Fully Solution‐Processed n–i–p‐Like Perovskite Solar Cells with Planar Junction: How the Charge Extracting Layer Determines the Open‐Circuit Voltage

Fully solution-processed direct perovskite solar cells with a planar junction are realized by incorporating a cross-linked [6,6]-phenyl-C61-butyric styryl dendron ester layer as an electron extracting layer. Power conversion efficiencies close to 19% and an open-circuit voltage exceeding 1.1 V with negligible hysteresis are delivered. A perovskite film with superb optoelectronic qualities is grown, which reduces carrier recombination losses and hence increases V oc .


DOI: 10.1002/adma.201604493
at the metal contact (often gold). Further development has been driven by empirical optimization strategies and testing of a variety of architectures. One of these promising architectures uses fully organic selective contacts. This choice brings, first of all, the advantage of low temperature processing; additionally, a beneficial effect in terms of electrical stabilization of the device with organic electron extraction layers in a flat junction configuration, with respect to the use of TiO 2 , has been repeatedly observed. [6][7][8] So far, architectures with fully organic selective contacts are either limited to the so-called inverted architecture, as a consequence of difficulties in the solution processing of multilayers, or they have to rely on vacuum deposition methods. [7][8][9] Solution-processed, flat junctions, multilayered direct architectures, which would open the path toward large-area, low-temperature processing of high-efficiency direct cells, have been poorly explored, and their power conversion efficiencies are still limited. [10][11][12] Intuitively, the development and optimization of different architectures need a full engineering of different interfaces. This is especially critical for metal halide perovskites that, unlike more widely studied inorganic semiconductors, have fluctuating ionic structures where tilting, distortions, and polarizability of the lattice strongly affect their optoelectronic properties. This makes the reliability of perovskite-based devices strongly dependent on the control of the microstructure of the active material and of its response to external stimuli, such as chemical interactions upon interface formation, [13] electric field, [14,15] light, [16] and environmental agents. [17,18] It is also worth noting that, while perovskite solar cells have made spectacular progress in terms of solar-to-electrical current conversion efficiency in very short time, such advances are mainly based on an empirical approach, which has led to a myriad of architectures, whose advantages and limitations cannot be attributed yet to a fundamental property of the active material or to the design and processing of the device.
Herein we present the fabrication of a fully solutionprocessed direct perovskite solar cell with planar junction, reaching power conversion efficiencies (PCEs) close to 19% and an open-circuit voltage (V oc ) larger than 1.1 V. Such a result is achieved by using a fullerene derivative, the [6,6]-phenyl-C61-butyric styryl dendron ester (PCBSD) functionalized with a dendron containing two styryl groups as thermal cross-linkers, as a selective contact for electrons. In situ cross-linking of Solution-processable hybrid perovskite semiconductors have risen to the forefront of photovoltaics research, offering the potential to combine high power conversion efficiency with low-cost fabrication and the shortest energy payback time. [1,2] The first architectures made use of nanocrystalline TiO 2 deposited on the transparent electrode, both in the form of mesoporous and compact films, as selective contact for electrons alike the so-called Grätzel solar cells. [3][4][5] Such a structure is generally indicated in the field as direct architecture as electrons are collected at the front transparent electrode and holes (2 of 7) 1604493 PCBSD by thermal treatment forms a robust, adhesive, and solvent-resistant thin electron accepting layer, on top of which a high quality perovskite film processed from solution can be grown. We show that the choice of the electron extracting layer (EEL) is of utmost importance to maximize the solar cell V oc as interface engineering accounts not only for the energy level alignment between the EELs and the perovskite, but also for the quality of the microstructure of the perovskite bulk film that is driven by the substrate surface. We further demonstrate that in this device architecture, at V oc conditions, the high carrier density is sustained not only by the charge extracting contacts but also by the semiconducting active layer. This rationalizes the optimal optoelectronic performance of the proposed devices as a beneficial effect of the engineering of the solution-processed EEL/perovskite interface in terms of microstructure and reduced carrier recombination/ charge extraction improvement.
