Growth of MoS₂–Nb-doped MoS₂ lateral homojunctions: A monolayer p–n diode by substitutional doping

The recent development of two-dimensional (2D) layered transition metal dichalcogenides (TMDs) has opened a new research field for nanotechnology.^1,2 In particular, 1H-phase monolayer MoS₂, MoSe₂, WS₂, and WSe₂ are direct-gap 2D semiconductors with a bandgap of 1.5–2.0 eV, which is an important feature for developing various optoelectronic devices, such as light-emitting devices and photodetectors.^3–6 Furthermore, even in monolayer limits, TMDs demonstrate excellent carrier mobility and a high on/off ratio when used in field-effect transistors (FETs), making them promising next-generation semiconductor materials.^7,8

In general, as-grown TMDs tend to be p- or n-type semiconductors even without intentional doping because of the defects formed and their energy levels.^9–11 For example, MoS₂ and WS₂ show strong n-type behaviors because of the heavy electron doping caused by surface sulfur vacancies.^12–15 Furthermore, at the metal-electrode–sulfur-based-TMD interface, there is strong Fermi-level pinning close to the conduction band minimum with a pinning factor of 0.11–0.15, which is attributed to the sulfur vacancies and the metal–chalcogen interactions, resulting in the difficulty of hole injection.^16–18 These properties of sulfur-based TMDs make their p-type operation challenging, thus hindering the formation of lateral homojunctions such as monolayer TMD-based p–n diodes. Therefore, the development of effective p-type doping techniques is important for exploring their potential for device applications.

To dope carriers in TMDs, the substitutional doping method, which is based on carrier doping by substituting metal or chalcogen sites with impurity atoms with different electronic numbers, has been investigated.¹⁹ For example, group-V transition metals (V, Nb) and group-XV elements (N, P) would work as electron acceptors, whereas group-VII transition metals (Re) would be electron donors when substituted in metal or chalcogen sites.^20–24 In particular, Nb atoms are neither chemisorbed nor physisorbed on MoS₂ surfaces (i.e., intercalated in bulk MoS₂); however, they preferentially replace the Mo sites in MoS₂ and act as single acceptors.²⁵ Thus, Nb doping is promising for fabricating various devices based on monolayer TMD.

One method for obtaining substitutionally doped monolayer TMDs is chemical vapor deposition (CVD). The advantages of this method are the availability of large-area monolayer TMDs and its simplicity (only by adding compounds containing dopant elements). Thus, various CVD experiments using various dopant sources have been performed.^24,26–36 For Nb doping, by adding Nb compounds, such as Nb₂O₅, Nb(HC₂O₄)₅ · xH₂O, and NbCl₅, into a chamber, large-scale monolayer Nb-doped MoS₂ and WS₂ have been successfully obtained, thus showing p- or ambipolar behaviors.^31–34 In these studies, the crystal was uniformly doped by Nb atoms; however, the formation of p–n junctions has not been achieved to date. The formation of TMDs with uniform Nb doping is attributed to the stable supply rate ratio between Mo (or W) and Nb during growth. For example, Nb sources coated on the substrates or metal sources with high sublimation properties were used.^33,34 Nevertheless, there are certain studies on the CVD growth of TMDs with localized Nb-doped regions where an Nb source with a lower supply rate than the W source or a two-step growth method (doped and undoped regions were successively grown) was used.^34,35 However, the properties of interfaces showing true 2D p–n junctions, such as rectifying behaviors, have not been investigated yet.

