Conductive Atomic Force Microscopy of Semiconducting Transition Metal Dichalcogenides and Heterostructures

Abstract

Semiconducting transition metal dichalcogenides (TMDs) are promising materials for future electronic and optoelectronic applications. However, their electronic properties are strongly affected by peculiar nanoscale defects/inhomogeneities (point or complex defects, thickness fluctuations, grain boundaries, etc.), which are intrinsic of these materials or introduced during device fabrication processes. This paper reviews recent applications of conductive atomic force microscopy (C-AFM) to the investigation of nanoscale transport properties in TMDs, discussing the implications of the local phenomena in the overall behavior of TMD-based devices. Nanoscale resolution current spectroscopy and mapping by C-AFM provided information on the Schottky barrier uniformity and shed light on the mechanisms responsible for the Fermi level pinning commonly observed at metal/TMD interfaces. Methods for nanoscale tailoring of the Schottky barrier in MoS₂ for the realization of ambipolar transistors are also illustrated. Experiments on local conductivity mapping in monolayer MoS₂ grown by chemical vapor deposition (CVD) on SiO₂ substrates are discussed, providing a direct evidence of the resistance associated to the grain boundaries (GBs) between MoS₂ domains. Finally, C-AFM provided an insight into the current transport phenomena in TMD-based heterostructures, including lateral heterojunctions observed within Mo_xW_1–xSe₂ alloys, and vertical heterostructures made by van der Waals stacking of different TMDs (e.g., MoS₂/WSe₂) or by CVD growth of TMDs on bulk semiconductors.

1. Introduction

In the last years, transition metal dichalcogenides (TMDs) have attracted an increasing scientific interest because of their unique and tunable electronic structure, holding great promise for next generation applications in electronics and optoelectronics [1,2]. These layered materials are composed by the stacking of MX₂ trilayers (M = Mo, W, etc. and X = S, Se, Te, etc.) held together by van der Waals bonds. Each trilayer consists of a layer of hexagonal-packed transition-metal atoms (M) sandwiched between two layers of chalcogen atoms (X), with a strong intralayer covalent bond between the different atoms. Depending on the composition and internal structure of the individual trilayers, the electronic behavior of TMDs can range from insulating to semiconducting and semi-metallic. Furthermore, the optical and electronic properties of these layered materials depend on the stacking order and number of layers [3]. For instance, bulk or multilayer MoS₂ exhibit an indirect band gap of ~1.3 eV, whereas monolayer MoS₂ presents a direct band gap of 1.8–1.9 eV. A similar behavior also occurs in other semiconducting TMDs. The tunable band gap along with the high quantum efficiency and optical absorption make TMDs interesting for optoelectronics [1]. In addition, due to the superior electrostatic control of channel conductance in field effect transistors (FETs) with an ultra-thin TMD body [4,5], these 2D semiconductors are currently regarded as strategic materials for the post-Si complementary metal oxide semiconductor (CMOS) technology [6]. Very interesting performances in terms of the on/off current ratio (10⁶–10⁸) and a subthreshold swing approaching the ideal value (60 meV/decade) have been already demonstrated in the early studies on single layer [4] and multilayer MoS₂ FETs [5]. More recently, ultra-scaled MoS₂ FETs with a 1 nm gate length [7] or with a sub −10 nm channel length [8] have been reported, further supporting the expectations for MoS₂ transistors to further extend the validity of Moore’s law [9]. In addition, as replacements of silicon for the CMOS technology, TMDs are also currently employed as building blocks of new transistor concepts based on vertical van der Waals heterostructures [10,11,12], or in-plane heterojunctions obtained by engineering the TMDs composition [13] or even the dielectric environment of the 2D semiconductors [14].

To date, many of these TMD-based (opto)electronic devices have been fabricated using small flakes obtained by mechanical exfoliation from the parent bulk crystals, due to the ease of preparation in academic laboratories [4,10]. However, the lateral size (few micrometres) and the low density of flakes produced by this approach are unsuitable for practical applications, and large area deposition methods of TMDs are required for a scale-up of these technologies. In the last years, many progresses have been made in the growth of TMDs by chemical vapor deposition (CVD). The direct CVD growth of TMDs, in particular MoS₂, on noncatalytic insulating or semiconducting substrates (including SiO₂ [15,16,17], sapphire [18], and GaN [19]) is possible at relatively low temperatures in the range from 700 to 900 °C. This represents an important aspect for the integration of this process in the CMOS devices fabrication flow. The CVD growth of highly uniform monolayer TMDs on 100 mm [20] and, more recently, 150 mm wafers [21] have been demonstrated. Furthermore, alternative chemical deposition methods, e.g., atomic layer deposition (ALD) [22], or physical depositions, e.g., molecular beam epitaxy (MBE) [23] and pulsed laser deposition (PLD) [24], are currently explored to achieve layer-by-layer growth of MoS₂ on a large area, up to the wafer scale.

Table 1. Recent progresses in the transition metal dichalcogenide (TMD) materials growth and devices.

