Cite this Article
Du Hui-Jing, Wang Wei-Chao, Zhu Jian-Zhuo. Device simulation of lead-free CH3NH3SnI3 perovskite solar cells with high efficiency. Chinese Physics B, 2016, 25(10): 108802  
Permissions
Device simulation of lead-free CH3NH3SnI3 perovskite solar cells with high efficiency
Du Hui-Jing, Wang Wei-Chao, Zhu Jian-Zhuo†,
College of Science, Yanshan University, Qinhuangdao 066004, China

 

† Corresponding author. E-mail: zhujz@ysu.edu.cn

Project supported by the Graduate Student Education Teaching Reform Project, China (Grant No. JG201512) and the Young Teachers Research Project of Yanshan University, China (Grant No. 13LGB028).

Abstract
Abstract

The lead-free perovskite solar cells (PSCs) have drawn a great deal of research interest due to the Pb toxicity of the lead halide perovskite. CH3NH3SnI3 is a viable alternative to CH3NH3PbX3, because it has a narrower band gap of 1.3 eV and a wider visible absorption spectrum than the lead halide perovskite. The progress of fabricating tin iodide PSCs with good stability has stimulated the studies of these CH3NH3SnI3 based cells greatly. In the paper, we study the influences of various parameters on the solar cell performance through theoretical analysis and device simulation. It is found in the simulation that the solar cell performance can be improved to some extent by adjusting the doping concentration of the perovskite absorption layer and the electron affinity of the buffer and HTM, while the reduction of the defect density of the perovskite absorption layer significantly improves the cell performance. By further optimizing the parameters of the doping concentration (1.3× 1016 cm−3) and the defect density (1× 1015 cm−3) of perovskite absorption layer, and the electron affinity of buffer (4.0 eV) and HTM (2.6 eV), we finally obtain some encouraging results of the Jsc of 31.59 mA/cm2, Voc of 0.92 V, FF of 79.99%, and PCE of 23.36%. The results show that the lead-free CH3NH3SnI3 PSC is a potential environmentally friendly solar cell with high efficiency. Improving the Sn2 + stability and reducing the defect density of CH3NH3SnI3 are key issues for the future research, which can be solved by improving the fabrication and encapsulation process of the cell.

PACS: 88.40.H–;88.40.hj;88.40.fc
Keyword:CH3NH3SnI3;perovskite solar cells;device simulation;high efficiency
1. Introduction

In recent years, lead halide (CH3NH3PbX3, X = Cl, Br, I) perovskite solar cells (PSCs) have drawn a great deal of attention because they have lower cost and simpler processing techniques[1–3] than traditional silicon based solar cells. In the past few years, the power conversion efficiency (PCE) of CH3NH3PbX3 PSC has been significantly enhanced from 3.8% in 2009[4] to 221% in 2016.[5] Despite the significant progress of the PSCs, the toxicity of lead remains a major obstacle to the wide application of these PSCs. Experimental and theoretical studies show that CH3NH3SnI3 has a narrower band gap of 1.3 eV,[6,7] which can cover a wider range of the visible spectrum, than the band gap of the lead halide perovskite (1.55 eV). First-principles study indicates that CH3NH3SnI3 is a promising perovskite absorber in the highly efficient solar batteries with the best optical properties and the widest light-adsorption range in all the CH3NH3BX3 (B = Sn, Pb; X = Cl, Br, I) compounds.[8] As a lead-free alternative, tin-based perovskites have been applied to solar cells with planar heterojunction architecture.[9–14] However, these devices degrade rapidly in the air. This is mainly attributed to the easy oxidization of Sn2+ into Sn4+, which greatly restricts the development of the tin-based PSCs. Recently, with the development of the fabrication and encapsulation process, the stability of the CH3NH3SnI3 based cells has been improved. The addition of SnF2 in the system reduces the Sn4+ caused by the oxidation of Sn2+.[14] FASnI3 PSC with high duplicability has been fabricated with SnF2 as an inhibitor of Sn4+, and the encapsulated device has exhibited a stable performance for over 100 days, maintaining 98% of its initial efficiency.[9] This strategy is now commonly applicable to the fabrication of Sn-based PSC. Moreover, by chemical substitution of I by Br in the form of the solid solution CH3NH3SnI3−xBrx, the band gap can be tuned to cover a wide range of the visible spectrum. Thus, the Sn-based hybrid perovskites could be a promising alternative to the Pb-based light-harvesting materials,[14,15] and this provides an opportunity to develop lead-free solar cells with high efficiency.

