Steep switching devices for low power applications: negative differential capacitance/resistance field effect transistors
Simply including either single ferroelectric oxide layer or threshold selector, we can make conventional field effect transistor to have super steep switching characteristic, i.e., sub-60-mV/decade of subthreshold slope. One of the representative is negative capacitance FET (NCFET), in which a ferroelectric layer is added within its gate stack. The other is phase FET (i.e., negative resistance FET), in which a threshold selector is added to an electrode (e.g., source or drain) of conventional field effect transistor. Although the concept of the aforementioned two devices was presented more or less recently, numerous studies have been published. In this review paper, by reviewing the published studies over the last decade, we shall de-brief and discuss the history and the future perspectives of NCFET/phase FET, respectively. The background, experimental investigation, and future direction for developing the aforementioned two representative steep switching devices (i.e., NCFET and phase FET/negative resistance FET) are to be discussed in detail.
KeywordsSteep switching device Negative capacitance Phase FET Low power application Field effect transistor
In order to implement better speed/power consumption and integration density (of course, for given cost) in integrated circuit (IC) chip, metal oxide semiconductor field effect transistor (MOSFET) has been scaled down with the help of various process/device solutions over the last a few decades; in fact, the Moore’s Law  has been alive for long time. Three-dimensional (3D) device structure, i.e., fin-shaped FET (FinFET), has been adopted for implementing 10 nm-scale technology platform in industry . Although the scaling of semiconductor device has been successfully achieved, the electric field intensity in device (e.g., in the channel region of device) has been hardly controlled (i.e., the electric field has not been constant but increasing every new technology generation). Thereby, the miniaturization of device such as MOSFET and FinFET has evenetually soared up the power density per unit area of IC chip. As of now, many groups have put an emphasis on addressing the technical issue (i.e., the power crisis in IC chip). The root-cause of the ever-increasing power density comes from the fact that threshold voltage (Vth) has not been appropriately scaled down as much as the physical size of device has been scaled [In the end, power supply voltage (VDD) in IC chip has not been scaled down as proportional to the integration density of IC chip]. This is primarily because thermionic emission process (i.e., the physical mechanism of how current in MOSFET forms and flows) causes the fundamental limit of subthreshold slope (SS), which is defined as the gate-to-source voltage (VGS) to increase the drain-to-source current (IDS) by a factor of 10 in the unit of mV/decade. Note that the value of SS at room temperature is, at best, 60 mV/decade.
In order to overcome the aforementioned bottleneck and to find the alternatives for sub-5-nm technology nodes, various material and/or device solutions have been sought. Among many new exploratory proposals, steep switching device (i.e., transistor/device which has the feature of sub-60-mV/decade SS) is one of the promising suggestions. For example, using the band-to-band tunneling mechanism (instead of the thermionic emission process), tunnel FET (TFET) [3, 4] can achieve sub-60-mV/decade SS. Unfortunately, TFET has more or less “low” on-state drive current as well as the layout penalty due to its “p-i-n” device structure, so that it should be used for a certain specialized application [not as the alternative for conventional complementary metal oxide semiconductor (CMOS) logic technology platform]. Nanoelectromechanical relay (NEM relay)  has taken advantages of using the mechanical connection between the channel and the source/drain, to achieve sub-60-mV/decade SS. Although NEM relay has “almost zero” off-state leakage current, it has more or less “poor” endurance characteristics as well as its intrinsic switching time issue (because of its mechanical movement).
In this review paper, we are going to discuss the conspicuous features of the CMOS-compatible steep switching devices for low-power applications; the first one is the negative capacitance field effect transistor (NCFET), and the second one is the phase transition field effect transistor (phase FET). NCFET takes advantages of using the “negative (differential) capacitance” effect of single ferroelectric layer, to achieve the sub-60 mV/decade SS. On the other hand, phase FET uses the “negative (differential) resistance” effect of threshold selectors for implementing the steep switching characteristic. The operational principle, recent results, and challenges of those two steep switching devices are to be shown and discussed.
