Ionic Liquid Gating of Suspended MoS2 Field Effect Transistor

Letter pubs.acs.org/NanoLett

Ionic Liquid Gating of Suspended MoS2 Field Effect Transistor Devices Fenglin Wang,† Petr Stepanov,† Mason Gray,† Chun Ning Lau,*,† Mikhail E. Itkis,‡ and Robert C. Haddon*,‡,§ †

Department of Physics and Astronomy, ‡Department of Chemistry, and §Department of Chemical and Environmental Engineering, University of California, Riverside, Riverside, California 92521, United States S Supporting Information *

ABSTRACT: We demonstrate ionic liquid (IL) gating of suspended few-layer MoS2 transistors, where ions can accumulate on both exposed surfaces. Upon application of IL, all free-standing samples consistently display more significant improvement in conductance than substrate-supported devices. The measured IL gate coupling efficiency is up to 4.6 × 1013 cm−2 V−1. Electrical transport data reveal contact-dominated electrical transport properties and the Schottky emission as the underlying mechanism. By modulating IL gate voltage, the suspended MoS2 devices display metal−insulator transition. Our results demonstrate that more efficient charge induction can be achieved in suspended two-dimensional (2D) materials, which with further optimization, may enable extremely high charge density and novel phase transition. KEYWORDS: Ionic liquid gating, suspended structure, molybdenum disulfide, metal−insulator transition

A

In this article, we demonstrate IL gating of suspended fewlayer MoS2 field effect transistors. In comparison with substrate-supported devices, suspended devices display dramatic enhancement of conductance and (to a lesser extent) mobility upon application of IL. Electrical characteristics reveal that for ionic liquid gate voltage VILg < 1 V, transport in these devices is dominated by Schottky emission at the MoS2− electrode interfaces. We estimate that the dielectric constant of the IL is ∼14.5, and the IL coupling efficiency of the suspended device is as high as 4.6 × 1013 cm−2 V−1, which is several times higher than that of substrate-supported devices in other work5 under similar conditions. Lastly, as IL gate varies, metal− insulator transition is observed. Our results demonstrate that suspended devices allow formation of EDL on both exposed surfaces and can be extended to other 2D materials. Few-layer MoS2 flakes are mechanically exfoliated onto Si/ SiO2 substrates from the bulk material (source: synthetic crystal from 2D Semiconductors), and their thicknesses are determined by atomic force microscope (AFM) and/or Raman spectroscopy. Figure 1a,b shows an optical and an AFM image of a 5 nm MoS2 sheet. The MoS2 sheet is coupled to titanium/gold electrodes that are covered by a thin protective layer of chromium,22 released from the SiO2 layer via wet etching in hydrofluoric acid, and dried in a critical point dryer. The width and length of the devices are 2.0 and 0.9 μm,

n ionic liquid (IL) is a liquidized salt at room temperature, containing separable and mobile cations and anions. It has been demonstrated as an effective gating electrolyte on many different materials.1−7 Under an externally applied voltage, the ions in the liquid will separate and a layer of cations or anions accumulates on the sample surface, forming an electric double layer (EDL) that is essentially a nanogap capacitor with enormous capacitance. The charge density that can be induced by IL gating is usually orders of magnitude larger than that of a conventional silicon back gate. For instance, strong electromodulation of interband exciton transitions are observed in semiconducting carbon nanotube films under IL gating,8 and superconductivity induced by the exceedingly high charge density (up to 1014 cm−2) under ionic liquid gating has been achieved in a 20 nm thick molybdenum disulfide (MoS2) flakes.4 To date all the IL gating experiments are performed on substrate-supported devices, which limit the exposure of the device to EDL to a single surface. On the other hand, suspended device structures based on nanotubes9,10 and graphene11−15 have been widely studied. Removing underlying substrate not only eliminates the external influence from the Si/ SiO2 substrate but also opens another channel for ionic liquid gating to form a second EDL on the bottom surface of the atomic layer. Additionally, the relatively high dielectric constant16 of ILs may also improve device performance, as shown by prior works using HfO217−19 and Al2O3.20,21 Thus, IL gating of suspended devices based on 2D materials presents an interesting yet unexplored venue of research. © XXXX American Chemical Society

