Solution Deposition-Conversion for Planar Heterojunction Mixed Halide Perovskite Solar Cells




Solution-deposited-converted perovskite solar cells are studied by converting PbI2 planar films into the phase pure, mixed-halide perovskite (H3CNH3)PbI3-xClx. These solar cells exhibit very high photovoltaic performance and close to unity internal incident photon-to-electron conversion.

The alkylammonium metal trihalide perovskite absorbers first used in working photovoltaic devices were based on liquid electrolyte sensitized solar cells. Introduced by Kojima et al., the devices exhibited a starting point power conversion efficiency of 3.8% and, with further work, they were quickly improved to reach over 6%.[1] It was not until a solid-state configuration was employed, however, that high device efficiencies were achieved.[2] Initial results were reported at 9% for perovskite sensitized titania-based devices[2b] and further improvements were simultaneously achieved in a “meso-superstructured” configuration by replacing the mesoporous TiO2 scaffold with an electronically inactive mesoporous Al2O3 layer, exhibiting device efficiencies of over 12%.[2c],[3] Some of the key advantages for this material system over other competing device concepts are that they are compatible with solution-processing techniques and can be fully processed at low temperatures, thus enabling their use in flexible device applications.[4]

Recently, Burschka et al. have demonstrated a method whereby an initial PbI2 film is deposited over a mesoporous TiO2 structure, which is then fully converted into the methylammonium lead triiodide (MAPbI3) perovskite via a second step.[5] The lead iodide coated substrates are immersed in a methylammonium iodide (MAI) solution in isopropanol for a short time (<1 min), resulting in the conversion of PbI2 into the perovskite phase. The resulting films were coated with the hole transporter 2,2′,7,7′-tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9′-spirobifluorene (Spiro-MeOTAD) and a metal cathode, resulting in solar cells that approach the 15% benchmark.[6] Recently, this fabrication method was extended by Liu et al. for planar heterojunction based devices in which a planar PbI2 film was deposited over a ZnO blocking layer and was then converted into the MAPbI3 perovskite in a second step.[7] This resulted in perovskite crystal sizes ranging from 100 to 1000 nm and an average thickness of ≈300 nm. The resulting device performance of 15.7% is currently the highest performance achieved for perovskite solar cells, pointing towards planar heterojunction devices as a promising device architecture for further technological improvements.

The short circuit currents demonstrated for the devices prepared by Liu and co-workers of 20.4 mA cm-2,[7] while high, are still short of the maximum current of over 22 mA cm-2 reasonably achievable, taking into account other light capture losses for this material.[3a] A crucial limitation in this respect is the low diffusion length of around ≈100 nm of the photoexcited species in the MAPbI3 perovskite.[8] This parameter can be greatly extended to over 1 μm with the inclusion of chloride in the precursor solution.[8a],[9] Furthermore, it has been recently shown that the inclusion of chloride is beneficial for charge transport in the photoactive layer.[10] It is expected that the addition of chloride results in improved short circuit currents and thus overall photovoltaic performance. It is worth noting here that for devices incorporating mesoporous TiO2 photoanodes, the neat tri-iodide perovskite functions efficiently without the need for the extended diffusion length of the photoexcited species.[11] This is a result of the interpenetrated nature of the collection photoanode, which exhibits pore sizes at the order of tens of nanometers, and in effect reduces the distance electrons must travel to this magnitude before being collected. In the case of planar heterojunctions, electrons must travel the entire thickness of the film, which can sometimes exceed hundreds of nanometers and thus extended diffusion lengths are a requirement for efficient operation.

Here we present planar, fully solution-processed heterojunction solar cells based on the solution deposition-conversion technique. We highlight that chloride is critical in MA lead halide perovskites via a controlled addition of methylammonium chloride (MACl) to the MAI immersion solution. The resulting devices exhibited power conversion efficiencies approaching 15%, and more importantly, showed short circuit currents of over 22 mA cm−2, representing a gain of over 10% over state-of-the-art devices.[7] The parameter most influenced by the presence of chloride is the photoluminescence lifetime of the photoexcited species in the device, which reaches values exceeding 300 ns, matching previously reported results for the solution processed mixed halide perovskite films.[8a] Additionally, a reduction of series resistance from 14 to 7 Ω cm2 was observed.

