betaFeedback

WorksPhysics and Astronomy

Two-dimensional MoSi2N4: An Excellent 2D Semiconductor for Transistors

Abstract

We report the performance of field-effect transistors (FETs), comprised of mono-layer of recently synthesized layered two-dimensional MoSi2N_4 as channel material, using the first principles quantum transport simulations. The devices' performance is assessed as per the International Roadmap for Devices and Systems (IRDS) 2020 roadmap for the year 2034 and compared to advanced silicon-based FETs, carbon nanotube-based FETs, and other promising two-dimensional materials based FETs. Finally, we estimate the figure of merits of a combinational and a sequential logic circuit based on our double gate devices and benchmark against promising alternative logic technologies. The performance of our devices and circuits based on them are encouraging, and competitive to other logic alternatives.

Introduction

The two-dimensional (2D) semiconductors are promising channel materials for future technology nodes, owing to the ultrathin thickness (<1< 1 nm), no surface dangling bonds, and sharp turn-on of the density of states (DOS) at band edges [1]–[4]. In the last five years, there has been remarkable progress on synthesizing novel 2D semiconductors and several promising device concepts [5]–[7] have been demonstrated based on them. Also, the shortest MoS2_2 transistor with 1 nm-gate length has been fabricated, which shows a near-ideal sub-threshold swing of ∼65\sim65 mV/decade and an ON/OFF current ratio of ∼106\sim10^{6} [8].

The 2D semiconductors materials library has been enriched due to their bulk, arranged in layered form, in which intra-layers bonds are strong covalent bonds and weak Van der Walls (vdW) force connects inter-layers. Advanced experimental techniques have been used to isolate their layers from bulk. But, most bulk materials are non-layered (strong covalent bonds connect all three dimensions). Thus, the exfoliation process can not create their 2D structure. Recently, Silicon has been introduced as a passivator during the CVD growth of non-layered molybdenum nitride (MoN2_2) [9]. This process results in the growth of the centimeter-scale film of layered 2D MoSi2_2N4_4.

Mono-layer (ML) MoSi2_2N4_4 has excellent mechanical, electronic, optical, and thermal properties [9]–[13]. It is also a promising photocatalyst for water splitting and CO2_2 reduction [10], [12]. It has been shown that ML-MoSi2_2N4_4 has excellent stability (far better than other 2D semiconductors), using phonon, molecular dynamics (MD) calculations, and experimental testing. Even it can be handled, processed, and stored without any protective environment, unlike black phosphorus (In ambient conditions, black phosphorus (BP) can be easily etched due to chemical degradation [14] and MoS2_2 (In moist air below 373 K, its surface starts oxidizing [15]). The Young’s modulus and breaking strength of ML-MoSi2_2N4_4 are ∼479\sim479 GPa and ∼49\sim49 GPa, respectively. These values are more than double of ML-MoS2_2. ML-MoSi2_2N4_4 is an indirect band gap semiconductor with the experimental band gap value of 1.94 eV. Its elastic constant is ∼4\sim4 times of ML-MoS2_2, and the carrier mobilities in it are ∼4\sim4 times and ∼6\sim6 times of ML-MoS2_2 (the most widely studied 2D material for FETs application [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28] [8]). Its lattice thermal conductivities [11], [12] are approximately 1.6 times silicon (Si) and much higher than other widely known 2D semiconductors [29, 30, 31] (ML-MoS2_2, As, Sb, silicene, and, ML-BP), but much lower than graphene [29, 32]. The high lattice thermal conductivity of ML-MoSi2_2N4_4 ensures a high rate of heat removal through nano-electronic devices comprised of this material. Its optical transmittance is high (∼97\sim97 %) and comparable to graphene. The metal contacts to the ML of MoSi2_2N4_4 show exceptional physical properties with a large Schottky barrier height slope parameter, outperforming most other 2D semiconductors. CMOS compatible metals (Sc and Ti) also show excellent ohmic contact to ML-MoSi2_2N4_4 with zero interfacial tunneling barrier [33, 34]. Overall, Mono-layer MoSi2_2N4_4 is an excellent semiconductor for logic applications. However, there is a need to investigate the transport properties of FETs based on ML-MoSi2_2N4_4, and performance of integrated circuits (ICs) comprised of MoSi2_2N4_4 FETs.

Here, we exploit the capabilities of maximally localized Wannier functions (MLWFs) [35, 36, 37] to model electronic structure of ML-MoSi2_2N4_4 and generate tight-binding (TB) like Hamiltonian for the targeted device dimensions. Next, we compute transport properties of n- and p-type devices based on this ML by solving coupled Poisson and Schrödinger equations in non-equilibrium Green’s functions (NEGF) formalism. We assess the performance of MoSi2_2N4_4 based FETs as per the requirements of International Roadmap for Devices and This work was supported in part by the Humboldt Foundation and in part by the Swarnajayanti Fellowship (Grant No. – DST/SJF/ETA-02/2017-18) of the Department of Science and Technology (DST), Government of India.

K. Nandan and Y. S. Chauhan are with the NanoLab, Department of Electrical Engineering, Indian Institute of Technology Kanpur, India, 208016, e-mail: [email protected], [email protected].

B. Ghosh and A. Agarwal are with the Department of Physics, Indian Institute of Technology Kanpur, India.

S. Bhowmick is with the Department of Material Science and Engineering, Indian Institute of Technology Kanpur, India. Systems (IRDS) 2020 for the year 2034 [38]. The channel length scalability of devices is also studied with their switching performance. Finally, the figure of merits (FOMs) of a combinational circuit (32-bit adder) and a sequential circuit (ALU) are estimated and benchmarked against promising logic technologies (CMOS and beyond-CMOS).

