Temperature-Dependent Electrical Properties of Al₂O₃-Passivated Multilayer MoS₂ Thin-Film Transistors

Abstract

It is becoming more important for electronic devices to operate stably and reproducibly under harsh environments, such as extremely low and/or high temperatures, for robust and practical applications. Here, we report on the effects of atomic-layer-deposited (ALD) aluminum oxide (Al₂O₃) passivation on multilayer molybdenum disulfide (MoS₂) thin-film transistors (TFTs) and their temperature-dependent electrical properties, especially at a high temperature range from 293 K to 380 K. With the aid of ultraviolet-ozone treatment, an Al₂O₃ layer was uniformly applied to cover the entire surface of MoS₂ TFTs. Our Al₂O₃-passivated MoS₂ TFTs exhibited not only a dramatic reduction of hysteresis but also enhancement of current in output characteristics. In addition, we investigated the temperature-dependent behaviors of the TFT performance, including intrinsic carrier mobility based on the Y-function method.

1. Introduction

With the discovery of graphene, two-dimensional (2D) layered materials have been drawing intense research interest. Despite its excellent mechanical, optical, and electrical properties, however, graphene is considered unsuitable as an active component of field effect transistors due to its lack of a band gap [1,2,3]. In efforts to overcome this limitation, i.e., to produce a band gap in graphene, a great deal of research has been carried out, but the results thus far have only added to the complexity of the processes and reduced mobility [4,5].

In this context, layered 2D transition metal dichalcogenides (TMDs)-based thin-film transistors (TFTs) are drawing considerable attention as promising candidates to lead the next-generation transistor technology. Among the 2D TMDs, molybdenum disulfide (MoS₂) is at the center of attention, and relevant research is underway [6,7,8]. MoS₂ TFTs have excellent carrier mobility, a high on- and off-current ratio (I_on/I_off), mechanical flexibility, and a relatively large band gap [9,10,11,12]. Despite these properties, when the MoS₂ channel is exposed to atmospheric environments, MoS₂ TFTs exhibit a hysteresis phenomenon in the transfer characteristic. Various research groups have already shown that this hysteresis phenomenon occurs in MoS₂ TFTs due to the effects of oxygen and water in the air [13,14,15]. To suppress this phenomenon, the surface of a semiconducting active channel must be isolated from the air by passivation layers, e.g., atomic-layer-deposited (ALD) aluminum oxide (Al₂O₃) or hafnium dioxide.

It is known that the surface of pristine (i.e., without any surface modifications) MoS₂ does not react well with trimethylaluminum (TMA), which is used as a precursor to form Al₂O₃ layers, due to the absence of dangling bonds [16,17,18,19]. However, high-k dielectric layers without pinholes and/or cracks are indispensable for nano-electronic devices. To improve the wettability and reactivity between the MoS₂ surface and TMA and to ensure complete isolation of the surface from atmospheric environments, some strategies prior to deposition of the passivation layer were proposed, for example, oxygen plasma or ultraviolet-ozone (UV-ozone) treatment [18,19,20,21,22,23]. Conventional MoS₂ TFTs without passivation layers exhibited different electrical properties at high temperatures with respect to the height of the Schottky barrier between the metal electrodes and MoS₂ [8,24,25]. For a relatively high Schottky barrier, charge carriers (i.e., electrons) were blocked by the barrier and thus could not pass at a low temperature. However, as the temperature increased, those carriers started to transport into the MoS₂ channels through thermionic emission [26]. This led to an increase in carrier mobility with respect to increased temperature. Unlike previous reports on MoS₂ TFTs, little research has been conducted on the transport mechanism at high temperatures in passivated MoS₂ TFTs.

The present study investigated how passivated MoS₂ TFTs with a high Schottky barrier change in a high-temperature environment. First, cross-sectional transmission electron microscopy (TEM) was utilized to visualize the quality of Al₂O₃ deposited on the MoS₂ surface through UV-ozone treatment. The surface morphology of the Al₂O₃ on the MoS₂ channel was investigated by atomic force microscopy (AFM), depending on UV-ozone treatment. Also, the electrical properties of the devices were characterized to assess the effectiveness of Al₂O₃ passivation. As a result, it was confirmed that carrier mobility improved and that the hysteresis effect decreased significantly; thus, the desired effect of passivation was achieved. After that, the same tests were performed while the temperature was increased from room temperature to 380 K, and the results showed that the mobility also increased with increasing TFT operating temperature. To analyze the increment of the mobility, the Y-function method was employed to examine the intrinsic mobility of the MoS₂ TFT by excluding contact resistance between the metal electrode and MoS₂ channel.

