A vacuum flash–assisted solution process for high-efficiency large-area perovskite solar cells

See allHide authors and affiliations

Science  01 Jul 2016:
Vol. 353, Issue 6294, pp. 58-62
DOI: 10.1126/science.aaf8060


Metal halide perovskite solar cells (PSCs) currently attract enormous research interest because of their high solar-to-electric power conversion efficiency (PCE) and low fabrication costs, but their practical development is hampered by difficulties in achieving high performance with large-size devices. We devised a simple vacuum flash–assisted solution processing method to obtain shiny, smooth, crystalline perovskite films of high electronic quality over large areas. This enabled us to fabricate solar cells with an aperture area exceeding 1 square centimeter, a maximum efficiency of 20.5%, and a certified PCE of 19.6%. By contrast, the best certified PCE to date is 15.6% for PSCs of similar size. We demonstrate that the reproducibility of the method is excellent and that the cells show virtually no hysteresis. Our approach enables the realization of highly efficient large-area PSCs for practical deployment.

In the span of a few years, the power conversion efficiency (PCE) of perovskite solar cells (PSCs) has risen from 3.8% (1) to 22.10% (2), which is unprecedented in the field of photovoltaics. However, such high efficiencies have been achieved only with cells of very small size—between 0.04 and 0.2 cm2—and few investigators have attempted to fabricate larger-area cells (310). The use of small-area devices has raised some doubts on the remarkable progress of the PSC field because the measurement errors tend to increase as the active cell area becomes smaller. Thus, the development of PSCs with a mandatory minimum active area of >1 cm2 is required for the evaluation of this new photovoltaic (PV) technology (3, 8). At present, the best certified PCE of a cell with a size exceeding the critical threshold of 1 cm2 is 15.6% (11) because of the limitations of current preparation methods (1215). Top-performing PSCs often are made with an antisolvent such as chlorobenzene to precipitate the perovskite or its intermediate from its solution, which typically contains a solvent mixture of γ-butyrolactone (GBL), dimethylformamide (DMF), and dimethylsulfoxide (DMSO) (3, 7, 8, 14, 1619). The antisolvent induces oversaturation of the perovskite solution, but because the antisolvent is usually dripped in the center of the film during spin-coating, the result is a radial gradient in oversaturation; the spatially inhomogeneous nucleation of the perovskite or its intermediate ultimately leads to defects in the perovskite film (3, 20). In addition, the antisolvents currently used are toxic and harmful to the environment, hampering their large-scale application (8, 14). Thus, alternative procedures for preparing large-area PSCs are warranted if their performance on a large device area is to be competitive with that of inorganic thin-film photovoltaics.

We developed a simple and effective method to produce high-quality perovskite films for large-area PSCs by applying a vacuum-flash treatment during the solution processing of the perovskite. Our approach differs from previous studies that used high-vacuum methods for vapor deposition of the perovskite (13, 15) or for removal of reaction products (i.e., methylammonium chloride) by sublimation during the thermal annealing of the films (21). Vacuum flash–assisted solution processing (VASP) is a method that enables the sudden and well-controlled removal of solvent, thereby boosting rapid crystallization of a fibrous material that consists of a Lewis acid-base–type adduct representing the perovskite precursor phase (19). Upon thermal annealing, the precursor phase produces highly oriented, crystalline perovskite films of excellent electronic quality that can be grown on a variety of substrates (1719). Furthermore, VASP allows deposition of the perovskite films on large substrate sizes and can be turned into a continuous process.

We achieved a maximum PCE of 20.5% and a certified PCE of 19.6% for large cells with square aperture areas greater than 1 cm2, which is commensurate with the 21.0% reached by today’s best thin-film copper indium gallium selenide (CIGS) and CdTe devices of similar size (11). Our method also eliminates the hysteresis in the current-voltage (J-V) curves, a notorious problem with PSCs (22). We tested the method with state-of-the-art perovskites using formamidinium (FA) and methylammonium (MA) mixed-cation and iodide-bromide mixed-anion perovskite formulations (16, 23) of composition FA0.81MA0.15PbI2.51Br0.45. To demonstrate the versatility of the method, we also prepared the emerging cesium (Cs+) and FA mixed-cation perovskite FAxCs1–xPbIyBr3–y formulations (2426). The VASP method is also readily scalable to the industrial level.

