Improved performance and stability of perovskite solar cells with bilayer electron-transporting layers

Zinc oxide nanoparticles (NPs) are very promising in replacing the phenyl-C61-butyric acid methyl ester (PC61BM) as electron-transporting materials due to the high carrier mobilities, superior stability, low cost and solution processability at low temperatures. The perovskite/ZnO NPs heterojunction has also demonstrated much better stability than perovskite/PC61BM, however it shows lower power conversion efficiency (PCE) compared to the state-of-art devices based on perovskite/PCBM heterojunction. Here, we demonstrated that the insufficient charge transfer from methylammonium lead iodide (MAPbI3) to ZnO NPs and significant interface trap-states lead to the poor performance and severe hysteresis of PSC with MAPbI3/ZnO NPs heterojunction. When PC61BM/ZnO NPs bilayer electron transporting layers (ETLs) were used with a device structure of ITO/poly(bis(4-phenyl)(2,4,6-trimethylphenyl)amine) (PTAA)/MAPbI3/PC61BM/ZnO NPs/Al, which can combine the advantages of efficient charge transfer from MAPbI3 to PC61BM and excellent blocking ability of ZnO NPs against oxygen, water and electrodes, highly efficient PSCs with PCE as high as 17.2% can be achieved with decent stability.

Organic-inorganic hybrid perovskite solar cells (PSCs) have recently attracted tremendous attention because of their excellent photovoltaic efficiencies. [1][2][3][4] Since the initial results published in 2009 with efficiencies about 4% using a typical dyesensitized solar cell structure with liquid electrolyte, 5 signicant progress has been made in device performance through developing high quality lm processing methods, 6-10 tuning the perovskite composition, [11][12][13][14][15] optimizing the device architectures 16,17 and synthesizing new hole/electron transport materials. [18][19][20][21] Recently, a certied record power conversion efficiency (PCE) of 22.7% was achieved. 22 Despite of the success in obtaining dramatically improved PCE, there are certain concerns about the stability and cost towards commercialization. For the state-of-the-art PSCs, perovskites are susceptible to degradation in moisture and air, thus the charge transport materials should prevent the perovskite from exposure to such environments. 20,[23][24][25] One the other hand, PSCs also suffer from the high cost of widely used organic charge transport materials such as 2,2,7,7-tetrakis(N,N-di-p-methoxyphenylamine)-9,9-spirobiuorene (spiro-OMeTAD), phenyl-C 61/71 -butyric acid methyl ester (PC 61/71 BM). 3,18,26 As alternatives, inorganic materials such as CuSCN, 27 CuI, 28 CuGaO 2 , 20 and NiO x 29,30 which can be acted as hole transport materials and ZnO, 31,32 SnO 2 12,33,34 and TiO 2 10,35 which can be acted as electron transport materials are widely studied. Among them, metal oxide nanoparticles (NPs) are very promising in replacing the organic counterparts due to the high carrier mobilities, superior stability, low cost and solution processability at low temperatures. 16,31,33 The perovskite/ZnO NPs heterojunction has been demonstrated much better stability than perovskite/PCBM, 23 however it shows lower PCE compared to the state-of-art devices based on perovskite/PCBM heterojunction. [36][37][38] Thus in this paper, we systematically studied the charge transfer and recombination at CH 3 NH 3 PbI 3 (MAPbI 3 ) and ZnO NPs or PC 61 BM interfaces and tried to fabricate devices with high PCE and super stability simultaneously. We demonstrated that insufficient charge transfer from MAPbI 3 to ZnO NPs and signicant interface trap-states lead to the poor performance and severe hysteresis of PSCs based on MAPbI 3 /ZnO NPs heterojunction, while the devices based on MAPbI 3 /PC 61 BM show high PCE and negligible hysteresis due to the efficient charge transfer from MAPbI 3 to PC 61 BM and less recombination at the interface. On the other hand, the MAPbI 3 /ZnO NPs devices show excellent stability in air because of the excellent capping ability of ZnO NPs while the stability of MAPbI 3 /PC 61 BM devices is very poor. Thus, we fabricated the PSCs with bilayer electron-transporting layers (ETLs) with the device structure of ITO/poly(bis(4-phenyl)(2,4,6-trimethylphenyl)amine) (PTAA)/ MAPbI 3 /PC 61 BM/ZnO NPs/Al, trying to combine the advantages of efficient charge extraction ability of PC 61 BM and excellent blocking ability of ZnO NPs against oxygen, water and electrode, and nally device with PCE as high as 17.2% was achieved with decent stability.

