Confined van der Waals Epitaxial Growth of Two-Dimensional Large Single-Crystal In2Se3 for Flexible Broadband Photodetectors

The controllable growth of two-dimensional (2D) semiconductors with large domain sizes and high quality is much needed in order to reduce the detrimental effect of grain boundaries on device performance but has proven to be challenging. Here, we analyze the precursor concentration on the substrate surface which significantly influences nucleation density in a vapor deposition growth process and design a confined micro-reactor to grow 2D In2Se3 with large domain sizes and high quality. The uniqueness of this confined micro-reactor is that its size is ~102-103 times smaller than that of a conventional reactor. Such a remarkably small reactor causes a very low precursor concentration on the substrate surface, which reduces nucleation density and leads to the growth of 2D In2Se3 grains with sizes larger than 200 μm. Our experimental results show large domain sizes of the 2D In2Se3 with high crystallinity. The flexible broadband photodetectors based on the as-grown In2Se3 show rise and decay times of 140 ms and 25 ms, efficient response (5.6 A/W), excellent detectivity (7×1010 Jones), high external quantum efficiency (251%), good flexibility, and high stability. This study, in principle, provides an effective strategy for the controllable growth of high quality 2D materials with few grain boundaries.


Introduction
Two-dimensional (2D) materials have been considered promising candidates for miniaturized and highperformance electronic and optoelectronic devices due to their atomically flat and ultrathin nature and their lack of dangling bonds. Grain boundaries and defects in 2D materials can reduce charge transport, [1] mechanical, [2][3][4] and thermal properties [5]. Therefore, the controllable growth of 2D materials with large domain sizes and high quality with few grain boundaries and defects is important in order to achieve good device performance. To this end, different vapor deposition methods have been developed to grow 2D materials with large domain sizes, such as epitaxial growth on special substrates (e.g., sapphire [6] or mica [7,8]), reducing nucleation density by locally feeding the precursors [9,10], and passivating active sites during growth [11]. As summarized in Figure 1(a), the current domain sizes of semi-metallic graphene range from micrometers to meters [9,12,13], while semiconducting transition metal dichalcogenides (TMDCs) [14][15][16] and insulating hexagonal boron nitride (h-BN) have sizes up to hundreds of micrometers or larger [17][18][19]. Recently, another group of 2D materials with the structure A 2 B 3 , where A is a group III element and B is a group VI element, such as indium selenide (In 2 Se 3 , its structure is shown in Figure S1), has attracted increasing interest. First, In 2 Se 3 has a direct bandgap of 1.36 eV [20], which is close to that of silicon (1.10 eV). Second, although some monolayer TMDCs like MoTe 2 also have a direct bandgap of 1.10 eV, they become indirect bandgap 2 Research materials as the number of layers increases [21], while In 2 Se 3 is a direct bandgap material regardless of its thickness [22,23]. Third, unlike black phosphorus which is also a direct bandgap 2D material, thin In 2 Se 3 flakes are very stable in air, which is very important for practical applications. Recently, there has been pioneering work on the use of 2D In 2 Se 3 in piezoelectronics [24,25], optoelectronics [26], and photovoltaics [27]. We note that these reported 2D In 2 Se 3 materials have small domain sizes (from a few to tens of micrometers, Figure 1(a) and Table S1) [28][29][30]. Therefore, controllable growth of high quality 2D In 2 Se 3 with large domains is of great importance to further advance the use of 2D In 2 Se 3 .
