Epitaxial growth of CH₃NH₃PbI₃ on rubrene single crystal

Organometal halide perovskites have attracted much attention because of their wide range of applications, such as in solar cells¹ and light-emitting devices,^2–4 including lasers,^5,6 and solid-state memory.^7,8 In particular, perovskite solar cells have been a very popular research field, reaching a higher efficiency than other thin-film solar cells in just half a decade.⁹ Although many fabrication methods have been proposed, the research on the fundamental viewpoint of crystallization seems to be insufficient, although it is strongly required to realize the high yield and reliable production of commercial devices. The development of methods for the control of the crystal structure is very important to form the backbone for semiconductor technology and theory. The tools employed to control the crystal structure represented by heteroepitaxy open the way for this undeveloped research field, whereby further improvement of device characteristics, discovery of novel material properties, and the realization of new functions are expected.

In the field of III–V compound semiconductors, heteroepitaxy is an essential method to achieve high-performance solar cells, and the method of crystallization has been studied for many decades. In this study, we have attempted the heteroepitaxy of an organometal halide perovskite, as schematically illustrated in Fig. 1(a), for the construction of highly controlled perovskite crystals with an ordered interface between the perovskite and the carrier transport layer.

Heteroepitaxy of the perovskite has already been reported by Wang et al.,¹⁰ where mica was used as a templating substrate. Some reports related to optical and carrier-dynamics analyses were also published by the same group.^11–17 These can be considered cutting-edge results on the construction of highly ordered perovskite films. Other examples of heteroepitaxy are the growth of CsSnBr₃ on the NaCl single crystal,¹⁸ the growth of CsPbBr₃ on the StTiO3 single crystal,¹⁹ and the growth of CH₃NH₃Pb(Br_1−xI_x)₃ on the CH₃NH₃PbBr₃ single crystal.²⁰ However, for the further development of this method, such as energy alignment, carrier transport control, and construction of solar cells with high crystallinity, it would be better to use a semiconducting material such as an organic semiconductor as a templating substrate. Based on this concept, we have used rubrene single crystals as substrates that can be easily fabricated by physical vapor deposition. Rubrene is a representative of p-type organic semiconductors because of its high hole mobility characteristics and rubrene can be applied as a hole transport material for perovskite solar cells.²¹ In our previous study, we succeeded in the construction of the epitaxial C₆₀ film on a rubrene single crystal.²² The benefit of the rubrene single crystal as a templating substrate has thus already been verified. In this study, we attempt the epitaxial growth of CH₃NH₃PbI₃ on a rubrene single crystal.

With regard to the method used to construct organometal halide perovskites, the majority of researchers fabricate perovskite by a solution process, which is a low-cost process compared to conventional semiconductor processes such as vacuum deposition. However, vacuum deposition is a method that is even more powerful for the fabrication of perovskite films because of its potential for high controllability and high-quality crystal growth without a solvent.²³ We have developed methods for the vacuum deposition of perovskites to construct highly controlled perovskite crystals. We have reported that an infrared (IR) laser deposition method rather than the use of a thermal heater enables precisely controlled deposition of the perovskite film.²⁴ 

The crystal orientation of the resultant perovskite crystal was analyzed by grazing incident wide angle x-ray scattering (GIWAXS), where the diffraction patterns were collected from 360° to obtain a three-dimensional reciprocal mapping of the crystal. The crystallinity and orientation were investigated with respect to the deposition conditions. In addition to verification of the epitaxial growth of perovskite with an atomic-order flat surface on the rubrene single crystal, nanoscale insight into the crystallization process of the perovskite was obtained. Based on the results, the importance of the initial stage of perovskite formation was clarified, where the octahedral unit of PbI₂ rotates vertically while retaining its lateral orientation to form a highly ordered perovskite crystal. The perovskite film is grown layer-by-layer; even PbI₂ and CH₃NH₃I were supplied simultaneously.

Rubrene (Aldrich) single crystals were fabricated by physical vapor deposition. The (001) orientation of the rubrene single crystal was verified by x-ray diffraction (XRD, RIGAKU Smart Lab, Cu Kα1 x-ray source). Single crystal rubrene thin plates were mounted on a glass substrate and transferred to a vacuum chamber with a base pressure of 10⁻⁵ Pa. The CH₃NH₃PbI₃ film was deposited onto the substrate using the laser deposition method, where the source materials (PbI₂ and CH₃NH₃I, purchased from Tokyo Kasei) were irradiated with a continuous-wave IR laser and evaporated to supply materials to the substrate.²⁴ The use of laser deposition enables precise control of the material supply rate. Both the co-deposition method and the alternative deposition method were attempted to fabricate perovskite films. The precise crystal orientation was analyzed by GIWAXS at a synchrotron radiation facility (SPring-8, BL46XU). GIWAXS measurements were performed by rotating the crystal 360°, and a diffraction pattern was measured using a two-dimensional x-ray detector (PILATUS 300K). The details of the measurement set up and data conversion to the reciprocal space mapping are described in the supplementary material. The surface morphologies of the CH₃NH₃PbI₃ thin films were examined by atomic force microscopy (AFM; Nanonavi/E-sweep, SII nanotechnology).