Year	Achievements	Ref.
2011	Monolayer MoS2 n-type FET with a high-k top gate dielectric	[4]
2011	Integrated logic circuit based on a single-layer MoS2	[25]
2012	Monolayer WSe2 p-type FET with a high-k top gate dielectric	[26]
2012	Chemical vapor deposition (CVD) of MoS2 atomic layers on SiO2 substrates	[15]
2014	FET built from all 2D material components	[27]
2014	Vertical and in-plane heterostructures of WS2/MoS2 monolayers grown by CVD	[13]
2014	Atomic layer deposition (ALD) of MoS2 thin films	[22]
2015	CVD of epitaxial monolayer MoS2 on sapphire	[18]
2015	Metal organic CVD of a highly uniform monolayer MoS2 on 100 mm SiO2 wafers	[20]
2015	WSe2-based CMOS and integrated circuits	[28]
2015	Tunnel diodes based on MoS2/WSe2 vdW heterostructures grown by CVD	[12]
2016	MoS2 FET with a 1 nm gate length	[7]
2016	MoS2 FET with a sub −10 nm channel length	[8]
2016	CVD of epitaxial monolayer MoS2 on gallium nitride	[18]
2016	Pulsed laser deposition (PLD) of MoS2 thin films	[24]
2017	Ambipolar MoS2 FET by tailoring the Schottky barrier with oxygen plasma	[29]
2017	Ohmic contacts to monolayer MoS2 by van der Waals bonded metal/h-BN electrodes	[30]
2017	Demonstration of a microprocessor based on CVD grown bilayer MoS2	[31]
2017	Hot electron transistor with GaN emitter, graphene base, and WSe2 base-collector barrier	[32]
2018	Batch CVD growth of uniform monolayer MoS2 on 150 mm soda-lime glass wafers	[21]
2019	Van der Waals contacts between 3D metals and 2D semiconducting TMDs	[33]

summarizes some of the most relevant achievements in the TMD devices and materials growth in the last years.

In spite of these recent progresses, several issues need to be addressed to fully exploit the potential of TMDs in real applications. In addition to the request of optimized thickness uniformity, defectivity, and doping of the deposited films, a further challenge for materials growth is represented by the engineering of TMD vertical or lateral heterojunctions [12]. On the other hand, low resistance ohmic contacts and selective-area doping of TMDs are two major requirements for the fabrication of devices based on these materials.

High resolution structural, chemical, and electrical characterizations revealed that the overall electronic properties of TMD_S are ultimately determined by peculiar nanoscale defects/inhomogeneities (e.g., point or complex defects, thickness fluctuations, grain boundaries, wrinkles, etc.), which can be intrinsic of these materials or dependent on the growth method, on the substrates and on the device fabrication processes. In this context, electrical atomic force microscopy methods, such as conductive atomic force microscopy (C-AFM), scanning capacitance microscopy (SCM), Kelvin probe force microscopy (KPFM), and scanning microwave impedance microscopy (SMIM) proved to be essential tools to understand the nature and electrical activity of defects/nonuniformities in 2D materials, and more specifically in TMDs [34,35,36,37,38,39]. In particular, C-AFM allows performing a high-resolution current mapping and local current-voltage (I–V) characterization at the nanoscale [40,41,42,43]. Hence, it is the method of choice to investigate the mechanisms of current injection from contacts to 2D materials [44,45], the lateral homogeneity of conductivity of 2D semimetals (graphene) [46,47,48], and semiconductors (TMDs) [44], as well as the transversal current transport across thin dielectric films [49,50] or 2D insulators (such as h-BN) [51,52,53]. Furthermore, C-AFM allows investigating the vertical current injection across the van der Waals heterostructures of 2D materials [12] or their heterojunctions with bulk semiconductors [19,54,55,56]. Figure 1 schematically illustrates these possible applications of C-AFM in the characterization of 2D materials and heterostructures.

This paper reviews some relevant applications of the C-AFM technique to the investigation of nanoscale transport properties in TMDs, discussing the implications of the local phenomena in the overall behavior of devices fabricated on these materials.

Section 2 includes recent studies of the current injection mechanisms at the metal/TMDs junctions. Nanoscale resolution current spectroscopy and mapping by C-AFM provided information on the Schottky barrier height (SBH) uniformity and shed light on the mechanisms responsible for the Fermi level pinning commonly observed at the metal/TMDs interface. The application of this approach to nanoscale mapping of the SBH distribution in MoS₂ thin films subjected to oxygen plasma prefunctionalization is illustrated, and the practical implications of these results in the realization of ambipolar MoS₂ transistors are also demonstrated.

Section 3 presents experiments on local conductivity mapping in monolayer MoS₂ grown by CVD, providing a direct evidence of the role of grain boundaries (GBs) between MoS₂ domains on the overall current transport in these layers.

Finally, Section 4 presents C-AFM studies of the current transport in TMD-based heterostructures, including lateral heterojunctions present in TMDs alloys, and vertical van der Waals heterojunctions formed by the stacking of different TMDs and of TMDs with bulk semiconductors.

