Ruge
Quhe
*a,
Junchen
Liu
a,
Jinxiong
Wu
b,
Jie
Yang
c,
Yangyang
Wang
d,
Qiuhui
Li
a,
Tianran
Li
b,
Ying
Guo
e,
Jinbo
Yang
cf,
Hailin
Peng
b,
Ming
Lei
*a and
Jing
Lu
*cf
aState Key Laboratory of Information Photonics and Optical Communications and School of Science, Beijing University of Posts and Telecommunications, Beijing 100876, P. R. China. E-mail: mlei@bupt.edu.cn; quheruge@bupt.edu.cn
bCenter for Nanochemistry, Beijing National Laboratory for Molecular Sciences (BNLMS), College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, P. R. China
cState Key Laboratory for Mesoscopic Physics and Department of Physics, Peking University, Beijing 100871, P. R. China. E-mail: jinglu@pku.edu.cn
dNanophotonics and Optoelectronics Research Center, Qian Xuesen Laboratory of Space Technology, China Academy of Space Technology, Beijing 100094, P. R. China
eSchool of Physics and Telecommunication Engineering and Shaanxi Key Laboratory of Catalysis, Shaanxi University of Technology, Hanzhong 723001, P. R. China
fCollaborative Innovation Center of Quantum Matter, Beijing 100871, P. R. China
First published on 3rd December 2018
A successful two-dimensional (2D) semiconductor successor of silicon for high-performance logic in the post-silicon era should have both excellent performance and air stability. However, air-stable 2D semiconductors with high performance were quite elusive until the air-stable Bi2O2Se with high electron mobility was fabricated very recently (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, Nat. Nanotechnol., 2017, 12, 530). Herein, we predict the performance limit of the monolayer (ML) Bi2O2Se metal oxide semiconductor field-effect transistors (MOSFETs) by using ab initio quantum transport simulation at the sub-10 nm gate length. The on-current, delay time, and power-delay product of the optimized n- and p-type ML Bi2O2Se MOSFETs can reach or nearly reach the high performance requirements of the International Technology Roadmap for Semiconductors (ITRS) until the gate lengths are scaled down to 2 and 3 nm, respectively. The large on-currents of the n- and p-type ML Bi2O2Se MOSFETs are attributed to either the large effective carrier velocity (n-type) or the large density of states near the valence band maximum and special shape of the band structure (p-type). A new avenue is thus opened for the continuation of Moore's law down to 2–3 nm by utilizing ML Bi2O2Se as the channel.
Recently, atomically thin 2D Bi2O2Se has been successfully synthesized by chemical vapor deposition (CVD) on a mica substrate.16,17 The Bi2O2Se crystal is a semiconductor with a layer dependent band gap of 0.31–1.14 eV at the generalized gradient approximation (GGA) level. The electron mobility of the 2D Bi2O2Se is as high as 450 cm2 V−1 s−1 at room temperature and 29000 cm2 V−1 s−1 at 1.9 K.16 Moreover, 2D Bi2O2Se shows excellent stability against oxidation and moisture in air. These features make 2D Bi2O2Se very attractive for the next generation electronics. The fabricated top-gated FETs with 21 μm-long Bi2O2Se channels exhibit large on/off current ratios of up to 106 and near-ideal subthreshold swing values down to 65 mV dec−1 at room temperature.16 Three fundamental questions arise naturally: (1) What is the device performance limit of 2D Bi2O2Se based FETs? (2) How does this device performance compare with that of other 2D semiconductor FETs? (3) Can 2D Bi2O2Se FETs meet the HP device requirement of ITRS?
In this paper, we investigate for the first time the performance limit of the n- and p-type sub-10 nm double-gated (DG) monolayer (ML) Bi2O2Se metal–oxide semiconductor FETs (MOSFETs) using accurate ab initio quantum transport simulations. The on-currents of the sub-10 nm ML Bi2O2Se MOSFETs are much larger than those of their ML MoS2 counterparts across most of the size range tested. The optimized n- and p-type DG ML Bi2O2Se transistors can reach or nearly reach the ITRS goals for the on-state current, delay time and power-delay product (PDP) of the HP devices until the gate lengths are scaled down to 2 nm and 3 nm, respectively. On the basis of the available data, we reveal that either a very small or a very large effective electron (hole) mass favors a large on-state current for the 2D semiconductor FETs.
