Phenomenological mechanisms of hybrid organic–inorganic perovskite thin film deposition by RIR-MAPLE

Hybrid organic–inorganic perovskites (HOIPs) are a class of materials featuring organic molecules incorporated into a crystalline inorganic framework. The archetypal hybrid perovskite features a three-dimensional structure comprising corner-sharing metal halide octahedra with a small organic molecule, usually methylammonium, occupying the space created by these octahedra.¹ Such HOIPs are uniquely characterized by tunable bandgaps,² large absorption coefficients,³ long carrier diffusion lengths,⁴ and shallow energy levels for intrinsic defects,⁵ all of which make this class of materials an excellent candidate for optoelectronics.^6,7 HOIPs have already been demonstrated as thin film transistors,^8–10 photovoltaic solar cells,^11,12 light emitting diodes (LEDs)^13,14 and lasers,^15,16 photocatalysts,¹⁷ and photodetectors.¹⁸ The existence of simple, low-temperature processes to synthesize HOIP single crystals and thin films has contributed to intensive research in this area.¹⁹ In fact, a wide variety of synthesis techniques have been demonstrated to yield high-quality HOIP thin films.²⁰ 

One promising approach for the physical vapor deposition of HOIP thin films is resonant infrared, matrix-assisted pulsed laser evaporation (RIR-MAPLE), which first demonstrated thin-film deposition of methylammonium lead triiodide (MAPbI₃) in 2018.²¹ Subsequent work has shown that HOIPs deposited by RIR-MAPLE have similar crystallinity, morphology, and optical properties compared to those synthesized by solution processing.²² As an example of film morphology, Figs. 1(a) and 1(b) show top-view and cross-sectional scanning electron microscopy (SEM) images, respectively, of MAPbI₃ thin films deposited by RIR-MAPLE (4-h deposition time, annealing at 110 °C for 10 min). The MAPbI₃ films deposited by RIR-MAPLE demonstrate compact surface features that extend throughout the film thickness and are comparable to solution-processed films. In addition, Fig. 1(c) shows a photograph of MAPbI₃ films deposited by RIR-MAPLE to demonstrate the typical sample size (1 × 1 cm²) and the uniform surface coverage of deposition. From experience, it has been determined that the RIR-MAPLE deposited film thickness is uniform for a 2 cm radius at the center of the substrate, and the thickness decreases linearly with radial distance from the center. It is important to note that a common feature of thin films deposited by RIR-MAPLE is inherent surface roughness. For MAPbI₃, the root-mean-square surface roughness determined from a 50 × 50 μm² atomic force microscopy image was 62 nm in an RIR-MAPLE-deposited film (compared to 18 nm in a spin-cast film).²² Nonetheless, this previous work²² extensively characterized MAPbI₃ films deposited by RIR-MAPLE, compared the RIR-MAPLE films to spin-cast films, and used the RIR-MAPLE films to demonstrate photovoltaic solar cells with a stabilized power conversion efficiency of 12.2%, which is reasonable performance as a first demonstration without extensive device optimization.

Yet, the true potential of HOIP thin-film deposition by RIR-MAPLE lies in its application to two-dimensional, or layered, HOIP thin films featuring larger organic molecules that self-assemble with sheets of metal halide octahedra to form organic–inorganic quantum wells.^23,24 While three-dimensional HOIP systems, like MAPbI₃, can be deposited with excellent properties from a wide range of simple, solution-based processing techniques, the same cannot be said for unique HOIP materials featuring larger organic molecules that are functional. Such functional organic moieties are desirable because they can be used to control and tune the optoelectronic properties of the overall HOIP crystal. Unfortunately, the chemistry of such functional molecules is often more complicated, precluding the simple solution-based deposition that is so effective for the organic and inorganic constituents in MAPbI₃. Importantly, the utility and versatility of RIR-MAPLE growth for two-dimensional HOIPs was demonstrated in a study of oligothiophene-based hybrid perovskite thin films featuring different sized oligomers and different metal halides to achieve film compositions that can be difficult to realize with other synthesis approaches.²³ In fact, the RIR-MAPLE growth of oligothiophene-based HOIP films achieved previously unattainable, controlled deposition of different oligothiophene and halide compositions that resulted in the experimental confirmation of theoretically predicted, tunable quantum well band alignments.²³ 

