Metallic tellurium for p-type contacts of two-dimensional MoTe₂ field-effect transistors

Abstract

While significant progress has been made in the fabrication of n-type contacts for two-dimensional field-effect transistors (2D FETs), the development of high-performance p-type counterparts using compatible techniques remains insufficient to realize competitive complementary circuits. Here, we demonstrate the growth of metallic-phase tellurium (m-Te) on MoTe₂ via evaporation as an efficient p-type contact. The atomic arrangement at the Te/MoTe₂ interface stabilizes m-Te under ambient conditions, forming an atomically sharp van der Waals gap with optimal band alignment and suppressed metal-induced gap states. Combined with hole doping and tellurium vacancies compensation, the interface enables barrier-free hole injection. Bilayer MoTe₂ FETs employing m-Te contacts achieve a contact resistance as low as 1.6 kΩ μm, an on-state current up to 124 μA μm^-1, and a maximum on/off ratio of 10⁷, which are among the best values obtained for p-type 2D FETs. Our work unveils metallic-phase chalcogen as a promising approach for contact optimization.

Introduction

The atomically thin bodies and dangling bond-free surfaces of two-dimensional (2D) semiconductors make them promising candidates for advanced logic devices^1,2,3. However, these features also pose challenges for realizing high-performance field-effect transistors (FETs) using standard fabrication processes. One key issue is the high contact resistance. Charge carriers transport across metal-semiconductor interfaces via diffusion, thermionic emission, thermionic field emission, and pure field emission (tunneling), depending on the barrier height and width at the interface⁴. In silicon (Si) technology, ion implantation is commonly employed to heavily dope the regions beneath contacts, narrowing the barrier and facilitating low-resistance tunneling contacts⁵. Nevertheless, the high-energy impacts involved in this process can irreversibly damage the crystal structure of 2D materials^6,7, rendering it unsuitable for 2D devices.

From the perspective of barrier height, simply selecting a metal with work function matching the semiconductor’s band edge (i.e., metals with low/high work functions for n/p-type contacts) is insufficient to achieve ohmic contact. This is because metal-induced gap states (MIGS) and/or defect states at the contact interface can pin the actual Fermi level within the band gap, resulting in a high charge injection barrier and a severe deviation from the Schottky-Mott rule^8,9. One might expect that 2D semiconductors could avoid this issue because of their dangling-bond-free van der Waals (vdW) surfaces. However, interactions involving the d orbitals between the transition metal atoms in transition metal dichalcogenides (TMDs) and electrode metals lead to serious MIGS^10,11. Additionally, lattice damage on atomically thin 2D bodies during metal deposition further exacerbates Fermi level pinning¹², thus degrading contact performance. Recently, significant progress has been made in n-type 2D semiconductors using semimetal contacts. With energy alignments, suppressed MIGS, and moderate evaporation temperatures, bismuth (Bi) and antimony (Sb) have shown contact performance surpassing even advanced Si technologies^8,9.

In contrast, progress in p-type 2D FETs has lagged far behind, primarily for two reasons. First, achieving p-type ohmic contacts requires the use of high-work-function metals such as gold (Au), platinum (Pt), and palladium (Pd). These heavy metals tend to generate intense thermal radiation during deposition¹²; their high momentum resulting from high atomic mass transfers substantial kinetic energy to the surface atoms of 2D semiconductors, exacerbating lattice damage (Supplementary Fig. 1) and Fermi level pinning⁶. Although lowering deposition temperatures has yielded some improvement, the reported contact resistance for monolayer WSe₂ remains as high as 229 kΩ μm¹². Second, unlike n-type contacts, suitable high-work-function semimetals for p-type contacts are lacking. While a low contact resistance of 0.75 kΩ μm on WSe₂ has been achieved using Sb/Pt contact, it relies on a MoO_x capping layer for p-type doping and exhibits ambipolar transport behavior¹³. Therefore, developing a compatible and scalable strategy to improve p-type 2D FET contacts remains an urgent challenge.

Here, we demonstrate the growth of metallic tellurium (m-Te) with a monoclinic P2₁ phase on MoTe₂, which serves as a promising p-type electrical contact for 2D FETs. The nearly perfect match of interplanar/atom spacing between m-Te (021) and MoTe₂ ($\bar{1}$100) planes enables the stabilization of m-Te under ambient conditions. The interface is atomically sharp with a vdW gap of 0.162 nm (surface to surface)/0.332 nm (atom to atom). First-principles calculations indicate that the work function of m-Te is basically aligned with the valence band maximum of MoTe₂ with mild MIGS and a negligible tunneling barrier. Besides, m-Te p-dopes the underlying MoTe₂ and compensates for its tellurium vacancies, thereby enhancing p-type characteristics. The m-Te-contacted bilayer MoTe₂ transistors show a pure p-type transport behavior with a low contact resistance of 1.6 kΩ μm, an on-state current of up to 124 μA μm^-1, and a maximum on/off ratio of 10⁷. These parameters are among the highest-performance p-type 2D FETs reported to date. Our work presents a promising direction for optimizing contacts of 2D semiconductors via in-situ metallic chalcogens growth.