2. Materials and Methods

2.1. Device Fabrication

Multilayer MoS₂ flakes were mechanically exfoliated from bulk MoS₂ (SPI Supplies, West Chester, PA, USA) through the conventional Scotch tape method and transferred onto thermally grown silicon dioxide (SiO₂) with a thickness of 300 nm used as a gate insulator. A p-type doped silicon (Si) wafer (resistivity ≤0.005 Ω⋅cm) was used as a gate electrode and substrate. To remove the organic and inorganic residues resulting from the transfer procedure, the MoS₂-covered SiO₂/Si substrate was cleaned in acetone and isopropyl alcohol for 1 h 30 min and 30 min, respectively. Titanium/gold (20 nm/100 nm) layers were sequentially deposited by electron beam (e-beam) evaporation, and then source/drain (S/D) electrodes were patterned through conventional photolithography and wet etching processes. To improve the contact between the S/D electrodes and MoS₂ active channel, the devices were annealed at 200 °C for 2 h under mixed gas flow (100 sccm of Ar/10 sccm of H₂).

2.2. UV-Ozone Treatment

For uniform growth of the Al₂O₃ layer, as-fabricated MoS₂ TFTs were subjected to UV-ozone treatment (UVC-30, Jaesung Engineering Co., Anyang-Si, Korea) at power of 15 to 25 mW for 5 min with wavelengths of 185 nm and 254 nm.

2.3. ALD Al₂O₃ Passivation

An Al₂O₃ layer was deposited on the UV-ozone-treated MoS₂ TFTs by ALD (Lucida D100, NCD Co., Ltd., Daejeon, Korea). The deposition conditions of the unit ALD sequence consisted of TMA (0.2 s)/N₂ (10 s)/H₂O (0.2 s)/N₂ (10 s) with a chamber temperature of 200 °C. The thickness of the deposited Al₂O₃ per unit sequence was controlled to be approximately 0.118 nm. The unit sequence was iterated 340 times for a target thickness of 40 nm.

Then a selected area in the Al₂O₃ deposited on the S/D electrode was eliminated to create the electrical contact during measurements.

2.4. Device Characterization

The cross section of the Al₂O₃-passivated MoS₂ TFT was analyzed by TEM (Titan Cubed 60-300, FEI, Hillsboro, OR, USA). The surface morphologies of the Al₂O₃ on MoS₂ TFTs were investigated by AFM (XE7, Park Systems, Suwon-si, Korea) with noncontact mode operation. The electrical properties of the MoS₂ TFTs were measured by using a semiconductor characterization system (4200 SCS, Keithley, Cleveland, Ohio, USA) with a probe station. The temperature dependence of the TFT performance was characterized by using a home-made temperature-controlling vacuum chamber with temperatures ranging from 293 K to 380 K under a moderate vacuum environment (<10⁻³ torr). Before each measurement, the temperature was maintained for 10 min to minimize variations in device performance.

3. Results and Discussion

Figure 1a shows a three-dimensional (3D) schematic illustration of the back-gated TFT encapsulated with Al₂O₃ on top of the multilayer MoS₂ active channel, which was also confirmed in a top-view optical microscope image, as shown in Figure 1b. MoS₂ and contact holes in the S/D electrodes are indicated by red dashed and green dotted lines, respectively. It should be noted that we could not find any pinholes or cracks in the Al₂O₃ layer resulting from UV-ozone pre-treatment, which will be further discussed in relation to Figure 2. Channel length (L) and width (W) of the MoS₂ device were 13.06 and 20.32 μm, respectively, used for calculating the TFT performance in relation to various temperatures.

Figure 1c shows a cross-sectional TEM image of the Al₂O₃-encapsulated MoS₂ TFT with UV-ozone pretreatment. It can be seen that the Al₂O₃ with an average thickness of approximately 42.0 nm uniformly covered the entire MoS₂ surface. In addition, the thickness of the MoS₂ multilayer was estimated to be approximately 64.6 nm, indicating its nearly bulk-like energy band characteristics. The fast Fourier transform patterns and high-angle annular dark-field imaging of the MoS₂ are shown in Figure S1 in the Supplementary Materials.

