Contact Research Strategy for Emerging Molybdenum Disulfide and Other Two-Dimensional Field-effect Transistors

Layered two-dimensional (2D) semiconducting transition metal dichalcogenides (TMD) have been widely isolated, synthesized, and characterized recently. Numerous 2D materials are identified as the potential candidates as channel materials for future thin film technology due to their high mobility and the exhibiting bandgaps. While many TMD filed-effect transistors (FETs) have been widely demonstrated along with a significant progress to clearly understand the device physics, large contact resistance at metal/semiconductor interface still remain a challenge. From 2D device research point of view, how to minimize the Schottky barrier effects on contacts thus reduce the contact resistance of metals on 2D materials is very critical for the further development of the field. Here, we present a review of contact research on molybdenum disulfide and other TMD FETs from the fundamental understanding of metal-semiconductor interfaces on 2D materials. A clear contact research strategy on 2D semiconducting materials is developed for future high-performance 2D FETs with aggressively scaled dimensions.

Author to whom correspondence should be addressed; electronic mail: yep@purdue.edu 2 As the forefather of the layered 2D materials, graphene had been put under the spotlight and enjoyed its several advantages of its fundamental properties. Despite the short history of graphene research, it has already revealed a series of new physics and potential applications, and no longer requires any further proof of its importance in quantum physics, condensed matter, and electronic devices. [1][2][3][4] However, the performance of graphene based electronic device is barely satisfactory. The zero bandgap of single layer graphene limits its possible applications in electronics even the carrier mobility of graphene can be reached up to 10 6 cm 2 /Vs. [5][6][7][8][9] Alternatively, transition metal dichalcogenide (TMD) is another material family with layered structure. [10][11][12][13] TMD family materials are composed of strong X-M-X interlayer covalent bonds, where X indicates the transition metal Mo or W; and X represents Se, S, or Te. 14-15 Similar to graphene, the bonding between different layers is the van der Waals force, showing the weak interlayer interactions where the isolation of single layer can be achieved by standard scotch tape method. 16 Molybdenum disulfide, MoS 2 , one of the most studied TMD family materials, has been regarded as a promising candidate for field-effect transistors with relatively high on/off ratio and reasonable electron mobility. [16][17][18][19][20][21][22][23][24][25][26][27][28] With recent observation of the indirect into direct bandgap transition, MoS 2 based optoelectronic devices has attracted newest interest in optical society. [29][30][31][32][33][34][35][36][37][38] In addition, due to the atomically-thin, flexible, and biocompatible nature of MoS 2 , a completely new generation of electronic sensor devices can be envisioned [39][40][41][42][43][44][45] . Moreover, all those devices are based on individual MOSFETs, giving more demands on single transistor performance. In order to realize high-performance MoS 2 MOSFET and others, fundamental device physics of MoS 2 transistor is introduced 3 first by clearly understanding of the switching mechanism of a Schottky barrier transistor. Different approaches to reduce the contact resistance on 2D materials are reviewed and explored in the latter part of this review. Record low contact resistance and high drain current are achieved on both MoS 2 and WS 2 after effective molecule chemical doping technique.
The nature of MoS 2 transistor is a Schottky barrier transistor, where the on/off states are switched by the tuning of the Schottky barriers at contacts. 46 As shown in Figure 1, we have two metal contacts that serve as source and drain for a single MoS 2 FET, named as source barrier and drain barrier. The effective barrier heights for source and drain barriers are primarily controlled by gate and drain biases. The carriers path for n-type MoS 2 transistor has been defined from the source to drain, that the electrons would encounter the source barrier first, where the carriers would undergo a thermal-assisted tunneling process from the source metal Fermi-level to the channel. On the other hand, the electrons in the channel would go from conduction band back to drain metals. Notably, gate bias has an opposite impact on these two barriers. As the high gate bias applied, the effective barrier height for source barrier, Φ s , is reduced due to a sharper triangle at the source end, where the effective Schottky barrier height has been shrank. However, lowering the conduction band at the large gate bias also enhances the Φ d , the effective Schottky barrier height at the drain end. With further increase of the drain bias, the barrier at the source end remains constant. However, the drain barrier starts to vanish 4 with large magnitude of drain bias, facilitating electron carriers movement from the source to the drain.