Results

Main concept of m-Te for p-type contacts of 2D FETs

The key to achieving low-resistance p-type contacts for 2D FETs lies in identifying contact metals with aligned energy levels and a weak Fermi pinning effect. Here, we propose an m-Te as an effective p-type contact for 2H MoTe₂. MoTe₂ is an ambipolar 2D semiconductor that exhibits p-type dominated transport properties through contact and thickness engineering¹⁴. As a transition metal chalcogenide (TMD), each vdW layer of MoTe₂ consists of a Mo atomic layer sandwiched between two Te atomic layers (Fig. 1a). Each Mo atom exhibits a trigonal prismatic coordination with six surrounding Te atoms, forming triangular chalcogen prisms at both ends. These structural units establish a periodic vdW surface potential landscape (Fig. 1b). The lowest and second-lowest energy positions are located above the centers of hexagonal hollow sites and directly above the Mo atoms, respectively, exhibiting a periodicity that matches the equilibrium vdW interaction radius of chalcogen atoms (Fig. 1a, b). This unique configuration enables the controlled introduction of an additional metallic chalcogen layer¹⁵. Given that the fully filled d-orbitals of chalcogens can alleviate the possibility of MIGS generation¹⁰, this induced metallic chalcogen layer has the potential to serve as a good electrical contact for the TMDs.

Fig. 1: Schematic illustration of m-Te induced by 2H MoTe₂ as a p-type electrical contact.

a Top and side views of the MoTe₂ crystal structure. The black rhombus shows the periodicity of the hexagonal hollow sites of the MoTe₂ lattice. The black arrows mark the spacing of MoTe₂ ($\bar{1}$100) planes. b Top view of the energy landscape of Te atoms on MoTe₂. The system energy for Te adsorption was obtained by comparing the total energies of MoTe₂ with a single Te adatom placed at different surface sites. The white rhombus shows the periodicity of the hexagonal hollow sites, i.e., the lowest energy positions, of the MoTe₂ lattice. c Crystal structure of m-Te in a three-dimensional manner (left) and along [100] direction (right). The black rhombus and arrows show the periodicity of Te atoms and the spacing of (021) planes. d Schematic drawing of the bilayer MoTe₂ transistor structure using m-Te as p-type contacts. Blue and coral spheres: Te and Mo atoms.

Elemental Te typically adopts a trigonal phase consisting of helical chains along the c-axis (P3₁21 or P3₂21 space groups, depending on chirality), and has a band gap of 0.33 eV under ambient conditions (Supplementary Fig. 2). However, under high pressure (~45 kbar), Te transforms into a monoclinic metallic phase with a space group of P2₁ (Fig. 1c, d, and Supplementary Fig. 3)^16,17. This phase exhibits a layered, buckled structure resembling black phosphorous but lacks its armchair-chain configuration. Instead, each Te atom forms covalent bonds with four neighbors within zigzag (020) planes, which are stacked together by weak vdW force (the detailed unit parameters can be found in Supplementary Fig. 3e,f).

Notably, the interplanar spacing of m-Te (021) planes matches that of 2H MoTe₂ ($\bar{1}$100) planes, with aligned unit sizes (Fig. 1a, c). Thus, 2H MoTe₂ can be potentially employed as a natural template for m-Te growth. Moreover, m-Te has a work function slightly below the valence band minimum (VBM) of 2H MoTe₂¹⁸, rendering it a promising p-type contact for MoTe₂ FETs (Fig. 1d).

Structure and mechanism of m-Te for p-type contacts of 2D FETs

15/80 nm Te/Au contacts were deposited on MoTe₂ following the above hypothesis, with 15 nm Te serving as a hole injection buffer layer (see Methods). The surface potential induction effect of MoTe₂ was first verified by scanning electron microscopy (SEM) characterization of the Te buffer morphology (Supplementary Fig. 4). The Te film shows a textured morphology with three dominant orientations on MoTe₂ with different thicknesses (Supplementary Fig. 5), which is distinct from the randomly distributed spherical clusters on silicon dioxide (SiO₂), indicating the possibility of van der Waals oriented growth^19,20.

The interface structure of Te/MoTe₂ was further investigated using cross-sectional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) (Fig. 2a, b and Supplementary Fig. 6). Te forms clean and atomically sharp vdW contacts with MoTe₂, showing an interfacial spacing of 0.162 ± 0.01 nm (surface to surface)/0.332 ± 0.01 nm (atom to atom) (Fig. 2b)—smaller than typical vdW gaps in temperature-controlled deposition¹² or transfer contacts⁶. The interplanar distances of Te and MoTe₂ along the interface are 0.310 nm and 0.316 nm, corresponding to P2₁ Te (021) planes and MoTe₂ ($\bar{1}$100) planes, respectively (denoted by the top blue and bottom orange lines in Fig. 2b). The lattice mismatch, defined as $|{d}_{\left(021\right)}-{d}_{(\bar{1}100)}|/{d}_{(\bar{1}100)}$, is only ~1.90%. More importantly, since P2₁ Te belongs to the monoclinic crystal system, Te (021) planes and MoTe₂ (0001) planes are non-orthogonal. Therefore, the (021) interplanar distance differs from that of the evaporated Te atoms at the Te/MoTe₂ interface (0.316 nm, marked by the top orange line in Fig. 2b). This value is identical to MoTe₂ ($\bar{1}$100) planes spacing, manifesting the potential landscape of MoTe₂ surface as shown in Fig. 1b.

