Cite this Article
Liu Lu, Zhang Sheng-Li, Wu Jian-Yu, Wang Wei-Huang, Liu Wei, Wu Li, Zhang Yi. Current improvement in substrate structured Sb2S3 solar cells with MoSe2 interlayer. Chinese Physics B, 2020, 29(5): 058801  
Permissions
Current improvement in substrate structured Sb2S3 solar cells with MoSe2 interlayer
Liu Lu1, Zhang Sheng-Li1, ‡, Wu Jian-Yu1, Wang Wei-Huang1, Liu Wei1, Wu Li2, Zhang Yi1, †
Institute of Photoelectronic Thin Film Devices and Technology and Tianjin Key Laboratory of Thin Film Devices and Technology, Nankai University, Tianjin 300350, China
School of Physical Science, Nankai University, Tianjin 300071, China

 

† Corresponding author. E-mail: yizhang@nankai.edu.cn shenglizhang@nankai.edu.cn

Project supported by the National Key R & D Program of China (Grant Nos. 2019YFB1503500, 2018YFE0203400, and 2018YFB1500200), the National Natural Science Foundation of China (Grant No. U1902218), the YangFan Innovative and Entrepreneurial Research Team Project of China (Grant No. 2014YT02N037), and the 111 Project, China (Grant No. B16027).

Abstract

Sb2S3 solar cells with substrate structure usually suffer from pretty low short circuit current (JSC) due to the defects and poor carrier transport. The Sb2S3, as a one-dimensional material, exhibits orientation-dependent carrier transport property. In this work, a thin MoSe2 layer is directly synthesized on the Mo substrate followed by depositing the Sb2S3 thin film. The x-ray diffraction (XRD) patterns confirm that a thin MoSe2 layer can improve the crystallization of the Sb2S3 film and induce (hk1) orientations, which can provide more carrier transport channels. Kelvin probe force microscopy (KPFM) results suggest that this modified Sb2S3 film has a benign surface with less defects and dangling bonds. The variation of the surface potential of Sb2S3 indicates a much more efficient carrier separation. Consequently, the power conversion efficiency (PCE) of the substrate structured Sb2S3 thin film solar cell is improved from 1.36 % to 1.86 %, which is the best efficiency of the substrate structured Sb2S3 thin film solar cell, and JSC significantly increases to 13.6 mA/cm2. According to the external quantum efficiency (EQE) and C–V measurements, the modified crystallization and elevated built-in electric field are the main causes.

PACS: ;88.40.H-;;88.40.hj;
Keyword:;Sb2S3 thin film;;super thin MoSe2;;built-in electric filed;;Kelvin probe force microscopy;
1. Introduction

Recently, Sb2S3 has received great attention for the application in solar cell. It possesses many good properties, such as a proper bandgap of ∼ 1.7 eV, high light absorption coefficient ( > 105 cm–1), earth abundance, nontoxic source, etc.[1,2] Sb2S3, as a binary semiconductor compound, can keep a single phase and avoid forming the unfavorable secondary phase at high temperature,[3] which makes it a good alternative to Cu(In,Ga)Se2 (CIGS), Cu2ZnSnS4 (CZTS), and CdTe absorber layers. In particular, the large bandgap provides a promising utilization for tandem dual-junction solar cells.[4]