2 Negative (differential) capacitance field effect transistor (NCFET)
2.1 The debut of negative capacitance field effect transistor (NCFET)
2.2 Experimental demonstration of negative capacitance effect
After the concept of NCFET was suggested, a few pioneering studies have been done to experimentally identify the presence of negative capacitance. Khan et al. have directly measured the negative capacitance, by using a simple RC circuit . The transient response of the RC circuit [in which a 60 nm-thick ferroelectric (PZT) capacitor is connected in series with a resistor] has clearly showed the negative slope in the measured voltage vs. time plot. This should verify the existence and observation of the negative differential capacitance. Appleby et al. have experimentally demonstrated the capacitance of heterostructure capacitor [which is composed of SrTiO3 and BaTiO3 (BTO) materials], showing that the capacitance increases as a function of BTO thickness; this indicates that the negative capacitance of BTO would contribute to increase the total capacitance [12, 13] (note that the total equivalent capacitance of two series-connected capacitors is lower than the capacitance of any stand-alone capacitor itself). Gao et al. have fabricated “ferroelectric Ba0.8Sr0.2TiO3 + dielectric LaAlO3 (LAO)” superlattice heterostructure, which resulted in a higher capacitance than the capacitance of LAO capacitor itself . Similarly, Ku et al.  and Sharma et al.  have also veryfied the negative capacitance effect of P(VDF0.75–TrFE0.25) and Hf0.5Zr0.5O2 (HZO) ferroelectric material, respectively.
2.3 Negative capacitance field effect transistor: Saviour of sub-10-nm MOSFET
Since the theoretical proposal of negative capacitance field effect transistor (NCFET) and the direct observation of negative capacitance were published, a clear manifestation of steep switching feature of NCFET has been necessary. In 2015, Jo et al. has done the proof-of-concept study, and they demonstrated the sub-60-mV/decade characteristic of NCFET by conceptually and simply connecting MOSFET in series with ferroelectric capacitor [17, 18]. In [17, 18], it is discussed that the negative capacitance effect enables to amplify the surface potential in baseline device, and thereby, the slope of ID–VG curve in sub-threshold region becomes steeper (i.e., subthreshold slope ~ 18 mV/decade at 300 K). Furthermore, Khan et al. demonstrated the short-channel negative capacitance FinFET, with BiFeO3 (BFO) ferroelectric capacitor externally connected to baseline device with channel length of 100 nm . Afterwards, many articles which deal with negative capacitance MOSFET [20, 21, 22, 23, 24] and FinFET [25, 26, 27] demonstrated the reproducible NC effect. Based on state-of-the-art devices such as (i) 2-Dimensional field effect transistors [28, 29, 30, 31], (ii) nanoelectromechanical switch [32, 33], (iii) carbon nanotube , (iv) silicon-on-insulator (SOI) device , and (v) polymer ferroelectric FET , recent papers has experimentally and theoretically shown the benefits of using negative capacitance. The number of published studies on NCFET has sharply risen since early 2010s, indicating the increasing interest of NCFET in electron device community (especially, silicon CMOS device community). In next sub-chapters, a few major issues in NCFET design are to be addressed.