Computational Methods

Vienna Ab initio Simulation Package (VASP) [39], a tool based on DFT, is used to relax the atomic positions and calculate the electronic structure of ML-MoSi2_2N4_4. The Projector Augmented Wave (PAW) [40] pseudopotentials with plane-wave basis set are used for DFT calculations. The generalized gradient approximation (GGA) developed by Perdew-Burke-Ernzerhof (PBE) [41] is used to consider exchange and correlation effects. The energy cutoff of 400 eV is used for the plane-wave basis set, and the Brillouin zone integrations are performed using 12×12×112 \times12 \times1 k-mesh.

The obtained Bloch/plane-wave states from DFT calculations are mapped to MLWFs using the Wannier90 suite of codes [42]. The obtained TB-like Hamiltonian in MLWF basis is used to construct the Hamiltonian for targeted device dimensions. In the transverse direction (channel width direction), periodic boundary condition (PBC) is considered with 30 uniform wave-vector samples. The constructed device Hamiltonian is used as input to solve coupled Schrödinger and Poisson equations in Non-equilibrium Green’s functions (NEGF) formalism [43, 44, 45, 46]. Additional computational details are described in Appendix.

Results

Structural and Electrical Properties of ML-MoSi2_2N4_4

The two dimensional periodic replication of one Mo, two Si, and four N atoms, packed in honey-comb lattice, generate the ML of MoSi2_2N4_4 (see Figure 1 (a) and (b)). This ML can be viewed as MoN2_2 (2H − MoS2_2 like structure) sandwitched between two buckled honeycomb SiN layers (see Figure 1 (b)). The optimized lattic constant is a(=b)∼2.90a(= b) \sim2.90 Å and the thickness of ML is ∼7.01\sim7.01 Å. The optimized structural parameters agree well with the literature [47, 48] [10].

Top and side views of ML-MoSi2N4 and Brillouin zones

Figure 1. (a) Top view and (b) side view of ML-MoSi2_2N4_4. The hexagonal primitive cell and orthogonal supercell of ML-MoSi2_2N4_4 are shown by dashed magenta and red lines in (a), respectively. (c) The first Brillouin zone (FBZ) and high symmetry points associated with hexagonal primitive cell (in magenta color) and orthogonal cell (in red color). The blue dashed lines in (c) show the folding of K point associated with FBZ of hexagonal cell to FBZ of orthogonal cell.

The electronic band structure of MoSi2_2N4_4 is plotted along with the high symmetry points (Γ\Gamma-X-S-Y-Γ\Gamma-S) in the orthogonal Brillouin zone (BZ). Figure 1 (c) shows the BZ associated with hexagonal cell and orthogonal cell. Figure 2 shows the band structure of MoSi2_2N4_4 obtained from DFT and MLWF-TB Hamiltonian. The band structure from MLWF-TB Hamiltonian shows a good match with DFT near VBM and CBM. It is an indirect band gap semiconductor with the conduction band maxima (CBM) lies in the way from Γ\Gamma to X (equivalent to K in hexagonal BZ) and valence band maxima (VBM) at Γ\Gamma. The curvature of CBM and VBM are isotropic, but the curvature of CBM is larger than the curvature of VBM. Hence, larger effective mass for holes (mh∗m_h^*) than electrons (me∗m_e^*) i.e., mh∗>me∗m_h^* > m_e^* (see Table I).

Figure 2. Electronic band structure of MoSi2_2N4_4 calculated using DFT and MLWF-TB along the high symmetry path.

Device Structure and Electrical Characteristics

The Single gate (SG) and double-gate (DG) devices, using ML of MoSi2_2N4_4 as channel material, are investigated. Figure 3 (a) and 3 (b) show the schematic of SG and DG devices, respectively. The geometrical parameters of the devices are gate length (LGL_G), channel length (LChL_{Ch}), source/drain extension LS/DL_{S/D}, underlap length (UL), and oxide thickness (toxt_{ox}). For SG, tboxt_{box} is BOX layer thickness with tbox=10t_{box} = 10 nm. Silicon dioxide (SiO2_2) is used as gate oxide and BOX layer. This work aims to assess the intrinsic performance of FETs based on the mono-layer of MoSi2_2N4_4. Hence, the S/channel/D doping used in the device simulations are n+/undoped (intrinsic)/n+ and p+/undoped (intrinsic)/p+ for n- and p-type devices, respectively. The S/D is doped such that the Fermi level is 50 meV above/below the conduction/valance band for n-FET/p-FET. The metal gate work function is tuned to get Ids∼10−2 μA/μmI_{ds} \sim10^{-2}\ \mu\mathrm{A}/\mu\mathrm{m} (say it OFF-current (IOFFI_{OFF}) at Vgs=0V_{gs} = 0 V for all the simulated devices.

Schematic of single-gate and double-gate FETs

Figure 3. Schematic of (a) single gate (SG) and (b) double gate (DG) FETs. BOX layer and gate oxides are SiO2_2, of thickness tboxt_{box} and toxt_{ox}, respectively. LGL_G, ULUL, and LChL_{Ch} (=LG+2×UL= L_G + 2 \times UL) are gate length, underlap length, and channel length, respectively. LS/DL_{S/D} is source/drain extension length. The source and drain are heavily doped with donors/acceptors for n-/p-type devices. This doping results in the energy degeneracy of 50 meV.

We start the investigation by simulating SG and DG devices with LCh(=LG)=12L_{Ch}(= L_G) = 12 nm, equivalent oxide thickness (EOT) =0.60= 0.60 nm, and VDD=0.50V_{DD} = 0.50 V. The source-to-drain tunneling is negligible for the devices with LCh=12L_{Ch} = 12 nm, and the band-to-band tunneling is negligible due to the high value of band gap in this ML. These device parameters are considered according to IRDS 2020 roadmap for the year 2034. According to the roadmap, the expected channel materials are germanium (Ge) and 2D materials, and the expected devices are 2D materials based devices and FeFETs for the year 2034 [38].