Thus, the fundamental understanding of the unique growth mechanisms that constitute RIR-MAPLE deposition of HOIPs is critical to advancing this novel deposition technique and to expanding HOIPs as a general semiconductor material technology. While three-dimensional HOIPs are not the most advantageous application of RIR-MAPLE, these materials are well understood and provide a solid framework for elucidating the phenomenological mechanisms of HOIP thin-film growth by RIR-MAPLE.

RIR-MAPLE is a variation of pulsed laser deposition that uses a low-energy laser (2.94 μm Er:YAG) to irradiate a frozen target (solution or emulsion) in which the matrix solvent contains hydroxyl bonds to resonantly absorb the laser energy. The target material is dissolved or emulsified within this matrix such that it is transferred, intact, from the target to the substrate by the kinetic energy resulting from sublimation of the matrix during each laser pulse.^25,26 Matrix-assisted pulsed laser evaporation was originally developed for polymer thin films.^27–30 Since then, a substantial effort has been made to establish a wide range of capabilities for the RIR-MAPLE deposition of organic and hybrid nanocomposite thin films based on frozen oil-in-water emulsion targets.^31–37 This body of work has identified design rules for the emulsion target chemistry needed to achieve organic and hybrid nanocomposite thin-film deposition by RIR-MAPLE: (1) the matrix solvent must contain hydroxyl bonds to resonantly absorb the laser energy; thus, the continuous emulsion phase of water (with two hydroxyl bonds) serves this purpose; (2) the primary solvent must be a good solvent for the target material, and it should have a low solubility-in-water to yield meta-stable emulsions; and (3) a low vapor pressure solvent must be included to prevent sublimation of the target due to the active vacuum conditions of the growth chamber.

In order to deposit hybrid perovskites using RIR-MAPLE, the emulsion target chemistry had to be altered to eliminate the presence of water, which leads to degradation of HOIPs.²¹ As a result, the HOIP target chemistry uses solutions in a polyalcohol instead of emulsions in water. Specifically, monoethylene glycol (MEG), with two hydroxyl bonds, is used to provide multiple functions within the target solution (both the matrix solvent and the low vapor pressure solvent). The primary solvent for hybrid perovskite deposition by RIR-MAPLE is the commonly used dimethyl sulfoxide (DMSO). Thus, the typical hybrid perovskite target chemistry used for deposition by RIR-MAPLE comprises a stoichiometric ratio of precursor materials in a 1:1 (by volume) mixture of DMSO:MEG. Another important difference for HOIP deposition by RIR-MAPLE, compared to organic and hybrid nanocomposite deposition, is that instead of transferring fully synthesized materials to create blended composition films, HOIP precursor materials are transferred intact and the desired crystal forms on the substrate after deposition. It is important to note that stoichiometric ratios of precursor materials are used in RIR-MAPLE HOIP targets, in contrast to other physical vapor deposition techniques, because only the matrix solvent resonantly absorbs the low energy infrared laser. This selective absorption of the laser energy creates a gentle process that protects organic molecules and prevents the loss of organic molecules common to most physical vapor deposition techniques.²¹ Because RIR-MAPLE enables direct transfer of the target composition to the substrate, stoichiometric and non-stoichiometric precursor concentrations can be achieved in a controlled manner. Additional unique capabilities provided by RIR-MAPLE for HOIP thin film deposition include (1) RIR-MAPLE can use low concentration precursor solutions (∼10 mM or less), as needed, to mitigate solubility challenges that can arise from larger organic molecules in solution with metal halides;²³ (2) film thickness can be varied by controlling growth parameters (e.g., substrate-to-target distance or deposition time) instead of precursor concentration such that the morphology and chemistry of precursor species in target solutions are consistent; and (3) RIR-MAPLE enables high-quality interfaces of hybrid perovskite films due to the minimal solvent contamination of the substrate (compared to solution processing) and the active vacuum environment during growth.