Fig. 2: Structure of P2₁ Te/2H MoTe₂ heterostructure.

a,b Atomic resolution images of Te on MoTe₂. a High-angle annular dark field (HAADF) scanning transmission electron microscope (STEM) image showing a clean interface between Te and MoTe₂. Scale bar, 1 nm. b Enlarged dashed rectangle region from (a) in which Mo and Te atoms are visible. Scale bar, 5 Å. The plane and atom distances of Te are shown in blue and orange lines. The plane distance in MoTe₂ is also indicated by orange lines. c Selected area electron diffraction (SAED) patterns of MoTe₂ and Te circled in dotted and solid lines, respectively. Scale bar, 3 nm^-1. The predominant diffraction dots of polycrystal Te are circled. d Simulated diffraction pattern of MoTe₂ and monoclinic Te along [001] and [100], respectively. Scale bar, 3 nm^-1. e X-Ray Diffraction (XRD) pattern of MoTe₂ covered by evaporated Te. Inset, enlarged XRD pattern of P2₁ Te. f Top and side views of the relaxed Te/MoTe₂ structure from density functional theory (DFT) calculation. The black arrows mark the spacing of MoTe₂ ($\bar{1}$100) planes and of Te (021) planes, respectively. g The formation energy of P3₁21 Te and P2₁ Te on MoTe₂.

Multiple characterizations confirmed the existence of monoclinic P2₁ Te, including the selected-area electron diffraction (SAED) of P2₁ Te (021) planes (Fig. 2c, d and Supplementary Fig. 7a, b), fingerprint X-ray diffraction (XRD) peaks of P2₁ Te (021) and (110) planes at 2θ = ~ 30.09^o and 30.42^o (Fig. 2e), and the fast Fourier transform (FFT) of the cross-sectional STEM images (Supplementary Fig. 7c, d). This finding is intriguing because: (i) P2₁ Te, typically requiring high-pressure synthesis¹⁶, now can be obtained through simple evaporation, and (ii) P2₁ Te (02$\bar{1}$) planes, rather than its vdW (001) planes, align parallel to 2H MoTe₂ (0001) planes. Both are attributed to the lattice matching as discussed below.

The experimental diffraction pattern (Fig. 2c) matches well with theoretical predictions (Fig. 2d). The circle where the hexagonal pattern corresponding to MoTe₂ ($\bar{1}$100) planes (dashed mark) is located intersects with a monoclinic pattern marked with solid lines corresponding to m-Te (021) planes. Two features should be noted. First, the spots share almost identical radius (~0.298 nm for m-Te and ~0.303 nm for MoTe₂, corresponding to a mismatch of ~1.65%), aligning well with the STEM results. Second, the three sets of m-Te patterns bisect the hexagonal MoTe₂ pattern with a 60^o inter-angle (Supplementary Fig. 7a), consistent with the three-fold orientation observed in SEM.

To investigate the preferred atomic configurations of evaporated tellurium atoms on MoTe₂ and their structural stabilization mechanism, density functional theory (DFT) calculations were carried out. Figure 2f shows the top and side view of the relaxed heterostructure (see methods), which agrees with the cross-sectional STEM results: evaporated tellurium atoms at the interface are located directly above the hexagonal hollow sites of MoTe₂ surface with matched crystal spaces along the y direction (~0.3056 nm for both m-Te and MoTe₂). The structure of Te in the heterostructure (Supplementary Fig. 8a–d) shows slight distortion from pristine m-Te (Fig. 1c and Supplementary Fig. 3), indicating the interaction between Te and MoTe₂. Due to the spacing mismatch along the x direction, the Te atoms cannot align with the centers of the hexagonal hollow sites. Instead, they sit above positions between the hexagonal centers and Mo atoms in MoTe₂, corresponding to the lowest and second-lowest energy sites (Fig. 1b). The atomic alignment of Te on MoTe₂, combined with the threefold symmetry of MoTe₂ (0001) plane, explains both the three main orientations of Te in SEM and SAED pattern.

We also calculated the formation energy of monoclinic P2₁ Te and trigonal P3₁21 Te on MoTe₂ (see detailed crystal structures in Supplementary Fig. 8). Although P3₁21 is the stable phase under ambient conditions, the formation energy of P2₁ Te is lower than that of P3₁21 phase on MoTe₂ (Fig. 2g). This confirms that the lattice induction from MoTe₂ stabilizes the metallic phase.

This growth strategy for chalcogen allotropes can be extended to other chalcogen-TMD systems. For instance, R$\bar{3}$ Se consisting rhombohedral Se₆ rings can be achieved by depositing selenium on WSe₂ (Supplementary Fig. 9). Analogously, previous studies have demonstrated the formation of metallic S in S/MoS₂ heterostructures¹⁵. These findings suggest that such in-situ unusual-phase growth could serve as a versatile platform for exploring intriguing transport characteristics in chalcogen/TMD vdW heterostructures.

