Nanotip Contacts for Electric Transport and Field Emission Characterization of Ultrathin MoS2 Flakes

We report a facile approach based on piezoelectric-driven nanotips inside a scanning electron microscope to contact and electrically characterize ultrathin MoS2 (molybdenum disulfide) flakes on a SiO2/Si (silicon dioxide/silicon) substrate. We apply such a method to analyze the electric transport and field emission properties of chemical vapor deposition-synthesized monolayer MoS2, used as the channel of back-gate field effect transistors. We study the effects of the gate-voltage range and sweeping time on the channel current and on its hysteretic behavior. We observe that the conduction of the MoS2 channel is affected by trap states. Moreover, we report a gate-controlled field emission current from the edge part of the MoS2 flake, evidencing a field enhancement factor of approximately 200 and a turn-on field of approximately 40 V/μm at a cathode–anode separation distance of 900 nm.

Mono-and few-layer MoS 2 flakes can be produced either by mechanical exfoliation (from bulk material) [15], or can be synthesized by chemical vapor deposition (CVD) [16]. At present, for larger scale production, high-quality MoS 2 flakes are preferably produced by CVD.
The MoS 2 bulk form consists of a layered structure held by van der Waals interactions [17]. Three atomic layers with the molybdenum plane in between two sulfur planes constitute each monolayer. Bulk MoS 2 presents an indirect bandgap (1.2 eV), which becomes direct and larger, up to 1.9 eV, for monolayers [18][19][20]. The high direct bandgap makes this material suitable for FET applications with an on/off ratio exceeding 10 8 [21] and for selective light absorption in optoelectronic devices with high photoresponsivity [7]. Unfortunately, the carrier mobility for these devices is limited to a few tens of cm 2 V −1 s −1 [22] due to short-and long-range scattering caused by the presence of structural defects, such as S (sulfur) vacancies, or Coulomb traps and surface corrugations related to the substrate [23][24][25].
Due to its sharp edges and high aspect ratio, few-layers MoS 2 is also considered suitable for field emission (FE). Several technological applications require the controlled propagation of electrons 2 of 12 in a vacuum, such as flat-panel displays, microwave amplifiers, electron microscopy, and X-ray sources. Field emission is a quantum mechanical tunneling process by which the electrons are extracted from a surface (metallic or semiconducting) by the application of an external electric field so the electrons can flow in a vacuum from a cathode to an anode. For a flat cathode, FE is enabled by a strong electric field (several kV/µm), while if the cathode surface has sharp edges or protrusions, electrons may be extracted by a considerably lower applied electric field, since the physical geometry provides a field enhancement near the emitting surface. To date, several nanostructures have been investigated as possible field emitters, like metallic nanowires and nanoparticles [26][27][28], semiconducting nanowires and nanoparticles [29][30][31][32][33], nanodiamonds [34], carbon nanostructures [35], carbon nanotubes (CNTs) [36][37][38][39][40][41], and graphene [33,[42][43][44]. Instead, few studies have investigated FE from MoS 2 structures, such as sheets and nanosheets [45,46], nanotubes and nanoflowers [47,48], nanostructures [49], thin films [50], and bilayers [12].
The deposition of metal electrodes on 2D materials by standard electron-beam lithography (EBL) and lift-off processes is an expensive step, in economic and time terms, during the device's fabrication. In this work, we propose a facile method to easily realize metal contacts on MoS 2 flakes that can be used for a first assessment of their suitability for FETs and FE devices. Such a method is based on the use of piezoelectric-driven W (tungsten) tips inside a scanning electron microscope (SEM) chamber, connected to external source-meter units, which are gently approached onto the flake under the combined control of SEM imaging and electrical current monitoring. Moreover, the method can provide a reliable electrical characterization without exposing the flake to process-related damage.