2. Schottky Barrier Height Mapping at Metal/TMDs Junction

The lack of out-of-plane bonds in TMDs initially led to the expectation that the Schottky barrier at the interface between metal contacts and TMDs would approach the Schottky–Mott limit of weak pinning [57]. However, contrary to this expectation, experiments showed the opposite behavior, i.e., the SBH for most semiconducting TMDs is only weakly dependent on the metal work function, and this effect has been attributed to a Fermi level pinning [58,59,60,61]. Clearly, this phenomenon has strong implications on TMDs-based transistors, as the current injection/extraction from the source/drain contacts rules the overall device behavior [62]. As an example, in the case of MoS₂ thin films (produced either by exfoliation or by CVD) most of the elementary metals exhibit a Fermi level pinning close to the MoS₂ conduction band, resulting in SBH values for electrons injection ranging from 20–60 meV for low work function metals (Sc, Ti, etc.) to 150–250 meV for high work function ones (Ni, Pt, etc.) [58]. As a matter of fact, the low SBH for electrons translates into a high SBH for holes. Hence, n-type FETs with an electron accumulation channel can be easily obtained with unintentionally n-type doped MoS₂, whereas the fabrication of MoS₂ FETs with the complementary p-type behavior is challenging, due to the difficulty to inject holes in the inversion channel [63,64]. Contrary to the case of MoS₂, a p-type behavior is commonly observed in monolayer WSe₂ FETs [26], whereas ambipolar transistors have been demonstrated using a few layers of MoTe₂ [65]. In spite of these differences, the electrical behavior of TMDs-based FETs is generally dominated by the Schottky barrier of source and drain contacts.

The origin of the Fermi level pinning at the junction between metals and TMDs is currently an object of debate. According to the commonly accepted models of Fermi level pinning at metal/semiconductor interfaces, this effect can be ascribed either to metal-induced gap states (MIGS) [66] or to the disorder induced gap states (DIGS) [67]. In the DIGS model, the Fermi level pinning is due to states in the electronic structure of the host semiconductor associated to defects in its crystalline structure. On the other hand, the MIGS model predicts that the states in the gap are induced by the interaction with the metal, i.e., the metal dominates over the substrate in pinning the Fermi level. Recently, ab-initio simulations of defect-free TMDs showed that the metal/TMD junctions largely follow the MIGS model, similar to three-dimensional semiconductors despite their different structure and bonding [59]. In this context, Fermi level depinning by the insertion of an ultra-thin tunnel barrier (such as monolayer h-BN) between contacts and MoS₂ has been recently reported [30]. The main function of this barrier layer is increasing the physical separation between the MoS₂ and the contact electrode, thereby minimizing the metal/MoS₂ interfacial interaction responsible for the creation of MIGS. On the other hand, early nanoscale electrical investigations (based on scanning tunneling microscopy) highlighted the possible role of the defects present at the MoS₂ surface in the Fermi level pinning phenomenon [68].

As a matter of fact, SBH mapping with nanoscale resolution is required to disentangle the effect of surface (or near surface) defects in the TMD materials from that of metal-induced gap states. To this purpose, the conductive tip of C-AFM has been employed as a nanoscopic metal electrode to record local I-V characteristics on the TMDs surface, from which the SBH was quantitatively evaluated [44,45,69]. Contacting the TMD surface with a sliding metal tip presents the additional advantage of excluding eventual reactions that have been reported at the metal/TMD interface (for some metal species) when the contact is fabricated by evaporation or sputtering [70,71]. In this way, the effect of the interaction with a metal electrode in close proximity to the TMD surface can be investigated. In the following subsection, recent studies of the Fermi level pinning in TMDs using C-AFM are discussed.

2.1. C-AFM Investigations of Fermi Level Pinning in TMDs

The C-AFM technique was firstly employed in the current-voltage (I-V) spectroscopy mode to investigate the current injection to the surface of multilayer MoS₂ exfoliated on a SiO₂ substrate [44]. Local I-V measurements were performed under ambient conditions using a Pt coated Si tip (curvature radius r_tip ≈ 10 nm) connected to a high sensitivity current amplifier, while a bias was applied to a macroscopic front contact, as schematically illustrated in Figure 2a. Figure 2b shows a set of 25 I-V_tip curves (with V_tip as the bias referred to the tip) measured on the MoS₂ surface on a 0.5 × 0.5 μm area. All the curves are asymmetric from a positive (forward) to negative (reverse) bias, indicating a Schottky behavior of the Pt/MoS₂ contact. A representative forward I-V_tip characteristic from this array of measurements is reported in Figure 2c. In the semilog-plot, a linear increase of the current over more than two decades (from 1 × 10⁻¹⁰ to 5 × 10⁻⁸ A) is observed, followed by a saturation. The SBH (Φ_B) and the ideality factor (n) for the nanoscopic Pt/MoS₂ contact were evaluated by fitting the low voltage region of the forward I-V_tip curve with the thermionic emission law:

$$I=A{A}^{\ast }{T}^2\exp \left(-\frac{q{\Phi }_B}{k_{B}T}\right)\exp \left(\frac{q{V}_{tip}}{n{k}_{B}T}\right)$$

(1)

where q is the electron charge, k_B is the Boltzmann constant, T is the absolute temperature (T = 300 K), A = πr_tip² is the tip contact area, and A* is the Richardson constant of multilayer MoS₂ [44]. In particular, Φ_B = 307 meV and n = 1.61 were determined from the intercept and the slope of the linear fit of the ln(I)-V_tip characteristic for V_tip < 0.3 V. The current saturation observed on the semilog-scale at higher bias values is due to a series resistance contribution R.