In this study, we demonstrate the one-step growth of monolayer undoped MoS₂–Nb-doped MoS₂ lateral p–n homojunction (herein referred to as MoS₂–MoS₂:Nb). To fabricate MoS₂–MoS₂:Nb, we used metal precursors with a large vapor pressure difference. At the earlier stage of growth, a considerably faster Mo source supply rate compared with that of Nb resulted in the formation of almost undoped MoS₂. Then, the Mo source supply rate slowed down because of the reduction of the Mo source amount, and the Nb concentration became relatively high, resulting in the formation of MoS₂:Nb in the outer region to surround the MoS₂ grains. Furthermore, our results suggested that Nb atoms would segregate into the outer region to form a sharp MoS₂–MoS₂:Nb junction. This Nb segregation would be driven by the lower formation energies of MoS₂:Nb with higher Nb concentration. Nb doping in the MoS₂:Nb region was confirmed via elemental analysis using energy-dispersive x-ray spectroscopy (EDX) and by measuring the binding energy and work function modulation using synchrotron soft x-ray scanning photoelectron microscopy system, which can obtain three-dimensional spatially resolved XPS (x-ray photoelectron spectroscopy, i.e., electron spectroscopy for chemical analysis (ESCA)) datasets, called “3D nano-ESCA” and Kelvin probe force microscopy (KPFM), respectively.^37–39 As a large amount of spectral data were obtained in the 3D nano-ESCA measurements, peak fitting was conducted using the spectrum-adapted expectation–conditional maximization (ECM) algorithm developed to improve the efficiency of peak fitting.^40,41 The MoS₂–MoS₂:Nb crystal demonstrated n-type and degenerate p-type conductions, respectively, demonstrating the successful spatially selective hole doping by Nb into MoS₂. Furthermore, the p–n junction obtained demonstrated a current rectification behavior. These results show that the obtained junction is a true 2D diode and that the proposed CVD method can be used for fabricating various TMD-based devices with controlled carrier concentrations.

We grew MoS₂–MoS₂:Nb using the CVD method. As precursors and a growth promoter, MoO₃ (Sigma-Aldrich, 99.97%), metallic Nb (FUJIFILM Wako Chemical, 99.5%), elemental S (High-Purity Chemicals, 99.99%), and NaCl (FUJIFILM Wako Chemical, 99.5%) were used. The mixture of MoO₃ (6 mg)/Nb (160 mg) powder, NaCl (13 mg) powder, and a surface-oxidized Si (SiO₂/Si) substrate with an oxide thickness of 90–300 nm were placed in a quartz tube having an inner diameter of 26 mm. Sulfur powders were placed at 280 mm upstream of the MoO₃/Nb powder. The quartz tube was heated using an electric heating jacket and an electric furnace at 160 and 780 °C, respectively, for 15 min under an Ar flow of 200 SCCM: Note that S and other elements were placed at the coolest and hottest zones, respectively. After growth, the furnaces were rapidly cooled to avoid excess NaCl adhering to the crystal.

We obtained optical images using a standard optical microscope (Nikon Eclipse ME600). Raman and photoluminescence (PL) measurements were performed using a confocal Raman microscope (Renishaw inVia Qontor) with a continuous-wave laser excitation of 2.33 eV. Objective lenses of 100× and 0.85–0.90 NA were used for all measurements. Atomic force microscopy (AFM) and KPFM measurements were performed using a standard AFM (Park Systems NX10). All measurements were performed at room temperature and ambient pressure.

TEM observation, in addition to selective-area electron diffraction (SAED) and EDX measurements, was performed using a typical TEM (Hitachi H-9500) with an acceleration voltage of 200 kV. The typical polymethyl methacrylate-based wet transfer method was used to transfer the MoS₂ crystal on a copper TEM grid. Spatially resolved Mo 3d, S 2s, and Nb 3d core-level photoelectron spectra were obtained using a 3D nano-ESCA system composed of a Fresnel zone plate for x-ray focusing and an angle-resolved hemispherical electron analyzer (Scienta-Omicron R3000 EWAL) installed at the soft x-ray beamline, BL07LSU, in SPring-8. The photon energy used in the measurements was 1000 eV, and the lateral spatial resolution was ∼100 nm. The Au 4f_7/2 peak from the electrodes at 84.0 eV was used for the binding energy scale calibration. Before the measurement, Cr (5 nm) and Au (50 nm) electrodes were fabricated onto the sample by standard electron beam lithography to reduce the charge during measurements.

