Dianyi
Liu
,
Mahesh K.
Gangishetty
and
Timothy L.
Kelly
*
Department of Chemistry, University of Saskatchewan, 110 Science Place, Saskatoon, SK, Canada. E-mail: tim.kelly@usask.ca; Fax: +1 306 966 4730; Tel: +1 306 966 4666
First published on 20th October 2014
Recent advances in the development of perovskite solar cells based on CH3NH3PbI3 have produced devices with power conversion efficiencies of >15%. While initial work in this area assumed that the perovskite-based cells required a mesoporous TiO2 support, many recent reports have instead focused on the development of planar heterojunction structures. A better understanding of how both cell architecture and various design parameters (e.g., perovskite thickness and morphology) affect cell performance is needed. Here, we report the fabrication of perovskite solar cells based on a ZnO nanoparticle electron transport layer, CH3NH3PbI3 light absorber, and poly(3-hexylthiophene) (P3HT) hole transport layer. We show that vapor-phase deposition of the PbI2 precursor film produces devices with performances equivalent to those prepared using entirely solution-based techniques, but with very precise control over the thickness and morphology of the CH3NH3PbI3 layer. Optimization of the layer thickness yielded devices with efficiencies of up to 11.3%. The results further demonstrate that a delicate balance between light absorption and carrier transport is required in these planar heterojunction devices, with the thickest perovskite films producing only very low power conversion efficiencies.
While many of the perovskite solar cells reported to date have successfully adopted the mesoscopic design, the high temperatures associated with sintering a TiO2 mesoporous framework (>400 °C) have spurred the development of alternative cell architectures. Several examples of perovskite solar cells with a planar heterojunction structure have now been reported,14,31–37 with highly encouraging results. While initial reports of cells based on a CH3NH3PbI3/C60 planar heterojunction showed relatively modest efficiencies of 3.9%,31 organolead halide perovskite/fullerene heterojunctions have since achieved PCEs of up to 12%.32,33 Recently, vapor-phase deposition techniques have also been used to prepare perovskite solar cells with the structure ITO/TiO2/CH3NH3PbI3−xClx/spiro-OMeTAD/Ag, and displayed PCEs of up to 15.4%.13 Our recent work has also shown that planar heterojunction devices can be fabricated by the same two-step deposition procedure used by Burschka et al.,12 producing efficiencies of up to 15.7% on rigid substrates, and 10.2% on flexible supports.14 These devices are based on an ITO/ZnO/CH3NH3PbI3/spiro-OMeTAD/Ag structure, and are produced using solution-phase deposition techniques carried out at room temperature. The simplicity of the cell design and fabrication, when combined with their high performance and compatibility with flexible substrates, makes planar heterojunction perovskite solar cells a promising alternative to the original mesoscopic design.
One of the key variables in the fabrication of such devices is the thickness of the light harvesting perovskite layer. If the perovskite film is very thin, then the number of photons absorbed is low, resulting in low photocurrents; however, if very thick perovskite films are used, then the efficiency of charge carrier extraction is low, and recombination becomes problematic. Mesoscopic cells that use TiO2 photoanodes partially circumvent this issue by providing a large CH3NH3PbI3/TiO2 interfacial surface area; electrons are therefore extracted from the perovskite throughout the entire thickness of the light harvesting layer. Previous studies have shown that the PCE of devices of this type show a relatively weak dependence on perovskite layer thickness, with optimized photoanode thicknesses of ca. 650 nm.9 In planar heterojunction devices, the situation becomes more complex. Recent work has shown that film thicknesses of 400–800 nm can be used to successfully prepare high-performance devices based on CH3NH3PbI3−xClx in both conventional7,9,38 and inverted32 architectures. The very long charge carrier diffusion lengths observed for CH3NH3PbI3−xClx (ca. 1000 nm)7 mean that film thicknesses comparable to those of mesoscopic devices can be used. This is not the case for the parent CH3NH3PbI3 material, where the carrier diffusion lengths are an order of magnitude lower.7,8 Previous reports of planar heterojunction cells based on CH3NH3PbI3 have utilized perovskite layer thicknesses of only 100–350 nm,7,14,33–35 substantially thinner than those used in the chloride-doped analogues. Given the substantial differences in carrier diffusion lengths for these two perovskite compositions, a much deeper understanding of the role of film thickness and morphology on the efficiency of CH3NH3PbI3-based planar heterojunctions is required.