Figure 2 compares the surfaces morphologies of the Al₂O₃ deposited on MoS₂ with and without UV-ozone pretreatment, which were measured by AFM scanning in a 1 μm × 1 μm region. Many clusters and boundaries can be observed on the surface of the Al₂O₃ (thickness of approximately 40 nm) deposited on the pristine MoS₂ (Figure 2a). However, in the case of Al₂O₃ deposited on the MoS₂ with UV-ozone pretreatment, it shows complete coverage with high uniformity (Figure 2b). The root-mean-square surface roughness (R_RMS) of the directly deposited Al₂O₃ is 0.672 nm, but it decreases to 0.190 nm with a 5-min UV-ozone treatment on MoS₂. These results indicate that UV-ozone exposure is an efficient way to achieve uniform growth of Al₂O₃ on a MoS₂ surface.

Previous studies have found that the electrical properties and device performance of MoS₂ transistors can be enhanced by high-k dielectric encapsulation [27,28]. Therefore, we measured the electrical properties of the MoS₂ TFTs to elucidate the effect of Al₂O₃ passivation on multilayer MoS₂ with UV-ozone pretreatment. Figure 3a compares the transfer characteristic curves I_ds−V_gs of the MoS₂ TFTs without and with Al₂O₃ passivation layers with the application of a source-to-drain voltage (V_ds) of 1 V. Before Al₂O₃ passivation, the pristine MoS₂ TFT exhibits a large hysteresis (black-solid lines in Figure 3a). However, after Al₂O₃ passivation with UV-ozone pre-treatment, represented by the red solid lines in Figure 3a, the transfer curve obtained from a forward V_gs scan closely approached that obtained from a reverse V_gs scan and vice versa, which indicates a distinct reduction of hysteresis behavior (see also Figure S3). It should be noted that reduction of hysteresis was not observed in the MoS₂ TFTs with only UV-ozone treatment, i.e., without Al₂O₃ passivation (Figure S4). For quantitative comparison, we define the difference of threshold voltage (ΔV_th) by subtraction between the V_th values in the forward and reverse transfer characteristic curves, which are estimated to be 20.1 V and 0.5 V for the pristine and Al₂O₃-passivated devices, respectively. However, I_off in the Al₂O₃-passivated MoS₂ TFT increased by about one order of magnitude, resulting in a decrease in the I_on/I_off from 10⁶ to 10⁵. The results may be attributed to the fact that excess electrons could be induced in the MoS₂ active channel due to the positive fixed charges in the Al₂O₃ layer [23,28].

According to transport physics in TFTs, the relation among I_ds, V_gs and V_ds are expressed for a linear regime (|V_ds|<|V_gs−V_th|) as in [29]:

$$I_{\mathrm{ds}}=\frac{{\mu }_{\mathrm{eff}}W{C}_{\mathrm{GI}}}{L}\left[\left(V_{\mathrm{gs}}-V_{\mathrm{th}}\right)V_{\mathrm{ds}}-\frac{V_{\mathrm{ds}}^2}{2}\right] ,$$

(1)

where C_GI is capacitance per unit area of the gate insulator, and μ_eff is the field-effect mobility, one of the figure-of-merit for evaluating TFT characteristics. Based on the standard model of metal-oxide-semiconductor (MOS) FETs and a parallel plate model of gate capacitance [8,29,30], μ_eff of the TFT can be expressed in terms of transconductance ($g_{\mathrm{m}}\equiv \partial I_{\mathrm{ds}}/\partial V_{\mathrm{gs}}$) as

$${\mu }_{\mathrm{eff}}=\frac{L{g}_{\mathrm{m}}}{W{C}_{\mathrm{ox}}{V}_{\mathrm{ds}}},$$

