Particle-Catalyst-Free Vapor-Liquid-Solid Growth of Millimeter-Scale Crystalline Compound Semiconductors on Nonepitaxial Substrates.

Direct growth of single-crystal compound semiconductors on nonepitaxial substrates is a promising route for device processing simplification in electronic and optoelectronic applications. However, the nonepitaxial growth technique for 2D single crystals is still a fundamental challenge. Here, we demonstrate that the macroscopic 2D interface of liquid metals and nonepitaxial solid substrates could be universally designed for the chemical vapor deposition growth of crystalline compound semiconductors. By adopting a sandwiched solid metal/liquid metal/solid substrate environment, millimeter-scale 2D GaS, 2D GaSe, and 1D GaTe single crystals of high quality were synthesized at the interface of liquid gallium and nonepitaxial substrates. Evidence shows that the particle-catalyst-free vapor-liquid-solid growth is driven by screw dislocations. Furthermore, we successfully extend the growth strategy to various metal chalcogenides (Sn, In, Cu, and Ag) and pnictides (Sb). Our work opens up a new route for the direct growth of single-crystalline compound semiconductors on nonepitaxial substrates.

Introduction

Integration of compound semiconductors with desired substrates is fundamentally important for the development of electronic and optoelectronic applications. Vapor-phase[1,2] or liquid-phase[3,4] epitaxial growth techniques could produce high-quality crystalline compound semiconductors with excellent device performance but with limitation on lattice-matched substrates. Though wafer-scale transfer of semiconductors to target substrates has been developed,[5,6] direct growth of single crystals on nonepitaxial substrates is still highly needed for device processing simplification and cost reduction in industrial applications.

The vapor–liquid–solid (VLS) growth involving catalytic liquid droplets is a well-known approach for the synthesis of one-dimensional (1D) nanostructures.[7] A major advantage is that growth of high-quality 1D nanostructures does not need an epitaxial substrate, which has inspired fundamental research studies and practical device explorations for decades.[8−11] However, the existence of nanoparticle catalysts inevitably not only raises the problem of geometry control of 1D nanostructures but also in principle hinders the growth of 2D crystalline materials on nonepitaxial substrates. Particle-catalyst-free efforts have been made to address the challenges within the VLS framework, such as the thin-film VLS method[12−14] and templated liquid phase growth.[15] In addition, the size of the as-grown single crystals is still limited, and intricate wetting pre-engineering is required. A general method available for large-scale nonepitaxial growth of both 1D and 2D crystalline materials is yet to be demonstrated.