To evaluate the feasibility of m-Te as a p-type contact for MoTe₂ FETs, electronic band structures of m-Te and MoTe₂ were investigated, respectively. First-principles calculations and a clear Fermi step beyond the zero binding energy in Ultraviolet Photoelectron Spectroscopy (UPS) confirm the metallic nature of P2₁ Te (Fig. 3a, the reciprocal primary cell is shown in Supplementary Fig. 7e; UPS results are shown in Supplementary Fig. 11d), consistent with previous reports¹⁶, fulfilling the fundamental requirement for electrical contacts. In addition, the Kelvin probe force microscopy (KPFM) results in Supplementary Fig. 10 indicate that the electric potential of m-Te is higher than that of MoTe₂. The potential difference between m-Te/MoTe₂ heterostructure and MoTe₂ is approximately 45 meV. UPS results in Supplementary Fig. 11a-c indicate the work function of m-Te/MoTe₂ heterostructure is 4.57 eV, larger than the Fermi level of free-standing MoTe₂ (4.3 eV), confirming the positive charge transfer from m-Te to MoTe₂. The band offset of m-Te, 2H MoTe₂ and m-Te/MoTe₂ was further investigated by first-principles calculations in Supplementary Fig. 11e.

Fig. 3: Mechanism of m-Te as p-type contacts for MoTe₂.

a Band structure of P2₁ Te. b Projected density of states (PDOS) of free-standing MoTe₂ (upper panel), where the Fermi level (E_F) lies around the middle of the band gap, and of MoTe₂ in contact with m-Te (lower panel). Mo ${d}_{{z}^{2}}$ and Te ${p}_{z}$ orbitals of MoTe₂ extend into the band gap after MoTe₂ is in contact with m-Te. The E_F of the system approaches the valence band maximum of MoTe₂. c Electron loss of MoTe₂ with different numbers of Te vacancies per supercell after being in contact with Te. d Electron spin resonance (ESR) results of MoTe₂ and Te/MoTe₂. The g value determines at which magnetic field the resonance occurs for a fixed microwave frequency. The signal at g around 2.004 corresponds to unpaired electrons from Te vacancies. e Raman mapping of A_1g peak positions from MoTe₂ to Te/MoTe₂ and then to MoTe₂. Inset: the corresponding optical microscope image of the heterostructure. Scale bar, 3 μm. f, g Raman (f) and X-ray photoelectron spectroscopy (XPS) (g) spectra of pure MoTe₂ and Te/MoTe₂. The arrows indicate the shift directions of the spectra after introducing Te.

After in contact with m-Te, the density of states (DOS) of MoTe₂ (Fig. 3b, lower panel) exhibits minor modifications compared to its pristine counterpart (Fig. 3b, upper panel), in stark contrast to TMD/traditional metal contacts (such as Au, In, Pt, Pd)²¹, where high density of MIGS appears within the bandgaps. The alleviation of in-gap MIGS in the case of m-Te contact indicates a lower possibility of Fermi-level pinning. We zoomed in on DOS of the band gap in Supplementary Fig. 12 and found that the band gap of MoTe₂ in contact with m-Te becomes narrower than that of the pristine one, and the metal-induced states lie within the band as an extension of band edge. Specifically, the valence band edge of Te-contacted MoTe₂ composed mainly of Mo ${d}_{{z}^{2}}$ and Te p_z bends shifts upwards compared with pristine MoTe₂, and the conduction band edge of it composed mainly of Te p_z shifts downwards. As a result, the Fermi level (E_F) of the heterojunction approaches the VBM of MoTe₂, accompanied by electron depletion of it (Fig. 3c and Supplementary Fig. 11f). The above clearly demonstrates the p-type doping effect of m-Te on MoTe₂, which remains robust even in the presence of donor-like Te vacancies in MoTe₂ as shown in Fig. 3c (defect densities: 3.281 × 10¹³cm^-2 and 6.562 × 10¹³cm^-2 corresponding to one vacancy and two vacancies per supercell, respectively) -a common occurrence in practical MoTe₂ samples²².

Furthermore, Te exhibits a tellurium-vacancy compensation effect on MoTe₂. The DOS analysis shows that the energy of defect states shifts toward the VBM with reduced density after heterostructure formation (Supplementary Fig. 13a, b), facilitating electron acceptance and enhancing p-type characteristics. Electron spin resonance (ESR) measurements at 92 K provide direct evidence of this compensation effect, showing diminished Te-vacancy-related resonance (g ≈ 2.004) following Te deposition²³ (Fig. 3d). This compensation stems from interfacial charge transfer enabled by the favorable band alignment between m-Te and MoTe₂.

Besides, the tunneling barrier between m-Te and MoTe₂ was calculated to possess a width (w_t) of 1.27 Å and a height (Φ_t) of 2.95 eV (Supplementary Fig. 11g), both smaller than those of semimetal/MoS₂ n-type contacts^8,9. These results theoretically support the favorable conditions of m-Te as p-type contacts of MoTe₂.

