Integrating 2D layered materials with 3D bulk materials as van der Waals heterostructures for photodetections: Current status and perspectives

Abstract

In the last decade, two-dimensional layered materials (2DLMs) have been drawing extensive attentions due to their unique properties, such as absence of surface dangling bonds, thickness-dependent bandgap, high absorption coefficient, large specific surface area, and so on. But the high-quality growth and transfer of wafer-scale 2DLMs films is still a great challenge for the commercialization of pure 2DLMs-based photodetectors. Conversely, the material growth and device fabrication technologies of three-dimensional (3D) semiconductors photodetectors tend to be gradually matured. However, the further improvement of the photodetection performance is limited by the difficult heterogeneous integration or the inferior crystal quality via heteroepitaxy. Fortunately, 2D/3D van der Waals heterostructures (vdWH) combine the advantages of the two types of materials simultaneously, which may provide a new platform for developing high-performance optoelectronic devices. Here, we first discuss the unique advantages of 2D/3D vdWH for the future development of photodetection field and simply introduce the structure categories, working mechanisms, and the typical fabrication methods of 2D/3D vdWH photodetector. Then, we outline the recent progress on 2D/3D vdWH-based photodetection devices integrating 2DLMs with the traditional 3D semiconductor materials, including Si, Ge, GaAs, AlGaN, SiC, and so on. Finally, we highlight the current challenges and prospects of heterointegrating 2DLMs with traditional 3D semiconductors toward photodetection applications.

INTRODUCTION

Since the discovery of single-layer graphene in 2004,¹ two-dimensional (2D) layered materials (2DLMs) have been attracting widespread attentions from the scientific community. In contrast to three-dimensional (3D) semiconductor materials, 2DLMs have many unique physical properties. First, due to the atom level thicknesses of 2DLMs, the free motion of electrons in 2DLMs is intensively confined in a 2D plane, which induces many strange physical phenomena, like quantum confinement effect.^2–4 Additionally, the energy bandgaps of 2DLMs can be modulated by controlling their thicknesses or applying external field and/or strain.^5–7 Third, the huge specific surface areas and strong in-plane covalent bonds enable 2DLMs to strongly interact with incident light and exhibit great mechanical flexibility and bending resistance.^8–10 Furthermore, van der Waals (vdW) interlayer interaction endows the 2DLMs with a natural passivation surface without dangling bonds. Therefore, 2DLMs are easy to disassemble by mechanical exfoliation and reassemble with any other 2D or 3D materials regardless of the lattice mismatch issue.^11,12 Finally, 2DLMs are a large material system with numerous categories which make it possible to design diverse vdW heterojunctions with different energy band alignments.^13,14 Until now hundreds of 2DLMs have been discovered and this number is still increasing continually each year. According to their conductivities, 2DLMs can be categorized as gapless semimetals, semiconductor materials, and ultra-wide bandgap insulators. Thanks to the diverse categories, the absorption range of 2DLMs spans a very wide electromagnetic spectrum. These original features make 2DLMs an appealing candidate for a wide application prospect in optoelectronics. Among them, photodetectors and photovoltaic devices have important application value both in military and civilian fields, such as infrared imaging, missile guidance, environmental inspection, optical communication and energy conversion.^15–18 In the past decade, several major breakthroughs have been made in photodetection based on 2DLMs. In terms of device performance, some pure 2DLMs based photodetectors even outperform the photodetectors made of traditional 3D semiconductor materials. For example, an ultrasensitive monolayer MoS₂ phototransistor reported in 2013 demonstrated a remarkable response in visible light range with a peak responsivity up to 880 A W⁻¹ at a wavelength of 561 nm.¹⁹ In 2015, a polarization-sensitive photodetector based on vertical BP p–n junction was reported for the first time, which exhibited a broad detection range from 400 to 3750 nm.²⁰ Later on, Massicotte et al. designed an ultrafast photodetector with a photoresponse speed of 5.5 ps by stacking graphene/WSe₂/graphene heterostructure.²¹ In 2017, Deng et al. achieved the continuous tuning of the bandgap of black phosphorus (BP) film from 0.3 to below 0.05 eV through external electric field⁵ and extended the mid-wave infrared (MWIR) photoresponse range of BP to beyond 7.7 μm.²² These impressive results indicate the great potential of 2DLMs for photodetection. Nonetheless, there are several crucial issues that need to be solved urgently for pure 2DLMs based photodetectors. First, lack of stable and reliable technology to synthesize wafer-scale and high quality 2DLMs films is the biggest challenge for the industrialization of all 2DLMs-based devices. Additionally, it is difficult to control the doping type, carrier concentration and stoichiometry of 2DMLs precisely, which constrain the device polarity. Finally, the insufficient light absorption resulting from the ultrathin thickness of 2DLMs usually leads to a limiting external quantum efficiency. Conversely, the material growth and device fabrication technologies of traditional 3D semiconductors have undergone long-term development and become gradually matured. Photodetection and photovoltaic devices based on them, such as silicon based solar cells, GaAs-based near infrared (IR) photodetectors, AlGaN-based UV photodetectors and HgCdTe-based middle and long IR photodetectors have been widely investigated over the past few decades and completed the process of industrialization. However, the further improvement of the photodetection performance is limited by the difficult heterogeneous integration or the inferior crystal quality via heteroepitaxy. Furthermore, HgCdTe, InGaAs, InSb, and other narrow bandgap semiconductors based MWIR photodetectors generally suffer from the problems of large dark current and high noise. Hence, they usually operate in cryogenic environment,²³ which increases the volumes and power consumptions of the photodetection systems and is difficult to meet the requirements of integration, miniaturization and low cost for IR detection applications. Fortunately, the integration of 2DLMs and traditional 3D semiconductor materials on the same substrate may provide a new platform for the development of high performance and monolithic integration photodetectors. This integration strategy combines the unique properties of 2DLMs with the mature processing technologies of traditional 3D semiconductor materials, which can not only absorb the advantages of the two types of materials simultaneously but also can provide a promising strategy for the large-scale integration of 2DLMs. Thanks to the unique properties and characters of 2DLMs, the 2D/3D mixed-dimensional vdW heterostructures (vdWH) possess many advantages such as free of lattice mismatch, sharp, and clean heterointerface, easy construction of heterojunctions and device structures with diverse types, easy manipulation of electron states and wide coverage of detection spectrum. Therefore, 2D/3D mixed-dimensional vdWH based photodetectors have become a new research hotspot and extensive experiments focusing on their multifunctional applications have been carried out in the recent years, as shown Figure 1.

FIGURE 1. The 2D/3D vdW heterostructures and their potential function applications for photodetection.

In this review, we first simply introduce the structure categories, working mechanisms, and the typical fabrication methods of 2D/3D mixed-dimensional vdWH, then outline the recent progresses on 2D/3D vdWH based photodetection devices integrating 2DLMs with the traditional 3D semiconductor materials, including Si, Ge, GaAs, AlGaN, SiC, and so on. Finally, we highlight the current challenges and prospects of the heterointegration of 2DLMs with 3D semiconductors toward photodetection applications.

STRUCTURE CATEGORIES AND WORKING MECHANISMS OF 2D/3D VAN DER WAALS HETEROSTRUCTURE BASED PHOTODETECTORS

The vdW interactions allow 2DLMs to stack on the surfaces of 3D bulk materials (3DBMs) regardless of the limitation of lattice match, thus making it possible to design high performance 2D/3D vdWH photodetectors with different structures. In terms of the functions of the 2DLMs and 3DBMs serving in the device, 2D/3D vdWH photodetectors can be categorized as three basic structures presented in Figure 2, including heterojunction structure, photogating structure, and optical waveguide or cavity structure. This section will briefly introduce the three structures and their working mechanisms, respectively.

FIGURE 2. Schematic illustrations of structure categories of 2D/3D vdWH photodetectors. (A) heterojunction structure, (B) photogating structure, (C) optical waveguide structure, and (D) optical cavity structure.

Heterojunction structure

The 2D/3D vdWH photodetectors based on heterojunction structure are the most widely investigated devices which consist of a PN or Schottky junction. In this structure, both the 2DLMs and 3DBMs act as the light absorption materials, so a broad spectral response range can be achieved easily. Under illumination, the photogenerated electron–hole (e–h) pairs formed in the junction area are separated rapidly under the drive of the built-in electric field. Then, these separated charges are collected by the electrodes on the two sides of the junction to generate photocurrent. Due to the built-in electric filed, this device usually shows ultrafast response speed and low dark current at zero or reverse bias. However, the absence of internal gain mechanism leads to a low responsivity of several mA W⁻¹.

Photogating structure

How to achieve a high gain-bandwidth (or response time) product is a common issue for most photodetectors. Fortunately, the 2D/3D vdWH photodetectors based on photogating structure seem to provide a suitable balance for this problem. In this structure, one type of material (usually 3DBMs) acts as light absorption material while the other (usually 2DLMs) is used for charge carrier transportation. Under light illumination, the one type of photogenerated carriers are trapped in 3DBMs, while the other type of carriers transfers to the 2DLMs and circulate constantly in the transportation channel under a source-drain bias. The great absorptivity of 3DBMs and high carrier mobility of 2DLMs are fully exerted in this structure, resulting in a simultaneous optimization of gain and response time.

Optical waveguide or cavity structure

The poor light absorptive capacities of 2DLMs seriously constraint their application potential for photoelectric conversion. One effective way to solve this problem is to decouple the light-matter interaction length from the thickness of 2DLMs, that is, change the light propagation from the vertical to the in-plane direction, which can be achieved by integrating 2DLMs with optical waveguide structure. Moreover, it provides a solution for the detection system to achieve on-chip integration, thereby enabling more freedom in designing desired device geometry for performance optimization. Another way to enhance the absorption is to embedding the 2DLMs into an optical cavity structure. This structure enables the energy of the optical field to be confined in a tiny space so as to acquire a high external quantum efficiency (EQE) and gain.

ASSEMBLY OF 2D/3D HETEROSTRUCTURES

Reliable and efficient fabrication and assembly techniques are of great significance for the applications and further research of 2D/3D heterostructures devices. Currently, mechanical transfer and in situ chemical or physical synthesis are two main strategies to assemble high-quality 2D/3D heterostructures, including dry transfer, wet transfer, chemical vapor deposition (CVD), metal–organic chemical vapor deposition (MOCVD), molecular beam epitaxy (MBE), and pulsed laser deposition (PLD), as illustrated in Figure 3.

FIGURE 3. (A) Schematic of mechanical transfer, including dry transfer and wet transfer. (B) Schematic of in situ growth. (i) Chemical vapor deposition (CVD). The right panel is HRTEM image of MoS2/n-type GaN heterojunction by CVD; Reproduced with permission.33 Copyright 2021, Elsevier. (ii) Metal–organic chemical vapor deposition (MOCVD). On the right, HRTEM image of the MoS2/WSe2/n-GaN heterostructure via MOCVD; Reproduced with permission.40 Copyright 2017, Royal Society of Chemistry. (iii) Pulsed laser deposition (PLD). On the right, cross-sectional HRTEM image of MoS2 on n+ implanted 4H-SiC by PLD; Reproduced with permission.45 Copyright 2022, John Wiley and Sons. (iv) Molecular beam epitaxy (MBE). The right panel is the HRTEM cross-section image of 20-layer GaSe/Si and the SAED pattern (inset). Reproduced with permission.48 Copyright 2015, American Chemical Society.

