Stability Improvement of Perovskite Solar Cells for Application of CuInS2 Quantum Dot-Modified TiO2 Nanoarrays.

In this work, we developed a CH3NH3PbI3 perovskite solar cell with CuInS2 quantum dot-modified TiO2 nanoarrays (TiO2-CuInS2 QD-NAs) as a scaffold layer. Based on the suitable device configuration, we achieved improved power conversion efficiency (PCE) of 13.3%, which was 38.3% higher than that of the device without QD modification (8.2%). After exposure to air for 30 days, the TiO2-CuInS2 QD-NA-based device possessed a PCE of 5.4%, being 41% of the original performance, which was far superior to that of TiO2 nanoarray-based solar cells with a PCE of 1.1%. Our results showed that the crystallinity of perovskite, surface state, and interface for charge transport of TiO2-CuInS2 QD-NA-based perovskites all remarkably improved, indicating the improved air stability for TiO2-CuInS2 QD-NA-based solar cells.

Introduction

Recently, metal halide perovskites have been recognized to be most outstanding photoelectric materials due to their special performances[1−5] including the well-matched optical band gap,[1] prominent charge diffusion lengths,[6] low fabrication property,[7] very low temperature processability,[8] and high light absorbance.[9] This aroused great enthusiasm among scientific researchers for the rapid enhancement of power conversion efficiency (PCE). Nowadays, tremendous efforts have been made by various engineering techniques such as a crystal culture method,[10−12] element modification,[13] interfacial modification,[14−16] and the application of effective carrier-transport materials,[8,17,18] propelling the PCE to reach a high value up to 20%.[19−25]

However, the poor stability remains to be the major obstacle in the potential commercialization of perovskite photovoltaic devices. The hybrid perovskite material is found to be easy to decompose into PbI2 and volatile organic groups under moisture corrosion, photo-oxidation, and a high-temperature environmental due to the instability of the weak bonding effect.[26−28] Interfacial modification and controlling have been known as one of the key methods for enhancing stability in hybrid perovskite solar cells.[29−31]

Herein, we improved the iodine-based hybrid perovskite CH3NH3PbI3 (MAPbI3)/electron transport material (ETM) interface using TiO2 nanoarrays (TiO2-NAs) embellished with chalcopyrite CuInS2 quantum dots (CuInS2 QDs), which have been considered as excellent photovoltaic materials due to high light absorbance (ca. 105 cm–1), suitable band gap (ca. 1.6 eV), and low toxicity.[32] These have been applied in some fields such as polymer-based photovoltaic devices,[33] ternary hybrid devices,[34] and perovskite-based solar cells.[35,36] In our recent research, TiO2 nanoarrays decorated with CuInS2 QDs (TiO2–CuInS2 QD-NAs) were fabricated to be applied in MAPbI3 solar cells. These TiO2–CuInS2 QD-NAs are found to be advanced interfacial ETMs, which can effectively improve the stability of hybrid MAPbI3-based solar cells. The results show that the crystallinity of perovskite, surface state, and interface for charge transport of TiO2–CuInS2 QD-NA-based perovskites all remarkably improved in fluorine doped tin oxide (FTO)/TiO2–CuInS2 QDs/MAPbI3/spiro-MeOTAD/Au solar cells. Additionally, we observed that the TiO2–CuInS2 QD-NA-based devices greatly improved air stability compared to that of the pure TiO2-NA-based MAPbI3 solar cells.

