Hysteresis of Transfer Characteristics in Field-Effect Transistors with a Molybdenum Disulfide Channel

Recent studies on singleand multilayer molybdenum disulfide (MoS2) devices have revealed their promising characteristics as novel semiconductor devices. Here, we report the effects of environmental gases on the hysteresis in the transfer characteristics and observation of an anomalously large hysteresis below 1K for back-gated multilayered MoS2 field-effect transistors. Comparisons between different gases (oxygen, nitrogen, air, and nitrogen with varying relative humidities) revealed that water molecules acting as charge-trapping (dominantly hole-trapping) centers are the main cause of hysteresis. While the hysteresis persisted even after pumping out the environmental gas for longer than 24 h at room temperature, it disappeared when the device was cooled to 240 K, suggesting a considerable increase in the time constant of the charge trapping/detrapping at these modestly low temperatures. Below 1 K, we observed for the first time an anomalously large hysteresis, which is not attributed to charge trapping. We hypothesize that this hysteresis results from the slow injection of electrons via quantum tunneling through the Schottky barrier at the contacts. The size of the hysteresis increased with increase in the scan rate of the gate voltage, which is consistent with the possibility of very slow injection of electrons.


M
olybdenum disulfide (MoS 2 ) has attracted much attention as a novel two-dimensional material for nanoscale electronic and optical devices [1][2][3][4].To explore the intrinsic properties and enhance the performance of MoS 2 -based field-effect transistors (FETs), thorough understanding of extrinsic effects such as environmental gas [5][6][7][8] and contact resistance of the electrodes [9][10][11][12] is required.A notable difference between MoS 2 FETs and conventional Si FETs is the significance of the Schottky barrier (SB) at the source and drain contacts.Significant contact resistance is associated with the SB.In this study, we focused on the hysteresis in the transfer characteristics of the MoS 2 -based FETs.Understanding of the mechanisms behind the hysteresis is crucial for device applications because the hysteresis is associated with fluctuations in the transport properties.We investigated the influence of various environmental gases, such as nitrogen, oxygen, and water vapor, on the hysteresis [13].An anomalously large hysteresis in the transfer characteristics observed at very low temperatures (below 1 K) [14] is also reported in this paper.We hypothesize that this hysteresis results from the slow injection of electrons via quantum tunneling through the SB at the contacts.