Peak fitting was performed using the spectrum-adapted ECM algorithm.^40,41 The spectrum-adapted ECM algorithm is a method for efficient and stable peak fitting based on the expectation–maximization (EM) algorithm.^42,43 Iterative calculations between the E-step and CM-step were performed using the intensity of the spectrum as the weight of the corresponding energy step.

We used the PVMMB fitting model composed of the pseudo-Voigt distribution (P_PV) and components of linear background (P_uni and P_tri) as follows:

where

and

where $Lorentzxn|μ,2ln2σ$ and $Gaussxn|μ,σ$ are the Lorentz and Gaussian distributions, respectively. K is the number of decomposed peaks. μ, σ, η, and λ are the parameters of the fitting model: μ is the location parameter specifying the peak of the spectrum, σ is the standard deviation (σ > 0), η is the mixing parameter of Lorentz and Gauss (0 ≤ η ≤ 1), and λ is the mixing coefficient of the kth component of the peaks (0 ≤ λ ≤ 1 and $∑k=1K+2λk=1$⁠). These parameters were estimated using the spectrum-adapted ECM algorithm for each spectral data.

The algorithm requires to set the value of K and initial values of μ, σ, η, and λ. Herein, we set K = 6 and 4 in the analysis of the peak position mapping of Mo⁴⁺ 3d_5/2 and Nb⁴⁺ 3d_5/2 peak intensity mapping, respectively. For Mo⁴⁺ 3d_5/2 peak position mapping, the initial values of parameters were set as $μ$ = {768, 767, 766, 765, 764, 763}, $σ$ = {5, 5, 5, 5, 5, 5}, $η$ = {0.5, 0.5, 0.5, 0.5, 0.5, 0.5}, and $λ$ = {1/18, 1/18, 1/18, 1/18, 1/18, 1/18, 1/3, 1/3}, respectively. For the peak intensity mapping of Nb⁴⁺ 3d_5/2, the initial values of the parameters were $μ$ = {784, 787, 789, 792}, $σ$ = {5, 5, 5, 5}, $η$ = {0.5, 0.5, 0.5, 0.5}, and $λ$ = {1/12, 1/12, 1/12, 1/12, 1/3, 1/3}, respectively.

We evaluated the structural and electronic properties of Nb-doped MoS₂ through density functional theory (DFT) simulations,^44,45 as implemented in the Quantum Espresso software package.^46,47 Doping was performed by building 2 × 2, 3 × 3, and 4 × 4 supercells of the MoS₂ monolayer and replacing certain Mo atoms with Nb atoms. For each atomic configuration, the in-plane cell parameters and atomic positions were relaxed until all force components were <10⁻⁸ a.u. and the pressure was <10⁻⁵ kbar. We used the Perdew–Burke–Ernzerhof (PBE) exchange–correlation functional and optimized norm-conserving Vanderbilt (ONCV) pseudopotentials.^48,49 The cutoff for kinetic energy was set to 60 Ry, and the k-point sampling was set to 18 × 18 × 1 for the primitive cell and 9 × 9 × 1 for the 2 × 2 supercell with a k-point density of 0.1 Å⁻¹. To avoid the interactions of atoms with their periodic images, a vacuum region of 15 Å was added in the direction perpendicular to the MoS₂ monolayers.

A scanning electron microscope (SEM)-integrated semiconductor parameter analyzer (Hitachi NanoProber: Hitachi N-6000SS with Agilent Technologies B-1500A) was employed to measure the semiconductor properties. SEM observation was performed using an acceleration voltage of 0.5 kV. Tungsten probes were directly placed on the sample to make contact with it. The measurement was performed at room temperature and high vacuum (∼4 × 10⁻⁵ Pa). During the measurement, the outer wall of the chamber was cooled by liquid nitrogen to keep the sample away from carbon contamination.

