A Breakthrough Efficiency of 19.9% Obtained in Inverted Perovskite

coefficient and tunable direct band gap.1-5 The power conversion efficiency ..... Figure 4 illustrates the representative device architecture of the i...
1 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/JACS

A Breakthrough Efficiency of 19.9% Obtained in Inverted Perovskite Solar Cells by Using an Efficient Trap State Passivator Cu(thiourea)I Senyun Ye,†,⊥ Haixia Rao,†,⊥ Ziran Zhao,† Linjuan Zhang,‡ Hongliang Bao,‡ Weihai Sun,† Yunlong Li,† Feidan Gu,† Jianqiang Wang,*,‡ Zhiwei Liu,*,† Zuqiang Bian,*,† and Chunhui Huang† †

Beijing National Laboratory for Molecular Sciences, State Key Laboratory of Rare Earth Materials and Applications, College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, P. R. China ‡ Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai 201800, P. R. China S Supporting Information *

ABSTRACT: It is extremely significant to study the trap state passivation and minimize the trap states of perovskite to achieve high-performance perovskite solar cells (PSCs). Here, we have first revealed and demonstrated that a novel p-type conductor Cu(thiourea)I [Cu(Tu)I] incorporated in perovskite layer can effectively passivate the trap states of perovskite via interacting with the under-coordinated metal cations and halide anions at the perovskite crystal surface. The trap state energy level of perovskite can be shallowed from 0.35−0.45 eV to 0.25−0.35 eV. In addition, the incorporated Cu(Tu)I can participate in constructing the p−i bulk heterojunctions with perovskite, leading to an increase of the depletion width from 126 to 265 nm, which is advantageous for accelerating hole transport and reducing charge carrier recombination. For these two synergistic effects, Cu(Tu)I can play a much better role than that of the traditional p-type conductor CuI, probably due to its identical valence band maximum with that of perovskite, which enables to not only lower the trap state energy level to a greater extent but also eliminate the potential wells for holes at the p−i heterojunctions. After optimization, a breakthrough efficiency of 19.9% has been obtained in the inverted PSCs with Cu(Tu)I as the trap state passivator of perovskite.

■

INTRODUCTION Organometallic halide perovskite as a promising photovoltaic material has invoked a worldwide upsurge of research interest in recent years due to its remarkable advantages such as low exciton binding energy, high charge carrier mobility, large optical absorption coefficient, and tunable direct band gap.1−5 The power conversion efficiency (PCE) of perovskite solar cells (PSCs) has been unprecedentedly rapidly boosted from the initial 3.8% to over 20% by now.6−11 In improving the performance of PSCs, a series of advanced engineerings have been intensely studied and applied. The interface,12−22 material7−9,23−26 and fabrication10,27−31 engineerings have particularly drawn widespread and sufficient attention, and many significant breakthroughs are eventually achieved. In addition, the trap state passivation is also worthwhile and of great importance because the trap states in perovskite layer can induce serious charge carrier recombination and thus greatly restrict the performance of PSCs.32−38 Recently, an increasing number of researches have focused on the trap state passivation.39−54 For example, Huang et al.39,40 first demonstrated that the fullerene layers deposited on perovskite can effectively passivate the charge trap states and eliminate the notorious photocurrent hysteresis. Sargent and Wu et al.41,42 directly integrated the fullerene into perovskite layer to expand the extent of trap state passivation and greatly improved the device performance. Simultaneously, Snaith et al.43,44 found © 2017 American Chemical Society

that iodopentafluorobenzene (IPFB) and Lewis bases, such as thiophene, were effective for passivating the trap states on the surface of perovskite layer due to their interaction with the under-coordinated halide anions and metal cations acting as the charge carrier recombination centers. Furthermore, Al3+ and nitrogen-doped reduced graphene oxide (N-RGO)48,50 were even proved their merits in trap state passivation. Very recently, we49 also demonstrated the availability of passivating trap states via incorporating the hole-conductor CuSCN into perovskite layer. Though effective PSCs have been obtained via the trap state passivation, the mechanism is still complicated, and excellent trap state passivators (TSPs) are insufficient, which deserve to be further explored and studied. Here, we have first discovered that a novel p-type conductor Cu(thiourea)I [Cu(Tu)I] can function as a highly efficient TSP of perovskite, and the possible trap state passivation mechanism has also been revealed. It was demonstrated that Cu(Tu)I is able to not only shallow the trap state energy level of perovskite via interacting with the under-coordinated metal cations and halide anions at perovskite crystal surface but also establish the bulk heterojunction with perovskite, thus effectively accelerate hole transport and reduce charge carrier recombination. Moreover, Cu(Tu)I can play a much better role than that of Received: February 10, 2017 Published: May 14, 2017 7504

