Barrier-Free Carrier Injection in 2D WSe₂/MoSe₂ Heterostructures via Fermi-Level Depinning

1. Introduction

Among various two-dimensional (2D) materials, transition metal dichalcogenides (TMDCs) are attracting tremendous interest due to their promising applications in electronics [1,2], valleytronics [3], optoelectronics [4,5]and spintronics [6]. Specifically, dangling bond free surfaces of 2D TMDCs enable the flexible construction of van der Waals (vdWs) heterostructures, providing opportunities for developing new devices with advanced functions. The observation of moiré superlattice exciton states in WSe₂/WS₂ vdWs heterostructures experimentally provides an attractive platform for controlling excited states of matter [7]. Tunneling field effect transistors based on 2D TMDCs heterostructures with subthreshold swing smaller than 60 mV dec⁻¹ were demonstrated previously [8,9]. Intriguingly, selection of appropriate TMDCs layers allows type II heterostructures to be constructed, in which the ultrafast dynamics of charge transfer across the vdWs interface and the spatial segregation of photo-generated holes and electrons can be achieved. The electron transfer from WSe₂ to MoS₂ taken place within 470 fs with 99% charge transfer efficiency in MoS₂/WSe₂ p-n heterojunctions was reported [10]. Policht et al. observed an ultrafast interlayer electron transfer of up to 69 fs in WS₂/MoS₂ type II band alignment by using 2D electronic spectroscopy [11]. Similar sub-ps level charge transfer processes were also demonstrated in MoS₂/MoTe₂, MoSe₂/MoS₂ and WSe₂/WS₂ heterostructures [12,13]. However, the photoresponse times drop to the order of μs when made these heterojunctions into optoelectronic devices [14,15].

Inevitably, 2D TMDCs-based devices have contact interfaces with metal electrodes, and the quality of the electrical contacts becomes particularly important on their performances. 2D semiconductors generally exhibit unfavorable band edges due to the strong excitonic effect and quantum confinement [16], thus, for most commercially available metals (with the work function ranging from 3.5 eV to 5.7 eV) [17] the formation of low resistance ohmic contacts to mask the intrinsic exceptional properties of 2D TMDCs is still challenging to obtain [18]. The Schottky barrier height (SBH) at the metal-2D TMDCs interfaces, not simply depending on the difference between the work function of the metal and the conduction band minimum (CBM)/valence band maximum (VBM) of the TMDCs due to the complex Fermi level pinning (FLP), is a measure of the barrier to charge injection. A large SBH leads to high contact resistance at the metal-TMDCs interfaces, reducing the carrier injection efficiency [19,20]. Therefore, it is highly desirable to decrease SBH (a low resistance ohmic contact with vanishing SBH) to gain a high performance of a device.

A Schottky pinning factor S is normally introduced to describe the strength of the FLP, which is defined as the change in SBH (Φ_SBH) with respect to the metal work function (W_F), i.e., S=|dΦ_SBH/dW_F| [17,21,22]. Not that S is the slope of Φ_SBH versus W_F, typical values reported are in the range of 0~1 [17,21,22,23], where a value close to 0 corresponds to a strong pinning interface, and a weakly interacting metal-TMDCs system is achieved for S close to 1, indicative of approaching the ideal Schottky-Mott limit. The FLP effects can be normally attributed to the interfacial effects [24], surface traps [25] and metal-induced gap states (MIGSs) [24,26,27]. Several works reported that the FLP can be weakened by inserting a buffer layer between the TMDCs and metal experimentally and theoretically, such as hexagonal boron nitride (h-BN) [28,29], molybdenum trioxide (MoOx, x<3) [30], ultrathin TiO₂ [31], graphene [32,33], ZnO [34], monolayer (ML) NbS₂ [25], VS₂ [26], and so on. These results show that the introduction of buffer layers is an effective strategy to break the direct metal-TMDCs interaction and destroy the interface states, improving the contact properties.

