Vacuum ultraviolet radiation effects on two-dimensional MoS₂ field-effect transistors

Molybdenum disulfide (MoS₂) has been the subject of intense research as an n-type two-dimensional channel material for next-generation electronics. MoS₂ field-effect transistors (FETs) exhibit large current on/off ratios (10⁸–10⁹)^1,2 and sizeable room-temperature field-effect mobilities (30–480 cm²/V s),^3–7 which argue that MoS₂ is particularly well-suited for low-power electronics.⁸ The MoS₂ electronic band gap, atomic-scale thickness, and strong optical absorption have also attracted interest for vertical heterostructure devices^9–13 and photovoltaic active layers.^14–17 For these reasons as well as proven operation in harsh environments,¹⁸ the unique properties of MoS₂ motivate the study of its suitability as a material in space electronics applications, in which its electronic band gap, large switching ratio, and ultrathin nature are favorable for low-power, light-weight operation. Computational results indicate that within the transition metal dichalcogenides¹⁹ and black phosphorus,²⁰ MoS₂ is among the least susceptible to radiation-induced defect formation. The exceptionally small cross-sections of MoS₂ flakes are further advantageous for radiation hardness to single-event effects (SEEs),²¹ suggesting that the interactions of the surrounding materials with incident radiation in the form of total ionizing dose (TID) will play the greatest role in the radiation response of MoS₂ as a semiconducting electronic device component. Initial studies on the basic mechanisms driving the TID response of MoS₂ field-effect transistors (FETs) have revealed the extrinsic effect of atmospheric adsorbates on the evolution of device transport properties²² or have demonstrated radiation damage effects that are possibly convoluted with atmospheric adsorbates.²³ Because such atmospheric effects have also been shown to dominate the radiation response of other low-dimensional nanoelectronic systems,^24,25 the radiation response of MoS₂ must be isolated from extrinsic factors to reveal the intrinsic behavior of MoS₂ FETs.

In this letter, we investigate the vacuum ultraviolet (VUV) TID response of both single-layer (SL) and multilayer (ML) MoS₂ FETs having SiO₂ gate dielectrics under high vacuum. For decades, VUV exposure has been used to understand radiation-induced charge trapping in SiO₂ due to its spatially shallow absorption near the incident surface and lack of displacement damage in Si or its adjacent oxides.^26–28 Contrary to previous experiments conducted under ambient conditions, it is found that MoS₂ FETs exhibit threshold voltage shifts toward negative gate biases under VUV exposure. Furthermore, the evolution of the transport curves as a function of VUV reveals the relative contributions of oxide and interface trapped charge. The maximum drain current and field-effect mobility are observed to increase with increasing dose, which is unique to MoS₂ compared to other two-dimensional electronic materials.^29,30 Finally, Raman spectroscopy before and after VUV exposure reveals no measureable structural changes in the MoS₂ channel, establishing the gate dielectric as the sensitive component of the system and suggesting dielectric engineering as a promising strategy for enhancing the radiation hardness of MoS₂ FETs.

The semiconducting channels of the FETs were prepared by micromechanically exfoliating MoS₂ flakes using the Scotch tape method onto degenerately doped n-type Si substrates having 300 nm thick thermal SiO₂. Single-layer (SL) and multilayer (ML) flakes were identified optically for subsequent device fabrication. Gold source and drain contacts were patterned via standard electron-beam lithography, thermal evaporation, and liftoff processes. Figure 1 includes the optical images of the SL and ML devices. The SL MoS₂ channel is outlined in Figure 1(a) and has dimensions of 2.9 μm × 6.8 μm (L × W), while the ML channel in Figure 1(b) is 3.1 μm × 6.3 μm, the width of the latter being the average of its trapezoidal bases. Electrical contacts were deposited without an adhesion layer to facilitate quasi-ohmic contact to the MoS₂ channel.⁴ The Si substrate functions as a global back gate contact for three-terminal FET devices, and the fabricated samples were wire-bonded to a ceramic package to allow in situ measurements under vacuum. A device schematic of a SL device is depicted in Figure 1(c). The ML channel device in this study consists of a van der Waals stack of six MoS₂ layers as determined by Raman spectroscopy (supplementary material, Figure S1). Additional fabrication details were published previously.³¹ 