DOI: 10.1021/jacs.7b01439 J. Am. Chem. Soc. 2017, 139, 7504−7512

Article

Journal of the American Chemical Society

bond. Similarly, the νCS absorption peak of the as-prepared Cu(Tu)Cl further blue-shifts to 1118 cm−1 (Figure 2b), which is consistent with the results of the literature,55,56 probably caused by the different Cu−X (Cl, I) bond. In addition, the band gaps and VBM of Cu(Tu)I and Cu(Tu)Cl are different with that of CuI. The band gaps of CuI, Cu(Tu)I, and Cu(Tu) Cl extracted from their ultraviolet visible (UV−vis) absorption spectra (Figure 2c) are respectively 2.95, 3.56, and 3.66 eV, and their VBM estimated from their photoelectron spectra (Figure 2d) are −5.21, −5.45, and −5.60 eV, respectively. It is noted that the VBM of Cu(Tu)I is the same with that of the perovskite (−5.45 eV, Figure S1), which is probably a crucial reason for the huge trap state passivation capability of Cu(Tu)I, and more detailed discussion is below. The perovskite−TSP hybrid materials were prepared by directly spin-coating the corresponding hybrid precursor solution of both perovskite (MAPbI3−xClx, MA = CH3NH3, x ≈ 0.2) and the TSP additive (CuI, Cu(Tu)I or Cu(Tu)Cl). To confirm whether the TSP additive could remain in perovskite layer during the film preparation process, the inductively coupled plasma-atomic emission spectroscopy (ICP-AES) analysis for the perovskite-TSP hybrid films was first carried out. The results (Table S2) show that the Cu element in the asprepared hybrid films is assuredly detectable, and the mole ratios of Cu to Pb in the films are similar to those in the precursor solution. Furthermore, almost no offset of the main characteristic XRD peaks for (110) and (220) planes of perovskite is observed in the as-prepared perovskite−TSP hybrid films (Figure 3a, b), suggesting that the TSP additives do not participate in constructing the perovskite frameworks, which is accordant with the identical absorption onset wavelengths of the perovskite films without and with TSP (Figure S2). On the other hand, the characteristic peak at 2θ ≈ 25.44°, assigned to the (111) plane of CuI, can be clearly

the traditional p-type conductor CuI in both the two synergistic effects, probably on account of its identical valence band maximum (VBM) with that of perovskite, which is able to not only shallow the trap state energy level to a greater extent, but also eliminate the potential wells for holes at the p−i heterojunction. After optimization, a breakthrough PCE of 19.9% in the inverted PSCs with Cu(Tu)I as the TSP has been achieved.

■

RESULTS AND DISCUSSION The detailed synthesis process of Cu(Tu)I is described in the Experimental Section. As a comparison, Cu(Tu)Cl was also synthesized. The results of elemental analysis (EA) shown in Table S1 indicate that the synthesized Cu(Tu)I or Cu(Tu)Cl is 1:1 complex of thiourea and CuI or CuCl. The crystal structure of Cu(Tu)I is shown in Figure 1, and more detailed

Figure 1. Crystal structure of Cu(Tu)I. Hydrogen atoms have been omitted for clarity.

information is available in CIF format in the Supporting Information. The crystal structure of Cu(Tu)Cl is unavailable due to inability of growing its single crystals. But, Cu(Tu)Cl could be confirmed via its powder X-ray diffraction (XRD) pattern (Figure 2a), which is in agreement with that of the literature.55 The Fourier transform infrared (FTIR) absorption peak of CS vibrational mode (νCS) of the as-prepared Cu(Tu)I located at 1097 cm−1 is slightly blue-shifted compared with that of the pristine thiourea (1085 cm−1) (Figure 2b), which is probably attributed to the formation of the Cu−S

