Thin single crystal perovskite solar cells to harvest below-bandgap light absorption — TEI-tagged view

Zhaolai ChenpersName, Qingfeng DongpersName, Ye LiupersName, Chunxiong BaopersName, Yanjun FangpersName, Yun LinpersName, Shi TangpersName, Qi WangpersName, Xun XiaopersName, Yang BaipersName, Yehao DengpersName, Jinsong HuangpersName (2017date). “Thin single crystal perovskite solar cells to harvest below-bandgap light absorption.” Nature Communicationstitle·j.

Object of study — thin methylammonium lead triiodide single crystalsterm·key grown on hole-transport-layer covered substrates

Research question — how to get past the efficiency bottleneck set by the non-optimized bandgap of perovskiteterm·key absorbers

Contribution — a new direction to boost the efficiency and stability of perovskiteterm·key solar cellsterm·key

Full text — tagged

Thin single crystalterm·key perovskiteterm·key solar cellsterm·key to harvest below-bandgap light absorption Abstract The efficiency of perovskiteterm·key solar cellsterm·key has surged in the past few years, while the bandgaps of current perovskiteterm·key materials for record efficiencies are much larger than the optimal value, which makes the efficiency far lower than the Shockley–Queisser efficiency limit. Here we show that utilizing the below-bandgap absorptionterm·dom of perovskiteterm·key single crystalsterm·key can narrow down their effective optical bandgap without changing the composition. Thin methylammonium lead triiodide single crystalsterm·key with tuned thickness of tens of micrometers are directly grown on hole-transport-layer covered substrates by a hydrophobic interface confined lateral crystal growth method. The spectral response of the methylammonium lead triiodide single crystal solar cellsterm·dom is extended to 820 nm, 20 nm broader than the corresponding polycrystalline thin-film solar cellsterm·dom. The open-circuit voltage and fill factor are not sacrificed, resulting in an efficiency of 17.8% for single crystalterm·key perovskiteterm·key solar cellsterm·key.

Introduction Organic–inorganic halide perovskitesterm·dom (OIHPsterm·dom) have attracted tremendous attentions for solar cellterm·key application in the past few years 1 – 9 due to the superior optoelectronic properties, such as large extinction coefficient 10 , long carrier recombination lifetime 11 , large carrier mobility 12 , unique defect physics 13 , etc. The efficiency of OIHPterm·dom solar cellsterm·key has reached 22.1% quickly after development in a few years by the improvement of material quality and device interface 9 , but it is facing a bottleneck caused by the non-optimized bandgap of the current perovskiteterm·key materials. The most widely studied perovskiteterm·key solar cellsterm·key are based on CH 3 NH 3 PbI 3 (MAPbI 3term·key ), which has a bandgap too large for single-junction solar cellsterm·key 14 . To reduce the bandgap of MAPbI 3term·key , composition engineeringterm·mth has been developed by partially replacing MA + and Pb 2+ with FA + and Sn 2+ , respectively 15 – 18 , which however brings new challenges. The easy oxidation of Sn 2+ to Sn 4+ causes additional severe chemical instability of the Sn-containing perovskiteterm·key 17 . Introduction of excessive FA + ions can cause the instability of material structure 16 .

Here we exploit an approach to broaden the absorption spectrum of perovskitesterm·key by utilizing their below-bandgap transition. Below-bandgap absorptionterm·dom has been identified in MAPbI 3term·key in previous studies, which was attributed to the indirect-bandgap absorption transition with a bandgap of ~60 meV smaller than the direct bandgap 19 – 22 . We found other perovskiteterm·key materials, such as MAPbBr 3 , also showed below-gap absorption. Similar to silicon single crystal solar cellsterm·dom, further improvement of the device efficiency may be achieved by utilizing the below-bandgap absorptionterm·dom. However, the absorption coefficient corresponding to the below-bandgap transition is several orders magnitude smaller than that of the above-gap transition. Therefore, the thickness of the perovskiteterm·key films thus needs to be over several micrometers (µm) or more, which is however beyond the carrier diffusion lengthterm·dom in existing polycrystalline perovskiteterm·key films. Perovskiteterm·key single crystalsterm·key are shown to have much longer carrier diffusion lengthterm·dom well above tens of micrometer due to the absence of grain boundaries and significantly reduced defect density 23 – 26 . Current studies of the intrinsic properties of the perovskitesterm·key are mainly based on thick bulk single crystalsterm·key with thickness of millimeter 27 , 28 which are too thick for application in solar cellsterm·key. Growing thin perovskiteterm·key single crystalsterm·key with large area represents an effective approach to overcome this obstacle 29 – 31 , however there is no effective method handling the micrometer-thick iodide-based perovskiteterm·key single crystalsterm·key with large area. Recently, LiupersName et al. 32 reported an efficiency of about 4% for perovskiteterm·key single crystal solar cellsterm·dom. The low efficiency was due to the too large thickness of 50 μm for the perovskiteterm·key crystal and non-passivated surface defects.