The surface of the rubrene single crystal [Fig. 1(b)] has clear step-and-terrace structures in which the step height is ca. 1.3 nm, which is approximately half the length of the c-axis lattice constant of rubrene and thus corresponds to a single layer of rubrene molecules. A molecular-order flat surface can be advantageous as a base for crystal growth.

PbI₂ and CH₃NH₃I were co-deposited to form a 20 nm thick CH₃NH₃PbI₃ perovskite film on the rubrene single crystal. The crystal structure was investigated with respect to the substrate temperature. AFM images of the perovskite films co-deposited at room temperature and 120 °C are shown in Figs. 1(c) and 1(d), respectively. The crystal grains became larger at the higher temperature. The typical grain size was several hundred nanometers. The line profiles for the AFM images in Fig. 1(c) (the red line) and Fig. 1(d) (blue line) are shown in Fig. 1(e). Clear step-and-terrace structures were observed. The typical step height was 0.6 nm and multiples thereof. When a histogram (see Fig. S3 in the supplementary material) was plotted, peaks with an interval of ∼0.6 nm were detected. The value of 0.6 nm corresponds to the interval of the octahedral unit of the perovskite crystal. This is the evidence of layer-by-layer growth, which results in an atomically controlled flat surface. An atomic-order flat surface was a typical feature in this study, whereas such a surface cannot be observed in a perovskite film fabricated on a conventional substrate such as TiO₂ or a polymer-based hole transport layer.

The crystal orientation was analyzed by three-dimensional reciprocal space mapping with GIWAXS. The reciprocal space mapping clipped at k_z ≈ 0 of a rubrene single crystal and that for perovskite films grown at room temperature and 120 °C are shown in Figs. 2(a)–2(c). By comparing the GIWAXS data for the substrate and the thin films, we can discuss the orientation relationship between the substrate and the grown films. The spot pattern attributed to CH₃NH₃PbI₃ appeared at a certain direction in the reciprocal space, which verifies the epitaxial growth of the perovskite on the rubrene single crystal. The patterns of orientation for the perovskite were different between the film grown at room temperature and that grown at 120 °C. Six to seven different orientations (0°, 11°, 18°, 41°, 49°, 72°, and 79° from the rubrene a-axis, where the appearance of the 0° orientation is dependent on the sample) were observed for the film grown at room temperature, whereas the pattern was reduced to three orientations (15°, 45°, and 75°) for the film grown at 120 °C. The mechanism for the growth of the film is thus considered to be dependent on the temperature. When the orientation patterns for films grown at intermediate temperature (60–80 °C) were analyzed, the results were dependent on the sample, and the number of orientation patterns ranged from 3 to 5 (see Fig. S4 in the supplementary material).

The mechanism for growth at elevated temperature is discussed here. The 15°, 45°, and 75° orientation patterns of the perovskite on a rubrene (001) surface are schematically represented in Fig. 3. The lattice relationship can be described as follows:

where a_PVS is defined as the vector for the direction of PVS [110] with the length between Pb ions and b_PVS is defined as the vector for the direction of PVS [$11¯0$] with the length between Pb ions. The lattice mismatch for PVS-1 and PVS-2 is 0.2% and 1.0%, respectively.

The mismatch between the rubrene and perovskite lattices is 0.2% for 15° and 75° orientation and 1.0% for 45° orientation. A mismatch of less than 1% is sufficiently small to achieve epitaxial growth. The source molecules diffuse actively on the substrate at the elevated temperature, so crystal growth proceeds in the most suitable direction. That is, there are three suitable directions (15°, 45°, and 75°), where the mismatch is less than the appropriate value (1%) to achieve epitaxy.