With such an approach, we analyze the electrical transport properties of CVD-synthesized n-type monolayer MoS 2 FETs in a back-gate configuration and their modifications due to gate voltage sweeping range, sweeping time, and electron-beam irradiation. We investigate the effects of the trap centers on the gate hysteresis and we observe that the hysteresis width increases linearly with the gate voltage range and exponentially with the sweeping time. Furthermore, we demonstrate that after exposure to electron-beam irradiation, the n-doping of the channel increases. Considering the n-type doping, we can study the local FE properties from the edge part of the flake. We demonstrate that the gate voltage can modulate the FE current. At the cathode-anode distance of 900 nm, we find a maximum value of the field enhancement factor of 200 and a turn-on field of approximately 40 V/µm at the gate voltage of 40 V. This study demonstrates that it is possible to extensively and reliably characterize 2D material-based devices even without complex nanofabrication by EBL, metal sputtering/evaporation, and lift-off techniques. The proposed method enables a quick selection of flakes before engaging in the laborious production of field effect or field emission devices or the characterization of devices unaffected by process-related damages.
We note that a similar lithography-free method to perform electrical measurements of 2D materials, using a manually actuated microprobe station with carbon fiber microprobes for the direct electrical contact, has previously been proposed [51]. However, the method here demonstrated utilizes thinner probes (approximately 100 nm), fine-controlled by piezo-driven arms, which can be reliably placed with a separation less than 1 µm, thus enabling measurements even on very small flakes. Furthermore, the SEM imaging instead of the optical one allows higher-resolution monitoring of the contact regions.

Materials and Methods
MoS 2 flakes were synthesized by CVD at a temperature of 750 K, using S powder and a saturated ammonium heptamolybdate solution as precursors, on a SiO 2 /Si substrate. To evaluate the number of MoS 2 layers, we performed micro-Raman spectroscopy measurements (using a 532 nm laser source). In Figure 1a, we report the Raman spectrum, evidencing the standard (in-plane) E 1 2g and (out-of-plane) A 1g vibration modes. The frequency separation of the two peaks of approximately 20 cm −1 indicates a monolayer flake [52,53]. The inset in Figure 1a shows a scanning electron microscopy (SEM) image of the flake and of the two sharp piezoelectric-driven tungsten tips (curvature radius approximately 100 nm) used as electrical contacts. Nanomaterials 2020, 10, 106

of 12
The schematic layout of the device and the experimental setup is reported in Figure 1b. The silicon substrate was used as the back gate and metallic tips as the drain and the source electrodes for the FET characterization. For the field emission measurements, we exploited the same setup, using a tip as the cathode and retracting the other tip (anode) from the MoS2 flake at a controlled cathodeanode separation distance.

Transistor Characterization
The electrical characterization by the tip-contact method of a selected MoS2 flake, used as the channel of a back-gated FET, is reported in Figure 2. The source-drain  output characteristics as a function of the gate voltage , shown in Figure 2a, reveal a rectifying behavior, which can be attributed to the formation of asymmetric Schottky barriers between the channel and the contacts, as often occurs when contacting MoS2 [54,55]. Figure 2b shows the  transfer characteristics at the drain bias = −5 V, on a linear (black curve) and logarithmic scale (red curve). The FET reveals n-type behavior with a threshold voltage of ≈ −10.5 V ( corresponding to a = 1 nA), indicating a n-doped channel, as reported in several studies due to the chemisorption of oxygen on MoS2 surface defects and sulfur vacancies [56][57][58]. Electrical measurements were performed inside a SEM chamber (LEO 1530, Zeiss, Oberkochen, Germany) at room temperature and high vacuum (10 −6 Torr), using metallic tips mounted on piezoelectric-driven nanoprobes. Probes were electrically connected to a semiconductor parameter analyzer (Keithley 4200-SCS, semiconductor characterization system, Tektronix Inc., Beaverton, OR, USA), with a current sensitivity of 10 −13 A. The piezoelectric control of the probes allows the fine tuning of the movements with a spatial resolution of about 5 nm and enables a gentle and non-destructive approach on the flakes. The approach is performed with the following procedure. Using SEM imaging, the tips can be positioned above the flake with an incertitude of 50-100 nm. After that, one of the two tips is harshly pushed against the flake (by selecting an edge or a location of no interest) until a movement or a scratch is observed on it. After that, we gradually move the second tip in 5 nm steps and simultaneously monitor the current between the two tips. We systematically observe a sudden rise of the current from the noise floor of 10 −13 A without any apparent damage of the approach point on the flake. Finally, we detach the first tip and repeat the gentle approach of it to a desired location.