ENVIRONMENTAL EFFECTS ON HYSTERE-SIS IN THE TRANSFER CHARACTERISTICS
Thin MoS 2 flakes were exfoliated from a bulk crystal (SPI Supplies) using adhesive tape.The flakes on the tape were then transferred onto a gel sheet [15].Subsequently, the flakes on the gel sheet were deposited on a Si substrate with a 270-nm-thick SiO 2 layer.After locating a suitable MoS 2 flake using an optical microscope, the source and drain electrodes made of Ti(12 nm)/Au(75 nm) were fabricated using photolithography and electron-beam deposition.The highly n-doped Si substrate was used as a back gate.An atomic force microscope was used to measure the thicknesses of the MoS 2 flakes, and the transfer and output characteristics of the devices were measured.Figure 1(a) shows an optical image of the back-gated FET device with a 6-nm-thick MoS 2 flake.The channel length and width of the sample are L = 4.3 mm and W = 4.1 mm, respectively.The devices exhibited ohmic I ds -V ds behavior at room temperature, where I ds and V ds are the drain-source current and drain-source voltage, respectively.
Figure 1(b) shows the transfer characteristics (I ds -V g curves, where V g is the gate voltage) in vacuum and air at varying pressures.The increase in the hysteresis with increasing pressure indicates that the hysteresis is attributed to the adsorbed molecules in air that act as charge-trapping centers.In Figure 1(c), the transfer characteristics in different environmental gases are compared, such as dry nitrogen, dry oxygen, air (relative humidity RH = 18%), and humid nitrogen (RH=80%).The total pressure of the gases was approximately 740 Torr.The magnitudes of the hysteresis were compared in terms of the difference in the threshold voltages V th1 and V th2 in different sweep directions, where V th1 and V th2 are defined as the voltages at which the drain-source current is I ds =1 nA. Figure 1(c) indicates that the molecules, in decreasing order of con-doi: 10.11605/j.pnrs.201701008tribution to the hysteresis, are water, oxygen, and nitrogen.It is remarkable that the influence of water (RH = 80%) on the hysteresis is more than thrice that of dry oxygen.
We note that the environmental effect on the hysteresis was reversible, that is, the hysteresis decreased again when the environmental gas was pumped out for a few minutes without annealing the devices.We measured the transfer characteristics while the humid nitrogen was evacuated using a turbo-molecular pump.A gradual decrease in the hysteresis was observed for a period longer than 10 h.A small residual hysteresis remained even after pumping for > 24 h.This result indicates the slow desorption of water molecules in vacuum at room temperature, which is ascribed to the strong hydrogen bonds between the water molecules and sulfur surface of MoS 2 that has a considerable polarity and hydrophilicity [5], which is in clear contrast to the hydrophobicity of graphene and nanotubes [16].We note that the calculated adsorption energies of water and oxygen molecules on the monolayer MoS 2 are considerably higher than the thermal energy at room temperature [17,18].
To analyze the effect of water molecules on the hysteresis, the threshold voltages (V th1 and V th2 ) and DV th = V th2 -V th1 are shown in Figure 1(d).With an increase in the humidity, V th2 is nearly constant; however, V th1 decreases significantly.Therefore, the increasing hysteresis is caused by the variation in V th1 .Because the decrease in V th1 is caused by hole trapping in the forward sweep, this result indicates that the water molecules dominantly act as hole-trapping centers.This is also implied from the transfer curves shown in Figure 1(c).A comparison between the transfer curves in air (RH = 18%) and humid nitrogen (RH=80%) reveals that the curve swept in the positive direction (off-to-on sweep) considerably shifts to the negative V g direction with an increase in the RH, while the curves swept in the negative direction (on-to-off sweep) nearly coincide with each other.While it has been indicated that water and oxygen molecules act as charge acceptors in equilibrium [17,18], the slow dynamics of the trapped charges on these molecules under varying gate voltage, which should explain the mechanism of the hysteresis, is still to be investigated.
We observed a remarkable temperature dependence of the hysteresis, as shown in Figure 2. Cooling the sample below 273 K drastically reduced the hysteresis.At 240 K, the hysteresis ascribed to the charge trapping was completely suppressed.Similar temperature dependence was observed for the , and DV th = V th2 -V th1 (filled circles) as functions of relative humidity.These data were obtained using humid nitrogen at 760 Torr as the environmental gas.The flattening of DV th with decreasing humidity is explained in terms of the limited measuring range of the humidity sensor.The humidity dependence of V th1 and V th2 indicates the dominance of the hole-trapping effects of water molecules in the hysteresis of the transfer characteristics [13].doi: 10.11605/j.pnrs.201701008multiple devices we fabricated.The disappearance of the hysteresis at low temperatures is attributed to the prolonged time constant for charge trapping/detrapping.The suppression of the hysteresis or instability in the easily attainable temperature range without surface passivation is highly advantageous for the device application of this system.