Transfer characteristics of n-FET and p-FET using MoSi2N4 as channel material

Figure 4. Transfer characteristics of (a) n-FET and (b) p-FET using MoSi2_2N4_4 as channel material. LCh(=LG)=12L_{Ch}(= L_G) = 12 nm, tox=0.6t_{ox} = 0.6 nm, and VDD=0.50V_{DD} = 0.50 V. The IOFFI_{OFF} is fixed at 10−2μA/μm10^{-2}\mu\mathrm{A}/\mu\mathrm{m} for all devices.

Figure 4 shows the transfer characteristics for n-FETs and p-FETs. In both types, DG configuration shows a steeper sub-threshold slope (SS) and higher IONI_{ON} than SG. The SS for thermionic current is proportional to 1+Cq/COX1 + C_q/C_{OX} [49], where CqC_q is quantum capacitance and COXC_{OX} is oxide capacitance. For DG configuration, oxide capacitance is double than SG, hence DG devices show better SS than SG for LCh=12L_{Ch} = 12 nm. The stepper SS and higher value of COXC_{OX} in DG ensure higher d∣Ids∣/d∣Vgs∣d|I_{ds}|/d|V_{gs}| and higher mobile carrier concentration, respectively than SG, hence higher IONI_{ON} than SG. Defects and series parasitics resistance from contacts and S/D access region are not included in our simulations, usually limiting the device’s performance.

Channel length Scaling

The channel length scaling is performed to study the scalability and immunity to source-to-drain tunneling (SDT) of MoSi2_2N4_4 based FETs. The SDT plays a significant role in deteriorating device performance at short channel lengths. It is more severe for the low effective mass carriers than high effective mass, as the tunneling probability [50] is proportional to exp⁡(−m∗)\exp(-\sqrt{m^*}). Figures 5 (a), (b), (c), and (d) show the transfer characteristics of DG n-FET, DG p-FET, SG n-FET, and SG p-FET, respectively for various channel lengths LCh=12L_{Ch} = 12, 8, 5, 3 nm.

The impact of SDT is more for n-type devices than p-type because the curvature of CBM is greater than VBM. For each type, DG is more immune to SDT than SG as DG shows better gate controllability than SG. Hence, DG devices are more scalable than SG. The best scalable device is DG p-FET, and it can be scaled down to 5 nm. At LCh=5L_{Ch} = 5 nm, DG p-FET show SS ∼65\sim65 and ION/IOFF>106I_{ON}/I_{OFF} > 10^6 for LP applications; For HP applications, it shows SS ∼67\sim67 mV/decade and ION/IOFF>104I_{ON}/I_{OFF} > 10^4.

Figure 6 shows the output characteristics of DG n-FET and p-FET with LCh=5L_{Ch} = 5 nm. Figure 7 shows IONI_{ON} and SS of n-FET and p-FET for various LChL_{Ch} at three different values of IOFFI_{OFF}. Also, for a full assessment of DG devices at short channel lengths, DIBL is estimated from the change in threshold voltage (VthV_{th}) obtained by varying VdsV_{ds} from VlowV_{low} (= 50 mV) to VDDV_{DD}. Figures 8 (a) and (b) show the variations of SS and DIBL with LChL_{Ch}, respectively. Figure 9 shows the variation of IONI_{ON} vs. LChL_{Ch} for DG n- and p-FETs based on

aa (Å)hh (Å)EGE_G (eV)me∗m_e^*mh∗m_h^*VBM/CBM
MoSi2_2N4_42.907.011.8420.4781.184Γ\Gamma / Γ\Gamma-X

Table 1. Lattice parameters (a(=b)a(= b)), Electronic Band Gap (EGE_G), electron/hole effective mass (me∗/mh∗m_e^*/m_h^*) and location of CBM/VBM for ML of MoSi2_2N4_4.

Transfer characteristics of double gate and single gate FETs

Figure 5. Transfer characteristics of double gate: (a) n-FET (b) p-FET and single gate: (c) n-FET (d) p-FET for different channel lengths.

ION and sub-threshold swing of n-FET and p-FET versus channel length

Figure 7. IONI_{ON} and sub-threshold swing (SS) of n-FET and p-FET vs LChL_{Ch} for three different values of IOFFI_{OFF}.

ML-MoSi2_2N4_4 and other promising 2D materials. For both types, the ML-BP FETs show the highest IONI_{ON} [51, 52] among others (the data is not shown in the figure). But, the stability of BP remains the primary concern [14]. For n-type devices: (1) For LCh>5L_{Ch} > 5 nm, ML-As, and ML-Bi2_2O2_2Se have high IONI_{ON} than ML-MoSi2_2N4_4. However, the preparation of As can produce toxic arsenic trioxide, making its fabrication more complex, other than stability issue [53]. ML-Bi2_2O2_2Se shows good environmental stability [54]. But the short-channel effects are more in ML-As and ML-Bi2_2O2_2Se than ML-MoSi2_2N4_4. (2) ML-InSe shows the lower value of IONI_{ON} than ML-MoSi2_2N4_4 and the IONI_{ON} difference increases as we go below 5 nm channel length. However, compelling surface oxidation in InSe deteriorates mobility and causes uncontrollable current hysteresis in InSe FETs [55]. (3) The ML-MoSi2_2N4_4 shows superior ON-state current than ML-MoS2_2 FETs. For p-type devices: (1) Except InSe, the MoSi2_2N4_4 shows superior ON-state current than others for LCh<5L_{Ch} < 5 nm.

(2) The IONI_{ON} for ML-MoSi2_2N4_4 is very close to ML-Bi2_2O2_2Se and superior to others for LCh>5L_{Ch} > 5 nm. BL-Bi2_2O2_2Se suffers from high leakage current due to small bandgap (EG∼0.18E_G \sim0.18 eV), and IONI_{ON} is much lower than ML-MoSi2_2N4_4 for both n- and p-type devices.