2.3.1 How to remove “hysteresis” in NCFET?
2.3.2 CMOS-compatibility and scaling
In the early days of NCFET design, people has used typical ferroelectric materials, such as PZT, BFO, and PVDF [11, 12, 13, 14, 15, 16, 17]. There are a few reasons to use them: (i) it was easy to fabricate ferroelectric capacitors with these materials. (ii) The polarization switching of typical ferroelectric materials has supremacy, to induce the negative capacitance effect. Yet, those materials had to be fabricated to have the thickness of a few tens of nanometers or even more, so as to have excellent ferroelectric properties. Furthermore, those materials were not compatible to current CMOS process [e.g., a few tens of nanometers ferroelectric layer is too thick to be deposited/included in the current replacement gate (RMG) stack]. To address those technical issues, the hafnium-based ferroelectric material are actively under development. Since the work of fabricating a hafnium-based ferroelectric layer has been published in 2011 , many studies have showed that ferroelectric properties can be implemented without doping  or with doping various materials (materials such as Y, Al, Zr, Si, etc.) into the hafnium oxide [40, 41]. Lee et al. have announced that it was possible to verify the steep switching characteristic of NCFET with 1 nm-thick HfZrOx ferroelectric layer . Follow-up studies have showed that, by integrating the hafnium-oxide ferroelectric layer (note that the current CMOS process uses the hafnium-oxide as the gate insulation material in mass production), it is possible to improve the CMOS-compatibility of NCFET [43, 44, 45]. Moreover, the hafnium-oxide ferroelectric layer can be thin enough for aggressively-scaled transistors (specifically, sub-3-nm-thick hafnium-oxide has been demonstrated!). The last point in this sub-chapter is the effect of domains in ferroelectric layer [46, 47, 48, 49, 50, 51]. In developing NCFETs, the domain structure of newly studied hafnium-based ferroelectric should be discussed and scrutinized for accurate device operation.
2.3.3 Discussion on internal metal layer
Previous studies which have worked on demonstrating the steep switching feature of NCFET have used an external connection of ferroelectric capacitor to baseline device. To be adopted in aggressively-scaled transistors, the ferroelectric layer should be eventually integrated in the gate stack of baseline device. Numerous studies have reported the performance of NCFET with “integrated” ferroelectric layer. The ferroelectric layer can be fabricated/deposited (i) directly on the dielectric layer or (ii) on top of the internal metal layer (which is deposited on the dielectric layer or interlayer). These two different gate stack structures would have an impact on device reliability. Khan et al. discussed the reliability issues with the aforementioned two different gate stack structures . Depending on the presence of internal metal layer, the conditions required for satisfying the stable operation of device as well as the degree to which the leakage current affects are determined. In this regard, more studies on various issues such as device design, defects, device structural variations, etc., should be done in the near future.
2.4 Future outlook
As mentioned before, the exploratory study on NCFET has begun in 2008. Afterwards, many works have been followed, such as (i) the verification work of negative capacitance phenomenon and the steep switching effect [11, 12, 13, 14, 15, 16, 17, 18], (ii) the engineering work to control the hysteresis window [20, 26, 37], (iii) the work for compatibility to current CMOS processes [38, 39, 40, 41, 42, 43, 44, 45], (iv) experimental and theoretical studies of capacitance-matching conditions required for stable and reliable operation [53, 54], and (v) modelling for circuit designs [55, 56, 57, 58, 59, 60]. Many studies have supported how to adopt NCFET as future ultra-low power logic transistor. However, there are many issues to be tackled, for commercialization. For instance, (1) the recipe of hafnium-based 1 nm-thick ferroelectric layer (especially, with reasonable controllability of defects and thickness), (2) the thorough discussion on the process-induced systematic and random variation of NCFET , (3) the discussion on optimal device structure when using NC, (4) the step-by-step capacitance matching manual in wafer-level process, (5) experimentally-verified compact model for NCFET, and (6) AC response and NCFET-based circuit designs [62, 63, 64] should be studied in the near future.