Herein, we report the growth of perovskiteterm·key single crystalsterm·key directly onto hole transport layerterm·dom (HTLterm·dom) covered indium tin oxide (ITO) substrates with controlled thickness of tens of micrometer and area of tens of millimeter square (mm 2 ). Solar cellsterm·key based on the thin single crystalsterm·key show obviously broader spectral response compared to the polycrystalline thin-film solar cellsterm·dom, while the open-circuit voltage ( V OCterm·dom ) and fill factor (FF) of the solar cellsterm·key remain comparable to those of polycrystalline thin-film solar cellsterm·dom, which demonstrates the potential for the application of perovskiteterm·key single crystalsterm·key to further boost the efficiency of perovskiteterm·key solar cellsterm·key.

Results Thickness-dependent device performance We first simulate the thickness-dependent absorption of MAPbI 3term·key and device efficiency to find out the feasibility of increasing device efficiency by utilizing the below-bandgap transition in MAPbI 3term·key and the details are shown in the “Method” section. Figure 1a shows an absorption coefficient curve of MAPbI 3term·key from literature 33 , which does show two–four orders magnitude weaker absorption for below-bandgap transition than direct transition. Figure 1b shows the calculated thickness-dependent percentage of light absorbed by the MAPbI 3term·key films in devices where a full reflection of the light by a metal back electrode is assumed. When the thickness of MAPbI 3term·key films increases from 500 nm to 200 µm, the absorption spectrum expands significantly to near-infrared region, leading to the red-shift of the absorption edge to 850 nm. The enhanced absorption in this wavelength range is important to increase the short-circuit current density ( J SC ), because sunlight has a significant portion of energy in this wavelength range. As shown in Fig. 1c , as the thickness of perovskiteterm·key crystals increases from 500 nm to 200 µm, ( J SC ) increases from 23.0 mA cm −2 to 27.1 mA cm −2 due to the broadened absorption spectrum. The V OCterm·dom of the devices, however, has an opposite variation trend with respect to the thickness of absorber based on the diode equation: where T , k , q , and, J 0 are the absolute temperature, Boltzmann constant, elementary charge, and reverse bias saturated dark current density, respectively. The details for the simulation of crystal thickness-dependent efficiency limit of the single crystal solar cellsterm·dom can be found in the Supplementary Note. Though increased J SC slightly increases V OCterm·dom , larger thickness of perovskiteterm·key films increases the charge recombinationterm·dom during their transport and thus J 0 , which reduces V OCterm·dom . The charge recombinationterm·dom dominates the variation of V OCterm·dom with a clear reduction with increasing crystal thickness, as shown in Fig. 1c . There should be an optimal absorber thickness that the perovskiteterm·key single crystal solar cellsterm·dom reach the highest efficiency. To determine the optimal thickness of perovskiteterm·key single crystalsterm·key, we assumed a constant FF of 80%, 90% external quantum efficiencyterm·mth (EQEterm·mth) over the whole absorption spectrum, which are all reasonable and are justified in the “Method” section. The power conversion efficiencyterm·mth (PCEterm·mth) increases first and then decreases with increasing the thickness of single crystalsterm·key, resulting in an optimal crystal thickness of 200 μm. When the crystal thickness is over 10 μm, the PCEterm·mth is higher than 23.5%, which is far beyond the state-of-the-art values of MAPbI 3term·key polycrystalline thin-film solar cellsterm·dom.