Results and Discussion

Here, we developed a particle-catalyst-free VLS method to synthesize various 1D or 2D single-crystal compound semiconductors on nonepitaxial substrates. As schematically shown in Figure A, gallium chalcogenides are chosen as the demonstration due to their widespread applications in optoelectronics.[16−18] Gallium is used as a reactive liquid environment due to its ultralow melting point (∼29.7 °C) and relatively high boiling point (∼2204 °C), which provides a wide liquid window for VLS growth.[19−21] The key issue is the preparation and maintenance of a macroscopic wetting liquid–solid interface at relatively high growth temperatures. In order to address the above issue, we developed a general liquid-metal-sandwiched growth configuration, W foil/liquid Ga/Al2O3(0001), for the synthesis of gallium chalcogenides (photograph in Figure B). The covering of W foils could keep the liquid/solid interfaces from dewetting during growth as well as minimizing the exposure of liquid to a reactive gas phase (S, Se, Te, etc.). The detailed reaction conditions are illustrated in supplementary methods. Figure C shows the photograph of the W foil and the Al2O3 substrate after growth. The sandwiched configuration was easily taken apart due to the liquid nature of gallium. As expected, the W foils and Al2O3 substrate were wholly covered and wetted by liquid gallium. On comparison, liquid gallium would be dewetted from the Al2O3 surface into small droplets without the assistance of W foils (Figure S1). The liquid could be easily separated from solid substrates by high-speed centrifugation (∼9000 rpm) at a temperature of ∼50 °C (Figure D). Surprisingly, large crystals could be observed with the naked eye on both substrates. The optical images in Figure E show a large lateral dimension of 2D GaS, 2D GaSe, and 1D GaTe with sizes of 1.3, 0.8, and 1.4 mm, respectively, grown at 1000 °C (results at lower temperatures seen in Figure S2). Some residual Ga could be further removed by diluted NaOH etching. The well-defined crystal shapes indicate the single crystal nature of the as-grown samples, which is further shown in Figure . The maximum growth rate of the single-crystalline gallium chalcogenides could be estimated as ∼110 ± 30 μm/min, which is around one order of magnitude faster than that by previous methods (Figure F).[22−26] The thickness of the as-grown crystals could be tuned by the thickness of the liquid Ga layer. At a thickness of 2 mm, single crystals could be obtained with a height of ∼20 μm (Figure G and Figure S3). While the thickness was reduced to 0.4 mm, ultrathin single crystals were achieved with a minimum height of ∼100 nm (Figure H). The statistics of Figure I clearly show the trend relationship between the liquid layer thickness and the average sample height, proving the controllability of the growth method. Furthermore, growth could be realized on other solid substrates such as SiO2, STO(100) (Figure S4), indicating that this method is a general nonepitaxial growth technique. Moreover, the well-defined crystal shape and large size of the as-grown crystals strongly indicate that the liquid–solid interface provides a quasi-steady growth environment, so that the shape of the as-grown crystal is determined primarily by the intrinsic edge energy and diffusion kinetics of vapor precursors in liquid.

Figure 1

Particle-catalyst-free VLS growth of metal chalcogenides. (A) Schematic illustrations of the interfacial growth of 2D metal chalcogenides on nonepitaxial substrates. (B, C) Photographs of sandwiched samples before growth and after growth, respectively. (D) Photographs of as-grown gallium chalcogenide single crystals on W foil as well as the Al2O3 substrate. (E) Typical optical images of GaS, GaSe, and GaTe single crystals. (F) Comparison of the growth rates of gallium chalcogenides in this work and those reported in the literature. (G) Side-view SEM image of GaSe single crystals grown in a 2 mm-thick confined Ga liquid layer. (H, I) Optical and AFM images of GaSe single crystals grown in a 0.4 mm-thick-confined Ga liquid layer. (I) The thickness dependence of as-grown GaSe and liquid Ga layers. The error bars represent the thickness variation.

Figure 2

EDS and XRD characterizations of as-grown gallium chalcogenides. (A) EDS mapping of as-grown GaS, GaSe, and GaTe single crystals. (B) EDS spectra of as-grown GaS, GaSe, and GaTe single crystals confirming their chemical compositions of Ga and X(S/Se/Te) with an atomic ratio of 1:1. (C) XRD spectra of as-grown GaX on Al2O3 substrates with reference.

