Improving stability of organometallic-halide perovskite solar cells using exfoliation two-dimensional molybdenum chalcogenides

Organometallic-halide perovskite solar cells (PSCs) are emerging as the most promising next generation solar cell devices. However, the stability is still the main bottleneck of their further development. Here, we introduce two-dimensional (2D) molybdenum chalcogenides (MoS2 and MoSe2) (MCs) nanoflakes as a buffer layer between perovskite layer and hole transport layer (HTL) to improve the stability of the organometallic-halide PSCs. 2D MCs are obtained via liquid-phase exfoliated (LPE) approach, and Glass/FTO/compact-TiO2/ mesoporous-TiO2/FA85MA15PbI85Br15/2D MCs/Spiro-OMeTAD/Au structured solar cell devices are designed and fabricated. In this system, 2D MCs act both as a protective layer and an additional HTL of PSCs. This kind of PSCs achieve a relatively high-power conversion efficiency (PCE) of 14.9%, along with a much longer lifetime stability compared to the standard PSCs. After 1 h, PCE of the PSC adding a 2D MCs buffer layer could maintain 93.1% of initial value, while the PCE of the standard PSC dropped dramatically to 78.2% of initial efficiency. Our results pave the way towards the implementation of 2D MCs nanoflakes as a material able to boost the shelf life of PSCs and further provide the opportunity to fabricate large-area PSCs in view of their commercialization.


INTRODUCTION
Traditional solar energy conversion devices based on crystalline silicon face the challenge of resource limitations and their energy intensive production process 1 . Therefore, there is an urgent need to develop cheap and efficient solar cells 2,3 . High-performance, economically viable, abundant and scalable device components are key to developing the solar cell materials. Organometallichalide perovskite solar cells (PSCs) are among of the most promising candidates for next generation solar cell technology [4][5][6] . Organometallic-halide perovskite have intense light absorption coefficient (1.5 × 10 4 cm −1 at 550 nm), long electron and hole diffusion length (up to 1 mm) and high carrier mobility (1-10 cm 2 V −1 s −1 ) 7-10 . However, the stability of PSCs has become a major barrier for their further development in the market [11][12][13][14][15][16][17] . Because perovskites are very sensitive to environment such as air, UV light, water, thermal stress and other factors [18][19][20] . Normally, adding hole and electron transporting layers (HTL and ETL, respectively) in PSCs between the perovskite layer and the electrodes helps to separate and selectively collect the generated charge carriers (holes at the cathode and electrons at the anode electrode) 21 . To date, TiO 2 is the most widely used ETL material for PSCs to prevent electron-hole recombination at the anode 4,6,22 . This kind of ETL typically consists of compact TiO 2 layer (c-TiO 2 , 50-80 nm) and mesoporous TiO 2 layer (mp-TiO 2 , 200-300 nm) 4,6,22 . On the other hand, 2,2′,7,7′-tetrakis-(N,N-di-4-methoxyphenylamino)-9,9′-spirobifluorene (Spiro-MeOTAD) represents the reference hole transport material (HTM) used for PSCs due to its facile deposition and high performance 23 . It could effectively transfer the photogenerated holes from the perovskite to the cathode while blocking the electrons from reaching the cathode. Unfortunately, Spiro-OMeTAD also has serious degradation issues induced by UV light, temperature, additives, and environmental conditions 23 . It appears that Spiro-OMeTAD HTL contributes to the poor stability of PSCs solar cell devices as well. Therefore, many other HTMs, such as copper (I) thiocyanate (CuSCN) 24 , nickel (II) oxide (NiO) 25 , and poly (3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT: PSS) 26,27 , have been explored to improve the PV performance including the stability of PSCs. However, these alternative HTMs could hardly compete to replace the ideal band edge positions of the Spiro-OMeTAD with respect to the perovskite absorbing layer. Another strategy to improve the stability of PSCs is inserting a buffer layer into the device structure to optimize the interface between the perovskite layer and Spiro-OMeTAD HTL [28][29][30][31] .