Up to now, the highest efficient (7.5 %) Sb2S3 solar cell is achieved by the semiconductor-sensitized device structure with glass/TCO/TiO2/Sb2S3/polymer.[5] As sensitized solar cells inevitably suffer from interface degradation and device instability. To improve the device stability, much more attention is shifted to the planar structured Sb2S3 solar cell. A few techniques have been used to synthesize the Sb2S3 thin films, like chemical bath deposition,[6] solution spin coating,[7] pulse electrodeposition,[8] sputtering with post-annealing,[9] and thermal evaporation.[10] Owing to the low melting point of Sb2S3 (∼ 550°C), it can be synthesized at low temperatures (∼ 350 °C) with good crystalline quality in a short time.[11] Hence, for the planar heterojunction Sb2S3 thin film solar cells, rapid thermal evaporation (RTE) is a promising approach to implement the large scale industrial preparation. At present, most of Sb2S3 solar cells are based on the superstrate-structured solar cell: glass/FTO/TiO2(or CdS)/Sb2S3/Au. In 2016, Yuan et al. fabricated Sb2S3 thin film solar cells with large grains and preferential growth by optimizing the rapid thermal evaporation and cooling techniques, achieving a power conversion efficiency (PCE) of 3.5 %.[11] As a one-dimensional ribbon material, the intra-ribbons are dominated by the covalent bonds, and the ribbons are bonded with each other by Van der Waals force. Thus, Sb2S3 exhibits strong anisotropy in optical and electrical properties.[12] Theoretically, most of Sb2S3 based devices are orientation-dependent. For this concern, Deng et al. successfully obtained vertical (hk1) orientated Sb2S3 by engineering TiO2 exposure facets, which significantly improved the interface quality and intra-ribbon transport.[13] As reported previously, different substrates affect both the morphology and the crystal structure orientation of Sb2S3 grains.[14] The orientation of Sb2S3 can be tailored by modifying the substrate. The Sb2S3 solar cells with superstrate structure show good performances, but it is difficult to further improve the PCEs due to their simple device structures. For the purpose of thorough scientific study, substrate configuration of glass/Mo/Sb2S3/CdS/i-ZnO/Al-ZnO/Ni:Al is beneficial to engineer the junction interface and improve the spectral matching, which may have potential for the flexible and tandem application. In 2017, Zhang et al. prepared Sb2S3 thin film on Mo substrate by evaporating metallic Sb and then annealing in N2/H2S atmosphere. The best PCE of the resulting solar cells was 0.65 %.[15] Later, Pan et al. systematically studied the effect of the substrate temperature and evaporation time on Sb2S3 thin films deposited by rapid thermal evaporation (RTE), and the PCE was increased up to 1.75 %.[16] As mentioned in many reports, the Sb2S3 film suffers from the parallel orientation (hk0) to the substrate, especially on the Mo substrate, resulting in an unsatisfying carrier transport and large series resistance, which is detrimental for device performance.[10,11,15–18] To induce (hk1) preferred orientations, an extra MoSe2 layer is taken into consideration. It is reported that the MoSe2 possesses two-dimensional semiconductor property. Each monolayer is bonded by covalence in the order of Se–Mo–Se and all layers are stacked by weak Van der Waals interaction.[19] Wu et al. demonstrated that a few layers of MoSe2 exhibit good electrical conductivity.[20] Thereby, a MoSe2/Mo substrate is supposed to be a good alternative.

In this work, a MoSe2 layer is applied to the Sb2S3 thin film solar cell with the substrate configuration of glass/Mo(MoSe2)/Sb2S3/CdS/iZO/AZO/Ni:Al. Systematic characterizations are employed to reveal the influence of the thin MoSe2 layer on the Sb2S3 thin film growth and resulting solar cell. Finally, the PCE of the Sb2S3 thin film solar cell is increased from 1.36 % to 1.86 %, especially, the short circuit current JSC rises from 7.11 mA/cm2 to 13.6 mA/cm2.

2. Experimental details
2.1. Fabrication

A two-layer structured Mo film was deposited on soda-lime glass by DC-magnetron sputtering. A pre-selenization process was carried out by chemical vapor deposition (CVD) in a quartz tube with two heating zones. The 300 mg selenium powder (99.999 % purity, Alfa Aesar) and Mo substrate were located in the center of each hot zone with temperatures of 250 °C and 550 °C, respectively. The reaction was lasted for 10 min with 20 sccm Ar carrying gas flow.[20] After that, the Sb2S3 film was prepared by rapid thermal evaporation in a single temperature tube furnace, the substrate was maintained at 310 °C for 15 min, then rapidly rose up to 530 °C in 30 s, the evaporation duration was 100 s followed by natural cooling.[16]

Sb2S3 thin film solar cells were fabricated with substrate structure of glass/Mo (MoSe2)/Sb2S3/CdS/iZO/AZO/Ni:Al. The CdS buffer layer was deposited on the Sb2S3 absorber by chemical bath deposition (CBD) at 75 °C for 9 min followed by sputtering 50 nm intrinsic ZnO and 500 nm ZnO:Al as the widow layer. Finally, the front electrode Ni:Al grid was prepared by electron beam evaporation, and the active area of each Sb2S3 solar cell was 0.18 cm2.