Thickness-dependent perovskiteterm·key absorption and device performance. a Schematic illustration of the direct, below-bandgap transitions and absorption coefficient of methylammonium lead triiodide (MAPbI 3term·key ) from a polycrystalline film. b The calculated absorption of MAPbI 3term·key films with different thickness. c The calculated ideal dependence of J sc , V octerm·dom , and PCEterm·mth of the single crystal solar cellsterm·dom on the thickness of the thin single crystalsterm·key Growth and characterization of thin crystals A diffusion-facilitated space-confined method was developed to obtain perovskiteterm·key single crystalsterm·key with controllable thickness down to micrometer scale. The confined space was constructed by two substrates (Fig. 2a ), and the supersaturation of precursor solution was generated by the inverse solubility method 23 . The space-confined method was reported to grow thin perovskiteterm·key singe crystals 31 , however, the crystal thickness could hardly go down to 100 µm for large-area perovskiteterm·key crystals. The lateral crystal growth in this method can be limited by the inefficient long-range transportation of precursor ions along micrometer-size gap to continuously replenish the depleted precursors by crystal growth, which results in the formation of many small crystals or even polycrystals, as illustrated in Fig. 2d . When the gap is down to micrometer scale, the interaction between substrate surface and solvent in precursor solution becomes critically important for ion diffusion in solvent. Since the perovskiteterm·key precursor ions generally form complex with the solvent 34 , the diffusion rate of precursor ions is determined by how fast the solvent molecules can transport on the substrate, and the surface tension determines how fast the solvent molecules close to the substrate can move along in-plane direction 35 – 37 . On wetting substrates, such as glass or ITO, the large surface tension imposes a friction force that drags down the ion diffusion 38 , as illustrated in Fig. 2e . Interestingly, when the substrates are covered with hydrophobic materials, such as PTAAterm·key 39 , which is also a hole transport material for perovskiteterm·key solar cellsterm·key, the ion diffusion is dramatically accelerated. The dramatically enhanced ion diffusion rate in micrometer-size gap enables a continuous growth of thin perovskiteterm·key single crystalsterm·key up to several millimeters along in-plane direction, as illustrated in Fig. 2c . To illustrate the dramatically enhanced ion diffusion rate by the surface modification, a drop of MAPbI 3term·key precursor solution was placed at the end of two regular glass substrates (left in Fig. 2g ) and two ITO substrates covered by PTAAterm·key (right in Fig. 2g ), which had a same gap of 10 μm filled with pure γ-butyrolactone (GBL) solvent, and they were hold at the temperature for crystal growth (110 °C) or room temperature. Driven by the concentration gradient, the precursor ions diffused in the GBL solution from the bottom side to the topside, and the diffusion of ions could be visualized by the color change of the solution with diffusion rate estimated by motion of front edge at the interface of pure GBL and perovskiteterm·key precursor solution. As expected, the edge propagation was very slow (0.01 mm s −1 ) in the gap of two glass substrates, while was at least 100-fold faster in the gap of two PTAA-covered ITO substrates (see Supplementary Video 1 ). We speculate that the much faster ion diffusion can be ascribed to the excellent lubrication effect of the nonwetting PTAAterm·key surface 38 . The nonwetting surface weakens the solvent–substrate interaction at the substrate surface, which will promote the diffusion of solvent. Therefore, the ions that form complexes with GBL molecules would not be dragged during the diffusion, as illustrated in Fig. 2f .