The solar cells developed in this work are composed of a TiO2/perovskite/Spiro-MeOTAD planar heterojunction, deposited on a fluorine-doped tin oxide (FTO) electrode and capped with a gold electrode (Figure 1). The perovskite deposition was performed in two steps: firstly, an ≈200 nm PbI2 film was deposited via spin coating, followed by full conversion into the MAPbI3-xClx perovskite via immersion in a heated solution of a mixture of MAI and MACl in isopropanol (IPA). A cross section of the final device structure can be seen in Figure 2e. The initial PbI2 film is shown in the cross sectional image in Figure 2a,b in which we can observe distinct PbI2 sheets, oriented flat over the non-porous TiO2 layer. The uniformity of the layer can be assessed from Figure 2b where we show an extended view of the cross section. Additionally, from the top views shown in Figure 2c,d we can see that the layer is smooth and covers 100% of the TiO2-coated FTO glass substrate. After immersion, we can clearly observe the formation of the perovskite crystals, as shown in Figure 2g, with a mixture of crystal sizes ranging from ≈100 to almost 600 nm. As can be seen in the low magnification image in Figure 2h, the surface coverage is approximately 100%.

Figure 1.

Schematic of the device structure. The expected charge transfer process is denoted with arrows.

Figure 2.

a,b) Scanning electron microscopy (SEM) cross-sectional images of the as-deposited PbI2 film on TiO2-coated FTO. SEM top view images of the same PbI2 films, c) with high magnification and d) with low magnification. e,f) SEM cross-sectional images of the fully converted PbI2 film (from the top row) into MAPbI3-xClx, g) with high magnification and h) low magnification. All layers are labeled in the images.

To assess the impact of chloride addition on the structure, the films were analyzed via powder X-ray diffraction (XRD) after removal from the substrate. The perovskite films presented here are phase pure, crystallized in the expected tetragonal I4/mcm space group.[12] No impurities were found, in contrast with other solution-deposition methods, which usually exhibit a small fraction of crystalline PbI2 and MAPbCl3 (Supporting Information Figure S1–3).[2c],[10] However, the change of the lattice parameters reported for perovskites formed within mesoporous scaffolds, which was attributed to the presence of chloride, was not observed for the system presented here.[10] Moreover, samples of MAPbI3-xClx perovskite prepared according to the established protocol in previous studies (spin-coating from a mixture of MAI/PbCl2) also did not show a significant change of lattice parameters.[10, 13] A similar result was found for films grown within mesoporous Al2O3 templates (Supporting Information Figure S2). This suggests that the vast majority of the chloride ions present in the precursor solution are not incorporated in the final structure. The chloride content in the structure was under the detection limit of both energy-dispersive X-ray spectroscopy (EDX) and electron energy loss spectroscopy (EELS) and could not be quantified (Supporting Information Figure S4,5). We have also performed X-ray photoelectron spectroscopy (XPS) measurements to examine whether the chloride is located on the crystal surface; this is shown in Supporting Information Figure S6. Again, as was the case for other detection methods, the amount of chloride was below the detection limit of the instrument. Additionally, we have performed coulometric titration with silver ions and colorimetry with mercurythiocyanate and Fe3+ ions. However, the iodide signal present in all these measurements heavily interferes with the small chloride signal and therefore the amount of the latter in the structure could not be quantified. For experiments with more than 20 wt% of MACl in the immersion solution a secondary MAPbCl3 phase was formed and thus this concentration was chosen as the upper limit.

The photovoltaic performance of devices incorporating perovskite films fabricated with a range of MACl concentrations in the immersion solution is shown in Figure 3. We can clearly see major increases in all device parameters, particularly short circuit current and fill factor, even for low MACl concentrations. The maximum performance was obtained for concentrations of 5 wt% MACl, which resulted in a champion device exhibiting almost 15% power conversion efficiency, 22.8 mA cm−2 short circuit current and an open circuit voltage approaching 1 V under 104 mW cm−2 equivalent AM 1.5 sunlight.

Figure 3.

a) Current density–voltage (JV) curves measured under AM 1.5 solar irradiance and 104 mW cm−2 equivalent light intensity conditions for the best performing MA lead halide perovskite devices for a range of MACl concentrations in the immersion solution. The inset summarizes the photovoltaic parameters: short-circuit current (Jsc, mA cm−2), open-circuit voltage (Voc, V), fill factor (FF, %), power conversion efficiency (η, %), and series resistance (Rs, Ω cm2). A spectral mismatch factor of 1.02 was calculated for the devices,[18] resulting in an equivalent 104 mW cm−2 irradiation intensity. b) The efficiency values are shown as box plots for efficiency distributions for a range of MACl wt% concentrations in the immersion solution. Whiskers represent the 10/90 percentile while box edges represent the 25/75 percentile. Small square symbols inside the boxes represents the mean, while the line across the boxes represents the median. The × symbols represent the maximum and minimum values. Each box represents the statistical distribution of between 30 and 40 working devices prepared under similar conditions.