The PCBSD is prepared by optimizing the synthesis route described by Hsieh et al. in order to achieve a final yield of 81% (see the Supporting Information for details). [19] The cross-linking of PCBSD is a thermalinduced radical polymerization between the terminal vinyl bonds of PCBSD (see Figure 1a). [20,21] In order to optimize the timing of the cross-linking process at 160 °C (see Figure S16 of the Supporting Information for the temperature optimization), its progress is monitored with in situ infrared (IR) spectroscopy for 45 min. Figure 1b presents the initial and the final spectra. The bands at 987 and 908 cm −1 , which are assigned to the CH and CH 2 wagging of the vinyl groups in trans conformation (see Figure S11 in the Supporting Information for assignment through DFT calculations), show an absolute reduction in intensity in time during the thermal crosslinking. Their lowering is due to the conversion of the sp 2 carbon atoms of the vinyl groups into sp 3 carbon atoms, which occurs during the PCBSD cross-linking. [22] The band at 908 cm −1 is selected for quantitative analysis and monitoring of the cross-linking time. The peak area of the band at 908 cm −1 is determined every 5 min up to a final reaction time of 45 min. The values, normalized over the peak area at t = 0, show that the reaction follows a first-order exponential decay and 35% of the carbon atoms of the vinyl groups are converted after a 45 min thermal annealing (Figure 1c). To prove that the achieved cross-linking degree is sufficient enough for the cross-linked PCBSD (C-PCBSD) layer to be solvent resistant, UV-vis spectra of the substrates are measured before and after being rinsed with dimethyl formamidine (DMF). The as-cast PCBSD films are almost completely washed off (Figure 1d), whereas only a weak decrease in the intensity of the bands is shown after cross-linking induced with a 30 min thermal annealing at 160 °C ( Figure 1e).
Before incorporating such EELs in a device structure, we tested its electronic properties. One of the main challenges to face upon cross-linking procedure of an organic semiconductor is a possible degradation of its transport properties. [23] C-PCBSD has been tested in a field-effect transistor (FET) geometry, which is quite demanding device in terms of electronic properties, as it requires in-plane charge transport along tens of micrometers. Electron percolation pathways can be evidently formed and a neat field-effect modulation of the current can be observed, already in a nonoptimized device, exhibiting an electron mobility in the order of 10 −5 cm 2 V −1 s −1 for surface charge densities in the range 8 × 10 15 to 3 × 10 16 m −2 (1.3 × 10 −7 to 5.1 × 10 −7 C cm −2 ) ( Figure S1 in the Supporting Information). This result shows that our C-PCBSD has almost an order of magnitude higher mobility than what previously reported in undoped C-PCBSD-based FET devices. [10] Thus we can safely c) The first-order exponential decay of polymerization reaction determined by normalized peak area of the band at 908 cm −1 . The absorption spectra of as-cast c) PCBSD and d) C-PCBSD before (solid) and after (dash) being rinsed with DMF.
confirm that the C-PCBSD film can sustain in-plane electron transport over several micrometers, and it is therefore likely to show good transversal electron transport properties over tens of nanometres of its thickness thus being a potential candidate as a thin EEL.