Switching Performance of FETs

The delay and power dissipation product (PDP) are crucial figures of merits (FOMs) for logic applications. They determine switching speed and switching energy (energy per switching event), respectively. Figure. 10 (a) and (b) show the delay [τ=(QON−QOFF)/ION][\tau= (Q_{ON} - Q_{OFF})/I_{ON}] and PDP [=(QON−QOFF)VDD][= (Q_{ON} - Q_{OFF})V_{DD}] vs LChL_{Ch}, respectively for DG FETs. These switching parameters for advance Si FETs (Si FinFETs [61, 62] and Si Nanowire FETs [63, 64]) and carbon nanotube (CNT) FETs [65, 66] are also shown in Fig. 10. For MoSi2_2N4_4 FETs with LCh=3,5,8,12L_{Ch} = 3, 5, 8, 12 nm, τ\tau and PDP lies in range 0.022 − 0.082 ps and 7.33−51.7 aJ/µm, respectively. Overall, MoSi2_2N4_4 FETs switch faster with lower switching energy than advance Si FETs (0.22 − 0.48 ps, 55 − 182 aJ/µm) and CNT FETs

Output characteristics of DG n-FET and p-FET

Figure 6. Output characteristics of DG n-FET and p-FET for LCh(=LG)=5L_{Ch}(= L_G) = 5 nm.

Variation in sub-threshold swing and DIBL for double-gate devices

Figure 8. Variation in (a) SS and (b) DIBL for double-gate devices with channel length.

ON current versus channel length for DG n-FET and p-FET devices

Figure 9. ON current vs LChL_{\mathrm{Ch}} for DG (a) n-FET and (b) p-FET devices. The data for several promising 2D materials [56, 57, 58, 59] based DG devices, obtained from first-principle calculations (IOFF=10−1 μA/μmI_{\mathrm{OFF}} = 10^{-1}\,\mu\mathrm{A}/\mu\mathrm{m}), are also shown in same plot for comparison.

(0.046–0.184 ps, 26–58 aJ/μ\mum). It is worth to notice that the theoretical limit on delay for non-tunneling barrier controlled binary logic switch is 0.04 ps [63]. Our non-tunneling devices (LCh=8L_{\mathrm{Ch}} = 8 nm and 12 nm) have τ\tau in the range 0.041–0.082 ps.

The switching performance of DG MoSi2_2N4_4 based FETs (LCh=12L_{\mathrm{Ch}} = 12 nm) is benchmarked against other promising 2D materials based DG FETs [52, 56, 57, 58]. Other than 1 nm-MoS2_2 FET, the channel length of other 2D materials based FETs are in the range 7–9 nm. The five best performing devices are BP AD n-FET, InSe n-FET, InSe p-FET, MoSi2_2N4_4 n-FET, and MoSi2_2N4_4 p-FET, these have energy-delay product (EDP = PDP ×τ\times\tau) << (4×10−304 \times10^{-30} Js/μ\mum). The BP armchair direction (AD) n-FET has best switching speed among these, followed by InSe n-FET, InSe p-FET, MoSi2_2N4_4 n-FET, and MoSi2_2N4_4 p-FET. The MoS2_2 1nm-LGL_G FETs have lower switching energy than others, but p-FET and n-FET

Delay, power-delay product, and benchmarking of MoSi2N4 FETs

Figure 10. (a) τ\tau and (b) PDP vs LChL_{\mathrm{Ch}} for MoSi2_2N4_4 FETs. (c) Benchmarking of MoSi2_2N4_4 FETs against promising 2D materials based FETs. The performance of 1nm-LGL_G MoS2_2 FET is also shown for comparison.

are ∼23 times and ∼212 times slower than MoSi2_2N4_4 p-FET, respectively. The EDP of our devices are close to high performing InSe p-FET (2.64 ×10−30\times10^{-30} Js/μ\mum) and BP AD n-FET (2.15 ×10−30\times10^{-30} Js/μ\mum). The best performing device is InSe n-FET with approximately one third EDP than our devices, and EDP of BP zigzag (ZZ) n-FET is approximately three times than our devices. Others (Bi2_2O2_2Se and MoS2_2) have EDP (25–603 ×10−30\times10^{-30} Js/μ\mum) far from MoSi2_2N4_4 FETs. All the devices have IOFF=10−1 μA/μmI_{\mathrm{OFF}} = 10^{-1}\,\mu\mathrm{A}/\mu\mathrm{m}.

FOMs of 32-bit Adder and ALU

The FOMs of a combinational (32-bit adder) and a sequential logic circuit (32-bit ALU) based on MoSi2_2N4_4 FETs are estimated, using the methodology of [60, 67]. The n- and p-type DG MoSi2_2N4_4 FETs with LCh=12L_{\mathrm{Ch}} = 12 nm are considered, and are sized to deliver same current. The 15 nm metal half pitch and 60 nm contacted gate pitch are taken, which results in interconnect capacitance (CicC_{\mathrm{ic}}) = 0.0378 fF and interconnect delay (tict_{\mathrm{ic}}) = 0.1891 ps for IOFF=10−1 μA/μmI_{\mathrm{OFF}} = 10^{-1}\,\mu\mathrm{A}/\mu\mathrm{m} (HP applications).

