20.8% slot-die coated MAPbI3 perovskite solar cells by optimal DMSO-content and age of 2-ME based precursor inks


 Solar cells incorporating metal-halide perovskite (MHP) semiconductors are continuing to break efficiency records for solution-processed solar cell devices. Scaling MHP-based devices to larger area prototypes is a crucial step towards commercialization. This requires the development and optimization of scalable process technology for these devices. Here, we demonstrate a maximum power conversion efficiency (PCE) of 20.83% for slot-die coated gas-quenched small-area methylammonium lead iodide (MAPbI3) devices. Our ink is based on 2-methoxy-ethanol (2-ME) with the strongly coordinating solvent dimethyl-sulfoxide (DMSO) added in small amounts. We found that the amount of DMSO, as well as the age of the precursor solutions, are determining factors in achieving highly efficient and reproducible devices. Through in-depth insight into the film formation process as a function of DMSO content from in-situ X-ray diffraction experiments, we found that just the right amount of DMSO favorably affects thin film growth. Adding 11.77 mol% of DMSO prevents the formation of a crystalline intermediate phase related to MAPbI3 and 2-ME (MAPbI3 -2-ME), reported here for the first time, and inducing the formation of some (DMSO)2MA2Pb3I8 intermediate phase. These results demonstrate that ink composition and process control are critical to enable reproducible large-scale manufacturing of MHP-based devices for commercial applications.

A variety of scalable fabrication techniques such as blade coating9, vapor-assisted deposition10, inkjet printing11 and slot-die coating (SDC)12 have been successfully utilized for the deposition of MHP materials. Notably, SDC is considered as one of the promising deposition techniques in the fabrication of functional coating and solution-processed optoelectronics electronics elds due to its advantages of making uniform and compact thin lms on both rigid and soft substrates13. When coating crystalline (or crystallizing) materials such as MHP, the quality of the thin lm is determined by the coating process. Fabricating high-quality thin lms requires exact control over the process window and removal of solvents at a de ned period of time to induce oversaturation and crystallization in a controlled manner.
Solvent removal can be facilitated by thermal annealing14, vacuum12, gas quenching15 or the deposition of an anti-solvent, which for larger area samples is often accomplished by dipping into a bath. 16 Currently, the PCE obtained for slot-die coated devices is lower than record device PCEs achieved by spin-coating17. This is due to the fact that there has been considerably less optimization work on SDC for MHP deposition and record-performance spin-coated devices rely on rapid solvent removal and induction of crystallization by deposition of an anti-solvent. One of the most important aspects to optimize slot-die coating for MHP deposition is to develop reproducible coating processes where the MHP crystallization is selectively induced during the coating process. Process optimization and design of stable precursor inks that enable the deposition of high quality MHP layers that yield higher PCE in solar cell devices.
In this work, we successfully slot-die coated methylammonium lead iodide (MAPbI3) thin lms from a low boiling point and highly soluble 2-methoxy-ethanol (2-ME) ink as previously introduced by Hendriks et al. in spin-coating. 18 Deng et al. adopted this ink to realize blade-coated PSCs9 introducing the strongly coordinating solvent dimethyl sulfoxide (DMSO) to improve the quality of coated perovskite thin lms.
We here demonstrate, that record-performance slot-die coated PSCs can be manufactured following this procedure and provide a detailed investigation on how the exact amount of DMSO and age of the precursor ink affects thin-lm morphology as well as device performance. Record performance device were achieved for 11.77 mol% of DMSO in the precursor solution. Using in-situ grazing incidence X-ray scattering experiments, we can prove that adding a small amount of DMSO leads to the suppression of a crystalline intermediate phase related to MAPbI3 and 2-ME (here referred to as MAPbI3-2-ME) in favor of the formation of MAPbI3 seed crystals and a limited amount of the (DMSO)2MA2Pb3I8 intermediate phase. A larger amount of DMSO and with proceeding age of the precursor solution, the precipitation of the crystalline (DMSO)2MA2Pb3I8 solvate phase19 becomes detrimental to the perovskite thin lm morphology. We highlight that both the exact amount of DMSO as well as the age of the precursor solution critically affect the coating results and device performance. This is critical insight to enable the development of stable and reproducible coating procedures for large scale manufacturing of perovskite solar cell devices. At an optimal content of about 12 mol% DMSO in the 2-ME ink, we demonstrated champion MHP device with e ciency of 20.83%, which is among the highest value reported in slot-die coated PSCs.