Here we report the fabrication of high performance perovskite (CH3NH3PbI3) solar cells with a planar heterojunction structure, and carefully elucidate the effect of perovskite film thickness and morphology on device efficiency. The perovskite films were deposited using a two-step deposition technique, with thermal evaporation used to prepare the PbI2 precursor film. This circumvents any issues with the solubility or viscosity of the PbI2 precursor solution, which have complicated these sorts of studies in the past. The evaporation of the PbI2 precursor allows perovskite films of precisely tailored thickness to be produced. Poly(3-hexylthiophene) (P3HT) was used as the hole transport material, and devices with the structure ITO/ZnO/CH3NH3PbI3/P3HT/Ag were prepared. Optimization of the film thickness produced power conversion efficiencies of up to 11.3%, which are among the highest ever reported for perovskite devices using a P3HT hole transport layer.
Before the influence of perovskite layer thickness and morphology could be investigated, it needed to be determined whether the change in PbI2 deposition process would have any effect on the perovskite structure, morphology, or the performance of the resultant devices. As can be seen from the powder X-ray diffraction patterns of the PbI2 and CH3NH3PbI3 films (Fig. S1†), there are only minor differences in structure that arise as a result of the change in deposition process. Both PbI2 films are poorly crystalline (as evidenced by the weak, broad diffraction peaks), and data are consistent with the 2H polytype, in keeping with previous reports of solution-prepared PbI2 films.12 After dipping both samples in the CH3NH3I solution, both films are successfully converted into the corresponding perovskite. The diffraction patterns are consistent with the tetragonal phase of CH3NH3PbI3.41 The morphology of the perovskite films was also compared by SEM (Fig. S1†). Both films consisted of a closely-packed collection of block-like perovskite crystallites, with at most only very subtle differences in crystallite shape and size.
In order to gauge the effect of processing method on device efficiency, the perovskite films derived from both solution- and vapor-phase deposited PbI2 were used to prepare planar heterojunction solar cells. In both cases, the thickness of the perovskite film was determined to be ≈ 300 nm. A thin (ca. 30 nm) ZnO nanoparticle layer on ITO was chosen as the electron-selective contact, and P3HT was used as the hole transport material. The J–V curves of all devices were measured under AM1.5G illumination, and those of the highest-performing cells are shown in Fig. 1a. The results clearly indicate that the particular method of PbI2 deposition has little effect on the overall power conversion efficiency: the devices prepared using solution- and vapor-phase deposition techniques had PCEs of 11.8% and 11.3%, respectively. Both devices were also observed to have almost identical open-circuit voltages (Voc = 0.949 V and 0.941 V) and similar fill factors (FF = 66.1% and 68.2%). The only notable difference between the two devices was a slightly higher short-circuit current density for the solution-processed device (Jsc = 18.8 mA cm−2) compared to the one processed by thermal evaporation (Jsc = 17.6 mA cm−2). The lower Jsc in the device fabricated from vapor-deposited PbI2 is also reflected in the incident-photon-to-current efficiency (IPCE) spectrum (Fig. 1b). However, despite these small differences in Jsc, the difference in the overall PCE of devices constructed from both solution and vapor-deposited PbI2 is quite minor.