2.2. Characterization

The crystal structure of the Sb2S3 film was characterized by x-ray diffraction (XRD) with Cu Kα radiation source (PANalytical X’pert pro). The lattice vibration of MoSe2 was analyzed by Raman spectrum at room temperature with 532 nm excitation wavelength (Horiba JobinYvon, LabRAM HR800). And the surface and cross-sectional morphologies of the Sb2S3 thin films were measured by scanning electric microscopy (SEM, Hitachi S-4800). The atomic force microscopy (AFM) (Dimension Icon, Bruker) was used to characterize the surface topography and thickness of the MoSe2 film with an area of 5μm2. Kelvin probe force microscopy (KPFM) was used to detect the contact potential difference of the films, which was equipped with a Pt coated Si tip with 6 nm radius of curvature (HYDRA6R-100NG-10, APPNABQ) as the probe in a noncontact mode, and the AC voltage was adjusted by the instrument according to the films. The ultraviolet photoelectron spectroscopy (UPS; Kratos, UK) was conducted to detect the work function of the Sb2S3 and molybdenum electrode by using He I radiation (21.2 eV) as the excitation source. The work function of the film can be calculated by Φ = hν – (Ecutoff – EF). The current density–voltage (J–V) characteristics of the Sb2S3 thin film solar cells were measured by a solar simulator (SAN-EI XES-500T1), which was calibrated by a standard single crystal Si solar cell, under the standard AM1.5 spectrum with an illumination intensity of 1000 W/m2 at room temperature. The capacitance–voltage (C–V) measurement was performed with a HP 4284A LCR meter in darkness at room temperature, the frequency and bias parameter were 100 KHz and 30 mV, respectively. The external quantum efficiency (EQE) of the Sb2S3 thin film solar cells was measured by Beijing Zolix Solar Cell Scan 100, which measured the ratio of short circuit photocurrent to incident illumination intensity in a wavelength range of 300–1300 nm.

3. Results and discussion
3.1. Growth of MoSe2 and Sb2S3 thin films

In this work, a bare Mo substrate is exposed to the Se atmosphere for 10 min to form MoSe2 first. The Ar flow is used as the carrying gas during this process. Figure 1(a) shows the Raman spectrum of this pre-selenized substrate, the peaks at 167.6 cm–1, 239.2 cm–1, and 286.3 cm–1 are attributed to E1g, A1g, and E2g modes of the MoSe2 film, respectively.[20,21] With the help of AFM, the thickness of the MoSe2 layer is measured to be about 8 nm (Fig. 1(b)). From Figs. 1(c) and 1(d), the bare Mo substrate has large grains with distinct shapes and the surface Ra (arithmetic average of absolute roughness) is 3.26 nm. For MoSe2 covering on the Mo film, however, the surface Ra increases up to 4.91 nm, and grains' topography becomes much higher and steeper, which indicates that MoSe2 can obviously improve the roughness of the Mo substrate and refine the surface microstructure.

Fig. 1. (a) Raman spectrum of MoSe2. (b) AFM image of MoSe2 atomic layer, the inset shows MoSe2 thickness of around 8 nm, and the step between Mo (dark layer in the center) and MoSe2 was created by a blade. AFM images of (c) Mo surface and (d) MoSe2, their surface roughnesses (Ra) are 3.26 nm and 4.91 nm, respectively.