Beyond their established advantages in ultrafast optoelectronics and emerging quantum information applications [35,36,37], WSe₂/MoSe₂ heterostructures present unique opportunities for fundamental interface engineering studies. Through density functional theory (DFT) simulations of four conventional metals (Ag, Al, Au, Pt) interfaced with monolayer WSe₂/MoSe₂ heterostructures, we reveal universal Schottky-type contacts with characteristic Fermi-level pinning. No ohmic contact is observed, MoSe₂ forms an N-type Schottky contact with Ag and Al electrodes with electron SBH of 0.31 and 0.88 eV, and a P-type Schottky contact with Pt and Au electrodes with hole SBH of 0.42 and 0.76 eV, respectively. A similar trend is also observed in metal-WSe₂ contacts, where the Pt-WSe₂ exhibits a minimum hole SBH of 0.43 eV. Furthermore, we also demonstrated that inserting a 2D materials metallic layer (mMoSe/mWSe) between the metals and WSe₂/MoSe₂ layer indeed weakens the interaction at the contacting interfaces, which can effectively suppress the MIGSs, depinning the Fermi level, and thus reducing all the SBHs significantly.

2. Computational Methods

We use the Projector Augmented Wave (PAW) method and plane wave basis set implemented in the Vienna ab initio simulation package (VASP) code to optimize the geometries [23,24]. Using the Perdew Burke Ernzerhof (PBE) exchange correction function to modify the generalized gradient approximation (GGA) function [38]. To ensure accuracy, the plane-wave cutoff energy is set to 600 eV. The Brillouin zone are sampled with special k-points of 5×5×1 for optimizing supercell structures and 10×10×1 for densities of states (DOS) calculations. DOS calculations are implemented by using Gaussian drag method with a drag width of 0.02 eV [39,40]. The calculation is considered complete only when the energy difference between two consecutive steps is less than 10^-6 eV and the force acting on each atom is less than 0.01 eV/Å.

To accurately model metal electrode interfaces, six atomic layers of (111)-oriented Ag, Al, Au, and Pt slabs were employed based on converged surface properties established in prior studies [41,42]. The in-plane lattice constants of WSe₂ and MoSe₂ are a₁=3.29 Å and a₂=3.32 Å showed excellent agreement with experimental measurements [43,44]. A supercell with a lattice mismatch of less than 5% is needed to get stable structures of heterojunctions. Thus, the pristine metals/WSe₂/MoSe₂ supercells are constructed using a 2×2 expansion of Ag, Al, Au, and Pt (111) and a√3×√3 expansion of monolayer (ML) WSe₂/MoSe₂, resulting in lattice mismatches ranging from 0.21% to 3.56% as displayed in Tables 1. A vacuum layer exceeding 15 Å thickness was implemented along the z-direction to eliminate spurious interactions between periodic images.

3. Results and Discussion

Figure 1(a) and (b) illustrate the architecture of top-contact configurations where monolayer WSe₂/MoSe₂ interfaces with six-layer (111)-oriented metal slabs (Al, Ag, Au, Pt). The most stable configuration of the ML WSe₂/MoSe₂ on Ag and Au comes from the structure that the Se atoms are located at the center of a hexagon formed by six adjacent metal surface atoms, directly below the surface hollow site, where the W/Mo atoms are located above the center of a triangle formed by three neighboring metal atoms, as shown in Figure 1(a). Figure 1(b) displays the most stable configuration of Al- and Pt-WSe₂/MoSe₂ interfaces in which the W/Mo atoms sit above the center of the hexagons and Se atoms are above the centers of the triangles formed by metal atoms, respectively. Since the obtained results do not change appreciably beyond six layers of metal atoms [45], we restrict the simulation to this thickness. The equilibrium interfacial distances dz, defined as the difference between Se atoms and metal atoms along the z-direction (perpendicular to the interface) at the contact interfaces, varying from 2.488 to 2.912 Å, decreasing in the order of Al>Au>Ag>Pt (see Table 1). The binding energy of per interfacial Se atom is defined as

E_b = (E_WSe2/MoSe2 + E_metal − E _{metal-WSe2/MoSe2})/N_Se

(1)

Where E_metal, E_WSe2/MoSe2, and E_{metal-WSe2/MoSe2} are the relaxed energies for metal surface, WSe₂/MoSe₂, and the combined system per supercell, respectively, and N_Se is the number of Se atoms at the interface of each supercell. E_b ranges from 0.182 eV (Al-WSe₂) to 0.342 eV (Pt-MoSe₂) as listed in Table 1, demonstrating the bonding strength trend. It can be found that Pt has a relatively strong adsorption (E_b =0.318/0.342 eV) and a small distance (dz=2.53/2.488 Å) with WSe₂/MoSe₂, showing a relatively strong bonding compared to those of other metals.