To grow 2D materials with large domains, the control of nucleation density is critical, as has been extensively studied in graphene growth [31]. Usually, in the vapor deposition process, the gas flow is controlled to be viscous laminar in order to obtain a constant atmosphere [32]. As shown in Figure  S2a, there is a velocity gradient of gas in a reactor, and the velocity gradually decreases to zero near the substrate surface, forming a stagnant layer above the substrate. Due to the fact that the precursor concentration on the substrate surface can influence the nucleation density, controlling the precursor concentration on the surface is important to control the nucleation of the 2D materials ( Figure S2b). By analyzing the vapor phase deposition process, we have obtained the following formula for the precursor concentration on the surface ( s ): where g is the precursor concentration in the gas phase, s is the surface reaction rate, and d is the characteristic size of the growth reactor (Table S2). Based on formula (1), s decreases as d decreases, causing a lower concentration of precursor on the surface and consequently there is a lower nucleation density of 2D materials in a smaller growth reactor. Based on the above analyses and previous works [33,34], we designed a confined micro-reactor to greatly reduce the size of the growth space so as to grow 2D In 2 Se 3 with large domain sizes. Specifically, the micro-reactor is composed of two slices of mica with a face-to-face stacking feature. First, the size of growth space is effectively reduced by two to three orders, from 10 5 -10 4 m to 10 2 m, so that the nucleation density of 2D In 2 Se 3 is reduced. Second, the mica, used as a substrate for van der Waals epitaxial growth, benefits the growth of 2D In 2 Se 3 with large domain sizes and thin thickness, because the atomically smooth surface and lack of dangling bonds greatly reduce strain from lattice mismatch between the mica and In 2 Se 3 . As a result, in the confined micro-reactor, high-quality In 2 Se 3 with domain sizes larger than 200 m has been grown on the mica. Moreover, direct growth on flexible mica facilitates the fabrication of flexible photodetectors which have a response time of 140 ms for rise and 25 ms for decay, high responsivity (5.6 A/W), high detectivity (7×10 10 Jones), and large external quantum efficiency (EQE, 251%). After bending 1000 times, the photodetector showed a steady photocurrent with an 80% retention, indicating good flexibility.

Results
As shown in Figure 1(b), the confined micro-reactor is composed of two slices of freshly cleaved mica, stacked face to face. The space ( ) between them is only 140 m, as seen from a side-view scanning electron microscope (SEM) image (inset of Figure 1(b)). This is two to three orders of magnitude smaller than the diameter of the quartz tube in conventional growth experiments (∼20-50 mm). To grow the In 2 Se 3 , In 2 Se 3 powder was used as the precursor, and the two slices of mica served as a micro-reactor (see details in Supplementary Materials and Figure S3). Because of the design of this reactor (Figure 1(c)), we obtained 2D In 2 Se 3 with large domain sizes. As shown in Figure 1(e), the products are triangular with sharp edges and large domain sizes. In contrast, in In 2 Se 3 grown in a conventional reactor (d > 20 mm, Figure 1(d)), the surface of mica is dirtier and the In 2 Se 3 crystals grown have smaller domain sizes as shown in Figure 1(f). Moreover, the edge length and thickness of In 2 Se 3 samples grown in two different reactors were measured by atomic force microscopy (AFM, Figures 1(g), 1(h), and S4). The results show that the average edge length and thickness of In 2 Se 3 are 110 m and 3.6 nm, respectively, grown in the space-confined micro-reactor. This average edge length of 110 m is much larger than that of In 2 Se 3 grown in the conventional substrate, with an average edge length of 40 m (Figure 1(g)).
Insight into the structure and chemical composition of the In 2 Se 3 was obtained from spectroscopic characterization. Figure 2(a) shows Raman spectra of the In 2 Se 3 collected at different positions using a 532 nm excitation laser. All spectra show three typical peaks at ∼108, 180, and 203 cm −1 , corresponding to the A 1 (LO+TO), A 1 (TO), and A 1 (LO) phonon modes of In 2 Se 3 , respectively. We compared the Raman spectra with those of 2D exfoliated In 2 Se 3 by using a scotch tape (pink curve, Figure 2(a)) and data from recent literature [25], and they all match well. To confirm the uniformity of the In 2 Se 3 , we also performed Raman mapping (Figure 2(b)), which indicated good uniformity and homogeneity of the material grown in the confined micro-reactor. The X-ray photoelectron spectroscopy (XPS) spectrum of In 3d shows two peaks located at 452.20 and 444.80 eV from In 3d 3/2 and In 3d 5/2 , originating from In 2 Se 3 (Figure 2(c)). Similarly, the XPS spectrum of Se 3d shows two peaks at 54.80 and 53.80 