A gated two-probe model of the DG ML Bi2O2Se MOSFET is constructed with intrinsic ML Bi2O2Se as the channel and degenerately doped ML Bi2O2Se as the electrodes, as shown in Fig. 2(a). The underlap (UL) configuration, i.e. the spacer region between the gate and electrode, has also been considered to improve the device performance.19–22 We apply silicon dioxide as the dielectric material. The equivalent oxide thicknesses (tox) in the MOSFETs are set to 0.41–0.54 nm according to the ITRS 2.0 2013 edition requirement.
We first investigate the device performance without the UL configuration and with the electron/hole concentration of ρ = 5 × 1013 cm−2 in both the source and the drain. This level of doping concentration has commonly been applied in the previous 2D materials-based device simulations.23–27 Experimentally, the doping strategy for 2D layered materials includes substitutional doping,28,29 ion implantation,30,31 plasma treatment,32 surface charge transfer,33 and photoinduced doping.34 With the introduction of AuCl3 dopants, a p-type doped MX2 transistor with a hole concentration of 5 × 1013 cm−2 is achieved.35 By Nb atom substitution28 or UV photoinduced doping,34 a hole or electron concentration (about 1014 cm−2) that is even higher than our modeled value is achieved in MoS2. Theoretically, we model the effect of doping by changing the number of electrons of the system with respect to the neutral case. Correspondingly, the electrostatic potential acting on the Kohn–Sham one-electron states is modified.
The transfer characteristics of the Bi2O2Se n- and p-MOSFETs at a supply voltage Vdd = 0.64–0.72 V are shown in Fig. 2(b) and (c). At a given gate voltage Vg, the currents of the n-MOSFETs are generally higher than those of the p-MOSFETs due to the smaller effective mass (me = 0.20m0vs. mh = 1.40m0) and thus higher carrier mobility of electrons than holes. As the gate length Lg decreases, the current increases apparently in the subthreshold region, a result associated with the more severe source to drain direct tunneling.
From the transfer characteristics, we extract the subthreshold swing (SS) and transconductance (gm) values to evaluate the gate control ability in the ML Bi2O2Se MOSFETs in the subthreshold and superthreshold regions, respectively. A small SS value implies weak short-channel effects, and the smallest SS achievable in thermionic devices at room temperature is 60 mV dec−1. As shown in Fig. 2(d), the SS values of n- and p-MOSFETs show similar degradation behavior at shorter Lg, and the former degrade faster. At Lg = 9 nm, the nearly ideal SS value of 60 mV dec−1 is achieved in both the n- and p-MOSFETs. As Lg decreases to 1 nm, the SS values increase to 600 and 500 mV dec−1 in the n- and p-MOSFETs, respectively. The trend to larger SS values at shorter Lg is consistent with the stronger gate controllability at the band edge location (Fig. S1†). The gate control ability in the superthreshold region also decreases with the shorter Lg, as evidenced by the smaller gm values (Fig. 2(e)). Notably, the ML Bi2O2Se n-MOSFETs show generally larger gm values (7–12 mS μm−1) than the p-MOSFETs (2–5 mS μm−1), implying a better superthreshold gate control for the former over the latter.
In a digital device, a higher on–state current is beneficial for achieving a faster logic transition speed. The off–current is fixed to 0.1 μA μm−1 according to the ITRS 2.0 2013 edition standards for HP applications, and the on-current is evaluated under a supply voltage (Vdd = Vb) of 0.64–0.72 V (Table S1†). As shown in Fig. 4(a), the on-currents of the n- and p-MOSFETs without a UL fulfill the ITRS HP requirements of 900–1350 μA μm−1 until Lg = 7 nm and 5 nm, respectively. Remarkably, the on-current in the Bi2O2Se n-MOSFET at Lg = 9 nm reaches 3500 μA μm−1. This on-current is the highest among the simulated sub-10 nm transistors based on the stable 2D semiconductors at the ab initio quantum transport level, to the best of our knowledge.13,19,26
To further improve the device performance, utilization of the UL structure is a viable solution. The UL could improve the subthreshold electrostatics because it increases the effective channel length and thereby suppresses the source to drain leakage. We consider the symmetric UL structure in the ML Bi2O2Se MOSFETs with Lg ≤ 5 nm; the length of the UL is chosen under the precondition of keeping the whole channel length (Lch = Lg + 2 × UL) at less than 10 nm. As shown in Fig. S2,† the transfer characteristics of the 1 nm-gate-length n- and p-MOSFETs become steeper in the subthreshold region with the increasing UL. As the UL increases from 0 to 4 nm, the SS values decrease from 660 to 168 mV dec−1 and from 474 to 111 mV dec−1 for the 1 nm-gate-length n- and p-MOSFETs, respectively. This enhanced electrostatic effect by the UL also occurs in all the other studied cases. As shown in Fig. S3(a and b),† the longer UL always renders a smaller SS. The SS decreases by 28–75% with the aid of the UL, and the magnitude of the SS reduction is larger for devices with shorter Lg.