Figure 3. (a) XRD patterns of the pristine MAPbI3−xClx film and the perovskite−TSP hybrid film prepared from the mixed precursor solution with CuI, Cu(Tu)I or Cu(Tu)Cl as the TSP additive, and the corresponding zoomed-in patterns in the regions (b) 13−29° and (c) 25−26°. (d) XRD patterns of the pristine MAPbI3 film and the perovskite−TSP hybrid films are prepared from the mixed precursor solution with different mole ratios of MAPbI3 to Cu(Tu)Cl. (e) FTIR spectra of the pristine MAPbI3−xClx and the perovskite−TSP hybrid material prepared from the mixed precursor solution with Tu, Cu(Tu)I, or Cu(Tu)Cl as the additive, and (f) the corresponding zoomed-in spectra in the region 1000−1200 cm−1.

Figure 2. (a) XRD patterns, (b) FTIR spectra, (c) UV−vis absorption spectra, and (d) photoelectron spectra of Cu(Tu)I and Cu(Tu)Cl, with CuI or Tu as a reference. 7505

DOI: 10.1021/jacs.7b01439 J. Am. Chem. Soc. 2017, 139, 7504−7512

Article

Journal of the American Chemical Society identified in the enlarged XRD pattern of the perovskite−CuI hybrid film (Figure 3c), which indicates that CuI successfully remained in the perovskite layer. However, almost no obvious characteristic peaks of Cu(Tu)I could be observed in the XRD pattern of the as-prepared perovskite−Cu(Tu)I hybrid film, which is probably attributed to the low crystallinity of Cu(Tu)I during the film preparation process. It is interesting to note that the intensity of the characteristic XRD peaks of MAPbI3 decreases and that of MAPbCl3 increases simultaneously when enhancing the mole ratio of Cu(Tu)Cl to MAPbI3 in the hybrid precursor solution (Figure 3d), revealing the subsistent exchange of Cl in Cu(Tu)Cl and I in MAPbI3 during the film preparation process. Moreover, the νCS absorption peaks of both the perovskite−Cu(Tu)I and perovskite−Cu(Tu)Cl hybrid films are located at 1099 cm−1 (Figure 3e, f), which is much lower than that of Cu(Tu)Cl (1118 cm−1) and also higher than that of the perovskite− thiourea hybrid material (1085 cm−1), but quite approximates to that of Cu(Tu)I (1097 cm−1). This indicates that Cu(Tu)I probably kept its original structure, but Cu(Tu)Cl would actually transform into Cu(Tu)I due to the exchange of Cl and I in the hybrid films. To further ascertain the existence of Cu(Tu)I in both the perovskite−Cu(Tu)I and perovskite− Cu(Tu)Cl hybrid films, the extended X-ray absorption fine structure (EXAFS) analysis was performed. As shown in Figure S3 and Table S3, the Fourier-transformed (FT) EXAFS spectra of both the perovskite−Cu(Tu)I and perovskite−Cu(Tu)Cl hybrid materials could be well-fitted by using the Cu−S and Cu−I coordination shells, and their fitted spectra and parameters are similar to those of Cu(Tu)I, but different with those of Cu(Tu)Cl. Combining with the aforementioned XRD and FTIR characterization, we can conclude that Cu(Tu)I could remain unchanged in the perovskite−Cu(Tu)I hybrid film, while Cu(Tu)Cl would transform into Cu(Tu)I on account of the exchange of Cl and I in the hybrid film. To explore the distribution of the TSPs in perovskite layers, the profiles of Cu element determined by the energy dispersive X-ray (EDX) line scan across the cross sections of the perovskite-TSP hybrid films are shown in Figure S4. It is noted that the intensity of Cu element in the perovskite−TSP hybrid layers is relatively higher than that in the other region, which reveals that the TSPs are probably dispersed throughout the entire hybrid layers. Moreover, the TSPs probably fill in the interstices between the perovskite grains, which can be inferred from the detected larger intensity of Cu element in the interstice than that in the perovskite grain (Figure S5). Importantly, such a distribution of TSPs would make it possible for the TSPs to effectively passivate the trap states in perovskite layer. Figure 4 illustrates the representative device architecture of the inverted PSCs with Cu(Tu)I as the TSP, which consists of an indium tin oxide (ITO) layer deposited on a glass substrate as the transparent electrode, a MAPb3−xClx−Cu(Tu)I hybrid layer (about 300 nm) as the light absorber with trap states passivated, a 40 nm C60 layer as the electron conductor, an 8 nm bathocuproine (BCP) layer as the hole blocker, and a 100 nm silver (Ag) layer as the back electrode. Likewise, the reference PSCs without TSP and with CuI or Cu(Tu)Cl as the TSP additive possess a similar device configuration. The photovoltaic characteristics of the PSCs are summarized in Table 1. The reference PSC without TSP possesses quite poor photovoltaic performances including an average PCE of just 8.9% (11.1% for maximum) due to the low open-circuit voltage

