Largely Reducing the Contact Resistance of Molybdenum Ditelluride by In Situ Potassium Modification

Semiconducting molybdenum ditelluride (MoTe2) is widely reported owing to its favorable electronic and optoelectronic properties. The effective modulation of its electrical characteristics has garnered growing attention in regard to building high‐performance MoTe2‐based complementary devices. However, the inherent Schottky barrier (SB) in MoTe2‐based devices severely inhibits the charge‐carrier injection efficiency, leading to a high contact resistance between MoTe2 and contact metals. Here, an efficient method is presented for reducing the SB height of field‐effect transistors (FETs) based on MoTe2 by in situ potassium modification. Interestingly, the electrons transported from K continuously change the electrical characteristics of MoTe2 FET from ambipolar to n‐type with an improvement of electron mobility of over one order of magnitude. Meanwhile, the contact resistance of MoTe2 FET is significantly decreased from 11.5 to 0.4 kΩ µm. By regulating the modification region spatially, it is possible to create a complementary inverter with a high gain of ≈32 at VDD = 3 V. This research demonstrates a relatively simple method for optimizing the contact for MoTe2‐based devices and tuning the electrical properties of MoTe2 for future high‐performance complementary electronics.

In traditional circuits, unipolar p-type and n-type FETs are the fundamental building blocks of semiconductor devices such as p-n diodes, [17,18] logic inverters, [19][20][21] photodetectors, [22] and solar cells. [23,24] However, MoTe 2 demonstrates an ambipolar behavior in most scenes due to the mid-gap Fermi level pinning occurring at the metal/MoTe 2 interface. [25,26] On the other hand, the mid-gap Fermi level pinning effect will induce an inherent Schottky barrier (SB) between metal electrodes and MoTe 2 , [13,27] which dominantly will affect the carrier injection type and efficiency, as well as the contact resistance. Several orders of magnitude higher contact resistance have been observed in MoTe 2 -based FETs compared to the conventional semiconductor materials. [27] These high contact resistances face formidable challenges in controlling transport polarity of MoTe 2 semiconductors and improving the related charge transport, which will further affect the development of reliable MoTe 2 electronic devices in the future. Various effects have been utilized to modulate the electronic properties and reduce the contact resistance of MoTe 2 by decreasing the SBs at the metal/ semiconductor contacts, including electrostatic gating, [15,28] interlayer contacts, [29,30] dry transfer of contact metals, [25] phase engineering contacts, [31,32] and so on. [12,33] However, a sophisticated fabrication process is required in these strategies, which faces great technical challenges, and the preparation efficiency is greatly affected by the quality of dielectric or contact Semiconducting molybdenum ditelluride (MoTe 2 ) is widely reported owing to its favorable electronic and optoelectronic properties. The effective modulation of its electrical characteristics has garnered growing attention in regard to building high-performance MoTe 2 -based complementary devices. However, the inherent Schottky barrier (SB) in MoTe 2 -based devices severely inhibits the charge-carrier injection efficiency, leading to a high contact resistance between MoTe 2 and contact metals. Here, an efficient method is presented for reducing the SB height of field-effect transistors (FETs) based on MoTe 2 by in situ potassium modification. Interestingly, the electrons transported from K continuously change the electrical characteristics of MoTe 2 FET from ambipolar to n-type with an improvement of electron mobility of over one order of magnitude. Meanwhile, the contact resistance of MoTe 2 FET is significantly decreased from 11.5 to 0.4 kΩ µm. By regulating the modification region spatially, it is possible to create a complementary inverter with a high gain of ≈32 at V DD = 3 V. This research demonstrates a relatively simple method for optimizing the contact for MoTe 2-based devices and tuning the electrical properties of MoTe 2 for future high-performance complementary electronics.

