Direct Laser Writing of δ- to α-Phase Transformation in Formamidinium Lead Iodide.

Organolead halide perovskites are increasingly considered for applications well beyond photovoltaics, for example, as the active regions within photonic devices. Herein, we report the direct laser writing (DLW: 458 nm cw-laser) of the formamidinium lead iodide (FAPbI3) yellow δ-phase into its high-temperature luminescent black α-phase, a remarkably easy and scalable approach that takes advantage of the material's susceptibility to transition under ambient conditions. Through the DLW of α-FAPbI3 tracks on δ-FAPbI3 single-crystal surfaces, the controlled and rapid microfabrication of highly luminescent structures exhibiting long-term phase stability is detailed, offering an avenue toward the prototyping of complex perovskite-based optical devices. The dynamics and kinetics of laser-induced δ- to α-phase transformations are investigated in situ by Raman microprobe analysis, as a function of irradiation power, time, temperature, and atmospheric conditions, revealing an interesting connection between oxygen intercalation at the surface and the δ- to α-phase transformation dynamics, an insight that will find application within the wider context of FAPbI3 thermal phase relations.

The recent advent of lead halide perovskites—compounds based on a general APbX3 formula, where A denotes an organic cation and X = I, Br, Cl—defined a historic turning point in solar energy and semiconductor materials science.[1,2] The key characteristics of perovskites that make these materials excellent light harvesters have sparked interest in their application elsewhere in the field of optoelectronics and, specifically, in photonic devices.[3−11] Even polycrystalline lead halide perovskites show a very well-defined tunable[12] band gap that is typically observed only in near-perfect crystalline materials, such as GaAs and other members of the tetrahedral III–V family.[13] Thus, the appeal for synthetic perovskite-based optoelectronic devices stems from their ability to combine the high performance of crystalline semiconductors with the ease of solution processing, all at a low cost.[14]

While the methylammonum (MA; CH3NH3+) lead iodide provskite (MAPbI3) has thus far enjoyed the preponderance of research interest in the field, recent developments are shifting attention toward formamidinium (FA; HC(NH2)2+) lead iodide (FAPbI3) and, particularly, mixtures incorporating dilute amounts of MA.[15−18] Compared to the MAPbI3 system, FAPbI3 offers several notable advantages for its application within optoelectronic devices, from lower hysteresis during current–voltage measurements,[19] superior photo and thermal stability,[12,17,20,21] and longer carrier lifetime (484 ns) to the roughly 1 order of magnitude gain in conductivity (1.1 × 10–7 (Ωcm)−1).[22] Together, these improvements provide great promise for the production of device-quality FAPbI3-based materials.

However, phase stability issues plague FAPbI3 perovkites. Of the two main FAPbI3 polymorphs, by far the most interesting for optoelectronic applications is the high-temperature black perovskite α-phase (trigonal) because of its appealing optical and electronic properties,[12,22] over its low-temperature yellow nonperovskite δ-phase (hexagonal) counterpart. While thermal transformations of FAPbI3 are relatively simple and easily controlled, synthesis of the desirable α-FAPbI3 and its stabilization at room temperature are not straightforward.[16,23,24] Attempts to better understand phase transformation relations among the δ and α polymorphs are currently under way;[23,25−27] however, a comprehensive model of the transient stages of phase stability is yet to fully explain the structural diversity exhibited by this material.[28−30] The scattered formation of dark α-phase nucleation centers within a yellow δ-FAPbI3 crystal to the coalescence of a thermodynamically stable dark perovskite represents a scientifically challenging and technologically important terra incognita. In most studies of perovskite phase relations to date, excluding the recent in situ transmission electron microscopy work by Divitini etal.,[31] the only real-time in situ structural diagnostics have been through X-ray diffraction (XRD) measurements. Usually, XRD alone is not versatile enough to yield data suitable for process control or elucidating transformation dynamics.

Prolonged exposure to light has been shown to enact a variety of changes in perovskite materials,[32−38] not all of them bad. For example, under the right light exposure,[39] significantly enhanced photoluminescence and carrier lifetimes can be achieved. Therefore, following synthesis, light treatments tailored to improve and modify material properties offer a highly convenient and versatile postprocessing route. Considering the fact that the complicated phase properties of FAPbI3-based perovskites remain firmly in the spotlight, it is somewhat surprising that irradiation-induced phase restructuring is relatively unexplored.[25]

Herein, the palette of functional design parameters available to material scientists and engineers utilizing FAPbI3-based perovskites is significantly extended, as we report the direct laser writing (DLW) of localized and stable δ- to α-phase transformations, using 458 nm continuous-wave (cw)-laser light. Through the powerful innovative coupling of in situ Raman scattering experiments within our laser-driven approach, the transformation dynamics and kinetics are revealed, for enhanced procedural control, shedding light on how to exact the phase restructuring for micrometer fabrication. In terms of engineering functional elements for optoelectronic applications, the ability to switch the phase of FAPbI3 at the micrometer-level builds on several other perovskite microprocessing methods available (DLW synthesis,[40] mechanical micropatterning,[41] and laser-induced microstructuring[38,42]) and provides a promising avenue for the controlled preparation of FAPbI3 materials with specific crystal structure.

Results and Discussion

Structural, Vibrational, and Optical Properties of Pure δ- and α-Phase FAPbI3

We begin by overviewing the characteristic properties of polymorphic FAPbI3. The room-temperature stable yellow δ-FAPbI3 is known to undergo a reversible phase transition above 150 °C,[43] rearranging into its desired black α-FAPbI3 structure (see inset of Figure to visualize such a change) before turning back to its yellow phase following several days of storage under ambient conditions.[43]Figure a displays powdered XRD scans recorded from both the δ- and α-phase of FAPbI3, with their diffraction peaks aligning with hexagonal (P63mc space group) and trigonal (P3m1 space group) structures, respectively. XRD has thus far been widely used to identify the structural nature of FAPbI3,[25,28,43] with Raman spectroscopy far less utilized[22] as a structurally sensitive probe. As will be seen later, a high tendency for FAPbI3 to experience oxygen intercalation at the surface[44] means accurate Raman scattering characterization can be achieved only by employing relatively low excitation power densities (Iexc) and short exposure and integration times.[45]

Figure 1

Physical properties of polymorphic FAPbI3. Characterization of the pure α-phase (top traces) and δ-phase (bottom traces) of FAPbI3 by means of (a) powdered XRD, (b) Raman scattering (λexc = 458 nm; Iexc < 5 × 102 W·cm–2), and (c) PL (λexc = 532 nm). The inset here shows a typical FAPbI3 single crystal in its yellow phase (bottom) before transforming into its black phase (top) following a 10 min 160 °C thermal treatment.