Field-effect transistor built on ultra-thin few-layer MoS 2 is effectively the ultra-thin body FET, which has an optimal structure to immune the short channel effects. [47][48][49] Moreover, the heavier effective mass of the MoS 2 allows its transistors to have increased drive current, and enhanced transconductance when benchmarked against the ultrathin body Si transistors at their scaling limit. 50 Previous studies of MoS 2 transistors channel length scaling has aggressively pushed the channel length down to 50 nm, 51 where the device has demonstrated an inspiring characteristic in driving current, as shown in Figure 2(a).
However, drain current saturation at the short channel regions had also been observed, which is directly attributed to the large contact resistance. The substantial contact resistance does not scale with channel length but remains almost same in the devices. As the channel length scales down to short channel regime, the channel resistance becomes comparable to the sum of two contact resistances. With further decrement of channel length, however, would not result in a significant improvement of drain current, where the drain voltage has been mainly applied on the two contacts. 19,51 Maximum drain current varies with different channel lengths of MoS 2 FETs had been reported in Figure   2(b). In the long channel regime, MoS 2 transistors have followed the classical square-law model that the drain current is inversely proportional to the channel length, I d ~ 1/L ch .
With continuous channel length scaling down, the driving current starts to have a saturation at ~90 mA/mm at L = 100nm, which is due to the dominant contact resistance 5 at short channel regime, indicating the sum of two contact resistances is comparable or even larger than the channel resistance.
Although, MoS 2 has attracted great interest for transistor applications because its large bandgap allows for field effect devices with low off-current, however, one key bottleneck in MoS 2 based device is the realization of the low-resistivity Ohmic contact. The on-state performance of the short channel MoS 2 FETs is mainly limited by its large contact resistance formed by Schottky barriers at the MoS 2 /metal interfaces. [52][53][54][55][56][57][58] In this paper, we present a review of several approaches to reduce the contact resistance of MoS 2 and other 2D TMDs field-effect transistors, and improve device performance. By systemically analyzing the contact strategy among the 2D semiconducting materials, a roadmap for future high-performance TMDs FETs with low contact resistance is nearly approached.
The first approach which is widely studied now is to choose the low workfunction contact metals. Once the metal workfunction is close to the conduction band edge of the 2D materials, low resistivity contacts are expected. However, a number of recent articles have applied large work function metals, such as Ni or Au, as the contact metals on MoS 2 field-effect transistors, and yet reported decent n-type contact formation and drain current. 16 where both low and high workfunction metals are aligned near the conduction band of MoS 2 , resulting in a monotonously n-type electrical characteristic. Experimental study of contacts to MoS 2 using low work function metal scandium (Sc) has been conducted to form an improved contact with MoS 2 film, which helps the electron carrier injection and to lower contact resistances for n-type MoS 2 transistors. 63 In addition, recent studies in MoS 2 contact had revealed that the device characteristics can be changed to p-type Schottky barrier FETs using extreme high workfunction contact materials. Substoichiometric molybdenum trioxide (MoO 3-x ), a high workfunction material aligned deeply into the valence band of MoS 2 , has been demonstrated as a promising p-contact for MoS 2 transistor with a moderate drain current. 64  The third approach is to heavily dope the source/drain regions of 2D materials. Heavily doped channel would significantly reduce Schottky barrier width thus reduce the contact resistance. Engineering electronic performance via doping is still in its infancy for MoS 2 .