Our strategy for obtaining MoS₂–MoS₂:Nb by CVD growth is using precursors with large vapor pressure differences, as discussed later [Fig. 1(a)]. We used MoO₃ and metallic Nb as Mo and Nb sources, respectively. Furthermore, we used the “alkali-metal assistance” method for the growth, which is known to increase the supply rate of metal precursors and promote the lateral growth of TMDs.^50–52 The details of the growth procedure are presented in the section titled Methods, and a schematic of the CVD growth setup is shown in Fig. 1(b). Figure 1(c) shows a typical optical microscope image of the MoS₂–MoS₂:Nb grown on the SiO₂/Si substrate. There are certain triangular crystals with a size up to a few micrometers with different optical contrasts from the inner and outer regions. This optical contrast was not attributed to the difference in the layer number because the AFM image and the corresponding height profile depicted in Figs. 1(d) and 1(e) show that the thickness of the triangular crystal is ∼1 nm (monolayer),⁵³ irrespective of the color of each region. The AFM image shows that wrinkles existed in the outer regions of the crystal (the magnified AFM image can be found in Fig. S1). For the monolayer TMD lateral heterostructure with a significant lattice mismatch such as WS₂–WSe₂, wrinkle formation for relaxing the accumulated compressive strain in the layer with a larger lattice constant (WSe₂) was observed.⁵⁴ Because the lattice constant of MoS₂:Nb monotonically increased with the increase in the Nb concentration (Fig. S2), the wrinkle formation in our sample is attributable to the lattice-mismatch-induced compressive strain in MoS₂:Nb.

The growth of MoS₂–MoS₂:Nb occurred as follows: In the initial stage, the supply rate of the Mo source was sufficiently fast compared with that of Nb because the vapor pressure of MoO₃ was much higher than that of Nb, resulting in the formation of n-type MoS₂. As the growth proceeded, the Mo supply rate slowed down because of the reduced amount of the MoO₃ source by sublimation, where the growth temperature was higher than the MoO₃ sublimation temperature (∼700 °C). Consequently, the Nb concentration became relatively high, and p-type MoS₂:Nb growth occurred in the outer region of the crystal [Fig. 1(a)]. In this scenario, Nb concentration gradually increases from the center to the edge of the crystal. However, as discussed later, we reported that Nb exists only in the outer regions, thus forming a sharp MoS₂–MoS₂:Nb interface. This result shows that Nb atoms would segregate and be condensed into the outer region of the crystal during the growth, which could be driven by the decreased formation energy of MoS₂:Nb with increasing Nb concentration to up to 25%.²⁵ However, the lattice constant of MoS₂:Nb monotonically increased with the increase in Nb concentration, and the lattice constant difference between MoS₂ and MoS₂:Nb led to lattice strain at their interface. This strain would be relaxed by the formation of wrinkles at the MoS₂:Nb region, as shown in Fig. 1(d). Because of the high flexibility of MoS₂, the strain could be released by forming wrinkles. These mechanisms of Nb segregation and strain relaxation would result in a quite sharp MoS₂–MoS₂:Nb interface with wrinkles in MoS₂:Nb, as shown in the 3D nano-ESCA, AFM, and KPFM images [Figs. 1(d), S3, and S4, respectively].