Figure b presents Raman scattering spectra recorded from pure δ- and α-FAPbI3 single crystals using a excitation wavelength of 458 nm and low excitation intensity; for accurate Raman measurements, “low” will be considered Iexc < 0.1 kW·cm–2. The δ-phase in Figure b exhibits a strong Raman mode centered near 107 cm–1 with a weak low-energy shoulder extending down to roughly 70 cm–1 and a featureless high-energy background. The energy of this shouldering band is reminiscent of the PbI2 phonon density of states (see, for example, Figure b), suggesting it originates from I–Pb–I bending and stretching.[35,46] In fact, preliminary measurements indicate the intensity of this shoulder to be connected to the shade darkness of the material’s yellow color. On the other hand, the Raman spectrum of α-FAPbI3 possesses two well-separated vibrational signatures: a prominent low-energy mode at 137 cm–1, which is blue-shifted and comparably narrow (∼14 cm–1) relative to the δ-phase, and another much broader and weaker Raman band located near 525 cm–1. The strong blue-shift in the former vibration likely reflects the large change in lattice constant, and, given its relatively high energy, the latter feature is expected to arise from the in-plane bending of FA cations[47] (δ(H2N–C–NH2)) within the framework. The in-plane bending of FA cations similarly contributes to the Raman scattering spectrum of the cubic perovskite phase of FAPbBr3 (521 cm–1),[48] and since this band arises purely from the trapped organics, it should act as a sensitive marker for the orientational disorder of α-FAPbI3 single crystals.[46]

Figure 3

Oxygen-related changes in the Raman scattering spectrum. Raman spectra recoded from δ-FAPbI3 (a) while exposed to a very low IDLW of 10 W·cm–2 for an extended period of time and (b) before (bottom trace) and after laser heating (IDLW = 500 W·cm–2) for 10 min under ambient (air; 21% oxygen), pure oxygen (O2), and pure nitrogen (N2) atmospheres. Note that spectra in (b) were recorded using Iexc = 50 W·cm–2 and, for direct spectral comparison, the spectrum of a pure PbI2 crystal is also presented. For clarity both plots display offset-normalized spectra (normalization factors given in (a)) with spectral signatures of various vibrational species labeled.

Figure c displays the steady-state photoluminescence (PL) spectrum of our δ- and α-phase FAPbI3 single crystals over 615–880 nm, recorded using 532 nm excitation and low Iexc. For the low-temperature δ-phase, no emission is observed over this photonic range. The α-FAPbI3 single crystals, on the other hand, exhibit a strong singular emission centered near 820 nm, in sound agreement with other studies.[22,39] The comparatively long wavelength of the emission suggests that our single crystals are relatively free of structural defects or grain boundaries, which would otherwise act to increase the optical band gap.[22]

Laser-Induced Transformation Dynamics

Next, we explore the use of intense 458 nm laser light to locally heat and structurally modify FAPbI3 single-crystal surfaces. An excitation wavelength of 458 nm was chosen because it provides a relatively small DLW footprint (compared to longer visible laser wavelengths) and resides on an optical absorption feature exhibited by δ-FAPbI3 (see Figure S1 in the Supporting Information), enhancing the heat transfer efficiency.

To examine the systematic effect of exposing α-FAPbI3 single crystals to relatively high DLW optical power densities (IDLW), Raman spectra were acquired using a low Iexc following a 25 s exposure to increasing IDLW values, under an ambient atmosphere. Selected Raman spectra recorded at different stages of a rising IDLW are presented in Figure a, while the analytical fitting of many measurements made during a systematic increase in IDLW provides the power dependence data (of indicated modes) in Figure b. For completeness, a clear discussion on how the Raman spectra are fitted and analyzed throughout this work (including an example) can be found in the Supporting Information. Examining the full range of the IDLW values presented in Figure , a relatively complex spectral evolution is exhibited by δ-FAPbI3 when exposed to intense laser light. Generally, the surface modifications depicted in Figure b track the following sequence with a rising IDLW: (i) oxygen intercalation into the FAPbI3 crystal surface (IPb–O), (ii) a δ- to α-phase transformation (Iα), (iii) the near total degradation of existing Raman signals, and (iv) the rapid growth of a lead monoxide Raman line shape (IPbO) at relatively high IDLW. Note that the details of this sequence will come later.

Figure 2

Power dependence of laser-induced surface modifications. (a) Raman spectra recorded using low IDLW from the α-FAPbI3 crystal surface after 25 s of exposure to the increasing laser powers indicated. These data have been normalized and offset, with the vertical lines indicating spectral mode assignment, which is discussed in the text. (b) Laser power density dependence data of the measured Raman signals identified in (a). The lines fitting these data provide a guide for the eye, and the SEM image in the inset shows the extent of the α-FAPbI3 crystal surface damage following exposure to the highest IDLW value studied.