Since the downward curvature in the high voltage region of the I-V_tip curves depends both on n and R, the Cheung’s method [72] was applied to evaluate the R contribution. In this method, the function H is defined as H = V_tip−nk_BT/q ln[I/(AA*T²)], which depends on I as H = nΦ_B + IR. Figure 2d shows a plot of H vs. I obtained from the forward bias I-V_tip characteristic in Figure 2c. R = (8.02 ± 0.12) × 10⁵ Ω was evaluated by the slope of the linear fit for current values larger than 0.1 μA. The R value is mainly due to the spreading resistance R_spr, associated to current spreading into multilayer MoS₂ from the nanoscale contact. R_spr is related to the local resistivity ρ_loc of MoS₂ under the tip as R_spr = ρ_loc/(4r_tip), being r_tip the contact radius. Hence, the local resistivity ρ_loc could be estimated from the R values measured at each tip position. By performing the same analysis on the full set of I-V_tip characteristics of Figure 2b, the distributions of the local SBHs, ideality factors, and resistivity values at the different tip positions on MoS₂ were determined. The histograms of the Φ_B, n, and ρ_loc values are reported in Figure 3a–c, respectively.

The distribution in Figure 3a shows an average SBH of 300 meV with a standard deviation of 24 meV. This average value is in good agreement with the SBH evaluated for deposited Pt contacts on multilayer MoS₂ [58]. According to the Schottky–Mott theory, the ideal SBH value for a metal/semiconductor contact without Fermi level pinning is expressed as:

Φ_B = W_M − χ

(2)

where W_M is the metal work function and χ is the semiconductor electron affinity. In the specific case of Pt and multilayer MoS₂, Φ_Μ ≈ 5.4 eV and χ ≈ 4.1 eV, result in an ideal Φ_B ≈ 1.3 eV. Clearly, the experimental SBH values are much lower than the ideal one. The histogram in Figure 3b shows that n is close to unity only on ~10% of the investigated MoS₂ area, whereas the average value of n is 1.60 with a standard deviation of 0.23. Generally, the deviation of n from unity indicates that the current transport is not perfectly described by the thermionic emission theory and can be ascribed to the presence of surface states [73]. Finally, an average resistivity of 2.99 Ωcm with a standard deviation of 0.68 Ωcm was estimated from the distribution in Figure 3c. The lateral variations of ρ_loc can be ascribed to inhomogeneities in the carrier concentration and/or in the carrier mobility of MoS₂.

Since these C-AFM analyses were carried out under ambient conditions, the measured SBH distribution can be affected by the presence of a water meniscus under the tip, which is known to cause a degradation of the lateral resolution. In this context, performing C-AFM under a high vacuum or within an environmental chamber has been shown to allow current mapping and spectroscopy with a greatly improved resolution [74].

Recently, high resolution current mapping by an environmental chamber C-AFM, combined with atomic resolution scanning tunneling microscopy (STM), provided further insight on the nature of defects responsible for the Fermi level pinning in MoS₂ [45] and other TMDs (MoSe₂, MoTe₂, WS₂, and WSe₂) [75]. Figure 4a shows a lateral force microscopy (LFM) and the corresponding AFM topographic image (inset) of a freshly cleaved MoS₂ sample, showing a smooth surface in both images [45]. On the other hand, the simultaneously recorded C-AFM map (Figure 4b) shows a laterally inhomogeneous current injection through the nanoscopic contact between the conductive tip (made of boron doped diamond) and the MoS₂ surface. Multiple dark circular features (with radii ranging between 3 and 4 nm) are present in this image, with a higher current than the surrounding areas, indicating a lower tip/MoS₂ contact resistance. The surface density of these features was found in the range between 10¹⁰ and 10¹¹ cm^–2, depending on the sample and location. Atomic resolution scanning tunneling microscopy (STM) analyses of the MoS₂ surface were also carried out to understand the nature of the high current spots in the C-AFM images. The STM maps revealed the presence of characteristic features appearing as circular depressions on which the MoS₂ lattice is superimposed (see Figure 4c). The typical lateral size of these depressions was up to 5 nm and their areal density was (8 ± 3) × 10¹⁰ cm⁻², similar to the density of conductive features in the C-AFM images. The absence of such depressions in the topographic and LFM images (Figure 4a) suggested that the observed features in the STM maps are electronic in nature and induced by subsurface defects. They were associated to Mo-vacancies or Mo-substitutional defects, located below the outermost S layer [45]. In addition to these subsurface defects, sulfur vacancies (see Figure 4d) were found to be ubiquitously present in the topmost S layer of MoS₂, with an areal density of (7 ± 4) × 10¹² cm⁻². No specific features corresponding to these point defects were observed in the C-AFM map.