The p-type doping effect of m-Te on MoTe₂ was systematically verified through multiple characterizations. Raman spectroscopy revealed a uniform ~1.5 cm⁻¹ blueshift in the A_1g mode of m-Te-contacted bilayer MoTe₂ (Fig. 3e, f and Supplementary Fig. 14a, b), confirming moderate p-doping, while the unchanged E_2g mode indicated negligible interfacial strain. Note that no widening occurred of the A_1g full width at half maximum (FWHM) (Supplementary Fig. 14c, d), distinct from Pt/Pd-contacted MoTe₂ (Supplementary Fig. 1), indicating negligible surface damage of m-Te contacted MoTe₂. The A_1g peak even became narrower after Te deposition. This might be related to the compensatory effect we mentioned above. Besides, X-ray photoelectron spectroscopy (XPS) measurements demonstrate consistent 180 meV redshifts in both Mo 3d_3/2 and Mo 3d_5/2 peaks (Fig. 3g), with analogous shifts observed in m-Te 3d_3/2 and Te 3d_5/2 peaks (Supplementary Fig. 15), collectively revealing a lowering of the Fermi level when MoTe₂ is in contact with m-Te^24,25,26. KPFM analysis further corroborated these findings, showing a significant surface potential reduction in Te-contacted regions (Supplementary Fig. 16).

Furthermore, we analyzed the interaction between interfacial atoms based on the results of energy loss spectroscopy (EELS) and crystal orbital Hamilton population (COHP) calculations. In the EELS experiment, we observed that the edge shape, i.e., the curve slope of interfacial MoTe₂ Layer 1 in EELS profile (Supplementary Fig. 17) of the m-Te/MoTe₂ scenario is unparallel to those of the layers away from the interface, which refers to changed electronic structures, quite different from the parallel scenario in van der Waals Pt contact¹². Thus, the p-type doping and defect compensation effects brought by m-Te create distinct electronic environments at the interface, making the first-layer Te in MoTe₂ different from that of the inner parts. However, no obvious energy shift was observed, indicating that no new bonds were formed at the interface. In addition, to further assess the interfacial Te–Te interaction strength, −COHP (minus Crystal Orbital Hamilton Population) curves and −ICOHP (integrated −COHP) were further computed. In Supplementary Fig. 18, the −COHP curves show that the Te–Te interaction is much weaker than the Mo–Te interaction within 2H MoTe₂ and the Te–Te interaction within m-Te, manifesting −COHP values hovering close to zero, with only small oscillations. This sharp contrast is then quantified by −ICOHP values, i.e., 0.15 eV per bond for interfacial Te–Te pairs, 2.23 eV per bond for MoTe₂ Mo–Te pairs, and 1.12 eV for Te–Te pairs in m-Te.

High-performance MoTe₂ p-FETs with metallic Te contacts

Finally, we investigate the electrical performance of bilayer MoTe₂ FETs employing m-Te as contacts. The analysis of the thickness of MoTe₂ can be seen in Supplementary Fig. 19 and Supplementary Fig. 20. Figs. 4a, b show the transfer curve (i.e., source-drain current vs. gate voltage, I_ds-V_gs) and output curves (i.e., source-drain current vs. source-drain voltage, I_ds-V_ds), respectively. Notably, the device exhibits primarily p-type transport behavior with excellent linearity at full-range gate voltages (Supplementary Fig. 21) and at temperatures from 100 K to 300 K (Supplementary Fig. 22a–k), confirming the formation of high-quality p-type ohmic contacts. Furthermore, the devices maintained stable transfer characteristics after 10 months of storage, proving the long-term reliability of m-Te contacts for p-type MoTe₂ FETs (Supplementary Fig. 22i); and the m-Te contacts remain stable after 300 °C annealing and 10³s prolonged operation, as shown in Supplementary Fig. 23 and Supplementary Fig. 24. Devices with variable channel lengths (L_ch) were fabricated (Fig. 4c inset) on uniform MoTe₂ strips and the transfer-length method (TLM) was employed to quantitatively determine the contact resistance (R_C). It is worth noting that the contact resistance can generally be divided into the intrinsic part that cannot be controlled by the gate and the junction part that can be controlled by the gate²⁷. Both of these parts contribute to the contact resistance. Although the TLM method cannot provide the origin and mechanism of the contact resistance, it can still offer a benchmark reference for the total interface resistance. Remarkably, the bilayer MoTe₂ devices show R_C as low as 1.6 kΩ μm at a carrier density (p_2D) of ~1.0 × 10¹³cm^-2 (Fig. 4c), with consistent performance observed in multiple devices (Supplementary Fig. 25).

