Crystallisation control of drop-cast quasi-2D/3D perovskite layers for efficient solar cells

Introducing layered quasi-2D perovskite phases into a conventional 3D perovskite light-absorbing matrix is a promising strategy for overcoming the limited environmental stability of 3D perovskite solar cells. Here, we present a simple drop-casting method for preparing hybrid perovskite films comprising both quasi-2D and quasi-3D phases, formed using phenylethylammonium or iso-butylammonium as spacer cations. The film morphology, phase purity, and crystal orientation of the hybrid quasi-2D/3D perovskite films are improved significantly by applying a simple N2 blow-drying step, together with inclusion of methylammonium chloride as an additive. An enhanced power conversion efficiency of 16.0% is achieved using an iso-butylammonium-based quasi-2D/3D perovskite layer which, to our knowledge, is the highest recorded to date for a quasi-2D/3D perovskite solar cells containing a non-spin-cast perovskite layer prepared under ambient laboratory conditions. Perovskite solar cells have substantial potential for solar conversion, but developing simple and scalable fabrication processes is challenging. Here, a drop-casting process compatible with roll-to-roll production of quasi-2D/3D perovskite layers is developed, with a conversion efficiency of up to 16%.

A fter ten years of rapid development, the record power conversion efficiency (PCE) for perovskite solar cells (PSCs) now exceeds 25%, which is approaching the Shockley-Queisser limit and is competitive with the best singlejunction silicon solar cells [1][2][3][4][5] . Three-dimensional (3D) perovskite films demonstrate excellent optoelectronic properties, including strong light absorbance, low exciton binding energy and long charge-diffusion lengths, and they also allow for ease of deposition via solution-based processing methods. However, while the durability of 3D PSCs has been improved tremendously 6,7 , they remain prone to device performance degradation due to the presence of organic cations that are vulnerable to atmospheric moisture 8 . This has resulted in delays in the upscaling and commercialisation of this technology, requiring researchers to investigate related material classes that retain the same optoelectronic properties while also imparting greater stability to environmental exposure 9,10 .
Quasi-two-dimensional (2D) Ruddlesden-Popper perovskites, R 2 A n−1 M n X 3n+1 (n ≥ 1), where R and A represent larger and smaller organic cations, respectively, M refers to transition metal cation (e.g. Pb + or Sn + ) and X is a halide anion, are attracting increased attention for use in solar cells due to their higher operational stability compared with 3D perovskites [11][12][13][14][15] . The incorporation of bulkier organic cations with longer hydrocarbon chains between perovskite layers confers the material with hydrophobic properties that hinders moisture ingress, which ultimately improves environmental stability 16 . However, the use of 2D perovskite materials presents additional challenges in fabricating high-efficiency solar cells, as the crystal orientation in the photo-absorber film significantly affects the device performance. To overcome this, Tsai et al. developed a hot-casting spin-coating method for the preparation of films comprising highly oriented quasi-2D perovskites (n-BA 2 MA n−1 Pb n I 3n+1 ), resulting in an improvement in PCE to 12.52% 13 . Then quasi-2D perovskites with other cations such as phenylethylammonium (PEA) and isobutylammonium (iso-BA) were reported using spin coating with PCEs achieved around 11% 17,18 . Subsequently, a series of alternative methods have been developed to further improve device performance, such as using different spacer cations (fluoro-phenylethylammonium (fluoro-PEA + ), 1,4-butanediammonium (BEA 2+ )), optimising the n value of the perovskite composition, and employing slow post-annealing conditions. These methods have significantly improved the quality of films containing the quasi-2D perovskites, and have pushed the PCEs of PSCs comprising quasi-2D perovskites beyond 17% [19][20][21][22] .
Nevertheless, almost all PSCs comprising quasi-2D perovskites reported to date have been prepared by spin-coating in a glovebox, which is not suitable for upscaling. Recently, we reported a simple, low-waste, scalable method to prepare films based on quasi-2D perovskites in which a precursor solution was dropcasted on a heated substrate, and then spontaneously spreads to form a smooth, uniform film upon drying 23 . In that work, PSCs containing drop-cast n-butylammonium-based quasi-2D perovskites were the primary focus, with a PCE of 14.9% achieved for solar cells fabricated in an ambient environment. This result suggests that the drop-cast method is a promising route for producing high-efficiency solar cells in a scalable manner, and presented opportunities to increase PCE by using alternative organic cations and optimising the crystallisation process.
