Investigation on Temperature Dependency of Recessed-Channel Reconfigurable Field-Effect Transistor

Abstract

Current-voltage (I-V) characteristics of a recessed-channel reconfigurable field-effect transistor (RC-RFET) is discussed, herein, depending on the variation of temperature (T) to understand the operation mechanisms, in depth. Assuming that RC-RFET can be simply modeled as a channel resistance (R_CH) and a Schottky contact resistance (R_SC) connected in series, the validity has been examined by a technology computer-aided design (TCAD) simulation with different Schottky barrier heights (SBHs) and carrier mobilities (μ). As a result, it was clearly determined that the drain current (I_D) of RC-RFET is dominated by the bigger component, since R_CH and R_SC have an opposite correlation with T.

1. Introduction

Over the past five decades metal-oxide-semiconductor field-effect transistors (MOSFETs) have been aggressively scaled down for a low-power operation with high performance and an enhanced logic functionality with large integration density [1,2]. However, below the 10 nm node, the extension of Moore’s law by aggressive scaling of FETs becomes increasingly difficult due to several technical issues [2]. Therefore, there have been a lot of efforts for an appropriate successor to the conventional complementary MOS (CMOS) technology especially based on novel devices [3]. On the other hand, a reconfigurable FET (RFET) has been regarded as another candidate to address the issues by extending the logic functionality of switching elements [4]. It features dynamically programmable operations which allows an integrated circuit (IC) to reduce the required devices for a similar logic function result in circuit-level scaling down [4,5,6,7,8,9]. Although there are several studies about RFETs, most of them have mainly focused on strategies for improving electrical performance with the help of geometrical device structures, materials, etc. [4,5,6,7,10,11] and there is still a lack of understanding about their operation mechanisms. RFETs are programmable as n- and p-FETs by selecting carrier types injected from Schottky contact at source which depends on the temperature (T), sensitively [12]. Therefore, a rigorous study about the T characteristics of RFETs is an important research topic. In this letter, the T dependent current-voltage (I-V) characteristic is discussed and analyzed by technology computer-aided design (TCAD) simulation [13].

2. Device Structure and Simulation

In this work, a novel recessed-channel RFET (RC-RFET), which was proposed in [11] to improve scalability and short-channel-effect immunity of conventional RFET, was used for the study (Figure 1) [14,15]. The switching mechanism of RC-RFET (thermionic emission) differs from that of conventional RFET (Schottky barrier tunneling). It enables RC-RFET to overcome the fundamental limit of subthreshold swing (S) degradation as a function of gate voltage and promises higher ON–OFF current ratio (I_ON/I_OFF) [11]. The detail parameters used for the TCAD simulation are summarized in

Table 1. Simulated device parameters.

Definition	Abbreviation	Value
Silicon body thickness	TB	20 nm
Gate oxide thickness Gate length	TOX LG	1 nm 50 nm
Program gate thickness Control gate thickness Oxide thickness between PG and CG	TPG TCG TGAP	20 nm 20 nm 10 nm

and the following physical models are used: Shockley–Read–Hall (SRH) recombination, Schottky barrier tunneling (SBT), field-dependent mobility, drift-diffusion, and non-local band-to-band tunneling (BTBT) [13]. In order to exclude the effect of dopants and secure symmetricity for n- and p-FET operations, intrinsic silicon-on-insulator (SOI) is used for channel. The work function (W_FN) for both the polarity gate (PG) and control gate (CG) is 4.6 eV, while Schottky barrier height (SBH) for Si/metal contact is 0.56 eV unless otherwise noted. They are analogous to half of the Si band gap.

3. Results and Discussion

Figure 2 shows drain current (I_D) versus CG bias (V_CGS) for n- and p-FET operations as T increases from 300 K to 400 K with different PG bias (V_PGS). As shown in Figure 2a,b, the on-current (I_ON) defined as I_D at ±1.5 V-V_CGS is decreased for both n- and p-FETs as T increases in accordance with general expectation, since the mobility (μ) is decreased due to phonon scattering as expressed in Equation (1) [16]. The μ at 300 K-T (μ₀) and γ are 1417 cm²/V·s and 2.5 for electron, whereas 486 cm²/V·s and 2.2 for hole, respectively [13].

$$\mu ={\mu }_0{\left(\frac{T}{300\mathrm{K}}\right)}^{-\gamma }$$

(1)

The noteworthy points are shown in Figure 2c,d that the I_ON becomes an increasing function of the T as |V_PGS| is decreased from 5 to 2 V.

Figure 3a clearly shows I_ON has different tendencies on the T depending on the V_PGS, regardless of n- or p-FETs. It is not related to the subthreshold characteristics since the extracted S is exactly sitting on the 2.3k_B/q-slope (~0.2 mV/K) line similar to the conventional MOSFETs, where k_B and q are the Boltzmann constant and elementary charge, respectively (Figure 3b).

In order to analyze these phenomena, the RC-RFET is simply modeled as a channel resistance (R_CH) and a Schottky contact resistance (R_SC) connected in series, which have an opposite correlation with T. In other words, if T is increased, the R_CH is increased due to the increased phonon scattering, whereas the R_SC is decreased due to the increased carrier injection ‘over’ and ‘through’ the Schottky barriers at source and drain contacts (Figure 4). Accordingly, the total resistance (R_TOT) and I_ON is dominated by the bigger one. Since R_SC is exponentially decreased as a function of band bending, the R_TOT can be approximated to R_CH and R_SC with a small and a large |V_PGS|, respectively. From Equation (1), the μ of electron (μ_e) is more sensitive to T than that of the hole (μ_h) due to the different coefficient γ. Thus, R_CH of n-FET increases faster than that of p-FET as T increase which is well corresponded to the different slopes for 5 V-|V_PGS| in Figure 3a. On the other hand, in case of small |V_PGS| (= 2V), p-FET is more sensitive than n-FET to the T. This is because the effect of R_SC decreasing is cancelled out by the R_CH increase with the higher T for n-FET.

In order to confirm the hypothesis, the effects of R_SC and R_CH on n-FETs’ I_ON have been examined independently. For that, the change of doping concentration is inappropriate since both factors are affected at the same time. Thus, the R_SC and the R_CH of n-FET are changed by adjusting SHB and μ_e, respectively. First, if SBH decreases from 1.06 eV to 0.26 eV, the dominant factor changes from R_SC to R_CH; I_ON decreases as T increases as shown in Figure 5. On the other hand, if μ_e increases several times the default value (1417 cm²/V·s), the R_TOT is determined by the R_SC (Figure 6). As a result, there is a positive correlation between I_ON and T. Consequently, the assumption is clearly proven to be reasonable. It has to be mentioned that even if this approach is less practical in terms of device design, it is very meaningful to understand and confirm the operation mechanism and overall physics of RFET which is a milestone for device engineers.

4. Conclusions

The effects of T on the electrical characteristics of RC-RFET have been examined. There is an abnormal phenomenon that the I_ON decreases as T increases with a small |V_PGS|. Based on rigorous study with the help of TCAD simulation, this was attributed to the large R_SC which decreases the function of T and dominates the R_TOT. This needs to be further addressed for high-performance operation with low-power consumption. In order to decrease SBH for the lower R_CH, adopting a narrow band gap material (e.g., SiGe or Ge) at the source/drain and metal contacts could be a promising solution. In future works, the influences of band gap on SBH will be examined and the optimized RC-RFET will be demonstrated.