Although RIR-MAPLE has been used to demonstrate HOIP thin film deposition, the phenomenological mechanisms of film formation for these materials have not been articulated. Such knowledge could help explain and mitigate unique characteristics of HOIP films deposited by RIR-MAPLE, such as the inherent surface roughness. Previous work has described the kinetics of MAPLE deposition using an ultraviolet laser,³⁸ and the current phenomenological understanding of emulsion-based RIR-MAPLE has been described for polymer thin films.³⁷ However, the distinct properties of HOIP thin-film deposition by RIR-MAPLE are unknown, thereby motivating a careful investigation of the underlying film formation mechanisms. Therefore, it is important to note that this work focused on the initial stages of film growth. While HOIP films for characterization and device fabrication typically are grown for ∼4 h via RIR-MAPLE, the films in this study are grown for a maximum of 2 h. As a result, pinholes and incomplete substrate coverage were observed, but these films would not be used in applications. In addition, it is important to determine the role of the inorganic metal halide in film formation. Therefore, in this work, three different HOIPs [methylammonium lead triiodide (MAPbI), methylammonium lead tribromide (MAPbBr), and methylammonium lead trichloride (MAPbCl)] were deposited by RIR-MAPLE for different time intervals to investigate and characterize the temporal evolution of films.

The organic salts methylammonium iodide (CH₃NH₃I, 99.999% purity), methylammonium bromide (CH₃NH₃Br, 99.999%), and methylammonium chloride (CH₃NH₃Cl, 99.999%) were purchased from Dyesol Inc. (now GreatCell Solar). The metal halides lead (II) iodide (PbI₂, 99.999% purity), lead (II) bromide (PbBr₂, 99.999% purity), and lead (II) chloride (PbCl₂, 99.999% purity) were purchased from Alfa Aesar. Anhydrous dimethyl sulfoxide (DMSO) and anhydrous monoethylene glycol (MEG) were purchased from Sigma-Aldrich Corp. (now MilliporeSigma). Soda-lime glass or quartz substrates were purchased from VWR International, LLC. All materials were used as received without further purification.

Soda-lime glass substrates were used for deposition of MAPbI and MAPbBr, while quartz substrates were used for deposition of MAPbCl (to reduce UV absorbance by the substrate). For each RIR-MAPLE growth cycle, six substrates were loaded into a substrate holder at equidistant radial positions around the circumference of the substrate holder. Due to the radial dependence of film thickness, this substrate configuration ensured that the deposition geometry was consistent throughout the study. Substrates were diced to 1 × 1 cm² squares for each sample and were cleaned by the following protocol before being loaded into the RIR-MAPLE growth chamber: (1) submersion and sonication for 10 min in acetone, followed by methanol, and then isopropanol; (2) substrate drying by nitrogen gas; and (3) 10-min oxygen plasma cleaning in a plasma asher.

RIR-MAPLE target solutions were prepared in a nitrogen-filled glovebox with O₂ and H₂O levels <0.1 ppm. HOIP precursor materials comprising organic salts and metal halides, namely, methylammonium iodide (MAI) and PbI₂, methylammonium bromide (MABr) and PbBr₂, or methylammonium chloride (MACl) and PbCl₂, were measured at a molar ratio of 1:1 organic salt:metal halide with a 20 mM concentration and dissolved in a 1:1 mixture (by volume) of DMSO and MEG. Target solutions were loaded into a stainless-steel target cup within the RIR-MAPLE deposition chamber. A nitrogen purge cycle, in which the chamber pressure was cycled from 1.33 kPa to 13.33 kPa, was performed to prevent the condensation of moisture as the target was cooled by liquid nitrogen to −196 °C. After reaching this target temperature, nitrogen gas was introduced to maintain the chamber pressure at ∼33–36 kPa until the target solution was completely frozen. Once the target solution was frozen solid, the nitrogen gas was turned off and the chamber pressure was reduced to <1 × 10⁻⁴ kPa before initiating RIR-MAPLE deposition.

