Preparation of Low Grain Boundary Perovskite Crystals with Excellent Performance: The Inhibition of Ammonium Iodide.

For the study, we prepared a low grain boundary three-dimensional CH3NH3PbI3 crystal (3D-MAPbI3) on TiO2 nanoarrays by inhibition of ammonium iodide and discussed the formation mechanism of the crystal. Based on the 3D-MAPbI3 crystal, solar cells showed modified performance with a power conversion efficiency (PCE) of up to 19.3%, which increases by 36.8% in contrast to the counterparts. We studied the internal photocurrent conversion process. The highest external quantum efficiency is up to 92%, and the electron injection efficiency is remarkably facilitated where the injection time decreases by 37.8% compared to the control group. In addition, based on 3D-MAPbI3, solar cells showed excellent air stability, which possesses 78.3% of the initial PCE, even though they were exposed to air for 30 days. Our results demonstrate a promising approach for the fabrication of perovskite solar cells with high efficiency and stability.

Introduction

Organic–inorganic hybrid perovskites have been considered as potential materials in various fields due to their special chemical structures and distinctive properties.[1−5] Since the first methylammonium lead halide perovskite (CH3NH3PbX3, X = Br or I) in an electrolyte-based dye-sensitized solar cell was reported, scientists have researched the perovskite-based photovoltaic devices widely. Up to now, based on perovskite materials, solar cells have achieved outstanding improvements with a power conversion efficiency (PCE) reaching 25.6%.[1,6−20] The increasing device PCE heralds a new rapid development era of solar cells.

One of the main issues that hinder the development of perovskite solar cells is poor stability.[21−25] To overcome the difficulty, numerous efforts were devoted to improve the interface of the device, including developing various new electron transport materials (ETMs),[26−30] hole transport materials (HTMs),[31−35] and prepared all-inorganic solar cells.[20,36] Scientists expect using various carrier materials to improve the performance and even the stability. However, very few of the reported materials have been able to achieve gratifying stability. In fact, the general engineering of the interface facilitates the charging transfer behavior by providing effective charging transport channels.[18,28,37−40] The transporting process of the carrier, especially the transporting speed (injection rate) between the donor and the receiver, has not been greatly improved.

In addition, obtained by the one-step method, perovskite crystals usually have many morphologic defects because of the heterogeneous growth during the evaporation of the solvents. For perovskite, during the complete transformation from precursor materials to crystals, the morphology of the film presents a large number of pinhole defects, which is harmful to the performance of the device. To overcome the obstacle, a great variety of traditional technologies were attempted preliminarily, such as the roll-to-roll method, doctor-blading process, spray coating, inkjet printing, slot-die printing, and so on.[41−46] However, the perovskite materials obtained by these methods are either too expensive or poor on the performance, which makes it difficult to popularize in reality.

In addition, there are many defects in perovskite crystals synthesized by chemical methods. These defects are harmful to the inner carrier transport of the device. Therefore, as for perovskite, to reduce or repair the defects in the crystal effectively has become one of the key challenges. In previous reports, synthesized by the gas phase method, perovskite can reduce the crystal defects[47−49] effectively. However, there are no suitable gas phase methods to synthesize low-defect perovskite crystals except for the traditional chemical vapor deposition (CVD) method, which is reported to be the main one,[13,47] but this kind of method has many disadvantages, such as high energy consumption, high price, and inability to be applied or promoted.[50] In view of this, scientists gradually developed and utilized different gas phase methods, such as the spray deposition method,[51,52] gas post-treatment,[53] gas-assisted crystallization,[54] and so on. However, the obvious disadvantage of these methods is the simple use of methyl ammonia gas for repair. It has not been well studied on controlling the process of growth to improve the morphology and structure of crystals.

Hereinafter, we reported a three-dimensional (3D) perovskite crystal on TiO2 nanoarrays (TiO2-NAs) with low grain boundaries by introducing ammonium iodide into the film. We also studied the formation process and defect repair mechanism of the crystal. Consequently, we reported 3D-MAPbI3-based solar cells with improved performance and air stability. We also discussed the reasons for the improvement on the performance of the device. Notably, the electron transport layer is the TiO2 array rather than mesoporous TiO2. The main reasons for choosing the TiO2 nanoarray as the ETM are as follows: (i) TiO2 nanoarrays have good single-crystal properties and one-dimensional phase extractions, which can provide more contact sites and effective charging transport channels for fast electron transports due to the high uniform structures;[27] (ii) there are almost no deep insights into the healing process upon the perovskite/two-dimensional nanoarray interface; (iii) the mesoporous TiO2 (m-TiO2) film is normally used for the ETM in perovskite solar cells to increase the contact area,[55] but the crystallinity and the surface area of such TiO2 thin films are not satisfactory.[15] Additionally, by using the simple spin-coating methods, perovskite crystals formed on mesoporous TiO2 films obtain two-dimensional (2D) flat films, which reduce the contact interface between perovskite and TiO2 films as well.