Introduction
2D-layered materials, such as transition metal dichalcogenides (TMDs), have been a beyond "complementary metal oxide semiconductor (CMOS)" focus for some years because their atomic thin nature offers a great opportunity to overcome the scaling www.advelectronicmat.de interface. Hence, a heavy doping technique of 2D MoTe 2 devices is required to tune the electronic polarity and decrease contact resistance. Conventional doping techniques such as thermal diffusion and ion implantation are not suitable for 2D materials as their atomically thin nature is a great challenge to maintain highly energetic impinging ions and this could cause possible structural failure. Substitutional impurity doping has more stable performance compared to charge-transfer doping by surface adsorbates. [34] Electrostatic doping strategies exhibit excellent sensitivity in nondestructive, steady, reversible, and industrial availability. [35] The remote modulation doping has received great attention as an effective doping method. The recent study showed that remote doping modulates the chargecarrier density with the unchanged 2D lattice, retaining the mobility and suppressing the charge scattering. [36] However, the remote doping approach has also its own limitation including the sophisticated device fabrication and inevitable energy consumption. As a damage-free and nonvolatile doping technique, surface charge-transfer doping (SCTD) depends on the interfacial charge transfer without creating significant defects to damage the lattice structure of doped materials, [37][38][39] which offers valuable insights into the improvement of the device characteristics, including modulating the carrier type, concentration and mobility, and optimizing the contact quality.
In the following paper, we report a simple technology to effectively tune the electronic and contact characteristics of few-layer MoTe 2 FETs via in situ surface functionalization with potassium (K) adatoms. After K doping, a significant electron doping effect on the MoTe 2 devices was obtained with two orders of magnitude improvement in electron concentration. The contact resistance of MoTe 2 devices significantly decreased from 11.5 to 0.4 kΩ µm. In addition, we achieved complementary inverter devices with the high gain of ≈32. Our work establishes a critical step for building up high-performance complementary devices by improving the electrical properties of MoTe 2 , and represents a great potential in the future application of electronics and optoelectronics. Figure 1a shows the schematic diagram of a few-layer MoTe 2 FET on the SiO 2 /Si substrate. The thickness measured by atomic force microscopy (AFM) in Figure 1b is about 5.5 nm (approximately eight atomic layers [40,41] ). The Raman spectrum of the few-layer MoTe 2 exhibits three elevated peaks at 174, 236, and 290 cm −1 , corresponding to A 1g , E 1 2g , and B 1 2g vibration modes in MoTe 2 crystal lattice, [37] respectively, as shown in Figure S1 (Supporting Information). Figure 1c illustrates the typical transfer characteristic curve of pristine MoTe 2 FET (I ds -V g ) measured under a drain bias of 1 V, which reveals a p-type dominated ambipolar transport property. The carrier concentration of the device under a specific gate voltage V g can be calculated by the formula

Results and Discussion
where C i demonstrates the capacitance per unit area of the SiO 2 . The threshold voltage V th for electrons is estimated to be ≈27 V by selecting the linear region of electron side and extending until the intersection meets with the horizontal axis. Therefore, the electron concentration of this MoTe 2 FET is calculated to be 7.9 × 10 11 cm −2 at the gate voltage of 50 V. To assess the field-effect mobility, Equation (2) is adopted by extracting the linear regime of transfer curve where dI ds /dV g is the slope of the linear region in transfer plot, L and W are the length and width of the conduction channel, respectively. The hole and electron mobilities of the device in Figure 1c are calculated to be 32 and 12 cm 2 V −1 s −1 , respectively. Figure 1d shows the output characteristics (I ds -V ds ) of the same device. I ds and V ds indicate a superior linearity in the V ds range from −1 to 1 V under various V g , suggesting the Ohmic contacts between metal electrodes and MoTe 2 flake.