The inset in Figure b shows a SEM image of the crystal surface after high irradiance exposure (∼2 kW·cm–2), revealing localized degradation to the α-FAPbI3 surface. For relatively large IDLW values, severe laser-induced crystal damage is ultimately realized, with the Raman signals of polymorphic FAPbI3 species fading into the formation of a substantial lead oxide scale. Near 1.3 kW·cm–2 it is not that the entire Raman signal is lost (as signals near the zero baseline), but it is under transition with several species coexisting within the Raman microprobe at once. The moment that lead monoxide modes first appear in the spectra it is assumed that the status of the sample is moving toward that of significant degradation, visible in its morphology changes (inset of Figure b). We provide in Figure S3 of the Supporting Information a comparison of the Raman spectrum recorded from a laser-damaged α-FAPbI3 surface with that measured from pure PbO powder (99.999%, Sigma-Aldrich), under identical conditions. The observed vibrational energies and fwhm of the PbO-like modes in Figure a are far lower and higher, respectively, compared to the Raman spectra recorded from pure lead monoxide. Moreover, no notable change is observed in the relative oscillator strength when comparing modes within the same DLW-derived PbO spectrum. These differences suggest a lower degree of structural quality and/or the formation of small ablated PbO crystallites, leading to “phonon confinement”.[49]

The Role of Free Oxygen

Hereon we will focus on the interesting modifications arising from exposure to lower irradiances and specifically those that foster the formation of α-FAPbI3 crystal. For IDLW incident with a power of 0.2 kW·cm–2, modes at 160 and 240 cm–1 are the first to be introduced into the Raman line shape of Figure a. To study this first structural change, Raman spectra were recorded in situ every 5 min using an extremely low Iexc value (∼10 W·cm–2) for a total of 17 h. The Raman characterization results of this experiment can be found in Figure a. Interestingly, the same new bands at 160 and 240 cm–1 arise in the Raman line shape after a 3 h exposure, paralleled by a degradation to δ-FAPbI3 Raman signals. Following much longer exposure times (>6 h) within an ambient atmosphere, these modes intensify, before the system equilibrates and the changes significantly slow. We note that the normalizing factors used to scale these data grow over time, reflecting a drop in the α-FAPbI3 Raman cross-section. This feature is in fact also contained in Figure b, where Iδ decreases rather quickly with increasing laser powers, relative to the rise of other laser-induced species. In explaining this, we suspect the weakening of δ-phase Raman signals under light exposure is driven in multiple parts: (i) oxygen intercalating into the perovskite (forming Pb–O), reducing the number of Pb–I oscillators contributing δ-FAPbI3 signals, (ii) light-induced surface restructuring,[32−35] and (iii) temperature changes, which act to shift both resonant Raman effects (via perturbing the electronic structure) and the Stokes scattering rate (i.e., phonon populations).

To provide further data points on the role of oxygen, similar measurements to Figure were recorded using a strongly transformative power density (IDLW ≈ 0.9 kW·cm–2) under atmospheres ranging from pure oxygen to anoxic (pure nitrogen); see Figure b. Exposed to pure N2 gas, the vibrations at 160 and 240 cm–1 never appear. Conversely, within a pure oxygen atmosphere, these vibrations re-emerge and they are confirmed to be oxygen-related. Both the vibrational energies and relative intensities of the vibrations at 160 and 240 cm–1 are indicative of Pb–O bands within the Raman microprobe, due to oxygen intercalation at the surface.[37] This is not unprecedented, as parallels can be directly drawn here to the MAPbI3 system, which suffers from a strong susceptibility to oxygen intercalation.[44] Moreover, this effect is optically driven,[37] explaining the response of δ-FAPbI3 to long exposures of low-power 458 nm laser light (see Figure a). We assign these modes to oxygen intercalation at the δ-FAPbI3 surface (IPb–O) and point out that they are in fact inadvertently contained within the δ-FAPbI3 Raman spectrum reported by Han etal.,[22] highlighting the strong propensity for this phenomenon to occur.

Again we examine the role of oxygen by acquiring low-power Raman spectra ex situ for a series of δ-FAPbI3 single crystals heated from 110 °C up to the phase transition temperature[43] at 150 °C. Surprisingly, the introduction of Pb–O vibrations at the surface is not limited only to an optically driven process, as these peaks also appear within the Raman line shape of materials heated as low as 110 °C FAPbI3 (see Figure S4 of the Supporting Information). As well, α-FAPbI3 phonon modes are introduced into the scattering volume as low as 110 °C. This is in comparison with powdered XRD measurements, which did not begin to reveal the emergence of extremely weak α-phase Bragg peaks until reaching an annealing temperature of 130 °C (see Figure S4 of the Supporting Information). In unravelling this discrepancy the kinds of information the two techniques yield must be considered; Raman scattering intrinsically probes surface structure, rather than the overall average, like in XRD. This result suggests that when exposed to temperatures well below 150 °C, oxygen finds its way into the δ-FAPbI3 crystal surface and is paralleled by the thermally premature transformation of α-FAPbI3 near the surface. We assume here that oxygen migration is surface limited and that the subsurface crystal is unaffected in this manner, explaining the XRD results.

DLW of δ- to α-Phase Transformation in FAPbI3 Single Crystals

For Raman spectra recorded following exposure to IDLW = 0.41 kW·cm–2 and slightly higher in Figure a, a δ- to α-phase transformation is clearly observed, manifested via the steady growth of α-like modes at 137 and 525 cm–1 up to a maximum relative intensity. Given that 458 nm argon-ion laser light is used for excitation, the photolytically activated influences (photon-induced chemical decomposition) are assumed negligible (DLW photon energy = 2.7 eV) compared to the thermally driven process (radiative transfer). For the specific case δ-FAPbI3 irradiated by relatively low intensity 458 nm laser light (see, for example, the long exposure data contained in Figure a), Raman measurements demonstrated no evidence of α-FAPbI3 synthesis following long exposures; only once temperatures in the laser-heated volume increase will a δ- to α-phase transformation occur.

The relative weighting of the α-phase Raman bands exhibited at its height in Figure is archetypal of repeated experiments, covering a variety of α-FAPbI3 single crystals and IDLW values, a feature likely defined by the optical penetration of 458 nm laser light incident on the (δ/α-)FAPbI3 surface[50,51] and the transformation kinetics of the laser-driven phase transformation. When comparing the pure δ- and α-phase Raman signals measured under identical experimental conditions, the Raman scattering cross-section (λexc = 458 nm) of the δ-phase is found to be over an order of magnitude larger than that of the α-phase, a resonant Raman effect likely incited by our choice to position the excitation wavelength on a prominent absorption feature. Thus, it follows that there must exist a substantial α-phase deposit within the total scattering volume to account for the relative intensity of δ- and α-phase Raman signals. Determining the exact transformation depth here is difficult; however, depth profiling by near-IR (NIR) fluorescence microspectroscopy using a high NA oil objective (1.4) reveals it to be less than the diffraction-limited probe depth of 500 nm.