To directly observe the development of HOIP thin films during RIR-MAPLE deposition, temporal film evolution was observed under the same growth conditions, like chamber pressure (described above), laser irradiance, and substrate temperature. An Er:YAG laser with a peak wavelength of 2.94 μm, a pulse frequency of 2 Hz, and a laser fluence of 125–135 mJ/cm² was used for all depositions. The target-to-substrate distance was 7 cm. The target rotation was maintained at 4 RPM to achieve uniform irradiation of the target by a laser raster pattern (from center to outer edge of target cup and back). Substrates were exposed to the irradiated target for defined time intervals before being removed from the vacuum chamber. All film deposition was conducted at a substrate temperature of 25 °C to prevent condensation when removing samples after breaking vacuum at specified time intervals (30 s and 1, 2, 3, 4, 5, 10, 15, 20, 40, 60, 80, 100, 120 min). These time intervals were chosen to provide a comprehensive picture of film formation that focuses on the crucial first steps. Based on previous experience, the reasonable assumption was made that after 2 h, mostly continuous and uniform films are achieved.

The morphological, crystallographic, and optical properties of the films at the specific time intervals for RIR-MAPLE deposition were characterized as follows. Scanning electron microscopy (SEM) images were obtained using an FEI XL-30 system with the accelerating voltage set at 5 kV. X-ray diffraction (XRD) spectra were measured using a Panalytical Empyrean powder X-ray diffractometer with Cu K-α radiation using operating voltage and current set at 40 kV and 40 mA, respectively. Photoluminescence (PL) spectroscopy measurements were obtained using a Horiba Jobin Yvon LabRam ARAMIS system with an excitation source appropriate for each HOIP: 633 nm HeNe laser (MAPbI), 442 nm HeCd laser (MAPbBr), or 325 nm HeCd laser (MAPbCl). Ultraviolet-visible (UV-Vis) absorbance spectroscopy measurements were conducted using a Shimadzu UV-3600 spectrophotometer. Contact angle measurements of HOIP target solutions used in RIR-MAPLE were conducted using a Rame-Hart Instrument, Co. contact angle goniometer.

SEM images collected at different time intervals for MAPbI deposition by RIR-MAPLE are shown in Figs. 2(a)–2(n). Circular features with a large size distribution are observed to be scattered randomly across the substrate surface, and these features overlap as the deposition time increases such that a continuous film is achieved at t = 120 min. XRD spectra of the MAPbI samples for different time intervals [Figs. 3(a) and 3(b)] provide a description of how the deposited material evolves into the desired HOIP crystal structure. While Fig. 3(a) shows a wide range of 2theta values for each time interval in stacked plots, Fig. 3(b) shows two peaks of interest in overlaid plots. In Figs. 3(a) and 3(b), the first indications of crystalline phases are observed at t = 3 min; that is, XRD peaks greater than the noise floor (on a linear scale) are observed. Initially (t = 3 min to t = 5 min), the stronger of these peaks occurs around 12.6°, corresponding to PbI₂ (denoted by @), while a weaker peak corresponding to the (110) plane of MAPbI is observed around 13.8° (denoted by #). At t = 10 min, the MAPbI (110) peak becomes dominant, and for t > 20 min, the XRD peak corresponding to the PbI₂ crystalline phase is no longer observed. In addition, for longer deposition times (t ≥ 80 min), lower angle XRD peaks (denoted by *) are observed that correspond to MAPbI–solvent complexes in which MAPbI crystalline phases contain solvent molecules. These intermediate phases result from trapped solvent in the film and are highly sensitive to vacuum conditions during growth and post-deposition annealing. The appearance of these solvent complex phases for longer deposition times may indicate that it is more difficult for solvent to evaporate once the crystal is mature. The optical properties for MAPbI depositions of different time intervals were measured using PL and UV-Vis absorbance spectroscopies. The peak wavelengths of the normalized PL spectra [Fig. 3(c)] demonstrate a slight redshift with increasing deposition time to ∼764 nm for t ≥ 60 min (Table I). The reported bandgap for MAPbI is ∼1.6 eV, corresponding to a wavelength of ∼775 nm.³⁹ Meanwhile, the full-width at half-maximum (FWHM) linewidths for the PL spectra at different time intervals have an average value of 37.9 nm and no strong trend with increasing deposition time (Table I). The first measurable UV-Vis absorbance spectrum [Fig. 3(d)] is observed at t = 10 min, and the absorbance steadily increases with deposition time. Strong band edge absorbance is observed for t ≥ 40 min, which is the same time interval for the observation of a strong MAPbI (110) XRD peak without the presence of a metal halide peak, indicating that the MAPbI crystal is fully formed in some regions of the film.

