Experimental methods

MoS₂ FETs were fabricated from mechanically exfoliated bilayer flakes (SPI Supplies, West Chester, PA, USA) in a back-gated geometry on 285 nm-SiO₂/Si chips pre-cleaned with acetone and isopropyl alcohol. The thickness of the samples was confirmed by optical contrast microscopy against cross-calibrated atomic force microscope data. Electrical contacts were fabricated by thermal evaporation of a Ti adhesion layer (5 nm) followed by Au (40 nm) on areas pre-defined by electron beam lithography (EBL), utilizing poly(methyl methacrylate) as resist. Fabricated on-chip MoS₂ devices were modified in a Fischione Instruments 1020 plasma cleaner operating on a 1:3 mixture of O₂:Ar gas at a constant chamber pressure of ~5 mbar. The samples were always placed in the same position in the chamber (to within ±1 mm).

After each plasma exposure, the device was characterized electrically in a two-probe configuration in an in situ cryogenic stage environment (Kammrath & Weiss), cooled by liquid N₂ inside a scanning electron microscope (SEM) chamber maintained at a pressure of 10⁻⁴ mbar. Tungsten nanomanipulator tips (Imina miBot) served as source and drain terminals powered by an Agilent B2912A dual channel source meter. The measurements were taken starting at 83 K and finishing at room temperature. Before each characterization at a new temperature, the system was left to equilibrate so that the temperature variation was <1 K for longer than 5 min. The samples were then removed from the system and exposed to the ambient when transferring to the plasma cleaner for further plasma treatment. High-resolution transmission electron microscopy (HRTEM) was performed in an FEI Titan 80–300 microscope at a beam energy of 300 keV and chamber pressure of ~10⁻⁷ mbar. The samples were first exfoliated onto SiO₂ substrates and transferred onto TEM grids by stamping with poly(vinyl alcohol), followed by dissolution in water. They were then treated in the plasma cleaner while in the TEM sample holder and immediately placed into the microscope for imaging.

Results and discussion

The morphology of the device was revealed by using a scanning electron microscope (SEM). Figure 1(a) shows a false-colored SEM micrograph of a typical bilayer device after EBL processing. The IV curves measured at zero gate bias shown in Figs. 1(b) and 1(c) demonstrate the effect of plasma treatment on the transport properties of the bilayer MoS₂. Before treatment, the device outputs tens of microamps (I _ds) at drain–source biases of V _ds =  ± 1 V, even at cryogenic temperatures. The semiconducting nature of the material is evident from the transport gap observed around the zero-point at low temperatures, and the non-linearity of the sweeps. After 4 s of plasma treatment, the output current has dropped several orders of magnitude (~nA) and subsequent treatments to 6 and 8 s drop the current levels to the noise floor of the instrument (see Supporting Information; Fig. S1). Importantly, for the 4-s treated sample, the output characteristics exhibit a clear linear relationship between the applied bias V _ds and the output current I _ds.

Figure 1.

(a) False-colored SEM micrograph of contacted bilayer MoS₂ device. Scale bar is 1 µm. (b) IV plots of the device before plasma treatment at different temperatures. (c) IV plots of the device after 4 s of plasma exposure. Curves are colored according to the temperature scale bar on the right (in Kelvin). (d)–(g) Gate curves taken at temperatures mapped to color scheme on the scale bar (in Kelvin) at plasma treatment times of 0, 4, 6, and 8 s, respectively. (h) Threshold voltage shift with the temperature at each of the exposure times.

The transfer characteristics of the same device were also tested at different temperatures and are presented in Figs. 1(d)–1(g). The untreated MoS₂ gate sweeps follow standard literature observations, with inversion onset in the negative gate bias region, and a qualitative temperature dependence for the sweeps noted in both monolayers^[Reference Jariwala, Sangwan, Late, Johns, Dravid, Marks, Lauhon and Hersam^10] and bilayers^[Reference Baugher, Churchill, Yang and Jarillo-Herrero^11] of MoS₂. After plasma exposures, the current magnitude drops significantly as exposure time increases while the threshold voltage, V _th, is seen to drastically shift to larger positive gate biases with decreasing temperature. This dependence is charted in Fig. 1(h) for all exposure times, indicating an almost linear increase in V _th after each exposure time below ~155 K. This scaling is in line with the approximate dependence often utilized for MOSFETs^[Reference Filanovsky and Allam^12], and may originate in the temperature-dependent charge carrier density variation in the 2D MoS₂ channel.

