Control of the Schottky barrier height in monolayer WS₂ FETs using molecular doping Open Access

Atomically thin two-dimensional (2D) materials received widespread attention due to their superior intra-layer transport of fundamental excitations, such as charge, heat, spin, and light.¹ Graphene is one of the most intensively investigated materials since its discovery in 2004.² Despite its excellent carrier mobility, graphene lacks a bandgap that limits its use in electronics.^3–5 In contrast, several 2D transition metal dichalcogenides (TMDs) possess sizable bandgaps around 1–2 eV, and their band structure can be tuned through strain engineering and thickness control. WX₂ or MoX₂ (X = sulfur and selenide), in particular, have been widely studied in transistors,^6–8 tunneling devices,^9,10 sensors,^8,11 flexible electronics,¹ and optoelectronics.^12,13 MoS₂ is particularly the most representative and intensively studied of the TMD materials because it is relatively stable and readily available. Monolayer MoS₂ has a relatively high mobility, and a direct bandgap makes it interesting for applications in integrated circuits and optoelectronics.¹⁴ Similar to MoS₂, WS₂ possesses a direct bandgap of 2 eV in the monolayer regime,¹⁵ spin and valley coupling,¹⁶ and band structure tunability with strain.¹⁷ Due to its reduced effective mass, WS₂ is predicted to have the highest mobility among TMDs based on density functional theory calculations.¹⁸ 

Despite these promising properties, several significant challenges still limit the potential of TMD-based devices: production of films in the monolayer regime by mechanical exfoliation and large-area synthesis are generally limited to small sizes or poor film quality and uniformity, respectively;¹⁹ significant sheet resistance and contact resistance exist at the metal and semiconductor interfaces which decrease the overall device performance.^20,21 The role of contacts and interfaces of 2D materials are significant in terms of impact on device performance.^22,23 Since the pristine surface of 2D materials does not tend to build covalent bonds, the interfaces between the metal and 2D materials in top-contact situations are formed by a van der Waals gap, which can act as a carrier tunnel barrier. Previous research has shown that the important criteria for an efficient charge injection formation include: (1) strong orbital overlap between the channel materials and contacts, (2) a low Schottky barrier (SB) height, and (3) a narrow tunnel barrier.²⁴ The SB height can be modulated by switching the contact metal through contact engineering. However, researchers have shown that the range to adjust the SB height is relatively small because of Fermi level pinning;^25,26 consequently, simple work function engineering alone is not enough for high-performance nanoelectronics.

As in the traditional semiconductor industry, controlled doping is essential for achieving desired device characteristics and improving interfaces and device performance. Doping can be used to effectively tune the electrical and optical properties of semiconductors through the modulation of the carrier concentration, defect vacancies, and interface contact performance. In our previous study, molecular reductants and oxidants were introduced onto the surface of exfoliated flakes of 2D materials to achieve strong and nondestructive n- and p-doping effects.²⁷ Here, we focus on the molecular n-doping of monolayer (1 L) WS₂ using pentamethylrhodocene dimer (RhCpCp*)₂ as the n-dopant. (RhCpCp*)₂ is moderately air-stable in the solid-state and is applied to the surface n-doping of 2D graphene,²⁸ carbon nanotubes,²⁹ metal and metal-oxide electrode materials,³⁰ organic semiconductors,³¹ and various TMDs.²⁷ (RhCpCp*)₂ can donate two electrons to acceptors through bond cleavage accompanied with the formation of two monomeric cations.^32,33 The contact resistance got reduced by more than two orders of magnitude, and SB heights decreased from about 112 to 57 meV after n-doping. We investigated the temperature dependence of the mobility, observing a peak at ≈120 K, limited by charged impurity scattering at low temperatures and electron–phonon scattering at higher temperatures. Controlled doping is a powerful tool for tuning the fundamental properties of 2D materials, and effective interface modification and contact engineering can be used to realize effective charge carrier injection and extraction, thus improving the overall device performance.