In spite of the progress of the stability improvement in the Sn-based PSCs, the PCE of the CH3NH3SnI3 based PSCs is still very low, so it is necessary to understand the relationship between the structure parameters and the cell performance deeply. Yet to date, there has been no report on the device simulation of CH3NH3SnI3 based PSCs. Simulation methods allow intuitive examination of each parameter in solar cells and thus identify the optimal conditions for operating.[16–20] In this paper, the factors affecting the lead-free CH3NH3SnI3 PSC efficiency are analyzed by one-dimensional device simulation with SCAPS (ver.3.3.02) under AM1.5G illumination. The solar cell capacitance simulator (SCAPS) is a general solar cell simulation program that is based on three basic semiconductor equations and it is well adapted to modeling various hetero- and homo-junctions, multi-junction, and Schottky barrier devices.[18–20]

2. Device simulation parameters

In the CH3NH3SnI3 based solar cell adopted is a planar heterojunction architecture with layer configuration of glass substrate/TCO/buffer layer TiO2(ETM)/absorption layer CH3NH3SnI3/hole transport material (HTM) spiro-OMeTAD /metal back contact (see Fig. 1). Note that the main material parameters are carefully selected from those reported in experimental data and other theoretical results.[7–10,17–21] Table 1 summarizes the primary parameters for each layer in the simulation. The unmarked parameters of TCO, buffer and HTM layers are cited from the simulation literature about PSCs.[7,17–21] Thermal velocities of the electron and hole are both set to be equal to 107 cm/s. The defects in the absorption layer are set to be in the neutral Gaussian distribution with a characteristic energy of 0.1 eV, and the defect energy level is at the center of the band gap. To consider the interface carrier recombination, we insert defect interfaces into the buffer/absorber and the absorber/HTM respectively. The interface defects are set to be single and neutral defect with a total density of 1×1017 cm−3, located at 0.6 eV above the top of valence band Ev. The parameter setting of the defects in the simulation is shown in Table 2. Pre-factor Aα is 105 to obtain the absorption coefficient (α) curve calculated by α = Aα (hνEg)1/2. The optical reflectance neither at the surface nor at the interface of each layer is considered in this simulation.

Fig. 1. Schematic diagram of the CH3NH3SnI3PSC.
Table 1.

Simulation parameters of CH3NH3SnI3 PSC.

.
Table 2.

Parameters setting of interface defect and the defect in the absorber.[18,20]

.

Our study focuses on the effects of the doping concentration, the defect density and the thickness of the perovskite absorber layer, and the electron affinity χ of the ETM and the HTM. The control variable method is used in the study during the parameter optimization. The initial parameters are shown in Table 1. The initial defect density Nt of the absorber is set to be 4.5×1017 cm−3, because on this condition the simulated carrier diffusion length of 30 nm is similar to the experiment value of Noel et al.’s research on CH3NH3SnI3.[8] The tin perovskite shows a p-type conducting behavior caused by the self-doping process of Sn2+’s easy oxidization into Sn4+,[11] so we assume that the absorber is a p-type semiconductor doped with an initial carrier density of 3.2×1015 cm−3.