Received: April 24, 2015 Revised: July 7, 2015

A

DOI: 10.1021/acs.nanolett.5b01610 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 1. (a,b). Optical and AFM images of a typical few-layer MoS2 sheet, scale bar: 2 μm. Inset: height profile indicates 5 nm thickness. (c). SEM image of a suspended MoS2 device, scale bar: 1 μm. (d) I−V characteristics of as-fabricated devices (without IL) at Vbg = 0 (blue), 5 V (green), 10 V (red), respectively. Inset: G(Vbg) Conductance versus back gate voltage.

Figure 2. (a,b) Schematics and optical image of Ionic liquid gating on suspended device. (c). Zero-bias sheet conductance at different VILg of nine substrate-supported (blue) and nine suspended (red) samples.

respectively. A “side gate” lead that is separated from the MoS2 sheet by 20 μm acts as the counter electrode for the ionic liquid. As a control, a number of substrate-supported devices are also fabricated. For all measurements, the leakage current between the IL side gate and the sample is 1 V, the I−V curves are linear for V up to ±1 V. In contrast, at VILg = 0 the I−V curves are linear for relatively small source-drain bias (V < 0.1 V) (Figure 3a,b) but become strongly nonlinear for large V range (Figure 3a,b inset). Insight into transport in intrinsic MoS2 devices can be obtained by studying the I−V characteristics at VILg = 0. Figure B

DOI: 10.1021/acs.nanolett.5b01610 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters 3c,d replots the same data in Figure 3a,b as I versus √V in log− linear scale for devices 1 and 2, respectively. Both data sets, spanning 4 decades in I, fall nicely onto a straight line, indicating that I ∝ e√V. Such a dependence may arise from two different mechanisms:23 (1) Schottky emission, in which the thermionic emission dominates the carrier transport across the metal−semiconductor interface, or (2) Frenkel-Poole emission, which is due to field-enhanced thermal excitation of trapped electrons into the conduction band of the semiconductor. For both mechanisms, one expects

⎡a V − Φ ⎤ B I ∝ exp⎢ ⎥ kBT ⎣ ⎦

(1)

where a = ηe(e/4πϵ0ϵrd)1/2. Here e is the elementary charge, ϵ0 and ϵr are the permittivity of vacuum and device, respectively, d is the distance over which the electric field is applied, V is the bias voltage, ΦB is the barrier height, and kB is the Boltzmann constant. η is a constant and is 1 and 2 for Schottky and Frenkel-Poole emission, respectively. To distinguish between these two transport mechanisms, we note that Schottky emission occurs at the metal−MoS2 interface, so d is given by the thickness of the MoS2 sheet, ∼ 2−10 nm. In contrast, transport via Frenkel-Poole emission takes place laterally in the bulk of the MoS2 sheet, and the electric field drops across the length of the samples, thus d ∼ 1 μm. Fitting the data in Figure 3c,d to eq 1, we obtain a slope for a/kBT0 of 6.81 for device 1 and 6.04 for device 2. This yields εrd ∼ 70 nm for both devices and excludes Frenkel-Poole emission as the leading transport mechanism, because d ∼ 1 μm would lead to the unphysically small value of εr < 1. We thus conclude that transport in IL-gated MoS2 devices is dominated by Schottky emission at the metal−MoS2 interfaces. In fact, it is very likely that Schottky barrier is also dominant in conventional samples on SiO2 and constitutes one of the mobility bottlenecks of MoS2 field-effect transistors. Using values of the slope obtained from the fitting, T0 = 300 K (265 K) and d = 6.2 nm (10 nm) for device 1 (device 2), we estimate εr to be 10.9 and 11.1 for device 1 and device 2, respectively. Taking εr as the average of the dielectric constants of MoS2, its substrate and superstrate (the latter two being the IL), we extract the dielectric constant of DEME-TFSI to be εIL ∼ 14.5 near room temperature, which is in reasonable agreement with that reported in literature.16,24 With the extracted value of εIL, we can then estimate the charge density induced by IL by measuring the ratio α/β, where α and β are coupling efficiencies of IL gate and back gate, respectively. To this end, we couple the sample to a given VILg value at room temperature, allowed a sufficient time lapse so that the equilibrium state is reached, then cool the sample slowly to T= 150 K while maintaining constant VILg. Figure 4a displays G(VILg) of the device at T = 265 K. As the IL freezes at ∼200 K, the IL-induced charge density stays constant, while the back gate can be employed to further tune the density of charge carriers. The ratio α/β can then be inferred by plotting the G(Vbg) curves at various VILg values at T = 150 K (Figure 4b,c) and calculated by comparing the gate voltages required to induce the same magnitude change in conductance. For instance, Figure 4b displays G(Vbg) in linear scale for VILg = 2.4, 2.2, and 2.0 V, respectively. (ΔG/ΔVbg) ≈ 2.4 μS/V, whereas (ΔG/ΔVILg) = 217 μS/(0.2 V) = 1085 μS/V. Hence at the average value VILg = 2.3 V, α/β = 450. The calculated values of α/β for VILg ranging from 0 to 2.3 V are displayed in Figure