Figure 1(a) illustrates the schematic depiction of the device structure. A bottom-gate, bottom-contacted device structure was used. The monolayer TMDs were synthesized through chemical vapor deposition (CVD) and transferred using a PDMS stamp onto pre-patterned metal contacts on the Si substrate with a 300 nm oxide layer. Channel areas of the patterned 1 L WS₂ were obtained with an average area between 10 (L) × 20 (W) μm² and 20 (L) × 20 (W) μm². The detailed synthesis, device fabrication, and transfer technique are described in the Methods section. Fabricated WS₂ devices were then exposed to molecular solution by immersing in 2.5 mM solutions in an argon-filled atmosphere glovebox.

The WS₂ was characterized through Raman spectroscopy and electrical measurements initially. The quality of the CVD-grown WS₂ was confirmed by Raman, photoluminescence (PL), and fluorescence. Figure 1(c) shows the Raman spectra for the pristine and doped monolayer WS₂ obtained with a 514 nm excitation at room temperature. Two strong features are observed: one single resonance peak at 418 cm⁻¹ corresponds to the A_1g mode. The other feature comprises two peaks, a second-order Raman peak (2LA) at 351 cm⁻¹, overlapping with the E¹_2g peak at 356 cm⁻¹. The 2LA mode is stronger than the E¹_2g mode. The E¹_2g phonon mode comes from the in-plane vibrations between W and S atoms, whereas the A_1g peak represents the out-of-plane vibration of S atoms. These data are in agreement with the calculations and experimental measurements.³⁴ After doping, we observe a significant downshift of the A_1g peak position by ∼1 cm⁻¹ (“softening” or redshift), while the E¹_2g peak position shows a similar change (∼0.7 cm⁻¹). The peak shift is also accompanied by the intensity decrease and peak broadening after the treatment. The n-doping results agree with the previous results on electrostatic and chemical doping of the monolayer WSe₂ flakes.³⁵ The high electronic concentration in the n-doped WS₂ increases the electron–phonon scattering and affects the full width at half maximum (FWHM) and frequency of phonons via the renormalization of their self-energy. Our experiment leads to a Raman mode softening and peak broadening after n-doping.

The WS₂ FETs were fabricated by transferring the CVD grown 1 L onto pre-patterned electrical contacts and treated with the n-dopant by dipping into the 2.5 mM solutions. The 1 L WS₂ film studied exhibited field-effect mobility of ≈15 cm² V⁻¹ s⁻¹ prior to molecular doping, comparable to a previous study on few-layer WS₂ based FETs.^36,37 Figure 1(d) shows representative transfer characteristics (I_sd vs V_bg curves) of the WS₂ FETs at room temperature after doping with the dimer. The threshold voltage, V_th, is extracted through linear extrapolation of the drain current measured as a function of gate voltage. When Au is used as the contact metal, pristine WS₂ FETs show n-type behavior despite its high work function (WF ∼ 5.1 eV). After n-type doping, the Fermi level of WS₂ moves toward the conduction band minimum, where the contact metal Au also accommodates the shift of the Fermi level. After the first dopant solution treatment(≈30 s), I_sd-V_bg curve shifts significantly to the negative bias side, as shown in Fig. 1(d), which indicates that the WS₂ was n-doped consistent with previous demonstration data.²⁷ As the doping treatment duration increases, the I_sd-V_bg curves shift toward a higher negative voltage, indicating a more substantial n-doping effect. The shift of the I_sd-V_bg curve can be quantified by using the threshold voltage shift (⁠$Δ$V_th). The $Δ$V_th is defined as the difference between the applied V_bg corresponding to an I_sd for the dopant treated device and that for the untreated pristine sample (⁠$Δ$V_th,pristine = 0). A trend of negative $Δ$V_th is observed for dimer treated devices, and a significant increase of $Δ$V_th up to 50 V is achieved after just 1 min of treatment. Longer treatment times lead to the degeneration of doping of the WS₂. As shown in Fig. 1(d), the doping of the WS₂ also impacts the off-state current (I_min). The WS₂ conductance can be effectively increased by more than two orders of magnitude using molecular n-doping for about 10 min. The WS₂ FET devices are in their off state under this gate bias condition, and the number of carriers induced under this gate bias is also negligible. Thus, the change of the I_min is mostly attributed to the electron transfer between the WS₂ and the dopants and the increases in the carrier concentration.

