Defect Passivation via Isoxazole Doping in Perovskite Solar Cells.

To improve perovskite solar cell (PSC) performance, which is deeply related to perovskite layer quality, researchers have explored numerous strategies. Additive doping into perovskite precursors has been widely used to improve the PSC performance. In this study, we used isoxazole-a Lewis-base small molecule-as an additive for the CH3NH3PbI3 (MAPbI3) precursor and explored how isoxazole effectively passivates defects in the perovskite structure. We found that isoxazole interacted with undercoordinated Pb2+ ions from an X-ray photoelectron spectroscopy survey and verified that isoxazole doping improved the device performance. When the optimized concentration of isoxazole was doped in the MAPbI3 precursor, the power conversion efficiency increased from 15.6 to 17.5%, with an improved fill factor and short-circuit current density. In addition, an isoxazole-doped device sustained 94% of its initial performance after 8 days under ambient air conditions (10 ± 5 RH %, 25 °C), whereas a device without isoxazole doping only maintained 64% of its initial performance.

Introduction

To solve the fuel energy problems, solar energy has been considered an alternative future energy source.[1] Perovskite solar cells (PSCs) have been widely researched for solar energy generation applications due to advantages such as cost-effective fabrication,[2] easy fabrication processes,[3] tunable band gaps,[4] long carrier lifetimes,[5] and high absorption coefficients.[6] In particular, the inorganic–organic hybrid PSCs based on CH3NH3PbI3 (MAPbI3) have been studied extensively.[7−16] The power conversion efficiency (PCE) of PSCs has increased tremendously from 3.9 to 25.7% over the past 11 years.[17,18] Furthermore, an inverted PSC with the p–i–n structure has attracted considerable attention due to its metal oxide-free layer and low processing temperature.[19]

Meanwhile, defects in perovskites are responsible for nonradiative recombination,[20] and thus, their passivation has attracted much attention for use in suppressing defect-induced PSC degradation. Many researchers have explored defect passivation through molecular interactions between defects and various functional groups.[15,21−26] Particularly, Lewis-base molecules contain N,[27−29] S,[27,28,30−32] O,[33−35] and P[36,37] and hence can donate nonbonding electrons, reacting with and passivating undercoordinated Pb2+ ions or Pb clusters, which act as a nonradiative recombination center and induce perovskite phase degradation by reacting with O2 and H2O.[38] Compared with large Lewis-base molecules, small Lewis-base molecules might easily diffuse into the interior of bulk perovskites and hence are efficient for bulk defect passivation, whereas they can easily escape from the crystal, unable to passivate defects. Therefore, it is crucial to discover small molecules that improve the performance of PSCs. Previous studies have reported that poly(1-vinyl-3-ethyl-acetate) imidazole tetrafluoroborate,[39] bithiophene-based n-type conjugated small molecules,[40] 4-dimethylaminopyridine,[41] and dicyandiamide[42] can serve as small-molecule additives for PSCs to improve their performance.

In this study, we used isoxazole—a small Lewis-base molecule—as an additive to the perovskite precursor for improving the PSC performance. Isoxazole is a weak base heterocyclic compound containing both N and O atoms,[43] whose molecular structure is shown in Figure S1, and it has been widely used in drug synthesis.[44,45] We explored how this compound played a role in the perovskite structure and demonstrated that a small amount of isoxazole doping enhanced the PSC performance and reduced nonradiative recombination by the defect passivation. Doping 0.4 M isoxazole into the perovskite precursor improved the PCE of the best device from 15.63 to 17.50% with a great increase in JSC. In addition, the isoxazole-added device retained 94% of its initial PCE under ambient air conditions (10 ± 5 RH %, 25 °C) after 8 day storage, indicating that it had better stability than the control device without isoxazole doping, which retained only 64% of its initial PCE.