With these initial parameters in Table 1, we study the current density–voltage (J-V) characteristic of the cell (see curve 1 in Fig. 2). The short-circuit current density (Jsc) of 18.67 mA/cm2, open-circuit voltage (Voc) of 0.68 V, fill factor (FF) of 47.43%, and power conversion efficiency (PCE) of 6.09% are obtained. The simulated device performance is consistent with the experimental values of the Tin-based PSCs,[9–11] certifying that the device simulation is valid and the input parameters that have been set are close to those for a real device. Due to the narrower band gap 1.3 eV of CH3NH3SnI3 than 1.55 eV of CH3NH3PbI3, the optical absorption edge of tin perovskite is red-shifted to 950 nm in the external quantum efficiency (QE) curve (see Fig. 3). The QE covers the entire visible spectrum and reaches a broad absorption maximum over 50% from 400 nm to 850 nm accompanied by a notable absorption onset up to 950 nm, which is in good accordance with the measured PCE spectrum of Ref. [22]. The red-shift of the QE curve is more beneficial to the light absorption at infrared wavelengths.

Fig. 2. J–V curve of the PSCs during the optimization.
Fig. 3. External quantum efficiency curve of the PSC with the parameters in Table 1 and without considering the variation of the parameters.

In Section 3, based on the parameters above, the factors affecting the solar cell performance, such as the doping concentration, the defects density and the thickness for the perovskite absorber layer, the electron affinity χ of the ETM and the HTM, are studied in depth.

3. Results and discussion
3.1. Influence of doping concentration of perovskite absorption layer

CH3NH3SnX3 (where X = Cl, Br, I) is unstable in ambient atmosphere: the Sn2+ ion will rapidly oxidize into more stable Sn4 + analogue,[9] which acts as a p-type dopant within the material in the “self-doping” process. Earlier studies on bulk CH3NH3SnI3 show that the doping level can be varied greatly in a range of 1014 cm−3–1019 cm−3 due to the presence of Sn4+ impurities.[9,10] In order to make it clear how the acceptor doping concentration (NA) of the perovskite absorption layer can affect the performances of solar cells, CH3NH3SnI3 layers with the values of NA ranging from 1014 cm−3 to 1017 cm−3 are considered. Figure 4 provides the PCEs of solar cells with different acceptor densities of the perovskite, and the maximum value of PCE appears at the NA of 1.3×1016 cm−3. The Jsc and Voc are also related to the change of NA for the perovskite, and both of them reach their maximum values when the NA is approximately 1×1016 cm−3. The external quantum efficiency (QE) significantly increases with the NA of perovskite increasing from 1014 cm−3 to 1016 cm−3 (see Fig. 5), which implies that the generation rate of the photo-generated carriers increases under the same incident photon number. Hence, an appropriate doping concentration of the perovskite absorption layer is beneficial to the improvement of the photo-absorption efficiency and the Jsc. However, the Voc drops rapidly when the NA exceeds 1×1016 cm−3. The variation in the cell performance with the doping concentration can be explained from the perspective of the built-in electric field which is enhanced with the increase of doping concentration. The enhancement of the electric field promotes the separation of carriers and then the improvement of the cell performance. However, further increasing the doping concentration will cause a higher Auger recombination rate, which is not beneficial to the increase of Voc. It can be found from Fig. 6 that the recombination rate R increases significantly when the NA exceeds 1×1016 cm−3. Besides, the hole transportation will be greatly suppressed with the increase of p-type doping concentration of the perovskite absorption layer because of the enhanced impurity scattering and recombination. The quasi Fermi level of the hole (EFp) becomes far from the valence band top with the increase of NA (see Fig. 7), even flattens into the EFp of the HTM. This affects the hole transportation strongly from the absorption layers to HTM, and the reduction of the hole density can also be seen from the EFp uplift of the perovskite absorption layer.

Fig. 4. Performance parameters of PSC with different acceptor densities of perovskite.
Fig. 5. Variations of external quantum efficiency (QE) with NA of perovskite
Fig. 6. Variation of recombination rate (R) with NA of perovskite.
Fig. 7. Distributions of quasi Fermi level of the hole (EFp) for different values of NA.

Consequently, only appropriate doping concentration can commendably improve the Jsc and Voc, and then obtain a larger PCE value, and excessive doping concentration is unfavorable because of the increase of the recombination. The deposited films of tin perovskite must have low carrier concentration to maximize the carrier mobility within the active perovskite. Under the doping concentration of 1.3×1016 cm−3 for the perovskite absorption layer, the optimum cell performance with Jsc of 20.55 mA/cm2, Voc of 0.70 V, FF of 55.58%, and PCE of 8.03% is obtained (see Fig. 4).