In order to ensure the reproducibility of our results, we fabricated and tested 34 separate devices using evaporated PbI2 layers, and 29 separate devices using spin coated PbI2 layers. The tabulated device performances are shown in Table 1. In this case, the average Jsc of the solution processed devices was found to be slightly higher (17.1 ± 1.1 mA cm−2) than that of the devices derived from evaporated PbI2 films (16.0 ± 0.9 mA cm−2), suggesting that the differences in Jsc observed in Fig. 1 simply reflect batch-to-batch variation in the perovskite layer, rather than any significant difference between the two fabrication processes. Similarly, the very small differences in Voc and FF (and ultimately, PCE) for the two fabrication methods are well within their associated standard deviations, indicating that the two methods produce cells of similar performance.
PbI2 deposition method | V oc (V) | J sc (mA cm−2) | FF (%) | PCE (%) | Best PCE (%) |
---|---|---|---|---|---|
a Averages derived from measurements on 29 separate devices. b Averages derived from measurements on 34 separate devices. | |||||
Spin coatinga | 0.94 ± 0.01 | 17 ± 1 | 62 ± 4 | 10.0 ± 0.9 | 11.8 |
Thermal evaporationb | 0.94 ± 0.02 | 16.0 ± 0.9 | 65 ± 4 | 9.7 ± 0.6 | 11.3 |
Fig. 2b shows the absorption spectra of the perovskite films as a function of film thickness. All perovskite films exhibited a broad range of light absorption from 300 to 760 nm, with the optical density of the film increasing monotonically with film thickness. The linear increase in optical density demonstrates both the excellent control over film thickness provided by the evaporation of PbI2, and the overall consistency of the two-step deposition process.
In order to understand how film thickness affects the morphology of the perovskite layer, both PbI2 and CH3NH3PbI3 films were evaluated by SEM (Fig. 4). Fig. 4a–c shows a top–down view of vapor-deposited PbI2 films of ca. 50, 150 and 300 nm thickness. The 50 nm thick film was relatively smooth, with only very small nanocrystallites visible under the electron microscope. As the film thickness is increased to 150 nm, the PbI2 crystallites begin to develop a flake-like texture, eventually becoming large nanoplates in the thickest films. After conversion to the perovskite (Fig. 4d–f), the morphology of the films is substantially altered; the film is still composed of a closely-packed collection of nanocrystallites, but the individual crystallites adopt a more block-like morphology, in keeping with the tetragonal perovskite unit cell. The same trend in morphology is also observed in the perovskite films. In the thinnest (ca. 110 nm) film, the block-like crystallites are small and densely packed, forming a film that, while not smooth, has very few gaps between the crystallites. As the perovskite film grows thicker, the individual nanocrystallites grow larger, in turn producing a rougher surface morphology. For the thickest (ca. 580 nm) film, many of the crystallites are over 100 nm in size, with equally large gaps between them; this produces a highly textured morphology.
In order to more quantitatively evaluate these trends, we plotted both the particle size (as measured by SEM) and the minimum grain size (as determined from a Williamson–Hall analysis of the pXRD data, Fig. S3†) as a function of perovskite layer thickness (Fig. S4†). As suggested by Fig. 4, the average particle size increases as the perovskite layer is made thicker; however, this evolution in particle size is modest when compared to the pronounced morphological differences observed for the PbI2 films (Fig. 4). This suggests that a substantial degree of structural reorganization occurs within the crystallites during the CH3NH3I intercalation step, resulting in a highly crystalline perovskite film. This is supported by the Williamson–Hall analysis of the pXRD data, which separates out the effects of size and lattice strain on peak broadening (Fig. S3†). This analysis suggests that there is a slight dependence of the minimum grain size on the film thickness, with the thickest films having slightly larger crystalline domains than the thinnest samples. The degree of lattice strain is essentially negligible (<0.1%) for the majority of samples (Fig. S5†). Combined with the excellent agreement between the observed CH3NH3PbI3 thicknesses and those predicted based on the PbI2 precursor thickness and the subsequent increase in molar volume, this suggests that the Pb2+, I−, and CH3NH3+ ions are sufficiently mobile throughout the two-step deposition sequence to produce highly relaxed, crystalline films. This, in part, explains the excellent performance of the perovskite films fabricated using this technique.14,42 The one exception to this appears to be the thinnest films, which show a small amount of additional lattice strain, likely as a result of interactions with the underlying ZnO interface.