After doping the device, the on/off current ratio increased from ≈10⁵ to ≈10⁶ and the on-current increased by more than one order of magnitude at V_ds = 1 V. Figure 1(e) shows the output characteristics of the same transistor before and after doping. In the low drain bias region (V_ds < 3 V), the I_sd-V_ds for the doped samples show good linearity, indicating the small contact resistance of the device, while the undoped sample shows prominent rectifying characteristics indicating a sizable Schottky barrier at the contact. Meanwhile, the drive current of the doped WS₂ sample has been improved by more than six times compared with the undoped devices.

To better understand the nature of the interface and investigate the Schottky barrier that forms between the contact metal (Au) and the 2D channel material (WS₂), we measured the two-probe I–V curves as a function of temperature and different gate bias. Figure 2(a) shows the I–V curves measured under different temperatures (77–283 K) for pristine WS₂ at V_bg = 20 V. As shown in the figure, the current increases as the temperature increases. The Schottky barrier height was extracted by measuring the activation energy in the thermionic emission regime by using the following 2D thermionic emission equation:^37,38

where A is the contact area, $A2D*$ is the two-dimensional equivalent Richardson constant, q is the magnitude of the electron charge, $Φ$_B is the Schottky barrier height, k_B is the Boltzmann constant, n is the ideality factor, and V_ds is the source–drain bias.

For three-dimensional semiconductors, the typical Arrhenius plot is ln(I_sd/T²) vs 1000/T. Here, a reduced power-law T^3/2 is employed for the Arrhenius plot for two-dimensional semiconductors. As shown in Fig. 2(b), ln(I_sd/T^3/2) is plotted against 1000/T under different source–drain bias (V_ds). The data are linear at each bias, and the slope is subsequently plotted in Fig. 2(c) as a function of drain–source bias. The intercept of the fitted linear curves of the Arrhenius plot in Fig. 2(c) yields −0.834. The Schottky barrier height was calculated from the intercept: [−q$Φ$_B/1000k_B]. For the pristine 1 L WS₂ transistor, the $Φ$_B is 112 meV. This is close to the previously reported Schottky barrier height for 1 L WS₂.³⁹ Previous reports on multilayer MoS₂ for various metal contacts observed a clear dependence on the work function of the contact metals.⁴⁰ Similar analysis was conducted for the molecular-doped 1 L WS₂ transistors, where $Φ$_B is determined to be 57 meV when V_bg = 20 V, about three times smaller than the pristine device.

Such analysis was repeated for various n-doping treatment times, and the $Φ$_B as a function of the doping strength is plotted in Fig. 3(a). The $Φ$_B is reduced from 112 to 89 meV and to 57 meV for the high n-doping level. The numbers of electrons contributed per unit area were determined based on the X-ray photoelectron spectroscopy (XPS) results reported in previous work.²⁷ Here, we demonstrate that the Schottky barrier height can be effectively tuned through doping. The energy diagram of metal–WS₂ contacts before and after doping is shown in Fig. 3(b). Before doping, the charge neutrality level (CNL) in WS₂ is close to the middle of the bandgap, resulting in a large Schottky barrier. The height of the Schottky barrier is large enough to rectify the electron ejection from the metal to the semiconductor at low V_ds. Moreover, the barrier height cannot be efficiently modified by varying the work function of the contact metals because of the complicated metal–2D interfaces and the Fermi level pinning. Without doping, it is much harder for the electrons to inject from the metal contacts to the semiconductor WS₂ because the thermionic current exponentially decreases with the increased barrier height. However, when the tunneling current starts to dominate the current through the metal–semiconductor junction, the electron injection through the barrier becomes easier. The interfacial states and work functions play critical roles in the Schottky barrier height determination. After doping, the CNL in WS₂ shifts closer to the conduction band minimum. At the higher doping levels and increased electron density, the height of the Schottky barrier decreases, and the depletion region is reduced, resulting in an improved interface between the metal contact and 1 L WS₂. The change of $Φ$_B has a significant effect on the current flow across the device. For a highly conductive transport channel, the control of the barrier height through doping permits minimization of the depletion region.

