Novel Physical Vapor Deposition Approach to Hybrid Perovskites: Growth of MAPbI3 Thin Films by RF-Magnetron Sputtering

Solution-based methods represent the most widespread approach used to deposit hybrid organic-inorganic perovskite films for low-cost but efficient solar cells. However, solution-process techniques offer limited control over film morphology and crystallinity, and most importantly do not allow sequential film deposition to produce perovskite-perovskite heterostructures. Here the successful deposition of CH3NH3PbI3 (MAPI) thin films by RF-magnetron sputtering is reported, an industry-tested method to grow large area devices with precisely controlled stoichiometry. MAPI films are grown starting from a single-target made of CH3NH3I (MAI) and PbI2. Films are single-phase, with a barely detectable content of unreacted PbI2, full surface coverage and thickness ranging from less than 200 nm to more than 3 μm. Light absorption and emission properties of the deposited films are comparable to as-grown solution-processed MAPI films. The development of vapor-phase deposition methods is of interest to advance perovskite photovoltaic devices with the possibility of fabricating perovskite multijunction solar cells or multicolor bright light-emitting devices in the whole visible spectrum.

industry-tested technique for large area film growth. The relatively lower deposition efficiency of sputtering is overcome by a magnetron-based device. To date, no reports about the deposition of hybrid perovskites thin films by sputtering has been reported.
Motivated by the above reported issues related to the preparation of hybrid organic-inorganic perovskites films, in this paper we demonstrate the successful one-pot growth of methylammonium lead iodide CH 3 NH 3 PbI 3 (MAPI) films by means of RF-magnetron sputtering starting from a single target made of a CH 3 NH 3 I (MAI) and PbI 2 mixture with a 30% w/w excess of MAI. Sputtering technique allows to finely tune the deposition conditions by adjusting different parameters such as, for example, RF-power, gas pressure, and target to substrate distance thus providing an excellent platform to further optimize perovskite films as well as to extend the approach presented here for MAPI to any other material of interest in the field of Perovskite Solar Cells (PSCs).  Table 1.

Film
Thickness (nm) Lifetime (ns)  Figure 1A presents a sketch of the MAPI thin films growth method used in the present work, i.e. the RF-magnetron sputtering starting from a target made of MAI and PbI 2 with a MAI excess of 30 wt%. Depositions were carried out at a RF-power of 40 W, with argon (P = 2 × 10 −2 mbar) as the sputtering gas in the DC-bias mode by setting its value to 80 V. Figure 1B shows the x-ray diffraction patterns (XRD) of a series of representative MAPI films with variable thicknesses from below 200 nm (Film 1) up to about 3.2 μm (Film 9) (details are reported in Table 1). Optimal growth conditions were obtained after extensive optimization work and are reported in the Experimental Section. The starting target for the sputtering deposition was a mixture of MAI and PbI 2 with a 30% w/w excess of MAI.