To investigate the basic optical properties of the crystal, we measured the Raman and PL spectra. The measurements were performed at room temperature with a 2.33 eV continuous-wave laser source. Figure 2(a) shows the typical Raman spectra of the obtained MoS₂–MoS₂:Nb, which were measured in the inner and outer regions. In the inner region, two intense peaks at 385.3 and 401.9 cm⁻¹ were observed, which originated from the E′ and A′₁ modes of MoS₂, respectively.⁵⁵ The peak separation of these two peaks was 17 cm⁻¹, which agreed well with that of the monolayer MoS₂.⁵⁶ However, in the outer region, the observed Raman spectrum was similar to that observed in the heavily Nb-doped MoS₂ with a Nb concentration of >10%:⁵⁷ In addition to the E′ and A′₁ peaks, the broad peaks at ∼100 to 240, 343, and 370 cm⁻¹ were observed. This result shows that heavy Nb doping was achieved in the outer region of the MoS₂ crystal. A similar tendency of the change in the Raman spectrum was reported in heavily V- and Sn-doped MoS₂,^24,30 suggesting that these broad peaks originated from the disorder and decreased symmetry of the MoS₂ lattice because of substitutional doping. In all cases, the dopant concentrations in MoS₂ were greater than 1% (∼10% in our case), which is much higher than that in the case of conventional substitutional doping for Si (<0.1%). Thus, lattice disorder and the decrease in the resultant symmetry of the host MoS₂ are not negligible. The Raman peak positions of E′ and A′₁ modes in MoS₂:Nb were 382.4 and 402.2 cm⁻¹, respectively; they were red- and blue-shifted compared with those of the inner region. This peak shift can be attributed to three factors: high hole doping by Nb substitution, alloying with NbS₂, and compressive strain. A detailed discussion is presented in the supplementary material. Figures 2(b)–2(e) show an optical image of MoS₂–MoS₂:Nb, the corresponding Raman spectra mapping images of the E′ peak frequency, the full-width half-maximum (FWHM) of the E′ peak, and the integrated peak intensity of the Raman spectra from 80 to 240 cm⁻¹ of MoS₂–MoS₂:Nb, respectively. These Raman mapping images show a clear contrast between the inner and outer regions, corresponding to the spatially localized Nb doping. However, by only focusing on the inner (outer) region of the crystal, the mapping images were uniform, indicating that the crystal was homogeneous in each region and that there was no local mixing of the p- and n-regions. Thus, a p–n junction was expected to exist at the interface with a sharp optical contrast.

The PL spectra of the crystal are shown in Fig. 3. In an inner region (undoped MoS₂), the obtained crystal demonstrated a typical PL spectrum of the MoS₂ monolayer: A- and B-exciton PL peaks could be observed at 1.80 and 1.91 eV, respectively.⁵⁸ However, the PL peak of the outer region (MoS₂:Nb) was significantly smaller and broader than that of the undoped MoS₂, and additional peaks appeared at ∼1.6 eV. These results indicate that mid-gap states and/or excess carriers existed in the outer region, which significantly decreased the A-exciton PL intensity and was the origin of the low-energy PL peak at ∼1.6 eV. When Nb was doped into MoS₂, they formed gap states near the valence band maximum (VBM) and acted as acceptors. Because some of the excitons were bound to these neutral acceptor states, PL peak broadening and redshifting occurred.⁵⁹ Moreover, the increased hole concentration decreased PL from the excitons because of the increased trion-related PL. Similar results could be observed for the exfoliated monolayer MoS₂:Nb (Nb concentration of ∼0.5%), where additional PL peaks in the lower energy side of the A-exciton appeared (Fig. S6). From the above-mentioned Raman and PL measurements, we concluded that Nb was selectively doped in the outer region of the MoS₂ crystal, resulting in the self-organized formation of p–n lateral junctions, as discussed later. Note that this weak and broad PL emission from MoS₂:Nb was also attributed to the bandgap fluctuation caused by the wrinkles in the MoS₂:Nb. The wrinkles changed the angle of the Mo–S chemical bonds, resulting in the modulation of the Mo ligand field.^60,61 This modified the local band structure, leading to the spatial distribution of the bandgap and its direct or indirect nature.⁶¹ Thus, PL peaks in wrinkled MoS₂:Nb should be broader and weaker. The wrinkled region contains compressive strain, which could cause PL blueshift. However, the calculated value of PL blueshift from the parameter in the previous report (45 meV/% strain)⁶² and our Raman result (∼0.3% compressive strain, shown in the supplementary material) was ∼15 meV, which is smaller than the bandgap narrowing caused by Nb doping (∼170 meV, from the DFT calculation shown in Fig. S7). Thus, the effect of compressive strain could not be observed in our wrinkled MoS₂:Nb.