Experimental Section

Materials and Methods

All the materials were bought from commercial sources for use. We recorded the UV–vis measurements to test the absorption properties of the samples (Agilent Cary 300) and X-ray diffraction to test the crystal of the samples (D/MAX 2500, Cu Kα radiation, λ = 1.5405 Å). We characterized the morphology by transmission electron microscopy (TEM) measurements (Tecnai G2 F20 S-TWIN transmission electron microscope) and scanning electron microscopy (SEM) measurements (JEOL S-4800 field-emission scanning electron microscope). Time-resolved photoluminescence measurements were performed on a fluorescence spectrometer (FLS980, excitation source: 560 nm). The obtained decay curves were fitted with exponential functions (χ2 = 0.9–1.1).

Preparation of TiO2 Nanoarrays

We prepared the TiO2 nanoarrays by using hydrothermal methods.[27] Typically, we added 50 mL of 37% HCl and 50 mL of distilled water into a 250 mL flask with stirring for 15 min to form homogeneous solutions. Then, we transferred the solutions into an FTO-embedded Teflon-lined hydrothermal reaction vessel. Afterward, we added 1 mL of isopropyl titanate into the vessel and sonicated the mixture for 5 min. Finally, we heated the vessel at 180 °C for 90 min to obtain the TiO2 nanoarrays. For comparison, we prepared m-TiO2 by referring to a previous report.[56]

Preparation of MAPbI3 Precursor Solution

We synthesized CH3NH3I by the one-step method in the literature.[57] Methylamine (24 mL, 33% in absolute ethanol) was reacted with hydroiodic acid (10 mL, 57% in water) under a nitrogen atmosphere in 100 mL of ethanol for 2 h. After the evaporation of the mixture, a white-colored powder was formed to obtain the CH3NH3I crystal. We prepared the MAPbI3 precursor solution by blending 0.463 g of PbI2 and 0.447 g of CH3NH3I (mole ratio 1:3) in 3 mL of DMF solution.

Fabrication and Characterization of the Device

First, we added 0.1 mL of MAPbI3 precursor solution onto the TiO2-NAs/FTO thin film and directly annealed at 80 °C for 30 min under an air atmosphere to assemble the MAPbI3 crystal. Then, we spin-coated 50 μL of NH4I in ethanol (0.1 M) onto the films and heated at 60 °C for 12 h to form a uniform film (the synthesis process is displayed in Figure ). After that, we spin-coated a chlorobenzene solution containing 68 mM Spiro-MeOTAD, 55 mM tert-butylpyridine, and 9 mM lithium bis(trifluoromethylsyfonyl)imide salt on the active layer at a rate of 2000 rpm for 40 s. Finally, we deposited 200 nm Au onto the Spiro-MeOTAD surface (a device area of 0.12 cm2). We measured the current–voltage (J–V) characteristics of the devices on a solar simulator under a simulated AM1.5G spectrum at 100 mW cm–2 (Newport Oriel Sol 2A) and the external quantum efficiency (EQE) of the devices to test the inner photocurrent (QTest Station 500 USA).

Figure 1

Synthesis process of the 3D-MAPbI3 crystal.

Results and Discussion

Figure a shows the top-view SEM image of the untreated MAPbI3 crystal on TiO2-NAs. MAPbI3 is rough with lots of holes, and the crystal structure is loose. However, after 3 h of treatment with NH4I, the perovskite crystals began to aggregate and grow into large films (Figure b). When the treatment continued for 24 h, the perovskite crystal was restored to a more complete dense layer (Figure c). This indicates that NH4I possesses the ability to repair the crystal into a block film. Figure d shows the cross-sectional image of NH4I-treated perovskite on TiO2-NAs. Obviously, the TiO2 nanoarray has fully penetrated the perovskite crystals and the thickness of the compact perovskite on TiO2-NAs is calculated to be about 500–600 nm. Thus, the crystal forms a dense, smooth, and uniform three-dimensional structure. However, the MAPbI3 crystal on TiO2-NAs without NH4I treatment (Figure S1) is rather rough with lots of grain boundaries and not completely filled in TiO2-NAs. Notably, although grain boundaries are still observed on the surface of 3D-MAPbI3 crystals, the number has been greatly reduced.