Figure 4. (a) Cross-sectional scanning electron microscope (SEM) image of the device with Cu(Tu)I as the TSP and (b) the corresponding schematic illustration of the device architecture.

(VOC), short circuit current density (JSC), and fill factor (FF), which is in accordance with the results of most literature.57−60 Interestingly, the average PCE was improved to 14.3% (14.6% for maximum) effectively when utilizing CuI as the TSP with an additive concentration of 10 mmol L−1 in the hybrid precursor solution. Surprisingly, the average PCE was further boosted over 18% with Cu(Tu)I incorporated in perovskite layer to passivate the trap states with an identical additive concentration in the hybrid precursor solution. As predicted, Cu(Tu)Cl as another TSP additive with a same additive concentration in the hybrid precursor solution could also enhance the average PCE exceeding 18%, which is probably attributed to the trap state passivation for perovskite layer by the same Cu(Tu)I formed during the film preparation process due to the exchange of Cl and I, as demonstrated above. The maximum PCE of the PSCs with Cu(Tu)I as the TSP reached 20.0% measured in the reverse scan direction including a VOC of 1.12 V, a JSC of 22.3 mA cm−2, and a FF of 79.8%, which is almost consistent with the PCE measured in the forward scan direction (19.9%) including a VOC of 1.13 V, a JSC of 22.3 mA cm−2, and a FF of 78.9%, suggesting negligible J−V hysteresis (Figure 5a). The integrated JSC extracted from the incident photon to current conversion efficiency (IPCE) spectrum approximates to the measured JSC (Figure 5b), and the maximum steady-state PCE measured at a bias of 0.93 V under 100 mW cm−2 AM 1.5G irradiation is also as high as 19.9% (Figure 5c), which is almost the same with the measured PCE. Furthermore, the maximum measured PCE has also been certificated by an independent institute (Photovoltaic and Wind Power Systems Quality Test Center, Chinese Academy of Sciences) (see the appendix in Supporting Information), and the certificated PCE (19.9% in reverse scan direction and 19.8% in forward scan direction) is almost the same with the measured one, which indicates a good reliability of the measured results. To explore the reason why the inverted PSCs with Cu(Tu)I as the TSP of perovskite layer exhibited the best outstanding photovoltaic performance, the transient state photoluminescence (PL) spectra (Figure 5d) of the perovskite films without and with TSP deposited on glass substrates were performed to qualitatively compare the degree of trap state passivation. It was found that the PL lifetime of the pristine MAPbI3−xClx film was just about 182 ns calculated by using the single exponential fitting, which would sharply increase to 336 ns when using CuI as the TSP. Surprisingly, the PL lifetime could be further improved over 600 ns by passivating the trap states with Cu(Tu)I prepared from the hybrid precursor solution containing either Cu(Tu)I (611 ns) or Cu(Tu)Cl (698 ns). This indicates that the trap state passivation capability of Cu(Tu)I is much huger than that of CuI. In addition, the hole mobilities of the perovskite films with trap states passivated or not were also analyzed via the space-charge limited current 7506