3 Negative differential resistance FET (Phase FET)
Overview of experimental results of negative capacitance field effect transistors
Internal metal layer
Minimum subthreshold slope (mV/decade)
MOSFET (Lg = 1 μm)
First experimental demonstration of steep switching feature
FinFET (Lg = 100 nm)
Experimental demonstration of negative capacitance FinFET
MOSFET (Lg = 1 μm)
Hysteresis-free negative capacitance FET by controlling the drain voltage
FinFET (Lg = 30 nm)
Negative capacitance FinFET with integrated 5 nm thick hafnium-based ferroelectric layer
MOSFET (Lg = 10 μm)
Integration of 100 nm-thick PZT ferroelectric layer in MOSFET
FinFET (Lg = 70 nm)
Hyeresis reduction by adjusting the FinFET’s layout parameter
MOSFET (Lg = 30 μm)
Integration of 1.5 nm-thick HZO ferroelectric layer in MOSFET
FinFET (Lg = 14 nm)
Doped Hf ferroelectric material with 3 ~ 8 nm thickness, which is integrated in FinFET
MoS2 2D FET (Lg = 1 μm)
2D negative capacitance FET with integrated hafnia ferroelectric layer
3.1 The debut of negative differential resistance FET (or Phase FET)
There exists materials of which resistivity is varied by external stimulus. For example, Morin.  has demonstrated that the resistivity of VO2, one of the resistive switching materials, can be significantly modified at a critical temperature. Afterwards, many people have confirmed that the resistivity of VO2 can be engineered by (i) carrier concentration, (ii) electrical stimulus (bias), (iii) illumination, (iv) etc. [66, 67, 68, 69, 70, 71]. The characteristics of many resistive switching materials have been still under study [72, 73, 74, 75, 76, 77, 78, 79]. Thanks to using the resistive switching materials, novel transistors have been invented. In 1988, D. M. Newns et al. proposed a new type of field effect transistor using a Mott insulator (which is used in the channel region) . The Mott insulator channel is directly turned “ON”, by gate bias (i.e., the carrier concentration in channel is significantly increased by gate bias). However, more or less “high” gate voltage is necessary to sufficiently accumulate the carrier concentration for turning “ON” the channel. Instead of using the Mott insulation material, transistors with ionic liquid gate dielectric material have been investigated [81, 82]. It is known that ionic liquid (vs. solid-state-gate-dielectric material) helps to accumulate more carriers to turn “ON” the Mott insulator channel. However, the ionic liquid is susceptible to electrochemical reactions, and it has slow response characteristic. In 2012, Zhou et al. invenstigated the relaxation dynamics of ionic liquid with VO2 .
3.2 Phase FETs using various materials
Phase FET with VO2 material has been introduced because the material (VO2) has stable resistive switching as well as low “ON” resistance. VO-2 is transformed to low resistance state at 340 K. This means that the resistive switching cannot be appropriately used for high temperature applications (e.g., CPU is usually running beyond 340 K). Moreover, the resistance of VO2 is lower than that of filament-based TS device (which is explained below, to suppress the off-state leakage current of state-of-the-art transistor). For those aformentioned reasons, many new materials have been proposed to address the technical issues.
3.3 Future outlook
The phase FET has showed its promising characteristics for low power applications, because of its superior switching feature. However, there is still room for more improvement. The characteristic of filament-based TS device can be determined, most likely by the material, likewise CBRAM. Hence, various metal oxides need to be investigated, to look for best option. For example, ferroelectric tunnel junction (FTJ)  can be worked as the TS device in phase FET. Recently, Pb(Zr0.52Ti0.48)O3-based phase FET was introduced by Shin et al. .
As described above, for using negative differential resistance effect, the TS device in phase FET can be connected in series to three terminals (i.e., source, drain, and gate terminal) of baseline transistor. The source-connected phase FET (noted as S-phase FET, herein) decreases the effective gate-to-source voltage and drain-to-source voltage. The drain-connected phase FET (noted as D-phase FET, herein) only decreases the drain-to-source voltage. Lastly, the gate-connected phase FET (noted as G-phase FET, herein) only decreases the gate-to-source voltage. The S-phase FET most decreases not only the off-state leakage current but also the on-state drive current. The D-phase FET least decreases the off-state leakage current and the on-state drive current. The G-phase FET shows the intermediate characteristics of both D-phase and S-phase FETs. However, it needs an external resistor. In 2017, Vitale et al. investigated VO2-based S- and G-phase TFET . However, VO2 cannot show all pros and cons of the proposed device structures, because of relatively high off-state leakage current. Therefore, the best device structure for phase FET should be further scrutinized in the near future.