Device fabrication and testing
Prior to fabrication, the substrates were cleaned by sonication using detergent, deionized water, acetone, and isopropanol sequentially for every 15 min followed by 15 min of ultraviolet ozone (UV-ozone) treatment. The substrates were transferred to a glovebox. PTAA lm was fabricated by spin-coating a toluene solution with a concentration of 5 mg mL À1 on the ITO substrates in glove-box.
PbI 2 (1 M) and DMSO (1 M) were dissolved in DMF under stirring at 70 C. The solution was then spin coated on the PTAA lm at 3000 rpm for 60 s. Then a solution of MAI in 2-propanol (IPA) (50 mg mL À1 ) was dropped and spin-coated at 3000 rpm for 60 s. Aerwards, the as prepared lms were heated at 90 C for 15 min. Aer cooling down, a layer of PC 61 BM (20 mg mL À1 in chlorobenzene) was spin-coated at 2000 rpm for 45 s for MAPbI 3 /PCBM junction solar cells. While for MAPbI 3 /ZnO junction solar cells ZnO nanoparticles in isopropanol was spincoated at 4000 rpm for 30 s. Subsequently, samples were loaded into a vacuum deposition chamber (background pressure z 5 Â 10 À4 Pa) to deposit a 100 nm thick Al cathode with a shadow mask. To specify the illuminated area, we used an aperture with an area of 0.06 cm 2 , whereas the total device area dened by the overlap of the electrodes was approximately 0.12 cm 2 .
The J-V characteristics were measured with Keithley 2400 measurement source units with the devices maintained at room temperature in glove-box. The photovoltaic response was measured under a calibrated solar simulator (Enli Technology) at 100 mW cm À2 , and the light intensity was calibrated with a standard photovoltaic reference cell. The devices were stored in glove-box in dark overnight before measurement. The forward J-V scans were measured from À0.1 V to 1.2 V with a scan rate of 0.05 V s À1 and a voltage step of 0.01 V while the reverse J-V scans were measured from 1.2 V to À0.1 V with a scan rate of 0.05 V s À1 and a voltage step of 0.01 V. The EQE spectrum was measured using a QE-R Model of Enli Technology.