Fig. 4: Electrical properties and benchmark of bilayer MoTe₂ with m-Te contacts.

a,b P-type bilayer MoTe₂ FET electrical characteristics. Transfer curves (a) of bilayer MoTe₂ with m-Te contacts showing pure p-type device characteristics. Linear output characteristics (b) of bilayer MoTe₂ with Te contacts. V_gs, V_ds and I_ds refer to gate voltage, bias voltage and current density, respectively. W = 2.63 μm, L = 788 nm. c Transfer length method (TLM) results of bilayer MoTe₂ with m-Te contacts when the hole density is around 1 × 10¹³cm^-2. The inset shows the optical microscope image of the TLM device. Scale bar, 2 μm. d Arrhenius plots of the Te–MoTe₂ at V_gs from -10 V to -25 V and V_ds of -1 V. T is temperature, with the unit being Kelvin. The barrier for the Te–MoTe₂ interface is negligible. e I_ds–V_ds curves of the Te-bilayer MoTe₂ (exfoliated) field-effect transistor (FET) with L_ch of 130 nm. V_gs: 0 V to -40 V, step: -1 V. I_on is the current density at minimum V_gs (-40 V here) and V_ds of -1 V. f Statistics of I_on for 25 Te-contacted bilayer MoTe₂ transistors. Data are extracted at the hole density around 1×10¹³cm^-2 and at the same V_ds of -1 V. g, h Benchmark of current density and on/off ratio of p-type MoTe₂ FETs (g) and of current density and contact resistance of p-type FETs based on MoTe₂ FETs and undoped WSe₂ FETs (h), with values reported in the literature using different methods. The blue and yellow marks represent the performance of p-type MoTe₂ FETs^{42,43,44,45,46,47,48,49,50,51,52} and WSe₂ FETs^{7,12,28,31,32,53,54,55}, respectively. The symbols for solid and hollow, respectively, represent devices with 1-2 layers and three or more layers. All current densities are collected at V_ds of -1 V.

The excellent p-type contact properties originate from the ideal band alignment between m-Te and MoTe₂, as confirmed by Schottky barrier analysis. Unlike conventional Schottky contacts that exhibit negative slopes in Arrhenius plots (ln(I_ds/T^1.5) vs. 1,000/T, where T is temperature), m-Te/MoTe₂ FETs consistently show positive slopes when the device is turned on, providing direct evidence for barrier-free hole injection (Fig. 4d, Supplementary Fig. 26 and Supplementary Fig. 27). For the details of the Arrhenius curves, please refer to Methods. The absence of an effective hole barrier suggests weak Fermi-level pinning, which can be attributed to the following factors: (1) MoTe₂ remains intact after Te evaporation, ruling out the introduction of interfacial defect states; (2) The vdW gap between m-Te and MoTe₂ minimizes the overlap between the orbitals of m-Te and MoTe₂ and the MIGS induced by m-Te are negligible; (3) The tellurium-vacancy compensation effect of m-Te above MoTe₂ can suppress the defect-induced gap states.

A 130 nm L_ch bilayer MoTe₂ FET was fabricated to further increase the on-state current (I_on). The device shows good linear output curves with an I_on achieved 124 μA μm^-1 at a V_ds of -1 V (Fig. 4e). Figure 4f presents the statistics of I_on measured on 25 MoTe₂ devices with channel length at a p_2D of around 1×10¹³cm^-2, indicating good repeatability of the high-performance p-type contacts using metallic Te (see detailed characteristics of each device in Supplementary Fig. 28).

We finally benchmark our devices with the state-of-the-art 2D p-type FETs based on MoTe₂ (blue) and undoped WSe₂ (orange) (Fig. 4g, h). Given that few-layer 2D semiconductor devices generally exhibit lower performance compared to those based on multilayer counterparts, we distinguish the data between 1-2 layers (solid) and multilayer devices (hollow). It is worth noting that all MoTe₂ samples used in this work were exfoliated from bulk crystals, which are closer to intrinsic semiconductors compared to those grown by chemical vapor deposition (CVD). Our devices yield one of the highest current densities among MoTe₂ transistors and an on/off ratio reaching ~1.2×10⁷, which is comparable to 100 nm-channel WSe₂/Pt devices (CVD-grown) on 5 nm SiO₂²⁸. Although the NO doping method has improved the p-type contact, the thermal stability of which still needs to be enhanced^29,30. To the best of our knowledge, the achieved contact resistance of 1.6 kΩ·μm represents one of the lowest reported values for undoped mono-/bilayer p-type MoTe₂. Compared to metal transfer or CVD-grown epitaxial approaches^31,32, our direct thermal evaporation of Te offers superior compatibility with existing silicon technologies since the transfer process is omitted and there is no need to consider the etching issue after m-Te growth. As shown in Supplementary Fig. 29, the m-Te contact strategy also has the potential to be extended to the fabrication of devices on large-scale MoTe₂, facilitating subsequent integration.

Discussion

In summary, we have demonstrated that in-situ metallic Te can serve as a promising p-type electrical contact for MoTe₂ FETs. This strategy yields a low contact resistance of 1.6 kΩ μm, an on-state current of up to 124 μA μm^-1, and an on/off ratio of 10⁷ in bilayer MoTe₂ transistors, which are among the highest-performance p-type 2D FETs reported to date. The success of this approach stems from three key factors. First, the nearly equal atomic spacing at m-Te/MoTe₂ interface stabilizes m-Te under ambient conditions, bypassing the need for high-pressure synthesis. Second, with aligned energy level, moderate orbital hybridization, hole doping and tellurium-vacancy compensation effect, the interface shows minor MIGS and a minimal hole injection barrier. Third, direct thermal evaporation of Te avoids lattice damage and strong Fermi-level pinning typically associated with high-energy metal deposition. This work not only presents a key step forward in p-type 2D electronics but also establishes a promising approach for vdW engineering through in-situ chalcogens on TMDs.