In this work, 2D MCs (such as MoS 2 and MoSe 2 ) nanoflakes were prepared via liquid-phase exfoliation (LPE) approach, which is a non-destructive, insensitive (to the environment such as air and water) and solution-processable method 37,38 . Then, these 2D MCs nanoflakes were utilized as a buffer layer material between perovskite layer and HTL (i.e., Spiro-OMeTAD) of the device to enhance the stability of PSCs. Based on the previous studies, Glass/FTO/compact-TiO 2 /mesoporous-TiO 2 /FA 85 MA 15 PbI 85 Br 15 /2D MCs/Spiro-OMeTAD/Au structured solar cells were designed and demonstrated. (a sketch of the solar cell and its band energy diagram is shown in Fig. 1a and b) 22,29,31,39 . We investigated various preparation conditions of 2D MC HTL buffer layers followed with a characterization of the devices' power conversion efficiency (PCE) and stability.

RESULTS
Preparation of 2D MCs 2D MCs nanoflakes (i.e., MoS 2 and MoSe 2 ) were obtained via sonication-assisted LPE approach. The principle of LPE method is quite simple. Because of the characteristic of layered structured materials (i.e., strong chemical bonds in-plane but weak interaction out-of-plane), ultrasonic energy can easily break up weak van der Waals interactions between each layer to obtain exfoliated 2D nanoflakes. As shown in Fig. 2a, taking MoS 2 as an example, first, bulk MoS 2 was exfoliated in N-Methyl-2-pyrrolidone (NMP), whose Hansen solubility parameters and surface energy are suitable for most 2D layered materials,using a probe sonic tip 37 . The sonic tip was pulsed for 6 s on and 2 s off with 60% amplitude of power to avoid damage to the processor and reduce solvent heating, and thus, degradation. During sonication, the temperature of the system was cooled to around 5°C by continuous flow of cooling water. Once sonicated, the dispersion was subjected to centrifugation to remove any unexfoliated materials. Then, the top 60% supernatant was collected. However, although NMP is the most effective solvent for the MoS 2 exfoliation, it is toxic (Health code ≥ 2, NFPA 704) and has a high boiling point (202°C). Moreover, it dissolves the perovskite crystals during the processing. Therefore, exfoliated MoS 2 nanoflakes were transferred from NMP to isopropanol (IPA, Health code 1; NFPA 704, boiling point: 82.6°C), forming environmentally friendly colloidal solutions (inset of Fig. 2b) with concentration up to 3 mg mL −1 . As shown in Fig.  2c, UV-vis spectra of exfoliated MoS 2 nanoflakes in NMP and IPA dispersions show typical feature absorption peaks of MoS 2 (one absorption peak onset about 450 nm and two peaks at around 674 nm and at around 612 nm), however, at the cost of concentration compared to they in NMP dispersion 37,40 . In addition, scanning electron microscopy (SEM, Fig. 2b) image confirms the efficient exfoliation into few-layered MoS 2  HAADF-STEM, HRTEM and its corresponding FFT pattern of exfoliated 2D MoSe 2 nanoflakes presented in Supplementary  Fig. 1a, b. Furthermore, different size of 2D MC nanoflakes obtained by centrifuging dispersions at different spin rates (1000 rpm, 3000 rpm, and 5000 rpm) called size selection. According to their TEM images and corresponding size distribution histogram ( Supplementary Fig. 2a-f), we could know that when the centrifuging rate increased from 1000 rpm to 5000 rpm, the average size of the exfoliated 2D MCs nanoflakes decreased from 248.27 nm to 42.61 nm. This is expected that increasing the centrifuging rates decreased the size of the MCs nanoflakes. And the thickness of liquid-phase exfoliated 2D MC nanosheets relates to their size 37 . The relationship between the dispersed concentration, C, and mean flake dimensions for exfoliated 2D nanosheets is where L is the length, w is the width and t is the thickness. This implies that reducing the concentration and size by increasing the centrifugation speed will result in thinner nanosheets.