Before performing the metal-TMDCs systems investigation, we obtain the band structures of pure ML MoSe₂ and WSe₂ layers (see Figure S1 in the supporting information). The results show that ML WSe₂/MoSe₂ is a semiconductor with a band gap of 1.63/1.51 eV, which is consistent with previously reported values [46,47]. Figure 2 shows the projected band structures of metal-ML WSe₂ combined systems where the band structures of metal-MoSe₂ systems are presented in Figure 3. Both majority bands of ML WSe₂ and MoSe₂ are still identifiable when contacted with Ag, Al, Au and Pt surfaces. Although the conduction bands of MoSe₂/WSe₂ is reserved well when contacted with Pt, as shown in Figure 2(d) and Figure 3(d), the valence bands exhibit a slight hybridization, which are related to the d-orbital radius and occupied level of metals [47]. Actually, the band hybridization degree in different metal-MoSe₂/WSe₂ interfaces can be explained by the d-band model [48]. Pt has partially occupied d-orbitals leads to stronger bond with MoSe₂/WSe₂ than that of Ag and Au (full d-shells), which can mix with the band-edge d-orbitals of W and Mo, resulting in a better electron injection efficiency. This result is in good agreement with the binding energy calculation. Accordingly, Al is not a good choice for metal-MoSe₂/WSe₂ contact owing to the absence of d-orbitals [41,46], and a high resistance contact of Al-TMDCs has been verified experimentally [49].

After formation of the contacts, the Schottky barriers appear in the vertical direction, and thus vertical Schottky barriers height can be precisely determined by Φ_n= E_CBM-E_F (electrons) and Φ_p= E_F-E_VBM (holes), where E_CBM, E_VBM, and E_F are the positions of the CBM, VBM, and Fermi level in the metal-MoSe₂/WSe₂ junctions, respectively. Φ_SBH is crucial for the optical and electrical properties of MoSe₂/WSe₂ device. As shown in Figure 2(a) and (b), in Ag- and Al-WSe₂ contacts, the Schottky barriers are N-type with electron SBHs of Φ_n= 0.43 and 1.1 eV, respectively. While in Au- and Pt-WSe₂ contacts (see Figure 2(c) and (d)), the vertical Schottky barriers exhibit P-type, giving hole SBHs of Φ_p =0.66 eV and 0.43 eV. The extracted SBHs values for various metal-WSe₂ combinations are listed in Table 1. Using a same approach, as displayed in Figure 3, SBHs for all of the metal-MoSe₂ systems are calculated, and the extracted electron/hole SBHs as given in Table 1. Similarly, the Schottky barriers are N-type with Φ_n=0.31 and 0.88 eV in Ag- and Al-MoSe₂, and are P-type with Φ_p = 0.76 eV and 0.42 eV in Au- and Pt-MoSe₂, respectively. In these contacts, the large Φ_n of Al-MoSe₂/WSe₂ indicating a high resistance in the contact interfaces, which is consistent with the previous analysis. Actually, the contact polarity of Au-WSe₂ has generally been reported to be ambipolar or P-type [50,51,52]. However, a predict N-type contact polarity could be realized with selenium vacancies [24]. A reported drop of P-type SBH from the transferred Au electrode (0.62 eV) to the evaporated Au electrode (0.19 eV) for WSe₂ ML devices [51], indicating that a SBH value can be tuned by different fabricated approaches. Moreover, For Pt-MoSe₂/WSe₂ contact systems, the vertical hole SBHs (0.42/0.43 eV) is comparable with the one (0.55/0.34 eV) reported by other works [46,47].