Methods

Device fabrication

Al₂O₃ deposited on highly doped silicon through atomic layer deposition (ALD) was used as a back-gate substrate. The process was performed at 150 °C with a growth rate of 1.4 Å per cycle. Bilayer MoTe₂ samples were mechanically exfoliated from bulk crystals and identified by optical contrast. Before exfoliating MoTe₂, Al₂O₃ substrates were cleaned by O₂ plasma (20 sccm, 30 W, 5 s) to increase the exfoliation yield. The 15/80 nm Te/Au electrodes were made by electron-beam lithography (EBL) and thermal evaporation. Then, MoTe₂ FETs were annealed at 100 °C for 30 min in an inert gas environment to enhance the contact performance.

We noticed that p-type 2D semiconductors are sensitive to electron beam lithography (EBL) acceleration voltage, which may explain the persistent challenges in achieving high-performance p-type 2D contacts. Although high acceleration voltage (100 kV) enables high resolution and good pattern fidelity (Supplementary Fig. 30a, d), it simultaneously brings a certain degree of n-type doping on 2D semiconductors. As shown in Supplementary Fig. 30, bilayer MoTe₂ FETs with Te contacts exposed at 100 kV exhibit nonlinear output and n-dominated bipolar transfer characteristics. To mitigate this effect, we employed a low acceleration voltage of 10 kV for all device fabrication in this work, which effectively suppressed electron beam-induced n-type in MoTe₂ (Fig. 4a, b). Moreover, the contact areas were exposed using a beam current of 150 pA and an areal dose of 120 in the system to minimize thermal effects and charge accumulation³³.

For TLM devices, MoTe₂ was patterned by EBL and etched to a bar geometry by reactive ion etching with CF₄ and O₂ plasma. The subsequent processing steps are the same as for other normal devices.

Note that the process of adding a protective layer actually prolongs the time that MoTe₂ exposing to the air or introduces more unstable sources, like water during the deposition of dielectrics. As shown in Supplementary Fig. 31, we observed the performance degradation even in multilayer devices after capping an additional layer of Al₂O₃. Therefore, in this work, the m-Te/MoTe₂ FETs did not employ any protective capping layers.

The deposition of m-Te

The Clausius-Clapeyron relation of tellurium’s saturated vapor pressure P and temperature T can be written as³⁴:

$${lg}P\left({torr}\right)=\left(10.45\pm 0.10\right)-\frac{8021\pm 71}{T},T < 723K$$

(1)

Before evaporation, the chamber should be vacuumed to 1 × 10^-5 ~ 1.8 × 10^-5Pa. After starting to heat the tungsten boat, the evaporation rate is controlled at 0.1 Å s^-1, and the chamber soon reaches a stable pressure of about 2.3 × 10^-5Pa (1.73 × 10⁻⁷ torr). The “dynamic equilibrium pressure” here is used as the approximation of the saturated vapor pressure of Te powder, then the temperature of 466.0 ± 6.1 K in a tungsten boat can be obtained by the formula. 15 nm Te was deposited onto MoTe₂ in 25 min and radiative heating of the 2D materials from the crucible during metal evaporation was negligible. For Au deposition, the evaporation rate was 0.2 Å s^-1.

It is worth noting that lattice matching is the principle for obtaining m–Te on MoTe₂, and temperature impacts the m-Te phase formation. To be specific, m-Te used as MoTe₂’s contacts is deposited at room temperature, while semiconducting Te can be grown on MoTe₂ at high temperature or other substrates like SiO₂/Si without the lattice matching effect.

As shown in Supplementary Fig. 32a, after being deposited at room temperature, Te on MoTe₂ and Te on SiO₂/Si show distinct morphology, indicating the significance of lattice match and the inadequacy of SiO₂/Si as the substrate for m-Te growth due to the lack of lattice matching effect. We used to heat the sample stage with MoTe₂ on SiO₂/Si attached to 90 °C to deposit Te, and found the same morphology of Te on SiO₂/Si and MoTe₂ (Supplementary Fig. 32b). According to the XRD results (Supplementary Fig. 32c), these forms of Te refer to trigonal Te (t-Te), i.e., the semiconducting Te. This may be because heating increases the energy of Te adatoms and overcomes the energy barrier induced by MoTe₂ lattice.

Calibration of carrier density

The hole density of MoTe₂ FETs was extracted according to:

$$h=\frac{{C}_{{ox}}}{q}\left({V}_{{th}}-{V}_{{gs}}\right)$$

(2)

By evaporating 200 × 200 μm² pads of Au on the Al₂O₃ substrates, C_ox (F cm^-2) was obtained by capacitance measurements using Keysight 1500 and area normalization. According to the transfer characteristics, the V_th of each device was extracted in the linear region when V_ds = 0.1 V:

$${I}_{{ds}}=\mu {C}_{{ox}}\frac{W}{L}({V}_{{gs}}-{V}_{{th}}){V}_{{ds}}$$

(3)

By fitting

$${I}_{{ds}}=a{V}_{{gs}}+b,$$

(4)

a and b were gained, so

$${V}_{{th}}=-\frac{b}{a}$$

(5)

After obtaining C_ox and V_th, the hole density was derived under certain V_gs for each device.