The average power conversion efficiency for the optimum fabrication protocol including MACl is 10.5%, as shown in Figure 3b, which is much more efficient than the 5.5% found for neat triiodide. A histogram of device power conversion efficiencies is shown in Supporting Information Figure S7. An incident photon to current conversion efficiency (IPCE) spectrum is shown in Supporting Information Figure S8. Once this peak MACl concentration is exceeded, the devices begin to exhibit both lower short circuit currents and lower voltages. This is very likely due to the observed morphology changes of the films, which exhibit average crystal sizes exceeding those of the neat triiodide by approximately 50% (Supporting Information Figure S9,10), and the appearance of a larger number of gaps between the crystals.[8a]

Additionally, we quantified the series resistance present in the devices (see inset of Figure 3a) and observe a decrease from 14 ± 1 to 7 ± 1 Ω cm2, upon addition of chloride. This clearly accounts for the improvement of the fill factor measured and is consistent with previous observations of mixed halide perovskites.[10] The neat MAPbI3 solar cells presented here perform significantly worse than state-of-the-art devices presented by Liu and co-workers.[7] This is very likely due to the larger grain size present in our films, as well as the use of a TiO2 blocking layer, which increases series resistance and does not collect charges as efficiently at higher biases. However, the important parameter in terms of this work, the short-circuit current, is comparable between the two studies and thus represents a valid platform to compare neat and mixed halide perovskite solar cells.

The measured short circuit currents are consistent with those estimated from light absorption measurements inside an integrating sphere (Supporting Information Figure S11). We also find that increasing the MACl content in the immersion solution improves light absorption at the bandgap edge due to the marginally higher scattering of these films. However, there is little room for improvement in this regard for the devices prepared here, as only ≈5% extra photocurrent could be gained if 100% of the incoming light were to be absorbed. All the additional losses in overall light harvesting arise from parasitic absorption in the FTO and reflection at the glass/air and glass/FTO interfaces. If no losses were present in the system, a maximum short circuit current of 27.15 mA cm−2 could be achieved. Both films fabricated from mixed halide or neat MAI absorb light similarly, however, neat MA lead triiodide films exhibit a 10% lower short circuit photocurrent.

In order to understand this difference, we performed time-resolved photoluminescence measurements, as shown in Figure 4, and fitted the resulting curves with a biexponential decay, where Afast and τfast correspond to the amplitude and lifetime respectively for the faster decay component, and Aslow and τslow correspond to the amplitude and lifetime respectively for the slower decay component. The fitted parameters are summarized in Table 1. This technique yields important information about the lifetime of excited species in photoactive materials and can be used to extract their diffusion length by comparing films with and without a selective charge quencher.[8] Here, slower decays can be correlated with longer diffusion lengths, indirectly giving a measure of charge collection efficiency for a given film thickness. For the excitation intensity and wavelength used, the PL is expected to arise mostly from recombination of free charges, as the exciton binding strength for this system is relatively low (Eb ≈ 50 meV).[14] For all samples measured, we clearly observe biexponential decay dynamics, which may be a result of the widely varying crystal sizes present in the films, ranging from very small crystals of around 100 nm to some that approach the micrometer scale. Neat MA lead triiodide films exhibit lifetimes for the fast component of approximately 18 ns and a slow component of ≈55 ns (Table 1). We note that these decay dynamics are significantly slower than those measured previously for the same tri-iodide system deposited via spincoating,[8] even for the fast component measured in the devices presented in this work.

Table 1. Summary of the biexponential fit decay parameters for the PL decays shown in Figure 4
MACl content in solution [wt%]Afast [A.U.]τfast [± 1 ns]Aslow [A.U.]τslow [± 1 ns]
Figure 4.

Time-resolved PL decay plots of MA lead mixed halide perovskite systems for a range of wt% MACl concentrations in the immersion solutions. The excitation wavelength was chosen to be 507 nm, the laser head was pulsed at 500 kHz with an excitation fluence of 0.3 μJ cm−2/pulse, and the resulting PL was measured at 778 nm.