Results and discussion
Inverted perovskite solar cells with the device architecture of ITO/PTAA/MAPbI 3 /PC 61 BM or ZnO NPs/Al were fabricated, which were shown in Fig. 1a. Fig. 1b shows the corresponding energy level diagram of the devices. The reported valance band of ZnO NPs and PC 61 BM are similar with a value of 4.2 eV, which is 0.3 eV lower the valance band of MAPbI 3 , and thus the electrons in perovskite lm can transfer to both ETLs and be collected by electrodes. Fig. 2a shows the current-voltage (J-V) characteristics of PSCs based on ZnO NPs and PC 61 BM as ETLs under 100 mW cm À2 AM1.5G solar illumination with reverse and forward scans. The corresponding photovoltaic parameters are summarized in Table 1. The device employing ZnO as ETL exhibits an open-circuit voltage (V OC ) of 0.98 V, a short-circuit current density (J SC ) of 12.5 mA cm À2 , and a ll factor (FF) of 0.64, yielding a PCE of 8.0% at reverse scan, and suffers severe hysteresis with a much lower PCE of 6.3% (V OC ¼ 0.99 V, J SC ¼ 12.0 mA cm À2 and FF ¼ 0.52) at forward scan, respectively. While the device using PC 61 BM as ETL shows a signicant improvement in PCE up to 15.0% at reverse scan with a V OC of 1.06 V, a J SC of 19.1 mA cm À2 and a FF of 0.72, and more importantly, with negligible hysteresis (14.70% PCE at forward   scan with a V OC of 1.05 V, a J SC of 19.2 mA cm À2 and a FF of 0.71). It exhibits improvement on all three parameters simultaneously compared to those of ZnO NPs based devices. The much enhanced J SC was also demonstrated by the external quantum efficiency (EQE) spectra shown in Fig. 2b. We also tested the stability of devices which were stored in air under dark with a humidity of 20% for 120 days. The corresponding J-V curves were shown in Fig. 2c. We found that the device with ZnO NPs as ETL retained 95% of the initial PCE, with a V OC of 1.06 V, a J SC of 15.6 mA cm À2 , a FF of 0.45 and a PCE of 7.6% aer 120 days storage, while the device with only PC 61 BM as ETL almost died only aer 10 days, showing a V OC of 0.56 V, a J SC of 0.77 mA cm À2 , a FF of 0.13 and a PCE of 0.06%. The much worse stability was attributed to the poor blocking ability of PC 61 BM against oxygen, water and the electrode. 23 In order to obtain both high PCE and excellent stability, we also fabricated device with bilayer ETLs in a structure of ITO/ PTAA/MAPbI 3 /PC 61 BM/ZnO NPs/Al shown in Fig. 1a. The device shows highest V OC , J SC , FF and PCE which ups to 17.2% at reverse scan with a V OC of 1.11 V, a J SC of 19.6 mA cm À2 and a FF of 0.79, and also shows negligible hysteresis with a V OC of 1.11 V, a J SC of 19.5 mA cm À2 , a FF of 0.78 and a PCE of 16.9% at forward scan. Aer stored in air 120 days, the device also shows a decent PCE of 11.4%, with a V OC of 1.09 V, a J SC of 19.2 mA cm À2 , a FF of 0.54, which is 66% of the initial value. Thus we demonstrated that this type of device show the advantages of the MAPbI 3 /PC 61 BM device with high PCE and the MAPbI 3 /ZnO NPs device with good stability.
The effects of different ETLs on the charge extraction and recombination process at perovskite/ETL interface were investigated in details to nd out the reasons of different behaviours of corresponding devices. Steady-state photoluminescence (PL) was performed to compare the electron transfer efficiency from perovskite to ETLs. As shown in Fig. 3a, only 54% of the PL intensity was quenched by depositing ZnO NPs on top, while more than 90% PL intensity was quenched when PC 61 BM or PC 61 BM/ZnO NPs were deposited on top of the perovskite layer. This means insufficient charge transfer from perovskite to ZnO NPs, while electrons in perovskite lm can be efficiently transferred to PC 61 BM layer. This was further conrmed by timeresolved PL (TRPL) (Fig. 3b). The TRPL curve was tted to a biexponential equation: Y ¼ A 1 exp(Àt/s 1 ) + A 2 exp(Àt/s 2 ) and the detailed data are shown in Table 2. In the absence of ETL quencher, the pristine perovskite lm showed a relatively long PL lifetime of 56.8 ns, while it decreased to 25.6, 5.4 and 5.3 ns for the ZnO NPs, PC 61 BM and PC 61 BM/ZnO NPs-based lms, respectively. This implies that faster and more efficient electron extraction was achieved at the perovskite/PC 61 BM interface. The insufficient charge transfer from perovskite to the ZnO NPs layer compared to PC 61 BM could due to the shallower conduction band and lower electron mobility of ZnO NPs, or the interfacial traps at the perovskite/ZnO NPs interface, 23 and the worse contact at perovskite/ZnO NPs interface. These could cause charge accumulation at the perovskite/ZnO NPs interface and thus leads to poor performance and severe hysteresis in the corresponding device. 23,33,40 The recombination kinetics were also studied carefully by measuring J SC and V OC at various light intensities (I) from 130 to 2.8 mW cm À2 (Fig. 4a and b). A power law dependence of J SC upon illumination intensity is generally expressed as J SC f I a , where I is the light intensity and a is the exponential factor. At short circuit condition, the bimolecular recombination should be minimum (a z 1) for maximum carrier sweep out. Any deviation from a z 1 implies bimolecular recombination. [41][42][43] Fig. 4a shows that J SC f I a , where a ¼ 0.96 AE 0.01 for the device using ZnO NPs as ETL while a z 1 for both devices using PC 61 BM or PC 61 BM/ZnO NPs as ETLs, indicating weak bimolecular recombination at short-circuit condition in the latter two types of devices. 43 For the MAPbI 3 /ZnO NPs device, the lower a could be attributed to the bimolecular recombination during sweep-out. 41,42 At open-circuit conditions, the current is zero, all carriers recombine within the cell. Thus,   23 The reduced recombination at both interfaces renders the highest V OC and performance of this type device.

Conclusions
We demonstrated that the insufficient charge transfer from MAPbI 3 to ZnO NPs and signicant interface trap-states lead to the poor performance and severe hysteresis of PSC with MAPbI 3 /ZnO NPs heterojunction, but the device shows super stability in air. While the device based on MAPbI 3 /PC 61 BM heterojunction shows high PCE and negligible hysteresis due to the efficient charge transfer from MAPbI 3 to PC 61 BM and less recombination at the interface, however the device show very poor stability in air. When PC 61 BM/ZnO NPs bilayer ETLs were used with a device structure of ITO/PTAA/MAPbI 3 /PC 61 BM/ZnO NPs/Al, high efficient PSCs with PCE as high as 17.2% can be achieved with decent stability. Our study also showed the possibility of obtaining highly efficient perovskite/metal oxide NPs heterojunction solar cells by interface engineering without high cost PC 61 BM.

Conflicts of interest
There are no conicts to declare.