DOI: 10.1021/jacs.7b01439 J. Am. Chem. Soc. 2017, 139, 7504−7512

Article

Journal of the American Chemical Society Table 1. Photovoltaic Parameters of the PSCs without and with TSP Additives TSP additive

VOC (V)

without CuI Cu(Tu)I Cu(Tu)Clb

± ± ± ±

1.00 1.01 1.11 1.11

0.00 0.01 0.01 0.01

JSC (mA cm−2) a

20.5 21.1 21.6 21.7

± ± ± ±

FF (%)

0.3 0.2 0.2 0.2

47.6 67.6 77.3 77.6

± ± ± ±

PCE (%)

5.7 1.7 1.1 1.1

8.9 14.3 18.5 18.6

± ± ± ±

2.7 0.3 0.2 0.4

a

The statistical data including average values and standard deviations are calculated from at least 10 separate devices. bCu(Tu)Cl as the TSP additive in the hybrid precursor solution would transform into Cu(Tu)I during the film preparation process due to the exchange of Cl and I.

Figure 5. (a) J−V curves of the device with Cu(Tu)I as the TSP in the perovskite layer measured in the forward and reverse scan directions. (b) The corresponding IPCE spectrum of the device. (c) Steady-state JSC and PCE of the device measured at a bias of 0.93 V under 100 mW cm−2 AM 1.5G irradiation. (d) Transient state PL spectra of the pristine MAPbI3−xClx film and the perovskite−TSP hybrid film deposited on glass substrates prepared from the mixed precursor solution with CuI, Cu(Tu)I, or Cu(Tu)Cl as the TSP additive. (e) J−V curves of the hole-only devices ITO/ PEDOT:PSS/MAPbI3−xClx (without or with TSP)/Au. (f) Rrec of the devices ITO/MAPbI3−xClx (without or with TSP)/C60/BCP/Ag. 9

V2

recombination of charge carriers in perovskite layer, and Cu(Tu)I plays a much better role than that of CuI. To reveal the passivation effect of the TSPs more directly and clearly, the admittance spectroscopy (AS) was conducted on the devices without and with TSP. AS is an effective technique for estimating both the energy level of trap states and the distribution of trap state density, which has been extensively applied to many photovoltaic systems, such as organic solar cells,64 Cu2ZnSnS4 solar cells,65 and PSCs.39,40,66−69 As the literature reported,66 for a p-type perovskite semiconductor, the defect activation energy (Ea) is approximately the depth of the trap state energy level (ET) relative to the VBM energy level (EVBM) of perovskite (Ea = ET − EVBM). Ea and the characteristic transition angular frequency (ω0) can be

(SCLC) model with the Mott−Gurney law ( J = 8 ε0εμ d3 )5 based on the hole-only devices {ITO/PEDOT:PSS [poly(3,4ethylenedioxythiophene):polystyrenesulfonate]/MAPbI3−xClx (without or with TSP)/Au}. As shown in Figure 5e, it can be inferred that the hole mobility of the perovskite−Cu(Tu)I hybrid film prepared from the hybrid precursor solution with either Cu(Tu)I or Cu(Tu)Cl as the TSP additive is relatively higher than that of the perovskite−CuI hybrid film by qualitatively contrasting the SCLC regions (J ∝ V2) in the J− V curves of the hole-only devices, which further demonstrates the huger trap state passivation capability of Cu(Tu)I than that of CuI. To further verify and contrast the role of TSPs on the device performance, the electrochemical impedance spectroscopy (EIS) was carried out to evaluate the charge carrier recombination resistance (Rrec) of the corresponding devices. The Nyquist plots shown in Figure S6a−e exhibit two diacritical characteristic arcs (a high-frequency one and a lowfrequency one), probably ascribed to the resistance of selective contacts and the Rrec of charge carriers, respectively, as most literatures reported,61−63 which can be well-fitted with the equivalent circuit model shown in Figure S6f. The detailed fitting parameters for the representative Nyquist plots measured at 0 V are listed in Table S4. As a result, the extracted Rrec of the device with Cu(Tu)I as the TSP prepared from the hybrid precursor solution with either Cu(Tu)I or Cu(Tu)Cl as the TSP additive is much higher than that of the device using CuI as the TSP, which is yet relatively higher compared with that of the device without TSP (Figure 5f). This reveals that the trap state passivation by TSPs can effectively decrease the