Around 200 K a clear gap, ΔV _th, develops in the gate sweeps, separating the blue (low temperatures) curves from the red (high temperatures) ones along the gate bias axis. At the same time, the treated devices are seen to deviate from standard MoS₂ FET transfer characteristics. The different regions of the transfer curve cannot now be easily distinguished. With increasing plasma treatment time, the shape of the curves indicates that the channel cannot be strongly inverted in the tested gate bias range. This is especially evident at low temperatures. As this shift is also seen in the pristine sample, we do not attribute the origin of ΔV _th to the plasma exposure process. It is, however, exacerbated with continuing plasma treatment and correlates with the increasing concentration of surface defects. The exact nature of this ΔV _th observation in layered 2D semiconductors necessitates further investigation in the future; though it is known that the physical origin of V _th is closely linked to the concentration of defects on the surface of 2D FETs^[Reference Shu, Wu, Guo, Liu, Wei and Chen^13].

We track the conductance drop in the sample with increasing plasma exposure time at different carrier densities in the high- and low-temperature regimes in Figs. 2(a) and 2(b). From the semi-logarithmic plots, we find good linear fits confirming an exponential drop in conductance with increasing plasma exposure time. The layer number-normalized exponential drop in electron mobility over time, μ(t), in oxygen plasma-treated MoS₂ has previously been reported to scale as^[Reference Khondaker and Islam^14]:

$$\displaystyle{{\mu (t)} \over L}\; \propto {\rm e}^{ - \varphi t/L}, $$

where L is the MoS₂ layer number, φ = μ′/μ and μ′ is the plasma-time derivative of the mobility. Following this convention, we can extract the plasma-time rate of change of the conductance, G, in our bilayer sample, with φ _G = 2.02 s⁻¹. The contribution of MoO₃ species formed by the oxygen-containing plasma may lead to the formation of an effective semiconducting medium in the FET channel. This, in turn, may give rise to the exponential dependence of current tunneling through the newly-formed heterostructure-like interface^[Reference Khondaker and Islam^14, Reference Choudhary, Islam, Kang, Tetard, Jung and Khondaker^15].

Figure 2.

(a) Conductance, G, plotted on semi-log plot as a function of plasma treatment time at a carrier density of the order of 10¹¹ cm⁻². The solid lines are linear fits at high (red) and low (blue) temperatures. (b) Similar plot from data extracted at a carrier density of the order of 10¹² cm⁻². Both sets of data fit the exponential plasma time decay model (equation included on the plot), regardless of carrier density. (c) HRTEM image of MoS₂ flake before plasma treatment. (d) HRTEM image of MoS₂ flake following plasma treatment of 6 s. Noticeable amorphization occurs on the surface of the material due to the formation of molybdenum oxides.

We note that at this rate of plasma exposure, the bilayer MoS₂ device becomes substantially damaged after 6 s of treatment, as field effect mobility values at 0 V gate bias drop below 1 cm²/(Vs) and output current at no gate bias is below the nA range (see Supporting Information; Fig. S1). Figures 2(c) and (d) are high-resolution transmission electron micrographs of the same few-layer MoS₂ sample before and after 6 s of plasma treatment, respectively. Plasma-induced damage to the surface of MoS₂ is evident from the contrast change after 6 s of treatment. The amorphous oxide which forms on the surface through a sulfur-replacing reaction creates severe defects in the material^[Reference Cui, Brown and McKinley^16]. It can then act as a source of short-range scattering which reduces the device conductance^[Reference Khondaker and Islam^14, Reference Islam, Kang, Bhanu, Paudel, Erementchouk, Tetard, Leuenberger and Khondaker^17].

We go on to analyze the temperature-dependent emission current in the bilayer MoS₂ system as it undergoes modification by the O₂:Ar plasma. Charge injection from the metal electrode into the 2D semiconductor (at moderate doping levels) is mostly governed by thermionic emission over the Schottky barrier at sufficient temperatures. Furthermore, Fermi level pinning occurs near the bottom of the conduction band of the MoS₂, independent of metal work function^[Reference Kim, Moon, Lee, Choi, Ahmed, Nam, Cho, Shin, Park and Yoo^18]. When considering current injection for this 2D system, the Richardson equation is modified as follows:

$$I_{{\rm ds}} = A_{2{\rm D}}^* T^{3/2} \exp \left( {\displaystyle{{q\Phi _{\rm B}} \over {k_{\rm B} T}}} \right)\left[ {1 - {\rm exp}\left( {\displaystyle{{qV_{{\rm ds}}} \over {k_{\rm B} T}}} \right)} \right],$$

where $A_{2{\rm D}}^* $ is the Richardson constant for a 2D system, T is the temperature, q is the Coulombic charge, k _B is the Boltzmann constant, Φ_B is the effective SBH and V _ds is the drain–source bias. We note that effective Φ_B can thus be extracted from the slope of the ln[I _ds/T ^3/2] versus 1000/T curve.