To start the solar cell fabrication, a very thin layer of C-PCBSD (less than 10-20 nm; see Figure S14 in the Supporting Information) is deposited onto the FTO/TiO 2 substrates as thicker films eventually lead to an increase of the device series resistance (see Figure S15 in the Supporting Information) . Its surface work function is measured in air by Kelvin probe and it results more electronegative with respect to that of TiO 2 and comparable to the [6,6]-phenyl-C 61 -butyric acid methyl ester (60-PCBM) (see Table S1 in the Supporting Information). Subsequently a methylammonium lead iodide (CH 3 NH 3 PbI 3 ) thin film is directly synthesised from solution on top of FTO/TiO 2 /C-PCBSD via a two-step procedure (see the Materials and Methods section in the Supporting Information for details). The final thickness of CH 3 NH 3 PbI 3 depends on the thickness of the PbI 2 layer (see the Supporting Information for the layer optimization). Figure 2a shows X-ray diffraction spectra measured before and after CH 3 NH 2 I spun cast on top of PbI 2 followed by thermal annealing on top of TiO 2 and TiO 2 /C-PCBSD substrates. In both cases, we do not observe any trace of the crystal precursors and the sharp peaks suggest the formation of large perovskite grains. Figure 2b-e shows the top-view images of the perovskite films grown on TiO 2 and TiO 2 /C-PCBSD, respectively, at different magnifications. The perovskite films show a good coverage on both substrates; however, the film grown on TiO 2 /C-PCBSD presents much larger grains (>1 µm) and a lower roughness. The static contact angle of water on C-PCBSD is more than twice larger with respect to the case of TiO 2 (as shown in Figure S2 in the Supporting Information), indicating a more hydrophobic surface. Thus, we conclude that the different morphology is the result of a lower wettability of the C-PCBSD surface to dimethyl formamidine/ PbI 2 solution with respect to the TiO 2 surface, which drives the formation of crystallite with larger aspect ratio. [7] To complete the devices, we deposited spiro-MeOTAD by spin coating as hole-extracting layer followed by gold via thermal evaporation. In Figure 3a,b, we show a cross section of the optimized planar devices employing TiO 2 and TiO 2 /C-PCBSD as EELs. Please note that such layers are extremely thin, of the order a few tens of nanometres, nearly conformal to the rough FTO substrate, while the perovskite layers have a different optimized thicknesses, approximately 400 and 320 nm, respectively. Figure 3c,d shows the current density versus voltage (J-V) characteristics measured in air under air mass 1.5 global (AM 1.5 G) conditions. To check for the possible hysteresis phenomena, which are known to strongly influence the device testing, especially for the planar junctions reported so far in the literatures, [13,24,25] we show the J-V curves as a function of scan rate, in forward and reverse scan directions. Table 1 summarizes the main figures of merit of the tested samples, i.e., short-circuit current density (J sc ), V oc , fill factor (FF), and PCE. As expected, the device based on bare TiO 2 as EEL has a characteristic that depends on the polarization record of the device (Figure 3c), and the photocurrent, upon polarization, has a transient time of a few seconds (Figure 3e). The figures of merit of such a device correspondingly depend on the testing conditions and are reported in Table 1. On the other hand, the TiO 2 /C-PCBSD interface electrically stabilizes the device. Not only the J-V characteristics have very limited dependence on scan directions and rates (Figure 3d), but also the photocurrent has no transient time upon polarization (Figure 3f). Such a stable device delivers a PCE as high as 18.7%, with a J sc of ≈21 mA cm −2 , V oc over 1.1 V and FF close to 80%. The reliability of such numbers is confirmed by a statistical study made over 61 devices, which can be found in Figures S3 and S4 in the Supporting Information.
It is worthy to notice that flat junction solar cells, especially those with direct architecture, hardly achieve a V oc of 1 V. [4,26,27] However, here we show a V oc that is consistently higher than 1.1 V (see statistics in Figure S3 in the Supporting Information). In this regard, these devices even outperform those made with 60-PCBM ( Figure S5, Supporting Information), despite the two EELs presenting very similar energy levels, which will likely align in the same way once interfaced with the perovskite semiconductor, given the same nature of the material. Recently, it has been shown that the structural order of the 60-PCBM layer has a significant impact on the enhancement in V oc due to a concomitant reduction of the energy disorder, as deduced from the significant decrease in the DOS of the organic semiconductor. [28] However, here, while the 60-PCBM clearly shows a sharp X-ray diffraction peak at 20.5° as a signature Adv. Mater. 2017, 29, 1604493 www.advancedsciencenews.com www.advmat.de (4 of 7) 1604493 of order and crystallinity, there is hardly any diffraction peak in C-PCBSD (see Figure S6 in the Supporting Information), thus excluding any direct role of the electron extracting layer morphology. Note that, for a fair comparison, we grew the perovskite film by thermally evaporating PbI 2 followed by MAI deposition by spin coating from 2-propanol, a procedure that some of us have recently developed [8] to preserve the soluble 60-PCBM layer and we noticed, also in this case, a tendency of forming larger crystallites over the C-PCBSD (see Figure S7 in the Supporting Information). Overall, beyond the interfacial energetics, the recombination dynamics will contribute to the definition of V oc of solar cells. Since the perovskite film is grown on top of the EEL, the latter's choice will determine its microstructure, which may influence the degree of defect states and the carrier recombination losses within the perovskite film. [29] In the following we will try to disentangle all these effects by combining excitation density dependent steady state photocurrent measurements and photoinduced transient techniques.
