IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 57, NO. 9, SEPTEMBER 2010 2317 Hetero-Gate-Dielectric Tunneling Field-Effect Transistors Woo Young Choi, Member, IEEE, and Woojun Lee Abstract—A tunneling field-effect transistor (TFET) is considered one of the most promising alternatives to a metal–oxide– semiconductor field-effect transistor due to its immunity to short-channel effects. However, TFETs have suffered from low ON -current, severe ambipolar behavior, and gradual transition between ON- and OFF-states. To address those issues, the authors have proposed hetero-gate-dielectric TFETs. The proposed device enhances ON-current, suppresses ambipolar behavior, and makes abrupt ON–OFF transition by replacing the source-side gate insulator with a high-k material, which induces a local minimum of the conduction band edge at the tunneling junction. Index Terms—Ambipolar behavior, gate insulator, heterogate-dielectric, subthreshold swing (SS), tunneling field-effect transistor (TFET). I. I NTRODUCTION R ECENTLY, various kinds of novel electron devices have been studied to overcome the scaling limit of MOSFETs [1]–[3]. Among them, a tunneling field-effect transistor (TFET) is considered one of the most promising alternatives since it is immune to subthreshold swing (SS) degradation at shortchannel length. On the other hand, it has been reported that TFETs have low ON-current Ion , which limits operation speed, and severe ambipolar behavior, which increases leakage current Iamb [4], [5]. Additionally, it is problematic that the SS values of experimentally demonstrated silicon TFETs are larger than expected values. In theory, the SS of silicon TFETs can be reduced smaller than 60 mV/dec at room temperature, which is the minimal SS of conventional MOSFETs. However, only a few research groups have succeeded in demonstrating sub60-mV/dec SS at room temperature experimentally [6], [7]. Thus, to improve Ion and SS, a high-k material has been introduced as a gate insulator [8]. However, using a high-k material as a gate insulator may increase leakage current Iamb due to severe ambipolar behavior. To alleviate ambipolar behavior, TFETs without gate–drain overlap have been proposed [9]. However, this approach actually increases drain-to-source channel length, which significantly reduces chip density. In this brief, we have proposed hetero-gate-dielectric TFETs (HG TFETs) for higher Ion , lower Iamb , and smaller SS without Manuscript received February 9, 2010; revised May 18, 2010; accepted May 20, 2010. Date of publication July 15, 2010; date of current version August 20, 2010. This work was supported by the National Research Foundation of Korea, funded by the Ministry of Education, Science and Technology, under Grants 2009-0082439 and 2009-0084522. The review of this brief was arranged by Editor C. McAndrew. The authors are with the Department of Electronic Engineering, Sogang University, Seoul 121-742, Korea (e-mail: wchoi@sogang.ac.kr). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TED.2010.2052167 Fig. 1. Schematic of the proposed HG TFET. The HG TFET features different gate insulators at the source (high-k material) and drain (silicon oxide) sides. Since gate-to-channel coupling strength is different between channel regions overlapped by the high-k material and silicon oxide, HG TFETs have a local minimum of Ec at the tunneling junction, which improves Ion and SS. In addition, because low-k silicon oxide is located at the drain side where ambipolar behavior occurs, Iamb can be suppressed. sacrificing chip density, as shown in Fig. 1. The HG TFET features different gate dielectric materials at the drain and source sides, which can be formed by isotropic etching of silicon oxide followed by high-k material deposition. The proposed HG TFETs will be compared with two kinds of conventional devices, namely, TFETs