Experimental Section

Materials

Greatcell Solar Materials Pty Ltd. (Queanbeyan, Australia) produced methylammonium iodide (MAI). Bathocuproine (BCP) and lead(II) iodide (PbI2) were obtained from TCI (Tokyo, Japan). Indium tin oxide (ITO)-patterned glass was obtained from Zhuhai Kaivo Optoelectronic Technology Co., Ltd. (Zhuhai, China). Sigma-Aldrich (St. Louis, MO, United States) provided chlorobenzene (CB), isoxazole, N,N-dimethylformamide (DMF), poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA), dimethyl sulfoxide (DMSO), Hellmanex III for detergent, and isopropyl alcohol (IPA). Nano-C Inc. (Westwood, MA, United States) provided [6,6]-phenyl-C61-butyric acid methyl ester (PCBM). All these chemical materials were used without any purification.

Device Fabrication

About 1.2 M PbI2 (553 mg) and 1.2 M MAI (190 mg) were added to DMF (700 μL) and DMSO (300 μL) for the precursor preparation. We added 0.2, 0.4, or 0.6 M isoxazole into the precursor solution for the experimental sample preparation, whereas no isoxazole was added for the control sample. All precursors were stored for 12 h at room temperature. For the hole-transport layer, 1.5 mg of PTAA was dissolved in 1 mL of toluene. For the electron transport layer, PCBM (20 mg/mL in CB) and BCP (1.5 mg/mL in IPA) were heated for 2 h on a hot plate at 60 °C. For device fabrication, ITO-coated glass was cleansed by ultrasonic treatment sequentially with a detergent, deionized water, and IPA for 25 min each. After cleaning, the substrates were treated with ultraviolet (UV) ozone plasma for 30 min. We coated the PTAA layer on a substrate via a spin-coating process at 4000 rpm for 33 s (initial 3 s for acceleration), followed by a heating procedure at 120 °C for 10 min. The perovskite layer was deposited on the PTAA-coated substrate at 1000 rpm for 10 s and 5000 rpm for 35 s, for which acceleration times were 1 and 5 s, respectively. 15 s after the second acceleration, CB (350 μL) was dripped on the substrate. After depositing the active layer, we put the sample on a hot plate at 60 °C for 30 s and 90 °C for 20 min. The next step was spin-coating PCBM (60 μL) and BCP (60 μL) at 1500 rpm for 30 s and 4000 rpm for 30 s, respectively, with an acceleration time of 5 s. From the PTAA deposition to BCP fabrication, all steps were performed in a glove box filled with nitrogen gas. As the last step, Ag (100 nm) deposition on the electrode was conducted using a thermal evaporator under high-vacuum conditions (<8.0 × 10–6 Torr).

Characterization

We employed scanning electron microscopy (SEM) to observe the surface morphologies using JSM-7500F (JEOL, Tokyo, Japan). We used SmartLab (Rigaku, Tokyo, Japan) to study the X-ray diffraction (XRD) pattern of our MAPbI3-coated films. The Fourier transform infrared (FTIR) spectra were recorded using a Spectrum Two instrument (PerkinElmer, MA, United states) with a resolution of 2 cm–1. The current density–voltage (J–V) curve results and space-charge-limited current (SCLC) measurement were obtained using a Keithley 2400 source meter (Keithley, Cleveland, OH, United States) under illuminated and dark conditions. The standard 1 sun illumination condition (AM 1.5 G, 100 mW/cm2) was controlled using HAL-320 (Asahi spectra USA, Torrance, CA, United States). To probe the incident photon-to-current efficiency (IPCE), we employed Solar Cell Scan 100 (Zolix Instruments, Beijing, China). For X-ray photoelectron spectroscopy (XPS), we used AXIS-Nova (Kratos Inc., San Diego, CA, United States) to assess the binding energy of each perovskite film surface molecule. Time-resolved photoluminescence (TRPL) and photoluminescence (PL) measurements were obtained using Fluorolog3 (HORIBA, Kyoto, Japan); the excitation wavelength for PL and TRPL was 467 nm and the emission wavelength for TRPL was 750 nm. UV–visible (UV–vis) spectroscopy analysis was obtained using Ultra-3660 (Rigol, Beijing, China). Electrochemical impedance spectroscopy (EIS) was performed using VersaSTAT3 (Ametek Scientific Instruments, Berwyn, PA, United States) under dark-room conditions. All these characterizations and analyses were obtained under ambient air conditions without the capsulation process. To estimate the average grain size, we calculate the average area of the grain by dividing the entire SEM image area by the total number of grains, assuming that the grain shape is a perfect circle (in two dimensions), and regard the diameter as the lateral size.