Using these data, the τ\tau and PDP are calculated for inverter with fan-out of four (τ×PDP=0.3984\tau\times\mathrm{PDP} = 0.3984 ps ×\times 21.0382 aJ), NAND gate with fan-in of four (0.6379 ps ×\times 19.4914 aJ), 32-bit adder (111.32 ×102\times10^{2} ps ×2.1885×103\times2.1885 \times10^{3} aJ), and ALU (267.41 ps ×2.2136×104\times2.2136 \times10^{4} aJ) comprised of MoSi2_2N4_4 FETs. Figure 11 (a) and 12 (a) show τ\tau vs PDP for 32-bit adder and ALU, respectively based of MoSi2_2N4_4 and other spintronics and electronics logic devices (these data are taken from [67]). The spintronics devices (ASL, CSL, MEMTJ, mLogic, CoMET, and SWD) based adder and ALU switch slower and consumes more energy than electronics devices. Among electronics devices, switching speed of MoSi2_2N4_4 based adder and ALU is comparable to CMOS HP and ThinFET (WTe2_2/SnSe2_2 heterojunction interlayer TFET), but the switching energy of MoSi2_2N4_4 is one order less than CMOS HP and comparable to TFET devices (ThinFET and GaSb/InAs heterojunction TFET). Further, we characterize the power consumption parameters of adder and ALU. Figure 11 (b) and Figure 12 (b) show the standby power vs. active power for adder and ALU comprised of MoSi2_2N4_4 FETs and other promising electronic devices. The standby power and active power of the circuits consisting of MoSi2_2N4_4 devices are close to CMOS HP and more than TFET devices (ThinFET and GaSb/InAs heterojunction TFET).

PDP versus delay and active power versus standby power for a 32-bit adder

Figure 11. Comparison of (a) PDP vs τ\tau and (b) active power vs standby power for a 32-bit adder based on ML-MoSi2_2N4_4 FETs against other promising logic technologies [60].

Comparison of PDP versus tau and active power versus standby power for a 32-bit ALU

Figure 12. Comparison of (a) PDP vs τ\tau and (b) active power vs standby power for a 32-bit ALU based on ML-MoSi2_2N4_4 FETs against other promising logic technologies [60]. The performance of ALU comprised of MoSi2_2N4_4 FETs is comparable to CMOS-HP.

Conclusion

Two-dimensional semiconductors are promising candidates as a channel of next-generation electronic devices. In recent years, a new 2D semiconductor with the formula MoSi2_2N4_4 has been discovered and gaining attention owing to its excellent physical and electronic properties. In this work, the performance of FETs based on recently discovered ML-MoSi2_2N4_4 is assessed, using the first principles based quantum transport simulations. The upper-performance limit is reported, as the transport is assumed to be ballistic in nature. The scalability and impact of source-to-drain tunneling are investigated by performing channel length scaling study. The double-gate devices are scalable down to 5 nm. However, the p-type devices are more immune to short channel effects (SCEs) than n-type. The performance is estimated as per the IRDS roadmap for the year 2034. The key FOMs for logic switches are calculated and benchmarked against other promising 2D materials-based FETs. It is found that the switching parameters of double-gate devices are better than advanced Si-FETs and CNT-based FETs. Finally, we calculate the FOMs of a combinational and a sequential logic circuit based on our double gate devices and benchmark against CMOS and beyond-CMOS logic technologies. The performance of 32-bit adder and ALU are promising among other alternative logic technologies.

Appendix

At each bias (VgsV_{gs}, VdsV_{ds}), the transmission coefficient can be expressed as,

T(E,k,Vgs,Vds)=Trace⁡[ΓSGRΓDGA].(1)T(E,k,V_{gs},V_{ds}) = \operatorname{Trace}[\Gamma_{S}G^{R}\Gamma_{D}G^{A}]. \tag*{(1)}

Here, GRG^{R} and GAG^{A} are retarded and advance Green functions, respectively. ΓS/D\Gamma_{S/D} is the broadening from source/drain contacts. GRG^{R}, ΓL/R\Gamma_{L/R}, and GAG^{A} can be expressed as,

GR(E,k,Vgs,Vds)=[EI−H(k)−Σ]−1,(2)G^{R}(E,k,V_{gs},V_{ds}) = [EI-H(k)-\Sigma]^{-1}, \tag*{(2)}
ΓS/D(E,k,Vgs,Vds)=i[ΣS/D−ΣS/D†],(3)\Gamma_{S/D}(E,k,V_{gs},V_{ds}) = i[\Sigma_{S/D}-\Sigma^{\dagger}_{S/D}], \tag*{(3)}
GA(E,k,Vgs,Vds)=[GR(E,k,Vgs,Vds)]†(4)G^{A}(E,k,V_{gs},V_{ds}) = [G^{R}(E,k,V_{gs},V_{ds})]^{\dagger} \tag*{(4)}

where, E and I are energy and identity matrix, respectively. Σ\Sigma (=ΣS+ΣD= \Sigma_{S}+\Sigma_{D}) is the sum of source and drain contact self-energy matrix, and H(k) is the channel Hamiltonian. Next, the drain current (IdsI_{ds}) is calculated using the Landauer-Buttiker approach [68]. For a given gate-to-source voltage (VgsV_{gs}) and drain-to-source voltage (VdsV_{ds}), it can be expressed as,

Ids(Vgs,Vds)=eπℏ∫−∞∞∑kT(E,k,Vgs,Vds)[f(E−μs)−f(E−μd)]dE(5)I_{ds}(V_{gs},V_{ds}) = \frac{e}{\pi\hbar}\int_{-\infty}^{\infty} \sum_{k} T(E,k,V_{gs},V_{ds}) \left[f(E-\mu_s)-f(E-\mu_d)\right]dE \tag*{(5)}

where, ee is the electron charge, ℏ\hbar is the reduced Planck constant, T(E,Vgs,Vds)=∑kT(E,k,Vgs,Vds)T(E,V_{gs},V_{ds}) = \sum_k T(E,k,V_{gs},V_{ds}) is the transmission coefficient at Energy EE for a given bias (Vgs,Vds)(V_{gs},V_{ds}), μs/d\mu_{s/d} is the chemical potential at source/drain, and f(E−μs/d)f(E-\mu_{s/d}) is the Fermi-Dirac distribution function at source/drain.