Fabrication of PSCs with a 2D MCs buffer layer As shown in Fig. 1, a 2D MCs buffer layer was added between the perovskite layer and Spiro-OMeTAD layer (so called "HTL buffer") to comprise a Glass/FTO/compact-TiO 2 /mesoporous-TiO 2 / FA 85 MA 15 Table 1) because of the solvent (i.e., IPA) existing in 2D MCs dispersions. However, this factor could be improved by optimizing deposition process. As shown in Fig. 3, three different conditions were investigated: (P-1) MoS 2 deposited on annealed perovskite without further annealing; (P-2) MoS 2 is deposited on annealed perovskite film and further annealed at 100°C in the glove box; (P-3) MoS 2 is deposited on as deposited perovskite film then both films are annealed at 100°C. Supplementary Fig. 5a, b show the crosssection TEM images presenting the different components of the solar cell device from the front contact (FTO) to the back contact (Au) of the standard PSC and the PSC with a 2D MoS 2 HTL buffer layer prepared by (P-1), (P-2), and (P-3) methods, respectively. Comparing to the standard solar cell, the perovskite layer of the solar cell spin-coated the 2D MoS 2 dispersion by (P-1) process obviously damaged. This is because the small amount IPA solvent in 2D MoS 2 dispersions could induce degradation of the perovskite layer. Therefore, adding an annealing process (P-2 and P-3) could reduce the damage of IPA on the perovskite layer. However, as shown in Supplementary Fig. 5c, d, the interface between ETL and the perovskite layer of the PSC with a MoS 2 HTL buffer layer annealed separately from the perovskite layer (P-2) is much smoother than the one with a MoS 2 HTL buffer layer annealed together with the perovskite layer (P-3). It implies a MoS 2 could influence the growth of the perovskite, so it can reduce the quality of the perovskite layer. Therefore, it's better to do the deposition of the MoS 2 layer onto the complete perovskite layer by (P-2) method. Figure 4 shows the cross-section TEM image of the PSC with a 2D MoS 2 HTL buffer layer prepared by (P-2) approach. It shows less roughness then expected at interfaces perovskite / 2D MoS 2 / Spiro-OMeTAD which indicates that the inclusion of the MoS 2 layer between the perovskite and the Spiro-OMeTAD did not engender big cracks. This could be seen as a good sign for the solar cell operation.
PV performance of PSCs with additional 2D MCs buffer layer Figure 5e shows the IV curves of the four devices including standard PSC (without HTL buffer) and PSCs with a 2D MoS 2 buffer layer prepared by three different approaches as described above.
It is clear that the device prepared with sequential annealing (P-2) performs as good as the standard PCS, while the PCS prepared without annealing of MoS 2 (P-1) or annealing MoS 2 together with pristine perovskite film (P-3) performed less efficiently. Interestingly, the fill factor (FF) was strongly affected (45.9% instead of 64.2%, Fig. 5c, f), followed by the open-circuit voltage (V oc ) (0.92 V instead of 1.01 V, Fig. 5a, f) but the short-circuit current density (J sc ) was only slightly affected (21.64 mA cm -2 instead of 22.22 mA cm -2 , Fig. 5b, f). This indicates that there are serious interface issues under these two preparation conditions (P-1 and P-3). It is worth noting that IPA can partially dissolve the perovskite, so adding 2D MoS 2 buffer layers may lead to a damage of the surface of perovskite and this damage can be more serious in cases of (P-1) and (P-3). In (P-1) because the 2D MoS 2 film is not annealed so the residues of IPA will have more time to get in contact with the perovskite surface. In case of (P-3), the 2D MoS 2 film is added on the as deposited perovskite film so not in form of grains (crystals) which is make it surface more soluble than in the crystalline case. This could be the reason behind the low fill factor of the solar cell devices prepared under these conditions. This is in good agreement with the cross-section TEM images as mentioned before ( Supplementary Fig. 5c, d). It is worth noting that compared to a standard PSC, a PSC with a MoS 2 (P-2) layer has slightly higher V oc and J sc so that the efficiency of the device could maintain at 14.9% (Fig. 4d, f). This could be attributed to conducting MoS 2 nanoflakes improved charge transport property of the device 36 . What's more, it is clear from the energy band alignment (Fig. 1b) that MoS 2 can act as an electron blocking layer, which suppresses interfacial charge recombination 36 . In addition, using 2D MoSe 2 nanoflakes we studied the effect of the nanoflake's size to devices' PV performances. According to the IV curves ( Supplementary Fig. 6a-d), we could see that smaller size gave the better PV performance due its higher J sc and FF than the larger size nanoflakes. Therefore, the remaining of this manuscript focuses on small size 2D MCs nanoflakes integrated into PSC.