Figure 2

(a) Top-view SEM image of MAPbI3 on TiO2-NA films without ammonium iodide; (b) MAPbI3 on TiO2-NA films treated with NH4I for 3 h; (c) MAPbI3 on TiO2-NA films treated with NH4I for 24 h; (d) cross-sectional SEM images of MAPbI3 films on TiO2-NAs after NH4I treatment for 24 h.

Figure a and Figure b show the SEM and TEM images of the as-synthesized TiO2 nanoarrays, respectively. The average diameter of the nanorods is estimated to be 50–80 nm, and the length is about 500 nm. Figure c shows the corresponding high-resolution TEM (HRTEM) of the TiO2 nanoarrays. The interspace (0.335 nm) of the crystal is the reflection of the rutile TiO2 (101) plane.[27] We recorded the structures by XRD patterns (Figure d). TiO2-NAs/FTO showed three peaks at 26.3, 35.8, and 37.7°, which agrees well with rutile TiO2.[58] For the CH3NH3PbI3 crystals, there are three new peaks that appeared at 14.2, 28.5, and 31.8° in both of the 3D-MAPbI3 films before and after the NH4I treatment, which can be attributed to the (110), (220), and (310) planes, respectively.[59] It is worth noticing that all the peak intensities in the 3D-MAPbI3 films after the NH4I treatment are extremely enhanced. This means that the crystallized process that occurred in the healing process is better. The XRD peak at 52° in all films can be attributed to the subtraction of FTO. We tested the absorption spectra to distinguish the absorption features of MAPbI3 films. As shown in Figure e, the 3D-MAPbI3 crystal showed broad absorption bands at regions from 300 to 800 nm, which is in good agreement with the previous work.[60] Compared to the untreated films, the absorption intensities of 3D-MAPbI3 crystals after the NH4I treatment are significantly increased, indicating that a further assembly process occurs during the NH4I treatment process. The formation of compact bulk crystals of perovskite should be responsible for the high absorption effects.

Figure 3

(a) Cross-sectional SEM, (b) TEM, and (c) HRTEM images of TiO2 nanoarrays; (d) XRD patterns of MAPbI3 on TiO2 nanoarrays and (e) absorption spectra of MAPbI3 thin films on glasses.

Based on the above results, we proposed the healing mechanism. Actually, the repairing process of the crystal is the reversible absorption/release of the MA gas.[47] Therefore, in an effective healing process, we should take two aspects into consideration: (i) the transformation of the rough film (Figure a) into a smooth one (Figure c) through self-leveling; (ii) the slow release of MA to produce a highly uniform and compact film. The whole process involves an intermediate state (HPbI3) that is unstable to release a small amount of HI molecules.[49] The relative equations are as follows

Therefore, if we control the slow formation of intermediate products (HPbI3), then the whole chemical kinetic process will be effectively controlled. According to this, we added ammonium iodide to films to destroy the dynamic equilibrium process, as shown in formula . The main reason for choosing ammonium iodide is that it is unstable and easy to decompose into HI and ammonia gas under heating conditions. Because of the release of HI from NH4I, the equilibrium process of the chemical reaction in formula moves toward the direction of reactants. This inhibited the decomposition of intermediate products and slowed down the release rate of methylamine to a certain extent. With the slow decomposition and recrystallization of perovskite molecules, new smooth, homogeneous, and large-scale perovskite films are gradually formed. The whole phase transition process can be reflected by SEM from Figure a–c.

It is believed that the loose structure of MAPbI3 before the treatment provides an effective channel for the HI gas to enter the array, which makes the perovskite phase transitions among the titanium dioxide arrays proceed smoothly. Thus, the pore fillings of the MAPbI3 molecules in TiO2 nanoarrays are improved effectively. This effect can be easily observed by the cross-sectional SEM images (Figure d).

We fabricated the solar cells based on 3D-MAPbI3. The energy-level diagrams of the solar cells are displayed in Figure a. All the energy levels of the materials refer to our previous work.[27] In Figure b, we measured the J–V curves of the champion solar cells scanned in forward and reverse directions. The 3D-MAPbI3 device after the NH4I treatment shows less hysteresis than the untreated device. Theoretically, in the perovskite solar cells, accumulations of chargings or ions and transferring imbalances of chargings at the interface between perovskite/ETM should be responsible for the hysteresis. Based on MAPbI3 before NH4I treatment, solar cells achieved performance of an open-circuit voltage (Voc) of 1.04 V, a short-circuit current (Jsc) of 20.7 mA cm–2, a fill factor (FF) of 0.65, and a PCE of 14.1%. Compared to that, based on 3D-MAPbI3 after the NH4I treatment, the performance of the device showed significant improvement where the Voc, Jsc, FF, and PCE are measured to be 1.04 V, 23.8 mA cm–2, 0.78, and 19.3%, respectively. The PCE increased by nearly 36.8%. In addition, the Jsc and FF (Table S1) based on 3D-MAPbI3 are remarkably improved, indicating that 3D-MAPbI3 is the photovoltaic material that has more potential. It should be noted that the Voc in both of the devices is the same, suggesting that the Voc is mainly dominated by the energy levels between MAPbI3 and TiO2 nanoarrays. To confirm the performance of the device, we recorded 30 devices, and the average statistics are displayed in Table (Figures S3 and S4). The statistical data demonstrates that the average performance has been improved for NH4I-treated 3D-MAPbI3-based devices.