The improved gate control in the subthreshold region by using the UL is also reflected in the enhanced modulation of the band edge location by Vg. Take the ML Bi2O2Se p-MOSFET at Lg = 1 nm as an example. In Fig. 3, we define a maximum hole barrier height ϕmax as the energy barrier for the holes at the VBM of the source (E = −0.32 eV) to transport to the drain. To keep the same off-current of 0.1 μA μm−1 in the off-state, ϕmax is largest at UL = 0 nm followed by UL = 2 nm and UL = 4 nm (Fig. 3(a)–(c), respectively), and the peaks of the three spectral current curves are of the same order of magnitude (Fig. 3(d)). Under the gate modulation of 0.64 eV, the VBMs of the ML Bi2O2Se MOSFETs in the channel move upward, and the devices are turned to the on-state. ϕmax decreases slightly by 0.10 eV at UL = 0 nm, while it decreases significantly by 0.31 eV at UL = 2 nm and by 0.25 eV at UL = 4 nm (Fig. 3(e)–(g)). The enhanced subthreshold electrostatics induced by the UL are advantageous for boosting the on-current. As shown in Fig. 3(h), the peak of the spectral current increases apparently with the longer UL. Remarkably, the current chiefly comes from those transmissions with the energy below the chemical potential of the source μs (namely out of the bias window) in terms of the spectral current. The maximum enhancement of the on-current by the UL is 21 times in the ML Bi2O2Se MOSFET with Lg ≤ 5 nm (Fig. S3(c and d)†). After applying the UL, the on-currents of the ML Bi2O2Se p-MOSFET at Lg = 3 nm and n-MOSFET at Lg = 2 nm reach 1127 and 871 μA μm−1, respectively, fulfilling or almost fulfilling the ITRS requirement of 900 μA μm−1 (Fig. 4(a)).
Next, we check the intrinsic gate delay time (τ) and PDP, which are two key figures of merit reflecting the switching speed and power efficiency for the transistor operation, respectively. The fringing capacitance Cf is included by assuming it is twice the intrinsic gate capacitance Cg according to the ITRS requirements. The total capacitance Ctotal is the sum of Cg and Cf. The delay time is calculated by τ = CtotalCdd/Ion. Without the UL, the calculated delay times (0.10–42.00 ps) of the ML Bi2O2Se n- and p-MOSFETs meet the ITRS HP requirements of 0.423–0.446 ps until Lg = 7 nm and 5 nm, respectively (Fig. 4(b)). The introduction of the UL brings τ of the ML Bi2O2Se n- and p-MOSFETs below 0.26 ps at Lg ≤ 5 nm, thus meeting the ITRS HP goal even at Lg = 1 nm. PDP is calculated according to the equation PDP = VddIonτ = CtotalVdd.2 Even without the UL, PDPs of the sub-10 nm ML Bi2O2Se n- and p-MOSFETs range from 0.03 to 0.37 fJ μm−1 and decrease with shorter Lg values (Fig. 4(c)). The use of the optimal UL makes PDP decrease by 20%–110% in the ML Bi2O2Se MOSFETs at Lg < 5 nm. Therefore, the sub-10 nm ML Bi2O2Se n- and p-MOSFETs require low energy during on–off state switching, in terms of the smaller PDP compared with the ITRS requirements for HP application (0.24–0.45 fJ μm−1).
We have also explored the potential of the sub-10 nm ML Bi2O2Se n- and p-MOSFETs for LP application. The details are provided in Table S2 and Fig. S4.† For the n-MOSFETs, the leakage current is too large to be lower than the ITRS off-current requirement of 10−5 μA μm−1, due to the small effective mass and short channel barrier. For the p-MOSFETs, only those with 7- and 9 nm gate lengths could fulfill the ITRS LP requirements for the on-current and the application of a UL does not help. Apparently, ML Bi2O2Se is less satisfactory for LP applications in the sub-10 nm region.