The evolution of the electronic transport characteristics of the SL and ML MoS₂ FET devices was measured as a function of sequentially larger VUV doses at pressures below 10⁻⁶ Torr and under a photon flux of 3.4 × 10¹⁰ photons/cm² s. This flux includes only those photons having energies large enough to be absorbed by the SiO₂ (E_g ≥ 9 eV). A calibrated silicon photodiode with enhanced extreme and vacuum ultraviolet responsivity (AXUV100G, Opto Diode) was used to quantify the irradiance of the vacuum ultraviolet light source (L10366 series, Hamamatsu Photonics K.K.). This quantification was performed by measuring the photocurrent at the sample position and converting it to irradiance (0.58 μW/cm²) based on the calibrated photodiode responsivity and VUV bulb spectrum. I_d-V_g transport measurements were carried out before irradiation and after each VUV exposure. During VUV exposure, the devices were held at a constant drain bias (5 mV) and constant forward gate bias (+1 MV/cm) to facilitate separation and collection of VUV-liberated charge carriers in the dielectric. In situ measurements are critical to gauging the radiation response of these unencapsulated devices in the absence of atmospheric adsorbates. All measurements and exposures were performed at room temperature. The linear field-effect mobilities for the as-fabricated SL and ML devices were found to be 6.4 cm²/V s and 42 cm²/V s, respectively. The I_d-V_g curves prior to VUV exposure for both devices are included in black in Figure 2. The accumulated VUV dose has the effect of increasing the maximum drain current and subthreshold swing while shifting the transfer characteristics toward negative gate bias (red curves). The shift toward negative gate bias for both devices indicates an overall accumulation of positive charge between the MoS₂ channel and the gate electrode.

Analysis of the collective changes in threshold voltage and subthreshold swing can be used to quantify the number and type of VUV-induced trapped charges in the MoS₂-SiO₂ FET. These trapped charges are expected to be located predominantly within the gate oxide layer or at the MoS₂-SiO₂ interface. Effects related to surface adsorbates or top-surface interface charges are not considered because these devices are not encapsulated by dielectric layers and the irradiations are performed under vacuum. The observed shift in threshold voltage, ΔV_th, is given by separate contributions from oxide trapped charges and interface trapped charges, ΔV_OT and ΔV_IT, respectively. For a given VUV dose, the net effect is ΔV_th = ΔV_OT + ΔV_IT.³² For two states having different subthreshold swings (i.e., before and after exposure), the induced interface trapped charge, ΔN_IT, is computed as $Δ N I T = C o x ln 10 ϕ b k T Δ S S$⁠, where ΔSS is the observed difference in subthreshold swings. At room temperature, kT = 26 meV and ϕ_b is the energy difference between the midgap energy and the Fermi level, 0.4 eV and 0.15 eV, for the SL and ML devices, respectively.³³ Given ΔN_IT, the corresponding ΔV_IT is computed from $Δ V I T = q C o x Δ N I T$⁠.

The values of ΔV_th, ΔV_OT, and ΔV_IT are computed for each VUV dose and are plotted as a function of dose for both devices in Figure 3. Due to the biasing conditions during irradiation (+1 MV/cm across the SiO₂ gate dielectric), holes liberated by incident photons will migrate toward the MoS₂-SiO₂ interface, where they are trapped in defect centers known to exist in SiO₂.³⁴ Additionally, the positive gate bias brings the MoS₂ channel into accumulation, where the large electron density can supply electrons to fill energetically favorable interface states. Thus, we assign the signs of ΔV_OT and ΔV_IT to be positive and negative, respectively. For a total VUV dose of 2.2 × 10¹³ photons/cm², we observe that ΔV_OT and ΔV_IT shift in opposite directions, which is the net effect of hole trapping in the bulk SiO₂ (shifts toward negative gate bias) and electron trapping at the MoS₂-SiO₂ interface (shifts toward positive gate bias). The overall effect is a comparatively small magnitude shift toward negative gate bias.

While the qualitative similarity of the trends in the charge trapping and voltage shifts in the present SL and ML devices suggest similar mechanisms affecting transport in both devices, the magnitudes and behaviors are quantifiably different. The maximum values of ΔV_OT and ΔV_IT for the SL device (−54 V and 41 V, respectively) are significantly larger than those of the ML device (−30 V and 16 V), an observation that is explained by the larger electronic transport band gap and reduced screening in single-layer MoS₂ versus multilayer MoS₂ (E_g = 2.7–2.8 eV and 1.3 eV, respectively^33,35). The larger transport band gap is reflected in the ϕ_b term in ΔN_IT, allowing a higher quantity of trap states, and thus larger voltage shifts. While the voltage shifts in the ML device (Figure 3(b)) occur monotonically as a function of dose, the SL device (Figure 3(a)) exhibits ΔV_OT and ΔV_IT values that vary without an identifiable trend until the high dose regime is reached (>10¹³ photons/cm²). This variability might arise from the “all-surface” nature of SL MoS₂, where current is strictly confined within the two-dimensional plane. The ML device also has a thickness comparable to the Debye length of 7 nm in ML MoS₂, suggesting that its current flow hot spot is removed from the MoS₂-SiO₂ interface.³⁶ The result is that charged substrate or adsorbate impurities are less effectively screened from the electronic channel in the SL device, an effect illustrated by the superlinear dependence of the drain current on the gate voltage in the transfer characteristics of the SL device compared to the ML device (Figure 2).³⁷ Additionally, the SL device exhibits larger fluctuations with charge redistribution under low radiation doses and stronger voltage shifts at higher doses.