Figure 4

(a) Energy-level diagram of the solar cells based on 3D-MAPbI3; (b) J–V curves and (c) dark J–V curves of the devices; (d) corresponding EQE and calculated Jsc of the solar cells.

Table 1

Average Performances of 30 Devices Based on MAPbI3 Crystals

devices	Voc (V)	Jsc (mA cm–2)	FF	PCE (%)
before treatment	1.04 ± 0.1	19.6 ± 2.0	0.63 ± 0.3	12.8 ± 2.1
after NH4I treatment	1.04 ± 0.1	22.5 ± 1.8	0.76 ± 0.3	18.8 ± 2.0

To better understand the transfer and recombination kinetics of the charging, we studied the dark J–V data (Figure c), in which the curves are in good accordance with the Shockley diode equationwhere n is the ideality factor, k is the Boltzmann constant, J0 is the reverse saturation current density, and T is the temperature in Kelvin. Compared with the device before the NH4I treatment, solar cells treated with NH4I showed lower J0 values and inflection points, suggesting a charging contact, which is more effective, for which the lower charging recombination and leakage at the interface between 3D-MAPbI3 and TiO2-NAs should be responsible.[61]

Considering the enhancement in Jsc, we recorded the external quantum efficiency (EQE) spectra of the solar cells (Figure d) to further investigate the inner photocurrent conversion efficiency. Both of the two MAPbI3-based devices showed a broad band in the whole region from 400 to 800 nm. Based on MAPbI3, the highest EQE before treatment is 84%, while the EQE increases dramatically for the devices based on 3D-MAPbI3 after the NH4I treatment. The highest EQE reaches 92%, indicating the improvement of Jsc. The integrated Jsc values from the EQE-based devices for MAPbI3 before and after the NH4I treatment are calculated to be 19.9 and 23.1 mA cm2, respectively, which are quite consistent with the J–V curves.

The resistance (Rs) of the device mainly reflects the FF. Generally speaking, a lower Rs indicates a higher FF. To conform the Jsc as well as the FF, we got deep investigations in the resistance of the series by the typical equation (eq ),[62] where IL and I0 represent the photocurrent (n = 1) and dark saturation current (reverse polarization), respectively. q, T, K, and V are the values of the charging, absolute temperature, Boltzmann constant, and bias potential, respectively.

Based on MAPbI3 before and after NH4I treatment, the Rs values of the devices are calculated to be 9.55 and 4.6 Ω cm2. The change in Rs is likely to be related to the structure of the perovskite. In heterostructure solar cells, the grain boundaries have strong influences on performance parameters, and Rs increases with increasing numbers of grain boundaries. The 3D-MAPbI3 solar cells are ascribed to the low Rs, probably as a result of there being fewer grain boundaries.[1]

To study the characteristics of the charging transport, we recorded the time-resolved photoluminescence spectra (Figure a). For the pure MAPbI3 thin films before and after the NH4I treatment, we measured the lifetimes to be 14.2 and 18.1 ns, respectively. The longer lifetime in 3D-MAPbI3 thin films after the NH4I treatment suggests a longer survival time of the exciton. In addition, we calculated the charging transfer time (τCT) by a typical formula (formula ) τfilm = 1/τMAPbI3 + 1/τCT,[63] where the τfilm and τMAPbI3 are the decay lifetimes of MAPbI3 on TiO2 nanoarrays and pure MAPbI3 films, respectively. The τCT for MAPbI3 on TiO2 nanoarrays before the treatment is estimated to be 15.0 ns, while the 3D-MAPbI3 crystal after the NH4I treatment shows a short lifetime of 9.5 ns (Table ), which indicates that the electron injection between 3D-MAPbI3 and the layer of TiO2 nanoarrays is significantly facilitated, and the injection time is improved by 37.8%.

