PCBM/CsPbI3 QD hybrid film
CsPbI3 QDs were synthesized and purified according to the procedures previously published19. PCBM was added into the CsPbI3 QDs dispersed in chlorobenzene (CB) to obtain a hybrid solution. The CsPbI3 QDs with and without PCBM have similar cubic shape with an average size of about 10 nm, characterized by the transmission electron microscopy (TEM) shown in Fig. S1. Fig. 1a depicts two different kinds of QD film deposition and solid-state ligand removal process. In general, the QD solution was spin-coated onto a substrate under ambient conditions with low relative humidity (RH<10%) at room temperature. Then the native long oleic acid (OA) and oleylamine (OLA) ligands were removed by soaking the as-cast QD film in anhydrous methyl acetate (MeOAc)20. This procedure was repeated 3-5 times to build up ~300 nm thick films, resulting one control CsPbI3 QD film and another hybrid PCBM/CsPbI3 QD film.
We first comprehensively characterized the properties of CsPbI3 QD films with or without PCBM using the Fourier transform infrared (FTIR), femtosecond transient absorption (fs-TA) and PL spectroscopy. In the FTIR spectra (Fig. 1b), the signal intensities of C-H modes at 2851 and 2921 cm-1 for both control and hybrid films were remarkably reduced comparing to the pristine film (without MeOAc treatment), which indicates the majority of native long chain ligands were removed after the MeOAc treatment and the addition of PCBM into the QD matrix had little effect on the process of surface ligand removal. To investigate the charge transfer dynamics in control and hybrid QD films, fs-TA measurements were carried out to collect the detailed information for both radiative and nonradiative processes41-42. As shown in Fig. 1c-d, the hybrid film exhibits more obvious bleaching shifts as increasing delay time than the control one, demonstrating the charge transfer occurs in hybrid film is under a stronger drive. The corresponding time-resolved components extracted from the dynamic curves at bleaching peak of around 670 nm are shown in Fig. 1e. The time constants τ1, τ2 and τ3 calculated through a three-exponential fitting are generally attributed to the hot phonon bottleneck, the Auger recombination and charge transfer process, respectively.41 Clearly, the reduced charge transfer time in the hybrid film (312.8 ps) indicates more efficient charge extraction compared with that in the control film (483.2 ps). The slightly longer Auger recombination process in the hybrid film (57.6 ps vs. 52.3 ps) suggests less trap density, which possibly results from the passivation of PCBM.
Steady-state PL and time-resolved PL (TRPL) measurements were used to confirm the improved carrier dynamics observed in fs-TA measurements. As shown in Fig. S2, compared to the pristine film, the control QD film exhibits red-shifted PL peak with reduced intensity, indicating improved electronic coupling after surface ligand removal41. Meanwhile, the PCBM/CsPbI3 QD hybrid film demonstrates obvious PL quenching with more red shift, suggesting the charge transfer between PCBM and CsPbI3 QDs,22, 32 and the further enhanced electron coupling in the hybrid blends. In addition, the faster PL decay profile (Fig. S3) for the hybrid CsPbI3 QD indicates PCBM in the hybrid QD film could provide an effective transport channel for accelerating charge extraction, which is also consistent with the fs-TA analysis.
Performance of CsPbI3 QD photovoltaics
The schematic diagrams of the control and the target PCBM/CsPbI3 QD solar cell architectures are given in Fig. 2a-b, respectively. To fabricate the control devices without hybrid films, an ultra-thin PCBM was prepared on top of the SnO2 layer. Each layer of both devices is clearly identified in the cross-sectional SEM image presented in Fig. S4. To achieve the best optimized PV performance, as shown in Table S1, we first tuned the thickness of the QD absorber layer for control QD devices. A decent efficiency of 12.4% is obtained using four depositing cycles. Then we optimized the thickness of the hybrid QD layer in the HIA, concluding the device with the 1:4 thickness ratios of hybrid QD layers to control QD layers delivered a highest efficiency up to 15.1%, which is among the highest reported efficiency for CsPbI3 QD solar cells (Table S2). As shown in Fig. 2c and Table 1, the enhancement of PCE mainly contributed from the significantly improved short-circuits current density (Jsc). External quantum efficiency (EQE) was then carried out on the optimal control and target devices, showing the EQE of the target QD device is enhanced across the entire response region (Fig. 2d), which could result from the improved charge collection efficiency through introducing HIA. Furthermore, we increased the thickness of hybrid QD layers as the control QD layers without any change. The device having thicker hybrid QD layers exhibits decreased efficiency down to 13.8% (Table S1) that could be mainly resulted from the unfavourable effects of CB solvent during upper hybrid layer deposition,43 leading to PCBM exfoliation from CsPbI3 QDs in the bottom layer.