Figure 4. (a). G(VILg) at T = 265 K. (b). G(Vbg) at T = 150 K and VILg = 2.0, 2.2, and 2.4 V, respectively. (c). G(Vbg) at various VILg values at T = 150 K in log scale. (d). Ratio of coupling efficiencies between IL gate and back gate as a function of VILg.

4d. When the device is highly doped at VILg > 2 V, α/β ∼ 450; as the Fermi level moves toward the band edge, this ratio decreases to 51 at Vbg ∼ 1.3 V and 25 at Vbg ∼ 0.25 V. Such dependence of gate efficiency ratio on VILg has been observed in prior experiments6 and arises from the fact that as the Fermi level is tuned into the band gap, the diminishing quantum capacitance becomes the dominating term, as it is added in series with the geometric capacitance. Finally, the back gate’s coupling efficiency β can be estimated by using parallel plate capacitance that consists of 180 nm of SiO2 and 120 nm of IL. Using εIL ∼ 14.5 for IL as discussed above, β ∼ 1.1 × 1010 cm−2 V−1. Thus, IL’s coupling efficiency α is estimated to be 4.9 × 1013 cm−2 V−1, which is 4−5 times larger than that reported in prior experiment,5 and consistent with ion accumulation on both sides of the suspended sheet. One caveat in the above measurements of α and β is our assumption that εIL stays constant for the temperature range, 150−300 K. Albeit not directly verified, this assumption is consistent with the G(T) data at VILg = 0, which in Arrhenius plots appear as straight lines with almost zero curvature, which would otherwise be introduced by the temperature dependence of εIL (see Figure 5a−d and eq 1). Additionally, due to the relatively small serial capacitance of SiO2, the values of α and β are quite insensitive to the exact values of εIL: for instance, doubling εIL only changes α and β by ∼7%. Thus, our measurement of charge density induced by IL should be reasonably robust. We also note that we do not discern any significant dependence of α on the thickness of MoS2 sheets, though such dependence cannot be conclusively ruled out and warrants further experimental studies. We now return to the observed dramatic enhancement of conductivity of IL-gating on suspended devices relative to that on substrate-supported samples (Figure 2c). On the basis of our analysis above, such enhancement may be attributed to two C

DOI: 10.1021/acs.nanolett.5b01610 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

constant), particularly considering that two-terminal resistivity measured here may include contact resistance. In conclusion, by applying ionic liquid gating to suspended MoS2 FET devices, we observe dramatically improved conductivity and mobility, which suggest Schottky barrier reduction and dielectric screening by the IL environment. Their transport characteristics indicate that Schottky emission is the dominating transport mechanism. We estimate that the dielectric constant of DEME-TFSI is ∼14.5 with a coupling efficiency of 4.6 × 1013 cm−2 V−1. Metal-insulator transition is observed in these IL-gated samples. With further optimization in choosing the ionic species and in IL handling and storage, we expect even more efficient coupling can be reached. IL gating of suspended devices may enable potentially unprecedented charge density for exploration of novel phenomena such as unconventional superconducting phase transition.26

■

ASSOCIATED CONTENT

S Supporting Information *

Sample preparation, device fabrication, device characterization. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.5b01610.