Since both the change in particle/grain size and the associated changes in surface roughness have important implications for device performance, we further probed the surface roughness of the samples using atomic force microscopy (Fig. S6†). The root-mean-square surface roughness (Rrms) of the perovskite films was found to change with increasing film thickness, with the thickest film (560 nm) showing a slightly higher Rrms (49 nm) compared to the rest of the samples (23–33 nm). The implications that this observation has for device performance are discussed in more detail below.
The second factor leading to the observed increase in PCE is the steady increase in film optical density (Fig. 2b) with increasing thickness. For the thinnest films studied, the light harvesting efficiency at longer wavelengths is relatively low: at 700 nm, only ca. 69% of the incoming light is absorbed (assuming a 4% reflection loss at the air/glass interface, and 100% reflection at the silver back-contact), and when combined with high levels of recombination, this results in an IPCE of only ca. 40%. As the film thickness is increased, the light harvesting efficiency increases steadily; for the optimum film thickness of 330 nm, approximately 86% of the light at 700 nm is absorbed, leading to the stronger IPCE spectrum and improved Jsc. Since at this film thickness, only ca. 14% of the long wavelength light is lost to transmission, there is relatively little benefit to be realized from the use of thicker perovskite films.
For the devices with perovskite film thicknesses >400 nm, the decline in device performance is particularly sharp. As the film thickness continues to increase, there is a significant increase in the series resistance of the devices (Table S1†); the series resistance of the cells with the thickest perovskite layers is over an order of magnitude higher than for optimized devices. EIS measurements suggest that this is due to an increase in the contact resistance between the perovskite and interfacial layers (Table S2†); since the ZnO/CH3NH3PbI3 interface does not change as the perovskite thickness is increased, we ascribe this increase to changes in the CH3NH3PbI3/P3HT interface. This may suggest that for the thickest films, the polymer hole-transport material has difficulty penetrating into the porous surface; this would lead to an increased contact resistance, and contribute to the higher series resistance of these devices. Additionally, as the thickness is increased, there is also a drop in the shunt resistance (Table S1†). This may be related to the changing morphology of the perovskite layer; as the CH3NH3PbI3/P3HT interface becomes rougher (Fig. 4), there are also more opportunities for surface recombination to occur, and the efficiency is adversely impacted. Analogous behavior has been observed for perovskite solar cells with a high surface area ZnO/CH3NH3PbI3 interface.46 The combination of these two effects explains the observed drop in fill factor (64% to 42%).