Then the Sb2S3 absorbers are deposited on the Mo and MoSe2/Mo substrates by RTE. Figure 2(a) displays the XRD patterns of the Sb2S3 films. All the diffraction peaks of these two thin films can be indexed to the orthorhombic Sb2S3 phase (space group Pbnm, #JCPDS 42–1393) without any impurity phase. Besides, as shown by the Raman spectrum of Sb2S3 on MoSe2 substrate in the inset of Fig. 2(a), the peaks at 189.7 cm–1 and 208.3 cm–1 are assigned to the B1g mode of S–Sb–S vibration and the peak at 251.4 cm–1 corresponds to the A1g mode of S–Sb–S vibration, suggesting a pure phase of Sb2S3. Clearly, the XRD results of both samples show good crystallization, but their orientations differ due to the difference in their substrates. The Sb2S3 on the Mo substrate prefers (120) orientation, then moves to (310) due to the introduction of MoSe2. To find out more about the variations of orientation between the Sb2S3 films deposited on Mo and MoSe2/Mo substrates, the texture coefficient is calculated based on the following equation:[22]

where I(hkl) is the observed peak intensity of the (hkl) plane, I0(hkl) is the intensity of the standard XRD pattern, and N is the total number of counted peaks for calculation. The diffraction peaks with larger texture coefficient demonstrate better crystallization in that direction. As shown in Fig. 2(b), the Sb2S3 on Mo substrate presents [hk0] preferred orientation, as indicated by the strong diffraction peaks of (120), (220), and (420), while the above peaks are suppressed in the Sb2S3 on MoSe2/Mo substrate, except for the stronger (310) peak. Notably, the intensities of (221) and (211) significantly increase, which become more than twice of those for the Sb2S3 on Mo substrate. The calculated (hk1)/(hk0) for these two samples increases from 0.18 to 0.51, which proves that the MoSe2 can effectively increase the (hk1) orientation. It suggests that the refined MoSe2 layer with larger surface roughness could enhance [hk1] orientation by providing more nucleation sites for Sb2S3 growth. As is well known, Sb2S3 exhibits a one-dimensional structure, which combines with covalent bonds along the ribbon, and the ribbons are held together by the weak Van der Waals force.[11,12] As a result, it leads to the optical property and carrier mobility being greatly anisotropic. When the films grow along (hk1), the ribbons will stand up or tilt on the substrate, which can offer transport tunnels to carriers. Thus, the modification of the absorber orientation would play a vital role in improving the Sb2S3 device performance. The Sb2S3 films grown on MoSe2/Mo possess higher intensity of (hk1) than the film grown on Mo substrate, which would facilitate the transport of carriers and reduce the covalent bonds breaking at the lattice terminal.

Fig. 2. (a) XRD patterns, the inset shows the Raman spectrum of Sb2S3 on MoSe2 substrate. (b) Texture coefficients of diffraction peaks of Sb2S3 thin films deposited on Mo substrate with and without MoSe2. SEM images: (c) top view and (d) cross section of Sb2S3 film deposited on Mo substrate, and (e) top view and (f) cross section of Sb2S3 film deposited on MoSe2/Mo.

The top view and cross-sectional SEM images of the Sb2S3 films deposited on Mo and MoSe2/Mo substrates are shown in Figs. 2(c)–2(f). It is observed that both of them exhibit compact and continuous grains without obvious voids, but the Sb2S3 film on the MoSe2/Mo substrate has smaller grains and some bright edges. From the cross-sectional SEM images, it can be seen that the Sb2S3 films are composed of large grains extending through the whole absorber, and the adhesion of the absorber to the substrate is good for these two samples.

To explore the influence of the modified Sb2S3 orientations on the transport of carriers, KPFM is characterized and used to detect Sb2S3 surface contact potential difference (CPD).[23,24] Figure 3 shows the surface topographies and CPD maps of the Sb2S3 thin films on different substrates from KPFM, and the line profiles reveal the relationship between the topography and CPD. From Figs. 3(c) and 3(f), we find the CPD at grain-interiors is higher than that at grain boundaries for both samples. As reported, the high potential can accumulate electrons and repel holes, the carriers can be separated by the CPD difference between grain-interior and boundaries.[25,26] For the Sb2S3 film on the MoSe2/Mo substrate, the whole CPD is elevated, and the relative difference in CPD between peaks and dips is larger (D2 > D1), revealing an enhanced capability of carrier separation, electrons and holes can effectively be collected by the corresponding electrodes.[27] The overall CPD distributions of the Sb2S3 films are shown in Fig. 4. Obviously, the CPD of the Sb2S3 film on the MoSe2/Mo substrate exhibits narrower distribution than that of the Sb2S3 film on the bare Mo substrate. It means that the surface potential is distributed with less fluctuation,implying that the surface defects and dangling bonds decrease.[28] As discussed above, the property of the Sb2S3 thin film is orientation dependent, so the improvement in carrier transport can be ascribed to the increased (hk1) channels and less dangling bonds on the surface of Sb2S3 induced by the MoSe2 layer.