Growth mechanism of perovskiteterm·key thin single crystalsterm·key. a – d Schematic illustrations of correlation between ion diffusion and thin single-crystal growth. Schematic illustrations of ion diffusion rate in the confined gaps using e hydrophilic and f hydrophobic substrates. g Photographs of the diffusion process of methylammonium lead triiodide (MAPbI 3term·key ) precursor solution in the confined gaps after different durations using hydrophilic glass and hydrophobic PTAA-covered ITO substrates The fast ion transportation in the gap between two PTAA-covered substrates enables continuous growth of thin perovskiteterm·key single crystalsterm·key to a large area. The thin single crystalsterm·key were grown on a PTAA-covered ITO substrate and the other PTAA-covered ITO substrate could be easily peeled off. The method is universally applicable to both iodide- and bromide-based perovskitesterm·key. Figure 3a, b shows the photographs of as-grown millimeter-size MAPbI 3term·key and MAPbBr 3 single crystalsterm·key with thickness of 10–20 μm. As shown in Fig. 3c–e , thin MAPbI 3term·key single crystalsterm·key are obtained with tuned thickness from 10 to 40 μm. For single crystalsterm·key with thickness smaller than 10 μm, small-area multicrystals with lots of cracks were usually formed, which led to poor PCEterm·mth for solar cellsterm·key. The X-ray diffraction (XRD) pattern of a thin MAPbI 3term·key single crystalterm·key in Fig. 3f shows only two sharp diffraction peaks which can be assigned to (200) and (400) planes, indicating that the crystal is (100) oriented. The corresponding rocking curve displays a small full-width at half-maximum of 0.33° for the (400) plane (Supplementary Fig. 1a ), confirming the good quality of the thin single crystalterm·key. The 2D XRD shown in Supplementary Fig. 1b consists of scattered diffraction spots rather than a continuous arc, proving the single-crystal character of the grown materials. XRD pattern of the powders scratched from the thin single crystalsterm·key (inset of Fig. 3f ) confirms the tetragonal structure of these crystals at room temperature. The top-view SEM image of the thin single crystalterm·key (Supplementary Fig. 2 ) shows that the surface of the thin single crystalterm·key is smooth. Figure 3g shows the single-path absorption spectra of a single crystalterm·key with thickness of 10 μm and a polycrystalline thin film with thickness of 500 nm, which manifests an obvious redshift of the absorption spectrum edge from 790 to 810 nm. The redshift of the absorption spectrum is general for other kinds of perovskitesterm·key, such as MAPbBr 3 , Cs 0.05 (FA 0.85 MA 0.15 ) 0.95 Pb(I 0.85 Br 0.15 ) 3 (Supplementary Fig. 3 ) and FAPbI 3 24 . The hole and electron mobility of the thin single crystalsterm·key measured by the space charge limit current (SCLC) method (Supplementary Fig. 4 ) are 121 ± 15 cm 2 V −1 s −1 and 36.8 ± 3.7 cm 2 V −1 s −1 , which are comparable to the values of bulk single crystalsterm·key, indicating the good quality of the as-grown thin single crystalsterm·key. The carrier recombination lifetime was measured by the transient photovoltaic (TPV) measurement under one-sun illumination (Fig. 4b ). The obtained carrier lifetime is 2.2 ± 0.5 μs, which is shorter than that of bulk single crystalsterm·key measured by the same method, because the smaller volume of the thin single crystalsterm·key increases the carrier density, thus increasing the bimolecular recombination. In addition, carriers in the thinner single crystalsterm·key are more susceptible to surface recombination by a quicker transit to the crystal surface. Combining the measured mobility and lifetime, the hole diffusion length and electron diffusion length under one-sun light are calculated to be 24.7 and 13.6 μm, respectively.

Characterization of perovskiteterm·key thin single crystalsterm·key. Photographs of a methylammonium lead triiodide (MAPbI 3term·key ) thin single crystalterm·key and b methylammonium lead tribromide (MAPbBr 3 ) thin single crystalterm·key using the hydrophobic interface confined lateral growth method. Cross-sectional SEM images of the MAPbI 3term·key thin single crystalsterm·key with different thickness: c ≈10 μm, d ≈20 μm, e ≈40 μm. f X-ray diffraction patterns of a MAPbI 3term·key thin single crystalterm·key and the grounded powder (inset). g Absorption spectra of a 10-μm-thick MAPbI 3term·key thin single crystalterm·key and a 500-nm-thick polycrystalline film Characterization of perovskiteterm·key single-crystal solar cellsterm·key. a Device structure of the single-crystal solar cellsterm·key. b Transient photovoltaic curve of a single-crystal solar cellterm·key under one-sun illumination. c Current density–voltage and d external quantum efficiencyterm·mth (EQEterm·mth) curves and integrated current density of the optimal single-crystal solar cellsterm·key using methylammonium lead triiodide thin single crystalterm·key with different thickness. e EQEterm·mth of single-crystal solar cellsterm·key and polycrystalline thin-film solar cellsterm·dom. f Trap density of states of the single-crystal solar cellsterm·key before and after methylammonium iodide (MAI) treatment. g Current density–voltage curves of the single-crystal solar cellsterm·key after MAI treatmentterm·mth measured under different scanning directions. h Stabilized J sc and PCEterm·mth of single-crystal solar cellsterm·key with MAI treatmentterm·mth. The thicknesses of the single crystalsterm·key in Fig. e – h and polycrystalline thin film are 10 and 500 nm, respectively Perovskiteterm·key single-crystal solar cellsterm·key Single-crystal solar cellsterm·key with device structure of ITO/PTAAterm·key/MAPbI 3term·key /phenyl-C 61 -butyric acid methyl ester (PCBM)/C 60 /bathocuproine (BCP)/copper (Cu) (Fig. 4a ) were fabricated by depositing electron transport layers (ETLs) and metal electrodes on the top of the single crystalsterm·key with the easy handling of ITO substrates. Figure 4c , d shows the thickness-dependent J–V and EQEterm·mth curves of the single-crystal solar cellsterm·key. The highest device efficiency is obtained when using single crystalsterm·key with thickness of 10 μm, which is consistent with the measured carrier diffusion lengthterm·dom. The best device has a J SC of 20.5 mA cm −2 , a V OCterm·dom of 1.06 V, a FF of 74.1%, and a PCEterm·mth of 16.1%. As shown in Fig. 4e , the EQEterm·mth spectrum of this device shows an EQEterm·mth cutoff at 820 nm, corresponding to 20-nm redshift compared to that of the polycrystalline thin-film devices. The larger redshift of EQEterm·mth spectrum than the absorption spectrum should be attributed to the reflection of light by the electrode, which doubles the effective crystal thickness for light absorbing. The statistics of PCEterm·mth, FF, J SC , and V OCterm·dom for devices using single crystalsterm·key with thickness of 10 μm are shown in Supplementary Fig. 6 .