Results and Discussion
The films reported in Fig. 1B have a crystal structure compatible with that of tetragonal MAPI (vertical red bars in Figure -main reflections indicated in the figure) and are single-phase or present very low PbI 2 impurities (below 5% -the only peak of PbI 2 observed is marked with an asterisk in the Figure). In addition, the diffraction peaks are quite narrow, indicating a good crystallization due to the sputtering process which is a significant result considering that the substrate is not heated during the deposition (Full Width at Half Maximum for the peak at about 14° are reported in Table 1 of SI). The possibility of growing crystalline MAPI thin films on any substrate without in-situ and/or ex-situ thermal treatments, together with the use of a single target, are unique advantages of the present deposition process. In general, from Fig. 1B, it can be observed a slight preferential growth along the (00 l) directions, as suggested by the relative intensity of the experimental peaks corresponding to the (002) and (004) reflections compared to the calculated intensities. No significant shifts of the peaks as a function of film thickness are evident in the XRD patterns.
The surface morphology of the MAPI films has been investigated by Atomic Force Microscopy (AFM). Figure 2 reports some selected images of films with variable thickness.
From the images of Fig. 2, it is possible to see complete coverage of the substrate surface starting from film 4 ( Fig. 2B) which has an estimated thickness of about 340 nm. The morphology of the deposited layers is made of small spherical grains with size in the 100-200 nm range which tends to increase by increasing the film thickness. In particular, Fig. 2D shows that, for a bulk-like film (thickness ~3200 nm), the grain size is comparable to that of a polycrystalline powder. The average roughness estimated from the AFM maps, and defined as root mean square (R RMS ) of surface height, is around 20 nm for film 1 (thickness <200 nm), 7 nm for film 4 (thickness ~340 nm), 10 nm for film 7 (thickness ~640 nm) and 20 nm for film 9 (thickness ~3200 nm), values low enough to suggest a deposition mechanism based on layer growth on the fused silica substrate used here. Further insight into the morphology of the deposited film has been achieved by SEM microscopy. Figure 3 reports some representative images for film 4 (ca. 350 nm) and film 7 (ca. 640 nm) collected at two different magnifications (i.e., 20 kX and 100 kX). More images are displayed in the Supplementary Information. SEM inspection reveals and confirms the very good coverage of the substrate obtained with the present technique. In addition, a more compact structure is observed in the images of the more thick sample, i.e. figures C and D, suggesting a coalescence of the film as the thickness increases. The morphology observed in the present films has some similarities with the morphology reported for MAPI films deposited from different vapor methods (such as CVD and thermal evaporation) 10,11 . However, in the present case, a more compact layer is always observed with respect to the other techniques together with a better coverage of the substrate 10,11 .
Optical and excited-state properties of the deposited film have been assessed by absorptance (A) and photoluminescence (PL) measurements. Figure 4 shows a representative absorptance spectrum of a thin MAPI film (4) grown by sputtering. The spectrum of thicker samples clearly showed saturation phenomena for increasing photon energies above the band-gap, caused by the strong increase of the band-to-band absorption coefficient (see Fig. 1, Supporting Information). As expected, a sharp absorption edge is found in the typical region of MAPI  No evidence of enhanced light absorption at 550 nm due to PbI 2 is present, confirming its absence or, when present, its relatively low amount according to the XRD analysis. The measured photoluminescence spectrum is also reported in Fig. 4. The spontaneous emission is almost resonant with absorption, suggesting an intrinsic origin. Small variations of the emission peak energy from sample to sample, in the range of 10-20 meV, were observed (see inset of Fig. 4); an analogous effect was indeed reported on solution-processed MAPI thin films, too 13 . The measured photoluminescence spectrum was compared with the expected spectrum according to the reciprocity relation between absorption and emission: As the strong absorptance is due to direct band-to-band transitions, the photoluminescence spectrum derived in this way stems from spontaneous photon emission involving the very same intrinsic states. The excellent agreement between the expected and the measured spectra (red and black curves in Fig. 4) further supports the absence of any extrinsic contribution to light emission in MAPI films grown by sputtering. Figure 5 reports the dependence of the photoluminescence intensity on the excitation power level of two films (4 and 12). The reported behavior is representative of the whole set of samples investigated in this work: in all films, the spontaneous emission scaled linearly (m = 1) with the laser intensity at low excitation regimes, then grew superlinearly at higher pumping intensities, following a m = 3/2 power-law. In absence of nonradiative processes, light emission intensity is expected to linearly scale with the excitation level, independently of the nature of the involved electronic states, as all absorbed photons are subsequently reemitted. Nonradiative channels change this functional dependence. We have recently developed a simple approach to relate nonradiative processes with the dependence of the photoluminescence intensity on the excitation power 15 . In simple terms, if the concentration of electrons (holes) as a function of the laser intensity scales as a power law with index m e (m h ), the ensuing photoluminescence intensity also follows a power-law, but with index m = m e + m h . We found that the kinetics triggered by deep traps leads to m = m e + m h = 1 + 1/2 = 3/2, under the assumption (Shockley-Read-Hall model) that traps in their ground state can be filled by only one type of carrier (in the following we assume to be electrons) 15 . This result can be intuitively understood by looking at the recombination paths sketched in Fig. 5. Electrons undergo a conventional first order decay process; their density is thereby linearly dependent on the laser intensity in steady-state (m e = 1). Holes recombine with trapped electrons, following a true bimolecular process; the square of the hole density is thus proportional to the laser intensity, which leads to m h = 1/2. The hole lifetime is much longer than that of electrons, owing to the fact that the bimolecular recombination is a slow process at low carrier concentration. As a consequence, the densities of electrons and holes are unbalanced: the concentration of holes is much larger, and the semiconductor behaves as p-doped. m = 3/2 is indeed the most common response observed in solution-processed MAPI films [16][17][18][19] .
In order to explain the linear behavior observed at low excitation regime, we first note that m = 1 in our samples cannot be explained in terms of radiative recombinations. At low power levels, the measured photoluminescence quantum yield was in fact around 10 −3 , proving that nonradiative recombinations are the most efficient decays. Figure 4 shows the proposed model to explain the experimental findings. We assume that films contain a low concentration of shallow acceptors (unintentional chemical p-doping). As long as the excitation rate is low, the hole concentration is independent of excitation (m h = 0), while the dynamics of electrons remains unaffected (m e = 1) and thus m = m e . Increasing the excitation level, the concentration of photogenerated holes becomes dominant and the m = 3/2 behavior is recovered. This analysis is consistent with the observation that laser intensity, at which the transition from chemical doping to photodoping occurs, depends on sample, ranging from 10 mW/cm 2 (i.e., below the light intensity delivered by sun) to a few 10 2 W/cm 2 . Unintentional doping (both p-and n-type) is often reported also in solution-processed MAPI films 20,21 .
The time-resolved PL spectra of samples, representative of thin, intermediate and thick samples, are shown in Fig. 6, together with the spectrograms of the two films with the longest lifetimes. The spectrograms of all films show no spectral shift of the emission with time (see the spectra recorded at two different delays in the left panel of Fig. 6), as expected from intrinsic band-to-band emission following carrier thermalization at the band edges.
The photoluminescence lifetimes of all the films investigated in the present work are reported in Table 1 and provide a measure of trap concentration and non-radiative recombination rates associated to them. In general, the lifetimes are comparable to those of as-prepared solution-processed MAPI thin films [22][23][24][25][26] , but fall short with respect to state-of-the-art MAPI thin films optimized with post-growth treatments to passivate deep traps that have been demonstrated to enhance the photoluminescence lifetime and, consequently, quantum yield [27][28][29][30][31] . The results reported in this work refer to films which did not undergo any post-growth treatment and suggest that MAPI thin films produced by sputtering could be also substantially improved in their physico-chemical properties with future optimization work. By way of example, a possible morphology tuning and/or substrate heating to reduce the surface area of the film and/or to modulate the grain size, and therefore the surface defects, could already provide a significant increase in optical emission yield. Figure 6 shows that the longest carrier lifetimes have been obtained for intermediate thicknesses, which are close to the optimal values for PSCs absorbing layers. On other hand, for thinner and thicker films a reduction below 2 ns is observed. For thinner films this may be related to a major role played by surface defects, while for thicker film such reduction could be possibly related to an increase of the polycrystallinity which is known to enhance the nonradiative channels [27][28][29][30][31] .