To tune the electronic properties of MoTe 2 , K was in situ deposited into its surface and further electrical characterizations (Experimental Section) were performed under high vacuum (≈10 −8 mbar). As shown in Figure 2a, the transfer characteristic curves of MoTe 2 FET continually shift to the negative direction with the increasing K thickness from 0 to 16 Å, demonstrating a significant n-type doping effect. It is worth noting that after 0.64 Å K deposition, the hole transport completely disappears. Furthermore, after thick deposition (≈16 Å) of K, the semiconducting MoTe 2 eventually transfers to the metallic transport and the on-current of electron side indicates over two orders of magnitude improvement compared with the pristine MoTe 2 . MoTe 2 exists more than one phase even at room temperature, and 2H to 1T′ phase transition by electrostatic doping, [42,43] laser irradiation, [31,44] or strain [45,46] has been achieved. We have conducted in situ Raman measurements to characterize the MoTe 2 phase evolution, as shown in Figure 2b. Before doping, the pristine-exfoliated MoTe 2 shows the 2H phase with the characteristic peaks at 174 cm −1 (A 1g mode), 236 cm −1 (E 1 2g mode), and 290 cm −1 (B 1 2g mode). The intensity of the peaks displays a simultaneous decrease at 174, 236, and 290 cm −1 when K dopant increased, and the intensity of Raman spectra of the 1T′-MoTe 2 started to emerge at 86 cm −1 (A g ), 166 cm −1 (B g ), and 266 cm −1 (A g ) after doing 16 Å K. [47] Therefore, MoTe 2 transferred from semiconducting 2H phase to metallic 1T′ phase after thick K doping. Figure 2c exhibits the electron concentration (red dots) under the gate voltage of 50 V and mobility (blue dots) of MoTe 2 as a function of K's thickness. The estimated electron concentration monotonically increases from 7.9 × 10 11 to 1.9 × 10 13 cm −2 at V g = 50 V with the increasing of the K coverage. Interestingly, the electron mobility gradually increases from 12 cm 2 V −1 s −1 for the pristine MoTe 2 to 154 cm 2 V −1 s −1 for 8 Å K-decorated device and then gradually decreases with thicker K modification. We propose that in the first regime, the oxygen molecules exist on the surface of MoTe 2 as the electron-trapping sites by physical or chemical adsorption, and cannot be fully disappeared even under the high vacuum condition. [13,48] Therefore, K atoms can effectively occupy these trapping sites and quickly increase the electron concentration of MoTe 2 . However, further deposition K atoms over 8 Å enlarged the scattering probabilities and thus decreased the carrier mobility. Figure 2d shows I ds as a function of V ds for different values of the K thickness at V g = 50 V. The I ds -V ds curves are essentially linear for the pristine MoTe 2 FET. With the doping time increases, the currents of the I ds -V ds curves increase. Thus, at the increasing K thickness, the linear trend and the increased current represent the improving of the contact, implying SB reduction.
Besides modulating the electronic properties of MoTe 2 in a controlled manner, surface modification of K can also sufficiently improve the contact quality of MoTe 2 , that is, dramatically reducing the Schottky barrier. The I ds -V g curves of pristine and 16 Å K-modified MoTe 2 FET as a function of temperatures are shown in Figure S3a,c (Supporting Information), respectively. With the increase of temperature, a continuous increase of I ds is clearly observed, implying the significant dependence of the current on the temperature owing to the thermionic emission. In the output curves (see Figure S3b,d in the Supporting Information), the I ds -V ds curves are still linear even under 79 K, indicating the Ohmic contact between metal electrodes and MoTe 2 flake. The Schottky barrier of MoTe 2 is extracted by the thermionic emission equation [49][50][51] exp 1 exp ds 2 where I ds is the current through the channel, 2D * A is the 2D equivalent Richardson constant, q is the magnitude of the electron charge, Φ B is the Schottky barrier height, k B is the Boltzmann constant, T is temperature, and V ds is the drain-source bias voltage. For qV ds ≫ k B T, Equation (3) can be simplified as follow exp ds 2 The current flows in the subthreshold regime therefore can be expressed as exp ds 2 where E A = qΦ B + E C∞ − E C0 is the total activation energy, and E C∞ − E C0 is the difference between the conduction band www.advelectronicmat.de minimum in the bulk and at the interface. The so-called flatband condition E C∞ − E C0 = 0 is obtained in the case that V g is equal to V FB . Figure 3a,b illustrates the Arrhenius plots of pristine and 16 Å K-modified MoTe 2 FET. By the slope of ln(I ds /T 3/2 ) versus 1000/T in the Arrhenius plot, the extracted Schottky barrier height of pristine MoTe 2 is obtained (Figure 3c), which is ≈158 meV for electrons. In comparison, a negligible small Φ B of ≈20 meV is achieved after 16 Å K doping (Figure 3d). After K doping, the phase transition from 2H phase to 1T′ phase is formed, because of the semimetal 1T′ phase, the conductivity increases between 1T′ MoTe 2 and Pd/Au electrode, which reduces the SB. [52] The dramatic reduction of Schottky barrier height reveals the pre-eminent ability of K-modification in enhancing the carrier injection for MoTe 2 .