It is noted that thin single crystalsterm·key may still have large surface charge traps 40 , though the bulk trap density should be very small. The surface charge traps are most likely generated from the loss of MAI when removing the single crystalsterm·key out of the hot precursor solution 41 . Therefore, we treated the surface of the single crystalsterm·key by spin coating MAI solution to reduce the surface charge trap density. This surface passivation effect was confirmed by the enhanced photoluminescence intensity (Supplementary Fig. 7 ). As shown in Fig. 4f , the surface trap densityterm·dom of the thin single crystalsterm·key measured by thermal admittance spectroscopy is reduced by 2–5-folds after MAI treatmentterm·mth. The PCEterm·mth of the best device also increases to 17.8% with a J SC of 21.0 mA cm −2 , a V OCterm·dom of 1.08 V, and a FF of 78.6% (Fig. 4g ). The PCEterm·mth of the device is confirmed by measuring the steady-state photocurrent output at the maximum power point (Fig. 4h ). The increase of J SC is consistent with the increase of the EQEterm·mth values (Supplementary Fig. 8 ), and matches with the calculated J SC from EQEterm·mth. The corresponding statistics of PCEterm·mth, FF, J SC , and V OCterm·dom are shown in Supplementary Fig. 9 . By comparing with the calculated results in Fig. 1c , the V OCterm·dom is close to its predicted value of 1.14 V, and the FF is close to what is observed in high-efficiency polycrystalline perovskiteterm·key solar cellsterm·key. The main limiting parameter of the present single-crystal solar cellsterm·key is the smaller J SC than the predicted value of 25.8 mA cm −2 , which may be caused by the enhanced light reflection on the much flatter single-crystal surface (Supplementary Fig. 10a ), in addition to charge collection loss caused by the incompletely passivated surface defects on both surfaces of the single crystalsterm·key (Supplementary Fig. 10b ). Addressing these issues will have potential to further increase the efficiency.

Finally, we evaluate the stability of the single-crystal solar cellsterm·key. We have demonstrated that bulk single crystalsterm·key showed better moisture stability than the polycrystalline thin films due to the absence of grain boundaries 41 . Therefore, the air stability of the single-crystal solar cellsterm·key should be better than that of polycrystalline thin-film solar cellsterm·dom. To confirm this, the shell lifetime of the single-crystal solar cellsterm·key in air is investigated. As shown in Supplementary Fig. 11 , the PCEterm·mth shows almost no change after the device was stored in air under dark for 30 days, which is better than the polycrystalline thin-film solar cellsterm·dom.

Conclusion In conclusion, we demonstrated the use of a single crystalterm·key to broaden the photoresponse range of the perovskiteterm·key solar cellsterm·key without losing device photovoltage and fill factor. MAPbI 3term·key thin single crystalsterm·key were grown by a hydrophobic interface-confined lateral growth method, which yielded well control of crystal thickness at tens of micrometers. This work presents a new direction to further boost the efficiency and stability of perovskiteterm·key solar cellsterm·key. More efforts are needed to enhance the carrier diffusion lengthterm·dom and passivate the surface nonradiative recombination center density in order to boost the device efficiency to SQ limit. The extended absorption can be attributed to the indirect bandgap absorption or Urbach band tail. Though there is some study that MAPbI 3term·key has signature of indirect bandgap, there is no study whether the same conclusion can apply to MAPbBr 3 or other perovskiteterm·key materials yet. This study shows that these charges can actually be extracted out without being trapped.