Fig. 3. (a) Surface topography, (b) CPD map, (c) topography and potential line scans for Sb2S3 thin film grown on Mo substrate. (d)–(e) Those for Sb2S3 thin film grown on MoSe2 substrate. D1 and D2 in panels (c) and (f) show the difference in CPD between the peaks and dips.
Fig. 4. Difference between surface contact potentials of Sb2S3 films deposited on Mo substrate with and without MoSe2 derived by KPFM.
3.2. Performances of Sb2S3 thin film solar cells

Sb2S3 thin film solar cells are fabricated with substrate configuration of glass/Mo (MoSe2)/Sb2S3/CdS/iZO/AZO/Ni:Al. The device parameter distributions in Figs. 5(a)–5(d) reveal a total improvement of JSC from 8.32 mA/cm2 to 12.48 mA/cm2 on average. As for the open circuit voltage (VOC) and fill factor (FF), they are nearly the same. As a result, the efficiency is increased from 1.27 % to 1.62 % on average. Figure 5(e) displays the best J–V curves of the optimal Sb2S3 solar cells deposited on the Mo substrate and the MoSe2/Mo substrate. The best device parameters are summarized in Table 1. It is observed that the PCE of the Sb2S3 solar cells is improved from 1.36 % to 1.86 % by inserting a thin MoSe2 layer between the Sb2S3 layer and Mo layer. For the Sb2S3 solar cell with the MoSe2 interlayer, JSC dramatically increases from 7.11 mA/cm2 to 13.6 mA/cm2, though VOC declines from 490 mV to 466 mV and FF presents a slight drop. Considering the high work function of MoSe2, the band structure at back contact was first used to account for the improved JSC, but the work functions of the Sb2S3 film and Mo film measured by UPS show little difference. Specifically, their work functions are 4.27 eV and 4.86 eV, respectively (Fig. 6). Thus Sb2S3/Mo can naturally form ohmic contact, this means that a super thin MoSe2 has no influence on the rear band alignment. After that, we find that the enhancement of JSC can be explained by EQE in Fig. 5(f). A considerable improvement of quantum yield in the long wavelength range (550–750 nm) can be seen and the peak EQE is over 70 %, higher than that of the device reported in the literature. Besides, the limit of light absorption in long wavelength broadens to ca. 800 nm. As is well known, the increase in long wavelength reveals a correlation between the absorber crystallization and the back-interface modification. From this point, we conclude that the MoSe2 interlayer can not only modify the crystallization of Sb2S3 with (hk1) orientation but also serve as the texture of light trapping. In addition, a small amout of Se is supposed to diffuse into Sb2S3 during the heating process, because the band gap decreases from 1.73 eV to 1.65 eV (see the inset in Fig. 5(f)). As reported previously, sulfur vacancy (VS) is a deep-donor defect, which usually forms during Sb2S3 deposition at high temperature.[12] But VS can be easily replaced by Se, which helps to passivate the deep defects and improve the carrier density.[29] Considering the possibility of amorphous Se being absorbed on the MoSe2 surface, we speculate that some deep defects of Sb2S3 may be passivated by Se diffusion.

Fig. 5. Distributions of device parameters (a) VOC, (b)JSC, (c) fill factor, and (d) efficiency. (e) The J–V curves and (f) EQE spectra of optimal Sb2S3 solar cells deposited on Mo substrate with and without MoSe2, the inset shows the band gap extracted from the EQE data.
Fig. 6. UPS of (a) Sb2S3 film and (b) Mo film, showing the work functions of Sb2S3 and Mo being 4.27 eV and 4.86 eV, respectively.
Table 1.