( ) E

expressed in the relation ω0 = βT 2 exp − k Ta , where β is a B

temperature dependent parameter, T is the temperature and kB is the Boltzmann’s constant. The ω0 is determined by the derivative of the capacitance−frequency spectrum. According to ω0

( ) = ln β −

this equation, the Arrhenius plot (ln

T2

Ea ) kBT

is

fitted, and the value of Ea can be obtained from the slope of the Arrhenius plot line. The distribution of trap state density can be V dC ω derived from the equation, 6 6 NT(Eω) = − qWbi dω k T , B

βT 2 ω

( ), where V

Eω = kBT ln

bi

is the built-in potential, W is

the depletion width, q is the elementary charge, C is the capacitance, and ω is the applied angular frequency. Vbi and W can be extracted from the Mott−Schottky analysis through the 7507

DOI: 10.1021/jacs.7b01439 J. Am. Chem. Soc. 2017, 139, 7504−7512

Article

Journal of the American Chemical Society

Figure 6. Admittance spectra of (a) the reference PSC without TSP and the device with (b) CuI or (c) Cu(Tu)I as the TSP, measured at temperature between 230 and 300 K with a step of 10 K. (d) The corresponding Arrhenius plots of the characteristic transition frequencies determined from the derivative of the admittance spectra. (e) Mott−Schottky analysis at 1 kHz for the devices. The calculated Vbi are 0.719, 0.753, and 0.865 V, and the corresponding W values are 126, 195, and 265 nm, respectively. (f) Trap state density (NT) of the devices measured at 290 K.

Figure 7. Schematic illustration of possible mechanism for the trap state passivation. The size of graphic atoms is adopted randomly according to aesthetics.

the Mott−Schottky analysis was performed at a frequency of 1 kHz with the bias potentials ranging from 0 to 1.4 V. The results shown in Figure 6e suggest that the Vbi of the reference PSC without TSP and the device with CuI or Cu(Tu)I as the TSP are 0.719, 0.753, and 0.865 V, and the corresponding W values are 126, 195, and 265 nm, respectively. Then, the distributions of trap state density in these devices could be calculated, and the results are illustrated in Figure 6f. It has been found that the trap state energy level in the reference PSC without any TSP is relatively deep (0.35−0.45 eV), and the trap state density is also slightly large. As predicted, both CuI and Cu(Tu)I can effectively passivate the trap states, mainly in shallowing the trap state energy level and reducing the trap state density to a certain extent. It is noticed that Cu(Tu)I, compared to CuI, can pull down the initial deep trap state energy level to a much shallower one (0.25−0.35 eV vs 0.30− 0.40 eV), indicating the relatively huger trap state passivation capability of Cu(Tu)I than that of CuI. Herein, the possible mechanism for the trap state passivation is illustrated in Figure 7. As previous literature reported,43,44 the trap states of perovskite mainly derive from under-coordinated metal cations and halide anions at crystal surface. The resulting relatively deep trap state energy level would cause severe potential wells at perovskite grain boundaries for charge carriers

capacitance−voltage measurement. According to the depletion approximation,70 the C, Vbi, and W at the junction can be expressed in the relation,

C A

=

εε0 W

=

qεε0N 2(Vbi − V )

, where A is the

active area, ε is the static permittivity of perovskite,71 ε0 is the permittivity of free space, N is the apparent doping profile in the depleted layer, and V is the applied bias. A Mott−Schottky A2

plot ( C 2 =

2(Vbi − V ) ) qεε0N

describes a straight line where the

intersection on the bias axis determines Vbi and the slope gives the impurity doping density N. Then, the depletion width W=

2εε0Vbi qN

corresponding to the zero bias can be calculated.