To understand the effect of the fabrication route and the substrate on the quality of the thin perovskite film, we performed photocurrent measurements within a photoconductor device configuration. [30] Briefly, planar symmetric gold contacts separated by 1 mm are evaporated on top of the film and continuous wavelength (cw)-laser-induced carriers are extracted by applying a bias voltage of 10 V across them (Figure 4a). Note that the electric fields within this configuration are smaller by a factor of 100 in comparison to the usual electric fields within the solar cell architecture. In Figure 4b, we show the photocurrent plotted as a function of the photoexcitation density. All measured films are ≈300 nm thick and grown on a thin layer of TiO 2 , TiO 2 /60-PCBM, and TiO 2 /C-PCBSD on glass; this guarantees to have a thin film microstructure comparable to the one present in the device while observing mainly bulk processes upon top layer illumination. Since perovskite cannot be solution-processed on top of TiO 2 /60-PCBM we also compare two thin films grown on C-PCBSD, one solution-processed and the other grown through thermal evaporation.
While all the films show a monotonic growth of photocurrent with the excitation density, they exhibit distinct intensity trends. As a guide for the eyes, we draw hypothetical linear (≈I) and sublinear ( I ) intensity dependent photocurrent in dotted lines. The photocurrent trends in all the films follow sublinear behavior at low excitation densities and reach a linear trend at higher densities. Such a behavior can be explained within a trap-limited Shockley-Read-Hall (SRH)-like recombination scenario, where at low-excitation densities, electrons (holes) are most likely trapped within defects and thus do not contribute to the photocurrent. In such case, the photocurrent has a sole contribution from hole (electron) population, which can be shown to scale as I under steady state conditions, [31] subsequently giving rise to the sublinear behavior of the photocurrent. As the excitation density is increased, there Adv. Mater. 2017, 29, 1604493 www.advancedsciencenews.com www.advmat.de  is a substantial contribution from the untrapped electron (hole) current. Nevertheless, given that the effective defect density is around 10 16 cm −3 , the carrier dynamics at the investigated densities still lie within the trap-limited regime, [32,33] where the electron (hole) population can be shown to scale linearly with intensity within the SRH formalism.
Since the photocurrent is proportional to the number of free charge carriers and therefore composed of electron and hole current density, j = j e + j p , we expect that the general light intensity dependence in the simplest approximation can be described with a linear and square root contribution. The ratio between this sublinear and linear component is therefore an indication for the degree of trap-limited behavior of the perovskite films, with greater linear component suggesting lower degree of carrier trapping. The photocurrent extracted from films grown on C-PCBSD in Figure 4b follows a behavior closer to the linear limit at excitation densities of 10 13 −10 15 cm −3 relevant for the PV operation, in comparison to the films grown directly on TiO 2 and TiO 2 /60-PCBM, indicating a lower density of traps. We have already highlighted above that the wettability of PbI 2 dissolved in DMF on C-PCBSD is quite poor with respect to TiO 2 , which leads to the formation of larger and flatter grains and hence a lower density of defects. Interestingly, the intensity dependence of the photocurrent in evaporated MAPbI 3 on TiO 2 and TiO 2 /60-PCBM follows exactly the same behavior, proving similar film growth conditions on both bottom layers. Thus, the choice of the EEL on the bottom strongly influences the optoelectronic property of the perovskite films and thus must be taken into account in the device optimization.