Herein we report a facile method to improve the performance of PSCs comprising quasi-2D perovskites using known cations, such as PEA and iso-BA. The application of a stream of air or nitrogen is used to dry the drop-cast films, with the incorporation of additional methylammonium chloride (MACl) additive further improving film quality. The same method is also applied for the preparation of 3D perovskites, and the difference in film formation using a quasi-2D perovskite precursor solution and that using a 3D perovskite CH 3 NH 3 PbI 3 (MAPbI 3 ) precursor solution is examined. The effect of the drying conditions and the presence of MACl additive on film morphology, phase purity and crystal orientation using the quasi-2D perovskite precursor solutions based on PEA or iso-BA spacer cations is investigated, and the photovoltaic performance of devices incorporating the drop-cast films is evaluated. The highest PCE of 16% was achieved for PSCs comprising N 2 blow-dried iso-BA-based quasi-2D perovskite precursor solution.

Results and discussion
Film preparation methods. The preparation of films from (PEA) 2 (MA) 4 Pb 5 I 16 (PEA-2D) and (iso-BA) 2 (MA) 4 Pb 5 I 16 (iso-BA-2D) precursor solutions is illustrated in Fig. 1. The perovskite precursor solution is drop-cast onto a pre-heated substrate (Fig. 1a), with the solution spreading spontaneously to form a circular wet film. Two methods are employed to dry the wet films: drying under ambient laboratory conditions, hereafter referred to as 'naturally dried' (Fig. 1b, c), and drying under a stream of nitrogen or air to accelerate the evaporation of the solvent, hereafter referred to 'N 2 blow-dried' or 'air blow-dried', respectively, (Fig. 1d, e). These different drying processes lead to significantly different film qualities, as discussed in detail below.
Film morphology. Visual inspection of the drop-cast PEA-2D and iso-BA-2D films revealed significant differences in the appearance caused by the various drying techniques (Fig. 2). The N 2 blow-dried films (Fig. 2c, d) look smoother and more uniform than their naturally dried counterparts, which appear grey and have several fringes on the surface (Fig. 2a, b). The morphology of the films was studied in greater detail using scanning electron microscopy (SEM) (Fig. 2e-f), with the naturally dried PEA-2D film (Fig. 2e) appearing rough and comprising some cylinder-like grains. The naturally dried iso-BA-2D film (Fig. 2f) also comprises cylinder-like grains, but its surface is considerably smoother. Both of these thin films exhibited severe microstructural defects with large pinholes leading to incomplete coverage of the substrate. In contrast, the PEA-2D and iso-BA-2D films prepared using N 2 blow-drying present uniform, dense, and smooth surfaces (Fig. 2g, h). It should be noted that, as reported previously for spin-coated quasi-2D films [19][20][21] , the N 2 blow-dried quasi-2D perovskite films do not display the granular features seen in 3D perovskite films. In contrast to cracks and pinholes often observed in the spin-cast quasi-2D perovskite films, there is little evidence of cracks and pinholes in the N 2 blow-dried iso-BA-2D perovskite films were produced in the present work (see Supplementary Fig. 1). The flow rate used during N 2 blow-drying influences film quality, with optimisation of the N 2 flow rate performed on iso-BA-2D films indicating that flow rates of at least 5 L min −1 are required to achieve uniform films (see Supplementary Fig. 2).
The difference in film morphology can be understood by considering the process of crystal growth [24][25][26][27] ; the slower solvent evaporation during natural drying results in lower supersaturation and a low rate of nucleation, resulting in a low crystal nuclei concentration that causes the growth of fewer, larger grains. In contrast, N 2 blow-drying causes rapid evaporation of the solvent, leading to high supersaturation and a high rate of nucleation. This ultimately causes the formation of highly uniform, compact films without defects (Fig. 1).