Figure 6a−c displays the temperature-dependent admittance spectra of the devices without and with TSP measured at various temperatures (T = 230−300 K) in the dark from 10 to 105 Hz. A small alternating current (AC) voltage of 20 mV was used, and the direct current (DC) bias was kept at zero during the measurement. The corresponding Arrhenius plots of the characteristic transition frequencies determined from the derivative of the admittance spectra are shown in Figure 6d, and the calculated Ea of the pristine perovskite, perovskite− CuI, and perovskite−Cu(Tu)I are 0.388, 0.356, and 0.318 eV, respectively. To estimate the Vbi and W values of the devices, 7508

DOI: 10.1021/jacs.7b01439 J. Am. Chem. Soc. 2017, 139, 7504−7512

Article

Journal of the American Chemical Society

representative example, the optimization for the Cu(Tu)I concentration in the perovskite−Cu(Tu)I hybrid film prepared from the hybrid precursor solution with Cu(Tu)Cl as the TSP additive is shown in Table S5. The optimum additive concentration is about 10 mmol L−1, and a much higher concentration would increase the device series resistance and thus decrease FF and PCE of the device, probably due to the dominant negative effect on the electron transport. While a lower concentration of Cu(Tu)I could not fully exert its positive effect, which would also restrict the device performance. Furthermore, the influence on the device performance induced by different surface treatments for ITO substrates[without or with ultraviolet−ozone (UVO) treatment] was studied. The result (Table S6) indicates that the UVO treatment for ITO substrates has almost no effect on the performance of the device with Cu(Tu)I as the TSP, despite that the UVO treatment can improve the work function of ITO substrates from 4.85 to 5.13 eV (Figure S8). The surface morphologies of the perovskite−TSP hybrid films prepared from the mixed precursor solution with CuI, Cu(Tu)I, or Cu(Tu)Cl as TSP additive were also investigated. As shown in Figure S9a−d, it was found that the average perovskite particle size of the perovskite−Cu(Tu)I hybrid film is smaller than that of both the pristine perovskite film and the perovskite−CuI hybrid film, indicating more grain boundaries acting as trap state sites in the perovskite−Cu(Tu)I hybrid film, which is disadvantageous for improving the device performance. This also suggests that the outstanding performance of the devices with Cu(Tu)I as the TSP is not ascribed to the morphology variation but attributed to the huge trap state passivation of Cu(Tu)I and the formation of efficient perovskite−Cu(Tu)I bulk heterojunction, as demonstrated above. To further improve the robustness of the devices with Cu(Tu)I as the TSP, solvent (N,N-dimethylformamide, DMF) annealing treatment was performed as described in our previous work.49 As a result, the average PCE of the corresponding devices was successfully enhanced from 18.5% (traditional annealing) to 19.0% (solvent annealing) (Table S7), which is probably due to the increase of the perovskite particle size (Figure S9e). In addition, the stability of the device with Cu(Tu)I as the TSP was investigated. As shown in Figure S10a, the device could keep stable maximum power point (MPP) output under continuous 100 mW cm−2 AM 1.5G irradiation in the first 10 h, but the output power gradually declined to 89% of the initial in the later 14 h. More seriously, the device performance shrunk by almost half after test for 7 days under simulated 1 sun illumination (6 h per day) and MPP conditions in a N2-purged glovebox, though it could keep 89% of the initial PCE after storage for 55 days in the dark (Figure S10b). The unstable device performance measured under illumination and MPP conditions may be resulted from the unsatisfactory light stability of the perovskite−Cu(Tu)I hybrid material, which is very complicated and deserves to be further studied.