■

Figure 5. (a,b). Sheet conductance versus temperature at different VILg of device 1 and 2, respectively. (c,d). Same data in (a,b) plotted in Arrhenius scale. (e). Extracted Schottky barrier heights at different VILg from device 1 and 2.

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. Notes

The authors declare no competing financial interest.

factors: (1) The higher doping level is due to ion accumulation on both sides of the MoS2 sheets; in the nonphonon limited conduction regime, larger charge carrier density that participates in transport leads to larger conductivity. (2) Better screening is provided by the high-κ IL that encapsulates MoS2 on both sides, which reduces scattering from charged impurities, and possibly more importantly, the Schottky barriers at the metal−semiconductor contacts. Finally, we explore the temperature dependence of suspended devices at different VILg. Figure 5a,b presents such σs(T) data at different VILg values. For VILg < 2 V, the conductance decreases with temperature, indicating an insulating behavior. Figure 5c,d replots the data in an Arrhenius plot. For this insulating regime, the σs(T) curves can be described by thermal activation over an energy barrier that is VILg-dependent. From eq 1, σs ∝ exp(−b/T), where b = (−a(V0)1/2 + ΦB)/kB can be obtained from the slope of the curves in the Arrhenius plot, V0 = 10 mV is the applied bias, and a is obtained from the I−V curves near room temperature using eq 1. Thus, we can extract the Schottky barrier height ΦB from the data. Figure 5e shows the barrier height versus ionic liquid gate values. As VILg increases from 0 to 1 V, the barrier height lowers from ∼70 to 36 meV and is extrapolated to be 0 at VILg ∼ 1.8 V. This expected disappearance of Schottky barrier at large VILg is borne out by data; for VILg ≥ 2 V, the conductance increases with a decreasing T, that is, the device displays metallic behavior. Thus, the devices display metal−insulator transition tuned by VILg. The critical sheet resistance that separates the metal−insulator transition is 16 kΩ for device 1 and 28 kΩ for device 2. This is close to the theoretical value of h/2e2 = 12.9 kΩ for spin-degenerate thin films25 (where h is Planck’s

■

ACKNOWLEDGMENTS This work was supported in part by FAME, one of six centers of STARnet, a Semiconductor Research Corporation program sponsored by MARCO and DARPA, and by UC Lab Fees Research Program Award #237789. C.N.L. acknowledges the support by CONSEPT center at UCR. M.E.I. and R.C.H. acknowledge the support by NSF/ECCS-1404671.

■

REFERENCES

(1) Asanuma, S.; Xiang, P. H.; Yamada, H.; Sato, H.; Inoue, I. H.; Akoh, H.; Sawa, A.; Ueno, K.; Shimotani, H.; Yuan, H.; Kawasaki, M.; Iwasa, Y. Appl. Phys. Lett. 2010, 97, 142110. (2) Yuan, H. T.; Liu, H. W.; Shimotani, H.; Guo, H.; Chen, M. W.; Xue, Q. K.; Iwasa, Y. Nano Lett. 2011, 11, 2601−2605. (3) Ye, J. T.; Craciun, M. F.; Koshino, M.; Russo, S.; Inoue, S.; Yuan, H. T.; Shimotani, H.; Morpurgo, A. F.; Iwasa, Y. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 13002−13006. (4) Ye, J. T.; Zhang, Y. J.; Akashi, R.; Bahramy, M. S.; Arita, R.; Iwasa, Y. Science 2012, 338, 1193−1196. (5) Perera, M. M.; Lin, M. W.; Chuang, H. J.; Chamlagain, B. P.; Wang, C. Y.; Tan, X. B.; Cheng, M. M. C.; Tomanek, D.; Zhou, Z. X. ACS Nano 2013, 7, 4449−4458. (6) Chu, L. Q.; Schmidt, H.; Pu, J.; Wang, S. F.; Ozyilmaz, B.; Takenobu, T.; Eda, G. Sci. Rep. 2014, 4, 7293. (7) Jo, S.; Ubrig, N.; Berger, H.; Kuzmenko, A. B.; Morpurgo, A. F. Nano Lett. 2014, 14, 2019−2025. (8) Wang, F. H.; Itkis, M. E.; Bekyarova, E.; Haddon, R. C. Nat. Photonics 2013, 7, 460−466. (9) Deshpande, V. V.; Bockrath, M. Nat. Phys. 2008, 4, 314−318. (10) Deshpande, V. V.; Chandra, B.; Caldwell, R.; Novikov, D. S.; Hone, J.; Bockrath, M. Science 2009, 323, 106−110.