Mechanical transfer

Mechanical transfer is a relatively simple and convenient method to assemble 2D/3D heterostructures, which can be divided into dry transfer and wet transfer. The dry transfer is widely used to stack the 2DLMs obtained by mechanical exfoliation.^24–26 This method includes four key steps. First, in case of forming any wrinkles and cracks on the 2DLMs, a sticky polymer carrier as a structural support, like polydimethylsiloxane (PDMS) is sticked on the 2DLMs grown on the original substrate; second, peeling off the PDMS from the original substrate; next, the 2DLMs adhered to the PDMS are aligned under a microscope and then are stacked onto the specific position of 3DBMs; finally, releasing the PDMS through heating. By repeating this procedure, multilayered heterostructures can be constructed. Despite its simple and nondestructive process, this method is not scalable enough to meet practical application. Furthermore, the limited size of 2DLMs transferred by this method currently restricts the area of heterostructure interface, thereby inhibiting wider-ranging research studies in this realm.

To achieve the large-area and efficient transfer of the synthetic 2DLMs that has strong adhesion to the growth substrates, the wet transfer technique has been developed. Briefly, a polymethyl methacrylate (PMMA) membrane is spin-coated on the substrate at first. Second, after the substrate is etched by the corresponding etchant, the PMMA with 2DLMs will float on the etchant. At last, the 2DLMs can be transferred to the target substrates followed by dissolving PMMA in acetone. This method features high transfer efficiency and has been widely adopted to transfer the CVD-grown graphene on copper foil and the CVD-grown transition metal dichalcogenides (TMDCs) on SiO₂/Si substrate.^27–29 However, the intrinsic properties of transferred materials are easily affected by the etchant and polymer residues. In addition, it is unavoidable that some foreign contaminants are introduced in the heterostructure interfaces during the transfer process, leading to the degradation of device performance.

In situ growth

Compared with the mechanical transfer method, direct in situ growth of 2DLMs on target 3DBMs leaves out the transfer process, thus preventing contaminants from invading the heterostructure interfaces. So far, CVD is the most popular method for synthesizing high-quality 2DLMs, such as graphene and TMDCs, considering its high throughput, low cost, and good repeatability. In this process, take TMDC as an example, the chalcogen powder is placed in the low-temperature zone of a quartz tube; transition metal-based powder and the substrate are placed in the high-temperature zone. As the quartz tube is heated up under a carrier gas, the two kinds of precursors are transported to the substrate in the gas phase and react directly to form TMDCs. Through precise control of the growth parameters such as the temperature, pressure, flow rate, and the distance between precursors and substrate, 2DLMs films with different size, morphology, and thickness can be obtained.^30,31 Therefore, CVD has overall control over material properties contrast to other growth methods and has been proven to be a versatile tool for fundamental studies of 2DLMs. In recent decades, various types of 2D/3D heterostructures have been realized by this method, for example, MoS₂/GaN,^32,33 MoS₂/Al₂O_3,³⁴ graphene/Ge,³⁵ and graphene/AlGaN.³⁶ However, a wide grain-size distribution, random nucleation sites, and irregular grain orientations usually occur during the CVD synthesis and lead to the formation of grain boundaries and defects which would bring negative effects on the electrical property of 2DLMs.^30,37 Furthermore, it is difficult to control the flux of precursors and prepare continues monolayer films using this method.

In comparison with CVD, MOCVD has the similar working mechanism, but is more promising to grow wafer-scale uniform monolayer and few-layer films. Metal carbonyls [Mo(CO)₆, W(CO)₆], organo-chalcogen compounds [(C₂H₅)₂S, (C₄H₉)₂S, (C₂H₅)₂Se, (C₄H₉)₂Se] and hydrides (H₂S, H₂Se, CH₄) are common precursors used in MOCVD for TMDC and their heterostructures growth.^38,39 These precursors are taken into reaction chamber by carrier gases and then decompose into growth materials at high temperature. Since gaseous precursors are employed, MOCVD can accurately control the supply of precursors and growth rate. Using this method, Zhang et al. successfully realized the scalable growth of MoS₂, WSe₂, and their heterostructures on GaN with atomically sharp interface.⁴⁰ Recently, Tuqeer Nasir et al. achieved direct growth of highly uniform graphene nanostructures on SiO₂ substrate by remote metal catalyst-assisted MOCVD, evading the tedious transfer process of CVD-growth graphene on catalytic metal substrates.⁴¹ Moreover, the successive growth of MoS₂/WS₂, WS₂/MoS₂, MoS₂/graphene, and WS₂/graphene heterobilayers on sapphire substrates have been realized via MOCVD.^42,43 Despite these great progresses, limited usability for specific 2DLMs, complicated operation, high cost, environmental pollution caused by metal–organic precursors, which restrict the extensive use of MOCVD, are still primary issues that need to be solved.

PLD is a physical vapor deposition technique used to grow epitaxial films with the high energy pulsed laser. The growth process includes two steps. First, the target materials are ablated under the high energy pulsed laser irradiation. During this process, the plasma plumes are generated accompanied by a series of complicated physical phenomena. Next, the plasma plumes reach the cold substrate and crystallize into films. Benefited from the high energy of focused laser beam, PLD avoids the high growth temperature. Therefore, a large scope of substrates for direct growth of TMDC films are available to PLD, which creates a great possibility for the design of flexible and wearable devices.⁴⁴ Overall, featured with cleanliness, safety, and high throughput, PLD is particularly suitable for the in situ growth of 2D/3D heterostructures. For instance, by directly growing MoTe₂ on Si substrate, Lu et al. fabricated a MoTe₂/Si vertical 2D/3D heterojunction photodiode with the response range spanning from 300 to 1800 nm. Most recently, MoS₂/4H-SiC heterojunction diodes have been fabricated via PLD deposition of MoS₂ on 4H-SiC surfaces.⁴⁵ Impressively, this structure showed an electronic transport tunability by controlling the doping level of SiC, which paves the path for industrial development of MoS₂/SiC device.

MBE is another scalable method developed for the epitaxial growth of high-quality TMDC films and III–VI group 2DLM films. In a typical MBE process, the high-purity elemental sources are evaporated to form the molecular beams in the ultrahigh vacuum chamber (with pressure typically below 10⁻¹⁰ mbar) and then impinge on the heated substrate, giving rise to chemical reaction and generation of the desired materials.⁴⁶ Meanwhile, the thickness and quality of as-grown film can be monitored in situ by reflection high-energy electron diffraction (RHEED). Therefore, MBE can precisely control the thickness, doping concentration, and alloy composition. Thus far, MBE has been extensively used to assemble 2D/3D heterostructures such as MoSe₂/GaAs,⁴⁷ GaSe/Si,⁴⁸ and Te/Ge.⁴⁹

INTEGRATING GRAPHENE WITH 3D SEMICONDUCTOR AS VAN DER WAALS HETEROSTRUCTURES FOR PHOTODETECTION

Graphene is a 2DLM with a hexagonal honeycomb structure composed of sp² hybridized carbon atoms.⁵⁰ In 2004, Geim et al. used tapes to peel off graphite with only one atomic layer thickness for the first time, marking the advent of the first 2D material—graphene.¹ At the same time, this discovery also officially opened the door for researchers to study 2DLMs. Graphene is the first discovered two-dimensional layered material, and its structure and properties have been extensively studied, so it is also the most familiar 2DLMs. Among many 2D materials, zero band gap, high transmittance (up to 97.7% for incident light with a wavelength in the range of 300–1000 nm) and high carrier mobility (up to 10⁴ cm² V⁻¹ s⁻¹ at room temperature) are considered to be the three outstanding characteristics of graphene.⁵¹ In addition, graphene has the advantages of high environmental stability, adjustable work function, and mature manufacturing process. The above characteristics make graphene an important part of van der Waals heterojunction, and it has been widely used in the field of optoelectronics. In addition to the simplest and direct mechanical exfoliation method to obtain graphene, after recent years of research, many effective methods have also been developed for synthesizing high-quality graphene such as CVD,^52–54 PECVD,^55–57 and thermal decomposition of SiC.^58,59 Even though graphene has ultra-high carrier mobility, the light absorption rate of single-layer graphene in the ultraviolet to visible range is only 2.3%, and the large dark current caused by the zero band gap results in low light detectivity.^60–62 Fortunately, the combination of graphene and other semiconductor materials to form a graphene-based heterojunction can not only make up for these shortcomings, but also give full play to the advantages of graphene, showing huge application potential in the field of photodetection. In the following section, we will briefly introduce the research progresses of heterostructures constructed by graphene and some traditional 3D semiconductor materials and their applications in photodetection. Table 1 summarizes the critical performance parameters for graphene/3D semiconductor heterostructure based photodetectors reported in the last decade.

TABLE 1 Summary of the critical performance parameters for graphene/3D semiconductor heterostructure based photodetectors.

Material (configuration)	Wavelength (nm)	Response time (τr/τd)	Responsivity (A W−1)	D* (cm Hz1/2 W−1)	EQE (%)	Working bias	References
ZnO TG/Gr/Si	395–850	—	70	2 × 1013	—	VDS = 0.1	63
In2S3/Gr/Si	400–1200	33/40 μs	4.53 × 104	3.02 × 1011	—	VDS = −2	64
Gr/AlN/Si	365	1.9/1.4 ms	1.03	2.94 × 107	—	VDS = −10	65
Gr/GaAs	650	0.119/16.1 s	1321	3.24 × 108	—	VDS = −1	66
Ag NPs/Gr/GaAs	405	—	0.21	2.98 × 1013	>40	VDS = 0	67
Gr/nanoporous GaN	360	0.35/0.36 ms	~102	1.0 × 1017	>100	VDS = −1.5	68
Gr/nanopillars GaN	366	53/29 ms	~104	1.85 × 1017	1.1 × 107	VDS = 3	69
Gr/β-Ga2O3	254	9.3/10.3 s	2.6 × 103	9.7 × 1013	—	VDS = 10, VG = −8	70
Gr/4H-SiC	325	1.6/3.4 s	254.1	2.16 × 1010	9.6 × 104	VDS = −3, VG = 3	71
Gr/HgCdTe	1550	13 ns	2.5	2 × 1010	85	VDS = 0	72

Abbreviations: Gr, graphene; NPs, nanoparticles; TG, top gate.