The comparison between current density–voltage curves with and without NA optimization is shown in curves 2 and 1 of Fig. 2 respectively. At the optimum NA of 1×1016 cm−3, the PCE by 1.94% compared with that of the solar cell with an initial NA of 3.2×1015 cm−3. This consequence means that controlling the instability of Sn2+ oxidation state in perovskite layer and avoiding excessive Sn4+ in the perovskite are very important for optimizing the performance of the tin-based PSC. Experimental studies have shown that the addition of SnF2 in the system reduces Sn4+ caused by the oxidation of Sn2+,[9] and the encapsulated device showed a stable performance with no significant drop in the absorbance for over 100 days.[9] This strategy is now commonly applicable to the fabrication of Sn-based PSC. These successful experiments indicate that with an appropriate fabrication and sealing process, the tin-based PSC with better stability can be obtained.

3.2. Influences of electron affinity of buffer and HTM

The band offset between buffer/absorption layers/HTM is a decisive factor of carrier recombination at the interface, which determines the open-circuit voltage (Voc).[17] The band offsets are adjusted by varying the values of electron affinity (χ) of the buffer (3.8 eV–4.5 eV) and HTM (2.0 eV–2.9 eV). The variations of Jsc, Voc, FF, and PCE with electron affinity value are shown in Figs. 8(a) and 8(b). A better solar cell PCE can be obtained at the χ values of 4.0 eV–4.2 eV for the buffer and 2.3 eV–2.6 eV for the HTM, respectively. A barrier cliff is formed at the absorber/HTM interface as shown in the inset of Fig. 9. The proper electron barrier cliff around 0.0 eV–0.2 eV does not affect the photo-generated electron flow toward a back electrode, and Jsc is almost constant. When the Ec-buffer is more than 0.2 eV higher than Ec-absorber, the collection of electrons is impeded by the raised barrier formed at the conduction band offset (see Fig. 9). This is because of the double-diode like curvature at higher conduction band offset.[17,23]

Fig. 8. Variations of performance parameters of PSC with electron affinity of buffer (a) and HTM (b).
Fig. 9. Energy band diagrams of perovskite solar cells with charge barriers at the buffer/absorber /HTM interfaces. Insets show close-up images at the interface.

Figure 8(b) shows the variations of the Voc and the Jsc with the χ of HTM. The Voc decreases when the Ev–HTM is too low (lower than 2.2 eV), while the Jsc decreases when the Ev–HTM is too high (higher than 2.7 eV). The optimum position of the Ev–HTM is lower than that of the absorber by around 0.0 eV–0.3 eV and the χ of HTM is in a range of 2.3 eV–2.6 eV. The optimum properties of the solar cell of PCE of 7.21%, Jsc of 20.52 mA/cm2, Voc of 0.70 V, and FF of 50.33% are obtained when the electron affinity values of the buffer and HTM are set to be 4.0 eV and 2.6 eV respectively. Obvious improvement of the voltage-current characteristics can be seen from the J–V curves with the band offset optimization (see curve 3 in Fig. 2). The simulation results of the electron affinity of buffer and HTM are similar to those from the band offset study on CH3NH3PbI3PSCs in Ref. [17], and they also found that selecting buffers and HTMs with suitable electron affinity can inhibit the interface recombination and help to further optimize the solar cells performance.