To compare the crystallisation process of the drop-cast PEA-2D and iso-BA-2D films with a conventional 3D perovskite system, MAPbI 3 films were prepared using the same drop-casting methods (see Supplementary Fig. 3). Naturally dried MAPbI 3 films were rough, as found for the PEA-2D and iso-BA-2D films, and comprised needle-like grains, whereas N 2 blow-drying yielded smoother films comprising more compact, smaller needle-like grains. However, the nucleation rate in the MAPbI 3 wet disc is insufficient to give a pinhole-free film, and much higher N 2 flow rates (20 L min −1 ) were required to form smooth and uniform regions at the centre of the film (see Supplementary  Fig. 4a), with the middle and edge of the films remaining nonuniform (see Supplementary Fig. 4b-d). The different properties of films formed using drop-casting quasi-2D and 3D perovskite precursor solutions suggest that cation size has a considerable impact on the nucleation of the perovskite crystals. In this regard, high-quality films containing quasi-2D perovskite can be prepared using milder N 2 blow-drying. MACl is a popular additive used to improve the film quality of 3D perovskites [28][29][30] . However, addition of MACl has little effect on quasi-2D film morphology (see Supplementary Fig. 5) but has a profound effect on other film properties, as will be discussed below.
Phase distribution. UV-visible absorption spectroscopy was used to measure the optical properties of the films (see Supplementary  Fig. 6). The high baseline in the spectra of the naturally dried films is due to light scattering resulting from the high degree of surface roughness of these films, but a number of excitonic absorption bands can be discerned, including the n = 1 band in the naturally dried PEA-2D film.  display similar spectra with much less scattering and with distinct excitonic absorption peaks at 565, 608 and 643 nm, which are characteristic of n = 2, n = 3 and n = 4 phases, respectively 31 .
The absorption onset at 763 nm is attributed to the co-formation of quasi-3D perovskite phases (n → ∞), highlighting the fact that the films produced from the quasi-2D perovskite precursor solutions in this work actually comprise a mixture of quasi-2D phases having a range of n values, together with phases having sufficiently large n values for the optical properties to approximate that of a conventional 3D phase, which will be referred to hereafter as quasi-2D/3D films. The N 2 blow-dried films prepared with MACl additive show a slightly higher absorbance, which may be due to improved film quality. Similar results have been reported for spin-coated perovskite films 32,33 . The photoluminescence (PL) emission spectra of the quasi-2D/ 3D perovskite films are shown in Fig. 3. The naturally dried PEA-2D film displays PL maxima at 522, 574 and 619 nm, corresponding to the n = 1, n = 2 and n = 3 phases, respectively (Fig. 3a). The broad band extending between 650 and 750 nm is attributed to emission from multiple quasi-2D having n > 4 and quasi-3D phases. For the N 2 blow-dried PEA-2D film (Fig. 3b), the contribution from the quasi-2D phases decreases, with a corresponding increase in the PL from quasi-3D perovskite phases. For the PEA-2D film prepared using MACl additive (Fig. 3c), no n = 1 peak is observed, and the PL from the n = 2 and n = 3 phases are relatively stronger than the other quasi-2D phases. For the iso-BA-2D films, the PL spectra have similar peaks for n = 2 and n = 3 phases, and no signal for the n = 1 phase can be observed ( Fig. 3d-f). The broad band comprising overlapping peaks of the large n phases is lower intensity for the N 2 blow-dried film (Fig. 3e) than the naturally dried film (Fig. 3d). For the film prepared with MACl additive, the quasi-3D perovskite PL peaks become much stronger with an absence of any obvious peaks associated with quasi-2D n > 4 phases (Fig. 3f), suggesting that the phases in the film become purer, being mainly composed of the quasi-2D n = 2, n = 3, n = 4 phases and quasi-3D phases.