ANOMALOUS HYSTERESIS BELOW 1 K
For MoS 2 FETs with Au contacts, ohmic I d -V d curves are commonly observed at room temperature.This does not imply the absence of a Schottky barrier at the contact [10].The main carrier injection mechanisms at the Schottky barrier for the Au contact are known to be thermionic emission and thermally assisted tunneling of electrons [10,19,20].In the on state, thermally assisted tunneling current is dominant, which is responsible for the linear I d -V d curve.With decreasing temperature, the interface resistance increases exponentially.In the past, MoS 2 devices were studied down to ~2 K [21,22].In these studies at low temperatures, superconductivity with a transition temperature of ~10 K was observed.Here, we report the transport properties of MoS 2 FETs in a broader temperature range, from room temperature to 30 mK [14].Measurements at such low temperatures might enable the observation of novel and interesting phenomena caused by quantum effects; such observation is hindered by thermal effects at high temperatures.
The contact metal of the MoS 2 FETs, which was cooled down to 30 mK, is Cr (2 nm)/Au (50 nm).Cr is the adhesion layer between Au and the Si substrate.The MoS 2 flakes in the samples consisted of approximately 100 layers (thickness ≅ 70 nm).The samples were enclosed in an evacuated copper shield box that was thermally anchored to the mixing chamber of a dilution refrigerator.
Figure 3 shows the I ds -V g characteristics for V ds = 5 V at various temperatures.From these data, the field effect mobility was estimated to be 30-40 cm 2 V −1 s −1 .The temperature dependence of the field effect mobility was found to be small for the temperature range considered in this study.Note that the mobility estimated here is not the intrinsic mobility of MoS 2 , because the contribution of the contact resistance to the estimation must be significant.The threshold voltage in the I ds -V g curve increases with decreasing temperature, because of the decrease of thermally excited carriers in the channel.The current on/off ratio observed at 1 K was greater than 10 7 , and it decreased with an increase in temperature because of the increasing leak current that contributed to the measured I ds .
The inset of Figure 1 shows the I ds -V ds characteristics at various temperatures for V g = 10 V, where the FET is deep in the on state for large V ds .The I ds -V ds curve is nearly linear at room temperature, while nonlinearity becomes significant below 80 K.At temperatures below 4.2 K, I ds flows for V ds above the threshold voltage (~1 V).The temperature dependence of the I ds -V ds curve is not significant when the temperature is below 4.2 K. Irrespective of the temperature, the I ds -V ds curve is nearly symmetrical.The observed nonlinearity at low temperatures can be attributed to the SB between MoS 2 and Cr/Au.At very low temperatures (< 4.2 K), the current flow is nearly linear with respect to V ds above the threshold voltage because of direct tunneling through the SB at the source contact.With an increase in temperature, thermally assisted tunneling becomes significant, and eventually the I ds -V ds curve becomes linear at room temperature.Note that the I ds -V g curves shown in Figure 3 were obtained at V ds = 5 V, which is well above the threshold voltage of V ds .The hysteresis in the I ds -V g curves is very small.Because the sample is in vacuum in this measurement, the absence of hysteresis is consistent with the results presented in the last section.
At 30 mK, a remarkably large hysteresis was observed, as shown in Figure 4.In the upward scan of V g , the flow of I ds starts abruptly when V g exceeds a threshold voltage.However, in the downward scan of V g , I ds decreases continuously to zero  The inset shows I ds -V ds curves for V g = 10 V, where significant nonlinearity due to the Schottky barrier is observed, except for the curve corresponding to 290 K.The hysteresis is very small for these I ds -V g and I ds -V ds curves.The thickness of the MoS 2 flake of this device was measured to be ~70 nm [14].
(red line in Figure 4).We observed a considerable amount of scattering of the threshold voltage among repeated scans, e.g., the threshold voltage changed from 8.3 to 3.6 V in two consecutive scans.In spite of the scattering of the threshold voltage, the hysteresis as seen in Figure 4 was always observed in the repeated scans of V g .
To gain further insight into the nature of the hysteresis, we investigated its dependence on the scan rate of V g .Figure 5 shows scan-rate dependence of the I ds -V g curve measured at 1 K.The sample used for this result was different from that for Figs. 3 and 4. For this sample, we observed an abrupt jump in I ds at 30 mK, similar to the result shown in Figure 4.The hysteresis was still maintained at 1K, but the jump in I ds changed to a steep increase above the threshold value of V g , as shown in Figure 5.The threshold voltage did not display any scattering between repeated scans at 1 K.When the scan rate was varied from 0.24 to 4.8 V/s at 1 K, we found that the threshold voltage, and thus the size of the hysteresis, increased with scan rate.This implies that the hysteresis at 1 K is associated with a stochastic process with a time constant of the order of 1 s.At temperatures below 1K, the time constant associated with the stochastic process, which leads to the scattering of the threshold voltage, may be greater than 1 s.This conjecture is supported by the following observation.At 30 mK, V g was abruptly changed from −10 to 10 V, while V ds was maintained at 5V.Subsequently, I ds began to flow after approximately 20 s.This delay showed scattering among repeated attempts.
We now discuss the probable origin of the hysteresis as shown in Figure 4. First, we note that when V ds and V g are above threshold voltages, the system remains in the on state (i.e., a significant amount of I ds flows) at low temperatures (< 4.2 K).This indicates that under these bias conditions, direct tunneling effectively injects electrons from the source electrode at a sufficiently high rate.At such low temperatures, thermally assisted tunneling is not effective.The SB at the drain contact does not impede the electron flow, because it is sufficiently forward biased.The inset of Figure 4 shows the conduction band profile of the system.The tunneling probability through the source SB strongly depends on the barrier width, which is assumed to decrease with increasing carrier density in MoS 2 under the metal electrode [20].Therefore, for the system to be in the on state, the carrier density under the metal electrode must be above a threshold value, thereby increasing the tunneling probability.When V g is increased from −10 V, the carrier density is initially very low because of the very low tunneling probability at the source contact.When V g reaches ~5 V at a scan rate faster than the time constant of the slow carrier injection, the system has still not reached the on state.After a certain number of electrons are injected in the slow tunneling process, direct tunneling begins to dominate because of the increased tunneling probability, and thus, I ds flows abruptly.Furthermore, in the downward scan of V g from 10 V, the carrier density under the electrodes is sufficiently high at the beginning.This enables direct tunneling to be effective for the system to be in the on state because of the high tunneling probability.With decreasing V g , the sheet resistance of MoS 2 and the contact resistance associated with the SB at the source contact increase continuously.This explains the smooth decrease in I ds with decreasing V g .
Investigation of the low-temperature characteristics of MoS 2 FETs with different contact metals is of interest because SB heights should depend on the metals in contact with MoS 2 .It is likely that the appearance of the hysteresis at low temperatures will depend on the type of metal contact used.