3.3. Influence of defect density and thickness of perovskite absorption layer

The efficiency of the solar cell can be improved to a certain extent by adjusting the electron affinity values of buffer and HTM or the p-type carrier concentration of perovskite. In order to further improve the PSC performance, another affecting parameter, the defect density of the perovskite layer, should be considered. The morphology and quality of the perovskite film have been recognized recently as an important factor determining perovskite solar cell performance,[24,25] since the photoelectrons are mainly generated in this light-absorber layer. Poor film quality and coverage of the tin perovskite on the mesoporous TiO2 electrodes have been observed.[10] The charge recombination behaviors will become dominant in the light-absorber layer because of the larger defect density (Nt) caused by the poor film quality, which can determine the Voc of the device. The study about the influence of the defect density of the perovskite layer on cell performance is based on the Shockley–Read–Hall recombination model (SRH). The neutral defects with Gaussian distribution and characteristic energy of 0.1 eV are set to be at the center of the band gap. The SRH recombination model is as follows:[26,27]

where n and p are the concentrations of the mobile electrons and the holes, which can be obtained from solving the Poisson equation and the continuity equation for the charge carriers. For sufficient forward bias, i.e., qV > 3kT, we can neglect the term , which is from the thermal generation. Et and Nt are the energy level and concentration of the trap defects respectively. τn,p is the lifetime for the electrons and holes, and can be given by

where τn,p and σn,p are the lifetime and the capture cross-section of the electrons and holes respectively and vth is the thermal velocity. The carrier diffusion length l can be given by

where D is the diffusion coefficient and can be given by

where μ is the carrier mobility. With formula (2)–(4), combining with the relative parameters of the carriers, we can obtain the diffusion lengths of the electrons (ln) and holes (lp). The simulated values of ln and lp are the same, because all the parameters for the electron and hole in the absorber, such as the thermal velocity, the capture cross section, and the mobility (1.6 cm2/V·s, from literature[9]) are set to be the same as those in Table 2,[17] which is consistent with ambipolar characteristics of the carriers.[24] In the actual situation the diffusion length lp of the holes is slightly shorter than ln of the electrons because of the larger effective mass of the hole. Nonetheless, these values of ln and lp for perovskite are relatively balanced compared with typical values reported in bulk heterojunction solar cells, where the electron and hole transport lengths (proportional to their mobility) differ by orders of magnitude.[25] These balanced long charge-carrier diffusion lengths would account for the remarkable performances reported for these perovskite devices. Ignoring the difference between ln and lp, and only focusing on the effect of the defect on the carrier transportation, have little influence on the validity of the simulation.

The initial defect density Nt of the absorber is set to be 4.5×1017 cm−3, because on this condition the simulated carrier diffusion length of 30 nm is similar to the experiment value of Noel et al.’s research on CH3NH3SnI3.[8] Based on previous simulated studies of lead perovskite,[17,19,21] we change defect density Nt from 1014 cm−3 to 1019 cm−3 and depict the variation of photovoltaic property with Nt in Fig. 10. The performance of the device is improved significantly with the reduction of the defect density in perovskite, which is consistent with the numerical simulation of the lead perovskite.[21] When defect density is as low as 1×1015 cm−3 the cell performance is significantly improved attaining the Jsc of 28.48 mA/cm2, Voc of 0.79 V, FF of 67.60% and PCE of 15.28%. Further decrease of the Nt from 1015 cm−3 to 1014 cm−3, the improvement of the cell performance is slight. The realization of such low Nt of 1014 cm−3 in experiment is very difficult so we set the defect density of 1×1015 cm−3 as an optimized value, and all the values of Jsc, Voc, FF and PCE approximately reach their maxima with this defect density. Experimental studies have shown that the tin-based perovskite displays good charge-transport properties.[28,29] In order to understand the influence of the defect density Nt on the cell performance deeply, we study the effects of defect density on the carrier diffusion length l and the recombination rate R (see Table 3 and Fig. 11) based on the SRH recombination model and the formula of l. When the Nt is lower, the diffusion length l is longer and the recombination rate R is lower, which is conducible to the improvement of the cell performance.

Fig. 10. Current density-voltage curves, with increasing the value of Nt.
Fig. 11. Variations of the recombination rate with depth from surface for different values of Nt.
Table 3.

Variations of the diffusion length with defect density.

.