The slower decay dynamics measured here imply that the charge diffusion length is longer than measured previously, allowing efficient charge collection in thicker films, thus a larger fraction of the solar spectrum can be harvested. Additionally, we note that previous measurements on the PL lifetimes for neat tri-iodide films were performed on films consisting of very small crystals in the range of tens to a hundred nanometers.[8] We can clearly see that the addition of chloride has a dramatic effect on the dynamics of the photogenerated species, where long lifetime values of over 300 ns were found for films immersed in a solution mixture exceeding 5 wt% of MACl, which is consistent with recent results for mixed halide perovskites.[8a] Increasing the concentration of MACl beyond this point does not result in longer lifetimes. This is consistent with previous calculations, which predict that only a very small fraction of the chloride ions can in fact be incorporated in the structure.[10]

The differences in PL lifetime may account for the observed losses in short circuit photocurrent. Diffusion lengths could not be calculated for the presented films since the roughness of the perovskite film prevents a reasonable estimation using previously employed techniques.[8, 15] However, as a first approximation we may use literature values for the diffusion coefficient (De) in MA lead triiodide films and scale the diffusion length (LD) according to our measured PL lifetimes (τ) as inline image. Using this method, we can roughly estimate a 200 nm diffusion length for neat tri-iodide films and over 800 nm for samples that have been treated with chloride. A diffusion length of 200 nm, while longer than estimated previously,[8b] is not sufficiently long for the films presented in this work, where a large fraction of the crystals clearly exceed this thickness (Supporting Information Figure S8). Moreover, to achieve close to 100% charge collection, the diffusion length must approach three times the film thickness.[16] Thus, we can account for the photocurrent losses in samples that have not been prepared with chloride.

In conclusion, we have demonstrated that the short-circuit current of solution-processed planar heterojunction solar cells can be improved via the addition of chloride in the immersion solution. We have shown that planar PbI2 films can be fully converted to the MAPbI3-xClx perovskite structure within 5 min by immersion in a heated solution mixture of MAI and MACl. We find that the addition of chloride critically impacts the lifetime of photoexcited species in the active material, increases light absorption at the bandgap edge, and reduces the device series resistance, thus enabling nearly complete sunlight capture in 400 nm thick films. It is worth noting that the diffusion length of photoexcited species in the neat tri-iodide perovskite prepared via the deposition-conversion technique is roughly 200 nm; this is longer than what is estimated for films deposited via other methods. For this reason, films prepared in this way perform well compared to previous reports of solution-processed MAPbI3 planar heterojunction devices.[8, 17] This indicates that the photovoltaic properties of the material are highly sensitive to the method of film formation and crystallization.

Experimental Section

Preparation of the Methylammonium Salts: Methylammonium iodide was prepared following a previous report.[2c] In short, 24 mL of methylamine solution (33% in ethanol) was diluted with 100 mL of absolute ethanol. A 10 mL aqueous solution of hydriodic acid (57 wt%) was added to this solution under constant stirring. After a reaction time of 1 h at room temperature, the solvents were removed by rotary evaporation. The obtained white solid was washed with dry diethyl ether and finally recrystallized from ethanol.

To prepare the hydrochloride salt of methylamine, the hydriodic acid solution was replaced by 15 mL of concentrated hydrochloric acid (37% in water). The purification procedure was the same as described above.

Solar Cell Preparation: FTO coated glass sheets (7 Ω/, Pilkington) were etched with zinc powder and HCl (2 M) to obtain the required electrode pattern. The sheets were then washed with soap (2% Hellmanex in water), deionized water, acetone, and methanol and finally treated under oxygen plasma for 5 min to remove the last traces of organic residues. The substrates were then coated with a sol-gel derived TiO2 layer, prepared as previously described and calcined at 500 °C in air to achieve full anatase crystallization.[3a]

A 27.2 mM HCl solution in 2-propanol (Sigma Aldrich) (typical quantities of 2.53 mL of 2-propanol and 35 μL of 2 M HCl) was slowly added dropwise under vigorous stirring to a 0.43 M titanium isopropoxide solution in 2-propanol (typically 369 μL titanium isopropoxide in 2.53 mL of 2-propanol). The resulting solution was clear and transparent and was immediately discarded if cloudy.

A ≈ 200 nm layer of lead iodide was deposited via spincoating from a 1 M PbI2 solution in N,N-dimethylformamide (DMF) at 3000 rpm for 15 s. To achieve optimum performance, it was critical to ensure that both the substrate and precursor solution temperature when starting the spincoater was between 60 and 65 °C. The spincoating was performed dynamically (i.e., the solution was added while the substrate was spinning) with 100 μL of solution, without allowing the substrates or solution to cool.