A series of such plots is presented in Figs. 3(a)–(d) as the device progresses through increasing plasma treatment. The data are color-mapped to the scale bar above to denote the gate bias at which the current was extracted. The 0 s data fit well to the linear model below 233 K, while at higher temperatures and gate biases between ~−50 and ~0 V a different slope is extracted in the regime corresponding to thermionic emission only. When the device is exposed to the plasma for 4 s and the amorphous oxidized regions start to form on the surface, a significant break in the temperature dependencies occurs beyond ~150 K at higher gate biases. In this region, both the tunneling and thermionic components begin to contribute to transport between the metal and the effective semiconducting medium of oxidized MoS₂. Accompanied by decreased emission current even at higher gate biases, this break in the trend persists in the same temperature range after 6 and 8 s.

Figure 3.

(a)–(d) Richardson plots for the MoS₂ device as it progresses through plasma treatment. The curves are color-mapped to the extraction gate bias scale on the color bar above. (e)–(h) Plots of the effective SBHs (Φ_B) extracted from the corresponding above Richardson plots across the whole gate bias range. The values were obtained in the low- and high-temperature ranges labeled on each plot. The end of the linear dependencies on each respective plot marks the SBH extracted at the flat-band voltage.

Figures 3(e)–(h) present effective SBHs at each gate bias derived in the high (red) and low (blue) temperature regimes as the reaction with plasma progress. Following the convention of Kim et al.^[Reference Kim, Moon, Lee, Choi, Ahmed, Nam, Cho, Shin, Park and Yoo^18], we extract the true barrier heights at the flat-band voltage (V _fb) for each plasma time. At high temperatures, one can accurately extract the true SBH by considering the end of the linear region of the Φ_B dependence on gate bias. For biases below the flat-band voltage, the contribution of tunneling to the injection current should be negligible, ensuring only thermionic emission over the true barrier height. At low temperatures, the tunneling component will be more significant, and so true SBH values cannot be readily extracted for a semiconductor with a significant band gap such as MoS₂^[Reference Kim, Moon, Lee, Choi, Ahmed, Nam, Cho, Shin, Park and Yoo^18, Reference Wang, Yao, Huang, Zou, Liao, Chen, Fan, Zhang, Wu and Xiao^19]. With increasing plasma exposure time, a significant shift of V _fb occurs, naturally correlating with the positive V _th shift recorded in the gate sweeps in Figs. 1(d)–1(g). At the same time, Schottky barrier lowering is experienced by the MoS₂/Ti–Au interface with increasing plasma treatment. Altering of the SBH for MoS₂/metal interfaces has previously been performed with pure O₂ plasma, allowing for additional hole injection by effective depinning of the Fermi level^[Reference Giannazzo, Fisichella, Greco, Di Franco, Deretzis, La Magna, Bongiorno, Nicotra, Spinella, Scopelliti, Pignataro, Agnello and Roccaforte^8]. We do not observe ambipolar FET behavior in our devices across the tested sweep range of ±60 V, indicating that the argon-induced formation of sulfur vacancies is the dominant doping mechanism, and a much larger negative back gate voltage would be required to enter the hole branch on the transfer curves.

We investigate the temperature-dependent conductance mechanism for electrons in defective plasma-treated MoS₂ in Fig. 4. By extracting the parameter, W = −∂ln(ρ)/∂ln(T), where ρ is the device resistivity and T is the measurement temperature, the conduction mechanism can be determined from the slope of the plot of ln(W) versus ln(T), due to the exponential dependence of conductivity on temperature and device dimensionality^[Reference Yildiz, Serin, Serin and Kasap^20–Reference Kim, Kim, Zhao, Kim, Lee, Jin, Choi, Moon, Bae, Lee and Lim^22]. These plots are presented for the various plasma treatment times in Fig. 4. We note that negative slope values in these plots indicate semiconducting behavior for the sample after all treatment times. A good linear fit is found for the untreated sample with a slope indicating a conductance exponent of a = 0.48, i.e., conductance scales as T ^−0.48. Mott-like variable range hopping^[Reference Jariwala, Sangwan, Late, Johns, Dravid, Marks, Lauhon and Hersam^10, Reference Qiu, Xu, Wang, Ren, Nan, Ni, Chen, Yuan, Miao, Song, Long, Shi, Sun, Wang and Wang^23]_, which scales as T ^−1/3, has been considered the main conductance mechanism at low temperatures in semiconducting MoS₂. However, mixed transport^[Reference Kim, Kim, Zhao, Kim, Lee, Jin, Choi, Moon, Bae, Lee and Lim^22, Reference Stanford, Pudasaini, Gallmeier, Cross, Liang, Oyedele, Duscher, Mahjouri-Samani, Wang, Xiao, Geohegan, Belianinov, Sumpter and Rack^24, Reference Matis, Garces, Cleveland, Houston and Baldwin^25] regimes have been reported recently in polymorphic, defective, and encapsulated layered transition metal dichalcogenides, respectively, at various temperatures. The dependence we find for our mechanically-exfoliated bilayer is in line with ES hopping (T ^−1/2) which has recently been observed in polymorphic low-dimensional MoS₂^[Reference Papadopoulos, Steele and van der Zant^21]. This may be indicative of a highly sulfur vacancy-rich sample, where electron–electron interactions reduce the density of states near the Fermi level according to the ES model^[Reference Efros and Shklovskii^26].