To identify the structure and composition of the crystal, we employed TEM and EDX measurements. Figures S8(a) and S8(b) show typical bright-field TEM images of the MoS₂–MoS₂:Nb and EDX spectra, respectively. Only the spectrum measured at the outer region demonstrated a clear Nb Kα peak at 16.6 keV, indicating the successful selective doping of Nb at the outer region of the crystal. Furthermore, we characterized the SAED pattern to investigate the crystal structure of the lateral homojunction. The SAED patterns obtained from the undoped MoS₂ and MoS₂:Nb regions were demonstrated, as shown in Figs. S8(c) and S8(d), respectively. These two images show a similar set of hexagonal diffraction spots with the same angle, indicating the same crystallographic orientations. Moreover, these results confirm the formation of a MoS₂–MoS₂:Nb lateral homojunction. The SAED line profiles [Fig. S8(e)] along the lines shown in Figs. S8(c) and S8(d) describe that diffraction spots of MoS₂:Nb appeared at a slightly inner position compared with the undoped one. This result shows that MoS₂:Nb has a larger lattice constant than that of the undoped MoS₂ (100.6% of undoped MoS₂), which is consistent with the AFM images [Figs. 1(d) and S1] where wrinkles were observed because of the lattice expansion of MoS₂ by Nb doping.

To analyze the spatial elemental composition of MoS₂–MoS₂:Nb, we performed 3D nano-ESCA measurements. Figures 4(a) and 4(b) show an optical image and the ESCA spectra of the Mo 3d and S 2s core levels measured in the inner and outer regions of MoS₂–MoS₂:Nb, respectively. The Mo 3d and S 2s core levels of the MoS₂:Nb region had smaller binding energies than those of the pure MoS₂ region (Δ ∼ 0.3 eV). This difference was attributed to the Fermi-level (E_F) downshift by hole doping. Furthermore, we reported that the Nb⁴⁺ 3d peaks appeared only in the Nb-doped crystal (Fig. S6). The appearance of the Nb⁴⁺ peak suggests that the Nb atoms were substituted to the Mo sites of the MoS₂ crystal rather than the physisorbed sites on the surface of MoS₂. The Nb⁴⁺ concentration estimated from the ESCA spectra [Figs. 4(b) and S3(a)] was ∼13%, which is consistent with the Raman analysis. An optical image, a photoelectron peak position mapping of Mo⁴⁺ 3d_5/2, and an Nb⁴⁺ 3d_5/2 peak intensity mapping of MoS₂–MoS₂:Nb are shown in Figs. 5(b)–5(d), respectively. At the edge of the crystal, a low-binding energy shift in the Mo⁴⁺ 3d_5/2 peak and an Nb⁴⁺ 3d_5/2 peak were simultaneously observed. Thus, we concluded that the local hole doping into the single-crystalline MoS₂ monolayer by substitutional Nb doping was successful. The line profile of the integrated intensity at ∼203.0 to 205.0 eV (Nb⁴⁺ 3d_5/2) measured by 3D nano-ESCA is shown in Fig. S3(c). Signals from Nb⁴⁺ were observed only in the edge regions of the crystal, and the transition from undoped to Nb-doped regions was abrupt, forming a sharp interface. This shows the Nb segregation from the inner to the outer regions of the crystal during growth. The KPFM measurements supported the work function increase in the outer region of the crystal (Fig. S4).

The above-mentioned experimental results were confirmed by DFT calculations shown in Fig. S7. We calculated the band structure and density of states (DOS) of MoS₂:Nb with Nb concentrations of 0% and 12.5%. Nb concentration is defined as the fraction of Mo atoms substituted by Nb. The substitution of Nb atoms to Mo sites induced a downward E_F shift to under the valence band maximum (VBM) because Nb has one valence electron less than Mo, resulting in the p-type nature in MoS₂:Nb (i.e., the main contribution of Nb in DOS is in the valence band, as shown in Fig. S7). This E_F shift to the VBM made the binding energy smaller because of the magnified work function, as shown in the 3D nano-ESCA and KPFM results (Figs. 4 and S3). We examined the effect of Nb concentration on the lattice constant of MoS₂:Nb for an Nb concentration up to 50%. As the Nb concentration increased, the lattice constant of MoS₂:Nb monotonically increased (Fig. S2). The theoretical results are in good agreement with the 3D-nano ESCA and SAED measurements (∼13% Nb doping caused 0.6% lattice expansion on MoS₂:Nb).