that use only silicon oxide as a gate insulator (SiO2 -only TFETs) and TFETs that use only a high-k material as a gate insulator (high-k-only TFETs). The SiO2 only TFET corresponds to the HG TFET whose length of silicon oxide under the gate, i.e., LSiO2 , is equal to the gate length LG , whereas the high-k-only TFET corresponds to the HG TFET whose length of high-k material under the gate, i.e., Lhigh-k , is equal to LG . The HG TFET is expected to show higher Ion and smaller SS than the SiO2 -only and high-konly TFETs by modifying the band energy structure. A high-k material partially located at the source side induces a local minimum of the conduction band edge Ec at the tunneling junction, which will be discussed later from the following sections. In addition, since silicon oxide whose relative permittivity is low is located at the drain side, ambipolar behavior can effectively be suppressed, which lowers Iamb . II. R ESULTS AND D ISCUSSION To evaluate the merits of the proposed HG TFET, it has been compared with the SiO2 -only and high-k-only TFETs by twocarrier and 2-D device simulation using Silvaco ATLAS [10]. A nonlocal band-to-band tunneling model has been used [10]. In simulation, LG is 50 nm, which is equal to the sum of LSiO2 and Lhigh-k . In addition, the physical thickness of the gate insulator, i.e., tins , and that of the silicon-on-insulator (SOI) layer, i.e., tSOI , are 2 and 30 nm, respectively. The relative permittivity of the high-k material is assumed to be 25, which refers to that 0018-9383/$26.00 © 2010 IEEE 2318 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 57, NO. 9, SEPTEMBER 2010 Fig. 2. (a) Transfer curves of the HG, SiO2 -only, and high-k-only TFETs in the case of n-type doped polysilicon gates. HG TFETs follow SiO2 -only TFETs at low VG and high-k-only TFETs at high VG . (b) Transfer curves of the HG, SiO2 -only, and high-k-only TFETs when the gate workfunction is adjusted so that Vonset is 0 V for fair comparison. HG TFETs have the highest Ion due to their smallest SS. of HfO2 . The channel doping concentration is 1016 cm−3 . An abrupt junction profile is assumed. In addition, we assume that the interface between the high-k material and silicon oxide is abrupt. It is reasonable considering that the diffusion length of Hf into SiO2 is only ∼0.035 nm when the device is annealed at 1000 ◦ C for 5 s [11]. Ion is defined as the drain current ID when both the gate voltage VG and the drain voltage VD are 1 V, and Iamb is defined as ID when VG is −0.2 V and VD is 1 V. SS is defined as an average slope when ID increases from 1 fA/μm to 1 nA/μm. The onset voltage Vonset is defined as VG when ID is 1 fA/μm at 1-V VD . Fig. 2(a) compares the transfer characteristics of the HG TFET whose Lhigh-k is 6 nm with those of the SiO2 -only and high-k-only TFETs in the case of n-type doped polysilicon gates. Due to the heterogeneous gate dielectric, HG TFETs follow SiO2 -only TFETs at low VG and high-k-only TFETs at high VG . It is because the ON-state is determined at the sourceto-channel region overlapped by the high-k material, whereas the OFF-state ambipolar behavior is determined at the drain-tochannel region overlapped by silicon oxide. It should also be noted that the minimal leakage current level is determined by reverse-biased p-i-n diode leakage. For fair comparison, from now on, the gate workfunction will be adjusted so that Vonset becomes 0 V. The value of gate workfunction is ∼4.37 eV, which is attainable by using metal gates [12]. Fig. 2(b) shows a redrawn version of Fig. 2(a) when the gate workfunction is adjusted. It shows that HG TFETs have higher Ion than high-k TFETs, whereas they have as low Iamb as SiO2 -only TFETs, which leads to high a ON–OFF current ratio. It is because HG TFETs have smaller SS than SiO2 -only and high-k-only TFETs, which can be explained as follows: Fig. 3(a) shows that HG TFETs have a local minimum of Ec due to different gate-tochannel coupling strength between channel regions overlapped by the high-k material and silicon oxide. Fig. 3(b) compares HG TFETs with high-k-only TFETs in terms of source-to-channel tunneling barrier width as a function of VG . HG TFETs show a more abrupt change and a lower value of tunneling barrier width than high-k-only TFETs. It is because the tunneling barrier of HG TFETs abruptly narrows when a local minimum of Ec is aligned with the valence band edge Ev of the source, as shown in the inset of Fig. 3(b). In the case of SiO2 -only Fig. 3. (a) Band diagram of the HG TFET whose Lhigh-k is 6 nm when VG is 0 V and VD is 1 V. (b) Source-to-channel tunneling barrier width of the HG TFET whose Lhigh-k is 6 nm compared with that of the high-k-only TFET. The band diagram located in the box in (a) is magnified into two inset figures in (b) depending on VG . Fig. 4. (a) Ion and SS and (b) Iamb of HG TFETs depending on Lhigh-k compared with those of SiO2 -only and high-k-only TFETs. SiO2 -only and high-k-only TFETs correspond to HG TFETs whose Lhigh-k ’s are 0 and 50 nm, respectively. and high-k-only TFETs, source-to-channel tunneling barrier width gradually decreases since Ec gradually decreases from the source-to-channel region. On the other hand, in the case of HG TFETs, when the gate is biased around Vonset , tunneling barrier width is almost the same as that of high-k-only TFETs because a local minimum of Ec is not aligned with Ev of the source. It means that a local minimum of Ec has not been involved in the band-to-band tunneling process yet when VG is around Vonset . However, as VG becomes higher, a local minimum of Ec is shifted downward and eventually aligned with Ev of the source. Therefore, when VG exceeds Vonset , the length of the tunneling path abruptly decreases because bandto-band tunneling between a local minimum of Ec and Ev of the source begins to dominate the whole tunneling process. To optimize the design of HG TFETs, Lhigh-k has been optimized. Fig. 4(a) shows Ion and SS as a function of Lhigh-k . Note that HG TFETs whose LSiO2 or Lhigh-k is 50 nm correspond to SiO2 -only or high-k-only TFETs, respectively. It is observed that Lhigh-k can be optimized in terms of SS and Ion . The optimization process is related to the depth and width of the conduction band well where a local minimum of Ec is located. As Lhigh-k decreases, the conduction band well becomes shallower, which makes band-to-band tunneling difficult to occur. On the contrary, as Lhigh-k increases, the conduction band well becomes wider, which leads to a less abrupt transition between OFF- and ON -states. When Lhigh-k is optimized around 6 nm, CHOI AND LEE: HETERO-GATE-DIELECTRIC TFETs the optimized HG TFETs show ∼60% smaller SS and ∼30% higher Ion than high-k-only TFETs. In addition, the HG TFETs show ∼80% smaller SS and two orders of magnitude higher Ion than SiO2 -only TFETs. Further improvement is expected if the relative permittivity of the high-k material increases. Fig. 4(b) shows Iamb as a function of Lhigh-k . Iamb abruptly decreases as Lhigh-k decreases from 50 to 40 nm. It is explained by the fact that Iamb is originated from ambipolar behavior at the drain side. When the high-k material is replaced by silicon oxide at the drain side, the Iamb of the HG TFET becomes almost the same as that of the SiO2 -only TFET. Compared with highk-only TFETs, HG TFETs show seven orders of magnitude lower Iamb . 