Comparison of the PL spectra excited from the back (glass) face and front (perovskite) face provides information about the vertical distribution of perovskite phases in the film. All films display markedly different PL spectra depending on excitation from the back or front face. The PL signal from small n (n ≤ 4) quasi-2D phases is much more pronounced for back-face excitation than front-face excitation, suggesting that there is a higher concentration of small n quasi-2D phases forms at the bottom of the perovskite films, which is similarly observed in spin-coated films comprising quasi-2D perovskites 19 . Obvious PL quenching is observed for the films prepared on PEDOT:PSS, resulting from efficient hole transfer from the perovskite to PEDOT:PSS. The films prepared on glass and PEDOT:PSS show different peak positions for the quasi-3D phase, especially for the front excitation. The blueshift of PL peaks of quasi-3D phase suggests that the PL quenching of large n phases may be more efficient than for small n phases. There is no clear trend in the extent of PL quenching of the quasi-3D phase compared with the quasi-2D phases, with the exception of the front-face PL quenching of the quasi-3D phase in the iso-BA-2D film in which the PL quenching is improved significantly by using an MACl additive, suggesting improved hole transport through this film (Fig. 3f).
The effect of the MACl additive in the iso-BA-2D film was further investigated using time-resolved PL (TRPL) and transient absorption (TA) spectroscopy (Fig. 4). The iso-BA-2D perovskite film prepared on glass using the MACl additive shows an increased lifetime (2.6 ns) compared to the film without additive (0.9 ns), suggesting suppressed non-radiative charge-carrier recombination induced by trap-state passivation 34 . The TA spectra at selected pump-probe delay and the 2D contour plots ( Fig. 4b and see Supplementary Fig. 7, respectively) show prominent ground-state bleaching (GB) peaks corresponding to the n = 2, 3, 4, 5 quasi-2D perovskite phases, and a broad GB band at wavelengths longer than~700 nm associated with bleaching of the quasi-3D phase, in good agreement with the band positions in the UV-visible absorption and PL spectra. The presence of the MACl additive appears to make little difference to the TA properties of the films. The evolution of the onset of the GB bands with increasing n values seen in Supplementary Figs. 7 and 8 is consistent with energy cascading from the n = 2 to n = 5 quasi-2D phases before accumulating in the lowest energy quasi-3D phase, with the transfer of energy being essentially complete after about 100 ps, in good agreement with TA results reported by others for spin-cast films prepared from other quasi-2D (PEA) 2 (MA) 2 Pb 3 I 10 and (NMA) 2 (FA)Pb 2 I 7 precursor solutions (NMA = 1-naphthyl-methylammonium) 35,36 . The residual PL detected from each phase indicates the efficiency of energy transfer between phases is <100%. The progressive delay in the GB onset suggests that most of the 510-nm pump pulse incident at the back face of the film is absorbed primarily by the n = 2 phase, with little/no absorption by higher n phases. This implies there may be a degree of stratification throughout the perovskite layer, i.e. with n = 2 primarily deposited at the base followed by layers of larger-n phases, with the quasi-3D at the top.
Crystal orientation. The X-ray diffraction (XRD) patterns for the quasi-2D perovskite films are shown in Supplementary Fig. 9. The naturally dried PEA-2D film displays a weak peak at 5.3°, which is assigned to the 002 reflection of the n = 1 phase (PEA 2 PbI 4 ) 37 . This peak is not observed in the N 2 blow-dried film or the films prepared using MACl additive. All the films show many weak peaks which may be attributed to random crystal orientations. The iso-BA-2D films show nearly the same XRD patterns as the corresponding PEA-2D films but with only two main peaks (see Supplementary Fig. 9b), strongly suggesting that the crystals are highly oriented. Synchrotron-based grazing incidence wide-angle X-ray scattering (GIWAXS) was performed to further investigate the crystal orientation in the film (Fig. 5). All the PEA-2D films exhibit scattering rings or arcs (rather than discrete spots), suggesting predominantly random orientation of the crystals in the film. Since the charge carriers in the lead halide perovskite are unable to transport through the insulating large cations, a crystal random orientation complicates the transportation of charge carriers through the film. In addition to peaks indexable to 3D phase, multiple scattering rings at low q between 0.2 to 0.6 Å −1 are seen which correspond to various low n members of layered perovskite phases (see Supplementary Fig. 10). In particular, n = 1 phase is clearly observed in the naturally dried PEA-2D film 38 . The N 2 blow-dried PEA-2D film (without MACl) exhibits extra scattering features correspond to layered perovskite phase of n = 1 and 3, and additional intermediate or metastable phase 39 . Interestingly, these scattering peaks are greatly suppressed in the blow-dried PEA-2D film prepared with MACl, leaving relatively weak peaks correspond to n = 1 phase.