Graphene/Si van der Waals heterostructure based photodetectors

The integration of graphene and Si for photodetection has been under extensive investigation for more than ten years.^73–76 Combining the great transparency and high carrier mobility of graphene with the broad absorption spectrum of silicon, the graphene/Si heterojunctions present a natural superiority for photodetection applications. Moreover, with a good compatibility to complementary metal-oxide-semiconductor (CMOS) integrated circuits, graphene/Si heterostructures offer a great possibility for the monolithic integration of mixed-dimensional heterostructure photodetectors. However, in order to achieve high-performance graphene/Si heterojunction photodetector, there are two crucial problems needing to be solved. One is the low detectivity resulting from large dark current. Just as mentioned before, the electron potential barrier from n-Si to graphene is rather low, usually in the range of 0.6–0.7 eV.⁷⁷ Such a low barrier height is not enough to block the diffusion of the electrons in Si layer to graphene, which will cause an unexpected diffusion current in dark state. The other is poor photoresponse of conventional graphene/Si heterojunction due to the lack of photogain mechanism and inferior heterojunction interface quality. At first, since graphene is mainly used as a transparent electrode to extract charge carriers, the photon absorption is almost achieved by the silicon layer. Second, in contrast to photoconductor and phototransistor devices, there is no internal gain discovered in graphene/Si heterojunctions. In addition, the high density of interface trap states introduced in the process of device fabrication leads to severe carrier recombination at the heterojunction. Therefore, the early graphene/Si heterojunction photodetector reported in 2013 exhibited a poor detectivity of only 2.1 × 10⁸ cm Hz^1/2 W^−1.78 One way to alleviate these problems is inserting an interfacial passivation layer between graphene and Si. The interfacial layer can not only passivate the surface states but also tune the potential barrier height of the Schottky junction. In dark conditions, the insulating layer will block the electrons in the Si side through the high electron potential barrier. While, under light illumination, the photogenerated holes will accumulate at the insulating layer/Si interface and have a probability of tunnel through the insulating layer. So the dark current can be sufficiently suppressed and a large photocurrent is maintained at the meantime when an insulating layer with a proper thickness is chosen. The first reported interfacial layer is a native oxide layer on Si surface.⁷⁹ After introducing a thin SiO₂ layer, the dark current is reduced by two orders of magnitude with a photo-to-dark current ratio up to 10⁷. Correspondingly, the specific detectivity is improved to 5.77 × 10¹³ cm Hz^1/2 W⁻¹ at 870 nm irradiation. Later on, other insulating materials or wide bandgap semiconductors, such as Al₂O₃,⁸⁰ AlN,⁶⁵ and h-BN,⁸¹ were used as the interfacial layer. Recently, Yin et al. achieved the photocurrent multiplication in the graphene/AlN/Si device via impact ionization.⁶⁵ They introduced a AlN film with thickness of 15 nm into the interface of graphene/Si heterojunction as shown in Figure 4A. Under a high reverse bias, a strong electric field is formed in the thin AlN film since its high resistivity. As a result, the impact ionization occurs when the photogenerated carriers tunnel through the AlN layer, which gives rise to the carrier multiplication effect and an enhanced photogain. In addition, driven by the strong electric field in the AlN layer, plenty of hot carriers in graphene can easily overcome the barrier and inject to Si side. Benefited from the two processes, a peak response of 3.96 A W⁻¹ with a photogain of 5.8 was realized at the reverse bias of 10 V (Figure 4C). Apart from insulating materials, TMDCs and organic films were also proposed as the interfacial layer and demonstrated a series of promising results in recent few years.^82–84

FIGURE 4. (A) Schematic of the fabricated wafer-scale graphene/insulator/Si (GIS) heterojunction photodetectors. (B) The working mechanisms of the GIS heterojunction photodetector. (C) Spectra-dependent photocurrent responsivity of the GIS heterojunction photodetector at a bias of −10 V. (A–C) Reproduced with permission.65 Copyright 2021, Springer Nature. (D) Schematic of gate-tunable graphene-silicon photodetector. (E) Energy band diagram of graphene-silicon photodetector with gate modulation. (F) Dark and illuminated current of the device versus Vg. (D–F) Reproduced with permission.63 Copyright 2018, John Wiley and Sons. (G) Sketch of the graphene photodetector integrated with a Si microring resonator. (H) Photoresponse map demonstrating tunable photoresponse at zero bias on resonance (λ = 1555.87 nm) via VG1 and VG2. (I) Power-dependent junction temperature Te and voltage responsivity R[V/W]. (G–I) Reproduced with permission.93 Copyright 2021, Springer Nature.

To further improve the responsivity of graphene/Si heterojunction photodetectors, the photogating effect was introduced by applying a gating material.^63,76,85,86 For example, Chang et al. reported a gate-controlled graphene/Si hybrid structure (Figure 4D).⁶³ In this structure, ZnO is used as the top transparent gate, and the Fermi level of graphene can be lowered by applying a negative bias voltage to the gate, thereby increasing the Schottky barrier height and effectively suppressing the dark current to within the order of μA. At the same time, a large number of photogenerated holes accumulated at the interface, which increased the injection of holes into graphene. As a result, an ultra-high responsivity of 70 A W⁻¹ was achieved, which was 100 times higher than that of the traditional graphene field-effect photodetector (Figure 4F). Another graphene/Si hybrid structure was constructed by transferring a In₂S₃ nanoflake onto top of the graphene.⁶⁴ The photogain mechanism of this structure is based on the photogating effect. Under light irradiation, the photogenerated holes in the introduced In₂S₃ nanoflake will transfer to graphene at first and then arrive at p-Si side under the built-in electric field at the graphene/Si interface. These holes will recycle between the two electrodes on graphene and Si until they recombine with the electrons captured by the traps in In₂S₃ nanoflake. As a result, an ultrahigh responsivity of 4.53 × 10⁴ A W⁻¹ was obtained.

Another kind of graphene/silicon integrated photodetector aims to solve the problem of low light absorption of graphene, such as graphene integrated with silicon waveguide^87–90 and resonant cavity.^91–93 Contrary to graphene/Si heterojunction, in these structures, graphene is used as the absorption material and Si as light confinement material. Researchers from Cambridge University combined the graphene/silicon Schottky junction with an optical resonant cavity made of silicon to enhance light-matter interaction and absorption at a single layer graphene/silicon interface.⁹¹ Compared with graphene/silicon Schottky junction photodetectors in free-space, the reported device showed an excellent spectral selectivity and an enhanced responsivity of 20 mA W⁻¹ at 1550 nm. Recently, a graphene photodetector integrated on a silicon microring resonator shown in Figure 4G was proposed.⁹³ In this device, the type and density of the majority carriers in the graphene channel can be regulated continuously via varying the two gate voltages (V_G1 and V_G2) of the split-gate. By this way, four junction constellations (p–n, n–p, n–n, p–p) can be generated. At resonant wavelength of 1555.87 nm, the responsivity reaches its peak when the polarities of V_G1 and V_G2 are opposite and shows a polarity-switch at V_G1 = V_G2 (Figure 4H). This indicates that the photothermoelectric (PTE) effect dominates the photoelectric conversion mechanism. Given this, the junction temperature rises rapidly as the increased optical power and a maximum a voltage responsivity of 90 V W⁻¹ is achieved (Figure 4I).

Thanks to the fast progress of wafer-scale growth technology of graphene, imaging devices based on graphene/Si heterojunction arrays become possible.^94–96 A milestone work of integrating graphene with Si CMOS circuit was reported first by Goossens et al.⁹⁷ They transferred wafer-scale graphene grown by CVD on top of a CMOS wafer containing photodetection pixels of the 388 × 288 array and then deposited a sensitizing layer of PbS colloidal quantum dots on the graphene. In each pixel, the graphene channel is vertically connected with a CMOS readout integrated circuitry. Attributed to the photogating effect and high carrier mobility of graphene channel, this graphene-quantum dot image sensor realizes an ultrahigh gain of 10⁸ and a responsivity above 10⁷ A W⁻¹. Very recently, Xu et al. reported a graphene charge-injection (GCI) photodetector integrated with Si and achieved broadband imaging from ultraviolet to mid-infrared (around 375 nm to 3.8 μm) with a linear array device.⁹⁸ The key idea in this work is that the photocharge readout process is achieved by the photogating effect. The photocharges stored in the depletion well induced equal but opposite charges in the single-layer graphene (SLG) on top of the SiO₂, causing a real-time change of the drain current along with the periodic gate voltage pulses. Combining the charge integration feature of the charge coupled device (CCD) imagers with the independent pixel structure of the CMOS imagers, the GCI photodetectors integrate three functionalities—charge store, integration, and direct readout into a single pixel, which have great potential to build neuromorphic networks and become new generation imaging devices.

Graphene/GaAs van der Waals heterostructure based photodetectors

GaAs is the representative of the III–V group compound semiconductors with excellent properties such as a direct band gap of 1.42 eV, high optical absorption coefficient (10⁴ cm⁻¹) and high electron mobility (8000 cm² V⁻¹ s⁻¹ at 300 K).⁶⁷ Therefore, it is widely used in photovoltaic solar cells and near-infrared photodetectors. At the same time, the heterojunction composed of graphene and GaAs also exhibits excellent optoelectronic properties. In 2014, Li et al. tried to construct a graphene/GaAs Schottky junction solar cell and achieved a photoelectric conversion efficiency of 15.5%, which exceeded the highest level achieved by graphene/Si solar cells at that time (14.5%).⁹⁹ In the following year, they added a graphene/Al₂O₃/graphene gate structure on the basis of the traditional graphene/GaAs Schottky junction (Figure 5A).¹⁰⁰ According to the principle of electrostatic doping, the Fermi level of the graphene touching with GaAs shifts downward under a larger negative gate bias, leading to an increasing Schottky barrier height and a higher V_OC. Benefiting from the anti-reflection of the Al₂O₃ dielectric layer, high transmittance and adjustable Fermi energy level of graphene, a high-efficiency solar cell with V_OC = 0.96 V and PEC = 18.6% was achieved (Figure 5B).

FIGURE 5. (A) Schematic cross-section view and top view of the field-effect graphene/GaAs solar cell. (B) Experimental and simulated light J–V curves with different Vgate. (A,B) Reproduced with permission.100 Copyright 2015, Elsevier. (C) Schematic structure of the graphene-SAM-GaAs solar cells. (D) J–V curves for devices under illumination (AM 1.5G, one sun, 25°C). (C,D) Reproduced with permission.101 Copyright 2018, Royal Society of Chemistry. (E) Responsivity and (F) detectivity of the graphene/GaAs photodetector with and without 100 nm Ag NPs under zero bias voltage at five excitation wavelengths. (E,F) Reproduced with permission.67 Copyright 2018, Elsevier. (G) Top gate and back gate configuration of the graphene/GaAs interface-based photodetector. (H) Responsivity and gain at various back gate voltages under 148 nW light illumination. (G,H) Reproduced with permission.66 Copyright 2020, John Wiley and Sons.

In the process of constructing graphene/GaAs Schottky junctions, wet transfer technology is often used to transfer graphene to the GaAs surface, which can easily lead to oxidation of GaAs and introduce a large number of surface states at the interface. As recombination centers, these surface states will accelerate the recombination of surface carriers and cause surface scattering effects, which seriously affect the performance of the device. Therefore, the interface quality of the heterojunction is a key factor to determine the PEC of graphene/GaAs Schottky junction solar cells. In 2018, Wen et al. modified the GaAs surface with the self-assembled alkyl thiol monolayers so as to inhibit surface oxidation of GaAs (Figure 5C).¹⁰¹ Due to the insulating property of the organic molecular layer and improved interface quality, the leakage current is suppressed obviously and the PCE is increased to 15.9%.

In addition to solar cells, there are also a large number of reports on the application of graphene/GaAs heterojunctions in visible to near-infrared photodetectors.^{66,67,102,103} Due to the high absorption coefficient (10⁴ cm⁻¹), the light absorption mostly occurs near the surface of GaAs, thus the carriers are generated mainly at a distance of several hundred nanometers from the surface of the material. Therefore, the near-field enhancement effect can significantly enhance the light absorption of GaAs materials. Based on this idea, Lu et al. covered Ag nanoparticles on the graphene/GaAs heterojunction to confine the light field near the surface of GaAs, and promote the surface absorption of GaAs (Figure 5E).⁶⁷ Then the graphene/GaAs Schottky junction photodetector with a broad spectrum response (325–980 nm) and a detectivity of 2.98 × 10¹³ cm Hz^1/2 W⁻¹ was successfully manufactured (Figure 5F). However, unlike photoconductive detectors, the Schottky junction photodiode composed of graphene and semiconductor lacks a gain mechanism, resulting in low photoresponse limited to the mA W⁻¹ level. In order to solve this problem, Tian et al. proposed an interface-induced gain mechanism (Figure 5G).⁶⁶ Under the force of built-in electric field in the Schottky junction, the photogenerated holes in the GaAs drift toward the interface and are trapped by the interfacial trap states. The increasing hole concentration at the interface lowers the Fermi level of graphene via the level alignment mechanism, thus the conductivity of the graphene is significantly improved. This can be confirmed by the photoresponse test at various back gate voltages as shown in Figure 5H. At higher back gate voltage, the electric filed in GaAs is increased and the electron–hole pairs separate more effectively, so a high responsivity and gain can be achieved. Based on this, the fabricated graphene/GaAs Schottky junction photodetector demonstrated a high responsivity of 1321 A W⁻¹ at 650 nm.