Methods Growth of perovskiteterm·key thin single crystalsterm·key The perovskiteterm·key thin single crystalsterm·key were grown by a hydrophobic interface-confined lateral crystal growth method, as shown in Fig. 2 . The PTAAterm·key film was prepared by spin coating 0.2 wt% PTAAterm·key solution in toluene at 4000 r.p.m. The as-prepared film was then thermally annealed at 100 °C for 10 min. As illustrated in Fig. 2a , 1.5 M MAPbI 3term·key solution in GBL was inserted into two ITO/PTAAterm·key substrates, which were then placed on a hot plate with temperature of 100 °C for 1 h. Subsequently, the temperature was elevated to 110 and 120 °C to further promote the growth of the single crystalsterm·key. Finally, the two substrates were separated to obtain the thin single crystalsterm·key on ITO/PTAAterm·key substrates. The temperature was slowly decreased to room temperature (about 2 h) to ensure a good contact between PTAAterm·key and the single crystalsterm·key.

Device fabrication The device fabrication was completed by depositing the electron transport layer and electrode. PCBM solution of 1 wt% in chlorobenzene was spun on top of the thin single crystalterm·key and then annealed at 50 °C for 10 min. After that, C60 (10 nm), BCP (3 nm), and Cu (80 nm) were thermally evaporated in sequential order. The C60 and BCP were evaporated on the whole area of the surface of the single crystalterm·key. The Cu electrode was evaporated with a mask by covering the edge of the single crystalterm·key to avoid shorts. For MAI treatmentterm·mth, 0.5 mg ml −1 MAI in IPA was spin coated on the single crystalterm·key before depositing PCBM and then annealed at 50 °C for 10 min.

Characterization The absorption spectra were recorded by using an Evolution 201 UV-Visible Spectrophotometer. First, the transmission spectra of the thin single crystalterm·key and polycrystalline thin film were measured. Subsequently, the corresponding absorption spectra were calculated as Absorptance = 1-Transmittance. The reflection spectrum was measured with the integrating sphere. The photocurrent density–voltage (J–V) curves of the devices were measured under AM 1.5-G irradiation (100 mW cm −2 ), which was produced by a xenon-lamp-based solar simulator (Oriel 67005, 150-W Solar Simulator). The light intensity was calibrated by a Schott visible-color glass-filtered (KG5 color-filtered) Si diode (Hamamatsu S1133) before photocurrent measurement. The J–V testing was performed with reverse scan direction at 0.1 V s −1 and sweep delay time was 10 ms. The EQEterm·mth was obtained using a Newport QE measurement kit by focusing a monochromatic beam of light onto the devices. The X-ray diffraction patterns were obtained by a Rigaku D/Max-B X-ray diffractometer in the Bragg–Brentano parafocusing geometry. A conventional cobalt target X-ray tube equipped in the diffracted-beam monochromator was set to 40 kV and 30 mA. The SEM images were taken from a Quanta 200 FEG environmental scanning electron microscope. In the TPV measurement, the single-crystal solar cellsterm·key were illuminated by laser pulses (337 nm, 4-ns width from a SRS nitrogen laser) from the ITO side to generate a transient photovoltage signal which functions as a small perturbation to the background open-circuit voltage. The transient voltage signal was recorded by a 1-GHz Agilent digital oscilloscope.

Data availability The data that support the findings of this study are available from the corresponding author on reasonable request.

Electronic supplementary material Supplementary Information Acknowledgements The authors gratefully acknowledge financial support from Defense Threat Reduction Agency (HDTRA1-14-1-0030), Office of Naval Research (N00014-17-1-2163 and N00014-15-1-2713), and National Science Foundation (ECCS-1252623).

Author contributions J.H. conceived the idea and supervised the project; Z.C., Q.D., and Y.L. grew the thin single crystalterm·key and fabricated the devices; C.B. carried out the tDOS measurement; Y.F. carried out the PL measurement; Y.L. fabricated the polycrystalline film solar cellsterm·key; S.T. carried out the rocking curve measurement; Q.W. carried out the SEM measurement; X.X. carried out the TPV measurement; and Y.B. carried out the air stability test of the polycrystalline film solar cellsterm·key. Y.D. carried out the XRD measurement. J.H., Z.C., and Q.D. wrote the manuscript, and all authors reviewed it.

Competing interests The authors declare no competing financial interests.

Footnotes References Associated Data Supplementary Materials Supplementary Information Data Availability Statement The data that support the findings of this study are available from the corresponding author on reasonable request.