Figure 5a shows a map of the SBH distribution extracted from an array of 128 × 128 I-V curves collected by the conductive diamond tip on the freshly cleaved MoS₂ surface. More specifically, the local Φ_B values were obtained by fitting the low forward bias region of the individual I-V curves with the thermionic emission model, as already discussed in Figure 2c. This high resolution SBH map showed minimum Φ_B values of ~0.3 eV corresponding to the subsurface defect regions, whereas a nearly constant SBH of ~0.53 eV was observed in the surrounding areas. As a matter of fact, these SBH values are much smaller than the ideal one (~1 eV) predicted by the Schottky−Mott Equation (2) in the specific case of a boron doped diamond tip (work function Φ_M = 5.1 eV) in contact with MoS₂. Hence, even in the regions without the subsurface defects, a Fermi level pinning occurs, although less intense than on defects. The authors excluded that this Fermi level pinning can be due to the sulfur vacancies, and ascribed this effect to gap states induced by the interaction of the metal tip with MoS₂, as indicated also by ab-initio calculations [59]. Clearly, the high density of low Φ_B ≈ 0.3 eV patches (separated by a distance of ~10 nm from each other) dominates over the background with higher Φ_B ≈ 0.53 eV and determines the effective SBH. Based on these considerations, the results of this high resolution C-AFM investigation are in agreement with those of the C-AFM study in [44], as well as with SBH measurements on macroscopic contacts [58].

To get a more complete description of the Fermi level pinning effect, the dependence of the local SBH on the metal tip work function was also investigated [45]. Figure 5b shows a plot of the experimental Φ_B values as a function of the metal tip work function both on the defect-free MoS₂ areas and on the subsurface defect regions. In addition to the conductive diamond tip (Φ_M = 5.1 eV), a PtSi tip (Φ_M ≈ 4.9 eV) and a highly n-type doped Si tip (Φ_M ≈ 4.1–4.2 eV) were used in this experiment. The pinning factor S = dΦ_B/dΦ_M was evaluated as the slope of the linear fit of these data. The obtained values for S are ~0.3 and ~0.1 for the MoS₂ regions without defects and for the regions with subsurface defects, respectively. The observed pinning factor of the defect-free MoS₂ areas was found to be consistent with the theoretically predicted values for pristine MoS₂ considering gap states induced exclusively by the metal/TMD interaction [60,61]. On the other hand, the measured pinning factor on the defect sites matches well with the experimentally obtained values on metal/MoS₂ contacts [58,60,76], confirming the dominant role of these defects on the overall contact behavior.

This kind of investigation was also extended to other TMDs, i.e., MoSe₂, MoTe₂, WS₂, and WSe₂ [75]. Figure 5c,d shows the SBH values for the different TMDs evaluated in the defect-free (c) and in the defect regions (d) as a function of the metal tip work function. A similar behavior in the case of MoS₂ is observed also for these TMDs, i.e., the presence of Fermi level pinning in the defect-free regions, ascribed to gap states induced by the metal tip, and an even stronger pinning (30%–40% increase) on defect regions due to defects induced gap states.

2.2. Nanoscale Mapping of MoS₂ Schottky Barrier Tuned by Oxygen Plasma Treatments

As previously discussed, the Fermi level pinning at the metal/TMDs interface has important implications in electronic device performances, since the resulting Schottky barrier ultimately rules the device on-resistance, as well as the type of carriers (n-type or p-type) that can be injected in the channel of TMD-based transistors. In the specific case of MoS₂, which is an unintentionally n-type doped semiconductor, the Fermi level pinning close to the conduction band results in a relatively small SBH for electron injection and in a high SBH for holes injection. Hence, MoS₂ transistors commonly exhibit an n-type behavior associated with the formation of an electron accumulation channel for positive gate bias above the threshold voltage, whereas the observation of a p-type behavior related to the hole inversion channel at a sufficiently high negative bias is typically precluded. Clearly, this represents an obstacle for the realization of a complementary MOS technology all based on MoS₂.

Recently, soft O₂ plasma functionalization revealed an effective method to finely tailor the SBH of multilayer MoS₂, and this approach was exploited to demonstrate MoS₂ field effect transistors with ambipolar (i.e., both n-and p-type) behavior [29]. In this context, nanoscale electrical analyses by C-AFM showed how the SBH of MoS₂ can be tuned by increasing the plasma exposure time [29]. Figure 6a,d reports the SBH map and the histogram of the Φ_B values measured on pristine MoS₂, showing a narrow SBH distribution. Figure 6b reports the SBH map after the 300 s soft plasma treatment, which results in a broader SBH distribution, with Φ_B ranging from 0.21 to 0.58 eV, as shown by the histogram of the Φ_B values (see Figure 6e). Finally, Figure 6c shows the SBH map after the 600 s soft plasma treatment. In this case, the SBH distribution extends from ~0.2 to ~0.9 eV (see Figure 6f). Hence, it includes both regions with a low barrier for electrons and regions with a low barrier for holes, being the Schottky barrier for holes Φ_B,h = E_g − Φ_B, with E_g as the bandgap of multilayer MoS₂.