To identify and characterize as-grown crystals, we used energy dispersion spectroscopy (EDS), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), Raman, and photoluminescence (PL) spectroscopy on the samples synthesized on Al2O3. EDS confirms that gallium and chalcogens are well distributed in these crystals with homogeneity (Figure A), which all shows an atomic ratio of 1:1 (Figure B). It is worth noticing that the gallium-rich growth environment ensures the sole growth of the stable compound in the Ga-X (S, Se, Te, etc.) phase diagrams. Figure C shows typical XRD patterns of GaS, GaSe, and GaTe, respectively. (0002), (0004), and even higher-ordered diffraction peaks of layered hexagonal GaS and GaSe were observed, while the (210) and (420) peaks of monoclinic GaTe were indexed according to the standard PDF cards, indicating that the as-grown samples are single crystals (XPS spectrum shown in Figure S5). The identity and quality of gallium chalcogenides were also characterized by Raman and PL measurements. The characteristic peaks for hexagonal GaS were observed at 190.0 cm–1 (A1g1 mode), 298.0 cm–1(E2g1 mode), and 360.9 cm–1(A1g2 mode); those for hexagonal GaSe were observed at 133.0 cm–1 (A1g1 mode), 211.9 cm–1(E2g1 mode), 250.0 cm–1(E1g2 mode), and 306.9 cm–1(A1g2 mode); those for monoclinic GaTe were observed at 121.0, 160.0, 210, 270 cm–1 (Ag mode), and 176.9 cm–1(Bg mode) (Figure A). GaSe and GaTe show prominent excitation peaks at ∼2.0 and ∼1.64 eV, respectively, confirming that they are direct bandgap semiconductors.[27,28] Two faint excitation peaks were observed for GaS since it is an indirect bandgap semiconductor (Figure B). The gallium chalcogenides grown on Al2O3 can be easily transferred onto arbitrary substrates by an etching-free transfer method.[29] Thus, the as-grown samples could be transferred onto lacey carbon grids for transmission electron microscopy (TEM) and selective area electron diffraction (SAED) studies. SAED of GaS and GaSe shows a clear set of hexagonal reciprocal patterns along the [0001] zone axis, indicating high crystallinity and structural uniformity (left and middle of Figure C). The high-resolution TEM images of GaS and GaSe show clear hexagonal lattice fringes with lattice spacings of 0.31 and 0.32 nm (left and middle of Figure D), agreeing well with the atomic models (Figure S6). It is a little complicated for GaTe since the SAED shows two sets of reciprocal patterns: a stronger one of monoclinic structures (yellow dashed marks in Figure D right) and a weaker one of hexagonal structures (red ones). It has been reported that a spontaneous phase transformation occurs in the layered gallium telluride from a monoclinic to a hexagonal structure, when the bulk of GaTe is exfoliated to a few layers.[30] The HRTEM images clearly show hexagonal lattice fringes with a lattice spacing of 0.35 nm, confirming the existence of the hexagonal phase of GaTe. All the above characterizations prove that the as-grown metal chalcogenides are of high quality.

Figure 3

Raman, PL, and TEM characterizations of as-grown gallium chalcogenides. (A, B) Raman and PL spectra of as-grown GaS, GaSe, and GaTe single crystals on Al2O3 substrates. (C) SAED patterns of GaX crystals transferred on a Cu grid with lacey carbon films. (D) HR-TEM images of the three 2D materials with atomic resolution, and the insets show the FFT-filtered images of the areas marked by red squares.

Questions arise naturally that how does a large-crystalline 2D or 1D material nucleate and grow at the liquid–solid interface without the participation of catalytic nanoparticles. According to the classic crystal growth theory, the supersaturation (Δμ) of the system is the driving force for crystal growth:[31]where C is the precursor concentration and C0 is the equilibrium concentration. Either of the two growth modes, 2D-nucleation growth and spiral growth, dominates under different supersaturation conditions. The latter one describes a growth process as shown in Figure A: a screw dislocation defect serves as the initial nuclei of crystals, which offers a nonvanishing edge on the surface as the crystal grows. In principle, the growth at edges overcomes less energy barrier compared to the nucleation of new 2D layers (2D nucleation growth). Thus, spiral growth normally dominates under low supersaturation conditions. In this work, the chalcogen concentration at the liquid–solid interface is relatively low, which is limited by the diffusion rate of chalcogen in liquid gallium at the growth temperature. Therefore, we proposed that the particle-catalyst-free VLS growth is driven by screw dislocations on the nonepitaxial substrates.

Figure 4

Mechanism of particle-catalyst-free VLS growth. (A) Schematic illustration of screw-dislocation-driven growth. (B–D) Optical images of spirally grown GaS. GaSe. (E) DFT-MD simulation of the GaS growth process. The initial (left panel) and final (right panel) structures of GaS during a period of 20 ps at 1100 K.