As already mentioned, δ-FAPbI3 is susceptible to oxygen intercalation and α-phase modifications at the surface well below 150 °C. Their appearance in Figure confirms that there are two induced effects: oxidation intercalation at the δ-FAPbI3 surface (using ambient oxygen) followed by the restructuring into the α-FAPbI3 perovskite. This is highlighted in Figure b, where the introduction of IPb–O signals preludes the formations of α-FAPbI3 material, a seemingly fixed sequence across many experiments.

It is conceivable that the prior formation and growth of Pb–O bonds and the δ- to α-phase transformation are actually connected, with the energy barrier for the phase shift being lower in the presence of intercalated oxygen. For example, exposure to air during the fabrication of the MAPbI3 system was recently shown to lead to different restructuring pathways.[31] Within the inert anoxic (N2) atmosphere, there is little option for which FAPbI3 can restructure, and the material ultimately decomposes to PbI2 (see Figure b), akin to excessive thermal treatments.[20,52] Further, no IPb–O vibrations appear and α-FAPbI3 is never synthesized. On the other hand, under pure oxygen, there is an enhancement in the relative α-phase Raman signal, indicating that these conditions in fact foster more α-FAPbI3 synthesis compared to an ambient environment (∼21% oxygen). Developing a comprehensive physical model to aid interpretation is well outside the scope of the present study. However, these findings support a strongly plausible argument, one that points toward free oxygen atoms playing an influential role in FAPbI3 phase transformation dynamics. This aspect of (δ/α-)FAPbI3 phase transformation has so far been overlooked and is particularly significant. Moreover, it is also intrinsic to the transformation kinetics and will be revisited later.

Moisture effects—in the form of ambient humidity—are yet to be considered, with their influence on organic cations or perovskite mostly related to degradation processes by facilitating chemical reactions.[53] For our experiments, along with sample storage, a controlled ambient environment was implemented with a temperature of 20.5 to 21.5 °C and a relative humidity of 47% to 50%. In the case of FAPbI3, moisture can enhance photoluminescence;[39] however, the influence of moisture on FAPbI3 phase transitions is relatively unexplored. The high susceptibility to transform phase under ambient conditions is likely connected to an altered surface energy caused by oxygen incorporation, and recent density functional theory (DFT) calculations[54] have suggested that moisture too can assist in the surface rearrangement of the analogous MAPbI3 system. It is thus strongly possible that moisture is further playing a role here by reducing the phase transition barrier.

The stability of the α-FAPbI3 material transformed by DLW was investigated over time. Raman spectra were recorded before and after exposure to high irradiance, and, in both cases, an extremely low power density was employed. The absence of α-phase vibrational signatures before irradiation and their presence more than 2 weeks after storage under ambient conditions reveals a relatively stable DLW synthesis of α-FAPbI3. Furthermore, no major decrease in the relative Raman scattering intensities of these peaks is observed. Thus, DLW of α-FAPbI3 is not only demonstrated but also shown to be stable at room temperature. Note that our observations of anomalous long-term stability for the DLW α-FAPbI3 will be revisited in detail later.

DLW δ- to α-Phase Transformation Kinetics

A δ- to α-phase FAPbI3 transformation under an ambient atmosphere was investigated at room temperature on a α-FAPbI3 single crystal using a constant 0.5 kW·cm–2 laser power density and acquiring spectra once a minute. On the basis of preliminary measurements, this laser power was selected as it allowed the DLW phase transformation to occur on a well-resolved time scale. Figure a shows selected Raman spectra obtained in situ at different times during the exposure of laser light on the α-FAPbI3 surface, and the inset summarizes the temporal dependence of the modes indicated (chosen here for their relative spectral weight). The assignment of phonon signatures here receives the same treatment as in Figure a, as the process is initiated by the decrease in δ-phase signal and the early introduction of IPb–O modes, followed shortly after by the rapid growth of the two α-FAPbI3 vibrational bands. Up until 5 min (indicated by t′), the formation of Pb–O bonds within the Raman microprobe is steady and relatively quick, before slowing and rapidly equilibrating (see inset of Figure a). Conversely, for longer exposure times, the intensity of the α-phase bands continues to asymptotically approach their maximum limit, and by near 100 min temporal changes in the Raman spectra significantly slow.

Figure 4

In situ temporal Raman microprobe analysis of transformation kinetics. (a) Raman scattering spectra acquired in situ at different times (t) of a δ- to α-phase transformation using a laser power density of 500 W·cm–2. Spectra have been offset and normalized for clarity. The inset displays the temporal integrated intensities of selected modes from fitting all spectra measured during the 100 min DLW procedure, with the vertical line labeled t′ here indicating the frame at which the IPb–O modes significantly slow, corresponding to the same frame similarly indicated in the Avrami plot in (b). (b) Degree of laser-induced δ- to α-phase transformation obtained from normalized Raman intensity vs time t. The data fit is made using the Avrami equation (eq S1; see Supporting Information), and the inset shows the corresponding Avrami plot (ln[−ln(1 – X(t))] vs ln(t)) with its linear fit.