and thus lead to their serious trapping, accumulation, and recombination, which is very detrimental to the device performance. Significantly, both CuI and Cu(Tu)I can interact with the defective perovskite by not only supplying iodine to fill the halide vacancies and thus coordinate the under-coordinated metal cations (see red arrows in Figure 7), but also chelating the under-coordinated halide anions with the electropositive copper atoms (see red dashed lines in Figure 7). The former interaction can be inferred from the quite weak but detectable Pb−I bond in the MAPbCl3−Cu(Tu)I hybrid material by EXAFS analysis (Figure S7). Though the latter interaction is unable to directly characterize, it is also reasonable because the copper atoms on surfaces of both CuI and Cu(Tu)I particles are electropositive and partly coordinative unsaturated. Therefore, both CuI and Cu(Tu)I can effectively passivate the trap states of perovskite, which is also supported by the shallower trap state energy level of the perovskite film with CuI or Cu(Tu)I as the TSP, compared to that of the pristine perovskite film. This would reduce the depth of potential wells caused by the disparity between the trap state energy level and VBM of perovskite, and thus weaken the trapping, accumulation, and recombination of charge carriers. It is also noted that Cu(Tu)I can play a better role than that of CuI, which is probably resulted from the lower VBM of Cu(Tu)I (−5.45 eV) than that of CuI (−5.21 eV), in view of the analogous interaction between perovskite and CuI or Cu(Tu)I. On the other hand, the p-type conductor CuI or Cu(Tu)I incorporated in perovskite layer can also establish the perovskite−CuI or Cu(Tu)I bulk heterojunction, because the perovskite has been demonstrated to possess an ambipolar charge transport character that enables it to act as both n- and p-type conductors, depending on the type of junction formed with the neighboring semiconductor.72 This is also supported by the increasing depletion width of the devices without any TSP (126 nm) and with CuI (195 nm) or Cu(Tu)I (265 nm) as the TSP. So, the junction structure of the device with the ptype TSP (CuI or Cu(Tu)I) incorporated in perovskite layer can be described as (p−i)−n, where p represents the p-type conductor, i represents the ambipolar perovskite, n represents the n-type conductor (C60), and p and i form the p−i bulk heterojunction. It is worth to note that Cu(Tu)I can also play a better role than that of CuI, probably due to the exactly identical VBM of Cu(Tu)I with that of perovskite, which would completely eliminate the potential wells for holes caused by the VBM disparity between the p-type TSP and perovskite, thus greatly facilitate hole transport, and decrease charge carrier recombination. Accordingly, we can generally conclude that the p-type conductor CuI or Cu(Tu)I incorporated in perovskite layer can not only passivate the trap states of perovskite but also participate in constructing the bulk heterojunction with perovskite. The resulting two synergistic effects eventually improve the device performance, and Cu(Tu)I plays a much better role than that of CuI, probably on account of its identical VBM with that of perovskite. Although the identical VBM of Cu(Tu)I with that of perovskite is beneficial for improving the trap state passivation capability of Cu(Tu)I and accelerating hole transport at the perovskite−Cu(Tu)I heterojunction, the relatively high conduction band minimum (CBM) of Cu(Tu)I would simultaneously lead to high potential barriers for electrons, which probably has a negative effect on electron transport within perovskite layer. Therefore, it needs to optimize the p-type TSP concentration to balance its positive and negative effects. As a

■

CONCLUSION We have first proposed and revealed the possible mechanism of the trap state passivation for perovskite by the p-type conductors CuI and Cu(Tu)I. These p-type TSPs can effectively interact with the under-coordinated metal cations and halide anions at perovskite crystal surface and thus shallow the trap state energy level of perovskite. Furthermore, the incorporated p-type TSPs can form bulk heterojunctions with perovskite, which can further facilitate hole transport and 7509

DOI: 10.1021/jacs.7b01439 J. Am. Chem. Soc. 2017, 139, 7504−7512

Article

Journal of the American Chemical Society

and H2O). The IPCE spectra were measured on a Keithley 2400 source meter under irradiation of a 150 W tungsten lamp in cooperation with a 1/4 m monochromator (Spectral product DK240). Synthesis of Cu(Tu)I and Cu(Tu)Cl. The thiourea solution was prepared by dissolving 152 mg of thiourea (Alfa, 99%) in 10 mL of anhydrous acetonitrile (Acros, 99.9%). The CuI solution was prepared by dissolving 381 mg of CuI (Aldrich, 99.999%) in 20 mL of acetonitrile. The CuCl solution was prepared by dissolving 594 mg of CuCl (Alfa, 99.999%) in 10 mL of acetonitrile. The Cu(Tu)I solution was prepared by mixing 10 mL of thiourea solution and 20 mL of CuI solution. After standing the Cu(Tu)I solution for several days, the Cu(Tu)I crystals would precipitate. Afterward, the Cu(Tu)I crystals were filtered and dried at 80 °C in a N2-purged glovebox (