It should be noted that the HIA modification also leads to reduced J-V hysteresis (Fig. S5). The stabilized power output (SPO) of 14.61% for the target device and 11.93% for the control device were recorded (Fig. 2d), confirming the reliability of J-V measurements. We also characterized device stability (Fig. S6), and found that the control device lost 50% of its initial efficiency while the target device degraded by only 30% after storage in dry air-filled box for 14 days, indicating that the HIA is also beneficial for improving device stability. The contact angle of water drop on control and hybrid film is presented in Fig. S7, indicating the hybrid film has a more hydrophobic nature, which is beneficial for improved the device stability as well as depositing the upper QD layer.
To understand the enhancements of QD photovoltaic performances, a series of device characterizations were further performed. Electrochemical impedance spectroscopy (EIS) is introduced to investigate the quality of interfaces. As shown in Fig. S8, the target device exhibits a similar series resistance (Rs=28.5 Ω) but a larger recombination resistance (Rsh=3346 Ω) than that of the control device, indicative of enhanced charge transfer and reduced charge recombination process. Additionally, their different dark J-V characteristics in Fig. S9 show that the target device exhibits a lower leakage current under reverse bias compared to the control device, indicating decreased carrier recombination. Furthermore, transient photocurrent and photovoltage (Fig. S10) were performed on both devices. The target device has longer lifetime of 65 µs compared to that of 59 µs in the control one, indicating a smaller charge recombination rate. The transient photovoltage exhibits a similar trend and the target device has a longer lifetime of 4.6 µs (3.6 µs for the control one), demonstrating a faster charge transport rate. All these results indicate the intrinsic mechanism of enhancement of QD photovoltaic performance lies in the improvement of the QD/ETL interfaces, which is attributed to the improved energy level alignment. The values of valence band (VB) energy level were characterized by ultraviolet photoelectron spectroscopy (UPS) (Fig. S11), and their conduction bands (CB) were further calculated according to their optical bandgaps (Fig. S11). As shown in Fig. 2e-f, PCBM has the appropriate electron affinity serving as a modification layer between the SnO2 and the CsPbI3 QDs, which leads to the suppressed charge recombination at the interface between SnO2 and the CsPbI3 QD layer. The efficient charge transfer in hybrid device is enabled by the formation of a gradient CB alignment that efficiently injects the photogenerated electrons into the CB of PCBM and then extracted to the SnO2 ETL, as evidenced by the fs-TA and PL measurements. Hence, PCBM could be seen as a fast electron transport channel in the target QD device, which leads to the improved device performance.
CsPbI3 QD-PCBM molecular interaction
To further understand the molecular interaction between CsPbI3 QD and PCBM, density functional theory (DFT)-based molecular dynamic simulation and X-ray photoelectron spectrum (XPS) measurements were carried out. First, surface defects in the perovskite may lead to trap states that not only impede charge transport between the light absorber and the PCBM but also increase unfavorable interfacial recombination44-45. As shown in Fig. 3a-d, Pb-vacancy (VPb) on (100) surfaces is one potential defect type that has been shown in the bulk, leading to shallow levels near the valence band maximum (VBM), resulting in undesirable charge accumulation and potential recombination at the perovskite-PCBM interface. However, the projected density of states (pDOS) obtained from DFT simulations (PAW PBE-GGA) on a 2×2×4 perovskite slab with a PCBM molecule on the most commonly exposed (100) facet do not show detrimental electronic states for either charged or neutral VPb defects. The carbon p-states (C, p) of the PCBM are well aligned to receive carriers from the perovskites. The iodine p-levels that define the perovskite’s VBM do not encroach on the Carbon p-levels for the relatively large concentration of VPb simulated (25% atomic VPb/Pb), implying the perovskite is free of hole traps at the surface for this defect type. Second, it was found that the peaks of Pb 4f (Fig. 3e), I 3d (Fig. 3e) and Cs 3d/5 (Fig. S13) shifted towards a lower binding energy level in CsPbI3 QDs of the hybrid film, suggesting the formation of coordination bonds between carboxyl moieties and mostly under-coordinated Pb2+ ions on QD surfaces. The systematic characterizations of the hybrid QD blend film indicate the improved carrier dynamic process and molecular interaction. As illustrated in Fig. 3f, when PCBM blends with CsPbI3 QDs at the optimal ratio, the spherically symmetric C60 facilitates the efficient charge transfer at the interfaces. In addition, one problem with the conventional CsPbI3 QD film after the process of ligand removal is its sensitivity to moisture. The addition of hydrophobic PCBM could not only reduce the QD boundaries but also passivate the QD surfaces via forming strong bonding of the carboxyl group with the under-coordinated Pb2+ ions on the QD surfaces.