Hence, starting from a narrow distribution of low SBHs for electrons in the case of pristine MoS₂, the SBH map was modified after a 600 s O₂ plasma treatment into a broad distribution formed by nanometric patches with low SBH for holes in a background with low SBH for electrons. These SBH inhomogeneities in the O₂ plasma treated samples were associated to lateral variations of the incorporated oxygen concentration in the MoS₂ surface region [29]. Back-gated FETs were fabricated with Ni source and drain contacts deposited on pristine MoS₂ (see schematic in Figure 7a) or on areas selectively exposed to O₂ plasma functionalization for 600 s (see schematic in Figure 7c). Figure 7b shows the transfer characteristics I_D-V_G for different drain bias values (V_DS = 1, 2, and 5 V) measured on a FET with channel length of L = 10 μm fabricated on a pristine MoS₂ flake. The n-type transistor behavior typically reported for MoS₂ FETs can be observed, with a monotonic increase of I_D over more than five decades in the considered gate bias range. Figure 7d shows the transfer characteristics I_D-V_G (for V_DS = 1, 2, and 5 V) of a MoS₂ transistor with the same channel length and thickness, but with the source and drain contacts deposited on plasma O₂ functionalized regions. For V_DS = 1 and 2 V, the transfer characteristics exhibit a pronounced p-type behavior, with the hole current branch (for V_G < 0) significantly higher than the electron current branch (for V_G < 0), whereas for V_DS = 5 V an ambipolar behavior is observed, with the I_on/I_off current ratio of ~10³ for both the electrons and holes branches.

The ambipolar behavior in O₂ functionalized FETs was explained by the coexistence of regions at low SBH for electrons and regions at low SBH for holes within the same source and drain contact areas, as demonstrated in the SBH map of Figure 6c. In particular, for a positively biased drain contact (V_DS > 0), the injection of electrons from the source to the accumulation channel (for V_G > 0) occurs through regions with lower SBH for electrons, whereas the injection of holes from the drain to the inversion channel (for V_G < 0) is allowed by the regions with low SBH for holes. The possibility to have both the n- and p-type current transport in a single device structure is an important step towards the implementation of a CMOS technology with MoS₂.

3. Local Resistance Mapping at Grain Boundaries in CVD Grown MoS₂

The C-AFM experiments reviewed in the previous section demonstrated the enhanced current injection due to localized defects on the surface or near-surface region of multilayer TMDs [44,45,75]. These defects are responsible for the Fermi level pinning effect commonly observed at metal Schottky junctions with these semiconducting materials.

In addition to these localized defects, the presence of extended defects, such as grain boundaries (GBs), must be taken into account in the case of CVD grown TMDs. As an example, the CVD deposition of MoS₂ on a commonly used SiO₂/Si substrate typically results in the formation of a 2D polycrystalline material, composed by several domains with a triangular shape separated by GBs [77]. The structural and electrical properties of GBs have been the object of many experimental and theoretical investigations in the last years [78,79], due to their impact on the mobility of MoS₂ field effect transistors and their role in peculiar extrinsic charge transport phenomena in MoS₂ [80,81]. Recent investigations indicated that inter-domains scattering (due to GBs) can play a major role (as compared to the intra-domain scattering due to localized defects [68,82,83]) in the degradation of the MoS₂ mobility, especially for certain misorientation angles between MoS₂ domains [81]. To this purpose, complex transistors structures and modeling for electrical data interpretation have been employed to evaluate the impact of the specific GB configurations on the MoS₂ channel mobility [81]. In this context, C-AFM has been recently employed as a powerful tool for direct probing of the electrical properties of MoS₂ domains and GBs. As an example, the C-AFM analyses of a few layers of MoS₂ grown by CVD on sapphire revealed a reduced current injection in the GBs, located in the topmost and in the buried MoS₂ layers [84].

Nanoscale resolution current mapping by C-AFM has been also employed to evaluate the conductance drop associated with GBs in monolayer MoS₂ grown by CVD onto a SiO₂/Si substrate [85]. Figure 8a shows an optical microscopy of the as-grown domains, which exhibit typical triangular shapes and sizes up to 50 μm. The coalescence of MoS₂ domains, resulting in the formation of GBs, can be deduced from this optical image. Figure 8b illustrates the employed experimental configuration for C-AFM measurements, where the current is measured between the nanoscale contact formed by a PtIr tip and a macroscopic electrode partially covering some MoS₂ domains. Figure 8c,d shows the surface morphology and the corresponding current map measured with an applied bias V_tip = 1 V (referred to the tip) in the region indicated by a red box in Figure 8b. This area includes two MoS₂ domains (indicated as D1 and D2) separated by a GB, with D1 directly connected to the macroscopic contact. The GB is clearly visible both in the morphology (as a sharp peak with ~1 nm height) and in the current map. The current contrast is almost uniform within the individual domains, whereas a significant conductance drop can be observed from the domain D1 to the domain D2 separated by the GB. This is further elucidated by the two scan lines in Figure 8e,f, extracted from the morphology and current map, respectively. A reduction of the current by about a factor of 7 can be observed moving from D1 to D2, which is clearly associated to the resistance contribution of the GB. Furthermore, a dip in the current can be observed at the GB position. Noteworthy, the width of this current dip is ~150 nm, much larger than the width of the peak in the height profile. This can be explained by a potential barrier associated to the GB, with the formation of depletion regions at the two sides of this barrier. This picture is consistent with the scenario proposed in Reference [81], where a potential barrier ranging from ~0.2 to ~0.6 eV was calculated, depending on the misorientation angle between adjacent MoS₂ domains.

Hence, C-AFM represents a powerful tool to directly evaluate the impact of GBs in polycrystalline TMDs grown by CVD, without the need for fabricating ad hoc device structures.