Results and Discussion

Figure a–d shows top-view SEM images of perovskite films with different isoxazole concentrations. SEM results show that the control film and 0.2–0.6 M isoxazole-doped films showed similar film compactness. Notably, isoxazole-doped films had larger average grain sizes than the control film of 241 nm; that is, 310, 342, and 319 nm for 0.2, 0.4, and 0.6 M isoxazole-doping concentrations, respectively. Films filled with a larger grain have fewer grain boundaries, which may decrease the nonradiative recombination of carriers.[46]

Figure 1

(a–d) Top-view SEM images of the MAPbI3 perovskite film while increasing the isoxazole-doping concentration from 0 to 0.6 M.

To explore the effect of isoxazole doping on the formation and crystallinity of perovskite phases, we measured the XRD patterns for perovskite films fabricated on ITO-patterned substrates. Figure a shows that all XRD patterns have two characteristic peaks of the perovskite structure at 14.10 and 28.42°, which are consistent with the (110) and (220) lattice planes, respectively.[47] Besides, the isoxazole-coated films do not have any new characteristic peaks compared with the control film, indicating that the additive did not induce a new phase. The 0.4 M isoxazole-doped perovskite film has the highest peak intensity value at those peaks, indicating that it has the best crystallinity. As shown in Figure S1, isoxazole—the Lewis base—has N and O atoms and hence can react with and passivate the undercoordinated Pb clusters or Pb2+ ions on the surface, which behave as charge traps,[48] by providing lone pair electrons. We performed XPS measurements to examine the perovskite film surface with and without isoxazole doping (Figure b). High-resolution XPS results of Pb 4f for the control film displayed two specific peaks at 137.90 and 142.78 eV, corresponding to Pb 4f7/2 and Pb 4f5/2,[49,50] whereas the isoxazole-doped film had two peaks at 138.09 and 142.95 eV, respectively. Both peaks shift to higher binding energies by 0.19–0.17 eV, indicating that isoxazole anchored and reacted with Pb ions on the perovskite surface.[51] In the 0.4 M isoxazole-doped sample, the intensities of characteristic peaks were lower than those in the control sample. This may be due to the fact that isoxazole molecules on the surface block some of the incident X-rays. However, as discussed above, XRD data indicated that the crystallinity of the perovskite film was best for the case with 0.4 M isoxazole, although some isoxazole molecules might be located at the film surface.

Figure 2

(a) XRD patterns and (b) XPS data for Pb 4f7/2 and Pb 4f5/2 spectra of control and 0.4 M isoxazole-doped films. (c) FTIR spectra of isoxazole with and without PbI2.

To further verify the interaction between isoxazole and Pb2+ ions, we carried out FTIR spectroscopy measurements for 100.0 μL of isoxazole with and without 0.5 mg of PbI2. As shown in Figure c, the stretching vibrations of C=N, N–O, and C–O bonds were 1652, 1557, and 1218 cm–1 in pure isoxazole[23,52,53] and were red-shifted to 1645, 1553, and 1213 cm–1 upon adding PbI2, respectively. The weakening of C=N, N–O, and C–O bonds might be due to the formation of coordinate bonds between N or O atoms in isoxazole and Pb2+ ions in PbI2, implying the attractive interaction between isoxazole and MAPbI3, similar to previous studies.[54]