2319 [8] A. Vandooren, R. Rooyackers, D. Leonelli, F. Iacopi, E. Kunnen, D. Nguyen, M. Demand, P. Ong, L. Willie, J. Moonens, O. Richard, A. S. Verhulst, W. G. Vandenberghe, G. Groeseneken, S. D. Gendt, and M. Heyns, “A 35 nm diameter vertical silicon nanowire short-gate tunnel FET with high-k/metal gate,” in Proc. IEEE Silicon Nanoelectron. Workshop, 2009, pp. 21–22. [9] S. V. Anne, G. V. William, M. Karen, and G. Guido, “Tunnel field-effect transistor without gate-drain overlap,” Appl. Phys. Lett., vol. 91, no. 5, p. 053 102, Jul. 2007. [10] ATLAS Users Manual, SILVACO Int., Santa Clara, CA, 2009. [11] N. Ikarashi, K. Watanabe, K. Masuzaki, and T. Nakagawa, “Thermal stability of a HfO2 /SiO2 interface,” Appl. Phys. Lett., vol. 88, no. 10, p. 101 912, Mar. 2006. [12] H. Y. Yu, C. Ren, Y.-C. Yeo, J. F. Kang, X. P. Wang, H. H. H. Ma, M.-F. Li, D. S. H. Chan, and D.-L. Kwong, “Fermi pinning-induced thermal instability of metal-gate work functions,” IEEE Electron Device Lett., vol. 25, no. 5, pp. 337–339, May 2004. III. C ONCLUSION HG TFETs have been proposed for high performance and low-power consumption. In addition, device design has been optimized by modulating Lhigh-k . By using a local minimum of Ec at the tunneling junction and placing silicon oxide whose relative permittivity is low at the drain side, the optimized HG TFETs showed improved device performance than conventional TFETs such as SiO2 -only and high-k-only TFETs in terms of Ion , SS, and Iamb . R EFERENCES [1] M. R. William and A. J. A. Gehan, “Silicon surface tunnel transistor,” Appl. Phys. Lett., vol. 67, no. 4, pp. 494–496, Jul. 1995. [2] Z. Qin, Z. Wei, and A. Seabaugh, “Low-subthreshold-swing tunnel transistors,” IEEE Electron Device Lett., vol. 27, no. 4, pp. 297–300, Apr. 2006. [3] K. Boucart and A. M. Ionescu, “Double-gate tunnel FET with high-κ gate dielectric,” IEEE Trans. Electron Devices, vol. 54, no. 7, pp. 1725–1733, Jul. 2007. [4] O. M. Nayfeh, C. N. Chleirigh, J. Hennessy, L. Gomez, J. L. Hoyt, and D. A. Antoniadis, “Design of tunneling field-effect transistors using strained-silicon/strained-germanium type-II staggered heterojunctions,” IEEE Electron Device Lett., vol. 29, no. 9, pp. 1074–1077, Sep. 2008. [5] J. Appenzeller, Y.-M. Lin, J. Knoch, C. Zhihong, and P. Avouris, “Comparing carbon nanotube transistors—The ideal choice: A novel tunneling device design,” IEEE Trans. Electron Devices, vol. 52, no. 12, pp. 2568– 2576, Dec. 2005. [6] W. Y. Choi, B.-G. Park, J. D. Lee, and T.-J. K. Liu, “Tunneling field-effect transistors (TFETs) with subthreshold swing (SS) less than 60 mV/dec,” IEEE Electron Device Lett., vol. 28, no. 8, pp. 743–745, Aug. 2007. [7] J. Appenzeller, Y.-M. Lin, J. Knoch, and P. Avouris, “Band-to-band tunneling in carbon nanotube field-effect transistors,” Phys. Rev. Lett., vol. 93, no. 19, p. 196 805, Nov. 2004. Woo Young Choi (S’05–M’10) was born in Incheon, Korea, in 1978. He received the B.S., M.S., and Ph.D. degrees from Seoul National University, Seoul, Korea, in 2000, 2002, and 2006, respectively. From 2006 to 2008, he was with the Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, as a Postdoctor. Since 2008, he has been a member of the faculty of Sogang University, Seoul, Korea, where he is currently an Assistant Professor with the Department of Electronic Engineering. He has authored or coauthored more than 80 papers in international journals and conference proceedings. He is the holder of 10 Korean patents. His current research interests include fabrication, modeling, characterization, and measurement of CMOS/CMOS-compatible semiconductor devices and nanoelectromechanical memory cells. Prof. Choi was the recipient of the Humantech Thesis Prize from Samsung Electronics in 2005 and the Doyeon Paper Award from the Inter-University Semiconductor Research Center, Seoul National University. Woojun Lee was born in Seoul, Korea, in 1984. He received the B.S. degree in electronic engineering in 2009 from Sogang University, Seoul, Korea, where he is currently working toward the M.S. degree in electrical engineering with the Department of Electrical Engineering. His current research interests include CMOS and CMOS-compatible novel device modeling.