(2)

where C_ox is capacitance per unit area of gate insulator (1.15 × 10⁻⁸ Fcm⁻²). The maximum transconductance of the pristine MoS₂ TFT was 2.31 × 10⁻⁷ S, which was increased to 3.25 × 10⁻⁷ S after Al₂O₃ passivation. As a result, the μ_eff values for the pristine and Al₂O₃-passivated MoS₂ TFTs were calculated to be 40.9 cm² V⁻¹ s⁻¹ and 57.6 cm² V⁻¹ s⁻¹, respectively. Figure 3b compares the output characteristic curves (I_ds−V_ds) of the MoS₂ TFTs without and with Al₂O₃ passivation layers under the application of V_gs ranging from −30 V to 0 V with a step of 5 V. At a low V_gs range from −30 V to −15 V, the I_ds values of the Al₂O₃-passivated MoS₂ TFTs were slightly lower than those of the pristine MoS₂ TFT. However, the clear enhancement of the I_ds of the passivated device has been observed at higher V_gs than −10 V, which agrees well with Figure 3a.

It is well known that MoS₂ seems to be very sensitive to oxygen and water when exposed to an ambient environment [31]. Oxygen and water molecules can be adsorbed on the defect sites of MoS₂ to induce charge traps, thus leading to a relatively larger and clockwise hysteresis as well as mobility degradation [13,32]. High-k dielectric passivation through ALD can be suggested as an efficient method to isolate MoS₂ from atmospheric environments that include external contaminants [27,28,33]. The enhanced I_ds is also attributed to the annealing effect, which decreases the contact resistance between the S/D electrodes and MoS₂ surface. These results are consistent with those of previous reports on Al₂O₃-encapsulated MoS₂ devices, which indicates that no significant defect was introduced in the MoS₂ by the UV-ozone treatment [27,28]. The MoS₂ surface before and after UV-ozone treatment were investigated using Raman and X-ray photoelectron spectroscopy (shown in Figure S2).

Figure 3c,d show the statistical distributions of μ_eff and ΔV_th for 15 representative MoS₂ TFTs in order to confirm the effect of UV-ozone pre-treatment followed by Al₂O₃ passivation. The results of all the devices exhibit increments of the μ_eff as well as reduction of the ΔV_th. The average increment (reduction) rate of μ_eff (ΔV_th) is 40.4% (4.2%). The individual values of μ_eff and ΔV_th are listed in

Table 1. Comparison of μ_eff and ΔV_th values of 15 representative Al₂O₃-passivated MoS₂ TFTs. ¹

# of TFT	μeff (cm2 V−1 s−1)		ΔVth (V)
	Before	After	Before	After
1	31.58	36.57	15.16	1.79
2	24.79	40.55	22.33	0.98
3	15.49	23.58	19.66	0.87
4	32.07	42.59	19.32	0.64
5	36.99	48.52	16.32	1.04
6	51.19	75.53	20.05	0.94
7	32.79	42.18	12.25	0.5
8	40.93	57.62	20.09	0.54
9	31.87	37.79	16.18	0.51
10	23.49	35.55	19.43	0.17
11	36.83	57.22	24.37	0.72
12	33.14	41.69	13.08	0.87
13	10.82	18.19	20.61	0.53
14	74.57	96.67	8.58	0.15
15	32.11	46.63	14.6	0.47

¹ All the Al₂O₃ layers were deposited after UV-ozone treatment.

. These results are evidence that UV-ozone treatment of the multilayer MoS₂ TFTs not only improves the quality of the ALD-grown Al₂O₃ layer, but also enhances the mobility and stability of the MoS₂ devices.

To demonstrate the temperature-dependent behavior of the Al₂O₃-passivated MoS₂ TFT, the electrical properties of the device were measured at 10 different temperatures ranging from 293 K to 380 K under a moderate vacuum condition. Figure 4a presents a semi-logarithmic-scale plot of the temperature-dependent transfer characteristic curves at five different temperatures (293, 310, 330, 350 and 370 K). At the semi-logarithmic-scale of I_ds, variation of I_on was not clearly distinguished. To reveal the enhancement of I_on, linear scale plots of the transfer curves at five other temperatures (300, 320, 340, 360 and 380 K) are compared in Figure 4b.

As shown in Figure 4, I_ds increased with increasing temperature over the whole V_gs region. The results also confirmed the linear increment of the I_ds values at a V_gs of 40 V with respect to 10 different temperatures, as shown in the inset of Figure 4a. From the transfer characteristic curves, we extracted the V_th at 10 different temperatures, which linearly shifted toward the negative V_gs region with increasing temperature (inset of Figure 4b).