Device performance parameters of Sb2S3 thin film solar cells.

.

The C–V measurement is employed to get further insight into the JSC enhancement. The abrupt p–n+ heterojunction can be viewed as a parallel-plate capacitor model. We can obtain the carrier concentration, depletion width, and built-in potential (Vbi) from the following equations:[29,30]

where NA,p, V, Vbi, q, ε0, εr, Wd, and A represent the carrier concentration in the Sb2S3 thin film, bias voltage, built-in potential, elementary charge, vacuum permittivity, relative permittivity, depletion width, and device area, respectively. The C–V and 1/C2–V curves of the Sb2S3 thin film solar cells without and with a MoSe2 are plotted in Fig. 7. It is observed that the capacity increases with bias voltage increasing. As the depletion width decreases, the capacity gradually turns into a constant. The built-in voltage Vbi can be estimated by the x-intercept in the 1/C2–V plot and Ein = Vbi/Wd is used to calculate the built-in electric field (Ein).[31,32] The values of Vbi, Wd, NA,p, and Ein are summarized in Table 2.

Fig. 7. The C–V and 1/C2–V curves for Sb2S3 solar cells (a) without and (b) with MoSe2.
Table 2.

Summary of built-in potential Vbi, depletion width Wd, carrier density NA,p, and built-in electric field Vbi.

.

Compared with the Sb2S3 solar cell without MoSe2, the sample with MoSe2 shows short Wd (647 nm) and great carrier density (1.8 × 1016 cm–3). Furthermore, the Ein of this device is almost twice as large as that of the sample reported in the literature. The large Ein will facilitate the carrier separation, which is consistent with the result of KPFM, thus contributes to an enhanced JSC. We also note that the built-in potential of the Sb2S3 solar cell with MoSe2 is 538 mV, which is higher than 347 mV of the control-one, but the VOC improvement is not notable, which is considered to be neutralized by the decreased band gap.

4. Conclusions

We fabricate Sb2S3 thin film solar cells with substrate configuration of glass/Mo (MoSe2)/Sb2S3/CdS/iZO/AZO/Ni:Al grid. A thin MoSe2 layer is developed before the deposition of the Sb2S3 thin film. The schematical investigation indicates that a thin MoSe2 layer with refined micro-structure can increase the (hk1) orientation of Sb2S3, providing more carrier transport channels. The concentrated CPD distribution from KPFM suggests less interface defects and dangling bonds on the surface of Sb2S3 on the MoSe2/Mo substrate. What is more, the CPD variations in both samples confirm an improvement of electrons barrier at GBs, which will facilitate the carrier separation. Consequently, the PCE of the Sb2S3 solar cell is improved from 1.36 % to 1.86 % by inserting the MoSe2 layer between the Sb2S3 laye layer and the Mo layer. The JSC dramatically increases from 7.11 mA/cm2 to 13.6 mA/cm2. EQE and C–V measurements further disclose the role of MoSe2 in enhancing JSC for this optimal device. The MoSe2 can not only improve the crystallization of Sb2S3 with (hk1) orientation, but also serve as a light tapping texture. Moreover, a small amount of Se is supposed to diffuse into Sb2S3 during the heating process, which is considered to passivate the deep defects and promote the device performance.