Modeling of DLW Phase Transformation Kinetics

Because the Raman scattering cross-section is proportional to the total scattering volume, the evolving Raman peaks allow a direct measure of the transformation degree in real time and can be used to analyze the DLW procedure. Therefore, we replot the measured temporal integrated intensity of the α-FAPbI3-like phonon at 137 cm–1 in Figure b for detailed treatment. Our data exhibit an initially rapid reaction rate, which quickly slows, a behavior indicative of solid-state phase transformation kinetics (crystallization). The universal method to describe nucleation to coalescence of solids is the Johnson–Mehl–Avrami–Kolmogorov (JMAK) theory[55−57] (see Supporting Information for details). A central parameter in the theory is the Avrami exponent n, which acquires an integer or half-integer value in the range 0.5 to 4, depending on the dimensionality of the transformation and is defined as the local slope in a double-logarithmic Avrami plot: ln[−ln(1 – X(t))] vs ln(t). The corresponding Avrami plot and measured exponent are contained in the inset of Figure b, where below t′ = 5 min the Avrami plot reveals an initially large and time-dependent Avrami exponent. A large n value here indicates a healthy three-dimensional transformation rate that is experiencing a strong shift in dynamics. Beyond t′ = 5 min, however, a constant Avrami exponent of roughly unity emerges and signifies a low-dimensional transformation to unfold (i.e., limited to the upper surface). The low dimensionality here is likely mediated by both the attenuation of oxygen and low heat transfer into the subsurface; the rotational degrees of freedom within the “caged” organic sublattice of analogous MAPbI3 perovskites have been shown to manifest as an ultralow thermal conductance (0.5 W/(K·m)) for the whole system.[58] Interestingly, the onset of the Avrami plot data becoming linear coincides within the same measured time interval (i.e., spectral acquisition frame) as the sudden halt in IPb–O growth displayed in Figure a (identified also by a broken vertical line). The synchronicity of these two events must be explicated; before t′ both IPb–O and the value of n evolve with time, and following t′, the stoppage of IPb–O growth coincides with the α-FAPbI3 transformation adhering to regular crystallization theory: a time-independent Avrami exponent. Again, this provides compelling evidence for the strong influence that oxygen intercalation has on the phase relations of polymorphic (δ/α-)FAPbI.

DLW Micropatterning of Luminescent α-FAPbI3 Tracks on δ-FAPbI3 Single Crystals

The in situ Raman microprobe analysis proved extremely useful for understanding the δ- to α-phase transformation dynamics and kinetics, yielding powerful real-time diagnostics and elucidating parameters that define DLW control. It is within this context we move onto the rapid and controlled DLW of highly luminescent α-FAPbI3 phase microtracks onto δ-FAPbI3 single crystal and nanocrystalline thin film surfaces. Because laser-induced δ- to α-phase transformations occur readily under ambient conditions, we aim to simplify the DLW process as much as possible and directly expose an α-FAPbI3 single-crystal surface to focused 458 nm cw-laser light using air objectives and an ambient atmosphere, i.e., no special sample mounting apparatus. To minimize heat dispersion, cw-laser writing with the lowest dwell time (write speed = 10 mm/s) and relatively aggressive laser power was found to yield the best results, i.e., clean and narrow DLW lines with little production of other (luminescent) byproducts.

An overview of the key optical features of DLW surface micropatterns can be found in Figure . Figure a displays an optical micrograph of the DLW α-channel patterning: a 60 μm long line array periodically cycling every 10 μm. Here the characteristic yellow color of the δ-FAPbI3 crystal transmission image is interrupted by the dark lines of the DLW α-phase channels. The inset of Figure a displays an scanning electron microscopy (SEM) micrograph of the typical recessed/grooved contour of the DLW tracks, complete with a slightly raised lip at the ridge. Figure b provides a micro-PL map of the selected area in Figure a, with the analysis of the steady-state PL peak mapping contained in Figure c. Two emissions are seen to derive from our DLW procedure: (i) a sharp, intense α-FAPbI3 NIR peak near 780 nm, yielding the bright red tracks in the PL map shown in Figure b, and (ii) a weaker and much broader emission peak around 660 cm, contributing to the green pixels within the top DLW line. Due to its strong similarity to the steady-state PL spectra recorded from pure PbO powder (purchased from Sigma-Aldrich; see Figure S5 of the Supporting Information), we assign the 660 nm emission to the ablation of PbO particles ejected from the DLW area. Depending on IDLW parameters used the amount of ablated PbO debris produced on either side of the DLW track will differ a great deal, with a tunable landscape of resulting fluorescence morphologies existing between low and excessive IDLW values (as interpreted by PL mapping; see Figure S6 in the Supporting Information). The dual red line appearance in Figure b is attributed to mapping the emission from a grooved surface using confocal optics.

Figure 5

Optical properties of DLW micropatterning. (a) Optical micrograph of DLW patterning on a δ-FAPbI3 surface written with a 0.75 NA at 0.8 kW·cm–2. The inset shows a SEM image of the typical grooved microtrack created on the surface of the material following laser light exposure. (b) Micro-PL map of area identified in (a), with (c) PL spectral data showing the peaks that correspond to the dual color map defined in (b). Note the relative color intensity represented for the two measured emissions in (b) (i.e., red and green) has been normalized. The inset in (c) displays several normalized peaks recorded near 780 nm and highlights the variance of the NIR emission over the DLW area. (d) NIR fluorescence micrograph of the DLW pattern, with the inset comparing images rerecorded from the identified area several weeks after fabrication. (e) Normalized PL intensity of the pattern presented in (d) over the 4 weeks following microfabrication. Note the PL data in (d) and (e) were recorded over the spectral range 780 to 800 nm (see Figure S7 of the Supporting Information), and the sample was stored under ambient conditions.

The NIR emission measured from the DLW tracks is centered at ∼780 nm, consistently blue-shifted by approximately 40 nm compared to that of α-FAPbI3 single crystals (820 nm). Such a fixed PL peak energy has been shown experimentally to arise in a mixed-phase FAPbI3 system[28] as a consequence of band bending at the δ/α junction, via Fermi level pinning. Nonetheless, some variance is observed across all PL mapping experiments and is highlighted in the inset of Figure c. This is likely introduced as a result of structural defects or even grain boundaries forged during the rapid and aggressive (i.e., use of relatively high IDLW) fabrication.[22] Moreover, it is known that the properties of low-dimensional perovskites are further perturbed, compared to their bulk counterparts.[59]

Phase Stability of DLW δ- to α-Phase Transformed Material

The stability of the NIR fluorescence emitted from DLW arrays was inspected over time, with the findings presented in Figure d and e. The spectral detection range used to construct the fluorescence micrographs is shown in Figure S7 of the Supporting Information. In line with the results of the Raman characterization, fluorescence microscopy reveals a stable synthesis of α-FAPbI3, with bright clean lines measured repeatedly over 4 weeks. Notably, only a slight reduction in the fluorescence intensity is seen over this time, when compared to day 0 (see Figure d). In quantifying the temporal decay in NIR emission coming from the DLW micropattern, Figure e shows the emission intensity to reduce by roughly 30% over 4 weeks. Further, the accompanying emission spectra acquired during these experiments exhibit no change over this period (see Figure S7 of the Supporting Information). Again, we emphasize that this is an unexpected and stunning result, in that only pure FAPbI3 materials are present here, i.e., no mixed cations[21,60,61] (MA/FA or Cs/FA).