In contrast to the relatively random orientation of the PEA-2D films, all iso-BA-2D films show concentrated Bragg spots indicating highly textured films (Fig. 5d-f). To determine the crystal orientation, the experimental GIWAXS patterns were compared with the simulated patterns assuming two distinct crystallographic orientations (see Supplementary Fig. 11). We confirm the vertical orientation of the bulkier organic cation layers to the substrate from the excellent agreement of the strong as well as the weak Bragg reflections (see Supplementary Fig. 11). The missing in-plane scatterings correspond to the 0k0 reflections is possibly due to the presence of other low n phase disrupting the structural ordering of n = 4 phase, or being hidden by the substrate horizon due to the highly textured nature of the film. Compared to naturally dried iso-BA-2D film, the N 2 -dried films show Bragg spots with much lower azimuthal spreading of diffracted intensity, indicating improved crystal orientation. Further analysis on the pole figure along q = 1.7 Å −1 in Supplementary Fig. 12 confirms an improvement in the overall texture of the N 2 -dried film, with the fraction of crystallites exhibiting a high degree of texture increasing from 54% to 85% upon N 2 drying. Comparing the GIWAXS patterns of the iso-BA-2D films to the simulated patterns, additional peaks are seen as indicated by arrows in Fig. 5d-f that are not indexable to a perovskite or Ruddlesden-Popper phase. Comparing 1D sector plots in Supplementary Fig. 13 reveals that the additional phase is more prevalent in the blow-dried film as compared to the naturally dried film. Furthermore, the use of the MACl additive also appears to suppress the abundance of this phase. The additional phase is proposed to be an unresolved intermediate phase with the amount of residual intermediate phase present related to the crystallisation kinetics. N 2 drying accelerates the solvent removal process as compared to natural drying, resulting in a faster crystallisation kinetics and more residual intermediate phase. The addition of MACl had been reported to slow the perovskite crystallisation process, thereby resulting in less residual intermediate phase [28][29][30] . The near-perfect vertical orientation (see Supplementary Fig. 14) of the 2D perovskite crystals in the BA-2D films with respect to the substrate will be beneficial for chargecarrier transport 40 and may help to explain the improved PL quenching in the N 2 -dried iso-BA-2D film with MACl additive.
Photovoltaic performance. The performance of drop-cast PEA-2D and iso-BA-2D perovskite layers in solar cells was investigated using devices comprising ITO/modified-PEDOT:PSS (m-PEDOT:PSS)/quasi-2D perovskite/PC 61 BM/PEIE/Ag (Fig. 6a),  41,42 . N 2 flow rate was optimised using iso-BA-2D perovskite (see Supplementary Table 1). The best performance was obtained using a flow rate of 10 L min −1 , with little effect on device performance on using higher flow rates. The cells fabricated using naturally dried and N 2 blow-dried PEA-2D perovskite films show very low short-circuit current (J sc ) and PCE (Fig. 6b, Table 1). This may be due to the n = 1 phase at the base of the photoabsorber film blocking hole transport to the m-PEDOT:PSS layer. A demonstrable improvement in J sc and PCE was found for cells fabricated using N 2 blow-dried PEA-2D perovskite films with MACl additive, which is attributed to improved charge transport associated with the elimination of the n = 1 phase. Cells incorporating naturally dried iso-BA-2D perovskite films also showed very low PCE (2.2%), which is attributed to the discontinuous coverage of the perovskite layer allowing recombination between the hole transport layer (m-PEDOT:PSS) and the electron transport layer (PC 61 BM). The device comprising a compact N 2 blow-dried film generated higher J sc (16.3 mA cm −2 ), FF (71.9%) and PCE (11.9%), and the addition of MACl improved the PCE to 16.0%, with a higher V oc of 1.15 V, a J sc of 18.1 mA cm −2 and an FF of 77.2%. The PCE is much higher than the previous achieved PCE (10.6%) of iso-BA-2D PSCs prepared by spin coating 17 . This is ascribed to improved phase purity and crystal orientation, which facilitates charge transport in the perovskite layer. Furthermore, a low level of hysteresis is observed in the current density-voltage (J-V) curves obtained in forward and reverse scans (Fig. 6c). The steady-state PCE of the 'champion' cell recorded using maximum power point (MPP) tracking (Fig. 6d) was 15.9%, consistent with the J-V measurement. The PCE of an encapsulated cell retains 91% of the initial value after 500 h of continuous 1-sun illumination, indicating good device durability (see Supplementary Fig. 15).