Reference
[1] Ghosh C Varma B P 1979 Thin Solid Films 60 61
[2] Moon S J Itzhaik Y Yum J H Zakeeruddin S M Hodes G Grätzel M 2010 J. Phys. Chem. Lett. 1 1524
[3] Kondrotas R Chen C Tang J 2018 Joule 2 857
[4] Yu Z Leilaeioun M Holman Z 2016 Nat. Energy 1 16137
[5] Choi Y C Lee D U Noh J Kim E Im S H 2014 Adv. Funct. Mater. 24 3587
[6] Shaji S Garcia L V Loredo S L Krishnan B Aguilar-Martinez J A Das Roy T K Avellaneda D A 2017 Appl. Surf. Sci. 393 369
[7] Sung S J Gil E K Lee S J Choi Y C Yang K J Kang J K Cho K Y Kim D H 2017 J. Ind. Eng. Chem. 56 196
[8] Avilez-Garcia R G Meza-Avendaño C A Pal M Paraguay-Delgado F Mathews N R 2016 Mater. Sci. Semicond. Process. 44 91
[9] Medina-Montes M I Montiel-González Z Mathews N R Mathew X 2017 J. Phys. Chem. Solids 111 182
[10] Escorcia-García J Becerra D Nair M T S Nair P K 2014 Thin Solid Films 569 28
[11] Yuan S Deng H Dong D Yang X Qiao K Hu C Song H Song H He Z Tang J 2016 Energy Mater. Sol. Cells 157 887
[12] Tumelero M A Faccio R Pasa A A 2016 J. Phys. Chem. 120 1390
[13] Deng H Zeng Y Ishaq M Yuan S Zhang H Yang X Hou M Farooq U Huang J Sun K Webster R Wu H Chen Z Yi F Song H Hao X Tang J 2019 Adv. Funct. Mater. 29 1901720
[14] Kondrotas R Zhang J Wang C Tang J 2019 Sol. Energy Mater. Sol. Cells 199 16
[15] Zhang L Zhuang D Zhao M Gong Q Guo L Ouyang L Rujun S Yaowei W Lyu X Peng X 2017 Mater. Lett. 208 58
[16] Pan G Wang D Gao S Gao P Sun Q Liu X Zhou Z Sun Y Zhang Y 2019 Sol. Energy 182 64
[17] Ito S Tsujimoto K Nguyen D C Manabe K Nishino H 2013 Int. J. Hydrogen Energy 38 16749
[18] Wang X Li J liu w Yang S Zhu C Chen T 2017 Nanoscale 9 3386
[19] Xia J Huang X Liu L Z Wang M Wang L Huang B Zhu D D Li J J Gu C Z Meng X M 2014 Nanoscale 6 8949
[20] Wu Z Tai G Wang X Hu T Wang R Guo W 2018 Nanotechnology 29 125605
[21] Soubelet P Bruchhausen A Fainstein A Nogajewski K Faugeras C 2016 Phys. Rev. 93 155407
[22] Li Z Chen X Zhu H Chen J Guo Y Zhang C Zhang W Niu X Mai Y 2017 Sol. Energy Mater. Sol. Cells 161 190
[23] Chen Y J Zhang M J Yuan S Qiu Y Wang X B Jiang X Gao Z Lin Y Pan F 2017 Nano Energy 36 303
[24] Vishwakarma M Varandani D Andres C Romanyuk Y E Haass S G Tiwari A N Mehta B R 2018 Sol. Energy Mater. Sol. Cells 183 34
[25] Zhao X Kou D Zhou W Zhou Z Meng Y Meng Q Zheng Z Wu S 2019 J. Mater. Chem. 7 3135
[26] Xin H Vorpahl S Collord A Braly I Uhl A Krueger B Ginger D Hillhouse H 2015 Phys. Chemistry Chemical Physics : PCCP 17 23859
[27] Yun J S Seidel J Kim J Soufiani A M Huang S Lau J Jeon N J Seok S I Green M A Ho B 2016 Adv. Energy Mater. 6 1600330
[28] Park J Huang J Yun J Liu F Ouyang Z Sun H Yan C Sun K Kim K Seidel J Chen S Green M A Hao X 2018 Adv. Energy Mater. 8 1701940
[29] Yuan S Deng H Yang X Hu C Khan J Ye W Tang J Song H 2017 ACS Photon. 4 2862
[30] Deng H Yuan S Yang X Cai F Hu C Qiao K Zhang J Tang J Song H He Z 2017 Mater. Today Energy 3 15
[31] Zeng Y Zhang S Li X Ao J Sun Y Liu W Liu F Gao P Zhang Y 2019 Chin. Phys. 28 058101
[32] Zhang Y Li B Dang X Wu L Jin J Li F Ao J Sun Y 2011 Chin. Phys. 20 116802