Results And Discussion
As MHP tends to crystallize during deposition, the removal of solvents at a de ned point in time determines the thin lm quality. This provides some but limited control over the process window and the exact composition of the ink has a signi cant in uence on the quality of the MHP semiconductor. In the process discussed here, this was achieved by N2 gas quenching through an air knife at around 20 cm distance relative to the slot-die as illustrated in Figure 1 Figure 1.a. Therein, MHP lms were discontinuous and porous coated from 2-ME ink. We attribute this to the rapid MHP crystallization from inks without DMSO additives when 2-ME spontaneously evaporates during coating and air knife gas quenching.
Addition of small amounts of the strongly coordinating solvent DMSO led to dramatically changes of the ink properties and coating process. In Figure 1 b, the morphology of samples coated from inks with various amounts of DMSO, from 11.77 mol% to 58.88 mol%, are compared. Samples coated from a 2-ME ink, exhibit a disrupted morphology with pinholes density around 7% (Supplementary Figure 1). The addition of 11.77 mol% DMSO into 1M 2-ME ink leads to a denser thin-lm without pinholes while with higher DMSO contents the crystallite size becomes smaller as shown in Figure 1 c. Moreover, inks with added DMSO exhibited a yellow precipitate increasing with time and DMSO content when stored at room temperature as shown in Figure 1 d. The 2-ME precursor ink without DMSO and the 2-ME based ink comprising 11.77 mol% of added DMSO do not exhibit this precipitate. This suggests a chemical interaction between the DMSO and perovskite precursors. As previously reported9, without the addition of MAI, PbI2 has a limited solubility in 2-ME. As a strong coordinating solvent, DMSO can be expected to increase the solubility of the precursor salts in 2-ME. However, DMSO is known to interact strongly with MAPbI3 in the solid-state phase forming the crystalline intermediate phase ((DMSO)2MA2Pb3I8)20. As this DMSO intermediate phase has limited solubility in 2-ME, a higher content of DMSO in 2-ME solution leads to precipitation. The precipitate was identi ed by XRD as (DMSO)2MA2Pb3I8.20 After annealing the precipitate, the color converted to black and intensity of the diffraction peaks associated with the perovskite phase appeared (Supplementary Figure 2).
In-situ grazing-incidence wide-angle X-ray scattering (GIWAXS) was utilized to investigate the drying kinetics as a function of DMSO content with respect to the 2-ME ink ( Figure 2). The sample deposited from 2-ME ink initially (start to 0.5 minutes, see Supplementary Figure  the diffraction peak at 14.08° that correlates with the (110/002) peaks of the room temperature tetragonal MAPbI3 phase23. It can be noted that the MHP peak intensity is relatively high, indicating rapid crystallization of the MHP as 2-ME solvent starts evaporating.
When adding 11.77 mol% of DMSO into 1 M 2-ME ink as shown in Figure 2b, d, a diffraction peak at 5.92°s uggest the initial co-existence of the intermediate phases of MAPbI3-2-ME and the DMSO solvate phase with known diffraction peaks at 6.55°, 7.2°, 9.17° attributed to (DMSO)2MA2Pb3I8 during the rst 0.5 minute of the process20. There is no peak at 8.15° visible in this case suggesting that the formation of this second intermediate phase is suppressed. Instead, a diffraction pattern consistent with the formation of crystalline MAPbI3 is visible from the start as shown in Figure 2 c, d. We interpret these ndings that the DMSO additive avoids the formation of the 2-ME solvate phase and triggers the formation of MAPbI3 as well as DMSO solvate phase. These results indicate the formation of MAPbI3 seed crystals upon addition of DMSO, which favorably affects thin lm formation. We hypothesize, that the much higher binding constant of DMSO to lead-halide complexes in solution disrupts the formation of the 2-ME and other solvate intermediate phases to promote the formation of crystalline MAPbI3. Simultaneously, (DMSO)2MA2Pb3I8 is also formed, which, at small concentrations, may prove bene cial to thin lm formation as it may increase the processing window. That the formation of the (DMSO)2MA2Pb3I8 intermediate phase is detrimental at higher concentration becomes apparent when increasing to the amount of added DMSO to over 11.77 mol% (Figure 2d, Supplementary Figure 4), which leads to a disappearance of any MAPbI3-2-ME intermediate but also the MAPbI3 phase in