The performance of the cells using iso-BA-2D perovskite prepared by air blow-drying was also investigated. As discussed above, the formation of high-quality iso-BA-2D perovskite films does not require a high flow rate, and effective drying can be achieved with a conventional hair dryer that does not produce heat. The best cell incorporating an air-dried iso-BA-2D perovskite layer gave a V oc of 1.14 V, a J sc of 18.3 mA cm −2 , an FF of 75.5% and a PCE of 15.8%, which is very close to the values achieved using N 2 blow-drying (Fig. 6e, Table 1).
The performance reproducibility of the cells using N 2 blowdrying or air blow-drying (relative humidity 35%) was also investigated (Fig. 6f). Most of these cells displayed PCEs in the 14.5-15.5% and 14-15% ranges, respectively, demonstrating a high degree of reproducibility. The ease of using air-drying rather than an N 2 source not only significantly reduces the cost of the process but also addresses the operational safety concerns of using an asphyxiant, which is an important consideration in large-scale manufacturing.
A roll-to-roll (R2R) slot-die coating process was used to demonstrate the scalable production of the iso-BA-2D perovskite film (Fig. 7a). Some pinholes can be observed in the SEM image (see Supplementary Fig. 16), and the film was not as uniform as on a glass substrate, which is possibly due to the rougher and less hydrophilic surface of the ITO/PET substrate leading to dewetting on the substrate. Perovskite solar cells with the same device structure as the drop-cast method were produced using R2R slot-die process to deposit the perovskite layer; the 'champion' cell displayed a V oc of 1.04 V, a J sc of 13.6 mA cm −2 , an FF of 66.0% and a PCE of 9.3%, which is a 14% improvement from our previously reported PCE of 8% 23 . This demonstrates the potential for these materials to be incorporated into a facile, mass-production fabrication process, with higher PCEs anticipated from the use of substrates having a more hydrophilic and smoother surface, and lower sheet resistance. Cleaning the substrate using plasma treatment and protecting coatings from scratching may further improve the device performance.
In summary, we have demonstrated a simple drop-casting method to prepare quasi-2D/3D perovskite films with PEA and iso-BA cations. In comparison with the 3D MAPbI 3 perovskite, we found that the use of these large cations assists in the formation of uniform films. N 2 blow-drying and MACl additive were used to control the crystallisation process and improve the film quality. The N 2 blow-drying accelerates solvent evaporation, increasing the supersaturation and nucleation of solution, resulting in a compact and smooth film. The MACl additive can suppress the formation of the n = 1 phase of PEA-2D perovskite. For the iso-BA-2D perovskite, the MACl additive improves the phase purity and crystal orientation, leading to significantly improved device performance. GIWAX results suggest that the PEA-2D perovskite film is randomly orientated, while the iso-BA-2D film is highly vertically orientated, which is favourable for charge transport in high-performance devices. The staggered bleaching onset in TA measurements suggest that there may be a degree of stratification in the phase distribution in iso-BA-2D films, with n = 2 preferentially locating at the glass surface with n systematically increasing away from the glass surface, with quasi-3D phases concentrated at the top face of the film. PCEs of 9.2% and 16.0% were achieved using N 2 blowdried PEA-2D and iso-BA-2D films, respectively. Drying using a stream of humid air (relative humidity: 35%) was also investigated and a PCE of 15.8% was achieved. Translation from batch to a continuous R2R process was successfully demonstrated, achieving a maximum PCE of 9.3% for solar cells on a flexible substrate. This enhanced performance demonstrates the viability of using an R2R-compatible method to fabricate high-performance quasi-2D/3D perovskite layers under ambient laboratory conditions.  