Contact engineering is necessary to suppress the Φ_SBH of the metal-TMDCs systems with the goal to achieve a low resistance Ohmic contact for 2D devices. An ultrathin buffer layer is a technologically viable approach to reduce the Φ_SBH. It has been reported, for example, that a hBN buffer layer in Ni-MoS₂ contact can decrease the SBH from 0.158 to 0.031 eV [53]. Also, a graphene buffer layer is effective in reducing the SBH from 0.3 to 0.19 eV in Ag-MoS₂ configuration [54]. Moreover, thin oxides, such as Al₂O₃, have been used as buffer layer in Ti-MoS₂, reducing the SBH from 0.18 to 0.13 eV [55]. Here, we report the metal-doped MoSe₂/WSe₂ (mMoSe/mWSe) acts as a buffer that can be sandwiched between metal electrodes and 2D MoSe₂/WSe₂ semiconductors (as illustrated in Figure 1(c)). It is important to stress that the configuration of metal atoms replacing Se atoms required less formation energy and is more stable than that of replacing Mo/W atoms, and a mMoS layer has been realized via a three-step plasma-deposition-annealing yttrium-doping process experimentally [56]. The metal-doping means to covert semiconducting MoSe₂/WSe₂ into metallic mMoSe/mWSe is theoretically predicted. Au-doping is taken as an example (see Figure S2), the band structure of AuMoSe/AuWSe with the VBM as well as the CBM coinciding in the Brillouin zone, confirming the zero bandgap metallic characteristic. Figure 4(a)-(d) show the projected band structures of metal-mWSe-WSe₂ complexes. Obviously, after inserting an 2D mWSe layer into the contact interfaces, the barrier heights are reduced to 0.19, 0.04, 0.13 and 0.19 eV in the Ag-, Al-, Au- and Pt-mWSe-WSe₂ contacts, respectively. While for metal-mMoSe-MoSe₂ configurations, as shown in Figure 5(a)-(d), the corresponding vertical Schottky barriers are extracted to 0, 0.02, 0.03 and 0.29 eV, respectively. All the SBHs of electrons and holes are summarized in parentheses in Table 1. It is obvious that Φ_n is decreased significantly in the N-type contacts while Φ_p is relatively slight in the P-type contacts compared with those of without involving a buffer layer. Intriguingly, in contrast to the previous conclusion, the ultra-low Schottky barrier heights of Al-mMoSe/mWSe-MoSe₂/WSe₂ proves that Al becomes a good metal for low resistance contact after inserting the buffer layer. Furthermore, the transition from P-type contacts at Au-MoSe₂/WSe₂ interfaces to N-type contacts at Au-mMoSe/mWSe-MoSe₂/WSe₂ interfaces is observed. It is worth noting that a vanishing vertical N-type SBH in Ag-mMoSe-MoSe₂ combined system indicates a low resistance Ohmic or quasi-Ohmic contact (Figure 5(a)), implying a high carrier injection efficiency.

Apparently, the extracted Φ_SBH for various configurations are different from the predicted values by straightforward band alignment estimation, which are ascribed to the Fermi level pinning effect at the interfaces. To further quantify the strength of pinning behavior in different contacts, we plot the work function dependent Φ_n between semiconductor MoSe₂/WSe₂ and various metals systems, as shown in Figure 6. It is found that the SBHs of metal-WSe₂ interfaces lie on a line of slope S=0.41, while those of metal-MoSe₂ contacts lie on a line of slope S=0.33 (see Figure 6(a)), deviating from ideal Schottky-mott limit, indicating a strong pinning effect. Nevertheless, it is worth noting that these values are much larger than that of conventional semiconductor combines (GaAs (0.07) and Si (0.27)) [57]. For comparison, as exhibited in Figure 6(b), we also extract the value of S factors with mMoSe/mWSe buffer layers inserted between the metals and MoSe₂/WSe₂. A large pinning factor close to 1 (0.79 for metal-mMoSe-MoSe₂ and 0.76 for metal-mWSe-WSe₂) indicates a depinning Fermi level effect, which can be attributed to the reduction of the MIGS density in the contact stacks due to a weakened interaction between the metal and MoSe₂/WSe₂ layer. It is reported that an increased metal-TMDCs distance also can effectively suppress the MIGSs, and thus depinning Fermi level, reducing the SBHs [23].