Fig. 5 Comparison of the on-currents of the optimal ML Bi2O2Se, black phosphorus (BP),21,36 arsenene (As),19,26 antimonene (Sb),19,26 InSe,20 and MoS2 MOSFETs at different gate lengths obtained from the ab initio quantum transport simulations. The presented on-current has been optimized by using a UL. The detailed computational parameters are provided in Table S4.† |
Remarkably, the on-currents of the ML Bi2O2Se MOSFETs exceed those of the ML MoS2 ones, and the difference increases with increasing Lg. At Lg = 9 nm, the on-currents of the ML Bi2O2Se n- and p-MOSFETs are 15.6 and 3.6 times, respectively, those of their ML MoS2 counterparts. Besides the on-state current, the sub-10 nm ML Bi2O2Se MOSFETs also show generally superior performance to ML MoS2 in terms of the delay time and PDP (Table S3†). We are aware that the recently fabricated semiconductor tellurene (2D tellurium) also possesses both high carrier mobility and air stability,39 but its performance limit is unclear.
To interpret the difference in the on-current of the above 2D material-based transistors, it is helpful to check the dependence of the on-current on the effective mass of the channel materials along the transport direction (). A small effective mass of the channel material produces two effects: (1) it implies a large carrier effective velocity along the transport direction, which is beneficial for achieving a large on-current. (2) It implies a small density of states (DOS) near the VBM or CBM (, where the symbols ‖ and ⊥ indicate the transport and transverse directions, respectively). This is a disadvantage for achieving a large on-current, because a small DOS near the VBM or CBM indicates that a large variation of the gate voltage ΔVg is needed to drive a certain variation of the channel charge ΔQch, leading to a small transconductance gm. Another disadvantage, namely large tunneling leakage induced by a small , is not discussed here as this has been depressed by the aid of the UL.
In Fig. 6(a) and (b), we provide the on-current as a function of the effective mass in the 2D semiconductor n- and p-MOSFETs, respectively, with Lg fixed at 5 nm. In the range of small effective masses , the on-current shows an increasing trend with decreasing m* for both the n- and p-MOSFETs. In the range of large effective masses , the on-current shows an increasing trend with increasing for the p-MOSFETs. The relation between and on-current (Ion) in the n- and p-MOSFETs can be well fitted by the linear functions and , respectively (except for the armchair-directed ML phosphorene p-MOSFET).
Fig. 6 (a–c) On-currents of the n-MOSFET, p-MOSFET, and both n- and p-MOSFETs versus the effective masses of the ML channel material along the transport direction at Lg = 5 nm.19–21,26,36 All the data here are obtained from ab initio quantum transport simulations. The solid lines in (a,b) are linear fitted to the data (excluding BP in (b)). The solid line in (c) is for guidance to the eyes. (d–f) Total transmission spectra, k-point resolved transmission spectra, and transmission eigenstates of the ML Bi2O2Se p-MOSFET at the on-state. Lg = 9 nm and UL = 0 nm. The inset in (e) shows the 2D Brillouin zone of ML Bi2O2Se and the 1D Brillouin zone of the transistor. In (f), E = −0.5 eV. |
To establish an overall picture, we replot the data in Fig. 6(a) and (b) as a whole in Fig. 6(c). Among all the 2D semiconductors in Fig. 6(c), ML phosphorene is anisotropic while the other channel materials are isotropic . For the armchair-directed ML phosphorene p-MOSFETs, the noticeably large on-current (4500 μA μm−1) comes from the fact that it possesses both a small and large at the same time. Thereby, it can achieve both a large carrier effective velocity and a large DOS near the VBM. For the other cases, the two effects introduced by the effective masses are always in opposition, and the effective mass needs to be optimized so that a balance between the two effects is established. Compared with the case of MoS2, ML Bi2O2Se, InSe, arsenene, and antimonene n-MOSFETs give full play to the advantage of the large carrier effective velocity while ML Bi2O2Se, InSe and phosphorene p-MOSFETs maximize the advantage of the large DOS in the on-current. Therefore, not only those 2D semiconductors with very small but those with large are worthy of exploration for ultra-scaled electronics. The ML MoS2 MOSFET with shows the lowest Ion in Fig. 6(c). However, more studies are needed to precisely locate the Ion minimum point since no data is available in the range of 0.45m0–1.40m0. Another question that needs more exploration is to what extent the increasing trend of the on-current at continues.