favor of (DMSO)2MA2Pb3I8. The formation of the DMSO solvate phase occurs already in solution apparent by the formation of the yellowish precipitate in precursor solutions shown in Figure 1.c. In this case, MAPbI3 has to and increasing extent be formed by decomposition of the DMSO intermediate phase. This negatively affects the thin-lm morphology as shown in the SEM cross section images in Figure 1.b. To compare ink properties, ex-situ GIWAXS was employed to detect crystallinity and orientation of MHP semiconductors. We did not observe a change in orientation, indicating the crystal growth orientation wasn't affected by the addition of DMSO (Supplementary Figure 5). However, the 2ME-DMSO samples exhibit an increased intensity of the diffraction peaks associated with the perovskite crystal phase. We conclude that the addition of DMSO improves the crystallinity of the thin-lm sample (Supplementary Figure 6).
To study the stability of precursor inks. A series of devices were prepared from precursor inks with various amounts of DMSO (11.77 mol% -58.88 mol%) and time elapsed since preparation of the ink 1, 5, 10, 30 and 60 min in comparison to the reference ink. Figure 3.a show the device con guration is glass | ITO (150 nm) | 2PACz | MAPbI3 (~ 800 nm) | C60 (23 nm) | BCP (8 nm) | Cu (100 nm) where 2PACz is a selfassembled monolayer (SAM) of (([2-(9H-carbazol-9-yl)ethyl]phosphonic acid)). The results shown in Figure 3b clearly illustrate the dependency of the PCE obtained in solar cells on the DMSO content and age of the precursor ink. The devices made from the 2-ME reference ink exhibit fairly low and irreproducible performance metrics. For an increasing content of DMSO as an additive in 2-ME precursor inks, the average performance and reproducibility initially increases for intermediate DMSO concentrations of about 12 mol% and 24 mol% but then clearly decreases for higher DMSO content and also with increasing ink age. This is consistent with the observed increase in the formation of a precipitate shown in Figure 1c and UV-Vis measurements on precursor inks as shown in Supplementary   Figure 7. The formation of this precipitate inside the narrow tubing and ink reservoir of the slot-die coater will dramatically in uence the pump rate and solution feed during the coating process. This leads to random and unrepeatable deposition conditions along with the likely deposition of the (DMSO)2(MA)2Pb3I8 precipitate negatively affecting thin lm morphology. To investigate the in uence of precipitation on the lm nucleation, 23.56 mol% 2ME-DMSO ink was centrifuged at 6,000 rpm for 45 min with this ink without centrifuged were compared, which resulted in similar devices performance (Supplementary Figure 8) and indicates that the precipitate affect the coating process negatively. Optimal contents of DMSO is crucial to stabilize MHP inks. The 11.77 mol% DMSO ink exhibits better long-term stability than the others investigated herein. Even after storing this ink for 18 days, solar cells with a performance of 19.50% were manufactured as shown in Supplementary Figure 9. After storage for 18 days, the 23.56 mol% DMSO content ink exhibited precipitation and could not be used to produce good quality coatings at all. Figure 3.c show the solar cells prepared from the 11.77 mol% and 23.56 mol% 2-ME-DMSO precursor solutions yielded a maximum PCE of 20.83%, 20.25% compared to 15.42% for the pure 2-ME reference devices. These PCEs are among the highest values reported for slot-die coated PSCs (Supplementary Figure Table 1 Figure 10. We compared the device performance with and without shadow mask and found around 3% discrepancy in the JSC (Supplementary Figure  to short delay times. The device fabricated from the pure 2-ME precursor ink exhibits a more pronounced amplitude of the transient current response for fast J-V scan rates compared to the devices fabricated from inks containing small amounts of DMSO where this transient current response is fairly suppressed. This directly corresponds to a reduction of the hysteresis for fast scan-rates that suggest capacitive effects or sluggish charge carrier extraction for devices with inferior perovskite thin lm morphology. We conclude that the smaller grain size and higher degree of disorder and pinholes of samples prepared without the DMSO additive in the precursor solutions gives rise to more sluggish charge carrier extraction and hence larger amplitudes in the transient current response upon voltage perturbation.