To examine the electrical properties of obtained MoS₂–MoS₂:Nb crystals, we used an SEM-equipped prober station. The system has six tungsten probes in the SEM chamber, and we could manipulate the probes to directly measure the properties of the target positions in the crystal [Fig. 5(a)]. During the measurements, gate voltage was applied via the bottom SiO₂ dielectric. Figures 5(b) and 5(c) demonstrate the I_ds–V_g characteristics measured in the undoped MoS₂ and MoS₂:Nb regions, respectively. The undoped MoS₂ region demonstrated a typical n-type FET behavior with an approximate carrier mobility of ∼0.4 to 1 cm²/V s and an on/off ratio greater than 10⁵, which is consistent with previous reports, where MoS₂ crystals demonstrated n-type characteristics because of sulfur vacancies.^7,17 Nevertheless, the MoS₂:Nb region demonstrated a degenerate p-type behavior where the current was slightly reduced by the positive gate voltage and it was much larger than that of the inner region. The carrier mobility of MoS₂:Nb was ∼0.6 cm²/V s. Note that it was difficult to precisely calculate mobility in our measurements using direct probes as electrodes because of the difficulty in determining the width and length of the device. Furthermore, the I_ds–V_ds characteristics of MoS₂ and MoS₂:Nb demonstrated different characteristics. The MoS₂ region demonstrated a nonlinear I_ds–V_ds dependence, suggesting a Schottky barrier formation at the probe–MoS₂ interface [Fig. 5(d)]. However, in the MoS₂:Nb region [Fig. 5(e)], the characteristics demonstrated a nearly linear dependence, indicating the low-contact resistance at the probe/MoS₂:Nb interface regardless of the small contact area and Schottky barrier formed by the difference between ionization energy of MoS₂ and the work function of tungsten (5.77 and 4.18–5.25 eV, respectively^63,64). This linear dependence of the I_ds–V_ds characteristics of MoS₂:Nb results from the formation of a narrow depletion layer because of degenerate hole doping, which could allow a large tunneling current at the interface. These results confirmed the p-type dopant nature of Nb in the outer region of the MoS₂ crystal, which agrees with the following results. Therefore, we concluded that the crystal grown here had both n-type and degenerate p-type regions within the same single crystal.

Next, we measured the electric properties across the MoS₂–MoS₂:Nb interface. Figure 6(a) shows typical I_ds–V_ds characteristics across the interface. A clear rectifying behavior was observed, indicating that the obtained MoS₂–MoS₂:Nb is a true 2D diode. An ideality factor of the diode was 6.5 at V_g = 30 V. The ideality factor is limited by the huge contact resistance because of the direct probe measurements [Figs. 5(d) and 5(e)]. Here, the rectification degree was modulated by the gate voltage [Fig. 6(b)]. The current rectification ratio monotonically increased with an increase in gate voltage (Fig. S9), and the ratio increased from 53 (at V_g = −30 V) to 521 (V_g = 30 V). Because of the heavy Nb doping of MoS₂:Nb, the resistivity (or carrier density) of the heavily doped p-type region was not significantly increased (decreased) by the positive V_g as is the case with the lightly doped n-type region with a negative V_g [Figs. 5(a) and 5(c)]. Then, the rectification ratio increased with an increase in the positive V_g because of the decreased resistivity of the n-type region. The fabricated MoS₂–MoS₂:Nb p–n junction exhibits good long-term stability because the properties shown in Fig. 6 were measured four months after growth.