The thickness of the light-absorbing layer plays a crucial role in determining the performance of thin-film solar cells.[30] The variation of the cell performance with the thickness of the absorption layer is shown in Fig. 13. When the thickness is too low, it is not beneficial to the full absorption of the light and causes the lower cell PCE. With the increase of the thickness of the absorber, the PCE improves gradually. When the thickness is larger than 600 nm, the PCE of the cell remains invariable. If the absorber layer is too thick, the photogenerated carriers cannot be collected effectively because they must travel through the absorber to reach the carrier collecting layers before they are recombined. The charge diffusion length is critical in designing the perovskite layer thickness and structure. The influences of the thickness of the perovskite absorption layer with different charge diffusion lengths on the solar cell performance are shown in Fig. 14.

Fig. 12. Variations of cell performance, with increasing the diffusion length l.
Fig. 13. Variations of the cell performance with perovskite thickness.
Fig. 14. Variations of PCE with perovskite thickness for different values of diffusion length l of the carriers.

The PCE of the cell gradually increases with the increase of the thickness of the absorber, but when the thickness is larger than 600 nm, the growth of the PCE slows down. Taking into consideration the influence of the Nt and the thickness of the absorber, the performance parameters of devices are optimal when Nt is as low as 1×1015 cm−3 (l of 0.6 μm) and the thickness is 600 nm for the perovskite layer, and reach the Jsc of 30.14 mA/cm2, Voc of 0.79 V, FF of 68.96% and PCE of 16.50% (see line 4 in Fig. 2), resulting in better cell performance than the initial cell performance. This significant performance improvement of the cell with the Nt reduction is attributed to the lower recombination of the carriers and the longer l (see Figs. 11 and 12).

The simulation shows that Nt is an important factor determining the performance of perovskite solar cells, and this result consists with the research about CH3NH3PbI3 perovskite cell.[21] Experimental researchers have obtained the uniform, pinhole-free CH3NH3SnI3 perovskite films from a strongly coordinating solvent such as dimethyl sulfoxide (DMSO).[11] Introduction of pyrazine also can provide beneficial effects for improving the surface morphology and preventing the unwanted Sn oxidation.[31] Besides, by a new variant of pulsed excimer laser deposition: a room-temperature dry process, the growth of good quality lead-free CH3NH3SnI3 films is also demonstrated.[31] Finally, we consider all of the factors (NA, χ, Nt, thickness) discussed above in Subsections 3.1–3.3 (see Table 4) depict current density-voltage curves (see curve 5 in Fig. 2), and obtain encouraging results of the Jsc of 31.59 mA/cm2, Voc of 0.92 V, FF of 79.99% and PCE of 23.36%. Our simulation shows that the lead-free CH3NH3SnI3 perovskite solar cells with high efficiency can be achieved by adjusting the doping concentration and the defect density of the absorption layer, and the electron affinity of each of ETM and HTM. Especially the reduction of the defect density in the perovskite layer can improve the cell performance significantly.

Table 4.

Final optimized parameters of CH3NH3SnI3 PSC.

.

It should be noted that the PCE of the lead-free perovskite solar cell is still lower than that of the reported high-efficiency CH3NH3PbI3 perovskite solar cells in the present experimental studies, and this might be related to the poor quality of perovskite film and the low coverage on the mesoporous TiO2 electrode. The simulation in this paper shows that further efficiency enhancement of the lead-free PSCs would be expected by reducing the defect density and improving the stability of CH3NH3SnI3 in the future.

4. Conclusions

The lead-free CH3NH3SnI3 perovskite solar cells with different parameters are analyzed by using one-dimensional device simulation in this work. The results indicated that the appropriate p-type carrier doping concentration of tin perovskite could improve the PCE of the device because of the enhanced built-in electric field, but excessive concentration can lead to a higher recombination rate and poor cell performance; adjusting the electron affinity of buffer and HTM to provide an appropriate interface barrier of carriers can reduce the interface recombination and improve the PCE of the device to some extent. The defect density of the perovskite is the most critical factor in the simulation for high efficiency of solar cells, and reducing the density concentration as low as 1×1015 cm−3 can significantly increase the PCE of the device from 6.09% to 15.28%. The optimized perovskite thickness of 600 nm can absorb more light and increase the PCE of the device to 16.50%. Encouraging results with PCE of 23.36% are obtained by optimizing all of the factors mentioned above. The results show that the lead-free CH3NH3SnI3 perovskite solar cells have the potential to be highly efficient PSCs. By reducing the defect density and improving the stability of Sn2+ in CH3NH3SnI3through optimizing the fabrication and encapsulation process, further efficiency enhancements of the lead-free CH3NH3SnI3 perovskite solar cells would be expected.