4. Local Transport Properties in TMD Lateral and Vertical Heterojunctions

4.1. Lateral Heterojunctions of TMDs

Due to their unique electronic properties, TMDs offer various solutions to realize lateral, i.e., in-plane, semiconductor heterojunctions. As an example, the significant bandgap variation from monolayer (1L) to bilayer (2L) of TMDs has been exploited as a simple way to create lateral heterostructures at the boundaries between 1L and 2L regions of the same material [86]. Another promising way to tune the TMD electronic structure (i.e., the band gap and the conduction/valence band energies) is the formation of alloys, obtained by mixing Mo with W and S with Se in a controlled way [87,88,89]. As an example, single layers of Mo_xW_1–xSe₂ alloys have shown a tunable band gap, with the conduction/valence band levels depending on the exact ratio of Mo and W [90]. Although Mo and W exhibit a similar electronic structure, preferential segregation of Mo-rich and W-rich regions has been commonly observed during the growth process of the TMD alloys [91], resulting in the formation of one-dimensional (1D) lateral heterojunctions [92].

Recently, Bampoulis et al. [69] used C-AFM to characterize the conductivity of a Mo_xW_1−xSe₂ alloy at the nanoscale. They observed the segregation of Mo-rich and W-rich domains and demonstrated that these different regions exhibit distinct SBHs values, reflecting the different band structures of WSe₂ and MoSe₂. An atomic resolution C-AFM map of a Mo_xW_1-xSe₂ alloy (with x = 0.3) is reported in Figure 9a, showing distinct differences in the conductivity between neighboring regions with a nanometer size (indicated with (i) and (ii)). To get further insight into this observation, a local I-V spectroscopy was carried out on these differently conductive regions. Figure 9b shows a series of I-V characteristics recorded along the line indicated in the inset which depicts the boundary between the Mo-rich (more conductive) and W-rich regions (less conductive). The I-V curves appear to gradually change when moving from one region to the other. Figure 9c shows the SBH values along this line, extracted by thermionic emission fitting of the individual I-V curves in the forward bias regime. The transition to the smaller SBH (Mo-rich) regions happens gradually and on the length scale of about 3–4 nm. This gradual transition is due to band bending across the phase boundary. The Mo-rich and W-rich regions were expected to approach the electronic band diagram of the MoSe₂ and WSe₂ crystals, as schematically illustrated in Figure 9d. Based on this diagram, a type-II staggered gap heterojunction is expected when moving from a Mo-rich region to a W-rich region, and band bending occurs at the interface of the two regions. Finally, assuming that the Fermi level pins at the same energy level across the boundary, a conduction band offset (CBO) of ~0.13 eV was estimated from the SBH difference between the two regions.

4.2. Vertical Heterostructures of TMDs

In addition to the 1D lateral heterojunctions discussed in the previous paragraph, 2D semiconducting heterostructures obtained by vertical stacking of different TMDs are currently the object of increasing scientific interest, due to the wide range of potential applications in electronics and optoelectronics [10]. In particular, 2D heterostructures of TMDs with a type II band alignment, such as MoS₂ and WSe₂, offer the possibility to implement band-to-band tunneling diodes and transistors for ultra-low-power consumption logic applications.

Recently, Lin et al. [12] reported the growth of MoS₂/WSe₂ heterostructures onto epitaxial graphene (EG) on SiC by sequentially depositing monolayers of the two TMDs employing oxide powder vaporization or metal-organic CVD. In particular, WSe₂ was first grown on EG at 950 °C. Following this first growth step, the surface coverage of the WSe₂ on EG was typically >60%, with a lateral size of 2 µm for WSe₂ domains, as illustrated by the AFM image in Figure 10a. Subsequently, the growth of MoS₂ on WSe₂/EG was performed at 750 °C. The MoS₂ domains are smaller (~300 nm) and typically grow on WSe₂ starting from the domain’s edges, as illustrated by the AFM image in Figure 10b. Given the small size of the MoS₂/WSe₂ areas, C-AFM was employed to investigate a vertical current transport across the heterostructure [12]. In particular, local I-V measurements through the MoS₂/WSe₂/EG and WSe₂/EG heterostructures were carried out at room temperature according to the schematic in Figure 10c. Representative I-V curves collected on the two systems are reported in Figure 10d. The I-V characteristics of the WSe₂/EG heterojunction exhibit a slight rectifying behavior, associated to the p-n junction between p-type WSe₂ and n-type EG on SiC. On the other hand, the curves measured on the MoS₂/WSe₂/EG show a current peak (at V_peak ≈ 1.1 V) followed by a valley (with a peak-to-valley current ratio of 1.9) and, finally, an exponential current increase. The negative-differential-resistance (i.e., the Esaki diode behavior) is a clear evidence of band-to-band-tunneling occurring at room temperature, indicating the formation of an ultra-sharp interface at the MoS₂/WSe₂ n-p junction.