The schematic layout of the device and the experimental setup is reported in Figure 1b. The silicon substrate was used as the back gate and metallic tips as the drain and the source electrodes for the FET characterization. For the field emission measurements, we exploited the same setup, using a tip as the cathode and retracting the other tip (anode) from the MoS 2 flake at a controlled cathode-anode separation distance.

Transistor Characterization
The electrical characterization by the tip-contact method of a selected MoS 2 flake, used as the channel of a back-gated FET, is reported in Figure 2. The source-drain I ds − V ds output characteristics as a function of the gate voltage V gs , shown in Figure 2a, reveal a rectifying behavior, which can be attributed to the formation of asymmetric Schottky barriers between the channel and the contacts, as often occurs when contacting MoS 2 [54,55]. Figure 2b shows the I ds − V gs transfer characteristics at the drain bias V ds = −5 V, on a linear (black curve) and logarithmic scale (red curve). The FET reveals n-type behavior with a threshold voltage of V T ≈ −10.5 V (V gs corresponding to a I ds = 1 nA), indicating a n-doped channel, as reported in several studies due to the chemisorption of oxygen on MoS 2 surface defects and sulfur vacancies [56][57][58]. mobility is within the range (0.05 − 70 cm V s ) commonly reported in FETs with a MoS2 channel on SiO2 [22,[59][60][61]. Its relatively low value can be caused partially by the high contact resistance but is mainly an indication of a high density of scattering centers, such as intrinsic defects in the crystal structure of MoS2, extrinsic traps at the MoS2/SiO2 interface or into the SiO2 dielectric layer, and charged impurities such as adsorbates on the MoS2 surface [6]. In Figure 3a, we report the transfer characteristics measured for a gate-voltage sweep between  60 V and 60 V (forth and back), which induces Joule heating on the device that has been in high vacuum (at 10 Torr) for a long time. The combined effect of Joule heating and low pressure reduces/removes adsorbates and makes it possible to analyze the effect of intrinsic defects and traps only. The complete sweeping shows a right shift of the transfer curve, creating a hysteresis that can be explained in terms of negative charge trapping [25,56,62]. Hysteretic behavior can be analyzed by the hysteresis width (i.e., the difference of the gate voltage values corresponding to a channel current = 0.1 nA). Figure 3b,c shows two features of the gate-induced hysteresis as a function of the gate voltage range and the sweeping time, respectively. We observe that linearly increases with the sweeping range while it has exponential dependence on the sweeping time (see inset of Figure 3b,c). The linear behavior of with the sweeping range indicates a trapping process that is proportional to gate voltage and loads the capacitor formed by the MoS2 channel and the Si substrate. The exponential dependence of on the sweeping time [63], (time constant approximately 9 min), reveals a prevalent role of slow trap states related to either MoS2 or SiO2 defects compared to the contribution of MoS2 defects or MoS2/SiO2 interface (fast) traps [64,65]. These results are in agreement with previous experiments on similar devices contacted by Ti/Au (titanium/gold) metal contacts [6,66].
We remark that the backward sweeps of Figure 3b, which are unaffected by the gate-voltage range, well overlap each other, indicating that the proposed contacting method enables highly reproducible measurements. The repeatability of the measurements is further demonstrated by the Its relatively low value can be caused partially by the high contact resistance but is mainly an indication of a high density of scattering centers, such as intrinsic defects in the crystal structure of MoS 2 , extrinsic traps at the MoS 2 /SiO 2 interface or into the SiO 2 dielectric layer, and charged impurities such as adsorbates on the MoS 2 surface [6].