Graphene/AlGaN van der Waals heterostructures based photodetectors

Ultraviolet detectors have important application value in the fields of environmental detection, biomedicine, national defense, military, and aerospace research. At present, the traditional wide-gap semiconductor materials such as SiC,¹⁰⁴ Ga₂O_3,¹⁰⁵ GaN,¹⁰⁶ and its ternary alloy AlGaN¹⁰⁷ have been used to realize high-efficiency ultraviolet detectors. Among them, AlGaN has a tunable direct band gap of 3.4–6.2 eV and has excellent properties such as high electron mobility, high thermal stability, stable physical and chemical properties, and good radiation resistance, so it is widely regarded as one of the materials for making UV detectors.^106,108 However, due to the immature growth technology of GaN and AlN single crystal substrate, currently commercial AlGaN-based devices are generally obtained through heteroepitaxy where a large lattice mismatch is unavoidable. The lattice mismatch causes a large number of defects in the epitaxy film, resulting in relatively low carrier mobility and short carrier lifetime, which severely limits the responsivity of AlGaN-based UV detectors.¹⁰⁹ Moreover, the transparent electrode of indium tin oxide (ITO), which is widely used in optoelectronic devices, has strong absorption of ultraviolet light. Therefore, ITO is not suitable for AlGaN-based UV detectors. Graphene with nearly complete UV transparency, excellent conductivity and high carrier mobility, can not only serve as a transparent electrode, but also as an effective carrier transport channel in the graphene/AlGaN vdWH. The built-in electric field at the interface between graphene and AlGaN can promote the rapid separation of photogenerated carriers, which is helpful to the quantum efficiency and response speed. Many graphene/AlGaN van der Waals heterostructures based UV photodetectors have been reported recently and shown excellent detection performance even superior than the AlGaN-only UV detectors. For example, Tian et al. used a Schottky junction composed of an unintentionally doped GaN film and graphene to achieve a high-efficiency ultraviolet detector with a detection wavelength of 325 nm (Figure 6A).¹¹⁰ Because the Fermi level of GaN is higher than graphene, the energy band bends upward when it contacts graphene, and a built-in electric field pointing from GaN to graphene is generated at the Schottky junction interface.^68,110,111 Under UV light illumination, the photogenerated electron–hole pairs in GaN are effectively separated by the built-in electric field and then are collected by opposite electrodes. Compared with the GaN photodetector without graphene, the device exhibits a 10⁷ times higher responsivity and a high response speed of 4.6 ms (Figure 6B). In order to further increase the light absorption of GaN, in 2020, Li et al. prepared nanoporous GaN films by photoassisted electrochemical corrosion (PEC) and then transferred graphene microsheets onto the GaN films (Figure 6C).⁶⁸ Attributed to the enhanced light harvesting of nanoporous GaN film and high carrier mobility of graphene, the device performs a high sensitivity to 360 nm UV light with the specific detectivity of 1 × 10¹⁷ cm Hz^1/2 W⁻¹ shown in Figure 6D. Flexible photodetectors have important application value in implantable and wearable electronic devices, so they have attracted widespread attention and research. Han et al. embedded GaN nanowires horizontally in two layers of graphene films to prepare a flexible ultraviolet photodetector with a high response speed of <30 ms (Figure 6E,F).¹¹²

FIGURE 6. (A) Energy band diagram of the graphene/unintentionally doped GaN UV photodetector. (B) Photocurrent and responsivity of the device without and with graphene versus illumination power at −10 V. (A,B) Reproduced with permission.110 Copyright 2018, AIP Publishing. (C) Three-dimensional schematic of the graphene/nanoporous GaN based photodetector. (D) Specific detectivity of the device versus incident light power density at λ = 360 nm under different biases (V = −1.5, 0, 1.5 V). (C,D) Reproduced with permission.68 Copyright 2020, American Chemical Society. (E) Schematic of the process used to fabricate a flexible GaN–NW photosensor. (F) Photoresponse curves of the device at frequencies of 2.5 Hz with respect to the substrate bending radius. (E,F) Reproduced with permission.112 Copyright 2019, American Chemical Society.

Besides the graphene/GaN UV photodetectors, the vacuum ultraviolet (VUV) photodetectors based on graphene/AlN van der Waals heterostructures have also been demonstrated. The VUV refers to the UV light with the wavelength range of 10–200 nm. Since its extremely low sun background noise, the VUV shows huge application potential in the fields of aerospace technology and astronomical research, like observing the dynamic evolution of stars and the expansion of nebulae.^113,114 In 2020, Zheng et al. developed an in-plane enhanced heteroepitaxial method to grow the high-quality AlN single crystalline films on the GaN template and then a VUV photovoltaic detector based on Gr/AlN/GaN p–i–n double heterojunction was constructed.¹¹⁵ Under zero bias, the as-fabricated detector exhibited an ultrafast response time of 2.86 μs and a photovoltage of 2 V.

Graphene and other 3D semiconductors heterostructures integration for photodetection

Attributed to the unique 2D characteristics and ultra-high carrier mobility of graphene, mix-dimensional vdW heterojunctions composed of graphene and other 3D semiconductors display enhancing photodetection performance and even overcome some key technical problems suffered by pure 3D semiconductor photodetectors. For example, the fabrication of traditional HgCdTe infrared photovoltaic devices usually involves ion implantation to dope the HgCdTe materials. However, this process will cause irreversible damage to the crystal structures of HgCdTe materials and deteriorates the material quality which ultimately leads to the weakening of detectivity at room temperature. Therefore, current commercial HgCdTe infrared photodetectors still need to work in cryogenic environment to improve the photodetection performance. Recently, Wang et al. successfully designed a new ultrafast uncooled mid-wavelength infrared photodetector by constructing graphene/HgCdTe heterojunction (Figure 7A).⁷² In this device, graphene acts as a p type contact layer and contact with HgCdTe through vdW force, which avoids the introduction of defects in the HgCdTe materials. Thus, the peak detectivity is enhanced up to 2 × 10¹⁰ cm Hz^1/2 W⁻¹ under blackbody radiation at room temperature and 10¹¹ cm Hz^1/2 W⁻¹ at 77 K. Moreover, due to the high carrier mobility of graphene, this designed photodetector exhibited a fast response time up to 13 ns (77 MHz), which exceeded about one order of magnitude than that of currently commercial uncooled HgCdTe photovoltaic photodetectors (Figure 7B).

FIGURE 7. (A) Schematic diagram of the vdWs-on-MCT photodetector. (B) Relative response versus switching frequency, showing the 3 dB cutoff frequency of 77 MHz. (A,B) Reproduced with permission.72 Copyright 2022, John Wiley and Sons. (C) Schematic diagram of graphene/n-type Ge Schottky junction photodetector with a thin Al2O3 interfacial layer. (D) Wavelength-dependent detectivity of the photodetector measured at a fixed light intensity of 25 μW. (C,D) Reproduced with permission.118 Copyright 2021, De Gruyter. (E) Schematic diagram and microscopic photograph of the graphene/4H-SiC ultraviolet photodetector, scale bar: 10 μm. (F) Calculated responsivity and EQE, at 3 V gate voltage. (E,F) Reproduced with permission.123 Copyright 2020, John Wiley and Sons. (G) Schematic energy band structure and liquid phase epitaxy (LPE) diagrams of the graphene/Ga2O3 junction under 254 nm illumination at 0 V bias. (H) Dependence of Vx1 and Vx2 on the light spot position along the x direction at 30 and 200°C (in the inset). (G,H) Reproduced with permission.127 Copyright 2022, American Chemical Society.

For short-wave infrared (SWIR) detection, the graphene/Ge Schottky junction has become increasingly attractive in the past few years owing to its simple structure, mature manufacturing process as well as good compatibility with integrated circuits.^116–119 However, the dangling bonds on Ge surface are usually considered as a main source of leakage current, which seriously degrades the overall performance of the graphene/Ge photodetectors. In order to suppress the dark current, Kim et al. inserted a thin Al₂O₃ interfacial layer between the graphene and Ge Schottky junction to passivate surface states and extend the tunneling distance (Figure 7C).¹¹⁸ At room temperature, the reported device exhibited an improved normalized photo-to-dark current ratio (NPDR) of 4.3 × 10⁷ W⁻¹ and specific detectivity of 1.9 × 10¹⁰ cm Hz^1/2 W⁻¹ under illumination of 1550 nm wavelength (Figure 7D).

Apart from group-III nitrides, silicon carbide (SiC) is also regarded as an impressive candidate material for the development of UV photodetectors due to its large band gap of 3.26 eV, high electron mobility around 900 cm² V⁻¹ s⁻¹, outstanding thermal conductivity and good chemical stability.¹²⁰ In recent years, the integration of graphene and SiC has attracted extensive attentions and shown great potential for further improvement of the UV detection ability of SiC material.^71,121–123 Li et al. demonstrated a high responsivity UV photodetector with epitaxial graphene grown on 4H-SiC surface.⁷¹ The fabricated device has a metal-oxide-semiconductor structure, where the doping type and carrier concentration of the graphene channel can be controlled by gate voltage and UV light separately. Under 325 nm laser excitation, a planar n–n–n junction is formed at gate voltage of 3 V with the maximum responsivity of 254.1 A W⁻¹ and highest EQE of 9.6 × 10⁴%. Such an impressive detection performance is attributed to the high photoconductive gain in the planar n–n–n junction induced by the photogating effect of the Graphene/SiC van der Waals heterojunction. Besides, in 2021, Guo et al. developed a scheme of the in situ growth of an epitaxial graphene lateral p–n junction on a periodic boron ion implanted (BII) Si-face 6H-SiC(0001) substrate for the first time (Figure 7E).¹²³ They found that the work function of the SiC substrate increased about 1 eV after BII treatment. Therefore, the graphene grown on the SiC regions with BII treatment show p-type doping while the graphene grown on the SiC regions without BII treatment show n-type doping. Moreover, this p–n junction demonstrated an outstanding response to 325 nm UV light owing to the absorption of SiC and the charge exchange between graphene and SiC (Figure 7F).

Wide bandgap semiconductor ZnO is also a well-established material for UV light photodetection, for its bandgap of 3.37 eV, large exciton binding energy of 60 meV at room temperature, excellent mechanical strength as well as high temperature and pressure resistance.¹²⁴ The Mott–Schottky theory proposes a Schottky junction can be formed between graphene and ZnO. However, the Fermi-level pinning (FLP) effect often occurs due to surface defect states of ZnO, resulting in an ohmic or quasi-ohmic contact at the Gr/ZnO interface accompanied with a poor photovoltaic behavior. Chen et al. addressed this issue by treating the ZnO film with H₂O₂ solution to reduce surface states.¹²⁵ The heterojunction, composed of treated ZnO film and graphene, exhibited a rectification ratio of 36 and improved self-powered photovoltaic behavior with a responsibility of 50 μA W⁻¹ when exposed to 365 nm UV light under zero bias.