The basic steps of perovskite film fabrication by the VASP method are shown in Fig. 1A. The perovskite precursor solution, of composition FA0.81MA0.15PbI2.51Br0.45 containing DMSO with a nominal 1:1 ratio of lead to DMSO, was first spin-coated on top of a mesoporous TiO2 film prepared as described (16). The film was then placed for a few seconds into a vacuum chamber to boost rapid crystallization of the perovskite intermediate phase by removing most of the residual solvents, consisting mainly of GBL and DMF. We observed that the pressure applied during the vacuum-flash process has considerable influence on the perovskite film formation. We obtained optimal results with a pressure of 20 Pa and used this condition during further studies (fig. S1). The film darkened slightly after the VASP treatment and formed a transparent orange layer of the perovskite precursor. We confirmed the presence of DMSO in this transparent orange film with Fourier transform infrared spectroscopy (fig. S2). Thus, the intermediate phase represents a Lewis acid-base adduct of DMSO with the likely composition DMSO-PbI1.7Br0.3-(FAI)0.85(MABr)0.15 (14, 1719). Further analysis by scanning electron microscopy (SEM) showed that this intermediate product consisted of nanofibrous aggregates (fig. S3), which upon annealing for 30 min at 100°C transformed into the shiny, smooth, and highly crystalline perovskite layer.

Fig. 1 Perovskite film deposition and device structure.

(A) Schematic illustration of nucleation and crystallization procedures during the formation of perovskite film via vacuum flash–assisted solution processing (VASP). (B) Schematic illustration of the perovskite solar cell configuration, where a smooth and compact perovskite capping layer fully covers the mesoporous TiO2 layer (mp-TiO2) infiltrated with perovskite. bl-TiO2, TiO2 compact layer. (C) A high-resolution cross-sectional SEM image of a complete solar cell fabricated by VASP.

We spin-coated a hole-transporting layer onto the perovskite film [2,29,7,79-tetrakis(N,N-di-pmethoxyphenylamine)-9,9-spirobifluorene (spiro-OMeTAD) containing tert-butyl-pyridine (t-BP) and lithium bis(trifluoromethylsulfonyl)imide (LiTFSI) as an additive] (16). Finally, an 80-nm gold layer was evaporated on the hole transport layer to produce the full device. We illustrate the mesoscopic-planar bilayer device architecture in Fig. 1B and show a cross-sectional SEM image of this embodiment in Fig. 1C.

SEM images reveal stark differences between the FA0.81MA0.15PbI2.51Br0.45 perovskite films without and with vacuum-flash treatment (Fig. 2, A and B). The top view (Fig. 2A) of a film prepared by the conventional single-step solution deposition process without the vacuum-flash step (abbreviated as CP) (27) shows that the mesoporous TiO2 was not fully covered by the perovskite. Pigment aggregates formed islands surrounded by numerous pinholes, apparently because the film dewetted during the formation of the perovskite (28). The VASP method yielded homogeneous films without pinholes; the TiO2 was fully covered by the perovskite grains with sizes between 400 and 1000 nm, which greatly exceeds the grain size of films prepared by the CP method. This is illustrated by the top-view SEM image shown in fig. S1B. The cross-sectional SEM images (Fig. 2B) confirm the difference in morphology between films subjected and not subjected to VASP.

Fig. 2 Microscope images.

(A and B) Surface (A) and cross-sectional (B) SEM images of the perovskite films fabricated by the conventional process (CP) and VASP. (C) Left: Typical AFM image of VASP-treated perovskite film taken from one of the nine square-shaped spots, each 10 μm × 10 μm. Right: Optical image of the whole square-inch-size perovskite film. The yellow squares mark the locations of the nine spots on the surface of the perovskite layer examined by AFM.