Stability of PSCs with additional 2D MCs buffer layer
Dark storage studies (ISOS-D), 57 which estimated a cell's shelf life under ambient conditions when it is not exposed to light were conducted at room temperature (23 ± 4°C) in ambient air and humidity using sun simulator (AM1.5 G including UV). PCE of the standard PSC, PSC with a 2D MoS 2 HTL buffer layer obtained by (P-2) and (p-3) were measured after 1 day, 1 week and 3 weeks. All PSCs were stored in ambient air at room temperature in dark. It is known that the PCE of a standard PSC using Spiro-OMeTAD as a HTL increases after several days because of the oxidation of Spiro-OMeTAD 23 . We observed the same behavior for our PSCs with/ without a 2D MoS 2 HTL buffer layer. Interestingly, we observed that the efficiencies of these solar cells continue to increase even after 3 weeks as it is shown in Supplementary Fig. 7a What's more, we also tested the stability of PSCs with/without a 2D MoS 2 HTL buffer layer under continuous sun illumination (AM1.5 G including UV) in ambient air at room temperature for 1 h. As shown in Fig. 6a, the PCE of the standard PSC reached its lifetime (21.8% of its initial efficiency) in 60 min. However, the PSC with a 2D MoS 2 HTL buffer layer only loses 6.9% of PCE after 60 min. This result shows clearly that the 2D MoS 2 prepared by exfoliation and inserted as a HTL buffer is able to enhance the stability of the PSC devices. When looking at the variation of each solar cell parameter V oc , J sc and FF with the illuminated time, one could observe a gradual decrease of the values of these parameters with the time. However, in comparison to the standard PSC, the main difference is in the decay of the FF and J sc with time, which is more pronounced in the absence of 2D MoS 2 . This is because the 2D MoS 2 buffer layer can cause changes in conductivity, transparency, structure, and/or band alignment with other layers of PSCs 29,30,36,[41][42][43] . This would result in preventing drastically changes of FF, likely by reducing recombination losses at the interface through blocking electron and accelerating charge transport, simultaneously 12,15 . In addition, the improvement of the stability also attributed to the stabilization of J sc . which is related with the improved charge transfer efficiency.
In addition, IV curves of forward and backward scans for a standard PSC and a PSC adding a 2D MoS 2 HTL buffer layer are shown in Supplementary Fig. 8a, b. No appreciable difference was observed between backward and forward scans for the standard PSC ( Supplementary Fig. 8a), while J sc increased apparently leading to the increase of PCE for the PSC adding MoS 2 buffer layer ( Supplementary Fig. 8b). This is because of improvement of the charge transport property by adding a conducting 2D MoS 2 HTL buffer layer between the perovskite and Spiro-OMeTAD. It also supports the conclusion that adding a 2D MoS 2 HTL buffer layer gives better stability than a standard PSC. X-ray diffraction (XRD) tests of a standard PSC and a PSC adding an additional 2D MoS 2 HTL buffer layer after stability test (working under continuously sun illumination for 1 h at room temperature in ambient air) were conducted to see the structure degradation of the perovskite layer, respectively. As shown in Fig. 7, perovskite  characteristic peaks of a PSC with a MoS 2 buffer layer much stronger than a standard one. It implies less degradation of the perovskite in the PSC with a 2D MoS 2 HTL buffer layer 36,44 . What's more, according to the energy-dispersive X-ray spectroscopy (EDX) mapping of the PSC with a 2D MoS 2 HTL buffer layer after IV measurement, it's apparent that a 2D MoS 2 HTL buffer layer efficiently prevents migration of lead (Pb), iodine (I), and bromide (Br) from the perovskite layer to the Spiro-OMeTAD HTL (Supplementary Fig. 9). In the meanwhile, a 2D MoS 2 HTL buffer layer also prevents the migration of metals (i.e., Au) from electrode to the Spiro-OMeTAD layer to the perovskite layer ( Supplementary  Fig. 9).The migration of metal through the HTL can lead to high recombination losses, which in turn cause reduction of FF and V oc during the operation 12,15,29 .