Figure 5a shows the output characteristics of the Al₂O₃-passivated MoS₂ TFT at 293 K with those at 380 K, which exhibits almost double increments of I_ds. The results also imply that high temperature would influence the other aspects of TFT performance. However, our TFT architecture has a two-terminal configuration resulting in contact resistance between the S/D electrodes and the MoS₂ active channel, which should be minimized for further investigation of carrier transport mechanisms.

In this context, the Y-function method can be suggested as a proper solution for handling contact resistance in two-terminal systems [8,34,35,36,37], which underestimates the intrinsic capabilities of active materials. By adopting the mobility reduction coefficient η [34], Equation (1) can be rewritten as

$$I_{\mathrm{ds}}=\frac{W{C}_{\mathrm{GI}}}{L}\frac{{\mu }_0}{\left[1+\eta \left(V_{\mathrm{gs}}-V_{\mathrm{th}}\right)\right]}\left(V_{\mathrm{gs}}-V_{\mathrm{th}}\right)V_{\mathrm{ds}},$$

(3)

where μ₀ is the intrinsic mobility excluding any contact resistance component. Considering the definition of transconductance, $g_{\mathrm{m}}$ is also modified using η as follows:

$$g_{\mathrm{m}}\equiv \frac{\partial I_{\mathrm{ds}}}{\partial V_{\mathrm{gs}}}=\frac{W{C}_{\mathrm{GI}}}{L}\frac{{\mu }_0}{{\left[1+\eta \left(V_{\mathrm{gs}}-V_{\mathrm{th}}\right)\right]}^2}{V}_{\mathrm{ds}}.$$

(4)

Then, η is eliminated by combining Equations (3) and (4), and the Y-function can be expressed as

$$Y\equiv \frac{I_{\mathrm{ds}}}{\sqrt{g_{\mathrm{m}}}}=\sqrt{\frac{W}{L}{C}_{\mathrm{GI}}{\mu }_0{V}_{\mathrm{ds}}}\times \left(V_{\mathrm{gs}}-V_{\mathrm{th}}\right).$$

(5)

Finally, we can extract the intrinsic carrier mobility μ₀ from the slope of the Y−V_gs plot [8].

Figure 5b presents μ_eff and μ₀ of the Al₂O₃-passivated MoS₂ TFT at various temperatures, which were calculated from Equations (2) and (5), respectively. At all temperatures, the results showed that (i) all μ₀ values were higher than μ_eff, indicating the existence of relatively high Schottky barriers and subsequent contact resistances in our devices; and (ii) both μ_eff and μ₀ gradually increased with increasing temperature, which can be attributed to reduction of the contact resistance and increased thermionic emission due to high temperature [25,38]. The inset of Figure 5b presents the temperature-dependent variation of the difference between μ_eff and μ₀ (Δμ), which shows linear response behavior with respect to temperature increase. The details of μ_eff, μ₀ and Δμ are provided in

Table 2. Values of μ_eff and μ₀ at various temperatures.

Temperature	293 K	300 K	310 K	320 K	330 K	340 K	350 K	360 K	370 K	380 K
μeff	6.4	6.57	6.81	6.95	7.31	7.39	7.61	7.81	7.99	8.14
μ0	7.73	8.07	8.45	8.68	9.32	9.47	9.89	10.39	10.81	11.31
Δμ	1.33	1.5	1.64	1.73	2.01	2.08	2.28	2.58	2.82	3.17

.

4. Conclusions

In this paper, we reported on the effects of ALD-Al₂O₃ passivation on multilayer MoS₂ TFTs and the temperature-dependent variation of the TFT performance. High-k Al₂O₃ layers were uniformly passivated over the entire surface of MoS₂ TFTs with the aid of UV-ozone treatment leading to (i) huge reductions of the hysteretic transfer curves; (ii) saturation current enhancement of the output characteristics; and (iii) increases in μ_eff of approximately 40.4%. Based on investigations of the temperature-dependent TFT characteristics including intrinsic carrier mobility (μ₀) extracted by the Y-function method, we could conclude that the dominant transport mechanism is thermionic emission in our Al₂O₃-passivated MoS₂ TFTs with a considerable Schottky barrier between the S/D electrodes and active channel. In addition, the proposed approach at relatively high (i.e., realistic device working) temperatures can be employed to realize stable and reproducible electronic devices for robust and practical applications.