Without vacuum-flash treatment, a large fraction of the TiO2 remained exposed and only a part was covered by the perovskite, but if VASP was applied, the perovskite was well infiltrated into the mesoporous TiO2 film and formed a continuous TiO2-perovskite nanocomposite that was covered by a contiguous and compact capping layer ~400 nm thick. The size of most of the perovskite crystallites in the capping layer was commensurate with its thickness, and most of the grain boundaries were perpendicular to the substrate to minimize the grain boundary energy (29). Hence, very few grain boundaries are visible within the plane of the capping layer. As shown below, the small amount of grain boundaries retarded nonradiative charge carrier recombination, which enhanced the open-circuit voltage (Voc) of the cell (16).

To further scrutinize the uniformity of the VASP-treated FA0.81MA0.15PbI2.51Br0.45 perovskite film over large areas, we performed atomic force microscopy (AFM) to assess the surface roughness of nine square-shaped spots (each 10 μm × 10 μm in size) that were uniformly distributed over a film surface area of 1 cm2. A typical AFM topography of one of the spots is illustrated in the left panel of Fig. 2C. The apparent grain size observed in the AFM image is consistent with that seen in the SEM image in Fig. 2A. Furthermore, all nine spots showed about the same root mean square roughness value of 30 ± 5 nm, reflecting a promising scalability of the VASP technique with respect to morphological control of perovskite thin films.

The SEM top view of the fresh film treated by VASP (fig. S3) holds vital clues about the formation of the high-quality perovskite polycrystalline film from the intermediate phase. The image reveals the presence of particles consisting of agglomerated nanofibers. The nanofibers are likely a perovskite adduct with the DMSO solvent, of composition DMSO-PbI1.7Br0.3-(FAI)0.85(MABr)0.15 (fig. S1). Besides the nanofibers, a few bright particles were observed that probably represent perovskite grains. Upon annealing at 100°C, the fibrous intermediates rearranged and coalesced via the Ostwald ripening process to minimize surface energy, which led to a homogeneous polycrystalline film with large perovskite grains and an optimized crystal orientation (17). This mechanism agrees well with the cross-sectional SEM image of the VASP-treated perovskite film. During the conventional deposition process, the spin-coating of the perovskite precursor solution resulted in a metastable film containing residual solvent (such as GBL or DMF). The subsequent annealing of the film above its glass transition temperature increased the mobility of the perovskite precursor, which resulted in dewetting and uncontrolled morphological variations as well as a wide range of grain size distribution (28, 30). The VASP treatment prevented dewetting because flash evaporation of the solvent augmented the viscosity and glass transition temperature of the fluid thin film (apart from oversaturating the solution) and produced a burst of perovskite precursor crystals. The intermediate formation of the DMSO adduct retarded the crystal growth and increased the perovskite grain size during subsequent annealing (19).

The x-ray diffraction (XRD) spectra of the FA0.81MA0.15PbI2.51Br0.45 perovskite films (Fig. 3A) prepared by both methods show that for films prepared by the VASP method, the perovskite reflection at 2θ = 14.15° was 6.4 times as intense as the intensity of the peak from the fluorine-doped tin oxide (FTO) at 2θ = 37.73°, whereas for films prepared by CP, the ratio was only 1.4. This difference arises from the much superior shielding of the FTO serving as electron collector (Fig. 1A) by the contiguous and compact character of the perovskite capping layer formed via the VASP method. The CP reference sample lacked homogeneity, which is apparent from the photographs of the two films (Fig. 3B, inset). From the XRD spectra, we inferred that the vacuum-flash treatment also affects the crystal orientation. Thus, the ratio of relative intensity of Embedded Image to Embedded Image diffraction peaks (31) increased from 5.0 for the CP-deposited film to 9.3 for the VASP-treated film. The peaks are stronger and sharper for the films treated by VASP; the full width at half maximum (FWHM) of the Embedded Image peak decreased from 0.16° for CP to 0.11° for VASP, reflecting the expansion of the mean size of perovskite crystallites from 68 to 109 nm. These results indicate that the vacuum-flash treatment can simultaneously improve perovskite crystallinity and grain size.

Fig. 3 Additional analyses of perovskite films prepared by the CP and VASP methods.