To further investigate the isoxazole-doping effect, we measured the UV–vis absorption spectra of the control and isoxazole-doped perovskite films deposited on glass substrates (Figure a). Figure a shows that all films have almost the same absorption edge at ca. 770 nm, and the Tauc plot in Figure b shows that the band gaps of films are almost the same, that is, 1.56 eV, which is similar to the theoretically calculated band gap of the MAPbI3 perovskite (1.5–1.6 eV).[55] This indicates that isoxazole doping slightly affects the band gap, which agrees with the SEM and XRD results, where no noticeable changes upon isoxazole doping were observed in the structure and morphology. The light absorption (Figure a,b) and steady-state PL (Figure c) for 0.4 M isoxazole doping were the highest among the considered films, indicating that optimized isoxazole doping can reduce defects and improve the crystallinity of the perovskite[56,57] and hence may enhance the PSC performance. To further explore the isoxazole-doping effect on charge carrier recombination dynamics, we performed the TRPL measurement (Figure d). The TRPL spectrum value can be obtained from a second-order exponential decay formula: , where A1 and A2 denote the decay amplitudes and τ1 and τ2 denote the fast and slow decay times, respectively.[58] The 0.4 M isoxazole-doped film had a longer average PL decay time of 333.9 ns than the control film (282.3 ns) (Table S1), indicating that isoxazole doping can reduce the defects associated with nonradiative recombination.

Figure 3

(a) UV–vis absorption spectra, (b) Tauc plots, and (c) PL spectra of perovskite films with increasing doping concentration. (d) TRPL spectra of the control and 0.4 M isoxazole-doped films.

We fabricated PSCs (Figure a) with a configuration of glass/ITO/PTAA/perovskite with or without isoxazole/PCBM/BCP/Ag with a 0.04 cm2 active area to explore their photovoltaic performance. Figure S2 shows the cross-sectional SEM images of the device without isoxazole doping (the control device) and the 0.4 M isoxazole-doped device. The perovskite layer thicknesses of both devices are approximately 520 nm. Figure b shows that the energy band of each layer is well aligned for electron and hole transports, where the values of energy bands were obtained from previous studies.[59,60]Figure c shows the J–V characteristic curve of the best device for each experimental group; the 0.4 M isoxazole-doped device showed a photovoltaic performance with a PCE of 17.5%, whereas the control device achieved such with a PCE of 15.6%. The J–V curves of all different concentrations are shown in Figure S3, and the details of the photovoltaic parameters are shown in Table S2. The 0.4 M isoxazole-doped device had a 1.4 mA/cm2 higher average short-circuit current (JSC) than the control device, attributable to the suppression of electrically active defects responsible for nonradiative recombination through the reaction between isoxazole and Pb ions.[61]Figure S4 and Table S3 show the hysteresis results of each device and the detailed data under reverse scan (from Voc to 0 V) and forward scan (from 0 V to Voc). The hysteresis index (HI), which is given by ,[62] reduced from 0.04 to 0.02 using an isoxazole additive, where PCEreverse and PCEforward indicate the power conversion efficiencies measured under reverse and forward bias conditions, respectively. This is also explained by defect passivation by isoxazole.[63,64] As shown in Figure S5 and Table S2, after adding isoxazole, the standard deviation of PCE decreased, implying that using isoxazole is beneficial for obtaining better device uniformity in mass production. The 0.4 M isoxazole-doped device showed the highest average PCE of (17.1%), which was 1.9% higher than that (15.2%) of the control device. The increases in JSC and fill factor (FF) are mainly responsible for the PCE improvement, whereas the change in VOC is negligible upon isoxazole doping. The average PCE for 0.6 M isoxazole doping was 16.3%, which decreased by 0.8% compared with that for the 0.4 M doping group. This is consistent with smaller perovskite grain sizes and worse crystallinity for 0.6 M isoxazole doping than those for 0.4 M isoxazole doping (Figures and 2a). As seen from Figure d, the 0.4 M isoxazole-doped device shows a higher IPCE than the control device at wavelengths between 350 and 800 nm. The JSC values obtained by integration of IPCE are 20.5 and 18.4 mA/cm2 for the 0.4 M isoxazole-doped and control devices, respectively. These values are similar to the measured JSC in Figure c, within 1.4%, indicating the high reliability of our results. EIS was performed under dark conditions to explore the carrier transfer process. Figure e shows the corresponding Nyquist plot and equivalent circuit that includes Rs and Rrec; Rs comprises the electrode and contact resistances between each layer,[65] and Rrec is the charge recombination resistance.[66]Figure e and Table S4 show that the Rrec value (8369 Ω) of the 0.4 M isoxazole-doped device is significantly larger than that of the control device (5966 Ω), showing suppressed charge recombination upon isoxazole doping. To further probe into the charge recombination in the devices, we measured the light intensity dependence of VOC for each device. The VOC dependence on the light intensity (I) is given by VOC = nkTln(I)/e + constant, where T, e, k, and n are the absolute temperature, electron charge, Boltzmann constant, and ideality factor, respectively.[67,68] From the Shockley–Read–Hall recombination kinetics, a steeper slope (n > 1) denotes a higher trap state density of a device.[69]Figure f shows that the 0.4 M isoxazole-doped device has a lower slope (1.27 kT/e) than the control device (1.52 kT/e). The improved Rrec and the reduced n indicate that isoxazole doping can passivate charge traps in PSCs.