Materials characterisation. UV-visible absorption spectra were recorded using a Hewlett-Packard/Agilent 8453 UV-vis Spectrophotometer (300-1100 nm). PL spectra were recorded using a luminescence spectrometer (Edinburgh Instruments, FLSP920) with a CW Xe lamp (~0.1 mW cm −2 ) as the excitation source. TRPL measurements were performed using a TCSPC setup, which consists of 400 nm (1 MHz, 100 fs FWHM) pulsed-laser excitation with excitation pulse fluence of 40 μJ cm −2 and a commercial photon counting system (SPC150 card, B&H, Germany) as the detector. Femtosecond pump-probe TA measurements on films with 510 nm (FWHM < 20 nm) excitation were performed at room temperature using an in-house TA spectrometer system. A 60 fs laser pulse at 800 nm (96 kHz) was used to generate the 510 nm pump beam and white light probe beam separately. The pump and probe pulses were focused on the back (glass) face of film samples with an off-axis parabolic mirror, with the fluence of pump beam set to be 0.52 μJ cm −2 at the excitation point. The time-resolved TA spectra were recorded using the visible version of an Ultrafast Systems high-speed fibre-optic spectrometer. 'De-chirping' correction was performed using Surface Xplorer software by specifying four time zero-wavelength pairs in the chirp correction mode. XRD measurements were performed using a Rigaku SmartLab with Cu K α radiation. GIWAXS measurements were performed at the SAXS/WAXS beamline at the Australian Synchrotron 43 . The energy of X-ray incident beam was fixed at 15 keV. 2D scattering patterns were recorded by an in-vacuum Pilatus 2M detector placed~72 cm from the sample position, with a 0.40°incident angle and a total exposure time of 3 s. The sample-todetector distance was calibrated with a silver behenate sample. The scattering results were analysed using Igor Pro with an altered version of NIKA analysis package 44 . The simulation of GIWAXS patterns was done using GIXSGUI up to q = 2 Å −1 (ref. 45 ). Extremely weak Bragg reflections (theoretical intensity less than 1 based on the calculated powder XRD intensity using VESTA software 46  Perovskite solar cells made using R2R slot-die coating. The setup of the printer and the coating of the PEDOT:PSS layer can be found in our previous work 42 . A (iso-BA) 2 (MA) 4 Pb 5 I 16 precursor solution containing 0.7 M Pb 2+ was used. The speed of the substrate movement was 0.5 m min −1 and the solution feed speed was 30 μL min −1 . The coated film was passed over the first curved hot plate at 60°C to dry the film, and then over a second hot plate at 100°C, before being collected in a roll. After completing the perovskite layer, the substrate was cut into 25 mm × 25 mm to make solar cells. PC 61 BM, PEIE and Ag layers were deposited as described above.
Device PV measurements. The J-V curves were measured under an inert atmosphere without extra temperature control using a Keithley 2400 Source Meter under standard solar irradiation (AM 1.5G, 100 mW cm −2 ). The light source of Xe lamp with AM 1.5G solar filter is from Newport Model 67005. The light intensity was calibrated using a reference cell (Hamamatsu S1133 with KG5 filter, 2.8 × 2.4 mm of photosensitive area), which was calibrated by a certified reference cell (PV Measurements, certified by NREL with KG5 filter) under 1000 W m −2 AM 1.5G illumination from an Oriel AAA solar simulator fitted with a 1000 W Xe lamp. For the stability test under continuous illumination. The cell was encapsulated in an N 2 filled glovebox by covering a glass-lid onto the device with UV curable epoxy ( Supplementary Fig. 16). Stability tests were conducted using the Candlelight Maximum power tracker. The encapsulated devices were loaded in a device chamber and then were exposed to 1-sun conditions. J-V characteristics of the devices and the temperature and the relative humidity inside the devices chamber were recoded after every 10 min.
Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
The data that support the findings of this study are available on reasonable request from the corresponding author. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. © Crown 2020