Schottky barrier extraction

The thermionic emission current injected from a metal contact into a 2D channel through a Schottky barrier can be given by the Richardson-Dushman equation:

$${I}_{{thermo}}={A\!}^{*}{T}^{\alpha }\exp (-\frac{{\varphi }_{B}}{{k}_{B}T})$$

(6)

where A^∗ is the Richardson constant, T is the temperature, k_B is the Boltzmann constant, and φ_B is the effective contact barrier height. Considering the energy-independent density of states of a 2D system, α is 3/2 for 2D semiconductors. When a voltage V_ds is applied across the Schottky barrier, it modifies the barrier height and the carrier injection process. In a Schottky-barrier FET, the reverse-biased contact accounts for the majority of the voltage drop and plays a critical role in governing the transistor’s behavior. The drain current density thermally injected from the metal contact into the 2D channel through such a reverse-biased Schottky barrier can be expressed as:

$${I}_{{ds}}={A}_{2d}^{*}{T}^{1.5}\exp \left(-\frac{{\varphi }_{B}}{{k}_{B}T}\right)[1-\exp (-\frac{{V}_{{ds}}}{{k}_{B}T})]$$

(7)

When V_ds» k_BT, Eq. (7) can be simplified to:

$${I}_{{ds}}={A}_{2d}^{*}{T}^{1.5}\exp (-\frac{{\varphi }_{B}}{{k}_{B}T})$$

(8)

In this regime, the effective energy barrier φ_B at a specific V_gs can be determined by analyzing the slope of the Arrhenius plots. This is achieved using the following relationship:

$${{\mathrm{ln}}}\left(\frac{{I}_{{ds}}}{{T}^{1.5}}\right)=\frac{-{\varphi }_{B}}{{k}_{B}T}+c$$

(9)

As we can see, the Arrhenius plots of on-state Te-contacted MoTe₂ FETs display an opposite trend to the thermionic emission model (Eq. (9)). The absence of a negative slope implies the contact barrier-free nature of the Te–MoTe₂ FETs. It is worth noting that the total measured current is set by the series combination of contact-limited injection and channel resistance, which are altered by temperature-dependent scattering mechanisms³⁵, especially the latter one considering that the channel length of the device used to analyze barrier is 400 nm, manifesting as temperature-dependent current in Supplementary Fig. 26a. This causes the Arrhenius point to exhibit nonlinearity in Supplementary Fig. 26b, c. To make the physical picture clearer, we analyzed the gate-modulated band diagram in Supplementary Fig. 27. After m-Te and MoTe₂ contact, the band offset at the interface is fixed. Since the work function of m-Te is higher than the VBM of MoTe₂, as gate voltages change, the channel part of MoTe₂ can be aligned with respect to the m-Te when the positive gate modulation reaches a certain point, rather than the conventional flat band condition. The “barrier” energy of this pseudo flat-band condition might correspond to the energy required for holes to escape from the potential well at the contact interface.

Measurements

Transport characteristics were measured by applying voltages with the Keysight 1500 semiconductor parameter analyzer system. The low-temperature measurements were performed in a vacuum probe station with liquid nitrogen and a temperature controller.

ESR was tested by Brucker EMXplus-9.5/12. A mass of MoTe₂ was exfoliated to a 3 × 7 mm SiO₂/Si substrate and slid to the bottom of the tube for ESR test at 92 K. After the test, MoTe₂/SiO₂/Si was vacuumized immediately and ready for 15 nm Te evaporation. Then the ESR test of Te/MoTe₂/SiO₂/Si was conducted.

Raman data were collected by 633-nm laser excitation focused through a ×100 objective lens using Renishaw inVia-Qontor. XPS was performed using the Thermo Scientific EACALAB 250Xi. Atomic force microscopy and Kelvin probe force microscopy (KPFM) were performed using the Brucker multimode 8HR. To conduct KPFM, 15 nm Te was evaporated onto exfoliated MoTe₂ on PDMS. Then, Te/MoTe₂ was transferred to a Au substrate with MoTe₂ on the upper side.

DFT calculations

DFT calculations in this work were performed using the Vienna Ab initio Simulation Package (VASP)^36,37. The Perdew–Burke–Ernzerhof (PBE) functional within the generalized gradient approximation (GGA)³⁷ was used to describe the exchange-correlation interactions. The projector-augmented wave (PAW)³⁸ method was adopted, and the kinetic energy cutoff was set to 400 eV. The Brillouin zone was sampled using the Monkhorst–Pack method³⁹ with a k-point mesh density of 0.03 Å⁻¹. The DFT-D3⁴⁰ correction of Grimme with Becke–Johnson (BJ) damping⁴¹ was used to describe the interlayer van der Waals interactions. All structures were fully relaxed. The force and energy convergence criteria were set to 10^-2eV/Å and 10^-5eV, respectively. The heterostructure was modeled using a slab structure. A vacuum layer of 20 Å was added along the z direction to avoid interactions between periodic images.