Reference
1Kazim SNazeeruddin M KGratzel MAhmad S 2014 Angew. Chem. Int. Ed 53 2812
2Lotsch B V 2014 Angew. Chem., Int. Ed 53 635
3Wei Z HChen H NYan K YYang S H 2014 Angew. Chem. Int. Ed 53 13239
4Kojima ATeshima KShirai YMiyasaka T 2009 J. Am. Chem. Soc 131 6050
5StoumposC CMalliakas C DKanatzidis M G 2013 Inorg. Chem 52 9019 [6]
6Umari PMosconi EDe Angelis F 2014 Sci. Rep 4 4467
7Chen Q YHuang YHuang P RMa TCao CHe Y 2016 Chin. Phys. 25 027104
8Noel N KStranks S DAbate AWehrenfennig CGuarnera SHaghighirad A ASadhanala AEperon G EJohnston M BPetrozza A MHerz L MSnaitha H J 2014 Energy Environ. Sci 9 3061
9Hao FStoumpos C CCao D HChang R P HKanatzidis M G 2014 Nat. Photon 8 489
10Hao FStoumpos C CGuo PZhou NMarks T JChang R P HKanatzidis M G 2015 J. Am. Chem. Soc 137 11445
11Koh T MKrishnamoorthy TYantara NShi CLeong W LBoix P PGrimsdale A CMhaisalkar S GMathews N 2015 J. Mater. Chem. 3 14996
12Marshall K PWalton R IHatton R A 2015 J. Mater. Chem. 3 11631
13Kumar M HDharani SLeong W LBoix P PPrabhakar R RBaikie TShi CDing HRamesh RAsta MGraetzel MMhaisalkar S GMathews N 2014 Adv. Mater 26 7122
14Lang LYang J HLiu H RXiang H JGong X G 2014 Phys. Lett. 378 290
15Zhu J ZQi L HDu H JChai Y C 2015 Chin. Phys. 24 108501
16Minemoto TMurata M 2015 Sol. Energy Mat. Sol. Cells 133 8
17Kemp K WLabelle A JThon S MIp A HKramer I JHoogland SSargent E H 2013 Adv. Energy Mater 3 917
18Minemoto TMurata M 2014 J. Appl. Phys 116 054505
19Minemoto TMurata M 2014 Curr. Appl. Phys 14 1428
20Liu FZhu JWei JLi YLv MYang SZhang BYao JDai S 2014 Appl. Phys. Lett 104 253508
21Hao FStoumpos C CChang R P HKanatzidis M G 2014 J. Am. Chem. Soc 136 8094
22Scheer R 2009 J. Appl. Phys 105 104505
23Ball J MLee M MHey ASnaith H J 2013 Energ. Environ. Sci 6 1739
24Xing G CMathews NSun S YLim S SLam Y MGrätzel MMhaisalkar SSum T C 2013 Science 342 344
25Hall R N1953Proc. Phys. Soc24221
26Shockley WRead W T 1952 Phys. Rev 87 835
27Kagan C RMitzi D BDimitrakopoulos C D 1999 Science 286 945
28Mitzi D BWang SFeild C AChess C AGuloy A M 1995 Science 267 1473
29Liu D YGangishetty M KKelly T L 2014 J. Mater. Chem. 2 19873
30Lee S JShin S SKim Y CKim DAhn T KNoh J HSeo JSeok S I 2016 J. Am. Chem. Soc 138 3974
31Bansode UNaphade RGame OAgarkar SOgale S 2015 J. Phys. Chem. 119 9177