Conclusions
We reported the successful deposition of MAPI thin films by RF-magnetron sputtering. MAPI films were grown starting from a single-target made of MAI and PbI 2 (with a 30% w/w excess of MAI) and appeared to be single-phase, with full surface coverage and thickness ranging from less than 200 nm to more than 3 μm. The optical properties of the deposited films are comparable to as-grown solution-processed MAPI films and, in the future, the photoluminescence quantum yield could be substantially improved with post-growth passivation treatments. The development of vapor-phase deposition methods is of great interest in the current research on hybrid perovskites in view of a scale-up of device fabrication, the precise control of stoichiometry and the possibility of growing perovskite-perovskite heterostructures.

Methods
Film deposition. Thin films of MAPI have been deposited on amorphous silica substrates (MaTek, roughness ca. 1 nm) by means of radio frequency magnetron sputtering starting from a MAI/PbI 2 mixture (Aldrich, > 99.9%) with a 30% w/w excess of MAI. The target (diameter 5.08 cm, thickness 1 cm) was made of pressed powders of MAI/PbI 2 mixture. Depositions parameters were: i) target-to-substrate distance, 2 cm, ii) RF-power, 40 W, iii) argon pressure, 2 × 10 −2 mbar. The depositions have been carried out in DC-bias mode by setting the value to 80 V (with respect to the target electrode), thus assuring a good control over the deposition rate. Film thickness has been determined by means of a mechanical profilometer. Estimated film growth is about 30 nm/min. XRD Diffraction. The structural properties of the deposited thin films were characterized by X-ray diffraction (XRD) by means of a Bruker D8 Advance instrument (Cu radiation) in a Bragg-Brentano set-up.
Optical Properties Measurement. Reflectance (R) and transmittance (T) measurements were performed at 8° angle of incidence using a dual-beam spectrophotometer with an integrating sphere accessory (Agilent Cary 5000 UV-Vis-NIR). Absorptance (A) was calculated as 1-R-T.
Time-resolved photoluminescence. Samples were mounted excited by a regenerative amplifier laser (Coherent Libra) delivering 100-fs-long pulses at a repetition rate of 1 KHz. Photoluminescence was dispersed with a grating spectrometer (Princeton Instruments Acton SpectraPro 2300i equipped with a 50 gr/mm grating blazed at 600 nm), dispersed and detected by a streak camera (Hamamatsu).
CW photoluminescence. Samples were excited by a diode-pumped Nd:YVO 4 CW laser at 532 nm (Spectra Physics Millennia). Photoluminescence was dispersed by a grating spectrometer (Princeton Instruments Acton SpectraPro 2300i) and detected by a LN-cooled CCD camera (Princeton Instruments PIXIS). At the highest excitation intensities, laser beam was chopped to reduce the overall thermal loading. AFM Measurements. Surface topography and film roughness were measured by atomic force microscopy (AFM) with a NT-MDT Solver P47H-Pro instrument in semi contact mode at 1 Hz scan speed by a Etalon high-resolution non-contact silicon tip. Images were analyzed by WSxM software.