Results and Discussion

Figure a shows the absorption spectrum of CuInS2 QDs, which shows wide absorption in the region from 300 to 800 nm, indicating their excellent light absorption properties in the whole ultraviolet–visible region. Coincided with chalcopyrite CuInS2, the absorption band gap of as-synthesized CuInS2 QDs was calculated to be about 1.75 eV, which matches well with the energy band gap requirements.[37] The TiO2-NAs show a wide shoulder in the region of 300–360 nm, which is in good agreement with the previous work (Figure b).[38] The strong peak at the wavelength of 310 nm is based on Ti (3d)–O (2p) energy transition of TiO2 nanoarrays.[38]Figure c shows the absorption spectra of MAPbI3 films coated on TiO2 nanoarrays and TiO2–CuInS2 QD array films. Both TiO2-NAs/MAPbI3 and TiO2–CuInS2 QD-NAs/MAPbI3 exhibit two bands at about 480 and 760 nm (Figure S1), for which the optical transition of MAPbI3 should be responsible.[39] Obviously, the absorption intensity of the TiO2–CuInS2 QD-NAs/MAPbI3 is enhanced compared to that of the TiO2-NAs/MAPbI3 sample, which can be explained by two aspects. on the one hand, the wide absorption response of CuInS2 QDs can add to the absorption intensity of the films. On the other hand, the fine crystal changes of the MAPbI3 on TiO2–CuInS2 QD-NAs (X-ray diffraction (XRD) data discussed in Figure ) should be responsible for the higher absorption intensity. The XRD patterns of chalcopyrite CuInS2 QDs, TiO2 nanoarrays, and QD-modified TiO2 nanoarrays are shown in Figure d. From the XRD patterns, we can see that the chalcopyrite CuInS2 QD peaks recorded at 2θ = 28, 47, and 55° are attributed to the (112), (220), and (312) diffraction crystal planes, respectively (JCPDS card #86-0147).[40] To further study the CuInS2 QDs, we measured the size of as-synthesized CuInS2 QDs by Scherrer’s formula. The average size of the CuInS2 QDs was estimated to be 1–3 nm, consistent with that of the pure particles in transmission electron microscopy (TEM) data (Figure S2). It can be clearly seen in the TiO2-NA patterns that three diffraction peaks attributed to (101), (211), and (002), respectively, appeared, which is in good agreement with the rutile TiO2 crystal marked by JCPDS card #21-1276.[40] Different from TiO2-NAs, the new (112) peak originated from the chalcopyrite CuInS2 QDs in TiO2–CuInS2 QD-NA samples, suggesting the successful crystallization of CuInS2 QDs on the TiO2 nanoarray substrate. The crystal growth direction is easily implied to be preferential along the [112] direction.

Figure 1

Absorption spectra of CuInS2 QDs (a), TiO2 nanoarrays and CuInS2 QD-modified TiO2 nanoarrays (b), MAPbI3 thin films coated on TiO2 nanoarrays and QD-modified TiO2 nanoarrays (c), and XRD patterns of various thin films (d). The solid circle in (d) indicates the XRD signals of FTO. The inset in (a) is the optical energy band gap curve of CuInS2 QDs.

Figure 6

(a) XRD patterns of MAPbI3 films on TiO2-NA and TiO2–CuInS2 QD-NA substrates and (b) corresponding XRD patterns after 30 days. The XRD signals of TiO2 and FTO are indicated by different symbols.

Transmission electron microscopy (TEM) images were obtained to study the nanoscale surface morphologies and the interfacial state of the TiO2-NA and QD-modified TiO2-NA samples. The pure TiO2-NAs (Figures a, S3a, and S3b) are highly unique nanorods with average diameters of 40–50 nm and corresponding lengths of 500 nm. The selected electron diffraction patterns (SEDPs) of the as-synthesized TiO2-NAs are displayed in Figure b, which are in quite agreement with those of the rutile TiO2 samples. Figures c, S3c, and S3d show the SEM and TEM images of TiO2–CuInS2 QD-NAs. Clearly, the CuInS2 QDs evenly dispersed on the surface of TiO2-NAs with a size of 3–5 nm, which was slightly larger than that of the pure CuInS2 QDs, for which the effect of the surface energy of TiO2-NAs should be responsible. The high-resolution transmission electron microscopy (HR-TEM) image of TiO2–CuInS2 QD-NAs is displayed in Figure d. CuInS2 QDs are clearly visible with an interplanar spacing of 0.25 nm, matching well with the (101) plane of the rutile TiO2.[40] The thickness of the CuInS2 QD layer in the QD-modified TiO2-NAs was calculated to be 6.5 nm, which is almost 2 QDs in length.

Figure 2

TEM (a) and SEDP (b) images of TiO2-NAs. TEM (c) and HR-TEM (d) images of TiO2–CuInS2 QD-NAs.