In Figure 3a, we report the transfer characteristics measured for a gate-voltage sweep between 60 V and 60 V (forth and back), which induces Joule heating on the device that has been in high vacuum (at 10 −6 Torr) for a long time. The combined effect of Joule heating and low pressure reduces/removes adsorbates and makes it possible to analyze the effect of intrinsic defects and traps only. The complete sweeping shows a right shift of the transfer curve, creating a hysteresis that can be explained in terms of negative charge trapping [25,56,62]. Hysteretic behavior can be analyzed by the hysteresis width H W (i.e., the difference of the gate voltage values corresponding to a channel current I ds = 0.1 nA). Figure 3b,c shows two features of the gate-induced hysteresis as a function of the gate voltage range and the sweeping time, respectively. We observe that H W linearly increases with the V gs sweeping range while it has exponential dependence on the V gs sweeping time (see inset of Figure 3b,c). The linear behavior of H W with the sweeping range indicates a trapping process that is proportional to gate voltage and loads the capacitor formed by the MoS 2 channel and the Si substrate. The exponential dependence of H W on the sweeping time [63], (time constant approximately 9 min), reveals a prevalent role of slow trap states related to either MoS 2 or SiO 2 defects compared to the contribution of MoS 2 defects or MoS 2 /SiO 2 interface (fast) traps [64,65]. These results are in agreement with previous experiments on similar devices contacted by Ti/Au (titanium/gold) metal contacts [6,66].
imaging) for an exposure time of 51 s and with a constant beam current of 0.2 nA. In Figure 3d, we report the forward and backward transfer characteristics before and after the irradiation. After the irradiation, the curve shows a left shift, which reveals an increased n-type doping of the MoS2. Such doping is due to positive charge accumulation from beam-induced electron-hole pair generation in SiO2 [66,67]. In order to study the properties of the MoS2/W-tip interface and estimate the contact resistance, we varied the distance between the two tips to apply the Transfer Length Method (TLM) [68]. In Figure 4a, we show the  characteristics at the floating back gate as a function of the channel We remark that the backward sweeps of Figure 3b, which are unaffected by the gate-voltage range, well overlap each other, indicating that the proposed contacting method enables highly reproducible measurements. The repeatability of the measurements is further demonstrated by the inset of Figure 3a, which shows two transfer curves measured before and after the two tips were detached and reconnected.
Considering the expression of the SS in terms of the trap (C T ) and channel depletion layer (C DL ) capacitances per unit area: (where k is the Boltzmann constant, T is the temperature, and q is the electron charge) and assuming C DL is negligible with respect to C T (because of the low flake thickness), we estimated a trap state density of D T = C T q 2 ≈ 4.5 × 10 12 eV −1 cm −2 , a value consistent with existing data [6,67]. We also analyzed the effects on the transfer characteristics of electron-beam irradiation. The irradiation was performed at the electron-beam energy of 10 keV (energy typically used for SEM imaging) for an exposure time of 51 s and with a constant beam current of 0.2 nA. In Figure 3d, we report the forward and backward transfer characteristics before and after the irradiation. After the irradiation, the curve shows a left shift, which reveals an increased n-type doping of the MoS 2 . Such doping is due to positive charge accumulation from beam-induced electron-hole pair generation in SiO 2 [66,67].
In order to study the properties of the MoS 2 /W-tip interface and estimate the contact resistance, we varied the distance between the two tips to apply the Transfer Length Method (TLM) [68]. In Figure 4a, we show the I ds − V ds characteristics at the floating back gate as a function of the channel length. Since the characteristics have non-linear behavior, we estimated the total dynamic resistance from the linear fit of the curves in a small range of V ds around the value −5V. The total resistance for the two-probe configuration can be written as the sum of the channel resistance and the contact resistances, which we assumed for a rough estimation to have the same value for the two tips, i.e., R tot = 2R c + R ch . Writing R ch as R s W d, where d is the contacts separation, the previous becomes [69,70]: where R s is the MoS 2 sheet resistance and W the tips diameter (≈ 200 nm).
length. Since the characteristics have non-linear behavior, we estimated the total dynamic resistance from the linear fit of the curves in a small range of around the value −5V. The total resistance for the two-probe configuration can be written as the sum of the channel resistance and the contact resistances, which we assumed for a rough estimation to have the same value for the two tips, i.e., = 2 + . Writing as , where is the contacts separation, the previous becomes [69,70]: where is the MoS2 sheet resistance and the tips diameter (≈ 200 nm). From the linear fit of vs. (see Figure 4b), we extracted the specific area contact resistivity ρ ≈ 4 × 10 Ωcm from the intercept with y-axis and R ≈ 10 Ω/□ from the slope. The obtained value is comparable to that achieved with Au contacts [71], revealing that the W-tips can form good contacts with the MoS2 flake. Instead, the value is about four orders of magnitude higher than the values achieved in other works [69], likely caused by a possible oxidation of the flake for a long exposure to air or by the low quality of the MoS2/SiO2 interface, which can be further optimized.

