sensitizers for solid-state mesoscopic TiO2 solar cells. The two most promising QD sensitizers are PbS and Sb2S3; the PCEs of these QDSCs reached 7% in 2012. The most remarkable advancement in the development of all-solid-state mesoscopic solar cells was the discovery of light harvesters of the perovskite type. As Figure 19 shows, the development of perovskite-based solar cells has progressed expeditiously, with achieved PCEs of 10%11, 12 in 2012, followed by 12%2, 3, 103 and then 15%1 in 2013. Such striking and rapid progress in the development of organic-inorganic hybrid photovoltaic devices indicates the threshold of a new era for the commercialization of all-solid-state mesoscopic solar cells in the near future.

Table 1 Summary of the device performances for liquid-type QDSCs and perovskite solar cells
Figure 18
figure 18

Charts of the efficiency progress of liquid-type QDSCs and perovskite-based SCs developed during the period 2007–2013. The solid curve represents the average progress trend with the corresponding references cited in parentheses. The dashed and dashed-dotted curves represent those of DSSCs with sensitizers of ruthenium complexes and porphyrin dyes, respectively. The symbols MA and EA represent perovskites with organic components of CH3NH3 and CH3CH2NH3, respectively.

Table 2 Summary of the device performances for solid-state QDSCs and perovskite solar cells
Figure 19
figure 19

Efficiency progress charts of all-solid-state DSSCs, QDSCs and perovskite-based SCs developed during 2005–2013. The solid curve represents the average progress trend with the corresponding references quoted in parentheses. The symbol MA represents perovskites with an organic component of CH3NH3.