The photoluminescence (PL) spectra of the MAPbI3 thin films on FTO, TiO2-NAs, and TiO2–CuInS2 QD-NAs were investigated to confirm charge transfer behavior of CuInS2 QDs (Figure a). The pure MAPbI3 thin films on FTO showed a wide band located at 775 nm, which agrees well with the previous work.[40] The emission band was significantly quenched after coating of MAPbI3 on the TiO2-NA substrate, suggesting that an efficient charge transport channel existed between the MAPbI3 and TiO2-NA interface. What is more, the quenching intensity was remarkably enhanced when MAPbI3 contacted with TiO2–CuInS2 QD-NAs, implying the facilitated electron injection efficiency at the interface between the MAPbI3 layer and TiO2-NAs. Time-resolved photoluminescence spectrum was further recorded to test the charge transfer time (τCT), which can be calculated by the formula 1/τfilm = 1/τMAPbI + 1/τCT, where τfilm is the decay lifetime of MAPbI3 films on the ETM substrate and 1/τMAPbI is the decay lifetime of pure MAPbI3 films. As shown in Table S1, the τMAPbI of the pure MAPbI3 films was tested and estimated to be 42 ns, by which the τfilm and τCT for TiO2-NAs/MAPbI3 thin films were calculated to be 11.0 and 14.9 ns, respectively. In comparison, the τfilm and τCT in TiO2–CuInS2 QD-NAs/MAPbI3 were calculated to be 6.2 and 7.3 ns, which are much shorter than those of TiO2-NAs/MAPbI3 thin films, indicating that an improved charge transfer process occurred at the TiO2–CuInS2 QD-NA and MAPbI3 interface.

Figure 3

PL spectra (λex = 475 nm) (a) and time-resolved photoluminescence spectra (recorded at 775 nm, λex = 560 nm) of MAPbI3, TiO2-NAs/MAPbI3, and TiO2–CuInS2 QD-NAs/MAPbI3 thin films (b).

Figure S4 shows the cross section of solar cells. The corresponding device structure and energy-level diagram are displayed in Figure a,b, respectively. The J–V curve data of TiO2-NA-based and TiO2–CuInS2 QD-NA-based solar cells are measured in Figure c. For the TiO2-NA-based device, the performance with an open-circuit voltage (Voc) of 0.93 V, a short-circuit current (Jsc) of 13.4 mA cm–2, and a PCE of 8.2% was obtained. Compared to those of the solar cells based on TiO2-NAs, the TiO2–CuInS2 QD-NA-based device showed remarkably improved performance (Table ), where the Voc and Jsc were 0.98 V and 19.2 mA cm–2, respectively. The highest PCE of such a device was as high as 13.3%, which was 38.3% higher than that of the TiO2-NA-based device.

Figure 4

(a) Device structure and (b) energy-level diagram of the TiO2–CuInS2-NA-based solar cells; (c) J–V characteristics of TiO2-NA-based and TiO2–CuInS2 QD-NA-based solar cells and (d) corresponding external quantum efficiency (EQE) spectra.

Table 1

Device Performance of MAPbI3 Solar Cells Based on TiO2-NAs and TiO2–CuInS2 QD-NAs

devices	Voc (V)	Jsc (mA cm–2)	FF	PCE (%)
TiO2-NAs/MAPbI3	0.93	13.4	0.66	8.2
TiO2–CuInS2 QD-NAs/MAPbI3	0.98	19.2	0.71	13.3

The external quantum efficiency (EQE) spectra of the solar cells are recorded in Figure d. All of the devices showed a broad band in the region from 400 to 800 nm. The response peak at 730 nm can be attributed to the characteristic absorption of MAPbI3. The TiO2-NA-based device showed the highest EQE of only 62%. In contrast, the most efficient EQE of the TiO2–CuInS2 QD-NA-based solar cell was as high as 95%, much higher than that of the TiO2-NA-based solar cell. The results indicated that the QD-modified TiO2 nanoarrays can efficiently facilitate the charge transport process. To confirm this, we calculated the series resistance (Rs) using the typical eq ,[41] where the ideal diode behavior (n = 1) is assumed with IL and I0 representing photocurrent and dark saturation current (inverse polarization), respectively. Besides, q, T, K, and V correspond to the charge, absolute temperature, Boltzmann constant, and bias potential, respectively.Rs was estimated to be 2.81 Ω cm2 for the TiO2–CuInS2 QD-NA-based solar cell and 3.74 Ω cm2 for the TiO2-NA-based solar cell, indicating that the charge transport behavior is remarkably improved. Thus, the enhancement of the performance, especially the photocurrent in the TiO2–CuInS2 QD-NA-based solar cell, can be easily understood.