The estimated Nb concentration in our MoS₂:Nb crystal was ∼13%, much higher than the usual acceptor concentration in standard 3D semiconductors, such as p⁺-Si [e.g., the acceptor concentration (N_A) of <10²⁰ cm⁻³ is equivalent to 0.2%]. It is considered that this extremely high acceptor concentration is required for obtaining p-type monolayer MoS₂. This is because MoS₂ with sulfur vacancies on its surface shows strong n-type nature.¹⁴ Consequently, MoS₂:Nb demonstrated an n-type or ambipolar behavior with an Nb concentration of 0.5% when its thickness was <4 nm.⁶⁵ Thus, to achieve monolayer p-type MoS₂, N_A should be higher than the number of accumulated electrons on the MoS₂ surface (N_D^S). Previous studies revealed that N_D^S is 8.5 × 10¹² cm⁻² in MoS₂,⁶⁵ indicating that a higher N_A than 8.5 × 10¹² cm⁻²/(0.65 nm) = 1.3 × 10²⁰ cm⁻³ (∼0.7% of Nb, when the activation ratio of Nb is 1) is required for achieving p-type monolayer MoS₂. Furthermore, at the 2D limit, the quantum confinement effect and reduced environmental screening made the acceptor state deeper than that of the 3D case.^66,67 Thus, the 2D nature of the monolayer MoS₂ reduced the activation ratio of Nb; additional doping was required for achieving monolayer p-type MoS₂. However, in general, heavy doping degrades the carrier mobility of semiconductors, resulting in poor device performance. Thus, the passivating S vacancy of the MoS₂ surface (decreasing electron doping by the S vacancy) and increasing dielectric constant of the environment (increasing carrier activation ratio by shallowing the acceptor state) would be important for reducing N_A required for forming p-type MoS₂.⁶⁷ By reducing the Nb concentration, wrinkles in the MoS₂:Nb region reduce because the lattice constant of MoS₂:Nb approaches that of pristine MoS₂ as the amount of Nb reduces. Because wrinkles degrade crystal quality, reducing the Nb concentration would greatly improve the electrical properties of MoS₂:Nb.

In conclusion, we succeeded in synthesizing a monolayer MoS₂–MoS₂:Nb homojunction using the CVD method. The TEM and 3D nano-ESCA results confirmed that Nb was selectively doped in the outer region of the crystal. It was reported that the grown MoS₂–MoS₂:Nb had a built-in p–n lateral homojunction (i.e., a true 2D diode or one-dimensional p–n junction) because (i) the inner and outer regions demonstrated n-type and degenerate p-type I_ds–V_g characteristics, respectively. (ii) Moreover, at the p–n junction, a clear current rectification behavior was observed. We believe that the results of this study can provide a robust way for achieving high performance true 2D devices and obtaining a better understanding of the properties of 2D homojunctions based on TMDs.

See the supplementary material for the magnified AFM image, DFT calculation results of MoS₂:Nb, Nb 3d core-level spectra and corresponding line profile, KPFM image, supplementary discussion on the Raman shift of MoS₂:Nb, optical properties of exfoliated MoS₂:Nb, TEM observation results, and V_g-dependent rectification ratio of MoS₂–MoS₂:Nb.

This study was supported by JSPS KAKENHI Grant No. 19K15403; JST-CREST Grant Nos. JPMJCR16F3 and JPMJCR18T1; and JST-PRESTO Grant Nos. JPMJPR20T7 and JPMJPR17NB, Japan. ESCA spectral datasets were obtained with the support of the University of Tokyo outstation beamline, BL07LSU, at SPring-8 (Proposal Nos. 2020A7471, 2020A7486, 2019A7451 and 2021A7422). We are grateful to Y. Miyata (TMU), Y. Nakanishi (TMU), A. Ando (AIST), and Y. Okigawa (AIST) for the discussion on growth and characterization. We also thank W. Zhang (The University of Tokyo) for the support of measurements at SPring-8. M.O. acknowledges Y. Sasaki (JFCC) for the discussion on SAED analysis. Electrical measurements and lithography processes were conducted at AIST Nano-Processing Facility.

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

The authors have no conflicts to disclose.