It is noteworthy that the special shape of the valence band of ML Bi2O2Se contributes to the large on-current in the ML Bi2O2Se p-MOSFETs. In the valence band, there is a local maximum at point A with an energy of about 0.1 eV lower than that of the VBM at the point X (Fig. 1(a)). The hole effective mass of ML Bi2O2Se at the point A is 0.21m0, only 15% of the hole effective mass at the VBM (1.40m0). The key role of the point A in the transport properties of ML Bi2O2Se can be found by checking its contribution to the on-state transmission of the device.
Fig. 6(d) and (e) shows the on-state transmission spectrum and the k-point resolved transmission spectrum of the 9 nm ML Bi2O2Se p-MOSFET. The left inset in Fig. 6(e) shows the 2D Brillouin zone of ML Bi2O2Se, and the right inset shows the 1D Brillouin zone of the device, with K‖ indicating the transverse direction. Point A‖ is the projection of the point A to the 1D Brillouin zone. The transmission peak closest to the Fermi level is located at E = −0.5 eV, and the corresponding largest transmission contribution is from the point A‖, as shown in the k-point resolved transmission spectrum. The significant contribution of A‖ to the transmission can also be found in the transmission eigenstates (Fig. 6(f)). At E = −0.5 eV, the incoming wave function of both the points A‖ and X‖ spread over the whole transistor. The wave function of the point A‖ covers both the Bi and Se atoms, while the wave function of the point X‖ is mainly localized near the Se atoms. This is related to the fact that both the orbitals of the Bi and Se atoms contribute to the point A while only the orbitals of the Se atoms contribute to the point X in the band structure (Fig. S5†).
The excellent performance of the ML Bi2O2Se MOSFETs is predicted in the Ohmic contact limit, which is ensured by the degenerately doped source and drain. Experimentally, it is rather challenging to heavily dope the 2D semiconductors,3 and the metal is often applied as the source and drain electrodes in the devices. This metal-contacted FET is termed a Schottky barrier FET (SBFET). A Schottky barrier might exist at the metal-ML Bi2O2Se interfaces and hinder the carrier injections. To achieve the predicted excellent performance in practice, it is essential to find a proper metal electrode that forms a barrier-free Ohmic contact with ML Bi2O2Se. The ab initio simulations reveal that Pt, Sc and Ti electrodes would form a desirable n-type Ohmic contact with ML Bi2O2Se.40 With these metals as electrodes, the ML Bi2O2Se SBFETs are highly likely to approach the predicted performance limit of the corresponding MOSFETs.
(1) |
The transmission coefficient T(E) is the average of k-dependent transmission coefficients Tk‖(E) over the Brillouin zone. The k-dependent transmission coefficient at energy E is
(2) |
The electrostatics are treated by solving the Poisson equation self-consistently via a real space solver. The periodic and Neumann conditions are used on the boundaries along the transverse direction and the direction normal to the Bi2O2Se surface, respectively. To make sure of the charge neutrality in the electrode region, the Dirichlet boundary condition is applied on the electrode-channel boundaries along the transport direction. The k-point grids45 are sampled with a separation of 0.01 Å−1 in the Brillouin zone. The double-ζ plus polarization (DZP) basis set is adopted. The temperature is set to 300 K.
We utilize the generalized gradient approximation (GGA) in the form of the Perdew–Burke–Ernzerhof (PBE) exchange–correlation functional.46 Because the electron–electron interaction is significantly screened by doping carriers, the single electron approximation (DFT-PBE) is a good approximation to describe the electronic structure of the device.26,41,47–50 The reliability of the quantum transport simulation at the DFT-PBE level in predicting the sub-10 nm FET performance is validated by the comparison of the performance of the simulated 2D MoS2 transistors and the experimental ones. A general agreement in the overall transfer characteristics in the 1 nm-gate-length 2D MoS2 transistors with extremely thin effective oxide thicknesses is found.1,21 Especially in the subthreshold region, a SS of 66 mV dec−1 in the simulated device agrees well with the observed 65 mV dec−1.1,21 We further calculate the 5–9 nm-channel-length ML MoS2 n-type MOSFETs, and the drive currents in the ballistic transport limit are predicted to be 270–290 μA μm−1 at a bias of Vb = 0.64 V (Fig. S6†). These values are close to the drive current of approximately 180 μA μm−1 at Vb = 1 V in the fabricated 10 nm-channel-length ML MoS2 FET with the nearly Ohmic graphene-MoS2 contact and are still working in the diffusive regime.9
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8nr08852g |
This journal is © The Royal Society of Chemistry 2019 |