Supporting evidence is shown in Figure B,C. Beautiful growth spirals with signs of 6-fold or 3-fold symmetry could be occasionally observed on single crystals of GaS or GaSe, which strongly indicate that screw dislocations dominate the nucleation and growth process of 2D materials at the liquid–solid interface. Multiple growth spirals were also observed (marked in Figure D), and the aligned edges of the merged crystal suggest that each spiral self-collimated into alignment during the growth in liquid, which might explain the millimeter-scale crystal size.[32−34] Notably, the screw-dislocation-driven mechanism could also explain the separate morphologies (1 or 2D) in particle-catalyst-free VLS growth (Figure A). The key is the different growth rates around the dislocation core (vc) and those at the outer edges (ve):[35] when ve is larger than vc, the horizontal propagation of the as grown crystals is faster than the vertical growth at the dislocation core, resulting in the formation of 2D crystals, while 1D single crystals can be synthesized when ve is much smaller than vc. Density functional theory-based molecular dynamic (DFT-MD) simulations were performed on the growth of gallium chalcogenide crystals. The GaS nanoribbon with a Ga-terminated zigzag edge was exposed to the Ga–S mixed liquid and maintained at 1100 K for a period of 20 ps. The initial and final structures are shown in Figure E (also see Movie S1). The results show that almost one new row of GaS is grown in the initial zigzag edge within 10 ps. It is reasonable to believe that the simulations also work for the growth of other compounds.

The particle-catalyst-free VLS growth strategy can be universally applied for the synthesis of other binary compound semiconductors. Figure A shows a schematic of the periodic table highlighting the melting points and boiling points of some metals. In principle, metals with a large temperature window of the liquid phase could be applied to the liquid–solid interfacial growth. Figure B shows the optical images of eight as-grown 2 or 1D metal chalcogenides (InX, SnX, X = S, Se, Te) and pnictides (GaSb and InSb) with good crystal symmetries and sharp edges, suggesting that the liquid–solid interface provides a quasi-steady growth environment, so that the shape of the as-grown crystal is determined primarily by the intrinsic edge energy and diffusion kinetics of vapor precursors in liquid. The EDS data confirm the expected stoichiometries for each compounds, which is commonly of 1:1 atomic ratio for III(IV)–V(VI) chemical combinations (Figure S7). Raman spectra further confirm the identities of each product (Figure S8). Note that transition metals are also suitable for this growth strategy (Cu and Ag chalcogenides seen in Figure S9).

Figure 5

Library of metal chalcogenides and pnictides. (A) Overview of metals, chalcogens, and pnictogens with their melting points and boiling points. The ones used in this work are marked in red. (B) Optical images of 8 as-grown metal chalcogenides (InX, SnX, X = S, Se, Te) and pnictides (GaSb, InSb).

Conclusions

In conclusion, we demonstrate that the 2D interface of liquid metals and solid substrates could be universally designed for the CVD growth of various two-dimensional metal chalcogenides and pnictides. The new particle-catalyst-free VLS growth is fast and nonepitaxial, and the merits are well inherited from the classic VLS growth of one-dimensional nanostructures. The growth technique is scalable and compatible with a typical CVD system, which might provide an alternative pathway toward the synthesis of semiconductor single crystals for industrial optics and optoelectronics. Notably, it is reasonable to believe that not only liquid metals but also other suitable liquid compounds (ionic salts, for example) could serve as reactive liquid environments, which will enlarge the library of materials available.

Experimental Section

CVD Growth of GaX (X = S, se, Te, Sb)

The compounds were synthesized in a horizontal tube furnace (Hefei Kejing OTF-1200X CVD system). The width of the furnace is about 40 cm, the length of the horizontal tube is about 120 cm. H2 (50 sccm) was used as the carrier gas. A small quartz tube filled with moderate reaction source powders or pills (S ∼ 0.2 g, Se ∼ 0.2 g, Te ∼ 0.1 g, and Sb ∼ 0.1 g) was put on the upwind of the tube furnace. Before growth, a layer of liquid gallium (thickness ∼ 0.4 to 2 mm) sandwiched between a 1 × 1 cm W foil and a 1 × 1 cm Al2O3 (0001) substrate was placed on a quartz boat. The sample was heated to the growth temperature range of 700–1000 °C at a ramp rate of 15 °C min–1. When the reaction temperature was achieved, the precursor sources were heated to 150 °C (S), 250 °C (Se), 450 °C (Te), and 630 °C (Sb). The distances between the center of the quartz boat and the center of the reaction source were about 27, 25, 23, and 2 cm. Finally, after a 10 min growth, the system was cooled naturally to room temperature.