We further extracted the width-normalized contact resistance (R c ) from the total resistance (R tot ) by the transfer length method (TLM). [53,54] Figure 4a shows the transfer characteristics of the pristine MoTe 2 transistor with a different channel length. The current decreases monotonically with the extending channel length. With the 16 Å K doping treatment (Figure 4b), the MoTe 2 FET exhibits a high current with the orders of magnitude increasing compared with the pristine MoTe 2 transistor. As shown in Figure 4c,d, the 2R c is around 23 kΩ µm for the pristine MoTe 2 devices, which significantly reduces to 0.8 kΩ µm after 16 Å K doping. The phase change from 2H to 1T′ of MoTe 2 is generated after K doping, which enhances the current level and decreases the contact resistance of MoTe 2 FET based on 1T′ semimetal phase. [31] To the best of our knowledge, an R c of 0.4 kΩ µm is significantly lower than that of the conventional MoTe 2 ones. [33,55,56] Table 1 summarizes the device performance of different TMD transistors under various treatments. Compared with other FETs, the SB and contact resistance in our MoTe 2 FETs are pretty low, exhibiting excellent application potential in terms of device performance. With such a low R c , K doping is clearly important and holds promise for applications in next-generation electronics.
To investigate the interaction of K atoms in MoTe 2 , density functional theory (DFT; see the "Experimental Section") with the generalized gradient approximation (GGA) was carried out to exhibit the electronic structure of MoTe 2 with and without the K atoms. Structure of pristine MoTe 2 is shown in Figure 5a, where blue (yellow) solid ball represents Mo (Te) atom. The K atoms doping positions are considered by relaxing all atomic positions through the total-energy minimization. The ratio γ = N K /N Te , where N K is the number of K-doping atoms and N Te is Te atoms on the surface, is defined to help better understand the K-doped effect for the electronic properties of MoTe 2 . Figure 5b  www.advelectronicmat.de (see Figure S5 and Table S1 in the Supporting Information). The Fermi level of pristine MoTe 2 lies near the valence band maximum (VBM) and shifts toward the conduction band minimum (CBM) after K atom modification, indicating a significant n-doping effect, which is strongly consistent with the results of K-modified MoTe 2 FET. In addition, the charge distribution of MoTe 2 with the γ ratios of 0.33 and 0.66 is shown in the insets of Figure 5c,d. A significant vertical polarization can be observed after K coating at the γ ratio of 0.33, and the charge distribution becomes chaotic when γ increases to a ratio of 0.66, which is attributed to the induction of K atoms on the structural distortion and orbital hybridization in MoTe 2 . The density difference between the ratios of 0.33 and 0.66 is ascribed to the Mo and Te atoms on the top layer, indicating that the orbital hybridization only occurs on the surface of MoTe 2 .
In addition, we demonstrate a high-performance inverter, which the input voltage (V IN ) is applied to the bottom gate, while three top contacts are used as ground (GND), output voltage (V OUT ), and power supply (V DD ), respectively, as shown in Figure 6a. Figure 6b demonstrates the transfer characteristics of the MoTe 2 FETs with polymethylmethacrylate (PMMA) covering and with the 0.16 Å K doping. The red transfer curve (without K doping, PMMA covering) presents a V min at V g ≈ − 0.05 V. After K doping, the transfer curve (blue one, without PMMA covering) dominates a significant negative shift, indicating the improved electron transport. Thus, a clear intersection in the subthreshold region of these two curves is obtained, leading to the inversion of conductance ratio and the formation of reverse output signal V OUT corresponding to V IN . Figure 6c shows the output curves of the MoTe 2 inverter measurement with the V DD over three steps, 1, 2, and 3 V. During the voltage output measurement, for each V DD applied, V OUT presents a complete sharp transition from approximately V DD to ≈0 V. The inverter gain (defined as dV OUT /dV IN ) plot is extracted from the output curve, which shows the sensitivity of V OUT to continuous changes in the V IN . The device tends to show a higher gain value with the increase of V DD , and reaching up to the highest value of ≈32 at V DD = 3 V, providing new insights and potential of the inverter to enter into complex logic complementation systems. With the comparison and bench marking of MoTe 2 -based inverters, Figure 6d summarizes the previously reported various material inverters, such as MoTe 2 /MoS 2 or MoTe 2 /InGaZnO heterostructures or MoTe 2 with different electron and metal doping modification. A high-performance inverter solely using MoTe 2 by doping modification is formed in our work, offering the possibility of enabling more complex logic functions and circuits.