(A to C) XRD patterns (A), UV-vis spectra (B), and time-resolved photoluminescence (PL) decay (C) of representative perovskite films deposited on FTO covered by a compact blocking layer and a mesoscopic scaffold of TiO2 using VASP (red trace) or CP (black trace). Time-resolved photoluminescence dynamics was performed by monitoring the photoluminescence from the bulk perovskite (excited from the side of the capping layer) at 760 nm; the green lines are fits to Eq. 2.

Figure 3B shows ultraviolet-visible (UV-vis) absorption for the perovskite films. The smooth, compact morphology of the VASP perovskite films led to a much stronger absorbance than the CP reference sample in the 500- to 800-nm range. Time-resolved photoluminescence decay measurements (Fig. 3C) of the perovskite band-gap emission at 760 nm allowed us to derive quantitative information on the dynamics of charge carrier recombination in the perovskite films. Much faster photoluminescence decay was seen for the CP reference than for the VASP-treated film. Neglecting Auger-type recombination, we describe the kinetics of charge carrier recombination by the differential rate law,Embedded Image (1)where k1 and k2 are the first- and second-order rate constants for nonradiative (trap-controlled) and bimolecular radiative recombination of charge carriers, respectively. Integration of Eq. 1 yields Eq. 2, where n0 and n denote the concentration of charge carriers at time 0 and t:Embedded Image (2)From fitting the luminescence decays in Fig. 3C to this rate equation, drawn as solid line through the data points, we obtain the values of the rate constants for the nonradiative decay processes: k1 = 3 × 106 s−1 and k1 = 8 × 105 s−1 for the CP and VASP films, respectively. These values imply that the nonradiative recombination was slower by a factor of ~4 in the VASP-treated film. The slower nonradiative recombination arising from the lower defect concentration should entail an increase in Voc, and the average Voc of the perovskite cells prepared by VASP was 192 mV larger than that of the reference (Fig. 4A) (16).

Fig. 4 Photovoltaic characterization.

(A) PV metrics for 20 devices fabricated by the CP and VASP methods. (B) Current-voltage (J-V) curves for the best performing devices using perovskite films prepared by the CP (black) or VASP (red) method, measured under standard AM1.5 solar radiation. (C) Solid lines are IPCE curves of cells fabricated by the CP (black) or VASP (red) method. Measurements were taken with chopped monochromatic light under a white light bias corresponding to 5% solar intensity. Dashed lines show Jsc calculated from the overlap integral of the IPCE spectra with the standard AM1.5 solar emission. (D) J-V curves for the best cell fabricated by VASP, recorded in reverse (from Voc to Jsc) and forward (from Jsc to Voc) scanning directions. Inset: PV metrics derived from the two J-V curves as well as their average PCEs. (E) PV parameters for a representative FA0.81MA0.15PbI2.51Br0.45-based perovskite device fabricated by VASP measured from five different spots with an aperture area of 0.16 cm2 selected from the total active area of 1.2 cm × 1.2 cm under standard AM1.5 illumination. All J-V curves were recorded at a scanning rate of 50 mV s−1 in the reverse direction unless stated otherwise.

We fabricated mesoscopic-planar PSCs using perovskite thin films of composition FA0.81MA0.15PbI2.51Br0.45 as light harvesters and evaluated the PV metrics of the devices by measuring their photocurrent density versus voltage (J-V) curves and their incident photon-to-current conversion efficiency (IPCE). We also tested their stabilized maximum power output. We covered devices with an active area of 1.44 cm2 by a black mask with an aperture area of 1.0 cm × 1.0 cm. Tables S1 and S2 provide statistical data on the PV metrics of the PSCs. As shown in Fig. 4A, VASP greatly improved device performance and reproducibility relative to the CP method. The average PCE for 20 control PSCs fabricated by CP was 9.62 ± 0.67%, with an average short-circuit photocurrent density (Jsc) = 17.25 ± 0.49 mA cm−2, Voc = 938 ± 15 mV, and fill factor (FF) = 0.58 ± 0.03. For the VASP films, the average PCE more than doubled to 19.58 ± 0.37%, with an average Jsc = 23.12 ± 0.28 mA cm−2, Voc = 1130 ± 8 mV, and FF = 0.74 ± 0.01. J-V curves are shown for the best cells in Fig. 4B. The substantial performance improvement induced by VASP is reflected in the values for all the PV metrics. The best cell produced by CP gave a PEC of 10.79%, corresponding to JSC = 15.60 mA cm−2, VOC = 949 mV, and FF = 0.61. The values for the VASP perovskite film were JSC = 23.19 mA cm−2, VOC = 1141 mV, and FF = 0.76, reaching a PCE of 20.38% under standard AM1.5 solar illumination. Figure 4C shows IPCE over the spectral range from 400 to 800 nm. The VASP technique achieved much higher IPCE values than the CP method, matching the difference in device photocurrents obtained from these two procedures. Integration of the IPCE spectrum over the solar emission yields AM1.5 photocurrents of 23.2 and 17.8 mA/cm2 for the VASP and CP reference devices, respectively, in excellent agreement with the measured JSC values; this finding indicates that the spectral mismatch between our simulator and the true AM1.5 solar emission is negligibly small.