To deeply understand the effect of conventional metals or mMoSe/mWSe layers on the ML MoSe₂/WSe₂ semiconductor, we calculate the partial density of states (PDOS) of pristine MoSe₂/WSe₂ and the contact systems. As shown in Figure 7(a)-(d), after contacted with metal faces, significant MIGSs are induced in the original band gap of WSe₂ compared to that of the pristine WSe₂ without contact (see Figure S3), which is consistent with previous studies [17,46]. It is observed that much more overlap interface states distributed in the pristine band gap for Pt contact compared with those of in other contacts (Figure 7(d)), indicating a strong hybridization in Pt-WSe₂ combine, which is in agreement with the band structure hybridization degree. In contrast, as indicated in Figure 7(e)-(h), the corresponding metal-doped metallic mWSe in contact with ML WSe₂ has negligible metal-induced gap states, leading to almost no additional Fermi-level pinning, which can be attributed to the native ideal vdWs interfaces between the mWSe and WSe₂ layers. Moreover, similar contact nature trends are also obtained by comparing the PDOS in metal-MoSe₂ and mMoSe-MoSe₂ systems (see Figure S4 in the supporting information).

A prototype device (optoelectronic/electronic) based on MoSe₂-WSe₂ vdWs heterostructure consisting of a drain (D), a source (S) and a heterostructure channel is illustrated in Figure 8(a). d_Z measured from optimized geometries (Table 1) and a notably high ΔV, defined as the potential energy above E_F at the metal-WSe₂/MoSe₂ interfaces, is obtained (Figure S5), verifying the existence of tunneling barriers (TBs) [17,46]. In the light of TBs and Schottky barriers which are analyzed to evaluate the carrier injection efficiency of contacts, take N-type contact as an example, band diagrams for metal-WSe₂/MoSe₂ stacks and metal-interlayer-WSe₂/MoSe₂ stacks are shown in Figure 8(b) and (c), respectively. It is important to note that a narrow tunneling barrier and low Schottky barrier at metal-TMDCs interface can increase the carrier injection efficiency [17,18,46]. In the absence of buffer layer, a thin (even a vanishing) tunneling barrier is formed at the metal-WSe₂/MoSe₂ interface which is related to the band hybridization degree (Figure 8(b)). Although inserting a mMoSe/mWSe layer leads to a larger distance due to the nonbonding vdWs gap (~0.70 nm) between the mMoSe/mWSe and ML WSe₂/MoSe₂, resulting in a higher tunneling barrier, the weakened interaction at the interfaces can effectively suppress the formation of MIGSs (Figure 8(c)), thus reducing the SBHs.

4. Conclusions

The strategic modulation of Schottky barrier heights (SBHs) for both carrier polarities constitutes a critical advancement in developing high-efficiency two-dimensional optoelectronic devices. In this work, we have systematically investigated the contact interfaces between WSe₂/MoSe₂ heterojunction and a set of metals with a wide range of work functions (Ag, Al, Au, and Pt). Our calculation results show that the directly contacting WSe₂/MoSe₂ with various elemental metals leads to partial FLP due to the large density of MIGSs, and the electron/hole SBHs can be modulated effectively by varying metals (with the relatively large values ranging from 0.31 eV for Ag-MoSe₂ contact to 1.11 eV for Al-WSe₂ contact). A larger value of S close to 1 can be obtained by inserting a metal-doped mWSe/mMoSe layer as a metallic buffer between the metals and the 2D WSe₂/MoSe₂ semiconductor, suggesting a depinning Fermi level effect. In the presence of a buffer layer, both the polarities of Au-MoSe₂ and Au-WSe₂ contacts change from P-type to N-type, and all the SBHs can be reduced to a small even negligible value due to the suppression of MIGSs, implying a low resistance Ohmic contact. Our studies provide a theoretical reference for designing of high-performance 2D TMDCs devices.