Conclusion
In conclusion, we demonstrate that the electronic properties of MoTe 2 devices can be significantly modulated by in situ surface modification of K. The charges transferred from K induce a strong n-doping effect to MoTe 2 FET, and the pristine semiconducting MoTe 2 eventually transfers to the metallic transport behavior after 16 Å K doping. The Schottky barrier and the contact resistance are dramatically reduced from 158 to 20 meV and from 11.5 to 0.4 kΩ µm after 16 Å K doping, respectively, indicating the significant effectiveness of K modification in enhancing the contact quality. In addition, a highperformance complementary inverter device with a high gain of ≈32 is achieved. Our work demonstrates an effective method to improve the electrical properties of MoTe 2 , which has considerable promise for use in electronics and optoelectronics in the future.

Experimental Section
Preparation and Device Fabrication: MoTe 2 flakes were first mechanically exfoliated onto scotch tapes from bulk MoTe 2 single crystals (2H), and then the flakes were immediately transferred to the surface of a p-type-doped silicon substrate with 300 nm SiO 2 . PMMA photoresist was spin-coated on the surface of the silicon substrate covered with MoTe 2 . E-beam lithography (EBL) was used to pattern the source and drain electrodes on MoTe 2 , followed by E-beam evaporation to deposit 10 nm Pd and 50 nm Au onto the substrate. Then the as-fabricated sample was put into acetone for about 1 h to remove the photoresist and excess metal.
To fabricate inverter devices, a few-layer graphene sheet was initially isolated onto a p-doped Si wafer with 300 nm SiO 2 . h-BN and MoTe 2 flakes were then mechanically exfoliated on a viscoelastic stamp (polydimethylsiloxane (PDMS)) and sequentially transferred onto the graphene to create the vertically stacked MoTe 2 /h-BN/graphene structure. And then MoTe 2 devices were fabricated by a series of operations including PMMA spin coating, EBL patterning, and E-beam evaporation. A second EBL process was employed on the MoTe 2 channel to pattern the PMMA mask on the desired position. The masked devices were also bonded to a chip carrier before entering the vacuum chamber.
In Situ Device Characterization: All in situ film growth and cryogenic electrical measurements capabilities were performed under a homemade high vacuum system (≈10 −8 mbar). The electrical characteristic was measured by an Agilent 2912A source measure unit, and liquid nitrogen was used for cooling the sample. The potassium source was derived from a SAES Getter and directly deposited to the MoTe 2 devices. In particular, the thickness of K, that is, nominal thickness in this study, was calibrated by a quartz crystal microbalance (QCM: Kurt I. Lesker Co., model FTM-2400).
Other Characterizations and Device Measurements: AFM characterizations were performed on a Bruker Dimension Icon instrument (BRUKERMultiMode 8). Raman spectra were carried out in a Witec Alpha 300 instrument with a 532 nm laser wavelength as the excitation source. The electrical measurements (including temperaturedependent measurement) were conducted by a probe station (METATEST ScanPro Advance) configured with a Keithley-2636B Source Measure Unit parameter analyzer.
Theoretical Calculations: First-principles DFT calculations were determined by the Cambridge Sequential Total Energy Package (CASTEP) [73] code. The exchange-correlation interactions between electrons were carried out within the GGA of Perdew-Burke-Ernzerh (PBE). [74] Grimme's semiempirical DFT-D [75] was introduced in the computations to guarantee a better description of the electron interaction in a long range. The Vanderbilt ultrasoft pseudopotential [76]  www.advelectronicmat.de was used with a cutoff energy of 500 eV. Brillouin zone integration was generated as 3×3×1 according to the Monkhorst-Pack scheme. Geometric convergence tolerances were set for a maximum force of 0.03 eV Å −1 , a maximum energy change of 10 −5 eV atom −1 , a maximum displacement of 0.001 Å, and a maximum stress of 0.5 GPa.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.