Materials, Processes and Device Design for High Performance, Sub-thermionic MoS2 FETs
Abstract
The shrinking of the field-effect transistor (FET), commonly termed CMOS scaling, has revolutionised the semiconductor industry and impacted most aspects of human life. However, this decades-long successful trend of scaling the FET is now facing serious fundamental and technological challenges resulting in diminishing economic returns. The first challenge is that, as device dimensions shrink, electric fields in close proximity start to interfere with each other and disrupt the transistor’s operations. This phenomenon is called the short channel effect (SCE). Second, concurrent (quadratic) reduction in power consumption, an important aspect of scaling, is not possible because of the inability to reduce supply voltage below 1 V. This is primarily because the fundamental nature of charge transport governed by Boltzmann’s statistics restricts the sub-threshold swing (SS, abruptness between OFF to ON transitions) of FETs to the thermionic limit of 60 mV/dec at room temperature. Hence, as we cram in more transistors into the same footprint, energy dissipation and heat management have become fundamental bottlenecks. Clearly, the road ahead needs breakthroughs in new materials and device design.
In this thesis, we attempt to tackle both these challenges (SCE and power consumption) by developing high performance, sub-thermionic (SS<60 mV/dec) transistors on 2D semiconductors. Excellent electrostatics inherent to 2D semiconductors make them promising candidates for mitigating SCE, with recent demonstrations of sub-10 nm channel length molybdenum disulphide (MoS2) FETs. However, the development of sub-thermionic transistors on 2D semiconductors has been stymied by inefficient contacts, doping and dielectric integration, which motivates the initial portion of this work. First, we demonstrate how a facile sulfur-based solution can be employed to engineer surface states between MoS2/metal contacts, resulting in substantial reduction in Schottky barrier height. This enables record performance due to a ~6x/10x (Ni/Pd) reduction in contact resistance, along with complete mitigation in contact variability. In addition, through an empirical model we can predict the intrinsic limit of contact resistance in multilayer TMDs. Second, we develop an air-stable, area-selective, CMOS-compatible counterdoping strategy through vacancy engineering, enabling p-FETs and in-plane p/n junction photodioides. Third, we lay a two-part focus on dielectrics. In the first part, we examine the interplay of different scattering mechanisms in 2D charge transport and demonstrate ‘ideal’ nitride dielectric environments for large performance gains (field-effect mobility ~ 73 cm2V-1s-1). In the second part, we
Page | VII
break the paradigm of atomic layer deposition and employ functionalisation-free e-beam evaporated top gate high-κ HfO2 to achieve near-perfect SS~60 mV/dec operation. Finally, leveraging on the insights from the previous sections, we demonstrate, for the first time, sub-thermionic transport through tunable Schottky contacts in dual gated MoS2 FETs. Two device configurations, the GT3FET and DVATFET, are expounded. The GT3FET has the flexibility to operate either in the sub-thermionic tunnel regime, yielding steep SS<60 mV/dec OR thermionic high mobility regime. Combining the best of both tunnelling and thermionic regimes in the same operation cycle, the DVATFET, the closest to an ‘ideal transistor’, registers SS~29 mV/dec (3 dec) AND high mobility (100 cm2V-1s-1). This work is envisioned to pave a new path in the development of sub-thermionic, high performance FETs operating in the sub-0.5 V ‘green computing’ regime.