We used different J-V scan directions to examine the hysteresis of our cells fabricated by VASP (Fig. 4D). The PCE difference between the forward and reverse scan was as small as 1% in absolute PCE values, reflecting negligible hysteresis. The enlarged perovskite crystal size and improved crystalline quality eliminated hysteresis, likely by reducing the overall bulk defect density and suppressing charge trapping during solar cell operation (32, 33). A 20.32% stabilized PCE output with a photocurrent density of 21.25 mA cm−2 was achieved for the same large-area device in Fig. 4D (fig. S4), indicating that the PCEs obtained in forward and reverse scans at a routine scanning condition are near the real performance. We also measured J-V curves at five different small spots (each 0.4 cm × 0.4 cm) located at the center and the four corners of the device active area (Fig. 4E). All the PV metrics including JSC, extracted from these five curves by a black mask with a square aperture area of 0.16 cm2, were almost identical (table S3); the values of JSC show very small variation when the square aperture area of the mask is increased up to 1 cm2 (fig. S5), attesting to the uniformity of the perovskite film over the square-centimeter scale. One of our 1.0 cm × 1.0 cm devices was tested by an accredited PV test laboratory (Newport Corporation PV Lab, Bozeman, MT, USA), which certified its PCE to be 19.6%, with Jsc = 22.60 mA cm−2, Voc = 1.143 V, and FF = 0.76 (fig. S6). A preliminary stability investigation showed that the devices are stable in ambient conditions, their final PCE even increasing slightly after 39 days (fig. S7 and table S4). Figure S8 compares the stability of two typical devices prepared by the VASP and CP methods under light soaking with AM1.5 light of 100 W/m2 intensity. After 100 hours of continuous illumination, the VASP device and CP standard retained 90% and 70% of their initial performance, respectively, showing the superior light resistance of the VASP-based cell.

We also produced PSCs of 1 cm2 size using the antisolvent method as described (16) and measured their PV performance. Data on 10 cells are shown in table S5. The average PCE was 16.99 ± 0.71%, which is obviously lower and exhibits a larger standard deviation than that of 19.58 ± 0.37% for the devices prepared by VASP.

We found the VASP method to be versatile with respect to variations of precursor components and perovskite composition. As a demonstration, we tested it with the emerging perovskite material FAxCs1–xPbIyBr3–y (2426). We obtained an excellent PCE of close to 18.0% in the initial trial; detailed PV metrics are listed in table S6.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S8

Tables S1 to S6

References and Notes

  1. Acknowledgments: Supported by King Abdulaziz City for Science and Technology (M.G. and S.M.Z.), an EPFL Fellowship co-funded by a Marie Curie stipend from the European Union’s Seventh Framework Programme (no. 291771) (J.L.), the Swiss National Science Foundation, NRP 70 “Energy Turnaround,” SNF-NanoTera (SYNERGY), and the Swiss Federal Office of Energy. We thank F. Zhang for experimental aid with the preparation of some perovskite solutions and TiO2 films, I. Dar and G. J. Jacopin for assistance with the photoluminescence experiments, and W. Tress for help with the curve fitting. X.L., D.B., C.Y., S.M.Z., A.H., and M.G. have applied for a European patent for the VASP process.
View Abstract

Stay Connected to Science

Navigate This Article