Even more striking than the changes to the Voc and FF, however, is the dramatic loss (>75%) in Jsc as the perovskite film thickness begins to exceed 400 nm. This is by far the largest contributor to the reduced efficiency of the thickest devices, and is predominantly caused by the mismatch between the absorption depth of the perovskite and the carrier diffusion length. The electron and hole diffusion lengths for CH3NH3PbI3 have been reported to be ∼100 nm;7,8 however, for the thickest (580 nm) devices, the perovskite film thickness significantly exceeds the diffusion lengths of the charge carriers. As such, carriers generated near the center of the film will recombine before reaching the electrodes. This is consistent with the sharp decrease in IPCE observed for wavelengths > 500 nm (Fig. 5b). For wavelengths < 500 nm, the perovskite absorbs very strongly (Fig. 2b), and charge carriers are generated very close to the ZnO/CH3NH3PbI3 interface; as such, a reasonable number of electrons can be collected by the ZnO electron-transport layer before they recombine. For wavelengths > 500 nm, the lower extinction coefficient means that charge carriers are generated deep in the interior of the perovskite film and are more than 100 nm away from either interface. As such, the carrier collection efficiency in this spectral region begins to approach zero, resulting in the very low short-circuit current densities that are observed. Additionally, the less-compact perovskite morphology observed in thicker films (Fig. 4f) may result in poorer connectivity between adjacent crystallites, leading to a more tortuous pathway for carrier transport and a greater likelihood of recombination. Given the observed decrease in PCE, the increased light harvesting efficiency observed in the thicker films (96% at 700 nm for the thickest film, vs. 86% for the optimum thickness of 330 nm) is clearly insufficient to overcome these deleterious effects on carrier transport. This same argument would also hold true (although to a lesser degree) for the optimized perovskite film thickness of 330 nm. A large number of carriers would be generated in the innermost 130 nm of the perovskite film; since these are greater than an exciton diffusion length away from the electron- and hole-transport interfaces, many of these would be expected to recombine. However, given the good IPCE values achieved in these devices, this is clearly not the case. There are two possible explanations for this observation: (i) the carrier diffusion lengths are significantly longer than those originally reported by Stranks et al. and Xing et al.;7,8 or, (ii) carrier extraction is facilitated by the high surface area CH3NH3PbI3/P3HT interface. Although the measurements of Stranks et al. and Xing et al.7,8 are fairly conservative estimates, we believe that the high surface area CH3NH3PbI3/P3HT interface plays a key role in the high performance of these planar heterojunction devices. This can be understood more quantitatively by considering the highest performance devices, which have a 330 nm thick perovskite layer. The first 130 nm of this film is within one charge carrier diffusion length of the ZnO interface,7 and carrier collection is not expected to be problematic. Similarly, holes can readily be extracted from the 110 nm closest to the P3HT interface.7 This leaves ca. 90 nm in the interior of the film that is too far away from either interface to meaningfully contribute to the photocurrent. However, AFM measurements indicate that the CH3NH3PbI3/P3HT interface has a Rrms of 20–30 nm. Again, considering that the Rrms value represents the standard deviation of the sample height, the maximum peak-to-trough variation in the perovskite interface is closer to 60–90 nm. As a result, charge carriers can be extracted from much deeper within the perovskite film than would be the case if the CH3NH3PbI3/P3HT interface were completely planar (Fig. S7†). Although there is still a very small section of the perovskite film that still does not contribute to the overall photocurrent in this model (<30 nm), the discrepancy is quite small, and likely due to slightly longer charge carrier diffusion lengths in our films.
Hysteresis in the J–V curve is now a well-known issue for perovskite solar cells,47 and it results in a differential current response depending on the history of applied bias. In order to show that the trends observed in Fig. 6 are the result of changes in film thickness (and not due to changes in hysteretic behaviour), we measured the J–V curves using both forward (short-circuit to forward-bias) and reverse (forward-bias to short-circuit) scans for a series of representative devices (Fig. 8). As with many planar heterojunction devices, the hysteresis is quite pronounced, with the reverse scan having a substantially higher efficiency (Fig. 8a). However, the magnitude of the hysteresis was found to depend only very weakly on the thickness of the perovskite film (Fig. 8b). This suggests that batch-to-batch variability in sample preparation has as much or more influence over the hysteresis than the film thickness. Therefore, while these results suggest that the absolute efficiencies of these devices are lower than might be expected from measurements of just the reverse scan, the overall trends in device performance (Fig. 6) are actually due to changes in film thickness, and not due to changes in hysteretic behaviour.
Footnote |
† Electronic supplementary information (ESI) available: Powder X-ray diffraction patterns, SEM and AFM images, Williamson–Hall plots, crystallite size and strain data, tabulated resistance data and impedance parameters. See DOI: 10.1039/c4ta02637c |
This journal is © The Royal Society of Chemistry 2014 |