Finally, the delay time and the on-state drive current of phase FET must be deeply investigated. Various filament-based TS devices have showed more or less slow switching time property (i.e., 0.1–1 μs) as well as low on-state drive current (i.e., 10–100 μA). In 2016, Song et al. showed TiO2-based phase FET with AgTe electrode (instead of Ag electrode) . Telluride (Te) helps to dissolve the Ag filament, so that the TS device can be utilized at higher compliance current (~ 100 μA) . Moreover, the delay time is dramatically improved by a factor of ~ 10. Therefore, the method using chalcogenide materials needs to be developed and finely tuned.
To sum up, we have briefly discussed the concept, recent results, and future outlook for two steep switching devices, i.e., negative differential capacitance FET (a.k.a., NCFET) and negative differential resistance (a.k.a., phase FET). Since 2008, many studies have showed that, by using the effect of negative capacitance on conventional transistors, the switching metric of MOSFET, i.e., subthreshold slope, can become lower than its fundamental limit of 60 mV/decade at 300 K. Furthermore, recent studies have revealed that CMOS-friendly hafnium-based dielectric capacitor can have ferroelectricity, and thereby HfO2-based NCFET is currently receiving lots of attentions across countries. The device design for “hysteresis-free”, and the scalability of the ferroelectric layer in the gate stake of MOSFET are the most critical issue to be tackled in the near future, for having current CMOS technology to adopt the NC device. In the middle of 2010s, after the first introduction of resistive switching material-based phase FET, the steep switching characteristic of phase FET was experimentally demonstrated. Because of the simple fabrication process and superiorly low off-state leakage current, various experimental and theoretical studies are ongoing as of early 2018. Although there left many technical issues in material and process, NCFET & phase FET would have lots of potential as a CMOS replacement (or extension) device.
EK and JS equally contributed to this work. All authors read and approved the final manuscript.
Consent for publication
Ethical approval and consent to participate
The authors declare that they have no competing interests.
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. 2017R1A2A2A05069708). Further, this work was supported by the Future Semiconductor Device Technology Development Program (10067746) funded by the Ministry of Trade, Industry & Energy (MOTIE) and the Korea Semiconductor Research Consortium (KSRC).
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 1.G.E. Moore, Electronics 38, 114–117 (1965)Google Scholar