The contact resistance is closely related to the Schottky barrier height. To further quantify the effects of molecular doping on the metal–WS₂ interface properties, we extract the contact resistance by using a four-point measurement technique [device shown in Fig. 1(b)] before and after doping by n-dopant. Figure 4 depicts the gate-voltage dependences of the sheet resistance (R_s), and R_c extracted from the four-point measurement method. Both are significantly reduced after doping. At the low gate field (V_bg = 0 V), the R_c value decreases from 1140 to 220 k$Ω$ μm after the first, 30 s n-doping. At higher doping levels achieved through a 10 min treatment, a low R_c of 7 k$Ω$ μm was achieved. The reduced contact resistance arises from the Schottky barrier thinning induced by doping, enabling a much higher tunneling current to pass through the barrier. The reduction of R_c through this doping technique is significant for TMDs.

Here, we report the mobility measurements before and after doping monolayer WS₂ based on the FETs. The effects of contact resistance were removed by using the four-point measurement method to make sure the measurements of the sheet resistivity, and subsequently, the field-effect mobility are accurate. Figure 5 shows the temperature dependence of mobility in the device. Mobility is obtained from the conductance curves in the 50–60 V range of back-gate voltage using the expression for field-effect mobility $μ$ = [dG/dV_bg] × [L/(WC_ox)].⁴¹ The mobility shows pronounced temperature dependence before and after doping the WS₂ layer. This temperature dependence is characterized by a peak at ≈120 K. Below 120 K, we observe a decrease in the mobility as the temperature is lowered to 77 K. This behavior is consistent with mobility limited by charged impurity scattering. On the other hand, increasing the temperature above 120 K also results in a substantial decrease in mobility from the peak value of 4.2 cm² V⁻¹ s⁻¹, which is related to electron–phonon scattering is the dominant scattering mechanism at higher temperatures. We fit this part of the curve with the generic temperature of mobility μ ∼ T^−γ, where the exponent depends on the dominant phonon scattering mechanism. Monolayer WS₂ behaves as a classical semiconductor with the conductance decreasing as the temperature increases, whereas at a high doping level, the conductance increases as the temperature decreases, which is the hallmark of metallic behavior.

In summary, we have studied the contact and mobility of controllable doping of CVD-grown monolayer WS₂ with solutions of redox-active molecular dopants. The n-doping process adds extra carriers into the WS₂ by electron transfer from the dopant species to the semiconductor and dopant cations on the surface. Large carrier densities can be achieved (up to 6 × 10¹² cm⁻²) depending on the treatment time and the solution concentration. Electrical characterization of WS₂ FETs showed a shift of V_th and an improvement of the channel current after doping, further supporting strong n-doping effects. We measured the Schottky barrier height of direct contacted Au-WS₂ using the thermionic emission analysis of the current–voltage curves. We found that the Schottky barrier height was reduced by 60% with molecular doping. Additionally, our doping technique achieves more than two orders of magnitude reduction in the contact resistance. The significant improvement in the interface is due to the reduction of the Schottky barrier width and decreased Schottky barrier height. Overall, the solution-based surface doping described here provides a simple and readily scalable process for tailoring the electrical and interface properties of 2D semiconductors and controlling the electrical and interface properties.

Monolayer WS₂ was grown by chemical vapor deposition (CVD) on SiO₂/Si substrates (275 nm thickness of SiO₂). At ambient pressure, the synthesis is performed in a 2 in. diameter quartz tube furnace. Before usage, all SiO₂/Si substrates are cleaned in acetone, IPA, and Piranha etch and then thoroughly rinsed in DI water. A quartz boat containing ∼1 g of WO₃ powder is placed at the center of the furnace-heating zone with two SiO₂/Si substrates positioned face-down directly above the WO₃. The upstream substrate contains perylene-3,4,9,10-tetracarboxylic acid tetrapotassium salt, which promotes lateral growth. The downstream growth substrate is untreated. A separate quartz boat containing solid sulfur is placed upstream, outside the furnace-heating zone. Pure argon (65 SCCM) is used as the furnace heats to the target temperature. After reaching the target temperature of 825 °C, H₂ (10 SCCM) is added to Ar flow and maintained throughout the 10 min soak and subsequent cooling. To transfer the as-grown CVD WS₂ to the desired substrate, first, we select the desired CVD WS₂ flake using optical microscopy followed by the solution-assist-pick-up method⁴² to pick it up onto the PDMS carrier with a homemade transfer stage, and finally, aligning the WS₂ flake to the prefabricated Au contact to form WS₂/Au contact electrodes/SiO₂/Si substrate.