Figure 4

(a) Schematic structure and (b) energy band alignments of PSCs. (c) J–V reverse scan curves, (d) IPCE and the integrated current density (integrated J) from the IPCE measurement using the AM 1.5 G photon flux spectrum, (e) EIS measurements, and (f) VOC dependence on the light intensity of the control and 0.4 M isoxazole-doped devices.

We obtained the SCLC to measure the trap density (ntrap) of each device comprising ITO/perovskites with and without the isoxazole/Ag structure under dark conditions. Figure a,b shows a linear Ohmic-type response at low bias voltages; if this voltage is greater than the kink point known as the trap-filled limit voltage (VTFL), JSC starts to show a significantly higher slope increase with voltage. VTFL is given by VTFL = ,[70] where ε, ε0, e, ntrap, and L denote the relative dielectric constant, vacuum permittivity, elementary charge, trap density, and thickness of the perovskite, respectively; ε = 32.[71] The thickness of the perovskite film was 520 nm, as discussed above. The calculated value of the trap density of the 0.4 M isoxazole-doped device was 1.40 × 1015 cm–3—lower than that of the control device (1.76 × 1015 cm–3). This agrees with the results of crystallinity, EIS, and light-intensity-dependent VOC measurements and indicates that the isoxazole additive passivates defect-induced charge traps. Figure c shows the stabilized power output (SPO) and photocurrent density (JSPO) at the maximum power point voltage (Vmpp). The 0.4 M isoxazole-doped device showed a more stable SPO of 16.9% and a JSPO of 19.7 mA/cm2 than those of the control device (14.6% and 18.0 mA/cm2, respectively). We stored the devices without capsulation in ambient air to investigate PCE stability. As shown in Figure d, the normalized PCE of the 0.4 M isoxazole-doped device remained over 94% of the initial PCE after 8 days, whereas the control device retained only 64% of its initial PCE. It demonstrates that the isoxazole doping passivated the defects such as unsaturated Pb ions, which have coordinate bonds with O2 and H2O and hence cause perovskite film degradation.[38,72,73] The 0.4 M isoxazole-doped cell also showed the enhanced light soaking stability under illumination (see Figure S6), which was measured by a method similar to that in previous studies.[74,75] Our results indicate that isoxazole is an effective small-molecule additive for improving the efficiency and stability of inverted PSCs based on MAPbI3. Further study is required to prove that isoxazole additives also improve the performances of different kinds of PSCs.

Figure 5

Dark J–V curves of the (a) control device and (b) 0.4 M isoxazole-doped device; (c) stabilized photocurrent densities and power output; and (d) stability measurement of the devices with and without isoxazole doping.

Conclusions

In summary, we improved the PCE of the optimal PSC from 15.6 to 17.5% via a small Lewis-base molecule isoxazole-doping technique. We found that isoxazole reacted with undercoordinated Pb2+ ion defects via XPS measurements, which agrees well with the reduction of electrically active charge traps by an isoxazole additive observed by TRPL, EIS, VOC versus light intensity, and SCLC measurements. Furthermore, the isoxazole additive increased the PSC stability under ambient air conditions. Our results indicate that isoxazole is one of the promising small-molecule additives that may effectively passivate defects inside and near interfaces.