The fundamental mechanism by which the DLW α-phase is preserved here is yet to be discussed, and the role of oxygen intercalation at the crystal surface during the DLW procedure cannot be overlooked. To evaluate the influence of oxygen on the α-phase stability, a series of dark α-FAPbI3 single crystals were exposed to ozone and observed over time (experimental details and results can be found in Figure S8 of the Supporting Information). In every case, the ozone treatment had a dramatic destabilizing effect on the black α-phase of FAPbI3, as samples exposed to ozone exhibited a vivid premature yellowing, compared to their dark control counterparts. Therefore, while oxygen alone does appear to influence the δ- to α-phase transformation in FAPbI3, it does not stabilize the dark metastable perovskite phase, in fact, quite the opposite. An explanation for the stable NIR emission coming from our DLW pattern likely resides in the local microstructure of the system; studies of α/δ phase mixing and junctions in FAPbI3 recently[28] indicated that such microstructuring imposes strong stabilizing effects. As well, no matter the degree of phase mixing, mixed-phase FAPbI3 materials were shown to be concomitant with PL emission centered near 780 nm. That the same hypsochromic 780 nm emission (see Figure c) is seen in our DLW α-FAPbI3 material strongly suggests a similar mixed-phase configuration exists to account for its anomalous long-term stability.

Besides the obvious advantage of gaining enhanced structural stability for the α-phase material embedded in the δ-phase crystal, δ/α mixed-phase FAPbI3 systems were also recently shown to benefit from other unique physical properties that differ from either of the pure phases. For example, the mixed phase system gains a significant enhancement in NIR emissions near 780 nm, a lasting resistance to humidity, and an increased PL quantum yield.[28] In fact, when comparing the average PL intensity measured from our DLW α-FAPbI3 material with that recorded from the thermally annealed α-FAPbI3 single-crystal surface using the same acquisition conditions, the α-FAPbI3 DLW microline emission is around an order of magnitude larger. Collectively, these distinct physical properties add to an interesting microstructure worth intense investigation.

Microline Array Resolution and Precision

Through refining the DLW parameters explored thus far, we examine the spatial limitations of our DLW procedure in Figure . Figure a presents an optical transmission image acquired for a DLW line array using a 488 nm laser light, which also served as the excitation for the corresponding NIR fluorescence image in Figure b. Close inspection of Figure b reveals the fluorescence to overlay nicely with the transmission image in Figure a, with the fabricated structures exhibiting bright NIR emissions when compared to the regions that were unexposed to the writing beam. The variability in fluorescence intensities and line shape here is attributed to small localized topological differences across the scanned surface, resulting in variations in the DLW conditions (i.e., defocused DLW spot).

Figure 6

DLW cycling and resolution. (a) Bright-field microscopy image of a DLW line array drawn on a δ-FAPbI3 single-crystal surface using IDLW = 1.1 kW·cm–2 and a 10×, 0.4 NA objective. (b) Corresponding fluorescence image of the same area (λexc = 488 nm) with (c) fluorescence intensity profile determined across the whole fluorescence image in (b). (d) Comparison of the fluorescent fwhm measured from patterns written using different DLW optical spot sizes (measured as the in-focus diffraction-limited beam diameter; see Materials and Methods for details), as determined via line scans akin to that measured in (c). The inset here presents examples of the normalized intensity profiles, with the arrow indicating the employment of an increasing NA value. (e) Corresponding optical and fluorescent micrographs of a DLW pattern drawn on a δ-FAPbI3 thin film surface using IDLW = 3 kW·cm–2 and a 10×, 0.4 NA objective. The inset shows an SEM image of the DLW microtrack.

Figure c presents the intensity line scan of the array depicted in Figure b, highlighting the sharp periodic cycling of NIR fluoresce emanating from the DLW α-FAPbI3 microline array. The cross-sectional profile of the fluorescence intensity appears to be Gaussian-like, an aspect common across repeated experiments with varying DLW parameters (IDLW, spot size, etc.). If follows that the fwhm of this line shape may be used as a metric to assess the possible resolution of the DLW lines. Thus, to demonstrate the DLW cycling tunability of our fabricated structures, a range of α-FAPbI3 line arrays were drawn onto the δ-FAPbI3 single-crystal surfaces using optimized DLW parameters for three different air objectives (magnification/NA): (i) 10×/0.4, (ii) 20×/0.75, and (iii) 40×/0.9. An analysis similar to that presented in Figure b and c provides the basis of the microline fwhm data shown in Figure d. The inset in Figure d directly compares typical normalized line profiles recorded implementing different NA values. For DLW performed using the smallest optical spot size (NA = 0.9), a fluorescent line with a fwhm as low as ∼950 nm is achieved. Moreover, a well-defined trend for the dependence of the fluorescence fwhm on the NA of the objective employed emerges when assessing the fwhm across the entire NA range measured (see Figure d). This verifies a high level of control over patterning of different sized luminescent structures and indicates the approach detailed to scale effectively.

Examining respectively the optical transmission image and corresponding PL map in Figure a and b, it is clear that there is room left for refining the precision of our DLW procedure. The level of precision exhibited here stems from the fact that we have adapted a Raman microscope to perform the task of DLW. Without autotracking and correcting for topological features,[62,63] variations in surface height will act to defocus the DLW writing probe and, thus, lower the patterning precision over longer distances. Direct laser writing is of course an entire field in itself, with well-established techniques for dealing with such problems. However, there are two sides to consider when addressing topological influences: (i) employing Z-tracking to correct for topological height changes (ΔZ) across the DLW surface and (ii) improving the long-range flatness (over distances of 100s m) of the DLW surface. Implementing a system capable of tracking and correcting for changes in Z has already been covered, but addressing the latter will ultimately be of great benefit for the DLW precision, as well as offering a materials synthesis-based solution. Strong motivation for this stems from the fact that achieving single-crystal facets with sufficiently flat surfaces is difficult, as they often exhibit smooth, uneven angulating features (see Figure S9 of the Supporting Information for a vivid depiction of said features). This morphological aspect is seemingly unavoidable and likely arises due to soaking the fresh as-grown crystals in acetonitrile immediately after growth (see Materials and Methods section).