We now turn to the measurement of carrier recombination kinetics in working conditions (i.e., under 1 sun light bias), comparing the different device architectures presented in this work. First, we carried out the measurement of the device capacitances using photoinduced differential charging after subtraction of the geometrical capacitance (Figure 5a) as reported before. [34] Keeping the layers thickness comparable, these measurements allow us to compare the accumulated charge densities in solar cells at different bias. We report such densities in Figure 5b, showing that the solar cell with the C-PCBSD fullerene selective contact is capable to accumulate higher amount of charges. It is worthy to mention that the capacitance increases super linearly for voltages above 0.8 V and it is independent on the thickness of the C-PCBSD layer (see Figures S12 and S13 in the Supporting Information). This profile has been also observed in other thin film solar cells alike polymer [35] and small molecule organic solar cells. [36,37] While the linear increase of capacitance with the light bias is correlated to the accumulation of carriers at the device contacts, [38] the super linear behavior indicates that the contacts are not able to sustain the entire density of charges which accumulate in the bulk of the device as well. This is further supported by the fact that charge density, as well as the charges lifetimes, scale as the thickness of the perovskite active layer (see Figures S16-S18 of the Supporting Information). Importantly, this leads to a greater split of the quasi-Fermi levels for holes and electrons leading to a larger V oc -as long as strong recombination losses do not kick in which then points to the importance of the electronic quality of bulk semiconductor layer.
We now can compare the carrier recombination kinetics for the solar cells at equal carrier densities (n). As mentioned above, the V oc of solar cells will be influenced by the materials energetics (band alignment) and the carrier recombination kinetics. Figure 6 illustrates the measurements carried out using the same devices as in Figure 5 in identical light irradiation conditions. As can be seen, the solar cells using C-PCBSD as electron selective contact show longer carrier lifetimes and lower recombination losses. For instance at illumination intensities equivalent at 1 sun, the 60-PCBM-based devices display a total charge density of 20 nC cm −2 alike the TiO 2 -based ones and carrier lifetimes of 0.6 and 2 µs, respectively (as shown with a vertical line in Figure 6). Nonetheless, considering the same charge density for C-PCBSD perovskite solar cells the carrier lifetime is hundred times slower (τ > 200 µs). In fact, illuminating these solar cells at 1 sun the total charge was of hundreds of nC cm −2 and still the carrier recombination lifetime was the slowest Adv. Mater. 2017, 29, 1604493 www.advancedsciencenews.com www.advmat.de  between the three devices with a value close to τ = 10 µs. For a fair comparison with the decays recorded for the organic-based perovskite solar cells, we have used in Figure 6 the fastest component of the biexponential decay.
In conclusion, by employing a cross-linking strategy we demonstrate a fully solution-processed planar junction solar cell in direct architecture. The device delivers a PCE close to 19% and a V oc larger than 1.1 V. It is electrically stable, i.e., its J-V characteristics do not depend on the polarization history of the device, and accordingly the photocurrent has a virtually instantaneous response to polarization and remains stable over time. Such high reliability of the device is fully attributed to the high quality of the interface between the C-PCBSD EEL and the perovskite layer. We demonstrate that in such device architecture, under 1 sun illumination, the high carrier density achieved in opencircuit condition is not sustained by the extracting contact only, but accumulates within the active layer as well. Thus to enhance the V oc , where greater separation of the quasi-Fermi levels is desired, the carrier recombination losses via carrier trapping in the bulk perovskite films has to be minimized. Here, we show that the choice of the EEL layer not only determines the interface energetic but also the electronic quality of the active layer and, with c-PCBSD we achieve superior characteristics in both these aspects. These results, beyond the technological validity provided by the demonstration of a high efficiency, fully solution-processed, flat junction solar cell, which is electrically stable, define an important direction toward an engineered design of highly performing perovskite solar cells.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.