4.3. Vertical Heterostructures of TMDs with Bulk Semiconductors

As shown in the previous paragraph, the Van der Waals epitaxy of heterostructures entirely made of 2D materials is still in its infancy, and further developments of this approach will be required to realize these systems on a large area. On the other hand, the integration of 2D materials with conventional bulk semiconductors can represent an easier root toward the exploitation of these low dimensional materials in (opto)electronics. In fact, this approach allows adding new functionalities to the existing semiconductor devices, and opening the way to the demonstration of new device concepts [11,32,93]. In the last years, the integration of graphene and semiconducting TMDs with silicon and other semiconductors, such as GaN and related materials [94], has been explored by several research groups. Different approaches have been explored, from the transfer of 2D materials [95,96,97,98] to the direct growth on the semiconductor substrate [19,99]. In particular, the epitaxial growth of MoS₂ on the basal plane of GaN is especially favored by the low in-plane lattice mismatch (<1%) between the two hexagonal crystals. Furthermore, the small difference in the thermal expansion coefficients between the two materials is expected to result in a reduced strain during cooling down from the growth temperature of MoS₂ to room temperature [100,101].

Recently, Ruzmetov et al. [19] reported the CVD growth of epitaxially oriented MoS₂ islands on the GaN basal plane. Figure 11a reports a scanning electron microscopy (SEM) image of the as-grown MoS₂ on the GaN surface, consisting of triangular domains of monolayer MoS₂ with a typical size of ~1 μm. The sides of these triangles were perfectly aligned with the m-plane (1–100) of the wurzite GaN substrate, indicating the in-plane epitaxial alignment of the GaN and MoS₂ lattices. Thanks to this rotational order, no evidence of grain boundaries was observed in a larger size of monolayer MoS₂ islands formed by the coalescence of these small domains. This is a major advantage of MoS₂ grown on GaN with respect to the more commonly used CVD MoS₂ on amorphous SiO₂, which is a polycrystalline material with a large density of grain boundaries. In fact, as discussed in Section 3, grain boundaries are one of the main sources of electron mobility degradation in MoS₂ [81,85].

C-AFM has been applied in combination with KPFM to investigate the electrical properties of the heterojunction between CVD MoS₂ and n-GaN, specifically the vertical current flow across the heterointerface and the surface potential [19]. Figure 11b schematically illustrates the C-AFM setup for local current measurements, and Figure 11c reports a current vs. tip bias characteristic on an individual monolayer (1L) MoS₂ domain, showing a rectifying behavior of the tip/MoS₂/GaN junction. Figure 11d shows a KPFM surface potential map on a GaN region partially covered by the 1L MoS₂ domain. A line-scan of the surface potential along the red dashed line in the map is reported in Figure 11e, from which a surface potential difference of ≈360 meV between MoS₂ and n-GaN can be evaluated [19]. Finally, Figure 11f shows an illustrative energy band diagram showing the type I energy band alignment between 1L MoS₂ and GaN, as deduced from the surface potential map. This peculiar energy band alignment of the MoS₂/GaN heterojunction has been recently exploited in various device demonstrators, including Esaki diodes [102] and high responsivity deep-UV photodetectors [103].

5. Conclusions and Perspectives

The electronic transport properties of semiconducting TMD layers are strongly dependent on point and extended defects present in their crystalline structure. In this context, the C-AFM technique proved an essential tool to investigate the current transport in these materials at nanoscale. In this paper, recent C-AFM studies on TMDs have been reviewed, discussing the implications of local transport phenomena in the overall behavior of TMD-based devices. The current mapping and I-V spectroscopy by C-AFM allowed clarifying the mechanisms responsible for the Fermi level pinning commonly observed at the metal/TMDs interface. In particular, subsurface defects in the transition metal layer (with a density of 10¹⁰–10¹¹ cm⁻²) have been identified as the main origin of the pinning, although the metal/TMD interaction was also shown to partially contribute to this effect. The impact of extended defects, such as grain boundaries, on the conductivity of monolayer MoS₂ grown by CVD was also directly imaged by C-AFM. Methods for nanoscale tailoring of the Schottky barrier in MoS₂, such as oxygen plasma functionalization, have been illustrated, and the changes in the barrier height spatial distribution with the plasma exposure time was imaged at the nanoscale. The application of this approach for the realization of ambipolar MoS₂ transistors has been also discussed. Finally, C-AFM studies of the current transport in lateral and vertical heterostructures based on TMDs have been reported. As an example, a high-resolution current mapping of Mo_xW_1−xSe₂ alloys showed the formation of 1D lateral heterojunctions between the nanoscale Mo- and W-rich regions formed due to phase segregation. Finally, the C-AFM-based current-voltage spectroscopy provided insight on the current injection mechanisms in vertical heterojunctions formed by van der Waals epitaxy of different TMDs (such as the MoS₂/WSe₂ junction) and by the integration of TMDs with bulk semiconductors (such as the MoS₂/GaN junction). Potential applications of these heterojunctions for novel electronic and optoelectronic devices have been also discussed.

Further progress in the development of 2D/2D or 2D/3D van der Waals heterostructures is expected in the forthcoming years. As a matter of fact, realizing the full potential of these material systems will require understanding and controlling disorder, which can obscure intrinsic properties and hinder device performances [104]. Furthermore, atomic defects or disorder can also be harnessed to provide useful electronic, optical, chemical, and magnetic functions. In this context, the C-AFM technique, in combination with other electrical and optical scanning probe methods, will represent a valuable tool to elucidate the correlation between nanoscale and macroscopic properties of the heterostructures, and will provide a guidance to tailor materials properties.