Figure 5

(a) Time-resolved photoluminescence measurements of the thin films; (b) XRD patterns of MAPbI3/TiO2 thin films under an air atmosphere for 30 days; (c, d) SEM images of the thin films before and after the NH4I treatment under an air atmosphere for 30 days.

Table 2

Photoluminescence Lifetimes Measured for Pure MAPbI3 Thin Films and 3D-MAPbI3 Thin Films on TiO2 Nanoarrays (τfilm) and the Estimated Transfer Time of the Charging Carrier (τCT)

thin films on glass/FTO	τfilm (ns)	τCT (ns)
MAPbI3 (before NH4I treatment)	14.2
MAPbI3 (before NH4I treatment)/TiO2	7.3	15.0
3D-MAPbI3 (after NH4I treatment)	18.1
3D-MAPbI3 (after NH4I treatment)/TiO2	6.2	9.4

We further investigated the stability of the thin films. We recorded the XRD patterns of the crystal under an air atmosphere for 30 days (Figure b). For both of the MAPbI3 crystals before and after the NH4I treatment, a new XRD diffraction peak at 12.6° (PbI2) has appeared, suggesting the decomposition of crystals. Although they are exposed to air for 30 days, the diffraction peak intensity of PbI2 in 3D-MAPbI3 crystals after the NH4I treatment is much lower than that of MAPbI3 crystals before the NH4I treatment, indicating that the 3D-MAPbI3 crystal after the NH4I treatment possesses better air stability. To confirm that, we tested the SEM images of the MAPbI3 crystal after 30 days to observe the morphology (Figure c,d). Obviously, crystal exposure to air for 30 days before the treatment (Figure c) showed a higher degree of decomposition due to a large number of grain boundaries because the crystal decomposition usually occurs at the grain boundaries. On the contrary, the NH4I-treated crystals have better morphology stability. This observation is quite consistent with the previous discussion.

Considering the advanced stability property of the 3D-MAPbI3 crystals, we remeasured the J–V curves of the devices exposed to air for 30 days (Figure a). The device based on 3D-MAPbI3 before the NH4I treatment showed a Voc of 0.99 V, a Jsc of 14.6 mA cm–2, a FF of 0.63, and a PCE of 9.3% after exposure to air for 30 days. The PCE is 66% of the original efficiency. By contrast, the devices based on 3D-MAPbI3 after the NH4I treatment showed an enhanced performance with a Voc of 1.01 V, a Jsc of 20.1 mA cm–2, a FF of 0.74, and a PCE of 15.1%, which possesses 78.3% of the original efficiency, even though they are exposed to air for 30 days, which is much higher than that of the device based on 3D-MAPbI3 after NH4I treatment.

Figure 6

(a) J–V curves of the devices after exposure to air for 30 days; the PCE retention rate of the devices before and after the NH4I treatment (b, c) after exposure to air for 30 days.

To further investigate the PCE retention rate of the solar cells after exposure to air for 30 days, we counted 30 devices and obtained the statistics of the PCE retention rate for 3D-MAPbI3 before and after the NH4 treatment. For the devices based on MAPbI3 before the NH4 treatment, the average PCE retention rate is calculated to be 65.6%, and the highest PCE retention rate is 69%, while the average PCE retention rate for the 3D-MAPbI3 devices after the NH4 treatment is estimated to be 78.1%, and the highest PCE retention is measured to be 85%. The advanced air stability of the NH4I-treated solar cells suggests the potential application of 3D-MAPbI3 in photovoltaic devices.

Conclusions

In summary, we prepared the 3D-MAPbI3 crystal with low grain boundaries on TiO2 nanoarrays as well as proposed and discussed the formation mechanism. The inhibition of ammonium iodide came up for the first time. Solar cells based on 3D-MAPbI3 after the NH4I treatment showed a modified performance as high as 17.6%, which increased by 36.8% in contrast to the counterpart. Facts proved that, for the device, based on NH4I-treated 3D-MAPbI3 films, the EQE (92%) is significantly improved. Moreover, the electron injection between the 3D-MAPbI3 layers after the NH4I treatment and TiO2 nanoarrays is remarkably facilitated. The total time of the electron injection decreased by 37.8% compared to the control group. In addition, based on 3D-MAPbI, solar cells showed excellent air stability, and the device possesses 78.3% of the initial PCE, even though it is exposed to air for 30 days. The devices based on NH4I-treated 3D-MAPbI3 showed an average PCE retention rate of 78.1%, much higher than the counterpart. In view of the above results, it is believed that the inhibition of ammonium iodide has positive effects on the phase transformation and film formation for perovskite crystals, which is adopted to produce advanced materials for fabrication of perovskite photovoltaic devices with high efficiency and stability.