Conclusions
In conclusion, we were able to achieve highly PCEs of slot-die coated perovskite solar cells reproducibly when just the right amount of DMSO is added to 2-ME based precursor solutions. For an amount of about 12 mol% DMSO, a stable ink with a bene cial coating process window can be obtained enabling the deposition of high-quality MHP thin-lms and high-performance solar cell devices. were purchased from Tokyo Chemical Industry (TCI). Methylammonium iodide (MAI), was purchased from Dynamo. Bathocuproine (BCP) was purchased from Ossila. Ethanol (anhydrous) was purchased from VWR Chemicals and Copper Shots was purchased from Alfa Aesar. Other chemicals were purchased from Sigma-Aldrich. All the chemicals were used as received without puri cation.
SAM powders were dissolved in anhydrous Ethanol at a concentration of 0.5mg/ml and put into an ultrasonic bath for 15 min (~ 45 °C) before using 24 .
Slot die coating MHP semiconductors and devices fabrication.

SAM preparation
Before spin-coating SAM, the ITO glasses were pre-clean with 15 minutes UV-ozone treatment. Then SAM layer was spin coated at the speed of 3,000 r.p.m for 30 s, (ramping: 1,000/s) and then annealed at 100°C for 10 min.

SDC MHP semiconductors and devices making
The slot die coater (FOM technologies) which is composed of an ink reservoir, pump systems, slot-die head, and a temperature-controlled bottom plate (vacuum chuck) mounted on a conveyor belt system. An air knife for N 2 gas quenching can be optionally mounted in the system. In brief, when coating begins, the pump systems precisely control how much ink (pump rate) ows through a tube into the slot die head to substrates, and the conveyor system determines the coating speed and direction.
MHP semiconductors inks were pipetted into a syringe connecting to a tube, which is controlled by pump (pump rate x how much ink is pouring down out of the tube into die head per second). Coating speed is 20 cm/min, pump rate is 50 µl/min, gap is around 0.2-0.3 mm, air knife pressure is approximately 20 psi at room temperature. The MHP semiconductors were annealed at 100 °C for 20 minutes.
After cooling down to room temperature, the substrates were transferred to the evaporation system (MBRAUN ProVap 3G). The BCP (8 nm), C 60 (23 nm) and copper (100 nm) layers were subsequently deposition by thermal evaporate method. The evaporate rate of BCP is 0.1 Å/s, C 60 is 0.05 Å/s to 0.1 Å/s, and the rate of copper is 0.1 Å/s to 1 Å/s at base pressure of 1 × 10 -6 mbar. For the champion devices, a 100 nm NaF (Sodium Fluoride) layer was deposited upon the glass side at base pressure of 1 × 10 -6 mbar, with rate of 0.

EQE
The external quantum e ciency (EQE) measuremnts (from 300 to 850 nm) were performed using Oriel Instruments QEPVSI-b system with a Xenon arc lamp (Newport 300 W, 66902) chopped at 35.5 Hz and a monochromatic instrument (Newport Cornerstone 260) 26 .

SEM
The scanning electron measurements (SEM) images (top view and cross-section) were measured by Hitachi S-4100 at 5 kV.

GIWAXS
In-situ grazing incidence wide-angle X-ray scattering (GIWAXS) measurements were conducted at KMC-2 beamline 21 at the synchrotron source BESSY II (Helmholtz-Zentrum Berlin). The photon energy of the source is 8048 eV. Incident angle was xed at 2 deg. We use the area detector Bruker AXS. The detector was kept on 9 degrees to collect the scattering. An Anton Paar setup was used to control the substrate temperature and the temperature ramp is around 45 °C/min During measurement, the samples were kept under constant nitrogen gas ow ( Declarations