To investigate the stability of the solar cells, we studied the large-scale surface state of MAPbI3 films under an air environment for 30 days (Figures and S6). MAPbI3 coated on both TiO2-NA and QD-modified TiO2 nanoarray substrates showed smooth and compact surfaces, indicating a good crystallization process during annealing. After exposure to air for 30 days, the MAPbI3 on TiO2-NAs broken down into small particles with cracks on their surfaces. Compared to that on TiO2-NAs, MAPbI3 on the TiO2–CuInS2 QD-NAs substrate showed better air stability. The decomposition degree was greatly reduced with a more regular and smooth surface, indicating that TiO2–CuInS2 QD-NAs can effectively prevent MAPbI3 from decomposition.

Figure 5

SEM images of MApbI3 films coated on (a) TiO2-NAs and (b) TiO2–CuInS2 QD-NAs; the corresponding MApbI3 film SEM images on (c) TiO2-NAs and (d) QD-modified TiO2-NAs under an air environment for 30 days.

XRD patterns of MAPbI3 films on TiO2-NA and TiO2–CuInS2 QD-NA substrates were recorded to further corroborate the SEM results. Figure a shows the XRD patterns of MAPbI3 films on different substrates. For films on the TiO2-NA substrate, the diffraction peaks at 2θ = 14, 24, 28, 32, 40, and 43° can be assigned to the (110), (202), (220), (310), (224), and (314) planes of the tetragonal MAPbI3 crystal, respectively.[42] Notably, a new diffraction peak at ca. 36° appeared on TiO2–CuInS2 QD-NA samples that could be assigned to the (312) crystal plane of MAPbI3.[43] This implied that the MAPbI3 crystal was more inclined to grow along the [312] direction. It should be explained that there was almost no XRD signal of CuInS2 QDs in the films because of the very low content of QDs in TiO2–CuInS2-NAs/MAPbI3 films. To study the phase change of MAPbI3 films on TiO2-NA and QD-modified TiO2-NA substrates in air for 30 days, the corresponding XRD patterns were recorded (Figure b). For further comparison, we obtained the XRD pattern of PbI2. The diffraction peaks at 2θ = 12, 26, 34, and 39° can be assigned to the (001), (101), (102), and (110) planes, respectively, of the PbI2 nanocrystal.[44] The XRD pattern of MAPbI3 films on the TiO2-NA substrate after 30 days showed a sharp peak at 2θ = 12°, which could be assigned to the (001) plane of the PbI2 nanocrystal, indicating the decomposition of the MAPbI3 thin film. Interestingly, the PbI2 diffraction peak intensity of the (001) plane in MAPbI3 films coated on TiO2–CuInS2 QD-NA substrates was greatly reduced, indicating a more stable interface between MAPbI3 and CuInS2 QD-modified TiO2-NAs. Our results suggest that the CuInS2 QD-modified TiO2-NAs can effectively increase the stability of MAPbI3 films, which was in good agreement with the SEM data.

The performances of TiO2-NA-based and TiO2–CuInS2 QD-NA-based solar cells in air for 30 days were characterized (Figure S7 and Table ). After 30 days in air, the device based on TiO2-NAs showed a Voc of 0.73 V, a Jsc of 5.7 mA cm–2, and a PCE of 1.1%. In contrast, the corresponding TiO2–CuInS2 QD-NA-based device exhibited a Voc of 0.93 V, a Jsc of 10.1 mA cm–2, and a PCE of 5.4%. The PCE of the TiO2-NA-based device was only 12% of the original efficiency with a decreased fill factor of 0.27, for which the low stability of MAPbI3 should be responsible. Fortunately, the corresponding TiO2–CuInS2 QD-NA-based device showed better performance, which possessed 41% PCE of the original device. The results demonstrated that the CuInS2 QD-modified TiO2-NAs can effectively increase the stability of the solar cells.