Recently, Tan et al. demonstrated a solar-blind metal–semiconductor field-effect phototransistor (MESFEPT) with an exfoliated Ga₂O₃ microflake as the channel and a graphene thin film as the top gate.¹²⁶ The high-quality van der Waals contact between Ga₂O₃ and graphene gate results in an excellent gate modulation performance with negligible hysteresis and an extremely low subthreshold swing of 69.4 mV dec⁻¹. More impressively, by simply adjusting the gate voltage, the photocurrent generation mechanism of the device can be continuously switched from the photoconduction effect with fast response speed to the photogating effect with high gain. This research offered a new idea for designing multifunctional photodetectors suitable for diverse application scenarios. Chen et al. demonstrated a solar-blind position-sensitive-detector (PSD) based on graphene/Ga₂O₃ Schottky junction.¹²⁷ As shown in Figure 7G, under 254 nm illumination, the photogenerated electrons drift to the bottom electrode, while the photogenerated holes accumulated in the graphene layer, resulting in a lateral potential gradient parallel to the PSD surface. The number of holes collected by the top electrodes is correlative to the distance between the light spot and the top electrodes. Hence, the output voltages of top electrodes will change as the light spot moves along the x direction (Figure 7H).

INTEGRATING TMDC WITH 3D SEMICONDUCTOR VAN DER WAALS AS HETEROSTRUCTURES FOR PHOTODETECTION

Transition metal dichalcogenide (TMDC) is also an important member of the large 2DLMs family. The general chemical formula can be expressed as MX₂, where M represents the transition metal elements in Group 4 to Group 10 of the Periodic Table of the Elements, and X represents the chalcogen elements of S, Se, and Te. Compared with gapless graphene, the widely studied TMDCs generally show the bandgaps in the range of 1–2.5 eV,²³ and show a wide response range from visible to infrared light in the electromagnetic spectrum. Besides, their bandgaps will be wider with the decrease of their thicknesses because of the quantum confinement effect. The bulk TMDCs usually appear as indirect bandgaps. However, when the number of layers is reduced to one, they will be converted to direct bandgap semiconductors. More importantly, apart from the layer-dependent bandgap, their physical and chemical properties can also be adjusted through strategies such as vacancy engineering, phase engineering, heteroatom doping, and alloying different 2D TMDC nanosheets.² The suitable bandgaps, tunable properties and strong light-matter interaction and other diverse properties endow TMDCs a great potential for optoelectronics. Mechanical exfoliation, CVD, thermally assisted conversion (TAC), PLD, and other methods have been developed over the past few years to prepare large size and high-quality TMDC for photodetectors with high responsivity and sensitivity. Specially, integrating with traditional 3D semiconductors to further improve the detection performance have been widely exploited to accelerate the practical application of TMDCs. In the following section, we review the recent progresses and applications of photodetectors based on the heterostructures of TMDCs and some traditional 3D semiconductor materials. Table 2 presents a summary of critical performance parameters for TMDC/3D semiconductor heterostructure based photodetectors reported in recent years.

TABLE 2 Summary of the critical performance parameters for TMDC/3D semiconductor heterostructure based photodetectors.

Material (configuration)	Wavelength (nm)	Response time (τr/τd)	Responsivity (A W−1)	D* (cm Hz1/2 W−1)	EQE (%)	Working bias	References
V-MoS2/Si	808	56/825 ns	9.08 × 10−4	1.889 × 1013	—	VDS = −2	128
MoTe2/Si	980	150/350 ns	0.19	6.8 × 1013	24	VDS = 0	129
WS2/pyramid Si	980	5.2/22.3 μs	0.29	2.6 × 1014	—	VDS = 0	130
MoS2/GaAs	635	17/31 μs	0.321	3.5 × 1013	—	VDS = 0	131
WS2/GaAs	808	21.8/49.6 μs	0.527	1.03 × 1014	80	VDS = 0	132
MoS2/GaN	365	—	1.8 × 104	7.46 × 1012	6.19 × 106	VDS = 1	133
PtSe2/GaN	265	45.2/102.3 μs	0.193	3.8 × 1014	—	VDS = 0	134
MoS2/TiO2	400–800	—	2199	1.67 × 1013	—	VDS = 1, VG = 60	135
PdSe2/CdTe	780	4.9/8.3 μs	3.247 × 10−4	3.8 × 1012	51.6	VDS = −20	136
Gr/PtSe2/β-Ga2O3	245	12/212 μs	7.62 × 10−5	1.93 × 1013	—	VDS = 0	137
SnS2/4H-SiC	325	17/17 ms	2.42 × 104	7.3 × 1013	9.2 × 106	VDS = 5	138

Abbreviation: V-MoS₂, vertically oriented MoS₂.

TMDC/Si van der Waals heterostructure based photodetectors

Just as mentioned in the previous section, graphene has been a research hotspot over the past decade due to its novel structure and excellent conductivity. However, photodetectors based on graphene/3D heterostructures usually suffer the problems of high dark currents and short carrier lifetimes resulting from the semi-metallic property and lack of bandgap of graphene. Unlike graphene, TMDCs have proper intrinsic bandgaps to allow extremely high current on/off ratio which is essential for applications of high-sensitivity photodetection.^139–141 For instance, monolayer MoS₂ transistors demonstrate room-temperature current on/off ratios of 1 × 10⁸ and ultralow power dissipation.¹⁴² Thanks to that, photodetectors based on MoS₂/Si heterojunctions have attained outstanding responsivity up to 117 A W^−1.143 Besides, other TMDCs such as MoSe₂, MoTe₂, and WS₂ integrating with Si also show huge potential for high-performance photodetectors.^{129,130,144–147} However, since poor light absorption caused by atomic thickness of 2D layered materials and lack of high quality and large scale TMDC films, most of TMDC/Si based photodetectors are still far from meeting the standard of practical applications. For the sake of further improving the light absorption capability of TMDC/Si heterojunctions, Sun et al. introduced graphene quantum dots (GQDs) on the WSe₂/Si surface to function as sensitizing centers (Figure 8A).¹⁴⁸ They found that the I_ph dramatically increased at reverse bias after GQDs treatment and the responsivity improved from 2 to 707 mA W⁻¹ accordingly. When combined with GQDs, charge transfer occurred between the WSe₂ and GQDs, thus forming a built-in electric field at the WSe₂/GQDs interface. Therefore, GQDs provided an extra region for charge separation under illumination. When a reverse bias voltage was applied, electrons remained within the GQD layer, while the holes circulated through the WSe₂ film, resulting in an obvious increment of the I_ph, as the energy band diagram displayed in Figure 8B.

FIGURE 8. (A) Schematic of the GQDs/WSe2/Si heterojunction. (B) Energy band diagrams of GQDs/WSe2/Si heterojunction under reverse bias voltage while illuminated. (A,B) Reproduced with permission.148 Copyright 2018, Springer Nature. (C) Normalized photoresponse characteristic versus f curve of the FL-MoTe2/Si photodiode. Inset shows experimental setup for measuring the photoresponse speed of device. Reproduced with permission.129 Copyright 2020, John Wiley and Sons. (D) Schematic fabrication process of the ultra-flexible MoS2/Si heterojunction-based photodetector. (E) Dynamic photoresponse test under different radii and flat state as a function of time with a frequency of 5 s. (F) Device sensing performance variation with bending radii of flat, 5, 8, 10, 12, and 14 mm. (G) Stability of the MoS2/Si heterojunction with 1000 bending cycles. (D–G) Reproduced with permission.150 Copyright 2021, Royal Society of Chemistry.

In addition to responsivity and detectivity, response speed is also a key figure of merit to assess the performance of a photodetector. Unfortunately, how to achieve high responsivity and fast respond speed at the same time is a universal problem and a great challenge for researchers. Besides, the low carrier mobility and the poor crystal quality of TMDCs generally limit the respond speed at a level of milliseconds.^140,141,143 In light of this, Lu et al. started with improving material quality and optimizing device structure to solve the problem. They adopted PLD technique to fabricate the MoTe₂/Si heterojunction for high-speed and broadband photodiodes for the first time (Figure 8C).¹²⁹ The heterojunction photodiode possessed a 3 dB optical bandwidth as large as 1.0 MHz and ultrafast respond speed up to 150 ns along with a high responsivity of 0.19 A W⁻¹. In 2018, vertically oriented MoS₂ (V-MoS₂) nanosheets synthesized by CVD were transferred onto a silicon substrate to construct V-MoS₂/Si heterojunction. Owing to the enhanced light absorption efficiency and the high longitudinal intralayer carrier mobility of the V-MoS₂ nanosheets, an ultrafast response speed of 56 ns for rise time and 825 ns for fall time were observed.¹²⁸ So far, CVD has been proven to be a versatile method for the synthesis of 2DLMs with large scale and high quality. However, a high processing temperature is indispensable (>500°C) in this method which greatly restricts the scope of substrates.³⁰ To prepare 2DLMs on some flexible substrates like polyimide (PI) and polyethylene terephthalate (PET), researchers usually transfer the as-synthesized 2DLMs onto them that is prone to induce folds, cracks and contamination. Thus, there is an urgent need for developing a new growth technique with low-temperature, high-yield, and large-scale capability to realize 2D/3D semiconductors flexible optoelectronic devices. Recently, atmospheric-pressure plasma-enhanced chemical vapor deposition (AP-PECVD) technique was developed successfully to grow MoS₂ and WS₂ multilayers directly onto PET flexible substrates at temperatures of <200°C.¹⁴⁹ Three years later, an ultra-flexible MoS₂/Si heterojunction-based photodetector was successfully fabricated by the same way (Figure 8D).¹⁵⁰ Even at various harsh bending conditions, the as-fabricated device still remains an excellent photoresponse and exhibits exceptional flexibility, rollability, and durability (Figure 8E–G). As a classical example of 2D/3D semiconductor integration, TMDC/Si heterojunctions have been widely investigated.

TMDC/GaAs van der Waals heterostructure based photodetectors

Heterojunctions of MoS₂, WS₂, and other TMDCs with GaAs have also been reported successively in recent years. In 2016, Xu et al. reported a modified MoS₂/GaAs heterojunction photodetector (Figure 9A) with the responsivity and the detectivity reaching 321 mA W⁻¹ and 3.5 × 10¹³ cm Hz^1/2 W⁻¹,¹³² respectively. Herein, Si QDs were placed on the surface of MoS₂ as photoinduced dopants for improving the responsivity and a h-BN layer was inserted into the interface of MoS₂ and GaAs for suppressing the dark current. With the inserted h-BN layer, the barrier height increased for electron transferring from GaAs to MoS₂, while the transport of excess holes under illumination from GaAs to MoS₂ was almost unaffected. On the other hand, photoexcited holes produced by Si QDs inject into MoS₂ thin layer, resulting in p-type doping of MoS₂. Hence, a well suppression of dark current and high carrier collection efficiency can be expected. Lately, Jia et al. presented a WS₂/GaAs type-II van der Waals heterojunction for highly sensitive broadband photodetection (Figure 9C).¹³² The lager-area WS₂ films were synthesized by a two-step thermal decomposition method and then transferred on the prepared GaAs substrate. From the spectral response measurement, an obvious photoresponse beyond the limitation of bandgaps was observed, which can be ascribed to the interlayer excitations absorption in the WS₂/GaAs heterojunction (Figure 9D). A self-driven photodetector constructed by monolayer WS₂ films grown by CVD and GaAs was also proposed.¹⁵¹ This detector demonstrated an overall high performance in terms of a high responsivity of 65.58 A W⁻¹ at 365 nm and 28.50 A W⁻¹ at 880 nm, a large detectivity of 4.47 × 10¹² cm Hz^1/2 W⁻¹ and a fast response speed up to tens of millisecond.