Due to its nature of ultra-thin body structure, MoS 2 may not be doped as Si and III-V semiconductors by heavy ion implantation method; however, the ultra thin body nature allows the exploration of novel approaches, such as solid doping, 85 gas doping, 86 molecular doping, 87,88 and chemical doping. 89,90 One of organic chemicals, polyethyleneimine (PEI), has been proved to be an effective doping molecule in MoS 2 field-effect transistors. 87 The amine-rich aliphatic polymer, PEI is a widely used n-type surface dopant, for doping low dimensional nano-materials devices due to its strong electron-donating ability. [91][92][93][94] As shown in Figure 5 Moreover, this barrier height can't be efficiently modified by varying the workfunction of contact metals due to the complicated metal-to-TMD interface as described above. The difference of the R c between WS 2 and MoS 2 is due to the different alignment of the CNL 12 in the two materials. Compared with MoS 2 , the CNL in WS 2 is more close to the middle of the bandgap, resulting in a larger Schottky barrier. Without doping, it would be much harder for the electrons to inject from the metal to the semiconductor in WS 2 because the thermionic current exponentially decreases with the increasing of barrier height.
However, when the tunneling current starts to dominate the current through the M-S junction, the electron injection through the barrier becomes much easier. The effective electron density (induced by chemical doping and electrostatic doping) at V bg of 50 V is as high as 2.3 × 10 13 cm -2 and 2.9 × 10 13 cm -2 for WS 2 and MoS 2 , respectively. 95 As a result, both of the R c in the WS 2 and MoS 2 decrease significantly after doping. However, it is interesting to note that most of the electron density in WS 2 is attributed to the back gate bias rather the chemical doping because the electron density of WS 2 is determined to be only 6.0 × 10 11 cm -2 at zero back gate bias. 95 In another word, the Fermi-level (electron density) at the interface can be effectively modulated by the back gate bias.
Effective modulation via field-effect can be ascribed to the passivation of sulfur vacancy by Cl, given that the sulfur vacancy is the cause of the Fermi-level pinning on MoS 2 and WS 2 at M-S interface.
The R c of WS 2 can be significantly reduced after the Cl doping. Known as an ambipolar semiconductor, the undoped WS 2 shows large Schottky barriers for both electrons and holes, resulting an extremely large R c . 96 For such a larger Schottky barrier, it would be impractical to extract the R c by the TLM structure which is applicable to Ohmic or low resistivity contacts only. However, a simple estimation of the R c of the undoped WS 2 is on the order of 10 2 Ω·mm since the total resistance of the 100 nm device is calculated to 13 be 5×10 2 Ω·mm. After doping, an R c as low as 0.7 Ω·mm, 2-3 orders of magnitude reduction, can be extracted by linearly fitting the curve of total resistances. Figure 7(a) shows the TLM resistances of the Cl-doped WS 2 as a function of gap space at a back gate bias of 50 V. Since the low R c is achieved in WS 2 by Cl doping, high-performance WS 2 FET is expected. The output characteristics of the Cl-doped few-layer WS 2 FETs with 100 nm channel length are shown in Figure 7(b). The device exhibits promising device performance including a drain current of 380 mA/mm as well as good current saturation.
Due to a small R c , the linear region of the I ds -V ds curves shows excellent linearity. The drain current starts to saturate at V ds of 1.0 V due to the electron velocity saturation. To the best of our knowledge, such a low R c and a large drain current have never been achieved on WS 2 or other TMDs whose CNL is located in the middle of the bandgap.
With advances of its ultra-thin body, decent mobility and sizable bandgap, MoS 2 has been regarded as a typical semiconducting 2D material for the next generation channel material for thin-film transistor technology. However, one of the major road blocks for high-performance MoS 2 transistors is the exhibiting Schottky barrier at the metal/semiconductor interface thus large contact resistance. In this review, MoS 2 device physics had been firstly introduced to understand the contact involved switching mechanism in MoS 2 FETs. More importantly, contact research strategies to reduce R c on MoS 2 and other 2D TMDs transistors had been elucidated to help realization of the highperformance 2D FETs with low R c for future electronics applications.