eV from Se 3d 3/2 and Se 3d 5/2 , also attributed to In 2 Se 3 ( Figure 2(d)) [35]. In addition, the XPS results show that the In 2 Se 3 has good stoichiometry with an In/Se atomic ratio close to 2:3 ( Figure S5). Ultraviolet-visible-near infrared absorption spectroscopy (UV-Vis-NIR) was used to study the optical bandgap of the In 2 Se 3 (Figure 2(e)) which is obtained from the following equation: where is the effective absorption coefficient of the material, is incident photon energy, is a constant, g is the optical bandgap of the material, and n denotes the nature of  the sample transition, which equals 2 for indirect bandgap materials and 0.5 for direct bandgap materials. g was calculated to be 1.48 eV for n = 0.5 (Figure 2(f)), which is close to the value reported in previous work (1.36 eV) [20]. In total, all these spectroscopy results indicate that the synthesized materials are uniform In 2 Se 3 with good stoichiometry and optical quality. We also investigated the structure and crystal quality of the material using X-ray diffraction (XRD) and transmission electron microscopy (TEM). The XRD pattern of the In 2 Se 3 grown on the mica substrate in the confined reactor was compared with the patterns of bulk In 2 Se 3 and mica, and it was found that, except for the peaks from the mica substrate, all peaks originated from In 2 Se 3 ( Figure S6). Additionally, the In 2 Se 3 has a high crystallinity because the peaks are very sharp. The TEM samples were prepared using a modified transfer method assisted by polydimethylsiloxane (PDMS) and polymethyl methacrylate (PMMA, see details in the Supporting Information). After repeated trial and error, we developed this method to transfer 2D In 2 Se 3 onto arbitrary substrates such as plastics, SiO 2 /Si, a copper grid, and indium tin oxide (ITO) glass ( Figures S7a-d). Raman spectra show that the transferred samples have very similar spectra to the as-grown ones ( Figures S7e-i), indicating that negligible damage was done to the samples during transfer. Figure 3(a) shows a high-angle annular dark-field scanning TEM (HAADF-STEM) image of a triangular 2D In 2 Se 3 flake. The corresponding energy dispersive X-ray spectroscopy (EDS) elemental maps show the uniform distributions of In and Se atoms in the In 2 Se 3 (Figures 3(b) and 3(c)). In addition, quantitative analysis of the EDS results (Figure 3(d)) shows an In:Se atomic ratio of 2:3, in good agreement with the XPS results. have hexagonal symmetry with the same crystallographic orientation, confirming that it is a single crystal domain. All these results confirm that the 2D In 2 Se 3 grown in the confined micro-reactor is high quality and highly crystalline.
Generally, compared to monolayer semiconductors, fewlayer ones with direct and appropriate bandgaps are promising candidates for photodetectors due to their enhanced light absorption, lower energy loss during photo-electron conversion, and better carrier transfer between source and drain [41]. TMDCs like MoS 2 cannot satisfy all the requirements at the same time because it has an indirect-to-direct bandgap transition when the number of layers decreases from a few layers to a monolayer, so that few-layer MoS 2 has a low photo-electron conversion efficiency and monolayer MoS 2 has weak light absorption. [39,42] Hence, for MoS 2 and many other TMDCs, a sacrifice of properties is unavoidable to get a balance between light absorption and photo-electron conversion efficiency [43]. Fortunately, In 2 Se 3 is a direct bandgap material regardless of its thickness and shows clear advantages in optoelectronics [22,23]. Therefore, we fabricated two terminal devices on a flexible mica substrate using 2D In 2 Se 3 with large domain sizes and high quality as the channel material and studied their photoresponse behavior (Figures 4(a) and 4(b)). Several key parameters including dark current ( dark ), rise time ( r ), decay time ( d ), responsivity (R), detectivity ( * ), and EQE were systematically investigated under 660 nm incident light. Figure 4(c) shows that the photocurrent ( ph defined as ph = light − dark ) increases with the incident light power, and the largest on/off ratio reaches 460. Unlike other narrow bandgap semiconductors, In 2 Se 3 shows a large photo-induced on-off ratio, due to its appropriate bandgap and high quality, i.e., lack of defects. In order to investigate the response speed and stability of the photodetector, time-resolved photoresponse measurements were performed by turning on/off the incident light with a chopper, while a high-speed oscilloscope was used to monitor the device current. As shown in Figure 4(d), the In 2 Se 3 photodetector remains stable under several on/off light switching events (660 nm incident light at ds = 1 V) with a nearly constant "on-state" current of ∼25 nA (Figure 4(d)), with r and d calculated to 140 ms and 25 ms, respectively (Figures 4(e) and 4(f)). We note that the response time is much slower than commercial Si based photodetectors (∼5.9 s). Further engineering and optimization of the quality of materials and interface cleanness of devices should improve the response time. Note that, after bending 1000 times, the current remains steady with a retention of >80% (with an "on-state" current of ∼ 20 nA, Figure 4(d)). We also shined 850 and 940 nm incident light on the device and it showed good stability (Figures S8a and S8b). It is well known that   most semiconductors are sensitive to visible light, but few show an appropriate responsivity and response speed to NIR light. These results suggest that the In 2 Se 3 is a promising candidate for high-performance photodetectors in the UV-Vis-NIR region. Meanwhile, when we changed the power of the incident 660 nm light ( in ), different ph values were obtained ( Figure S9). The relationship between ph and in is fitted by ph = a . In our experiments, the parameters a and were calculated to be 0.89 and 0.68, respectively (Figure 4(g)). Moreover, as shown in Figure 4(h), we obtained a remarkable responsivity of 5.6 A/W (under P = 10.2 W cm −2 , ds = 1 V), which was calculated from the following formula: where is the responsivity and is the effective area of the photodetector. This responsivity is 10 3 times higher than the reported value of multilayer MoS 2 -based phototransistors (around 7.5 × 10 −3 A/W), presumably because few-layer MoS 2 has an indirect bandgap [37]. In addition, based on the following formula, where * is detectivity and e is the charge of an electron, we calculated the detectivity and found an identical trend with responsivity. Impressively, its maximum * reaches 7 × 10 10 Jones (under = 10.2 W cm −2 , ds = 1 V) as shown in Figure 4(h), and this value is three orders of magnitude higher than that of MoS 2 photodetectors [37]. This detectivity is superior to the vast majority of reported 2D materialbased phototransistors. Finally, we calculated EQE based on the different light photoresponses ( Figure S10), using the following: where ℎ is Planck's constant, c the velocity of light, and the wavelength of the incident light. The EQE of the In 2 Se 3 photodetector ( ds = 1 V) was calculated to be 1135% at 365 nm, 495% at 445 nm, 127% at 520 nm, 154% at 590 nm, 251% at 660 nm, and 18% at 850 nm. These results suggest that few-layer In 2 Se 3 grown on mica, as a direct bandgap semiconductor, is suitable for use in flexible broadband photodetectors. A comparison of the performance of photodetectors using our 2D In 2 Se 3 and other 2D materials is shown in Table 1. Overall, the combination of high responsivity, high detectivity, and flexibility makes In 2 Se 3 a promising material for flexible broadband photodetectors. We believe that the large size, high quality, and thin thickness of the In 2 Se 3 can prolong the photo-excited carrier lifetime and result in a high photoresponsivity. In addition, the use of a mica substrate which has an atomically flat surface may reduce trap states at the interface between the In 2 Se 3 and the mica substrate, leading to long photo-excited carrier lifetime. As shown from the formula R ∝ ph ∝ Δ = q( n + p )(Δp) 0 exp(−t/ ), it is clear that a longer photo-excited carrier lifetime ( ) leads to improved photoconductivity ( ), a larger photocurrent, and better responsivity.

Discussion
In summary, we designed a confined micro-reactor that greatly reduces the size of the growth space and thus the nucleation density of 2D materials. As a result, we achieved the growth of 2D In 2 Se 3 with large domain sizes and high quality. Because of the large domain size, high quality, and layer-independent direct bandgap, we have been able to fabricate In 2 Se 3 -based flexible photodetectors, which have a fast response speed, and high responsivity, detectivity, and EQE. The strategy used here could potentially shed light on the growth of other 2D materials, facilitating their application in a wide-range of devices. Vapor Phase Growth of 2D In 2 Se 3 . In our experiments, growth was conducted in a homemade atmospheric pressure vapor deposition furnace equipped with a 1-inch diameter quartz tube (TF55035C-1, Lindberg/Blue M). The quartz boat containing In 2 Se 3 powder was put at the center of the furnace, and two slices of freshly cleaved mica were placed downstream (8-10 cm), stacked face-to-face, and served as a confined micro-reactor for 2D In 2 Se 3 growth. The furnace was heated to the growth temperature of 850 ∘ C, which was determined by the thermo-gravimetric analysis ( Figure S3), with a ramp rate of 30 ∘ C min −1 , and kept there for 5-30 min for the growth. Ar was introduced during the ramping and growth periods at a flow rate of 50-100 standard cubic centimeters per minute (sccm, with a purity of 99.99%). After growth, the furnace was cooled to room temperature under 50 sccm Ar. In the controlled growth experiments, 2D In 2 Se 3 was grown on a freshly cleaved mica substrate placed in the same location but without the other mica sheet.

Materials and Chemicals. In
Transfer of As-Grown 2D In 2 Se 3 onto a TEM Grid. The TEM samples were prepared by transferring 2D In 2 Se 3 using a PMMA and PDMS assisted transfer method [44]. First, PMMA solution was spin-coated onto the mica substrate with the grown In 2 Se 3 (3000 rpm for 1 min). Second, the substrate was heated in air at 170 ∘ C for 5 min to form a PMMA film which served as a supporting layer and protected the In 2 Se 3 in the following steps. Third, the PDMS tape was placed on the PMMA and the substrate was heated at 180 ∘ C for 5 min to make a strong bond between the PDMS and PMMA. Fourth, the PDMS/PMMA/In 2 Se 3 was peeled off the mica substrate. Due to the hydrophobicity of PMMA and hydrophilicity of the mica, some water was introduced at the PMMA/mica interface to assist the separation of PMMA/In 2 Se 3 and the mica substrate during the peelingoff. Fifth, the PDMS/PMMA/In 2 Se 3 was attached to target substrates, followed by removing the PDMS tape. Finally, the PMMA was removed by hot acetone and the transferred In 2 Se 3 sample was dried naturally in an ambient environment for further characterization.
Characterization of 2D In 2 Se 3 . The side view of the space between the two mica sheets was checked by SEM (Hitachi SU8010, Japan). Optical images of the In 2 Se 3 crystal were taken using an optical microscope (Carl Zeiss Microscopy, Germany). The thickness of the In 2 Se 3 was determined by AFM (tapping mode, Bruker Dimension Icon, Germany). Raman spectroscopy was performed under a 532 nm laser excitation (Horiba LabRAB HR800, Japan). The laser spot was 1 m and the laser power on the sample surface was less than ∼100 W. Structural and chemical analyses of the samples were performed by XRD (Cu K radiation, = 0.15418 nm, Bruker D8 Advance, Germany), XPS (Thermo Scientific K-Alpha XPS, using Al (K ) radiation as a probe, USA), and TEM (FEI Tacnai F30, 300 kV acceleration voltage, USA) with an attached EDS unit. UV-Vis-NIR absorption was conducted to study the continuous In 2 Se 3 films (Perkin-Elmer Lambda 950 spectrophotometer, USA).
Device Fabrication and Measurements. The as-grown In 2 Se 3 sample was aligned with a shadow mask and titanium/gold (Ti/Au, 5 nm/50 nm) electrodes were then made by electron beam evaporation. Electrical measurements were conducted under a microprobe station and semiconductor property analyzer (Keithley 4200 SCS, USA) under ambient environment at room temperature. LEDs with different wavelengths ranging from 365 nm to 940 nm were used as incident light during photodetection measurements (CEL-LED535, LED Multiband and High-power Supply, China). Figure S1. Crystal structure of layered In 2 Se 3 with each layer composed of Se-In-Se-In-Se atomic sheets. Figure  S2. Schematics showing viscous laminar flow in the vapor deposition process and related physical parameters. Figure  S3. Thermo-gravimetric analysis of the In 2 Se 3 source in an Ar atmosphere, and the temperature window for the growth of 2D In 2 Se 3 in this work is shown by the green region. Figure S4. AFM images of the as-grown 2D In 2 Se 3 on mica by confined growth. Figure S5. Survey XPS spectrum of asgrown 2D In 2 Se 3 on mica. Figure S6. XRD patterns of 2D In 2 Se 3 grown on mica (red) with reference patterns from a blank mica substrate (blue), bulk In 2 Se 3 (green), and a simulated diffractogram (black). Figure S7. PDMS assisted transfer of 2D In 2 Se 3 from a mica substrate onto different substrates. Figure S8. Time-resolved photoresponse of the 2D In 2 Se 3 photodetector under 850 nm and 940 nm light. Figure S9. I-V curves of the 2D In 2 Se 3 photodetector under 660 nm incident light with different power values. Figure S10. I-V curves of the 2D In 2 Se 3 photodetector under different incident light wavelengths. (Supplementary Materials)