FIGURE 9. (A) Schematic structure and (B) electronic band alignment of the Si QDs/MoS2/h-BN/GaAs photodetector. (A,B) Reproduced with permission.132 Copyright 2016, Elsevier. (C) A schematic diagram of the WS2/GaAs heterojunction device structure. (D) Wavelength-dependent responsivity and absorption spectra of the WS2/GaAs photodetector. (C,D) Reproduced with permission.132 Copyright 2020, Royal Society of Chemistry. (E) The schematic illustration the PtSe2/GaAs heterojunction based photodetector. (F) Both responsivity and specific detectivity of the PtSe2/GaAs heterojunction under 808 nm light illumination with different intensities. (E,F) Reproduced with permission.155 Copyright 2018, John Wiley and Sons.

The photodetectors operating in near-infrared (NIR) light range play an indispensable role in military and civil fields including optical telecommunication, target imaging, night version, security inspection, and environmental monitoring. Nowadays, the commercial market of NIR photodetectors is largely dominated by narrow bandgap semiconductors, such as PbS, InSb, InGaAs, and HgCdTe.^152–154 Even though the above devices have high sensitivity, it is undeniable that the manufacturing processes of these compounds are costly and time consuming, due to involving complicated facilities for material growth.¹⁵² TMDC with narrow bandgap may provide a possible way for low-cost high efficiency NIR photodetectors. Recently, a PtSe₂/GaAs heterojunction has been fabricated for NIR light detection as shown in Figure 9E.¹⁵⁵ It exhibited a responsivity of 262 mA W⁻¹ and a detectivity of 2.52 × 10¹² cm Hz^1/2 W⁻¹ under 808 nm light illumination with light intensity of 0.11 mW cm⁻² (Figure 9F). In addition, the PdSe₂/GaAs mixed-dimensional vdW heterojunction photodetectors also showed an impressive response to 808 nm light with responsivity up to 171.34 mA W^−1.156

Polarization-sensitive photodetectors have lots of important applications in radiometry, biomedicine, remote sensing, and polarization sensing and imaging. SnSe and SnS are two typical post-transition metal dichalcogenides (PTMD) with high light absorption coefficients (~10⁵ cm⁻¹), high carrier mobility (~10⁴ cm² V⁻¹ s⁻¹) as well as in-plane anisotropic crystal structures.^157,158 Recently, Yang et al. synthesized the SnS_xSe_1−x nanosheets by alloying SnSe and SnS and then transferred onto a GaAs substrate to construct a SnS_xSe_1−x/GaAs mixed-dimensional heterostructure.¹⁵⁹ Benefit from the alloy engineering and strong build-in potential at the heterointerface, the device demonstrated not only an excellent photoresponse of 10.2 A W⁻¹, but also a high polarization sensitivity with the dichroic ratio of ∼1.25 at 405 nm and ∼1.45 at 635 nm.

TMDC/AlGaN van der Waals heterostructure based photodetectors

AlGaN and some common TMDCs such as MoS₂ and WS₂ have very similar lattice sizes with the lattice mismatch <1%.^160,161 Therefore, compared with other 2D/3D heterostructures, TMDC/GaN heterostructures can regard as a natural lattice-matched system. Based on this, many high-quality TMDCs have been grown successfully on the GaN substrates by CVD.^32,162,163 Moreover, the band alignments show that except WSe₂/GaN having type-I heterointerface, TMDCs, and GaN generally form type-II heterojunctions which are critical for effective carrier separation.^164–166 These create a broad prospect for designing 2D/3D heterojunction-based photodetectors. As so far, the WS₂ film grown directly on GaN via low pressure chemical vapor deposition (LPCVD) shows an EQE of near 60% for red and infrared, and above 50% in the violet region.¹⁶³ In recent years, strain-modulated photoelectric properties of TMDCs have gradually attracted a widespread attention. For instance, Liu et al. improved the performance of multilayer MoS₂ photodetectors by transferring continuous multilayer MoS₂ film to a patterned GaN substrate (PGS) (Figure 10A).¹⁶⁷ On the one hand, the strain induced by the patterned GaN substrate increased the band gap of the MoS₂ film, and extended its response range to shorter wavelength. On the other head, strong light scattering effect from the patterned GaN substrate enhanced the responsivity and detectivity of the device (Figure 10B).

FIGURE 10. (A) 3D schematic diagram of the MoS2/PGS photodetector. (B) Dark current and photocurrent of the two devices (FGS and PGS) as a function of voltage under 460 nm laser illumination. (A,B) Reproduced with permission.167 Copyright 2021, American Chemical Society. (C) A schematic illustration of a PtSe2/GaN heterojunction device. (D) Light intensity-dependent responsivity and specific detectivity. (C-D) Reproduced with permission.134 Copyright 2018, Springer Nature. (E) Schematic illustration of the PdSe2/GaN Schottky junction device. (F) Normalized photocurrent of the photodetector as a function of the polarization angle. (E,F) Reproduced with permission.171 Copyright 2022, American Chemical Society.

As a member of group-10 TMDCs, 2D layered PtSe₂ shows a widely tunable bandgap characteristic, that is, 1.2 eV for monolayer, 0.21 eV for bilayer, and zero bandgap for bulk PtSe₂.^168,169 Furthermore, PtSe₂ has a high room-temperature electron mobility (210 cm² V⁻¹ s⁻¹) and better air-stability than black phosphorus (BP).¹⁷⁰ Therefore, PtSe₂ exhibits huge competitive edge for optoelectronic applications. Zhou et al. fabricated the first PtSe₂/GaN heterojunction deep ultraviolet photodetector by in situ synthesis of large-area PtSe₂ film on n-GaN substrate, which demonstrated a high responsivity of 193 mA W⁻¹, an ultrahigh specific detectivity of 3.8 × 10¹⁴ cm Hz^1/2 W⁻¹ under 265 nm deep UV light illumination at zero bias voltage (Figure 10C,D).¹³⁴ Recently, Wu et al. grew the PdSe₂ film on GaN substrate to develop a polarization-sensitive UV detector (Figure 10E).¹⁷¹ PdSe₂ has a semi-metallic property similar to PtSe₂ with a low-symmetric crystal structure, thus a high dichroic ratio up to 4.5 is achieved (Figure 10F).

TMDC and other 3D semiconductors heterostructures integration for photodetection

Unlike bulk materials, 2DLMs featured by atomic thin structures and large surface to volume ratio can be regarded as interface-type materials. Therefore, the interfaces between 2DLMs and metals or dielectrics, to a great extent, control, and determine the device performance.¹⁷² For example, the direct deposition of metal electrodes on the surfaces of MoS₂ inevitably introduces many chemical disorders and defective states, which seriously degrade the channel mobility and the photodetection ability of MoS₂ photodetectors.¹⁷³ Aiming at this problem, Pak et al. inserted a TiO₂ interlayer between exfoliated MoS₂ and metal electrodes by atomic layer deposition (ALD) to enhance the performance of MoS₂ photodetectors.¹⁷⁴ Compared with bare MoS₂ device, the TiO₂/MoS₂ device exhibited a faster response speed and a higher response to light of 450–550 nm wavelengths. The enhanced performance can be ascribed to the smooth interface of TiO₂/MoS₂ and the absorption of polycrystalline TiO₂. Another instance of TiO₂/MoS₂ heterostructure for photodetection was proposed by Liu et al.¹³⁵ Herein, a 2-nm thick Ti film was first formed by e-beam-evaporate (EBE) on the top of MoS₂, followed by the natural oxidation of the Ti film to form an ultrathin TiO₂ layer. Under light illumination, both the threshold voltages of pure MoS₂ phototransistor and the TiO₂/MoS₂ device shifted to the negative direction when the light power densities increase. But the shift was more obvious for the TiO₂/MoS₂ device which suggested an enhanced photogating effect, thus leading to an increased photocurrent in the TiO₂/MoS₂ phototransistor.

Over the past few years, TMDCs integrated with SiC have been investigated for improving the optoelectrical performance and broadening the photoresponse spectrum of SiC. In 2019, Zhang et al. chose the undoped SiC as a substrate for mechanically exfoliated MoS₂ and elevated the responsivity of pure MoS₂ photodetectors by three orders of magnitude without any external assistances.¹⁷⁵ One year later, direct growth of multilayer MoS₂ on monocrystalline SiC substrate though CVD was reported.¹⁷⁶ The maximum responsivity, photoconductive gain, and normalized detectivity for the fabricated device under 365 nm illumination are 5.7 A W⁻¹, 79.8 and 3.07 × 10¹⁰ cm Hz^1/2 W⁻¹, respectively. Besides MoS₂, Yue et al. transferred the SnS₂ nanosheets onto a n⁻-type 4H-SiC substrate for the first time (Figure 11C).¹³⁸ The n⁻-type 4H-SiC layer can function as an insulating substrate in the dark and a semiconductor under DUV irradiation, respectively. Thus, the responsivity of this SnS₂/SiC heterostructure under 325 nm light reaches into 2.42 × 10⁴ A W⁻¹, which is one order of magnitude higher than that of SnS₂/SiO₂ structure (Figure 11D).

FIGURE 11. (A) Transfer curves in dark and under different illumination power densities at a bias of 1 V. The inset in top panel shows the schematic diagram of the TiO2/MoS2 phototransistor. The threshold voltage Vth was extracted from the Vgs-axis intercept obtained from the linear fitting (dashed lines) of the transfer curve; the inset in bottom panel shows the change of ΔVth with the increase of the power density. (B) Energy band diagrams of enhanced mechanism for the TiO2/MoS2 phototransistor. Energy band diagrams for TiO2/MoS2 devices under the illumination to depict the process of carrier excitation and trapping. (A,B) Reproduced with permission.135 Copyright 2020, Springer Nature. (C) Three-dimensional (3D) schematic diagram of the SnS2/SiC PDs. (D) Photocurrent and Responsivity of SnS2/SiC and SnS2/SiO2 PDs under 325 nm illuminations at Vds = 5 V. (C,D) Reproduced with permission.138 Copyright 2021, Royal Society of Chemistry.

In recent years, TMDC/ZnO heterostructures have been focused on self-powered and broadband detection.^177–179 In a recent study, a n-MoS₂/n-ZnO heterojunction with type II band alignment was developed by depositing MoS₂ film onto ZnO film.¹⁸⁰ The dominant response band of the designed device can be modulated by the applied bias. At a low forward bias, the photogenerated electrons in MoS₂ are hindered by the interface barrier, thus a dominant photoresponse in visible waveband from ZnO was observed. While, at a higher forward bias, the interface barrier for electrons injection from MoS₂ to ZnO disappears, resulting in an enhanced photoresponse in the NIR waveband. Based on this, the device demonstrated an electrically modulated dual-color photoresponse with a responsivity of 4.56 A W⁻¹ at 400 nm under 1 V bias and a responsivity of 6.04 × 10³ A W⁻¹ at 900 nm under 8 V bias.