The stock immersion solutions were prepared by dissolving 10 mg mL-1 methylammonium iodide or methylammonium chloride in dry isopropanol, the latter with heating to 60 °C. For mixing the desired concentrations, these stock solutions were combined in the desired ratio. Before immersion, 40 mL of the solution and the PbI2 films were heated to 60 °C on a hotplate. The temperature of the solution was monitored during the whole time with a thermometer. After immersion of the films into the solution in a square petri dish with the substrates face up, the solution was left undisturbed for 5 min. Once the conversion was finished, the films were washed with clean, anhydrous isopropanol and dried under a nitrogen stream.

After drying, the films were covered with a 400 nm layer of Spiro-OMeTAD (Borun Chemicals, 99.1% purity). 96 mg of Spiro-OMeTAD were dissolved in 1 mL of chlorobenzene and mixed with 10 mL 4-tert-butylpyridine (tBP) and 30 μL of a 170 mg mL−1 bis(trifluoromethane)sulfonimide lithium salt (LiTFSI) solution in acetonitrile. This solution was spincoated at 1500 rpm for 45 s. Before evaporating the gold electrodes, Spiro-OMeTAD was allowed to oxidize in air over night at room temperature and 15–20% relative humidity.

PL Sample Preparation: For the preparation of the PL samples glass slides were used instead of FTO substrates. The PbI2 deposition and immersion procedure in the MA salt solutions was the same as described above. It was noted that exposure of the glass slides to oxygen plasma was essential in order to achieve optically smooth films. After drying of the converted films in a nitrogen stream, they were covered with a poly(methyl methacrylate) (PMMA) layer to prevent degradation by ambient moisture. For this purpose, 10 mg of PMMA were dissolved in 1 mL of chlorobenzene and spin coated at 1000 rpm for 45 s.

Characterization Details: Solar simulated AM 1.5 sunlight was generated with an ABET class AAB solar simulator calibrated to give 102 mW cm−2 using an NREL calibrated KG5 filtered silicon reference cell. The spectral mismatch factor was calculated to be 2%. The JV curves were recorded using a Keithley 2400. The active area of the solar cells was defined with a metal aperture mask of about 0.08 cm2.

Steady-state absorption spectra of the perovskite films on glass/FTO/TiO2 substrates were acquired with a Varian Cary 300 UV/Vis spectrophotometer using an integrating sphere. To account for the highly reflective nature of the metal cathode, the optical densities were doubled as a first approximation in order to estimate the maximum short circuit currents achievable in complete devices. The FTO absorption was quantified as reported previously,[3a] where transmittance and reflectance were measured for the glass/FTO/air system in an integrating sphere.

IPCE spectra were obtained using a Fourier transform photocurrent spectrometer, incorporating a Fourier transform spectrometer (Vertex 80v, Bruker), current preamplifier (SR570, Stanford Research Instruments) and custom-built control electronics and software. The spectrometer was configured with a tungsten-halogen light source and CaF2 beam splitter. Photocurrent was recorded from short-circuited devices following the application of a 2 V forward bias. The spectra were calibrated against measurements taken with the same system on a reference silicon photodiode with a known IPCE spectrum. The solar cells and reference diode were masked with a metal aperture to define the active area of ≈0.0625 cm2.

Time-resolved PL measurements were acquired using a time-correlated single-photon counting (TCSPC) setup (FluoTime 300, PicoQuant GmbH). The samples were photoexcited using a 507 nm laser head (LDH-P-C-510, PicoQuant GmbH) pulsed at 500 kHz, with a pulse duration of 117 ps and fluence of ≈300 nJ cm−2/pulse. The samples were exposed to the pulsed light source set at 3 μJ cm−2/pulse fluence for ≈10 min prior to measurement to ensure stable sample emission. The PL was collected using a high resolution monochromator and hybrid photomultiplier detector assembly (PMA Hybrid 40, PicoQuant GmbH).

SEM images were obtained using a Hitachi S-4300 microscope. EDX spectra were acquired with an EDAX Si(Li) detector. The spectrum shown in Supporting Information Figure S4 was acquired over 32 min at a dwell time of 102.4 μs and 5 eV/channel. EEL spectra were recorded with a post column filter (Gatan Tridiem 863 P) at a dispersion of 0.2 eV in diffraction mode (camera length 102 mm).


P.D. and F.H. contributed equally to this work. The authors acknowledge funding from the Excellence Cluster “Nanosystems Initiative Munich” (NIM), the Center for NanoScience (CeNS) and from the Bavarian Research Network “Solar Technologies Go Hybrid”. The authors acknowledge support from the European Union through award of a Marie Curie Intra-European Fellowship. This article was modified after online publication to include the middle initial for author F.C.H.