In addition to the increase in maximum current, the field-effect mobility (⁠ $μ F E ∼ ∂ I d ∂ V g$⁠) is enhanced as a function of VUV dose for both devices. The changes in field-effect mobility relative to the pre-radiation values as a function of VUV dose for the SL and ML devices are shown in Figures 3(c) and 3(d), respectively. This trend stands in contrast to graphene FETs on SiO₂ gate dielectrics, which suffer significant mobility degradation on radiation exposure as a result of Coulomb scattering induced by dielectric trapped charges. For MoS₂ FETs, the increased field-effect mobility with increased oxide and interface-trapped charge indicates a transport mechanism that is not dominated by Coulomb scattering at room temperature. This conclusion is consistent with previous studies that showed phonon scattering limited transport at room temperature for MoS₂ FETs.⁴ Additionally, the increased field-effect mobility is the opposite of what is observed in MoS₂ devices exposed to X-ray²² and proton irradiation²³ under ambient conditions, emphasizing the importance of controlling the atmosphere during MoS₂ radiation studies. This observation is consistent with those for other low-dimensional electronic systems, such as carbon nanotubes²⁴ and graphene,²⁵ in which fundamentally different electronic responses are measured when irradiating under vacuum and ambient conditions. Finally, chemisorbed oxygen or the formation of insulating MoO_x species would be expected to decrease device field-effect mobility and maximum drain current, neither of which is observed.

Raman spectroscopy was next used to further investigate the structural evolution of MoS₂ as a result of VUV exposure. The Raman spectra of MoS₂ flakes have been extensively studied, and important structural data such as layer number³⁸ and oxidation content³⁹ are readily identified by characteristic positions of the Raman peaks corresponding to the E¹_2g and A_1g vibrational modes. Figure 4 shows the Raman spectra and spatial maps of a MoS₂ flake having multiple thickness regions before and after VUV exposure for 640 s. An optical image of this MoS₂ flake is provided in Figure 4(a). Raman spectra (532 nm excitation) are captured and peaks are fit to Lorentzian functions. Figure 4(b) includes pre-exposure and post-exposure Raman maps of the peak separation of the MoS₂ E¹_2g and A_1g vibrational modes, a quantity known to vary with MoS₂ layer number⁴⁰ and chemisorbed oxygen.³⁹ A comparison of the before and after Raman maps reveals no significant variation as a result of VUV exposure. Choosing isolated regions, the Raman spectra are shown for bulk-like (spot 1), 4-layer (spot 2), and 3-layer (spot 3) MoS₂ in Figure 4(c). Comparing these spectra before and after VUV exposure corroborates that the Raman signature of MoS₂ does not change with VUV dose for a range of flake thicknesses. This lack of significant structural changes in the MoS₂ channel indicates that the changes in electronic transport induced by VUV exposure are dominated by charge trapping in the SiO₂ and at the MoS₂-SiO₂ interface, rather than by severe damage to the MoS₂ crystal structure. These observations are consistent with TID experiments in other low-dimensional nanoelectronic systems such as graphene²⁹ and carbon nanotube²⁴ FETs.

In summary, this work delineates and quantifies the charge trapping behavior intrinsic to the MoS₂-SiO₂ system in the absence of atmospheric adsorbates following exposure to VUV radiation. Electronic transport characteristics are found to shift toward negative gate bias, consistent with the trapping of TID-generated holes known to occur in SiO₂. Additionally, a significant quantity of negative charge is trapped at the MoS₂-SiO₂ interface, and the field-effect mobility is found to increase with VUV exposure. The shift toward negative gate bias and increase in mobility with increasing radiation dose are observations that contradict earlier studies and can be attributed to the isolation of the MoS₂ FET from ambient exposure during irradiation. Finally, Raman spectroscopy data indicate no major structural changes or oxidation of the MoS₂ with VUV dose, suggesting that the interaction of TID with the MoS₂-SiO₂ system occurs primarily within the oxide layer. The similarity of the charge trapping behavior in the MoS₂-SiO₂ system compared to conventional Si-SiO₂ devices indicates that dielectric-based TID hardening strategies as employed in Si³⁴ and other nanomaterial^41,42 electronics should be applicable to two-dimensional MoS₂.

See supplementary material for Raman characterization of the multilayer MoS₂ FET.

This work was supported by the MRSEC program of the National Science Foundation (DMR-1121262), the National Aeronautics and Space Administration NSTRF Grant Nos. NNX12AM44H and NNX11AM87H, and the Defense Threat Reduction Agency (MIPR #HDTRA-15-15399). V.K.S. acknowledges support from the 2-DARE Program (NSF EFRI-1433510). This research was performed while S.W.S. held a National Research Council Associateship Award at the U.S. Naval Research Laboratory. This work made use of the EPIC, Keck-II, and SPID facilities of Northwestern University NUANCE Center, which has received support from the Soft and Hybrid Nanotechnology Experimental (SHyNE) Resource (NSF NNCI-1542205); the MRSEC program (NSF DMR-1121262); the International Institute for Nanotechnology (IIN); the Keck Foundation; and the State of Illinois.