Figure 4.

Plot of ln(W) as a function of ln(T) used to extract the conduction mechanism exponent a. The colored points are the experimental data for each exposure time, the green lines are linear fits to the data close to the ES regime, black lines are fits close to the T ⁻¹ regime. The extracted slopes are tabularized in Table 1.

As the sample is exposed to the plasma, a change in the conduction regime is observed, where the critical exponent a is now >0.5 at high temperatures (see Table 1). This indicates a mixing of the ES and nearest-neighbor hopping regimes. Below a critical temperature T* ≈ 181 K, the conduction mechanisms are fitted well with T ^−1.3, T ^−1.1, and T ^−1.1 dependencies for 4, 6, and 8 s, respectively. This scaling at low temperatures follows a quasi-Arrhenius relation, with an increased activation energy likely due to the presence of plasma-generated defective regions. A metal–insulator transition^[Reference Radisavljevic and Kis^27] due to electron–electron interactions is expected to occur at high carrier densities (n _2D ~ 10¹³ cm⁻²). We were not able to reach this order of n _2D in this experiment, as a top gate could not be fabricated without losing access to sample surface. We note here that although the conductivities were extracted at the same n _2D past the flat-band voltage, the break in the conductance regimes and the change in the SBH could not be decoupled for this analysis in the two-probe measurement scheme. The values of the exponent a for each temperature regime and plasma time are summarized in Table 1.

Table I.

Summary of extracted a exponents in each temperature regime for every plasma treatment time.

This type of conduction mechanism shift has previously been studied in hydrogenated amorphous Si films, where a reverse dependence was found for the conduction regimes^[Reference Yildiz, Serin, Serin and Kasap^20]. We note that our case cannot be treated as a simple amorphization of the whole material, as the reaction with the plasma proceeds from the surface only. The underneath MoS₂ layer in the bilayer system may not inherit as many defect centers, preserving a degree of crystallinity for the device and contributing to the unusual mixed transport regime. From Mathiessen's rule, it is known that the residual resistivity in semiconducting MoS₂ at low temperatures is due to the impurity scattering term, (μ _imp)⁻¹, which comprises both the long-range and short-range Coulombic scattering components^[Reference Cui, Lee, Kim, Arefe, Huang, Lee, Chenet, Zhang, Wang, Ye, Pizzocchero, Jessen, Watanabe, Taniguchi, Muller, Low, Kim and Hone^28]. In encapsulated and pristine MoS₂ FETs, this has been shown to dominate carrier transport below 100 K^[Reference Baugher, Churchill, Yang and Jarillo-Herrero^11, Reference Radisavljevic and Kis^27, Reference Cui, Lee, Kim, Arefe, Huang, Lee, Chenet, Zhang, Wang, Ye, Pizzocchero, Jessen, Watanabe, Taniguchi, Muller, Low, Kim and Hone^28]. In the presently-discussed case of plasma-modified 2D MoS₂, the impurity concentration may be sufficient to allow for dominant contribution of disorder-born scattering at temperatures as high as 181 K. This may have its origin in the increased presence of amorphous molybdenum oxides on the surface of the flake, as well as sulfur vacancy traps originating from plasma-etched regions^[Reference Jadwiszczak, O'Callaghan, Zhou, Fox, Weitz, Keane, Cullen, O'Reilly, Downing, Shmeliov, Maguire, Gough, McGuinness, Ferreira, Bradley, Boland, Duesberg, Nicolosi and Zhang^9]. These mid-gap trap states reduce output current at zero gate bias, and shift V _th to increasingly positive biases as plasma exposure progresses in time. A larger gate voltage is required to tune the Fermi level past the trap states and into the conduction band. This effective band gap increase has been reported previously with photoluminescence measurements in rapidly plasma-oxidized MoS₂^[Reference Ko, Jeong, Kim, Lee, Kim, Lee, Ryu, Park, Kim, Lee, Lee, Lee and Ryu^29, Reference Dhall, Neupane, Wickramaratne, Mecklenburg, Li, Moore, Lake and Cronin^30].