CONCLUSIONS
We investigated the influence of various environmental gases, such as nitrogen, oxygen, and water vapor, on the hysteresis of the transfer characteristics of the multilayer MoS 2 -based Figure 4.I ds -V g curves at 30 mK for V ds = 3 V with a scan rate of 0.59 V/s.An abrupt jump in I ds is observed in the upward scan of V g .The threshold voltage for the jump in I ds is scattered among repeated scans.In the downward scan, I ds decreases continuously.This anomalous hysteresis can be explained by quantum tunneling through the Schottky barrier.The inset shows the profile of the conduction band minimum together with the Fermi energies in the source and drain electrodes when V g and V ds are above their threshold values.
Figure 5. Scan-rate dependence of the I ds -V g curve measured at 1 K with V ds kept at 5 V.The sample used for this result was different from that for Figs. 3 and 4. The size of the hysteresis increases with an increase in the scan rate, suggesting that the hysteresis is related to a slow process with a time constant of ~1 s.The curve in the downward scan, which is denoted by the dashed line, did not show significant dependence on the scan rate.doi: 10.11605/j.pnrs.201701008 FETs.Water molecules are most influential on the magnitude of the hysteresis, while oxygen is less so.While the hysteresis persisted after evacuating the device at room temperature for longer than 24 h, it was completely suppressed by cooling down to ~240 K.This suppression is attributed to the temperature dependence of the time constant for charge trapping/detrapping.The suppression of the hysteresis by modest cooling without protective passivation or encapsulation is very promising for device applications by reducing the instability in MoS 2 devices caused by the environmental effects.The variation of the transfer curves at different levels of humidity indicates that the water molecules dominantly act as hole-trapping centers.These results provide a solid basis for understanding the environmental effects in the MoS 2 -based FETs.The applicability of this system to sensors for humidity and oxygen is also suggested.
The transport properties of MoS 2 FETs with Cr/Au contacts were measured from room temperature to 30 mK.The I ds -V ds curve exhibited significant nonlinearity below 80 K due to the SBs at the source and drain contacts.High on/off ratios (up to 10 7 at 1 K) were observed at all temperatures.Below 1K, an anomalously large hysteresis was observed in the I ds -V g curve, which is explained on the basis of slow injection of electrons via quantum tunneling through the source SB when the carrier density below the electrode is as low as that in the off state.The size of the hysteresis was found to increase with increase in the scan rate of V g .This observation supports the possible existence of a slow charge injection process.This novel finding might pave the way for an investigation into the quantum dynamics of electrons penetrating the SB.

Figure 1 .
Figure 1.(a) Optical image of an FET with a multilayer MoS 2 channel (thickness ~6 nm) on a SiO 2 /Si substrate.The transport properties between the source and drain contacts were measured as functions of the gate voltage.(b) I ds -V g curves for V ds = 0.1 V in vacuum and air with varying values of pressure.The scan rate for the measurement of the I ds -V g curves was approximately 1.0 V/s.The hysteresis increases with increasing air pressure.The sweep directions are indicated by the arrows.(c) I ds -V g curves for V ds = 0.1 V measured under different environmental conditions.The influence of nitrogen, oxygen, air (RH = 18%), and humid nitrogen (RH = 80%) are compared.Small temperature variation and environmental gas change the on-state current (I ds at V g = 40 V).Therefore, for the sake of enhancing the visibility of the hysteresis, some of the curves are vertically scaled such that the on-state currents shown in the figure nearly coincide with each other.The sweep directions are indicated by the arrows.(d) The threshold gate voltages V th1 , V th2 (open triangles), and DV th = V th2 -V th1 (filled circles) as functions of relative humidity.These data were obtained using humid nitrogen at 760 Torr as the environmental gas.The flattening of DV th with decreasing humidity is explained in terms of the limited measuring range of the humidity sensor.The humidity dependence of V th1 and V th2 indicates the dominance of the hole-trapping effects of water molecules in the hysteresis of the transfer characteristics[13].

Figure 2 .
Figure2.I ds -V g curves measured at various temperatures.In this measurement, V ds was maintained at 1 V.The sample was cooled in helium gas for heat exchange.At 240 K, the hysteresis due to the charge trapping related to the adsorbed molecules completely disappears.The curves are offset vertically for clarity.

Figure 3 .
Figure3.I ds -V g curves at various temperatures.V ds was kept at 5 V. n-channel behavior of the FET is clearly shown.The inset shows I ds -V ds curves for V g = 10 V, where significant nonlinearity due to the Schottky barrier is observed, except for the curve corresponding to 290 K.The hysteresis is very small for these I ds -V g and I ds -V ds curves.The thickness of the MoS 2 flake of this device was measured to be ~70 nm[14].