References

  1. [16]B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, and A. Kis, “Single-layer MoS₂ transistors,” Nature Nanotechnology, vol. 6, no. 3, pp. 147–150, Mar 2011, doi: 10.1038/nnano.2010.279.DOI
  2. [17]A. Nourbakhsh, A. Zubair, R. N. Sajiad, A. Tavakkoli K. G., W. Chen, S. Fang, X. Ling, J. Kong, M. S. Dresselhaus, E. Kaxiras, K. K. Berggren, D. Antoniadis, and T. Palacios, “MoS₂ field-effect transistor with sub-10 nm channel length,” Nano Letters, vol. 16, no. 12, pp. 7798–7806, 2016, doi: 10.1021/acs.nanolett.6b03999.DOI
  3. [18]F. Zhang and J. Appenzeller, “Tunability of short-channel effects in MoS₂ field-effect devices,” Nano Letters, vol. 15, no. 1, pp. 301–306, 2015, doi: 10.1021/nl503586v.DOI
  4. [19]J. Lee, J. N. Kim, W. T. Kang, Y. S. Shin, B. H. Lee, and W. J. Yu, “Schottky barrier variable graphene/multilayer-MoS₂ heterojunction transistor used to overcome short channel effects,” ACS Applied Materials & Interfaces, vol. 12, no. 2, pp. 2854–2861, 2020, doi: 10.1021/acsami.9b18577.DOI
  5. [20]D. Sarkar, W. Liu, X. Xie, A. C. Anselmo, S. Mitragotri, and K. Banerjee, “MoS₂ field-effect transistor for next-generation label-free biosensors,” ACS Nano, vol. 8, no. 4, pp. 3992–4003, 2014, doi: 10.1021/nn5009148.DOI
  6. [21]W. Liu, D. Sarkar, J. Kang, W. Cao, and K. Banerjee, “Impact of contact on the operation and performance of back-gated monolayer MoS₂ field-effect-transistors,” ACS Nano, vol. 9, no. 8, pp. 7904–7912, 2015, doi: 10.1021/nn506512j.DOI
  7. [22]X. Xie, D. Sarkar, W. Liu, J. Kang, O. Marinov, M. J. Deen, and K. Banerjee, “Low-frequency noise in bilayer MoS₂ transistor,” ACS Nano, vol. 8, no. 6, pp. 5633–5640, 2014, doi: 10.1021/nn4066473.DOI
  8. [23]B. Chakraborty, A. Bera, D. V. S. Muthu, S. Bhowmick, U. V. Waghmare, and A. K. Sood, “Symmetry-dependent phonon renormalization in monolayer MoS₂ transistor,” Phys. Rev. B, vol. 85, p. 16403, Apr 2012, doi: 10.1103/PhysRevB.85.16403.DOI
  9. [24]Y. Yoon, K. Ganapathi, and S. Salahuddin, “How good can monolayer MoS₂ transistors be?” Nano Letters, vol. 11, no. 9, pp. 3768–3773, 2011, doi: 10.1021/nl2018178.DOI
  10. [25]P. Rastogi, S. Kumar, S. Bhowmick, A. Agarwal, and Y. S. Chauhan, “Doping strategies for monolayer MoS₂ via surface adsorption: A systematic study,” The Journal of Physical Chemistry C, vol. 118, no. 51, pp. 30309–30314, Dec 2014, doi: 10.1021/jp510662n.DOI
  11. [26]C. Yadav, A. Agarwal, and Y. S. Chauhan, “Compact modeling of transition metal dichalcogenide based thin body transistors and circuit validation,” IEEE Transactions on Electron Devices, vol. 64, no. 3, pp. 1261–1268, 2017, doi: 10.1109/TED.2016.2643698.DOI
  12. [27]C. Yadav, P. Rastogi, T. Zimmer, and Y. S. Chauhan, “Charge-based modeling of transition metal dichalcogenide transistors including ambipolar, trapping, and negative capacitance effects,” IEEE Transactions on Electron Devices, vol. 65, no. 10, pp. 4202–4208, 2018, doi: 10.1109/TED.2018.2855109.DOI
  13. [28]K. Nandan, C. Yadav, P. Rastogi, A. Toral-Lopez, A. Marin-Sanchez, E. G. Marin, F. G. Ruiz, S. Bhowmick, and Y. S. Chauhan, “Compact modeling of multi-layered MoS₂ FETs including negative capacitance effect,” IEEE Journal of Electron Devices Society, vol. 8, pp. 1177–1183, 2020, doi: 10.1109/JEDS.2020.3021031.
  14. [29]B. Peng, D. Zhang, H. Zhang, H. Shao, G. Ni, Y. Zhu, and H. Zhu, “The conflicting role of buckled structure in phonon transport of 2D group-IV and group-V materials,” Nanoscale, vol. 9, pp. 7397–7407, 2017, doi: 10.1039/C7NR00838D.DOI
  15. [30]G. Qin, Q.-B. Yan, Z. Qin, S.-Y. Yue, M. Hu, and G. Su, “Anisotropic intrinsic lattice thermal conductivity of phosphorene from first principles,” Phys. Chem. Chem. Phys., vol. 17, pp. 4854–4858, 2015, doi: 10.1039/C4CP04858J.DOI
  16. [31]W. Li, J. Carrete, and N. Mingo, “Thermal conductivity and phonon linewidths of monolayer MoS₂ from first principles,” Applied Physics Letters, vol. 103, no. 25, p. 253103, 2013, doi: 10.1063/1.4850995.DOI
  17. [32]B. D. Dong, S. Paul, M. B. Nardelli, and K. W. Kim, “First-principles analysis of lattice thermal conductivity in monolayer and bilayer graphene,” Phys. Rev. B, vol. 80, p. 033406, Jul 2009, doi: 10.1103/PhysRevB.80.033406.