INTEGRATING BLACK PHOSPHORUS WITH 3D SEMICONDUCTORS AS HETEROSTRUCTURES FOR PHOTODETECTION

In 1914, black phosphorus(BP) was synthesized for the first time by conversion from white phosphorus at the high pressure of 1.2 GPa and the temperature of 200°C.¹⁸¹ However, at that time the bulk form of BP was just considered as a very common semiconductor and did not stimulate much interest from the scientific community. After about 100 years of silence, BP was rediscovered from the prospective of a 2D layered material. Since then, BP drew great attentions from the researchers around the world and the number of publications about single-layer, few-layer, and thin film form of BP began to surge. As a kind of elemental 2D material, BP has a similar layered structure to graphene and TMDC but with an unusual structural anisotropy. Each phosphorus atom bonds to three neighboring phosphorus atoms through sp³ hybrid orbitals, forming a puckered orthorhombic lattice structure.¹⁸² This unique atomic arrangement has two perpendicular and inequivalent lattice directions: the armchair direction and the zigzag direction, which make BP features strong in-plane anisotropic optical and electrical properties. For instance, the hole mobility of monolayer BP along armchair direction can reach 10 000 ~ 26 000 V⁻¹ s⁻¹ which is 16–38 times higher than that along zigzag direction.¹⁸³ Such high carrier mobility is the stem of high speed photoresponse of BP based photodetectors and wide electrical bandwidth of BP based radio frequency devices.^39,184,185 In addition, BP possesses a layer-dependent direct bandgap of 0.3–2 eV,¹⁸⁶ which means it has a wide response range from visible to mid-infrared light just filling the response gap between graphene and most TMDC materials (1–2 eV). However, due to the p-type nature of BP and the difficulty of n-doping, right now it is very challenging to build junction photodetectors based on BP only. The most reported BP photodetectors are photoconductive type but with very large dark current and low detectivity. Fortunately, the 2D nature of BP helps researchers to easily construct heterojunction with other semiconductors, especially the mature traditional 3D semiconductors, which shows great potential in wide spectral and dual-color photodetection.^187,188 Table 3 summarizes the critical performance parameters for BP/3D semiconductor heterostructure based photodetectors reported in the past few years.

TABLE 3 Summary of the critical performance parameters for BP/3D semiconductor heterostructure based photodetectors.

Material (configuration)	Wavelength (nm)	Response speed	Responsivity (A W−1)	Working bias (V)	References
BP/GaAs	514	—	3.7 × 10−5	VDS = 0	189
BP/Si WG	1550	3 GHz	0.657	VDS = 2	190
BP/Si WG	3680	—	23	VDS = 1	191
BP/Si WG	2000	1.33 GHz	0.307	VDS = 0.4	39
BP/MPG/Si WG	1550	150 MHz	10	VDS = 1.5	192
BP/Si PhCWG	3825	1.64 KHz	11.31	VDS = 0.5	193
BP/Si PPC	1547	1.42 GHz	0.125	VDS = 0.5	194

Abbreviations: MPG, metallic plasmonics grating; PhCWG, photonic crystal waveguide; PPC, planar photonic crystal.

In 2019, He et al. demonstrated a solar-blind UV and IR dual-band photodetector for the first time by transferring mechanically exfoliated BP onto the β-Ga₂O₃ film to construct a p–n heterojunction.¹⁸⁷ Under reverse bias, the photocurrent is attributed to the drift motion of the minority carriers generated in the β-Ga₂O₃ and BP. The device demonstrated a remarkable photoresponse of 88.5 and 1.24 mA W⁻¹ under 238 nm UV and 1030 nm IR irradiations, respectively. In following year, a heterojunction field-effect transistor with BP as the gate and β-Ga₂O₃ as the channel was reported.¹⁹⁵ The BP/β-Ga₂O₃ p–n junction exhibited excellent rectification characteristics with a high rectifying ratio about 10⁷ and remarkable gate-control ability with a low gate leakage current around pA as well as a minimum subthreshold swing of 260 mV dec^–1. This work further demonstrated that the BP/3D semiconductor vdW heterojunction shows great potential application value in high-performance photodetection. Recently, Wu et al. realized the uncooled dual-color infrared light detection by vertically stacked BP based vdW heterojunctions.¹⁸⁸ The device consists of a p–n–p back-to-back diode as show in Figure 12A. The p-Si substrate and BP function as the NIR and MWIR absorption layers, respectively. The MoS₂ sandwiched between them acts as an electron-collecting layer and hole-blocking barrier, which can effectively suppress the electrical crosstalk between the two heterojunction and enable them to work independently at the same time. After the NIR and MWIR photons are absorbed by the p-Si and BP, respectively, the photogenerated electrons are collected by the MoS₂ layer while the photogenerated holes are blocked by the hole barriers at the heterojunction interfaces (Figure 12B). This novel design delicately solved the signal crosstalk problem suffered by many pure 3D semiconductors based dual-color photodetectors. Besides the ultralow crosstalk of 0.05%, the specific detectivity (D*) of 6.4 × 10⁹ cm Hz^1/2 W⁻¹ at 3.5 μm and room temperature operation also verified the impressive performance of the as-fabricated device (Figure 12C). Compared with traditional dual-color IR photodetectors made of HgCdTe,¹⁹⁶ group III/V based quantum-well structures¹⁹⁷ and type II superlattices,¹⁹⁸ the vdW dual-color IR photodetector makes full use of the unique property of 2DLMs, that is, lack of dangling surface bonds which avoids the limitation of lattice matching in the process of conventional heteroepitaxial growth. Moreover, the non-refrigeration, low crosstalk, miniaturization and other advantages make it promising to become a new generation of dual-color IR photodetectors.

FIGURE 12. (A) Schematic of van der Waals dual-color infrared photodetector, showing the working mode and external circuit of the photodetector. (B) Circuit configuration of device to demonstrate the performance of dual-color photodetector. (C) Specific detectivity at different blackbody temperatures. (A–C) Reproduced with permission.189 Copyright 2022, Springer Nature. (D) Schematics of the waveguide-integrated on-chip system with BP photodetector. (E) Spectral responsivities of MIR waveguide integrated BP photodetectors with different BP's thicknesses and crystal orientation alignments. The armchair orientation of BP in both device A and B is aligned with the transverse direction in the waveguide, while the BP flake in device A is 23 nm thick and in device B is 40 nm. In device C, the armchair orientation is perpendicular to the transverse direction, while the thickness of BP is 40 nm. (D,E) Reproduced with permission.192 Copyright 2019, American Chemical Society. (F) Schematic diagrams of the PPC cavity-integrated BP photodetector, where D and S represent the drain and source electrodes (G) Photocurrent spectra (left axis) of the BP photodetector measured within a wavelength range from 1500 to 1560 nm at zero bias, where the transmission spectrum (right axis) of the finished device is also superimposed for comparison. (F,G) Reproduced with permission.195 Copyright 2021, American Chemical Society.

By right of the unique and wide bandgap range covering the electromagnetic spectrum from the visible to mid-infrared (MIR) light, BP stands out among other 2D materials. When the thickness of BP increases to eight layers (4.4 nm), its bandgap gradually gets into the MIR region (2.5–25 μm).²⁵ Covering not only the atmospheric windows but also molecular fingerprint region, photodetectors in this spectral range is widely applied in various fields, like environmental monitoring, biomedical analysis, water quality inspection, thermal imaging, and so on. Since the rediscovery of BP in 2014, it has been considered a strong candidate for MIR detection and a large number of BP based IR photodetectors have been reported successively. However, due to the insufficient light absorption, the photoresponsivity of these reported devices is usually in the range of tens to hundreds of mA W⁻¹, which is much lower than that of the state-of-the-art MIR photodetectors. Fortunately, the integration of BP with Si waveguides provides a promising solution to this problem. In comparison with traditional free-space photodetection, waveguide-integrated scheme can couple the vertical incident light into the waveguide which extends the matter-light interaction length so as to enhance the light absorption. More importantly, the waveguide-integrated photodetector enables the miniaturization of optical systems and monolithic integration with other photonic components such as lasers and modulators, which is, however, a great challenge for traditional IR detection materials due to the lattice mismatch limitation. In the past few years, BP photodetectors integrated with Si waveguides working in SWIR were first proposed.^190,192 In 2019, Huang et al. integrated MIR waveguide with BP photodetector for the first time. The noise equivalent power (NEP) is <1 nW Hz^−1/2, and the high responsivity of 23 A W⁻¹ at 3.68 μm and 2 A W⁻¹ at 4 μm are achieved at room temperature (Figure 12D).¹⁹¹ Considering the in-plane anisotropic optical property of BP, the alignment of BP crystal orientation with respect to the waveguide system seems a more critical factor to the performance of the device than the thickness of BP. As shown in Figure 12E, the device tends to show a higher responsivity, when the armchair crystal orientation of BP is parallel to the polarization direct of propagation mode in waveguide. Later, a high speed hybrid BP/Si waveguide photodetector was achieved with a high responsivity of 306.7 mA W⁻¹ at 2 μm and a 3 dB-bandwidth up to 1.33 GHz.³⁹ Another waveguide-integrated BP photodetector working at 3.825 μm was reported with a responsivity as high as 11.31 A W⁻¹ and a NEP of 0.012 nW Hz^−1/2 by utilizing the slow light effect in photonic crystal waveguides.¹⁹³ In 2021, a BP photodetector integrated with a Si planar photonic crystal cavity (PPC) was demonstrated as shown in Figure 12F.¹⁹⁴ The light guided into the PPC can circulate many times between the two air holes inserted in the waveguide. Unlike waveguide-integrated BP photodetectors, a long channel is necessary for enough light absorption, this PPC integrated BP photodetector only has a 7 μm long channel. Such a short channel ensures a dark current level of 20 nA and a high response speed with a 3 dB bandwidth exceeding ∼1.42 GHz. In addition, the photoresponsivity is improved by 36-fold at the resonant mode of the PPC cavity (Figure 12G).

OTHER EMERGING 2D LAYERED MATERIALS AND 3D SEMICONDUCTORS HETEROSTRUCTURES INTEGRATION FOR PHOTODETECTION

Beyond graphene, TMDC and BP, there are many other emerging 2DLMs with diverse properties exploited in various functional electronic and optoelectronic devices recently. Since the newly discovered 2DLMs suitable for photodetection are too much, herein, we just take the widely investigated and impressive 2DLMs such as tellurium (Te), Bi₂O₂Se, and MXenes as examples.

Te is a new member of elemental 2DLMs family emerging in 2017. Te possesses a widely tunable bandgap ranging from 0.35 to 1.265 eV similar to that of BP but has a better environmental stability.¹⁹⁹ Interestingly, the bulk Te is stacked by one-dimensional (1D) helical Te atomic chains along x and y axes and the 1D Te atomic chain consists of adjacent Te atoms connecting via covalent bonds.²⁰⁰ In this regard, the crystal structure of Te shows a sharp contrast with other 2DLMs that only have layered structures with interlayer vdW forces. The unique helical chain structure endows Te a high carrier mobility and in-plane anisotropic physical properties. In 2020, Tong et al. achieved polarized infrared imaging with the degree of linear polarization over 0.8 based on 2D Te under scattering environment, proving an outstanding polarized photodetection ability among all 2D materials.²⁰¹ Lately, a low-dimensional infrared photodetector based on 1D Te nanowires with an excellent blackbody responsivity of 5.19 A W⁻¹ was reported by Peng et al.²⁰² These results indicate Te has a great application potential for room temperature infrared photodetection. More recently, Zheng et al. constructed a p-Te/n-Si mixed-dimensional vdWH photodiode with a type-I band alignment (Figure 13A,B).²⁰³ Benefiting from the high-quality junction interface and large junction electric field, the device exhibited a photoresponse performance in self-powered mode with a high responsivity of 6.49 A W⁻¹ and a decent specific detectivity of 7.79 × 10¹² cm Hz^1/2 W⁻¹ under 808 nm illumination. Furthermore, a high polarization ratio of 2.1 was achieved under 635 nm laser illumination (Figure 13C).