D

DOI: 10.1021/acs.nanolett.5b01610 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters (11) Bao, W.; Jing, L.; Velasco, J.; Lee, Y.; Liu, G.; Tran, D.; Standley, B.; Aykol, M.; Cronin, S. B.; Smirnov, D.; Koshino, M.; McCann, E.; Bockrath, M.; Lau, C. N. Nat. Phys. 2011, 7, 948−952. (12) Bao, W. Z.; Zhao, Z.; Zhang, H.; Liu, G.; Kratz, P.; Jing, L.; Velasco, J.; Smirnov, D.; Lau, C. N. Phys. Rev. Lett. 2010, 105, 246601. (13) Bolotin, K. I.; Ghahari, F.; Shulman, M. D.; Stormer, H. L.; Kim, P. Nature 2009, 462, 196−199. (14) Du, X.; Skachko, I.; Duerr, F.; Luican, A.; Andrei, E. Y. Nature 2009, 462, 192−195. (15) Velasco, J.; Jing, L.; Bao, W.; Lee, Y.; Kratz, P.; Aji, V.; Bockrath, M.; Lau, C. N.; Varma, C.; Stillwell, R.; Smirnov, D.; Zhang, F.; Jung, J.; MacDonald, A. H. Nat. Nanotechnol. 2012, 7, 156−160. (16) Ueno, K.; Nakamura, S.; Shimotani, H.; Ohtomo, A.; Kimura, N.; Nojima, T.; Aoki, H.; Iwasa, Y.; Kawasaki, M. Nat. Mater. 2008, 7, 855−858. (17) Radisavljevic, B.; Radenovic, A.; Brivio, J.; Giacometti, V.; Kis, A. Nat. Nanotechnol. 2011, 6, 147−150. (18) Lembke, D.; Kis, A. ACS Nano 2012, 6, 10070−10075. (19) Radisavljevic, B.; Kis, A. Nat. Mater. 2013, 12, 815−820. (20) Liu, H.; Ye, P. D. D. IEEE Electron Device Lett. 2012, 33, 546− 548. (21) Sundaram, R. S.; Engel, M.; Lombardo, A.; Krupke, R.; Ferrari, A. C.; Avouris, P.; Steiner, M. Nano Lett. 2013, 13, 1416−1421. (22) Wang, F.; Stepanov, P.; Gray, M.; Lau, C. N. Nanotechnology 2015, 26, 105709. (23) Sze, S. M. Physics of Semiconductor Devices. In Physics of Semiconductor Devices; Wiley-Interscience: New York, 1981; pp 402− 404. (24) Fujimoto, T.; Awaga, K. Phys. Chem. Chem. Phys. 2013, 15, 8983−9006. (25) Licciardello, D. C.; Thouless, D. J. J. Phys. C: Solid State Phys. 1975, 8, 4157−4170. (26) Nandkishore, R.; Levitov, L. S.; Chubukov, A. V. Nat. Phys. 2012, 8, 158−163.

E

DOI: 10.1021/acs.nanolett.5b01610 Nano Lett. XXXX, XXX, XXX−XXX