  18. [33]Q. Wang, L. Cao, S.-J. Liang, W. Wu, G. Wang, C. H. Lee, W. L. Ong, H. Y. Yang, L. Kang, S. A. Yang, and Y. S. Ang, “Efficient ohmic contacts and built-in atomic sublayer protection in MoSi₂N₄ and WSi₂N₄ monolayers,” npj 2D Materials and Applications, vol. 5, no. 1, p. 71, Aug 2021, doi: 10.1038/s41699-021-00251-y.DOI
  19. [34]Q. Wu, L. Cao, Y. S. Ang, and L. K. Ang, “Semiconductor-to-metal transition in bilayer MoSi₂N₄ and WSi₂N₄ with strain and electric field,” Applied Physics Letters, vol. 118, no. 11, p. 113102, 2021, doi: 10.1063/5.0044431.DOI
  20. [35]I. Souza, N. Marzari, and D. Vanderbilt, “Maximally localized Wannier functions for entangled energy bands,” Phys. Rev. B, vol. 65, p. 035109, Dec 2001, doi: 10.1103/PhysRevB.65.035109.DOI
  21. [36]Y.-S. Lee, M. B. Nardelli, and N. Marzari, “Band structure and quantum conductance of nanostructures from maximally localized Wannier functions: The case of functionalized carbon nanotubes,” Phys. Rev. Lett., vol. 95, p. 076804, Aug 2005, doi: 10.1103/PhysRevLett.95.076804.DOI
  22. [37]N. Marzari, A. A. Mostofi, J. R. Yates, I. Souza, and D. Vanderbilt, “Maximally localized Wannier functions: Theory and applications,” Rev. Mod. Phys., vol. 84, pp. 1419–1475, Oct 2012, doi: 10.1103/RevModPhys.84.1419.DOI
  23. [38]IRDS 2020: https://irds.ieee.org/editions/2020.
  24. [39]G. Kresse and J. Furthmüller, “Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set,” Phys. Rev. B, vol. 54, pp. 11169–11186, Oct 1996, doi: 10.1103/PhysRevB.54.11169.DOI
  25. [40]P. E. Blöchl, “Projector augmented-wave method,” Phys. Rev. B, vol. 50, pp. 17953–17979, Dec 1994, doi: 10.1103/PhysRevB.50.17953.DOI
  26. [41]J. P. Perdew, K. Burke, and M. Ernzerhof, “Generalized gradient approximation made simple,” Phys. Rev. Lett., vol. 77, pp. 3865–3868, Oct 1996, doi: 10.1103/PhysRevLett.77.3865.DOI
  27. [42]A. A. Mostofi, J. R. Yates, G. Pizzi, Y.-S. Lee, I. Souza, D. Vanderbilt, and N. Marzari, “An updated version of Wannier90: A tool for obtaining maximally-localised Wannier functions,” Computer Physics Communications, vol. 185, no. 8, pp. 2309 – 2310, 2014, doi: 10.1016/j.cpc.2014.05.003.DOI
  28. [43]S. Datta, Quantum transport: atom to transistor. Cambridge university press, 2005.DOI
  29. [44]P. Rastogi, T. Dutta, S. Kumar, A. Agarwal, and Y. S. Chauhan, “Quantum confinement effects in extremely thin body germanium n-MOSFETs,” IEEE Transactions on Electron Devices, vol. 62, no. 11, pp. 3575–3580, 2015, doi: 10.1109/TED.2015.2477471.DOI
  30. [45]T. Dutta, P. Kumar, P. Rastogi, A. Agarwal, and Y. S. Chauhan, “Atomistic study of band structure and transport in extremely thin channel InP MOSFETs,” physica status solidi (a), vol. 213, no. 4, pp. 898–904, 2016, doi: 10.1002/pssa.201532727.DOI
  31. [46]K. Nandan, A. Agarwal, S. Bhowmick, and Y. S. Chauhan, “Performance investigation of p-FETs based on highly air-stable monolayer pentagonal PdSe2,” IEEE Transactions on Electron Devices, pp. 1–7, 2021, doi: 10.1109/TED.2021.3119552.DOI
  32. [47]S. Li, W. Wu, X. Feng, S. Guan, W. Feng, Y. Yao, and S. A. Yang, “Valley-dependent properties of monolayer MoSi2N4, WSi2N4, and MoSi2As4,” Phys. Rev. B, vol. 102, p. 235435, Dec 2020, doi: 10.1103/PhysRevB.102.235435.
  33. [48]Y.-L. Hong, Z. Liu, L. Wang, T. Zhou, W. Ma, C. Xu, S. Feng, L. Chen, M.-L. Chen, D.-M. Sun, X.-Q. Chen, H.-M. Cheng, and W. Ren, “Chemical vapor deposition of layered two-dimensional MoSi2N4 materials,” Science, vol. 369, no. 6504, pp. 670–674, 2020, doi: 10.1126/science.abb7023.DOI
  34. [49]M. Lundstrom and J. Guo, Nanoscale transistors: device physics, modeling and simulation. Springer Science & Business Media, 2006.DOI
  35. [50]D. J. Griffiths and D. F. Schroeter, Introduction to quantum mechanics. Cambridge University Press, 2018.
  36. [51]D. Yin and Y. Yoon, “Can bilayer black phosphorus outperform monolayer in field-effect transistors?” in 2015 73rd Annual Device Research Conference (DRC), ser. 2015 73rd Annual Device Research Conference (DRC), 2015, doi: 10.1109/DRC.2015.7175614. . ISSN 1548-3770 pp. 177–178DOI
  37. [52]F. Liu, Y. Wang, X. Liu, J. Wang, and H. Guo, “Ballistic transport in monolayer black phosphorus transistors,” IEEE Transactions on Electron Devices, vol. 61, no. 11, pp. 3871–3876, 2014, doi: 10.1109/TED.2014.2353213.DOI