FIGURE 13. (A) Schematic illustration of the Te/Si heterojunction with a patterned window. (B) Energy band diagram of the Te/Si vdWH under light irradiation. (C) Polar plot of normalized photocurrent for the wavelength of 635 nm at zero bias voltage. (A–C) Reproduced with permission.204 Copyright 2022, Royal Society of Chemistry. (D) Structure of a Bi2O2Se/p-Si heterostructure photodetector. (E) Schematic energy band diagram of the Bi2O2Se/p-Si heterojunction. (F) R, S, and D* of the Bi2O2Se/p-Si heterojunction photodetector at different wavelengths. (D–F) Reproduced with permission.209 Copyright 2023, American Chemical Society. (G) The cross-sectional schematic of the Ti3C2Tx/GaAs Schottky junction. (H) The energy band diagram of the Ti3C2Tx/GaAs Schottky junction. (I) The time-dependent photovoltaic photoresponse of the device under different wavelength light signals. (G–I) Reproduced with permission.210 Copyright 2021, Elsevier.

Bi₂O₂Se is another newly emerging 2DLM which has stimulated passionate research enthusiasm in recent years since its fascinating properties including an ultrahigh carrier mobility (~2.8 × 10⁵ cm² V⁻¹ s⁻¹ at 2 K), a great current on/off ratio (>10⁶) and a moderate bandgap of 0.8 eV.^204,205 Recently, Tong et al. reported a Bi₂O₂Se phototransistor with a wide response range (360–1800 nm) and an impressively high peak responsivity of 108 696 A W⁻¹ at 360 nm.²⁰⁶ Figure 13D displays the body-centered tetragonal structure of Bi₂O₂Se composed of alternating stacks of Bi₂O₂ and Se layers along c axis.²⁰⁷ It is reported that there is no clear-defined vdW gap in Bi₂O₂Se due to the neighboring Bi₂O₂ layers and Se layers are bonding via weak electrostatic interactions. Hence, when the Bi₂O₂Se crystals are cleaved, they will separate along the Se planes. Recently, Ling et al. developed a Bi₂O₂Se/p-Si heterojunction with a self-powered photoresponse characteristic at 365–1550 nm and a highest detectivity of 4.43 × 10¹² cm Hz^1/2 W⁻¹ at 980 nm (Figure 13D–F).²⁰⁸

MXenes is a relatively new category of 2DLMs which refers to transition metal carbides, nitrides and carbonitrides. This type of materials can be represented by a general formula, M_n+1X_n or M_n+1X_nT_z (n = 1, 2 or 3), where M, X, T_z represent transition metal (Sc, Ti, Cr, Mo, Ta, etc.), carbon and/or nitrogen, and surface functional groups (–O, –OH, –F), respectively. In the typical crystal structure of MXenes, the X atomic layers are generally sandwiched between the outer M atomic layers with the T_z functional groups connecting to the surface of MXenes.²⁰⁹ The most intriguing feature of MXenes is their surface functional terminations that enable to tailor the electrical and optical properties of MXenes. For instance, the MXenes without surface functional groups usually exhibit semimetal property. While, most of MXenes with surface functional groups have a variable bandgap range from 0.04 to 3.23 eV.²¹⁰ Currently, many kinds of MXenes combined with different 3D traditional semiconductors including Si,²¹¹ Ge,²¹² GaAs,²⁰⁹ Ga₂O₃,²¹³ and so on have been investigated for photodetection applications and shown impressive performances. Zhang et al. designed a self-driven Schottky photodiode by dripping Ti₃C₂T_x MXene solution on the GaAs substrate (Figure 13G).²⁰⁹ The I–V curve of Ti₃C₂T_x/GaAs heterojunction in dark condition indicates a pronounced rectifying behavior (Figure 13H). The photocurrent generation mechanism can be explained by the energy band diagram of the junction (Figure 13I). On one hand, the built-in potential across the depletion region will drive the photogenerated holes in GaAs into the external electrodes at Ti₃C₂T_x film side, giving rise to photocurrent. On the other hand, the disordered stack of few Ti₃C₂T_x nanoflakes on GaAs are able to excite the plasmon-induced hot electrons which are also responsible for the photoresponse to infrared wavelength exceeding the absorption edge of GaAs. The assembled Ti₃C₂T_x/GaAs Schottky junction exhibited an excellent detection ability with a broad response range up to 980 nm, a high responsivity of 1.46 A W⁻¹, a remarkable specific detectivity of 1.23 × 10¹³ cm Hz^1/2 W⁻¹ and a high photo-to-dark-current ratio of 5.6 × 10⁵.

SUMMARY AND OUTLOOK

The 2D/3D vdWH combine the advantages of both 2DLMs and traditional 3DBMs, providing a platform for design of multifunctional photodetectors that pure 2D or 3D materials cannot achieve. In this review, we first discuss the unique advantages of 2D/3D vdWH and give a brief introduction to the structure categories, working mechanisms, and the construction methods of 2D/3D vdWH photodetectors. Then we mainly focus on the applications of 2D/3D vdWH in photodetection and photocollection fields. Numerous experimental works have proved that the integration of 2D and 3D materials demonstrates a great application potential for broadband photodetection, dual-color photodetection, high-sensitive photodetection, polarization-sensitive photodetection, ultrafast photodetection and so on. However, the research on 2D/3D vdWH is still at a preliminary stage. Despite the great success that has been achieved, there are still many challenges awaiting us to break through.

1. The device performance depends on the material quality and heterojunction interface quality in a large degree. Although the assembly of 2DLMs and 3DBMs gets rid of the constraint of lattice matching, the transfer processes of 2DLMs inevitably introduce organic residue and other contaminations or winkles to 2DLMs which usually degrade the interface quality, leading to a deviation in actual device performances from expected performances.²¹⁴ Meanwhile, the direct in situ growth of 2DLMs on 3DBM substrates by CVD faces the problems of stability and controllability. Consequently, the primary challenge of 2D/3D vdWH based photodetector is the lack of a clean and efficient transfer technique and controllable large scale growth methods.
2. Doping is widely used to manipulate the electric property of semiconductors. Although some doping strategies, like substitutional doping, surface charge transfer doping, intercalation and electrostatic doping have been developed to achieve both nondegenerate and degenerate modulation of 2DLMs.²¹⁵ How to precisely control both doping dose and carrier density, and ensure stable doping still pose an unresolved challenge.
3. A good electrical contact between metals and semiconductors is of vital importance to the device performance. The rough and contaminated interface between deposited metals and 2DLMs created by conventional metallization process generally leads to high contact resistance and strong FLP effect that hinder the improvement of device performances.²¹⁶ Although several effective methods such as transferred metal contact,¹⁷³ alloy metal contact,²¹⁷ 1D edge electrical contact,²¹⁸ and vdW 2D metal contact²¹⁹ have been reported successively to obtain clear contact interface and relieve the FLP effect, complicated fabrication process and low efficiency are the common issues of these methods. Therefore, how to achieve a good metal–semiconductor contact is the third challenge of 2D/3D vdWH based photodetector.
4. Stability is also crucial for the practical application and commercialization of 2D/3D vdWH based photodetectors. Although the stability may not be a concern for most of traditional 3DBMs, it is a thorny problem for many 2DLMs such as BP, air exposure often causes serious device performance degradation. Therefore, how to improve device stability by suitable package is also an important challenge of 2D/3D vdWH based photodetector.

Although much progress has been achieved, considering the aforementioned challenges listed above, there is still a variety of research directions are worthy of being further explored in the near future as listed below.

1. The high-efficiency and high-precision assembly of 2D/3D vdWH is an essential prerequisite for both experimental research and industrial production. Nonetheless, the current assembly process relies too much on the personal experience and operating skills of the researchers. This greatly increases the obstacles in the sample preparation process and reduces the controllability, repeatability and efficiency of mass production. Therefore, it is a promising research direction to develop a high-efficiency and high-precision assembly technology, such as an automated material transfer and growth controlled by artificial intelligence.
2. Despite numerous high-gain photodetectors with responsivity over several A W⁻¹ based on 2D/3D vdWH have been reported. Few articles gave a detailed explanation and a comprehensive analysis to the origin of the photogain. Therefore, there should be more focus on the physical mechanisms behind the extraordinary phenomenon. The carrier dynamics, interlayer charge transfer at the heterointerfaces and energy band alignment regulation of 2D/3D vdW heterojunctions await more explorations by means of theoretical calculations and software simulations. In light of this, further investigations of detection mechanisms of 2D/3D vdWH based photodetectors could be another fruitful research direction.
3. For practical applications, the performance of a photodetector should be evaluated from multiple aspects, such as responsivity, specific detectivity, response speed as well as stability. However, most reported 2D/3D vdWH based photodetectors lack overall and systematic assessments of the performance at a standard measurement condition. Accordingly, a more comprehensive and standard characterization of the performance of 2D/3D vdWH based photodetectors should be the third research direction.
4. For imaging applications, it is highly desirable to fabricate large-scale arrayed photodetectors integrated with CMOS readout integrated circuit. Although, 2DLMs show a great potential to integrate with silicon materials, how to realize fully compatibility with the state-of-art silicon-based technology for 2DLMs remains a thorny issue. Second, high-quality and large-area 2DLM continuous films is also difficult to acquire. As a result, most reported 2D/3D vdWH based photodetectors only could work with single-pixel imaging mode. Consequently, exploration of relevant integration techniques for CMOS-compatible 2D/3D vdWH based photodetector arrays could be a stirring research direction.
5. 2D/3D vdWHs may permit a brighter prospect to a specific photodetection field due to their great integration capability and high-quality heterojunction interfaces without lattice mismatch. For room temperature infrared detection, the narrow-bandgap 2DLMs show a strong competitiveness in the aspect of widely tunable bandgaps as well as simple and low-cost synthesis methods. The large tolerance to lattice mismatch allows to design 2D/3D vdWH with many original structures used in dual-band or multiband detection. Additionally, anisotropic crystal structures of many 2DLMs further expand the potential of 2D/3D vdWHs for high-sensitivity polarization detection. Therefore, exploring the unconventional functions and applications of 2D/3D vdWH based photodetectors could be the fifth research direction.

ACKNOWLEDGMENTS

We acknowledge the support from the National Natural Science Foundation of China (Grant No. 62174063, 62174061, 61904184, 61974174), the National Key Research and Development Program of China (Grant No. 2022YFB3605104), the Key Research and Development Program of Hubei Province (Grant No. 2021BAA071), the Key Laboratory of Infrared Imaging Materials and Detectors, the Shanghai Institute of Technical Physics, the Chinese Academy of Sciences (Grant No. IIMDKFJJ-21-07), the Fundamental Research Funds for the Central Universities (Grant No. 2020kfyXJJS124) and the Director Fund of WNLO.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.