Field Emission Characterization
The n-type doping and the geometrical shape of the MoS2 flake [72] are excellent prerequisites for FE experiments. The FE measurements were implemented with a W-tip (anode) on the top of the MoS2 flake (cathode) in non-physical contact at a distance of 900 nm and we varied the gate voltage to observe a possible modulation of the FE current. In Figure 5a, we report FE characteristics (current−voltage), on a semi-logarithmic scale, in the voltage bias range from 5 to 120 V at given values. Over the setup sensitivity limit of 10 A, we observed an exponential increase of the current up to seven orders of magnitude as the applied voltage increased. The fluctuations and drops of the FE current can be attributed to the atomic modification of the flake edge, such as oxide or adsorbates removal by joule heating. For this reason, the initial sweep has an electrical conditioning effect and the following sweep results are smoother [39,73].
Remarkably, the FE current measured at = 40 V is significantly higher than the one at = 10 V . An increased gate voltage enhances the n-doping of the flake and favors field emission, similarly to what has been reported in WSe2 vertical field emission transistors [73].
The FE current can be analyzed using the traditional Fowler-Nordheim (FN) theory [74] based on the following equation: where is the work function of the emitter, is the emitting surface, is the local electric field, and and are constants. The local electric field is = / , where is the applied From the linear fit of R tot vs. d (see Figure 4b), we extracted the specific area contact resistivity ρ c ≈ 4 × 10 −2 Ωcm 2 from the intercept with y-axis and R s ≈ 10 8 Ω/ from the slope. The obtained ρ c value is comparable to that achieved with Au contacts [71], revealing that the W-tips can form good contacts with the MoS 2 flake. Instead, the R s value is about four orders of magnitude higher than the values achieved in other works [69], likely caused by a possible oxidation of the flake for a long exposure to air or by the low quality of the MoS 2 /SiO 2 interface, which can be further optimized.

Field Emission Characterization
The n-type doping and the geometrical shape of the MoS 2 flake [72] are excellent prerequisites for FE experiments. The FE measurements were implemented with a W-tip (anode) on the top of the MoS 2 flake (cathode) in non-physical contact at a distance of 900 nm and we varied the gate voltage to observe a possible modulation of the FE current. In Figure 5a, we report FE characteristics (current−voltage), on a semi-logarithmic scale, in the voltage bias range from 5 to 120 V at given V gs values. Over the setup sensitivity limit of 10 −13 A, we observed an exponential increase of the current up to seven orders of magnitude as the applied voltage increased. The fluctuations and drops of the FE current can be attributed to the atomic modification of the flake edge, such as oxide or adsorbates removal by joule heating. For this reason, the initial sweep has an electrical conditioning effect and the following sweep results are smoother [39,73].
Remarkably, the FE current measured at V gs = 40 V is significantly higher than the one at V gs = 10 V. An increased gate voltage enhances the n-doping of the flake and favors field emission, similarly to what has been reported in WSe 2 vertical field emission transistors [73].