  38. [53]S. Zhang, S. Guo, Z. Chen, Y. Wang, H. Gao, J. Gómez-Herrero, P. Ares, F. Zamora, Z. Zhu, and H. Zeng, “Recent progress in 2D group-VA semiconductors: from theory to experiment,” Chem. Soc. Rev., vol. 47, pp. 982–1021, 2018, doi: 10.1039/C7CS00125H.DOI
  39. [54]J. Wu, H. Yuan, M. Meng, C. Chen, Y. Sun, Z. Chen, W. Dang, C. Tan, Y. Liu, J. Yin, Y. Zhou, S. Huang, H. Q. Xu, Y. Cui, H. Y. Hwang, Z. Liu, Y. Chen, B. Yan, and H. Peng, “High electron mobility and quantum oscillations in non-encapsulated ultrathin semiconducting Bi2O2Se,” Nature Nanotechnology, vol. 12, no. 6, pp. 530–534, Jun 2017, doi: 10.1038/nnano.2017.43.DOI
  40. [55]P.-H. Ho, Y.-R. Chang, Y.-C. Chu, M.-K. Li, C.-A. Tsai, W.-H. Wang, C.-H. Ho, C.-W. Chen, and P.-W. Chiu, “High-mobility InSe transistors: The role of surface oxides,” ACS Nano, vol. 11, no. 7, pp. 7362–7370, 2017, doi: 10.1021/acsnano.7b03531.DOI
  41. [56]R. Quhe, J. Liu, J. Wu, J. Yang, Y. Wang, Q. Li, T. Li, Y. Guo, J. Yang, H. Peng, M. Lei, and J. Lu, “High-performance sub-10 nm monolayer Bi2O2Se transistors,” Nanoscale, vol. 11, pp. 532–540, 2019, doi: 10.1039/C8NR08852G.DOI
  42. [57]Y. Wang, R. Fei, R. Quhe, J. Li, H. Zhang, X. Zhang, B. Shi, L. Xiao, Z. Song, J. Yang, J. Shi, F. Pan, and J. Lu, “Many-body effect and device performance limit of monolayer InSe,” ACS Applied Materials & Interfaces, vol. 10, no. 27, pp. 23 344–23 352, 2018, doi: 10.1021/acsami.8b06427.DOI
  43. [58]J. Yang, R. Quhe, Q. Li, S. Liu, L. Xu, Y. Pan, H. Zhang, X. Zhang, J. Li, J. Yan, B. Shi, H. Pang, L. Xu, Z. Zhang, J. Lu, and J. Yang, “Sub 10 nm bilayer Bi2O2Se transistors,” Advanced Electronic Materials, vol. 5, no. 3, p. 1800720, 2019, doi: https://doi.org/10.1002/aelm.201800720.
  44. [59]Y. Wang, P. Huang, M. Ye, R. Quhe, Y. Pan, H. Zhang, H. Zhong, J. Shi, and J. Lu, “Many-body effect, carrier mobility, and device performance of hexagonal arsenene and antimonene,” Chemistry of Materials, vol. 29, no. 5, pp. 2191–2201, 2017, doi: 10.1021/acs.chemmater.6b04909.DOI
  45. [60]D. E. Nikonov and I. A. Young, “Benchmarking of beyond-cmos exploratory devices for logic integrated circuits,” IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, vol. 1, pp. 3–11, 2015, doi: 10.1109/JXCDC.2015.2418033.DOI
  46. [61]D. A. Antoniadis, I. Aberg, C. N. Cleirigh, O. M. Nayfeh, A. Khaki fiooz, and J. L. Hoyt, “Continuous MOSFET performance increase with device scaling: The role of strain and channel material innovations,” IBM Journal of Research and Development, vol. 50, no. 4.5, pp. 363–376, 2006, doi: 10.1147/rd.504.0363.DOI
  47. [62]B. Doris, M. Ieong, T. Zhu, Y. Zhang, M. Steen, W. Natzle, S. Callegari, V. Narayanan, J. Cai, S. H. Ku, P. Jamison, Y. Li, Z. Ren, V. Ku, T. Boyd, T. Kanarsky, C. D’Emic, M. Newport, D. Dobuzinsky, S. Deshpande, J. Petrus, R. Jammy, and W. Haensch, “Device design considerations for ultra-thin SOI MOSFETs,” in IEEE International Electron Devices Meeting 2003, ser. IEEE International Electron Devices Meeting 2003, 2003, doi: 10.1109/IEDM.2003.1296360. pp. 27.3.1–27.3.4.DOI
  48. [63]R. W. Keyes, “Fundamental limits in digital information processing,” Proceedings of the IEEE, vol. 69, no. 2, pp. 267–278, 1981, doi: 10.1109/PROC.1981.11959.DOI
  49. [64]H. Iwai, “End of the scaling theory and Moore’s law,” in 2016 16th International Workshop on Junction Technology (IWJT), ser. 2016 16th International Workshop on Junction Technology (IWJT), 2016, doi: 10.1109/IWJT.2016.7486661. pp. 1–4.DOI
  50. [65]S. E. Thompson and S. Parthasarathy, “Moore’s law: the future of Si microelectronics,” Materials Today, vol. 9, no. 6, pp. 20–25, 2006, doi: 10.1016/S1369-7021(06)71539-5.DOI
  51. [66]A. Javey, J. Guo, Q. Wang, M. Lundstrom, and H. Dai, “Ballistic carbon nanotube field-effect transistors,” Nature, vol. 424, no. 6949, pp. 654–657, Aug 2003, doi: 10.1038/nature01797.DOI
  52. [67]C. Pan and A. Naeemi, “An expanded benchmarking of beyond-CMOS devices based on boolean and neuromorphic representative circuits,” IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, vol. 3, pp. 101–110, 2017, doi: 10.1109/JXCDC.2018.2793536.DOI
  53. [68]M. Büttiker, Y. Imry, R. Landauer, and S. Pinhas, “Generalized many-channel conductance formula with application to small rings,” Phys. Rev. B, vol. 31, pp. 6207–6215, May 1985, doi: 10.1103/PhysRevB.31.6207.DOI

Paper details

vlnc:W_ya92seek
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents