Methods and optoelectronic device applications of semiconductor epitaxy assisted by two-dimensional van der Waals materials

ABSTRACT

This review summarizes research activities on two-dimensional (2D) materials-assisted epitaxy of inorganic semiconductors and their optoelectronic device applications. We presented the overall research related to the growth of epitaxial semiconductor layers on 2D van der Waals materials and discussed various methods to perform controlled growth of semiconductor nanostructures and microstructures on 2D layers. The 2D layers’ benefits in semiconductor technology and their role in non-destructive micro-crystallographic analysis, integration of heterogeneous semiconductor devices, and the fabrication of detachable semiconductor devices. This is followed by an in-depth discussion on 2D materials-assisted growth and non-destructive transfer of random and regular semiconductor arrays and their applications in free-standing, flexible, individually addressable semiconductor optoelectronic devices, including micro display applications. The research covered in this review promotes novel applications of semiconductors grown on 2D materials and exploits ways to synergistically combine their functionalities.

KEYWORDS:

• Semiconductor/2D heterostructure
• Non-destructive microstructural characterization
• Device integration
• 2D-assisted semiconductor lift-off
• Flexible semiconductor devices

1. Introduction

Fabricating flexible and wearable displays and/or sensors based on inorganic semiconductors is in high demand to realize future optoelectronics, including micro displays on curved eye gears [1], biomedical sensors on human or artificial skins [2–5], and infrared detectors for drones and rockets [6–8]. Semiconductor optoelectronic devices, such as light-emitting diode (LED) and laser diode (LD), have demonstrated superior characteristics over their organic counterparts in terms of high carrier mobility, high recombination rates, low power consumption, and long-term device stability and reliability [9–13]. Furthermore, these inherent characteristics are crucial in developing micro- and nano-optoelectronics where the single pixel becomes more compact to achieve higher spatial resolution while the micro-sized pixels should meet the requirements of device performances [14,15]. Still, current flexible optoelectronics have been widely studied and commercialized by organic materials because flexibility, wearability, and multiple-wavelength integration are relatively easier to achieve with such devices [16–26]. Nevertheless, semiconductor optoelectronics are gaining device transferability and flexibility by introducing two-dimensional (2D) layers into the semiconductor epitaxy. The main strategy for releasing semiconductor devices involves minimizing the strong bonding between the semiconductor overlayers and the growth substrate using the 2D van der Waals interlayer. Here, we examine techniques for producing epitaxial semiconductor materials in 2D, creating semiconductor devices that can be transferred, made flexible, addressed individually, and integrated with multiple colors.

Semiconductor epitaxy, originally developed to invent high-efficiency optoelectronic and electronic devices, can fabricate single crystalline semiconductor heterostructures and quantum structures with low defects and impurities. However, the conventional process of semiconductor epitaxy on lattice-matched 3D bulk substrates results in semiconductor devices that are brittle and rigid due to strong bonding. Thus, previous efforts to achieve freestanding and transferable semiconductor layers, such as laser-lift-off and chemical etching of sacrificial layers, required repeatable and time-consuming fabrication processes [27–30] and often damaged the epitaxial semiconductor thin film [31]. To overcome these issues, 2D van der Waals materials, such as graphene and hexagonal BN (h-BN), are introduced in semiconductor epitaxy. The 2D materials are mechanically strong, flexible, and thermally stable at high temperatures [32–37]. Many other interesting features of 2D materials have been extensively studied, which can also be utilized in conventional semiconductor technology. These include single-crystal growth and the capability to be produced on a large scale with atomic-scale thickness [38–41]. More importantly, epitaxial semiconductor layers can be grown on 2D layers directly or remotely [42–51], and the underlying van der Waals 2D materials enable the mechanical separation of semiconductor overlayers with a fast and nondestructive method.

This article describes various methods to introduce 2D materials in semiconductor epitaxy and provides an overview of progress in semiconductor/2D heterostructures and their device applications. The core topics covered in this article are also presented graphically in Figure 1. The review begins by comparing epitaxial growth of semiconductors on 3D bulk substrates, 2D materials, and 3D epitaxial substrates with 2D interlayers. The position and morphology control of microstructures and nanostructures fabricated on 2D layers is important for practical optoelectronic and electronic device applications. Therefore, the review discusses various methods currently used for controlled growth on 2D materials, which is followed by outlining the noticeable advantages that can be harvested by inserting 2D interlayers in semiconductor epitaxy. Here, non-destructive microstructural analysis of the initial stages of the semiconductor epitaxy using 2D materials, vertical integration of semiconductor devices using freestanding 2D substrates, and fabrication of detachable semiconductor devices using 2D-assisted lift-off are detailed. Then, the discussion moves into specific optoelectronic device applications of the semiconductor/2D heterostructures, with LEDs chosen as a representative example. The device transferability and flexibility assisted by 2D materials are first demonstrated using random semiconductor nanostructures and microstructures fabricated on 2D film. Next, the analysis focused on fabricating flexible and individually addressable micro-LED pixel arrays, utilizing position-controlled semiconductor devices on 2D materials. Moreover, high-density multi-wavelength LED pixels can be integrated into micro display applications using the semiconductor/2D heterostructure device. Both bottom-up and top-down approaches are utilized for the integration, with the semiconductors readily freestanding and flexible. Semiconductors/2D heterostructures can be used to fabricate high-performance LEDs, and epitaxially grown semiconductors could serve as essential building blocks for various optoelectronic applications not limited to LEDs.

Figure 1. Summary of the key aspects of semiconductor/2D heterostructures in this review.

2. Semiconductor growths on 2D van der Waals materials

Graphene and h-BN have been extensively investigated as 2D substrates for fabricating semiconductor/2D heterostructures due to their high-temperature tolerance and high chemical stability in reactive environments [43,44,52–59]. Numerous reports explained the growth of semiconductor layers, such as ZnO, GaN, GaAs, InAs, and Si nanostructures, on 2D materials like graphene or h-BN [60–67]. Particularly, 1D-like nanorods, nanowires, and nanoneedles on 2D materials have attracted significant interest due to their potential applications in flexible electronics and photonics [68–71]. This section introduces the growth of semiconductor layers on 2D materials and associated aspects by comparing the semiconductor layer growths on conventional 3D substrates.

The most intuitive method to integrate 2D layers into the semiconductor devices can be epitaxial growth since the 2D crystals are composed of atoms of periodic symmetry, similar to those of typical semiconductor 3D crystal substrates. For the epitaxial growth of semiconductor layers on 2D van der Waals materials, the 2D films are typically prepared on a rigid supporting substrate because the 2D materials are extremely thin and deformable. Despite the supporting substrates being poly-crystal or amorphous, the semiconductor overlayers establish an epitaxial relationship with the 2D layers.

Nevertheless, the surface reactivity of the 2D is much weaker than conventional 3D growth substrates. The 3D substrates, such as sapphire and Si, are prepared by cleaving single crystal ingots, which generates numerous unsaturated bonds on the substrate surface, as schematically shown in Figure 2a. Additionally, the miscut of the substrate yields many atomic cliffs. Therefore, the surface of conventional substrate possesses high chemical reactivity for semiconductor growth. Figures 2b and 2c respectively present the plan-view and tilted-view scanning electron microscope (SEM) images of ZnO nanorods grown on c-Al₂O₃ substrates through metal–organic vapor phase epitaxy (MOVPE), showing high-density growth [72]. An almost identical result of high-density growths of ZnO nanorods on Si substrates was also reported [73] where the growth density was estimated to be 10¹⁰ cm⁻² for both 3D crystal substrates.

Figure 2. Epitaxial growth of semiconductors on 3D and 2D materials. (a) Schematic illustration of a conventional substrate showing unsaturated bonds and atomic cliffs. (b) Plan-view and (c) tilted view SEM images of ZnO nanorods grown on c-Al₂O₃. Reprinted from [Park et al., Appl. Phys. Lett. 80, 22 (2002)] with the permission of AIP Publishing. (d) Schematic illustration of multiple stacks of 2D sheets in pristine condition. SEM images of ZnO nanorods grown on (e) pristine graphene layers and (f) CVD h-BN. The inset represents a high-magnification image of ZnO nanorods. Reprinted from [Kim et al., Appl. Phys. Lett. 95, 213101 (2009)] with the permission of AIP Publishing and from [J. Appl. Phys. 130, 223105 (2021)] with the permission of AIP Publishing. (g) Schematic of semiconductor epitaxy on 2D with underlying 3D epitaxial substrates. (h) Top view and (i) cross-sectional view SEM images of ZnO microrods grown on graphene-coated a-ZnO and c-ZnO substrates, respectively. The inset in (h) shows the graphene-coated a-ZnO substrate before the ZnO microrod growth. Reprinted from [Jeong et al., Nanoscale 10, 22970 (2018)] with the permission of RSC Publishing.

Meanwhile, the growth of high-density semiconductor nanostructures or full-coverage epitaxial semiconductor films on 2D films is not as typical as those on the 3D substrates. As schematically shown in Figure 2d, 2D materials comprise vertical stacks of 2D monolayers held together by van der Waals interactions, while the monolayer 2D sheet is a lattice array of strongly bonded atoms. Unsaturated bonds, capable of capturing adatoms for semiconductor nucleation, are formed by atomic cliffs or voids in 2D materials. Hence, only a few such unsaturated bonds are expected on the surface of pristine 2D films. Compared to conventional 3D substrates, 2D materials exhibit poor chemical reactivity for semiconductor growth.

Figure 2e shows the growth behavior of semiconductor layers on the 2D materials. The pristine graphene layers were prepared by mechanical exfoliations from the graphite powder, and then high-quality single crystalline ZnO nanostructures were grown on graphene layers by MOVPE [52]. However, growth density is significantly lower compared to that of ZnO nanorods grown on the 3D crystal substrate, strongly indicating inferior nucleation characteristics on the pristine 2D substrates. A nearly identical growth behavior was observed for ZnO nanostructures grown on mechanically exfoliated h-BN [58]. Figure 2e also displays a relatively high density of ZnO nanorods formed along the step edges of the graphene layers, implying the enhanced nucleation at the unsaturated bonds of graphene. Motivated by this feature, the density and position-controlled growth of semiconductor layers on 2D films were further demonstrated in the next section.

Figure 2f reveals ZnO nanorods grown on chemical vapor deposited (CVD) h-BN sheets, where the h-BN layer was synthesized on a Cu foil and then transferred onto SiO₂-coated Si substrates for the ZnO growth [59]. Compared to growth on the cleaved pristine 2D materials, a much higher density of ZnO nanostructures was grown on the CVD h-BN. The growth of CVD h-BN layers involves multiple and random nucleation and island coalescence, which leads to many atomic step edges in the 2D layers. So, the surface of CVD h-BN appears readily rough on an atomic scale, which was also confirmed by atomic force microscopy. Various semiconductors, including groups II−IV and III−V, exhibited relatively high growth densities on CVD graphene and h-BN [44,55,59,74]. Although the growth density was not as profound as seen on conventional semiconductor substrates, the limited chemical reactivity of pristine 2D film was partially alleviated when using CVD 2D film. Related studies of semiconductor growths on CVD 2D films were also reported using various growth apparatuses, including GaN [75–79], InP [80], InAs [81,82], Ga_xAs_1−xSb [83], Al_xGa_1−xN [84] and In_xGa_1−xAs [85] nanostructures grown on CVD 2D layers. More examples of semiconductor/2D heterostructures prepared by direct heteroepitaxy are summarized elsewhere with details [65,86,87].

Without using the epitaxial growth on 2D, high-quality semiconductor layers can still be grown on 2D materials by employing a single-crystal epitaxial substrate underlying the 2D materials, as indicated in Figure 2 g. An epitaxial relationship can form between the semiconductor overlayer and the supporting substrate using the extremely thin 2D gap, a technique referred to as remote epitaxy [49,88–90]. While it is possible for a polycrystalline 2D film to facilitate the growth of single-crystal semiconductor overlayers, the material choice and thickness of the 2D layers are critical. This critical 2D thickness depends on the polarity of the wafer and/or the epilayer. Many theoretical and empirical demonstrations of single-crystal-substrate-assisted epitaxial growths of semiconductors on 2D films have been reported for homoepitaxy [4,91–93] and heteroepitaxy [94–102], with a comprehensive list of materials and processes for remote epitaxy presented recently [49]. There is an argument that the micro- and/or nano-sized voids of the 2D film played a considerable role in transferring crystal information for the epitaxy [103,104]. A key agreement, however, is that semiconductor overlayers are readily mechanically detachable due to the underlying van der Waals 2D materials.

Figures 2h and 2i demonstrate the ZnO microrods grown on a-plane and c-plane ZnO films with an interlayer of graphene, respectively [91]. Polycrystalline CVD graphene layers were employed for the demonstration. Regardless of the CVD graphene crystallinity, the growth direction of ZnO microrods aligned with the crystal orientation of the underlying ZnO film, implying the establishment of an epitaxial relationship between the ZnO microrods and substrates. The studies further stated that the density of micro rods significantly decreased when increasing the epitaxial gap using thicker graphene layers because strong electric fields from the substrate can only be transferred through thinner graphene, which eventually attracts more atoms for a higher density growth of microrods.

In direct heteroepitaxy, since the epitaxial relation occurs between the semiconductor and the 2D layers, the underlying supporting substrates can be amorphous and/or polycrystal. In the meantime, preparing single crystal 2D layers on a large scale is necessary for practical device applications. Various materials of semiconductor nanorods can be grown on graphene without much considering the material compatibilities, but they are strongly affected by the growth of continuous thin films on graphene, similar to the heteroepitaxy on 3D bulk substrates, limiting the material choice. In remote epitaxy, the 2D film can be polycrystal to grow a single crystal semiconductor overlayer. Although using 3D single-crystal substrates is necessary, they can be reusable. Still, accurate control of the 2D film thickness is an issue to gain industrial scaling up of this technology. Damages to the 2D layer and/or the underlying material during the semiconductor growth can fail the remote epitaxy. However, to overcome these challenges, a comprehensive investigation of the parameters that affect these 2D-assisted growth processes is in progress, and it is expected that solutions will be developed in the future.

3. Controlled growth of semiconductor layers on 2D

To fabricate complex and high-density semiconductor devices, the position-controlled growth of semiconductor nanostructures and/or microstructures is often necessary [105,106]. This subsection provides several effective methods of controlling the distribution, density, and morphology of semiconductor overlayers on 2D films. The overall strategy involves engineering the inert surface of the 2D films to attain selective growth of semiconductor layers.

First, we highlight the density controls of the semiconductor layers on 2D films. As aforementioned, the amount of step edges on the 2D film strongly influences the growth density. On the inert surface of pristine 2D layers, numerous artificial step edges can be created using plasma treatments to enhance growth density, as schematically shown in Figure 3a(i) [42]. Figure 3a(ii) details the surface morphology of the ZnO nanostructures grown on the plasma-treated graphene layers. Due to the drastically increased number of nucleation sites resulting from plasma treatment, interconnected ZnO nanowalls were formed with high density. Because of the high-density nucleation and growth feature, uniform and full-coverage GaN films can also overgrow on the ZnO-coated graphene layers, as shown in Figure 3b [42].

Figure 3. Controlled growth of semiconductor structures in 2D. (a) (i) Schematic of high-density growths by generating artificial step edges on pristine 2D layers and (ii) SEM image of high-density ZnO nanowalls formed on the plasma-treated graphene layers. From [Chung et al., Science 330, 6004 (2010)] Reprinted with permission from AAAS. (b) Uniform and full coverage GaN thin film grown on the ZnO-coated graphene layers. From [Chung et al., Science 330, 6004 (2010)], reprinted with permission from AAAS. (c) (i) Schematic of the selective-area growth by activating preferable nucleation sites on pristine 2D layers and (ii) SEM image of ZnO nanotubes fabricated on graphene layers using the corresponding method. The inset shows a magnified image of the ZnO nanotube. Reprinted from [Kim et al., Adv. Mater. 24, 41 (2012)] with the permission of John Wiley and Sons. (d) (i) Schematic of position-controlled growth on 2D layers using a conventional mask layer and (ii) the ZnO nanotubes selectively grown on CVD graphene films using the corresponding method. The inset shows the magnified image of the nanotubes. Reprinted from [Park et al., APL Mater. 4, 106104 (2016)] with the permission of AIP Publishing. (e) (i) Schematic of the selective-area growth using patterned 2D arrays and (ii) the GaN microdisk arrays selectively grown on graphene dots using the corresponding method. The inset displays the magnified image of the GaN microdisk. Reprinted (adapted) with permission from [Baek et al., Nano Lett. 13, 6 (2013)]. Copyright {2013} American Chemical Society. (f) (i) Schematic of the selective-area growth using remote epitaxy with thickness-controlled 2D patterns and (ii) the ZnO micro rods selectively grown on GaN/sapphire substrate with patterned graphene interlayer using the corresponding method. Reprinted (adapted) with permission from [Jeong et al., ACS Appl. Nano Mater. 3, 9 (2020)]. Copyright {2020} American Chemical Society.

The inert surface of the 2D layer also offers a distinctive method to precisely control the position of the semiconductor microstructures and nanostructures on it by creating artificial step-edges at particular locations. Preferred nucleation sites can be activated by selective area plasma treatment, as specified in Figure 3c(i). The method involves a sequence of steps of preparing pristine graphene films, selective-area hole patterning using lithography, plasma treatment at the hole openings, and preferable nucleation and growth of the semiconductor overlayers on the plasma-treated 2D layers. The SEM image in Figure 3c(ii) shows the ZnO nanotube arrays grown on graphene layers via the selective-area plasma treatment [56]. The accurate position control of the ZnO nanotube arrays and the identical surface morphology of each ZnO nanotube suggest the excellence of the growth selectivity. Furthermore, since the growth selectivity was obtained on the plasma-treated region over the pristine graphene surface, no additional growth masks, such as SiO₂ or Si_xN_1−x layers, are necessary. Employing a similar strategy, ZnO nanostructures with controlled shape and morphology were fabricated on h-BN insulating layers [58].

Since CVD 2D films inherently possess a relatively high density of step edges and/or grain boundaries, a method to suppress the nucleation sites by a mask layer is more rational to grow semiconductor microstructures and nanostructures selectively on the CVD 2D film. Figure 3d(i) shows the corresponding method. The CVD 2D film is prepared on a supporting substrate. Then, a mask layer was patterned onto the CVD 2D film using a typical lithography process. The growth selectivity occurs on the 2D films over the mask layer. Figure 3d(ii) shows the position-controlled ZnO nanotube arrays grown on a CVD graphene film using a SiO₂ mask layer [74]. The ZnO nanotubes exhibited regular and uniform arrays with almost identical morphology. In addition to ZnO nanotubes, position-controlled GaN microstructure [107–109], InAs nanowire arrays [110], and Al_xGa_1−xN nanopyramids [111] were fabricated on CVD 2D films using the identical method. The idea also applies to fabricating flexible pressure sensors using ZnO nanotube arrays grown on graphene films [112].

Figure 3e(i) presents an alternative approach to growing regular semiconductor arrays using the growth selectivity between the 2D layer and the masking layer [75,113]. However, in this case, the underlying supporting substrate acted as a growth mask, while the patterned CVD 2D arrays serve as the nucleation sites. First, full-coverage large-size CVD graphene films are prepared on non-epitaxial supporting substrates, such as amorphous SiO₂, then the CVD graphene films can be patterned into discrete dot arrays using lithography and plasma etching. Figure 3e(ii) demonstrates the experiment in this selective-area growth where well-ordered GaN microdisks were grown on patterned graphene dots [75]. In addition to the position controls, the sizes of individual GaN microdisks were measured to be more than twice the size of graphene dots. This suggests epitaxial lateral overgrowth (ELOG) of the GaN microdisk. The same approach can be further extended to include controlled growth of other semiconductor microstructures and nanostructures on patterned graphene arrays. Examples include ZnO nanorods [114,115] and GaN nanowires [116].

In the case of remote epitaxy, nucleation on 2D materials occurs randomly, resulting in semiconductor growths without any precise spatial control. However, the remote epitaxy is significantly influenced by the thickness of 2D interlayers, providing a method to control growth density in specific regions. As shown in Figure 3f(i), the 2D layers can be prepared with two different thicknesses. Regions with thinner 2D films, such as monolayer graphene, enable remote epitaxial growth, while regions with thicker 2D films, such as multilayer graphene exceeding the critical thickness, inhibit remote epitaxy. In this way, the nucleation and crystal growth are confined to specific areas.

The corresponding experimental demonstration of the 2D thickness-dependent selective area growth is shown in Figure 3f(ii) [117]. The predominant nucleation and epitaxial growth primarily occurred on single-layer graphene, resulting in selective-area growth of vertically aligned ZnO micro rods. The growth selectivity was achieved using single-layer-graphene hole openings over four-layer-graphene masks. The underlying substrate was GaN films grown on c-Al₂O₃ where GaN films are suitable for the epitaxial growth of ZnO due to the same wurtzite crystal structure with a small lattice mismatch within 2%. For comparison, the same selective-area growth method was applied to non-epitaxial supporting substrates of SiO₂. Due to the amorphous nature of SiO₂ and poor growth selectivity between single-layer graphene and multilayer graphene, ZnO microrods were randomly grown on the 2D films, regardless of the 2D film thickness, strongly suggesting the selective-area growth by remote epitaxy.

4. Advantages of fabricating semiconductor/2D heterostructures

A new class of material system is developed by growing semiconductor layers of various dimensions on 2D materials and reducing the cost of using 3D single-crystal substrates normally required for epitaxial growth [60]. In these hybrid material systems, 2D layers such as graphene while acting as flexible and transparent electrodes [118] can simultaneously be utilized in various unconventional electronic and optoelectronic devices, including sensors [119–121], solar cells [122], and flexible displays [123]. Additionally, the hybrid structure can benefit from the synergistic characteristics of the constituent material systems. In this subsection, we will explore several noticeable advantages that can be exploited by inserting 2D interlayers in the semiconductor epitaxy. This includes non-destructive crystallographic analysis of the crystal growths, integration of heterogeneous semiconductor devices on double-sided 2D crystals, and fabrication of detachable inorganic devices.

4.1. 2 2D material-assisted non-destructive microstructural analysis

The idea of using 2D layers for the semiconductor epitaxy can be extended for the microstructural characterization. In particular, one of the most fundamental topics in material science is the study of the nucleation and crystal growth that forms nanomaterials [124–126]. Physical properties of the nanomaterials are significantly affected by the growth mechanism and defect formation and the ability to observe the arrangement of atoms in as-grown nuclei would help us investigate them [127,128]. High-resolution transmission electron microscopy (HR-TEM) imaging has traditionally been used to perform atomic-resolution analysis of the nucleation and initial growth. However, the destructive nature of the conventional TEM sampling process makes probing of the nanomaterials at atomic scale difficult [129]. This problem can be resolved by using graphene films as a substrate for nanomaterial growth and a supporting layer for TEM measurements, as suggested in Figure 4a [130]. Graphene films can be fully coated on TEM grids with good mechanical strength. Semiconductor nanomaterials grown on graphene remained stable after the growth and under TEM measurements. No destructive conventional TEM samplings were required due to the excellent electron beam transparency of graphene. Accordingly, direct imaging of the evolution of nanomaterials on graphene allowed the examination of the initial growth behavior with atomic resolution and high sensitivity.

Figure 4. Non-destructive analysis of the semiconductor nanomaterials using 2D substrates. (a) Schematic of the observation of ZnO nanomaterials using TEM. HR-TEM images of (b) the 10s-grown ZnO nucleus and (c) the ZnO nucleus after heat treatments of 6 min. Insets in (b) and (c) show corresponding FFTs and zone axes where the diffraction peaks of ZnO and graphene are indicated with green and yellow circles, respectively. Reprinted from [Jo et al., Adv. Mater. 26, 13 (2014)] with the permission of John Wiley and Sons.

For the demonstration, ZnO nanomaterials were grown directly on a graphene-coated TEM grid and their initial growth behavior was observed using HR-TEM [130]. This atomic resolution imaging enabled a detailed investigation into the crystal structure transition of the ZnO nucleus during the early growth stage. Figure 4b shows the ZnO nuclei grown for 10s, which exhibit rounded shapes, incomplete crystalline atomic arrangements, and often possess a cubic rocksalt structure. Contrarily, the ZnO nuclei grown for 1 min exhibited only a hexagonal wurtzite structure. The appearance of rocksalt ZnO structure during the initial growth was surprising, as previously, this structure appeared only under specific growth conditions [131,132]. Therefore, the transformation of rocksalt ZnO into wurtzite ZnO was further investigated. After the growth of ZnO for 10s, the sample was kept in the MOVPE reactor for another 6 min at the same growth temperature but without the precursor gas. Figure 4c indicates that the entire rocksalt ZnO structure disappeared and only the nuclei with the wurtzite structure and well-defined facets were observed. The corresponding Fast Fourier Transform (FFT) and zone axis are provided in the insets of Figures 4b and 4c, where the diffraction peaks of ZnO (green circles) and graphene (yellow circles) are parallel, implying the epitaxial growth. This nondestructive characterization also allows for the observation of the time evolution of the metastable rocksalt structure of ZnO transformation into the thermodynamically stable wurtzite structure. Various aspects related to nucleation and growth of nanomaterials on graphene can be investigated using graphene as a substrate for nanomaterial growth and a supporting layer for TEM measurements.

4.2. Integration of heterogeneous semiconductor devices on double-sided 2D materials

Various semiconductor materials have been epitaxially grown on graphene layers, while the suspended graphene, with consistent crystallinity on sides, enables the integration of heterogeneous semiconductor devices on the opposite sides of the graphene layers. The heterogeneous devices are already electrically connected by the graphene. The density and/or position of each heterogeneous device can be controlled independently. For example, we introduce the heteroepitaxial growth of single-crystalline wide-bandgap ZnO nanotubes and narrow-bandgap InAs nanorods on the top and bottom surfaces of graphene layers, respectively [133].

To fabricate a heterostructure comprising ZnO/graphene/InAs, CVD multilayer graphene was transferred onto perforated SiO₂-coated Si₃N₄ membranes with 300 nm-diameter hole arrays. The graphene substrate comprises one side of the as-grown graphene and the opposite side of the perforated-mask-coated graphene. The ZnO nanotubes were grown on the mask-coated graphene by selective-area MOVPE. Then, the InAs nanorods were grown on the opposite side of the graphene using molecular beam epitaxy (MBE). Figure 5a details the surface morphology of the free-standing ZnO/graphene/InAs heterostructure. Regular arrays of the ZnO nanotubes were selectively grown only on the graphene layers through the SiO₂/Si₃N₄ growth mask. The diameter and spacing of the nanotubes can be controlled by the mask. At the same time, vertical InAs nanorods were grown on the other side of the graphene film, exhibiting a random distribution. The nucleation of the InAs nanorods was rarely affected by the ZnO nanotubes, as the relatively thick multilayer graphene screened the interaction between the two semiconductors.

Figure 5. Integration of semiconductor nanodevices on double-sided graphene. (a) SEM image illustrating the overall cross-sectional morphology of the InAs nanorod/ graphene layers/ ZnO nanorod nanostructure, (b) Schematic of the dual-wavelength photodetector and spectral responses of the (c) ZnO-side and (d) InAs-side photodetectors. Reprinted from [Tchoe et al., NPG Asia Mater. 13, 33 (2021)] with the permission of Springer Nature.

Using the ZnO/graphene/InAs heterostructure, a dual-wavelength Schottky photodetector was fabricated, as graphed in Figure 5b. The Au layers were coated on the tips of the InAs and ZnO nanostructures for the Schottky contacts. To electrically isolate the electrodes, polyimide (PI) layers filled the spaces on both sides of the nanostructures. Figures 5c and 5d show the spectral responsivity of the ZnO-side and the InAs-side photodetectors, respectively. The responsivity peaks were observed at 3.6 eV for the ZnO side and at 0.6 eV peaks for the InAs side, indicating the dual-wavelength sensitivity. The work also suggests that the ZnO/graphene/InAs heterostructure device can combine the unique properties of each material and capture photons over a wide spectral range. Such characteristics are essential for broadband light-harvesting device applications. Moreover, various functionally integrated devices can be exploited using a similar approach. A freestanding ZnO nanorod/graphene/ZnO nanorod heterostructure device was proposed for piezoelectric nanogenerators [134]. Si/graphene/Si planar heterostructures have also been investigated for optoelectronic device applications [135].

4.3. Mechanical lift-off using semiconductor/2D heterostructures

In addition to long-studied lift-off methods, such as laser-lift-off and selective-chemical-etching, a new semiconductor lift-off method has emerged by employing the 2D van der Waals materials into the semiconductor epitaxy. Diverse semiconductor materials have been grown on 2D layers directly and remotely, while the weak van der Waals interactions on the 2D films enable the mechanical lift-off of the semiconductor devices. In particular, 2D layers such as graphene and h-BN, exhibited excellent high-temperature tolerance during the semiconductor growth. The 2D layer already demonstrated extreme large-size scalability over the conventional growth wafers. Accordingly, these layers facilitate an instant, cost-effective, and large-scale lift-off process.

The main mechanism of the 2D-assisted lift-off involves exploiting the van der Waals interactions of the 2D film. The corresponding lift-off process is displayed in Figure 6a. Initially, semiconductor device arrays can be fabricated on the 2D film, and the 2D films are attached to the underlying supporting substrate by van der Waals interactions, where the semiconductor overlayers are ready to be detached. Before the mechanical lift-off of the semiconductor devices, typical device fabrication processes, such as insulator filling and metallization, can be performed. Then, the whole device array can be mechanically peeled from the supporting substrate. A sticky tape, placed on the semiconductor devices, can assist in mechanical lifting off and handling the free-standing semiconductor devices to transfer onto target substrates. An almost identical lift-off process can be performed for the semiconductor devices prepared on 2D layers by direct heteroepitaxy and remote epitaxy. However, there is a difference in the strength of bonding between the semiconductor and the 2D layers. For the semiconductor/2D heterostructures grown by direct heteroepitaxy, the 2D film tends to be released from the supporting substrates while maintaining their union of semiconductor/2D heterostructures. Meanwhile, for the remote epitaxy, 2D films tend to remain on the growth substrate while the semiconductor layers are separated from the supporting substrate. This is presumably because the direct heteroepitaxy involves many chances to create strong bonding between the semiconductor layer and the 2D film, whereas the adhesion of 2D layers and the supporting substrates is only van der Waals interactions.

Figure 6. 2D materials-assisted semiconductor layer lift-off. (a) Schematic of the mechanical lift-off process where semiconductor devices can be fabricated on 2D by direct heteroepitaxy and/or remote epitaxy. (b) (i) A photograph, (ii) low-magnification and (iii) high-magnification SEM images of ZnO nanotube arrays, which are grown on CVD graphene by direct heteroepitaxy and then transferred on a plastic using the mechanical lift-off. The inset in (iii) shows the excellent lineup of the ZnO nanotube after the transfer. Reprinted from [Park et al., APL Mater. 4, 106104 (2016)], with the permission of AIP Publishing. (c) (i) As-grown GaN microstructures on graphene-coated sapphire by remote epitaxy. (ii) The released bottom surface of GaN microstructures encapsulated with PI, indicating that the microstructures are fully removed from the substrate. The inset shows the sapphire substrate after the mechanical lift-off. (iii) A photograph of the GaN microstructures exfoliated from a substrate using mechanical methods. Reprinted from Nano Energy, 86, Jeong et al., Transferable, flexible white light-emitting diodes of GaN p–n junction microcrystals fabricated by remote epitaxy, 106075., Copyright (2021), with permission from Elsevier. (d) Schematic illustration of the thru-hole epitaxy process, where 2D materials act as a lift-off layer, while epitaxial growth occurs through openings within the 2D materials. This process also involves a mechanical exfoliation. (e) Plan-view optical microscope images of the (i) ZnO microrod arrays grown on planner ZnO/MoS₂/GaN/sapphire, (ii) surface of GaN/sapphire after ZnO layers were fully detached mechanically and (iii) ZnO microrods transferred onto a foreign substrate without any observable damage. Reprinted (adapted) with permission from [Lee et al., Cryst. Growth Des. 22, 12 (2022)]. Copyright {2022} American Chemical Society.

Figure 6b reveals the experimental demonstration of the mechanical lift-off of ZnO nanotube arrays, fabricated on CVD graphene films by direct heteroepitaxy [74]. (i) of Figure 6b presents a photograph of the free-standing ZnO nanotube arrays after the mechanical lift-off where punched Kapton tape was used to handle the ZnO/graphene heterostructures. (ii) and (iii) of Figures 6b show SEM images of the ZnO nanotube arrays transferred onto a 3 µm-thick plastic sheet. While the thin plastic support was deformable to bend the nanotube arrays with a bending radius of 0.5 mm, the nanotube arrays showed their excellent line-up without any observable cracks or damage (inset of Figure 6b(iii)).

An almost identical lift-off process was utilized for the semiconductor/2D heterostructures fabricated by remote epitaxy [98,99,117]. Figure 6c(i) shows an SEM image of GaN micro-LEDs fabricated on a bi-layer-graphene-coated c-Al₂O₃ substrate. Random arrays of hexagonal GaN microstructures were grown by remote epitaxy. The gap of microstructure arrays was filled with a PI layer prior to the lift-off. The PI-coated GaN microstructures were mechanically exfoliated from the sapphire substrate. (ii) of Figure 6c shows the bottom side of the detached GaN microstructure array. No mechanical deformation or cracks were observed in the PI-coated microstructures. Although the graphene layer remained on the sapphire substrate, the inset of Figure 6c(ii) infers that the microstructures are fully removed from the c-Al₂O₃ substrate after the lift-off. A photograph in Figure 6c(iii) also depicts that the GaN microstructure arrays are fully detached after the mechanical lift-off.

Even when the 2D films have micro-voids, still the 2D-assisted epitaxial growth and mechanical lift-off are valid. Epitaxial semiconductor layers can be grown on 3D epitaxial substrates through the nano- and/or micro-holes distributed in the 2D interlayer. Due to the direct growth on the 3D epitaxial substrate, epitaxy through the hole is less sensitive to the crystallinity of 2D material, the polarity of the underlying crystal, and the thickness of the 2D material. The overgrown semiconductor layers and the underlying epitaxial substrates would have strong bonding at particular regions. However, most of the heterojunction surfaces are linked by van der Waals interactions, facilitating mechanical lift-off.

The experimental demonstration of mechanical lift-off by through-hole-epitaxy is schematically shown in Figure 6d [103]. A GaN-grown c-Al₂O₃ was used as an epitaxial substrate. Perforated MoS₂ layers were transferred onto the substrate to grow full-coverage ZnO films, where small-size through-holes in the MoS₂ allow the ELOG of the ZnO film. Lastly, the vertically aligned ZnO microrod arrays were fabricated on the ZnO film. (i) of Figure 6e portrays the plan-view optical microscope image of the as-fabricated ZnO microrod arrays. Using a sticky tape, the ZnO layers can be fully detached from the GaN/c-Al₂O₃ substrate through mechanical exfoliation. The optical image of Figure 6e(ii) expresses that no ZnO microrod arrays were left on the substrate. Furthermore, the detached ZnO microrod array preserves its morphology and intervals without any observable damages, as specified in (iii) of Figure 6e. Additionally, the ZnO film was crystallographically aligned with the underlying substrate, a characteristic commonly observed in remote epitaxy. However, in the case of lift-off by through-hole-epitaxy, the MoS₂ layer can be up to twelve atomic layers thick, surpassing the critical thickness observed in remote epitaxy.

5. Optoelectronic device applications using semiconductor/2D heterostructures

Considerable research efforts have been made to realize flexible and functionally integrated inorganic devices based on semiconductor/2D heterostructures since they can be applied to the direct heteroepitaxy and the remote epitaxy with efficient exfoliation and transfer onto foreign substrates. In this subsection, we analyzed and underscored several case studies in which semiconductor/2D heterostructures have been utilized in various optoelectronic applications, with LEDs as the representative example. First, we discussed the capability of transferring semiconductor devices onto diverse substrates, utilizing the advantageous properties of graphene such as thermal tolerance, flexibility, and conductivity. Next, we investigated the fabrication of flexible devices, with a particular emphasis on heterostructures comprised of semiconductor nanostructures and/or microstructures, grown on 2D layers through direct and remote epitaxy. In the latter case, the native substrate is reused. Additionally, using nanostructures ensures that the challenges associated with materials incompatibility, device flexibility, and structural defects are significantly reduced. Next, we described the fabrication of position-controlled micro-LED arrays, which can be further extended to demonstrate high-density, free-standing, and monolithically addressable micro-LED pixels. Lastly, this subsection concluded with a discussion on microdisplay applications that employed semiconductor/2D heterostructures. Using 2D materials in layer transfer enables the vertical stacking of wafer-scale red (R), green (G), and blue (B) LEDs, allowing further fabrication to achieve high-density, full-color pixel arrays with sub-10 µm pixels. Additionally, multicolor LEDs with controlled morphology were monolithically integrated, and fabricated using micro-LEDs grown on 2D films.

5.1. Device transferability

Transferable semiconductor devices using 2D layers were first proposed and demonstrated by fabricating planer GaN LEDs on mechanically exfoliated graphene layers [42]. The fabrication and transfer of the LED/graphene heterostructures are displayed in Figure 7a. The epitaxial GaN films were grown on the plasma-treated graphene layer with a ZnO nanowall interlayer. Subsequently, a Si-doped n-GaN layer, three-period In_xGa_1−xN/GaN multiple-quantum-wells (MQWs), and a Mg-doped p-GaN layer were deposited on a flat GaN film to build the LED structure. The graphene layers underneath the n-GaN and Ni/Au bilayers deposited on the top surface of p-GaN were utilized as n-type and p-type electrical contacts, respectively. After the device fabrication, the LED/graphene heterostructure was mechanically detached from the original substrate and then transferred as a united body onto various substrates, such as glass, metal, or plastic. The as-fabricated LED and the LED transferred onto foreign substrates emitted strong and comparable light emissions, as represented in Figure 7b.

Figure 7. 2D materials-assisted device transferability. (a) Schematic of the GaN film LEDs fabricated on graphene layers by direct growth and their transfer processes on foreign substrate. (b) Light emission images of the as-fabricated LED on the original substrate and transferred LEDs onto the foreign substrates. From [Chung et al., Science 330, 6004 (2010)]. Reprinted with permission from AAAS.

This demonstration provided a good reason to employ 2D materials in the semiconductor epitaxy. Following this report, 2D-assisted transferable LEDs were demonstrated by growing LED epilayers on h-BN-grown c-Al₂O₃ where the large-size 2D films become more practical [136]. Moreover, transferable LEDs were fabricated on monolayer-graphene-grown SiC substrate [137], a result that was later revisited to explore the concept of remote epitaxy.

5.2. Flexible devices using semiconductor/2D heterostructures

The 2D-assisted lift-off technique facilitates the fabrication of flexible and wearable devices based on single-crystalline inorganic compound semiconductors, offering superior device performances compared to organic and/or amorphous devices [10]. Most especially, 1D-like inorganic structures, such as vertically aligned micro rods and nanorods, have been considered a solution to achieve device flexibility since the spatially separated micro- and/or nano-device arrays rarely get stressed or damaged under bending conditions [53,57,138–140]. Many semiconductor nano-and micro-devices have been fabricated on graphene films for various flexible device applications [57,59,109,112,141–143]. Here, the fabrication and operation of flexible nano-LEDs comprising GaN/ZnO coaxial nanorod LEDs directly grown on CVD graphene were discussed [53]. In addition, the remote heteroepitaxy of GaN microrod LEDs on c- Al₂O₃ substrate through graphene was presented [98].

To fabricate nano-LED arrays on graphene layers through direct heteroepitaxy, ZnO nanorods were initially grown on CVD graphene films, followed by a coaxial coating of LED epilayers on the ZnO nanorods. All of the semiconductor layers were grown by MOVPE. Figure 8a shows the surface morphology of the coaxial nano-LEDs. The nano-LEDs were generally fairly aligned vertically. The device density was estimated to be 10⁸−10⁹ cm⁻². The individual nanostructures were single crystalline, as confirmed by TEM. After the growth, the metallization of p-GaN and the insulator filling processes were further performed to operate the LED device. Finally, the nano-LED arrays were transferred onto a flexible Cu-coated polyethylene terephthalate (PET) substrate to evaluate the device characteristics under various bending conditions.

Figure 8. Nanorod LEDs fabricated in 2D for flexible optoelectronics. (a) High (left) and Low (right) magnification SEM images of coaxial GaN nano-LEDs fabricated on CVD graphene films. (b) Light emission image at a bending radius of 3.9 mm. (c) Integrated EL intensity and EL spectra (inset) at various bending cycles. Reprinted from [Lee et al., Adv. Mater. 23, 40 (2011)] with the permission of John Wiley and Sons.

Figure 8b describes that the nano-LED array, driven at 10 mA, consistently emitted strong blue light emission under bending conditions. The 13 mm-wide PET substrate was bent to a width of 7 mm, where the corresponding bending radius was 3.9 mm. Furthermore, distinct light emission spots were observable in the images of the nano-LED array, indicating that the nano-LEDs function as individual light emitters. The reliable operation of the nano-LEDs was examined next by repeating the device bending up to 100 times. As portrayed in Figure 8c, the integrated electroluminescence (EL) intensity and EL spectra (inset of Figure 8c) of the nano-LEDs remained nearly constant, suggesting no observable damages or fractures in the nano-LEDs.

In addition to direct epitaxy, the strategy for fabrication of deformable micro-LEDs can also be initiated with remote heteroepitaxy, as shown in Figure 9a [98]. The substrate was prepared by transferring a single-layer CVD graphene twice onto a c-Al₂O₃ wafer. The discrete microrod LEDs were fabricated using MOPVE on the graphene-coated epitaxial substrate. Every microrod comprises n-GaN, three periods of coaxial In_xGa_1−xN/GaN MQWs, and the outermost shell of p-GaN. A PI layer provided electrical isolation between electrodes and structural support to the microrod LED arrays. The p- and n- electrodes were formed on top and bottom of microrods, respectively. The fabrication process involved several sequential steps, such as PI filling, p-contact formation, mechanical lift-off, and n-contact formation. The graphene layers remained on the wafer after exfoliation, making the n-contact formation directly on the bottom of the microrod LED. Finally, the microrod LEDs were transferred onto a Cu foil to complete the fabrication.

Figure 9. Flexible LEDs and substrate recycle through remote epitaxy. (a) Schematic and (b) light emission image of the microrod LEDs grown on a graphene-coated sapphire wafer. (c) Photograph of microrod LEDs deformed in 180°-folded form. (d) Comparison of EL spectra of the micro-LEDs fabricated by using virgin (blue line) and recycled (red line) substrates. The inset photographs show light emissions from the microrod LEDs fabricated on virgin (left) and recycled wafers (right). From [Jeong et al., Sci. Adv. 6, 23 (2020)], reprinted with permission from AAAS.

Figure 9b shows the light emission image of the micro-LEDs fabricated by remote heteroepitaxy. Every single microrod LED exhibited EL emissions with uniform colors. Although the growth was random, the micro-LEDs were fairly well distributed. Furthermore, the microrod LED arrays transferred onto a Cu plate tolerated large deformations, such as twisting, inward/outward folds, and random crumpling. For example, Figure 9c shows the 180°-fold micro-LEDs without observable degradations in the EL emissions, implying that the microrod LEDs can be attached to many curved static and even moving surfaces.

Since using 3D single crystal substrates is essential for remote epitaxy, wafer recycling becomes a significant consideration. Figure 9d implies that micro-LEDs produced from virgin and recycled wafers exhibit nearly identical EL spectra in terms of intensity and wavelength, indicating the viability of substrate recycling. Raman spectroscopy confirms the presence of graphene on the Al₂O₃ wafer after the micro-LEDs have been exfoliated. Residues of PI and thermal release tape are also typically present. Before wafer recycling, residues were completely removed through wet cleaning, thermal treatment, and reactive ion etching. This ensures the wafer is clean and smooth, ready for recycling [51,93,117,144,145].

5.3. Individually addressable micro-device arrays using semiconductor/2D heterostructures

Control of the dimension and position of semiconductor devices fabricated on 2D materials enables the realization of individually operational, high-density device arrays. This feature is essential for practical and complex device applications, including those in LEDs [56,108,109,113,146], pressure sensors [112,143], lasers [75], inorganic electronics [142], as well as biomedical applications [147]. Here, we discuss the fabrication of individually addressable GaN micro-LED arrays obtained through selective-area growths on the micro-size graphene dot patterns.

Figure 10a presents the SEM image of GaN micro-LED arrays, which were fabricated on graphene microdot patterns using selective-area ELOG [79]. The as-grown micro-LEDs were fully intact and firmly attached to the supporting substrate. They display uniformity and homogeneity over a large area in terms of size and spacing. The diameter of the GaN microstructure was 10µm over the 3 µm-diameter of the graphene dots, indicating lateral overgrowth. Since each GaN micro-LED was epitaxially grown on the graphene dots using the ZnO intermediate layer, as shown in the inset of Figure 10a, the underlying supporting substrates can be polycrystal or amorphous. Furthermore, a 100-µm-square contact pad was formed on the micro-LED array to operate the devices as an ensemble. In Figure 10b, the micro-LED arrays exhibit consistent and uniform blue light emissions, without dead regions observed. Each micro-pixel is distinguishable. Additionally, the inset of Figure 10b highlights that the blue emission appears stronger from the outer edge of the micro-LED compared to the center region, likely due to improved crystallinity at the ELOG region.

Figure 10. Individually addressable LED pixel arrays fabricated on graphene dots. (a) SEM image of the GaN microdisk arrays grown on graphene dots. The inset schematically shows the heterostructure and the (b) regular light emission spots from the micro-LED array. The inset exhibits a magnified optical image of light emission. Reprinted from Nano Energy, 60, Chung et al., GaN microstructure light-emitting diodes directly fabricated on tungsten-metal electrodes using a micro-patterned graphene interlayer, 82-86, Copyright (2019), with permission from Elsevier. (c) Fabrication of individually addressable micro-LED arrays. (i) Schematic and SEM images of the (ii) top and (iii) bottom electrodes for passive matrix addressing. EL images of (d) the individually addressable micro-LEDs and the micro-LEDs at bending radius D = 1 mm. (f) I−V curves of the micro-LED at different bending conditions. Reprinted from [Tchoe et al., NPG Asia Mater. 11, 37 (2019)] with the permission of Springer Nature.

To make the micro-LED arrays individually addressable [148], a passive matrix of metal layers was applied to the top and bottom surfaces of the free-standing micro-LEDs, as schematically shown in (i) of Figure 10c. The freestanding micro-LEDs were composed of discrete micro-LED arrays with PI surroundings to form a united structure. Metal strips of Ni/Au and Ti/Au were formed in a crossbar configuration to one another, on the top p-GaN and bottom n-GaN surfaces, respectively. Single-walled carbon nanotubes (SWCNTs) were incorporated into the metal strips, serving as flexible and stretchable backbones to the electrode. Figures 10c(ii) and 10c(iii) illustrate the metal strips on the top and bottom surfaces of the micro-LED arrays, respectively, where each metal strip continuously covered only a single row or column of the micro-LED arrays.

Figure 10d reflects an optical image of the freestanding passive matrix micro-LEDs, exhibiting localized light emission at a particular position. The EL emission spots can be freely moved to a new position by selecting the respective pair of top and bottom electrode strips. The passive matrix demonstrated control over individual pixels even at a bending radius of 1 mm, as depicted in Figure 10e. Figure 10f shows the current–voltage (I-V) curves of a single micro-LED pixel at various bending radii exhibiting nearly identical rectifying behavior, suggesting that the passive matrix rarely experienced mechanical damage or fractures under the device bending.

5.4. Integration of multi-color LEDs for microdisplay applications

The fast and large-scale lift-off methods by 2D van der Waals materials enabled the more practical integration of heterogeneous devices, such as the fabrication of full-color LED chips, especially for the microdisplay applications, sub-10µm full-color pixel arrays were demonstrated by 2D-assisted wafer-scale lift-off and integration of Al_xGa_1−xAs-based R-LED and In_xGa_1−xN-based G- and B-LEDs [149]. All of the LEDs were grown on latticed matched single-crystal substrates, while the 2D interlayers of graphene for the Al_xGa_1−xAs-based LED and h-BN for the In_xGa_1−xN-based LEDs were employed for the LED-wafer release.

The freestanding RGB LEDs were further united vertically by repeatable processes of LED bonding and lithography, as illustrated in Figure 11a. The vertical stack was composed of R-LED at the bottom, G-LED in the middle, and B-LED on the top by considering the optical transparency, especially, to fabricate the 10µm-size full-color pixel, the thickness of In_xGa_1−xN LEDs was controlled to be 1−2 µm, where the reduced aspect ratio enables a reliable lithography process for integrating RGB pixels. Green-light-absorber and blue-light-absorber PI layers were employed in between R/G and G/B LEDs to prevent the crosstalk. Additionally, the PI interlayers bond each LED wafer to be united. Figure 11b shows the optical and SEM images of the well-aligned vertical full-color arrays. Since the lithography process is only required for aligning the RGB pixels, the assembly of high-density pixels can be accurate and simultaneous. In this report, high-density full-color LED pixel arrays were demonstrated, with a pixel density of 1000−5100 pixels per inch. Moreover, the full-color pixel showed good color-mixing results, as shown in the optical image in Figure 11c. The dominant EL peaks for RGB colors were observed at 665, 535, and 463 nm, respectively, which wavelengths are desirable for full-color display applications. In the CIE 1932 color space, the full-color pixel covers 99.4 and 86.9% of the sRGB and PCI-P3 color gamut, respectively.

Figure 11. Fabrication of full-color LED pixel arrays using 2D-assisted layer transfer. (a) Schematic of the RGB pixel integration. (b) Optical and SEM images of the RGB pixels and (c) EL image of the color mixing. Reprinted from [Shin et al., Nature 614, 81 (2023)] with the permission of Springer Nature.

In addition to assembling different color pixels from different epitaxial wafers, monolithic approaches can be employed for generating multiple colors. Figure 12a shows the fabrication process flow of monolithically integrated multicolor LEDs on CVD graphene substrates [108]. The method of multiple color generations was selective-area growth. A 300 nm-thick SiO₂ mask layer composed of 600 nm-diameter hole arrays with different spacing was patterned on the graphene film. The ZnO nanostructures were firstly grown only on the hole openings, then the n-GaN layers were coated on the ZnO nanostructures, preserving the growth selectivity. Meanwhile, SEM images in Figure 12a demonstrate that the n-GaN microstructures exhibit rod-shape and pyramid-shape when the hole spacing is 2µm and 6µm, respectively. To build the LED structure, the In_xGa_1−xN QWs and p-type GaN layers were conformally coated on the n-GaN microstructures. The PI insulator filling and metallization were further performed for the device fabrication. Additionally, the graphene substrate allows for the transfer of the whole micro-LED arrays onto Cu foil, demonstrating reliable LED operations under various bending conditions.

Figure 12. Monolithic integration of multi-color LEDs on 2D films. (a) Schematic and corresponding SEM images of microrod and micropyramid LEDs simultaneously fabricated on CVD graphene films. (b) A photograph of multicolor light emissions, EL spectra of (c) the microrod and (d) the micropyramid LEDs at different current levels. Reprinted from [Lee et al., Sci. Rep. 10, 19677 (2020)] with the permission of Springer Nature.

All the growth and fabrication processes occurred simultaneously, while Figure 12b shows multicolor emissions, with blue from the microrod LEDs and green from the micropyramid LEDs. The EL spectra also exhibited distinct dominant EL peaks of 504−505 nm for the microrod LEDs (Figure 12c) and 570−580 nm for the micropyramid LEDs (Figure 12d). Since the semi-polar In_xGa_1−xN QWs have higher quantum confinement stark effects and in proportions compared to those of the nonpolar QWs, the semipolar micropyramids showed longer wavelength emissions. This report demonstrated the feasibility of integrating multicolor pixels on 2D films.

Since the conventional semiconductor LEDs are fabricated by epitaxy with the planner device structure, the uniform composition and thickness of the QW layers yielded in their monochromatic emissions. However, with the utilization of semiconductor nanostructures and microstructures, various interesting methods have been proposed for the monolithic integration of primary RGB colors into a compact chip, which includes selective-area growths [150–152], selective-carrier injections at the multifaceted QWs [105,153,154], and strain-engineering the QWs [155,156]. These approaches were originally demonstrated on 3D epitaxial substrates, while this result suggests a good feasibility that those monolithic approaches also can be adopted in the LEDs fabricated on 2D films.

6. Conclusion

This article discussed several topics, aiming to explore the extensive potential of heterogeneously integrated inorganic semiconductors on 2D materials. Followed by a comparison of the epitaxial growth of semiconductor layers on various kinds of substrates, different methods to perform controlled growth of semiconductors in 2D were introduced. The directly observable growth with TEM through 2D substrates provides an unparalleled opportunity to study the nucleation and initial growth mechanism of nanomaterials with high sensitivity and atomic level resolution. Multiple case studies regarding optoelectronic applications of semiconductor/2D heterostructures were also summarized where the intrinsic characteristics of 2D materials, possessing high mechanical flexibility and transferability, can be integrated into the semiconductor devices. We believe that establishing a general scheme for epitaxial growth of semiconductors on 2D materials is essential. This could revolutionize the inorganic semiconductor devices, making them flexible and wearable, and pave the way for assembling high-density nano- and micro-devices on a large scale. Still, a comprehensive understanding of the growth mechanism of 2D materials and potential adverse effects introduced in the semiconductor/2D system are further required to enhance the device's performance and manufacturing scalability. Also, various architectures and material combinations of semiconductor/2D heterostructures have remained to seek novel functionalities, improved device characteristics, and multifunctionality.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was financially supported by the National Research Foundation of Korea’s (NRF) grant funded by the Korean government (MSIT) (NRF-2021R1A5A1032996, NRF-2021R1C1C1010216, and NRF-2022M3H4A1A04096465) and Korea Institute for Advancement of Technology (KIAT) grant funded by the Korea Government (MOTIE) (P0023703, HRD Program for Industrial Innovation).

Notes on contributors

Aziz Ahmed

Aziz Ahmed obtained his PhD (Nano-Mechatronics) from Korea University of Science and Technology, Republic of Korea, in 2018. He is currently a postdoctoral researcher at the Research Center for Novel Quantum Architecture at Seoul National University, Republic of Korea. His research interests include the fabrication and characterization of inorganic semiconductor-based thin films and devices.

Kunook Chung

Kunook Chung received his PhD degree in Physics from Seoul National University, Republic of Korea, in 2015. From 2015 to 2019, he was a postdoctoral researcher at the Department of Electrical Engineering and Computer Science, University of Michigan, US. He is currently an assistant professor at Ulsan National Institute of Science and Technology, Republic of Korea. His research interest includes exploiting unique physical properties of various semiconductor quantum structures, nanostructures, van der Waals materials, and their hybrids, as key building blocks for fabricating high-performance and new functional optoelectronics.

Won Il Park

Won Il Park earned his Ph.D. in Materials Science & Engineering from Pohang University of Sicence and Technology, Republic of Korea, in 2005, followed by a postdoctoral position at Harvard University in Chemistry and Chemical Biology until 2007. Subsequently, he joined Hanyang University, Republic of Korea, as an assistant professor in the Division of Materials Science & Engineering. His research focuses on the synthesis and characterization of semiconductor nanomaterials, including nanowires, nanorods, graphene, and transition metal dichalcogenides. Specializing in the design and development of low-dimensional materials and structures, his work extends to nanoscale photonic and electronic devices, energy harvesting, and storage technology.

Gyu-Chul Yi

Gyu-Chul Yi is a professor in the Department of Physics at Seoul National University. He received a Ph.D. degree (1997) from Northwestern University. After working as a postdoctoral researcher at Oak Ridge National Laboratory, he joined Pohang University of Science and Technology, Korea, in 1999 as an assistant professor. Since 2004, he has been the director of the National CRI Center for Semiconductor Nanostructures. He has extensive experience in wide-bandgap semiconductor nanostructures and has published more than 250 referred articles in science, advanced materials, nano letters, nano energy, etc. He has also edited books and has written several book chapters as an author/coauthor.