Similar to RIR-MAPLE deposition of MAPbI, SEM images of MAPbBr deposited by RIR-MAPLE at different time intervals [shown in Figs. 4(a)–4(n)] demonstrate the overlap of circular features with a large size distribution that yield a mostly continuous film at t = 120 min. However, a notable difference is observed in the radial concentration of material within the circular features of MAPbBr compared to MAPbI. The MAPbBr circular features tend to have a sparse concentration of material [Fig. 4(b)], while those of MAPbI tend to have a dense concentration [Fig. 2(b)]. The XRD spectra of MAPbBr samples deposited for different time intervals are shown as stacked plots and overlaid plots in Figs. 5(a) and 5(b), respectively. The first observation of XRD peaks greater than the noise floor (on a linear scale) occurs at t = 2 min for the MAPbBr (100) and (200) crystal planes at ∼14.6° and 29.2°, respectively (denoted by #). In contrast to MAPbI, an XRD peak corresponding to the metal halide (PbBr₂ in this case) is only observed for t = 80 min (denoted by @). Otherwise, no additional crystalline phases (including solvent complexes with the perovskite) are observed in the XRD spectra. MAPbBr deposited by RIR-MAPLE demonstrated similar time development of the PL [Fig. 5(c)] and UV-Vis absorbance [Fig. 5(d)] characteristics compared to MAPbI. The PL spectra of MAPbBr samples show a slight redshift in the peak wavelength as the deposition time increases, with a value of ∼532 nm for t ≥ 60 min (Table I). The reported bandgap of MAPbBr is 2.3 eV, which corresponds to a wavelength of ∼539 nm.⁴⁰ The PL FWHM linewidth does not exhibit a strong trend with increasing deposition time, and the average value is 19.8 nm. The smaller FWHM linewidths for MAPbBr PL spectra compared to MAPbI are consistent with the crystalline phase purity observed by the XRD characterization of MAPbBr. The first measurable UV-Vis absorbance spectrum for MAPbBr [Fig. 5(d)] is observed at t = 10 min, and strong band edge absorbance is observed for t ≥ 10 min, indicating that the MAPbBr crystal is already fully formed in some regions of the film at this time. The absorbance continues to increase with increasing deposition time.

SEM images of MAPbCl deposited by RIR-MAPLE for different time intervals are shown in Figs. 6(a)–6(n). As in the case of MAPbI and MAPbBr, circular features of material with a large size distribution are deposited onto the substrate surface, and these features overlap as the deposition time increases such that a mostly continuous film is achieved at t = 120 min. The circular features of MAPbCl tend to have a sparse concentration of material, as seen in the MAPbBr samples. The XRD spectra of MAPbCl samples deposited for different time intervals are shown as stacked and overlaid plots in Figs. 7(a) and 7(b), respectively. The first observation of XRD peaks greater than the noise floor (on a linear scale) occurs at t = 4 min for the MAPbCl (100) and (200) crystal planes at ∼15.2° and 30.4°, respectively (denoted by #). In the case of MAPbCl, no additional crystalline phases are observed in the XRD spectra (including PbCl₂ or solvent complexes with the perovskite). MAPbCl deposited by RIR-MAPLE demonstrated similar time development of the PL [Fig. 7(c)] and UV-Vis absorbance [Fig. 7(d)] characteristics compared to MAPbI and MAPbBr. The PL spectra of MAPbCl samples show a slight redshift in the peak wavelength as the deposition time increases, with a value of ∼402 nm for t ≥ 60 min (Table I). The reported bandgap of MAPbCl is 3 eV, which corresponds to a wavelength of ∼413 nm.³⁹ The PL FWHM linewidth does not exhibit a strong trend with increasing deposition time, and the average value is 10.2 nm. The smaller FWHM linewidths for MAPbCl PL spectra compared to MAPbI and MAPbBr are consistent with the crystalline phase purity observed by the XRD characterization of MAPbCl. The first measurable UV-Vis absorbance spectrum for MAPbCl [Fig. 7(d)] is observed at t = 2 min, and the absorbance increases with increasing deposition time (features at 475 nm and 525 nm are related to the spectrometer system). Strong band edge absorbance is observed for t ≥ 10 min, which is the same time interval for the observation of a stronger and sharper MAPbCl (100) XRD peak, as seen in Fig. 7(a), indicating that the MAPbCl crystal is fully formed in some regions of the film.

As shown in the SEM images for MAPbI, MAPbBr, and MAPbCl deposited for 30 s [Figs. 2(a), 4(a), and 6(a), respectively], circular features are observed on the substrate surface at the very beginning of film deposition. Based on the understanding of RIR-MAPLE deposition of polymers,³⁷ these features are likely the result of target droplets that adsorb onto the substrate after being ejected from the target when irradiated by the laser. To reiterate, the frozen target comprises a solution of organic and inorganic precursors (organic salts and metal halides in a stoichiometric ratio) dissolved in a 1:1 mixture of DMSO and MEG. The MEG is the matrix solvent that resonantly absorbs the laser energy and sublimates into the vapor phase (the majority of which is pumped away by the active vacuum conditions of the growth chamber). In the area of the pulsed laser irradiation, the laser beam is not thermally confined and the frozen target melts slightly. The resulting target droplets containing precursors and DMSO are ejected by the transfer of kinetic energy as the MEG sublimates. Ideally, these target droplets contain an equilibrium concentration of precursors and DMSO as they leave the target surface. However, during the time-of-flight from the target to the substrate (7 cm distance), some amount of DMSO is pumped away due to the active vacuum condition of the chamber. In addition, because the target droplets are transferred intact to the substrate, if an area of the target were rich with organic precursors, inorganic precursors, or solvent, the transferred droplet from that area would reflect the same composition.

Once the droplet adsorbs onto the substrate surface, wetting behavior contributes to distinct droplet morphologies seen in SEM images (Figs. 2, 4, and 6), such as dense droplets observed primarily in MAPbI [Fig. 8(a)] and sparse droplets observed primarily in MAPbBr and MAPbCl [Fig. 8(b)]. It is important to note that a third morphology for the adsorbed droplet is observed in the SEM images of this study [Fig. 8(c)], identified as a solvent droplet. These droplets are considered solvent rich due to the contrast of the SEM image. Sample areas with greater atomic weight tend to scatter electrons efficiently and appear as bright areas in the resulting image (high contrast). Sample areas with lower atomic weight tend to scatter electrons poorly and appear as dim areas in the resulting image (low contrast). Therefore, the droplet morphology shown in Fig. 8(c) most likely results from the low atomic weights of DMSO and/or MEG. To test this hypothesis, a solvent-rich RIR-MAPLE target was prepared and irradiated for 10 s, and the resulting SEM images of the droplets were similar in appearance to that shown in Fig. 8(c).

The wetting behavior of the target droplets is determined largely by the coffee ring effect,^41–45 in which an immobile droplet (containing HOIP precursors and solvent) adsorbed to the substrate surface under vacuum experiences a faster rate of solvent evaporation at the circumference of the droplet. A fluid flow gradient is created by the anisotropic evaporation of solvent, thereby driving an increased precursor concentration toward the droplet circumference. As an example, the yellow arrow labeled “radial concentration” in Fig. 8(d) shows the increasing material concentration in a MAPbI droplet from the center toward the circumference (indicated by the brighter image contrast). This coffee ring effect is also clearly observed in the SEM images for MAPbI, MAPbBr, and MAPbCl (Figs. 2, 4, and 6, respectively) as brighter contrast regions at the droplet circumference (corresponding to larger material concentrations). As a result of the coffee ring effect, the precursors at the droplet boundary determine the contact angle and droplet wetting. Contact angle measurements of immobile droplets were conducted to describe the wetting behavior of the adsorbed droplet–substrate interface for the three different target solutions (MAPbI, MAPbBr, and MAPbCl precursors in 1:1 DMSO:MEG). The measured contact angles (average from three separate measurements) are shown in Table II and demonstrate a decrease in the contact angle with decreasing halide atomic number (from MAPbI to MAPbBr to MAPbCl). The larger contact angle for MAPbI means that the adsorbed droplets do not wet the surface as well as MAPbBr and MAPbCl. However, these differences in wetting do not fully explain the origin of the dense (MAPbI) and sparse (MAPbBr, MAPbCl) droplet morphologies observed by SEM.

Another important issue to consider in the explanation of the dense and sparse droplets is the Gibbs free energy of formation for each HOIP system, which decreases with decreasing halide atomic number.⁴⁶ Because the Gibbs free energy of formation is larger for MAPbI compared to MAPbBr and MAPbCl, the time interval at which the MAPbI crystal is fully formed is expected to be longer. The materials characterization at different time intervals for the HOIP materials supports this hypothesis. First, for MAPbI, XRD peaks corresponding to PbI₂ and MAPbI are initially observed at t = 3 min, which is defined as the time for the onset of MAPbI nucleation. However, the PbI₂ peak is dominant until t = 10 min, and this metal halide peak persists until t = 40 min, at which time it is no longer observed. Therefore, nucleation of MAPbI begins relatively early during deposition, but the crystal is not fully formed until 40 min have passed. This timeline is supported by the emergence of a strong band edge in UV-Vis absorbance measurements at t = 40 min for MAPbI.

In contrast, for MAPbBr, nucleation occurs at t = 2 min (defined by the first observation of the HOIP XRD peak). Importantly, the only non-perovskite crystalline phase observed is an XRD peak for the metal halide, PbBr₂, at t = 80 min. This metal halide peak most likely results from some type of degradation for the specific sample because of the late stage at which it is observed, and the metal halide peak is not observed for any other MAPbBr samples in the study. In addition, no solvent complex peaks are observed by XRD of the MAPbBr samples, and the PL FWHM linewidths are much narrower than those for MAPbI, consistent with better phase purity. Strong band edge absorbance is observed at t = 10 min, indicating that MAPbBr is already fully formed at this time, which is consistent with a lower Gibbs free energy of formation. Similarly, for MAPbCl, nucleation occurs at t = 4 min (defined by the first observation of the HOIP XRD peak). Neither PbCl₂ nor solvent complex peaks are observed at any time by XRD measurement of the MAPbCl samples. Furthermore, the PL FWHM linewidths are narrower than those for MAPbBr (indicative of better phase purity) and strong band edge absorbance is observed at t = 10 min, all of which indicate that MAPbCl is fully formed by t = 10 min, consistent with a lower Gibbs free energy of formation.

These observations help explain the dense (MAPbI) and sparse (MAPbBr, MAPbCl) droplets observed by SEM. The sparse droplets are formed, in part, because HOIP crystal nucleation disrupts wetting by the coffee ring effect. Regions of nucleation within a droplet behave similarly to the droplet circumference with rapid solvent evaporation and promote increased precursor concentration. In fact, bright contrast nucleation regions can be observed within MAPbBr and MAPbCl droplets as early as t = 30 s [Figs. 4(a) and 6(a), respectively]. Therefore, the dense droplets of MAPbI result from less surface wetting and slower crystal nucleation/formation, while the sparse droplets of MAPbBr and MAPbCl result from more surface wetting and faster crystal nucleation/formation.