Next, we discuss the growth mechanism at room temperature. The growth of PbI₂ on the rubrene single crystal plays an important role for the co-evaporation process of CH₃NH₃PbI₃; therefore, the growth of PbI₂ is discussed first. PbI₂ is reported to grow epitaxially on a layer-structured substrate with hexagonal crystals.²⁵ The epitaxial growth of PbI₂ on a mica substrate was reported recently.²⁶ We attempted the laser deposition of PbI₂ on a rubrene single crystal, and the epitaxial growth was verified (see Fig. S5 in the supplementary material). The PbI₂ crystal grows with its a₁ axis rotated ±3.75° from the rubrene a-axis (represented as PbI₂-1 and PbI₂-2 in Fig. 4). With this configuration, the lattice mismatch between the rubrene a + b lattice vector and PbI₂ 2a₁ + 2a₂ lattice vector is 1.9%. We assume that the PbI₂ crystal is converted into perovskite while maintaining its in-plane orientation, i.e., the octahedral unit rotates vertically, as schematically shown in Fig. 5. The in-plane crystal structure can be described as represented with PVS-4 and PVS-5 in Fig. 4. The original orientation of PbI₂ is ±3.75° from the rubrene a-axis; therefore, the [1 1 0] axis of the generated perovskite is 45° ± 3.75°. Considering the symmetry, both 15° ± 3.75° and 75° ± 3.75° orientations are also possible. The lattice relationship can be described as follows:

The lattice mismatch was 1.9% for PVS-4. The orientation pattern obtained in the room-temperature-deposition experiment can be explained with this model where the spot pattern appeared at around 15° ± 3.75°, 45° ± 3.75°, and 75° ± 3.75°. At room temperature, the diffusion of source molecules on the substrate is not active, so the perovskite crystal growth continues with the in-plane orientation of the PbI₂ crystal. Depending on the sample, the 0° orientation growth mode, illustrated as PVS-6 in Fig. 4, was observed,

The lattice mismatch was 1.4% for the rubrene a axis direction and 2.6% for the rubrene b axis direction. Based on these results, the initial-stage crystal orientation of PbI₂ is important for the successive perovskite crystal growth. The results suggest that the growth of the perovskite crystal can be expressed as a layer-by-layer mechanism, where PbI₂ crystallizes first and is then converted to the perovskite crystal structure as CH₃NH₃I is supplied, even under the co-evaporation conditions.

The initial stage of PbI₂ growth was determined to be important for the orientation control of the perovskite; therefore, alternative deposition can provide significant insight into the mechanism of perovskite crystal growth. Although alternative deposition or sequential deposition by the solution process has some difficulty in morphology control,^27,28 that by vacuum deposition is a promising method to construct highly controlled perovskite films.^29,30 We attempted an alternative growth of the perovskite on the rubrene single crystal, where 1 nm thick PbI₂ and 5 nm thick CH₃NH₃I layers were supplied alternatively ten times. The reciprocal space mapping for the perovskite film alternatively deposited at room temperature is shown in Fig. 6. The [110] diffraction of the perovskite appeared at 15°, 45°, and 75° from the rubrene a-axis. The orientation pattern was the same as that of the film co-deposited at elevated temperature. The perovskite crystal grows into the direction with the least lattice mismatch. Although PbI₂ and CH₃NH₃I were supplied alternatively and PbI₂ is considered to be epitaxially grown on the rubrene single crystal, the complete crystal grows into the most stable condition. This is because alternative deposition is a very slow process, and even at room temperature, the intermediate film has sufficient mobility to migrate into the most stable sites. On the other hand, when the reciprocal space mapping is closely observed, the diffraction that appeared at 45° is broad and weak compared with the spots at 15° and 75°. This can be due to the relatively large lattice mismatch (1% for 45°), which prevents migration to form the crystal at the minimal mismatch sites. During this process, the growth mode of PbI₂ on the rubrene single crystal (1st layer) and CH₃NH₃PbI₃ (2nd layer or higher) is predicted to be different. However, in this study, no specific orientation pattern of PbI₂ on CH₃NH₃PbI₃ was observed. In the reciprocal pattern for the final CH₃NH₃PbI₃ film, a weak spot of PbI₂ with the same orientation as that on rubrene was observed. The detailed growth dynamics for alternative deposition could be a topic for a future study. Regardless of the PbI₂ growth orientation, the experimental results indicate that the perovskite grows into the most stable sites.

