Mobility Deception in Nanoscale Transistors: An Untold Contact Story

Abstract

and energy harvesting and energy storage capabilities. Some of these materials are also providing opportunities for non-von Neumann and beyond Boltzmann computation, as well as offering unique platforms for hardware-based artificial intelligence and cyber security. [9][10][11][12] Although these materials are unlikely to cost-effectively compete with the performance level of Si-based complementary metal oxide semiconductor (CMOS) technologies, FET characteristics are routinely used to benchmark these novel materials and the extracted field-effect mobility (µ FE ) values are often used to champion the respective materials. The most widespread technique for such mobility extraction is the use of the peak transconductance (g max ) value, which represents the maximum slope of the source to drain current (I DS ) versus applied gate voltage (V GS ) characteristics in the ON-state of the FET operation. [13] This technique was primarily developed, tested, and justified for Si-based long channel FETs with contacts that are heavily doped, therefore not affecting the carrier transport. However, FETs based on the aforementioned novel nanomaterials are often limited by various intrinsic, extrinsic, and device geometry related contact effects that can manifest in the mobility extraction based on the g max technique. [14][15][16][17][18][19][20] The common assumption is that contacts can only play a detrimental role in carrier transport and, therefore, always underestimate the µ FE extracted from the g max value. Moreover, sandwiching these novel nanomaterials with a high-k dielectric layer (HfO 2 or Al 2 O 3 ) and a top-gate (TG) electrode can often lead to an overestimated mobility extraction due to improperly neglecting the coupling between the top-gate and the bottom-gate as mentioned by Fuhrer and Hone. [17,21,22] In fact, keeping the top-gate electrode disconnected can result in an increase of capacitive coupling by a factor of 53 according to Radisavljevic and Kis, which will lead to a large overestimation in the mobility value. [22] However, in this work, we experimentally demonstrate, and numerically validate, that contact effects can also lead to mobility overestimation under circumstances that are prevalent in FETs based on novel nanomaterials. Figure 1a shows the schematic of a standard back-gated (BG) FET geometry based on 2D layered semiconductors. Compared to a conventional BG Si FET shown in Figure 1b, there are two key differences: 1) in Si FETs, the metal/Si interfaces are mostly Ohmic in nature owing to the formation of metal

Stochastic resonance in MoS2 photodetector

“…For example, monolayer TMDCs are direct bandgap semiconductors, fostering enhanced light-matter interaction and thereby enabling photonic devices 31 . Furthermore, the atomically thin body nature of TMDCs allows for superior electrostatic control of charge carriers, which is critical for achieving low power operation 32 . While most of the preliminary device studies are performed on micromechanically exfoliated single-crystal flakes, transitioning to technology requires manufacturable solutions.…”

Schottky barrier lowering due to interface states in 2D heterophase devices

“…The first value corresponds to the estimated p-doping level reported by Sung et al 6 and the second value is comparable with more recent estimated doping levels in 1H phase TMDs. 25,26 The doping of a 2D material is extremely difficult to control and even to measure. Since almost the entire material is a surface, it is very sensitive to both the environment and local impurities.…”

Scalable Substitutional Re‐Doping and its Impact on the Optical and Electronic Properties of Tungsten Diselenide

“…[ 35,60,61 ] All three BGFETs demonstrate ON/OFF ratio >10 6 and extracted carrier mobility from the peak transconductance method equals 1 cm 2 V −1 s −1 which is highly underestimated due to the large Schottky barrier height that limits the ON‐state current. [ 62,63 ] Furthermore, device performance degrades as the doping concentration is increased up to 1.1% Re‐WSe 2 . Highly doped (1.1%) samples exhibit an order of magnitude decrease in ON current and complete loss in gate modulation where the BGFET no longer acts as a semiconductor (Figure 4d).…”