The FE current can be analyzed using the traditional Fowler-Nordheim (FN) theory [74] based on the following equation: where Φ is the work function of the emitter, S is the emitting surface, E S is the local electric field, and a and b are constants. The local electric field is E s = β V ds /d, where V ds is the applied potential, d is the cathode-anode separation distance, and β is the so-called field enhancement factor. When Φ is expressed in eV, S in cm 2 , and E s in V/cm, the constants a and b are 1.54 × 10 −6 AV −2 eV and 6.83 × 10 7 Vcm −1 eV −3/2 , respectively. According to FN theory, ln I ds /V 2 ds vs. 1/V ds is a straight line (known as the FN plot), whose slope can be used to estimate β. Figure 5b shows the corresponding FN plots to the curves in Figure 5a and their linearity confirms that the measured currents are governed by FN tunneling. At the back-gate voltage V gs = 40 V, we estimated a value of the turn-on field (here defined as the applied field necessary to extract a current of 10 pA) of E on = 40 Vµm −1 , and, assuming a MoS 2 work function of Φ = 5.15 eV [75], from the linear fitting we can estimate a maximum value of β ≈ 200, consistent with existing data at similar cathode-anode distances [67]. The obtained E on value can be considered a good result compared to what has been obtained with similar measurement setups on other layered materials [12,42,67].
We note that it has been recently suggested that the FN equation should be modified to account for the 2D nature of the emitting material [73,76,77]. A new model with E s , where c is a constant, here referred to as the 2D FN model, has been proposed [77]. The fit of the 2D FN model, shown in Figure 5c, yields a slightly better adjusted R-squared. However, the difference from the Equation (3) model is not significant and, based on the present measurements, we cannot make any sound conclusions in favor of this new model. Nanomaterials 2020, 10, x FOR PEER REVIEW 7 of 11 potential, is the cathode-anode separation distance, and is the so-called field enhancement factor. When is expressed in eV , S in cm , and in V/cm , the constants and are 1.54 × 10 AV eV and 6.83 × 10 Vcm eV / , respectively. According to FN theory, ( / ) vs. 1/V is a straight line (known as the FN plot), whose slope can be used to estimate . Figure 5b shows the corresponding FN plots to the curves in Figure 5a and their linearity confirms that the measured currents are governed by FN tunneling. At the back-gate voltage = 40 V, we estimated a value of the turn-on field (here defined as the applied field necessary to extract a current of 10 pA) of = 40 Vμm , and, assuming a MoS2 work function of = 5.15 eV [75], from the linear fitting we can estimate a maximum value of ≈ 200, consistent with existing data at similar cathode-anode distances [67]. The obtained value can be considered a good result compared to what has been obtained with similar measurement setups on other layered materials [12,42,67].
We note that it has been recently suggested that the FN equation should be modified to account for the 2D nature of the emitting material [73,76,77]. A new model with ∝ − , where c is a constant, here referred to as the 2D FN model, has been proposed [77]. The fit of the 2D FN model, shown in Figure 5c, yields a slightly better adjusted R-squared. However, the difference from the Equation (3) model is not significant and, based on the present measurements, we cannot make any sound conclusions in favor of this new model.

Conclusions
In conclusion, we proposed an easy and fast method to contact 2D flakes by the gentle touch of piezoelectric-driven tungsten tips inside a SEM chamber. We applied this method to a MoS2

Conclusions
In conclusion, we proposed an easy and fast method to contact 2D flakes by the gentle touch of piezoelectric-driven tungsten tips inside a SEM chamber. We applied this method to a MoS 2 monolayer flake that we fully characterized with measurements of parameters such as mobility, hysteresis, electron-beam effect, field emission, and so on.
Despite the facile contacting, we obtained good specific area contact resistivity ρ c ≈ 4 × 10 −2 Ωcm 2 , similar to the case of MoS 2 devices with metal contacts deposited using EBL techniques. We also demonstrated that the direct contacting by metallic nanoprobes allows the complete electrical characterization of the transport properties of 2D FETs. In particular, we reported a complete study of the hysteresis observed in the transfer characteristics. We demonstrated that the hysteresis width has a linear dependence on the gate voltage range and an exponential dependence (with a characteristic time of 9 min) on the sweeping time. Moreover we achieved a significant gate-controlled field emission current from the edge part of the MoS 2 flake under the application of a moderate turn-on field of approximately 40 V/µm with a field enhancement factor of approximately 200.
We have so proved that the use of tips directly on the flake, instead of metal leads, equally allows a rapid as well as a deep and complete investigation of the physical properties of the devices, avoiding an expensive and delicate EBL or optical lithography/lift-off processes.