The out-of-plane reciprocal mapping to the direction of k_x, k₁, and k₃ [defined in Fig. 6(c)] was represented in Figs. 6(d)–6(f), respectively. The diffraction spots of the rubrene single crystal and epitaxial CH₃NH₃PbI₃ were observed. Assignments of the diffraction spots marked in (d)–(f) are summarized in Table I. The discussion of the crystal phase, i.e., tetragonal or cubic, is interesting for perovskite, as discussed in Refs. 31 and 32. Based on the results in this study, the existence of [2 1 1] and [2 1 3] diffraction can be the evidence for the formation of tetragonal-phase perovskite.³³ 

In our previous report,³⁴ we investigated the crystallization dynamics for a solution process using the real-time GIWAXS analysis. The (0001) oriented PbI₂ crystal was converted to CH₃NH₃PbI₃. The resulting perovskite film first showed an oriented structure but subsequently transformed to a randomly orientated film. On the other hand, in the present study, the lateral crystal orientation was maintained during the conversion to the perovskite, and the PbI₆ octahedral unit rotates vertically to form well-oriented epitaxial perovskite crystals. Based on these results, the difference in the growth dynamics between the solution and vacuum processes was clarified. The perovskite can be crystallized through an atomically well-controlled pathway with the vacuum deposition process. The benefit of the vacuum process was demonstrated. The transformation from single crystal PbI₂ to the perovskite was reported in Ref. 35, and the difference between the solution process and vapor process was compared. It was reported that the perovskite crystal formed with the vapor process was randomly oriented with a rough surface morphology. This is due to expansion of the crystal where the reaction was started from bulky PbI₂ crystals with a thickness of ∼0.5 µm. In the present study, the formation of PbI₂ crystals and subsequent transformation into the perovskite proceeded layer-by-layer. Such nanoscale control of the crystal growth resulted in highly oriented crystals with atomic-scale flatness.

When we compared the results for deposition at room temperature and elevated temperature, a clear difference in the growth dynamics was evident and the transition temperature was relatively low (60–80 °C). The low transition temperature, where the perovskite growth maintains the initial PbI₂ orientation (low temperature) or growth occurs into the most stable orientation (elevated temperature), means that the diffusion kinetics for the perovskite crystal growth pass through a relatively low activation energy (∼30 meV). The difference between co-deposition and alternative deposition was also determined. With the alternative deposition method, where the perovskite crystallized slowly, the perovskite tended to be crystallized into the most stable sites, even at room temperature. Based on all the results obtained in the present study, there are three important elementary processes for perovskite epitaxial growth: epitaxial growth of PbI₂, vertical rotation of the octahedral PbI₆ unit, and migration of the intermediate structure into the most stable sites. We have obtained some insight into these elementary processes from both thermodynamics and kinetics points of view based on the temperature dependence and deposition method dependence experiments. In addition to obtaining nanoscale insight into the dynamics of perovskite growth, we have succeeded in the construction of highly ordered epitaxial films with atomic-order flatness. These results with the nanoscale dynamics of vacuum-process crystallization and atomic-order control of perovskite crystals represent an important achievement that should provide important insight into the mechanism of perovskite crystal growth, which is still under debate for both the solution³⁶ and vacuum processes. From a more general point of view, the reaction from the hexagonal crystal into the tetragonal crystal while maintaining the orientation is a new finding in the field of crystal growth. Although the crystal growth of GaN through the hexagonal crystal into the cubic crystal has been reported,³⁷ it is categorized as a phase transition. The transition of a crystal system by a chemical reaction while maintaining its crystal orientation, however, is original, to the best of the authors’ knowledge, and may thus attract significant attention in the field of fundamental crystal growth.

In conclusion, we have achieved the epitaxial growth of the CH₃NH₃PbI₃ perovskite with an atomically flat step-and-terrace surface structure using a rubrene single crystal as a template. Crystals with several orientation patterns were observed, where the orientation pattern was dependent on the substrate temperature and the deposition conditions. The perovskite crystals grow into the most stable sites at elevated temperature and with alternative deposition. The crystal orientation of the perovskite co-deposited at room temperature reflects the orientation of PbI₂, which was also epitaxially grown on the rubrene single crystal. The vertical rotation of the octahedral PbI₆ unit, while maintaining its lateral orientation, explains the crystal growth of the perovskite. This demonstration of atomic-order control of the perovskite and the revelation of microscopic insight into perovskite crystal growth dynamics are expected to be an important springboard for the further development of perovskites.

See the supplementary material for a detailed description of the XRD ϕ scan and additional data such as reciprocal space mapping.

GIWAXS measurements were performed at SPring-8 BL46XU with the approval of the Japan Synchrotron Radiation Research Institute (JASRI; Proposal Nos. 2016A1514, 2016B1861, and 2017A0136). This work was supported by a Kakenhi Grant-in-Aid (Grant No. 16H05978) from the Japan Society for the Promotion of Science (JSPS).