Thus, for improved DLW α-phase micropatterning, we control the flatness of the writing surface over relatively long distances (10’s of mm) by using spin-coated nanocrystal films. DLW experiments were performed using a 10×, 0.4 NA objective and IDLW = 3 kW·cm–2, on a 400–500-nm-thick δ-FAPbI3 nanocrystal thin film. The results of performing DLW on FAPbI3 thin films are presented in Figure e, where corresponding optical and fluorescence images of narrow α-FAPbI3 lines reaching 1 mm long are shown. The resulting fluorescence image demonstrates a significant enhancement to the DLW precision and overall aesthetics, with bright, continuous, and nicely confined 1-mm-long lines realized. Again, a similar DLW line morphology is revealed here via the SEM imaging shown here in the inset. In DLW of α-phase patterns on thin films, rather than single crystals, the local flatness (in the range of 100’s of nm) is ultimately reduced, as it is no longer a bulk flat surface, but rather a film of fused nanocrystals (see Figure S9 of the Supporting Information to visualize the film morphology). Importantly, however, the level of precision for long-range DLW patterning is substantially enhanced, without the need for Z-tracking, as demonstrated in Figure e. As well, use of thin films with thicknesses of 400 to 500 nm provides a limit in the Z direction for which the DLW patterns can form. As a result, the optical absorbing volume is reduced and the required writing laser power to induce the δ- to α-phase phase transformation is in fact increased (by a factor of 3 to 4 for the patterns drawn in Figure d), relative to that used for the single-crystal materials. The refined patterning shown here for DLW on flat δ-FAPbI3 thin films provides a promising avenue for applications requiring high precision, while also being able to lower the transformation depth to an in-plane α-FAPbI3 microstructure, via controlling the film thickness.

Conclusion

We have demonstrated an all-optical method for inducing localized δ- to α-phase transformations in δ-FAPbI3. A detailed in situ study of the laser-driven transformation dynamics and kinetics was achieved through Raman microprobe analysis and revealed the δ- to α-phase transformation dynamics to be strongly influenced by the availability of free oxygen. Through refining the DLW parameters, the rapid writing (10 mm/s) of α-FAPbI3 microtracks atop δ-FAPbI3 single crystals was shown possible. The method employed was extremely easy, using standard visible laser light (458 nm), air objectives, and ordinary mounting, being assisted by the fact that the DLW of α-FAPbI3 is promoted in an atmosphere possessing free oxygen. Refining of the writing parameters provided fine control over the DLW patterning dimensions, and highly fluorescent α-FAPbI3 line arrays with NIR emissions peaking near 780 nm and micrometer-sized spatial fwhm were realized. The α-FAPbI3 patterning exhibited anomalous long-term structural stability for at least 4 weeks, likely as the result of synthesizing a δ/α mixed-phase material. To the best of our knowledge, this is the first report on the laser-based fabrication of δ- to α-phase-transformed material in FAPbI3, let alone outlining a clear path for refining control over the DLW process for its potential applications. The ability to locally modify the structural, electrical, and optical properties of a wide range of optoelectronic materials is the focus of much current research. Within this context, local laser irradiation for engineering of a FAPbI3 phase and physical properties, bypassing conventional indiscriminate bulk thermal treatments, is important.

Materials and Methods

Chemicals and Reagents

Lead iodide (PbI2) (99.999% trace metal basis), formamidinium iodide (FAI), dimethylformamide (DMF), and γ-butyrolactone (GBL) (99%) were purchased from Sigma-Aldrich. All salts and solvents were used as received, without any further purification.

FAPbI3 Single-Crystal Preparation

FAPbI3 single crystals were prepared by the inverse temperature crystallization (ITC) method, as reported by Saidaminov etal.[59] Briefly, FAI and PbI2 were added into GBL followed by sonication to get a clear 1 M solution at room temperature. The solution was then positioned on a hot plate that was preheated to 70 °C. The temperature was then raised gradually (every 15 min) until several tiny dark nuclei could be observed at the bottom of the vial. The temperature to onset FAPbI3 crystallization strongly varies depending on the glass vial used and the volume of solution used. The typical temperature range to observe crystallite formation is between 100 and 120 °C. Once the crystallite started to form, the temperature was kept unchanged to allow the crystal to grow further. Usually it takes several hours to get a millimeter-sized single crystal of FAPbI3. After crystal growth, the single crystals were picked out using tweezers and were briefly soaked in 10 mL of acetonitrile to remove any precursor residue. The single crystals were then dried using dust-free wipes (Kimtech) and annealed at 50 °C for 3 min. For experiments on α-FAPbI3 materials, samples were placed on a 155 °C hot plate for 5 min prior to characterization.

FAPbI3 Thin Film Deposition

A 1 M precursor solution was prepared by dissolving FAI and PbI2 into an anhydrous DMSO/DMF (v/v 1:9) mixed solution and stirred overnight in a glovebox (H2O < 0.1 ppm; O2 < 1 ppm). The resulting clear, bright yellow solution was filtered with a 0.20 μm polyvinylidene difluoride filter right before spin-coating (2000 rpm for 20 s and 5000 rpm for 50 s). A drop of chlorobenzene was injected onto the spinning glass substrate (1.5 mm thickness) about 30 s prior to the end of spinning. After spin-coating, the film was annealed at 150 °C for 10 min and turned dark and become yellow in the following 2 to 3 days. The final δ-FAPbI3 film was roughly 400 to 500 nm thick and was then kept in a dark and inert environment until optical measurements were conducted.

Synthesis of PbI2 Nanocrystals

A precursor including 175 mg of lead(II) iodide and 10 mL of water was stirred at 70 °C for an hour to get a saturated solution. The solution was kept still at 70 °C for 10 min. Then the colorless supernatant liquid was sonicated in a water bath for 5 min at room temperature, which led to the formation of lead(II) iodide crystals. The resulting yellowish suspension was spin-coated onto a coverslip with a speed of 1500 rpm for 60 s.

X-ray Diffraction Experiments

Powder X-ray diffractions patterns of the ground perovskite were collected using a STOE STADI MP powder X-ray diffractometer. The diffractometer was running in Bragg–Brentano mode with 2θ–θ geometry, a linear position-sensitive detector, and a copper K-α1 (λ = 1.540 60 Å) source. The diffraction patterns were measured for an acquisition time of 1200 s per sample and collected values over −16° to 61° (2θ).

Raman Scattering Experiments

The Raman (micro)spectroscopy setup is a homemade system, employing an argon ion 458 nm laser excitation source. Scheme provides an overview of the important elements of the Raman experiments. An Ar+ cw-laser (Coherent Inc.) is tuned to operate at 458 nm, with its emission passed through a narrow bandpass filter to remove plasma lines from recorded spectra. Incident laser light is coupled into an inverted Olympus IX71 microscope (contrary to the simplified depiction in Scheme ), and scattered light is passed through a TriVista triple spectrometer setup (Princeton Instruments), which used to disperse optical signals that are detected using a liquid-nitrogen-cooled CCD camera. As it operates via three independent monochromators (see Scheme ), a so-called subtraction measurement was employed, whereby Mono 1 and Mono 2 have matched focal lengths (500 cm), gratings (G+, G– = 900 g/mm), and opposing dispersion geometries to act as a filter and negates the need for a Rayleigh filter (through extreme stray light rejection). Dispersion of Raman spectra is achieved through Mono 3, which has a focal length of 750 cm and groove density of 1800 g/mm geometry, facilitating a spectral resolution of ∼5 cm–1 down to Raman shifts of 40 cm–1. Correct instrument calibration was verified through checking the position of the Si band at ±520.7 cm–1, and laser power densities were controlled by neutral density filters and tuning the argon lasing current. All spectra are presented as raw data (no background subtracting or smoothing has been performed) and are recorded using the Raman laser power densities indicated in corresponding figure captions. A homemade gas flow sample-mounting stage fitted with an optical window was implemented for experiments performed under differing atmospheres.

Scheme 1

Simplified schematic diagram of the home-made Raman scattering and DLW instrument, whereby Raman spectra are recovered in situ from the backscattered 458 nm DLW laser light.

Direct Laser Writing

DLW experiments were performed on our homemade Raman scattering microscope, using the same incident 458 nm Raman excitation laser light. To ensure Gaussian microprobe spot sizes neared the diffraction limit, an optical beam expander was placed before the microscope optics used to focus light onto the sample via one of three air objectives with different magnification and numerical aperture (NA) values: (i) 10×, 0.4 NA, (ii) 20×, 0.75 NA, and (ii) 40×, 0.9 NA. The motorized XYZ stage (Märzhäuser Wetzlar) of the Raman microscope enabled precise sample manipulation, and laser light was visualized to focus on the sample surface (Z0) through a video camera/beam splitter assembly. With Z0 found, a potential DLW area was scanned using manual translations in X and Y while being directly observed through the video camera for adverse changes in the focal spot. This ensured the plane of the patterned area (typically on the order of 100 × 100 μm) would remain roughly in focus during a DLW procedure. Our ideal focus spot sizes were confirmed through a dual approach: (i) direct measurement across the edge of a cleaved Si wafer (line scan) and (ii) size and shape (Gaussian-like) were optimized before every DLW run using the video camera coupled to the microscope. The power of the DLW spot was controlled via a laser current supply module and selecting the appropriate neutral-density filter on a computer-controlled wheel and was measured using a calibrated power meter (ThorLabs photodiode S130VC). Power density values of the Gaussian beam spot were calculated for the 1/e2 (13.5% of peak) beam diameter. Following preliminary laser power tests, DLW patterning of α-FAPbI3 microtracks was achieved by moving the sample using the Märzhäuser Wetzlar XYZ stage. All kinetic movements of DLW patterning were defined by a series of XYZ coordinates controlled by a computer, permitting designs consisting of only straight orthogonal lines (in X and Y; no diagonals) with a maximum write speed of 10 mm/s. The limitations of this setup are evident at XY directional changes, where dwell time at these sites is maximized, forming a corner “node”. All DLW patterns (written with cw-laser light) commence and terminate with the laser focus at a relative positive Z value (300 μm away from the sample surface, Z0), permitting cleaner (no overexposed or redirected DLW lines) and discontinuous patterning with cw-laser light. Through this setup, it was further possible to acquire Raman scatting spectra in situ during the DLW experiments by redirecting backscattering light to the Raman spectrograph component of the system.

Scanning Electron Microscopy

SEM characterizations on the samples were carried out with a FEI Quanta FEG-250 SEM after optical measurements. An acceleration voltage of 2 or 5 kV was applied during the measurements to reduce charging on the sample surface.

Micro-PL Spectral Mapping

PL spectral mapping over the sample was carried out on a home-built inverted optical microscope (Ti–U, Nikon) in a confocal mode. An air objective of 0.92 NA and 60× magnification was used. A 532 nm diode laser was used as the excitation light source. The excitation power density was regulated with a set of neutral density filters. A pair of half- and quarter-wave plates was used for converting the linear-polarized laser light into circular-polarized for excitation. A 545 nm long-pass filter was used in front of the entrance of the spectrograph. A piezostage was used for sample scanning.

Confocal Fluorescence Microscopy

Confocal fluorescence micrographs were acquired on an inverted epi-fluorescence confocal microscope (Olympus IX81), λexc = 488 nm, equipped with an air immersion objective lens (Olympus, 40×, 0.9 NA). Detection of fluorescence emission (λ = 780–800 nm to avoid high-energy species; see Figure S7 of the Supporting Information) was done with a PMT and pixel dwell time of 20 μs. The images collected have fields of view of 158.72 × 158.72 μm2 with a pixel size of 248 × 248 nm2, which was zoomed in with a 2× magnification. Multiple images were taken at several focus planes and processed using ImageJ to obtain the average projection of the fluorescence signal. Background intensity was corrected by thresholding using ImageJ to remove signal contributed by the background.

Optical Absorption

The absorption spectra were recorded in the wavelength range 300 to 950 nm in a reflection geometry, using a PerkinElmer-Lambda 950 UV–visible spectrometer.