Dichloromethane and toluene were purchased from Sigma-Aldrich in an anhydrous grade solvent and used as received. Unless stated otherwise, all operations were executed under an atmosphere of argon or low pressure. The doping treatments were performed inside a MBraun glovebox under an atmosphere of nitrogen with low water and oxygen levels. The dopant (RhCp*Cp)₂ was synthesized according to the literature.^43,44 (RhCp*Cp)₂ was dissolved into the toluene for a 0.25 mmol L⁻¹ solution. The TMD flakes were immersed in the dopant solutions for different treatment times and were then cleaned and rinsed using fresh solvent a few times to remove excess dopants.

A bottom gate device structure was used with a thermally grown 300 nm SiO₂ dielectric on the highly doped silicon wafer (resistivity: 0.001 $Ω$ cm to 0.01 $Ω$ cm). The Ti/Au (3/15 nm) source and drain contact area were defined by photolithography, followed by the electron beam deposition at ∼10⁻⁶ Torr. The CVD-grown TMDs were transferred onto the patterned metal contacts using the PDMS/TMD patch previously mentioned. The active area of the channel was patterned through an Unaxis 790 reactive ion etcher (RIE) using a mixture of gas of SF₆ and O₂ at 100 W RF power for 20 s. I–V measurements were taken in the dark under vacuum (base pressure of 10⁻⁶ Torr) with an Agilent 4155C semiconductor parameter analyzer using a Lakeshore probe station. A bottom-gate device structure was used with a 300 nm SiO₂ dielectric. The back-gate voltage V_bg is applied to the highly doped Si wafer. The doping treatments were conducted inside the glovebox, but the samples still experienced a short exposure, shorter than 3 min, to air during the transfer to the measurement chamber under vacuum.

A FET with a special four source/drain contact geometry [as shown in Fig. 1(b)] is used in the four-point measurement method to determine the sheet resistance of the 2D conducting channel independent of the contact resistance. The current is forced between the large source and drain contacts on the left and right in the picture, and the voltage is measured between the two narrow contacts at the bottom. When combined with the device geometry, the resistivity of the WS₂ channel is determined without possible errors due to contact resistance. This sheet resistivity can be used with the geometry of the full channel to determine the total channel resistance, which is then subtracted from the total two-point resistance (measured when the current is forced between the large source and drain contacts and the voltage is measured with the same contacts) to determine the contact resistance of the two contacts. In addition, the inverse of the sheet resistivity is the conductivity of the WS₂ and is used to quickly determine the mobility of the WS₂ in the device.

S.Z. acknowledges the support from the National Institute of Standards and Technology Financial Assistance Award with the Federal Award No. 70NANB16H228. The authors are grateful to Dr. Karttikay Moudgil, Dr. Stephen Barlow, and Professor Seth R. Marder for providing dopant materials. The authors thank Dr. David J. Gundlach and Dr. Emily G. Bittle for their support with the electrical measurements and Erik M. Secula for his advice and help with revision. H.-J.C. acknowledges support from an American Society for Engineering Education fellowship. Certain commercial instruments, equipment, or materials identified in this paper is for the purpose of fostering understanding. It does not imply the endorsement or recommendation by the National Institute of Standards and Technology (NIST). It does not imply that the materials or equipment identified are necessarily the best available.

The authors have no conflicts to disclose.

Siyuan Zhang: Writing – original draft (equal). Hsun-Jen Chuang: Resources (equal); Writing – review & editing (equal). Son T. Le: Resources (equal). Kathleen M. McCreary: Supervision (equal); Writing – review & editing (equal). Kathleen M. McCreary: Resources (equal); Writing – review & editing (equal). Berend T. Jonker: Supervision (equal). Angela R. Hight Walker: Resources (equal). Christina A. Hacker: Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal).

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