Table 2

Device Performance of MAPbI3 Perovskite Solar Cells in Air for 30 days

devices	Voc (V)	Jsc (mA cm–2)	FF	PCE (%)
TiO2-NAs/MAPbI3	0.73	5.7	0.27	1.1
TiO2–CuInS2 QD-NAs/MAPbI3	0.93	10.1	0.57	5.4

Conclusions

In summary, solar cells based on CuInS2 QD-modified TiO2 nanoarrays were fabricated in this work. The solar cells based on TiO2–CuInS2 QD-NAs exhibited an improved PCE of 13.3%, much higher than that of solar cells using pure nanoarrays (8.2%). After 30 days in an air environment, the device based on TiO2–CuInS2 QD-NAs still possessed a PCE of 5.4%, which was 41% of the original device. The SEM and XRD results showed that the crystallinity, surface morphology, and interface for charge transfer of TiO2–CuInS2 QD-NA-based perovskites all remarkably improved, indicating the improved air stability of MAPbI3 films on TiO2–CuInS2 QD-NAs. On the basis of this, the CuInS2 QD-modified nanoarrays provide a novel channel for the preparation of high-performance and high-stability perovskite devices.

Experimental Section

Characterizations and Methods

UV–vis absorption spectra were measured to investigate the absorption properties (Agilent Cary 300). Time-resolved photoluminescence spectroscopic measurements (the obtained decay data was fitted by exponential functions with χ2 equal to 0.9–1.2) and PL spectra of the prepared films were measured to test the lifetime of the samples (Edinburgh FLS980). The X-ray diffraction patterns were recorded to measure the crystal characteristics (D/MAX 2500, Cu Kα radiation, λ = 1.5405 Å). Transmission electron microscopy (TEM) measurements (Tecnai G2 20 S-TWIN) and scanning electron microscopy (SEM) measurements (JEOL S-4800) were applied to test the morphologies of the samples. The MAPbI3 films were obtained by spin-coating the precursor solution on corresponding substrates at a speed of 1500 rpm. The films were then annealed at 100 °C for 30 min for further use.

Preparation of TiO2–CuInS2 QD-NAs

The TiO2-NAs were prepared by following the method in previous literatures.[45] First, a homogeneous solution was prepared in a 250 mL flask with 37% HCl and deionized water added (VHCl/Vwater = 1:1). The homogeneous solution was then shifted into a Teflon-lined chemical reaction vessel with impaction of FTO glass. Afterward, isopropyl titanate (1 mL) was added and the solution was sonication for 5 min. After heating at 180 °C for 90 min, the TiO2-NAs/FTO film was obtained, which was further annealed at 450 °C for 30 min to obtain TiO2-NAs.

The TiO2-NAs/FTO film was modified under room temperature by soaking into a cysteine solution (3 × 10–4 M) for at least 3 days, which was then added into a reaction mixture solution consisting of 60 mL of ethanol solution with Cu(Ac)2 (0.1 mmol), In(Ac)3 (0.1 mmol), CH3(CH2)16CH2NH2 (1.2 mmol), and CH4N2S (0.4 mmol).[32] The reaction mixture was further heated at 160 °C with a growth time of 9 h to form TiO2–CuInS2 QD-NAs/FTO thin film. The as-synthesized TiO2–CuInS2 QD-NAs/FTO thin film was finally annealed at 450 °C for 30 min, forming the chalcopyrite CuInS2 QD-coated TiO2-NAs/FTO thin film.

Fabrication and Characterization of Solar Cells

CH3NH3I was prepared by following a method reported previously.[46] Methylamine (ethanol solution, 24 mL, 33%) together with 10 mL of 57% HI (water solution) was dissolved in 100 mL of absolute ethanol for at least 2 h in a nitrogen atmosphere to achieve crystallization of CH3NH3I. After that, 0.463 g of PbI2 and 0.447 g of CH3NH3I (mole ratio 1:3) were added into 3 mL of dimethylformamide solution to form the CH3NH3PbI3 precursor solution. The precursor solution was then spin-coated onto the FTO/TiO2–CuInS2 QD-NA thin film (1500 rpm, 40 s). After annealing at 100 °C for 30 min, the MAPbI3 active layer was formed. Afterward, a spiro-MeOTAD (68 mM) chlorobenzene solution contains tert-butylpyridine (55 mM), and lithium bis(trifluoromethyl-sulfonyl)imide salt (9 mM) was prepared and spin-coated onto MAPbI3 at 2000 rpm for 40 s. Finally, a device (area of 0.12 cm2) was prepared by a 200 nm gold depositing onto the spiro-MeOTAD surface. The current–voltage (J–V) curves were tested under the AM 1.5G spectrum at 100 mW cm–2 (Newport Oriel). The EQE measurement was carried out to test the insight light–current conversion of devices using a Qtest station 500 instrument.