Because crystal formation is much slower for MAPbI, another important aspect of RIR-MAPLE film formation specific to this HOIP is droplet–droplet diffusion, that is, the diffusion of precursor materials from one target droplet to another enabled by physical overlap. To clarify, the Er:YAG pulse frequency used in RIR-MAPLE deposition is 2 Hz, so within 1 min of deposition, ∼120 laser pulses irradiate the frozen target and, for each pulse, target droplets are generated and distributed randomly across the substrate surface. In some regions, adsorbed target droplets spatially overlap, such as the droplets labeled “1” and “2” in the SEM image for MAPbI at t = 1 min [Fig. 8(d)]. The yellow arrow labeled “precursor diffusion” in Fig. 8(d) shows droplet–droplet diffusion from a bright contrast, dense droplet toward a dimmer contrast droplet that is less dense. This droplet-droplet diffusion also can be observed by the irregularly shaped droplets prominent in the SEM images for MAPbI during the early stages of film deposition [Figs. 2(b)–2(f)]. Droplet–droplet diffusion begins as soon as adsorbed droplets have sufficient overlap to create concentration gradients (∼t = 1 min, before crystal nucleation). Droplet–droplet diffusion likely continues throughout the remainder of deposition, even though it is more difficult to observe directly by SEM. In contrast, due to the faster crystal formation of MAPbBr and MAPbCl, droplet–droplet diffusion is not observed by SEM for these HOIP materials. While the overlap of adsorbed droplets still occurs, the circular droplet boundaries are clearly maintained throughout the deposition (Figs. 4 and 6, respectively).

Finally, after the initial crystal formation, as target droplet adsorption, wetting, diffusion (as applicable), and nucleation continue for the entire deposition time, precursor materials accumulate on the substrate and are incorporated in the HOIP crystal structure. The observed redshift in the PL peak wavelength with increasing deposition time for each HOIP material system supports such precursor accumulation as a final step of film growth. Previously, shifts in the PL spectra peak wavelengths have been associated with trap states along the band edges⁴⁷ or distortions of the HOIP crystal structure.⁴⁸ However, a recent study has shown there is a strong dependence between the redshift in PL emission and the increased thickness of MAPbI films.⁴⁹ Similarly, a redshift of the bandgap energy for MAPbBr has been attributed to an increase in film thickness and grain size.⁵⁰ The observation of increasing UV-Vis absorbance spectra over time for each HOIP material system, accompanied by sharper band edge and excitonic absorbance, also supports the idea that once the crystal is formed, additional target droplets adsorbed to the substrate are incorporated into the crystal due to the accumulation of precursor materials.

To conclude, the following phenomenological mechanisms are proposed to describe the film formation of HOIPs deposited by RIR-MAPLE: (1) adsorption of target droplets onto the substrate surface; (2) surface wetting by the droplet following the coffee ring evaporation model; (3) droplet–droplet diffusion of precursors due to the spatial overlap of adsorbed droplets (only for HOIPs with a large Gibbs free energy of formation); (4) nucleation of HOIP crystallites; and (5) accumulation of target droplets on the substrate and HOIP crystal growth. A schematic diagram of the phenomenological film formation timeline for each HOIP material is shown in Fig. 9, indicating the time intervals for which these mechanisms are dominant. The elucidation of these mechanisms not only informs process development for RIR-MAPLE deposition of three-dimensional HOIPs, but it helps provide experimental design parameters to determine the most favorable conditions for controlled crystallization and film growth. Such information can help improve the film quality of HOIPs deposited by RIR-MAPLE by directly addressing challenges, like increasing the crystal grain size and reducing surface roughness.²² The articulation of these phenomenological mechanisms also provides foundational understanding to continue exploration of more complex, two-dimensional HOIP materials featuring larger, functional organic molecules within the metal halide framework.

This work was supported as part of the Center for Hybrid Organic Inorganic Semiconductors for Energy (CHOISE), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences (BES). E.T.B. acknowledges a graduate fellowship through the Duke University Department of Electrical and Computer Engineering.

The data that support the findings of this study are available from the corresponding author upon reasonable request.