Moreover, the top-view SEM images in Fig. S14 clearly exhibit very similar morphology in the control and hybrid films, showing that both films are dense with similar surface roughness. Atomic force microscopy (AFM) height images in Fig. S15 indicate that the hybrid film is smoother than the control one. Both films have similar surface topography with roughness of 7.3 nm for the control film and 7.1 nm for the hybrid film, suggesting that the PCBM addition did not significantly alter the QD film morphology.
Flexible QD photovoltaics
To explore the advantageous of HIA strategy in fabricating efficient QD flexible solar cells, we fabricated perovskite QD flexible solar cells with same structure for rigid device using conventional PET (polyethylene terephthalate)/ITO substrates. Fig. 4a and Table 1 presents the optimized J-V characteristics and PCE distribution of the control and target flexible devices, both the efficiency and device reproducibility are significantly improved using HIA modification, with a record PCE of 12.3%, greatly outperforming both control and previous reports (Table S2). What is more exciting is that this HIA strategy enables us to practically demonstrate the better suitability of the perovskite QDs on flexible photovoltaics compared to extensively studied thin-film materials46-48. Fig. 4b shows the performance comparison between all-inorganic CsPbI2Br perovskite bulk film and QD device. It should be noted that the preparation of all-inorganic CsPbI3 bulk film needs high-temperature (>320 oC) annealing process49-50, which will damage the PET flexible substrate. Therefore, we select CsPbI2Br with decreased phase transition temperature down to ~100 oC. As shown in Fig. 4b, after 100 bending cycles at the radius of 0.75 cm, the efficiency of QD devices decreased to 90% of the initial value that may be attributed to the degradation of the rigid electrode ITO or Ag contact,51 while only ~75% of initial efficiency was remained for the CsPbI2Br thin-film based device. The detailed device parameters, i.e., open-circuit voltage (Voc), Jsc, and fill factor (FF) for CsPbI3 QD and CsPbI2Br thin film devices are shown in Fig. S16. These results first indicate the superior performance in QD-based flexible devices relative to its bulk thin-film counterpart.
To further investigate the degradation of device performance and understand the internal stress in films after bending, we carried out the film mechanical durability (Fig. 4c) test using SEM examination on a variety of materials (CsPbI3 QD, PbS QD, CsPbI2Br thin-film, and Cs0.05MA0.85MA0.1PbI2.55Br0.45 mixed perovskite thin-film) deposited on the same flexible PET/ITO substrate, as shown in Fig. S17. We don’t observe any apparent surface morphology change after bending up to 1000 times in both CsPbI3 (Fig. 4d) and PbS (Fig. 4e) QD films. However, crystal cracks propagation is apparently noticeable in the CsPbI2Br thin-film after 100 times bending, and gradually grow up to micrometer scale when further increasing the bending cycles (Fig. 4f). Similar trend was found that huge cracks and even severe crystal shedding appeared after 1000 cycle bending in the mixed perovskite thin-film with large crystal grains (Fig. 4g). These results allow us to conclude that QD films possess much better mechanical durability than bulk film since the intrinsic nanoscale boundaries and soft surface ligands in the low-dimensional QD materials are beneficial for releasing the thin film internal stress, as described in Fig. 4c. Based on the more durable performance of QD device, this is the first time to experimentally prove that QDs potentially offer higher mechanical flexibility than thin-film materials, and displaying their prominence in flexible devices.