DISCUSSION
Here, we demonstrated the use of 2D MCs nanoflakes (mainly focus on MoS 2 ) as a HTL buffer layer of PSCs (Glass/FTO/compact-TiO 2 /mesoporous-TiO 2 /FA 85 MA 15 PbI 85 Br 15 /2D MC/Spiro-OMeTAD/ Au). Guided by our computational simulation of the device performance, we reveal that inserting suitable 2D MCs HTL buffer results in better interface quality, absorption of photon energy, and improving the stability by working as a scaffolding layer between perovskite and HTL in PSCs. 2D MCs nanoflakes are obtained via sonication-assisted LPE approach. Then, these exfoliated 2D MCs in IPA dispersions are spin-coated onto the perovskite layer and immediately annealed to remove IPA solvent to obtain the 2D MCs HTL buffer layer. Interestingly, when a 2D MCs buffer layer is added between the perovskite layer and HTL (Spiro-OMeTAD), the stability of the PSC improved apparently. PCE of the standard PSC dropped to 78.2% of original value after endurance test of 1 h under solar simulator (AM1.5 G including UV) illumination in ambient air at room temperature. However, PCE of the PSC adding a 2D MoS 2 HTL buffer layer could maintain 93.1% of PCE after 1 h under continuously illumination. Forward and backward IV scans and XRD, and EDX measurements of PSCs with/ without a 2D MoS 2 HTL buffer layer support the conclusion that adding a 2D MoS 2 HTL buffer layer into the PSCs structure could efficiently enhance the stability of them. However, further investigations are underway to deeply understand the reasons behind this stability enhancement using prolonged tests followed by analytical experiments such as XPS and time-of-flight SIMS.

Preparation of 2D molybdenum chalcogenides nanoflakes
Here, a typical methodology for the preparation of an exfoliated 2D molybdenum chalcogenides dispersion is described. For MoS 2 as an example, in the initial MoS 2 experiments, 4 g of MoS 2 powder were added to 80 mL of NMP, and a probe sonic tip was utilized to sonicate the solution for a certain number of hours (power: 60% amplitude). The sonic tip was pulsed for 6 s on and 2 s off to avoid damage to the processor and reduce solvent heating, and thus, degradation. The beaker was connected to a cooling system that allowed for cold water (5°C) to flow around the dispersion during sonication. Then, exfoliated MoS 2 nanoflakes in NMP solvent were centrifuged at high spin rate (10,000 rpm) to obtain the MoS 2 nanoflakes. Then, these MoS 2 nanoflakes were redispersed in IPA through bath sonication, so exfoliated molybdenum chalcogenides nanoflakes dispersion in IPA were obtained. Afterwards, they were then centrifuged at different centrifugation rates (1000 rpm, 3000 rpm, and 5000 rpm) for 1 h, and the supernatant was retained (size selection of MoS 2 ). The same methodology was used to obtain the exfoliated MoSe 2 dispersions.