CVD Growth of InX (X = S, se, Te, and Sb)

The main growth process was similar to the growth method of GaX. The sandwiched structure is composed of a 1 × 1 cm W foil, a 1 × 1 cm Al2O3 (0001) substrate, and a 0.5 × 0.5 cm × 0.5 mm indium foil between them. The sample was heated to a growth temperature of 900 °C at a ramp rate of 15 °C min–1. It is worth noticing that indium would be re-solidified when cooled to room temperature. The sandwiched configuration could be taken apart on the heating table at about 160 °C.

CVD Growth of SnX (X = S, se, Te)

The main growth process was similar to the growth method of InX. The sandwiched structure is composed of a 1 × 1 cm W foil, a 1 × 1 cm Al2O3 (0001) substrate, and 0.15 g tin powders in between them. The tin would also be re-solidified when cooled to room temperature. The sandwiched configuration could be taken apart on the heating table at about 230 °C.

CVD Growth of M2X (M = cu, Ag; X = S, se, Te)

The main growth process was similar to the growth method of InX. The sandwiched structure is composed of a 1 × 1 cm W foil, a 1 × 1 cm Al2O3 (0001) substrate, and a 1 × 1 cm Cu or Ag foil in between them. The sample was heated to a growth temperature of 1100 °C at a ramp rate of 15 °C min–1. It is difficult to separate the as-grown metal chalcogenides from the metal/solid interface in an intact manner under ambient conditions due to their high melting points. However, millimeter-scale single crystals with varied thicknesses could be exfoliated on the Al2O3 substrate.

DFT-MD Simulation of the GaS Growth Process

DFT-MD simulations are implemented in the Vienna Ab Initio Simulation Package (VASP). The core electrons are treated by the projector-augmented wave method, and the exchange-correlation functional of the valence electrons is calculated within the generalized gradient approximation (GGA) using the parameterization of Perdew, Burke, and Ernzerhof (PBE). Considering the time-consuming nature of the DFT-MD simulation, the plane-wave cutoff is set to be 240 eV and the Brillouin zone is sampled at the Γ point only. The DFT-MD simulations are performed using the canonical ensemble (NVT), and the ion temperature is controlled around 1100 K using a Nosé–Hoover thermostat. The time step is 2 fs. The total time scales of the trajectories are more than 20 ps. In the simulation box, there is a vacuum layer along the growth direction, and the boundaries are periodic along the other two directions. The freestanding end of the GaS nanoribbon is fixed during the simulation.

X-ray Diffraction (XRD)

XRD measurements were performed using a Bruker D8 Venture XRD. The scanning angle was from 10 to 80°, the scanning rate was 0.1°/s, the tube voltage was 40 kV, and the tube current was 40 mA.

Scanning Electron Microscopy (SEM) and Energy Dispersive Spectroscopy (EDS)

SEM images with EDS measurements were obtained using a Hitachi S4800 FESEM system with an accelerating voltage of 20 kV.

Atomic Force Microscopy (AFM)

AFM measurements were performed using a Bruker Dimension Icon AFM. The images were taken in ScanAsyst-air mode with a ScanAsyst-air AFM tip.

X-ray Photoelectron Spectroscopy (XPS)

XPS measurements were performed using a Thermo Scientific ESCALAB 250 Xi system with monochromatized AlKa X-rays (1486.8 eV, 500 μm spot size) as the excitation source. All spectra were charge referenced independently to graphitic carbon (C1s = 284.8 eV).

Raman and PL Spectroscopy

The measurements were performed using a Horiba LabRAM HR-800 Raman spectrometer. A 532 nm laser source (intensity 0.45 mW at the sample position) was utilized.

Transmission Electron Microscopy (TEM)

TEM imaging was performed on an FEI Tecnai G2 F30 with an accelerating voltage of 300 kV.