Fabrication of perovskite solar cells
Substrate and ETL preparation: Solar cells were fabricated on FTO glass substrates with a sheet resistance of~8 Ω sq -1 . The impurities were removed from the FTO-coated substrate via a three-step cleaning process (soapy water, deionized (DI) water and isopropanol (IPA) for 15 min each).
The back side of the FTO glasses were covered with tapes. One side of the FTO glasses were etched using zinc power and 4 M HCl solution. Then, compact TiO 2 layer was deposited on FTO via dip-coating twice and annealing at 200°C for 10 min and 450°C for 30 min, respectively 39 . A precursor dip-coating solution was prepared by mixing 6 mL of titanium diisopropoxidebis(acetylacetonate) (TiAcac) solution with 54 mL of IPA. Then, mesoporous TiO 2 layer was deposited by spin coating for 20 s at 4000 rpm with a ramp of 2000 rpm s -1 , using 30 nm particle Ti paste (mixed titaniumoxide paste [Dyesol 30 NR-D] and ethanol with the ratio of 150 mg: 1 mL). After the spin-coating, the substrates were immediately annealed in a hot plate.
Perovskite precursor solution and film preparation: The perovskite solution was prepared first. Typically, 1.1 mM (508 mg mL -1 ) of PbI 2 , 0.183 mM (67.1 mg) of PbBr 2 , 1.05 mM (180.5 mg) of FAI, and 0.183 mM (20.4 mg) of MABr were dissolved into a mixture of DMF: DMSO (800 μL: 200 μL) solvent. Then, the perovskite solution was spin coated in a twostep program at 1000 and 4000 rpm for 10 and 30 s, respectively. During the second step, 300 μL of chlorobenzene was poured onto the spinning substrate 20 s prior the end of the program to smooth the film. Finally, the substrate was annealed at 100°C for 1 h in nitrogen-filled glove box.
Preparation of the 2D molybdenum chalcogenides buffer layer: Three different approaches were tried to prepare the 2D TMD buffer layer. First method is spin coating the 2D TMD layer at 2000 rpm for 20 s without any other treatment. Second and third methods added the heat treatment (100°C for 1 h) after spin coating the 2D TMD dispersion onto the substrate. In the second method, the TMD layer was spin coated and annealed after annealing the perovskite layer at 100°C for 1 h. But in the third method, the TMD layer was annealed together with the perovskite layer at 100°C for 1 h.

Materials and devices characterization
Characterization of 2D MCs nanosheets: The SEM images of all samples were acquired on a Zeiss Ultra Plus (Carl Zeiss, Germany) with an acceleration voltage of 2 keV. The TEM images of exfoliated 2D MCs nanosheets were performed on a FEI Titan at 300 kV (FEI, USA). Prior to the TEM study, the nanosheet dispersion was drop-casted onto holey carbon grids (Agar Scientific, UK). The UV-vis of 2D MCs dispersion was carried out with a Cary Varian 5000i UV-vis-NIR spectrometer. The dispersion was diluted to 1:10 in ethanol/water (60:40).
Characterization of PSCs: The numerical analysis has been carried out using 1D-solar cell capacitance simulator (SCAPS-1D) software. The computations were carried out using material parameters that were extracted from relevant literatures [48][49][50] . During the simulation, 2D MCs materials have been used as layered structured to mimic the real device structures. The highest reported electron mobility of MoS 2 is 200 cm 2 V -1 s -1,40,42 . However, the layered structure of these materials induces an anisotropy of the carrier transport. The electron mobility within the layers is excellent but it is much lower across the layers. In our device simulation, we have considered the effective mobility of 50 cm 2 V -1 s -1 as a typical mobility value based on at least two experimentally reported mobilities for MoS 2 layer having thickness varying from 1-23 monolayers 51,52 . FIB-TEM technique was used to get the cross-section TEM image and EDX mapping of the PSCs. XRD measurements were performed by means of Rigaku SmartLab X-Ray Diffractometer operating at room temperature. Current-voltage (I-V) tests were conducted using Newport Corporation IV tester under sun simulator (AM1.5G including UV).

DATA AVAILABILITY
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.