- 2.C. Auth, A. Aliyarukunju, M. Asoro, D. Bergstrom, V. Bhagwat, J. Birdsall, N. Bisnik, M. Buehler, V. Chikarmane, G. Ding, Q. Fu, H. Gomez, W. Han, D. Hanken, M. Haran, M. Hattendorf, R. Heussner, H. Hiramatsu, B. Ho, S. Jaloviar, I. Jin, S. Joshi, S. Kirby, S. Kosaraju, H. Kothari, G. Leatherman, K. Lee, J. Leib, A. Madhavan, K. Marla, H. Meyer, T. Mule, C. Parker, S. Parthasarathy, C. Pelto, L. Pipes, I. Post, M. Prince, A. Rahman, S. Rajamani, A. Saha, J.D. Santos, M. Sharma, V. Sharma, J. Shin, P. Shinha, P. Smith, M. Sprinkle, A.S. Amour, C. Staus, R. Suri, D. Towner, A. Tripathi, A. Tura, C. Ward, A. Yeoh, In: Proc. IEEE IEDM 29.1(2017)Google Scholar
- 16.P. Sharma, J. Zhang, K.N. Ni, S. Datta, IEEE Elect. Dev. Lett. https://doi.org/10.1109/LED.2017.2782261
- 22.S. Dasgupta, A. Rajashekhar, K. Majumdar, N. Agrawal, A. Razavieh, S.T. Mckinstry, S. Datta, IEEE J. Explor. SolidState Comp. Dev. Circ. 1, 43–48 (2015)Google Scholar
- 23.P. Sharma, K. Tapily, A.K. Saha, J. Zhang, A. Shaughnessy, A. Aziz, G.L. Snider, S. Gupta, R.D. Clark, S. Datta, IEEE VLSI, T154–T155(2017)Google Scholar
- 25.K.S. Li, P.G. Chen, T.Y. Lai, C.H. Lin, C.C. Cheng, C.C. Chen, Y.J. Wei, Y.F. Hou, M.H. Liao, M.H. Lee, M.C. Chen, J.M. Sheih, W.K. Yeh, F.L. Yang, S. Salahuddin, C. Hu, IEEE IEDM 22, 6 (2015)Google Scholar
- 27.Z. Krivokapic, U. Rana, R. Galatage, A. Razavieh, A. Aziz, J. Liu, J. Shi, J.H. Kim, R. Sporer, C. Serrao, A. Busquet, P. Polakowski, J. Muller, W. Kleemeier, A. Jacob, D. Brown, A. Knorr, R. Carter, S. Banna, IEEE IEDM 15, 1 (2017)Google Scholar
- 29.FA McGuire, YC Lin, B Rayner, AD Franklin, in IEEE 75th annual device research conference, 7999478(2017)Google Scholar
- 31.A.M. Ionescu, Nat. Nanotechnol. 13, 7–8 (2018). https://doi.org/10.1038/s41565-017-0046-2
- 34.T. Srimani, G. Hills, M.D. Bishop, U. Radhakrishna, A. Zubair, R.S. Park, Y. Stein, T. Palacios, D. Antoniadis, M.M. Shulaker, IEEE Electron Dev. Lett. https://doi.org/10.1109/LED.2017.2781901
- 35.H. Ota, S. Migita, J. Hattori, K. Fukuda, A. Toriumi, in Proceedings of the 16th international conference on nanotechnology, 2016Google Scholar
- 37.A. Saeidi, F. Jazaeri, F. Bellando, I. Stolichnov, C.C. Enz, A.M. Ionescu, in 47th European Solid-State Device Research Conference (ESSDERC) 78–81Google Scholar
- 42.M.H. Lee, S.T. Fan, C.H. Tan, P.G. Chen, Y.C. Chou, H.H. Chen, J.Y. Kuo, M.J. Xie, S.N. Liu, M.H. Liao, C.A. Jong, K.S. Li, M.C. Chen, C.W. Liu, IEEE IEDM 12, 1 (2016)Google Scholar
- 44.E. Ko, H. Lee, Y. Goh, S. Jeon, C. Shin, J. Electron Dev. Soc. 5, 306–309 (2017)Google Scholar
- 45.M.H. Lee, P.G. Chen, S.T. Fan, C.Y. Kuo, H.H. Chen, S.S. Gu, Y.C. Chou, C.H. Tang, R.C. Hong, Z.Y. Wang, M.H. Liao, K.S. Li, M.C. Chen, C.W. Liu, IEEE VLSI-TSA 7942466(2017)Google Scholar
- 54.S.C. Chang, U.E. Avci, D.E. Nikonov, I.A. Young, IEEE J. Elxplor. SolidState Comput. Dev. Circ. 3, 56–64 (2017)Google Scholar
- 58.S. Smith, K. Chatterjee, S. Salahuddin, IEEE Trans. Elect. Dev. 65, 295–298 (2018). https://doi.org/10.1109/TED.2017.2772780
- 61.T Dutta, G Pahwa, A Agarwal, YS Chauhan, IEEE Electron Dev. Lett. 39, 147–150 (2018). https://doi.org/10.1109/LED.2017.2770158
- 64.S.K. Samal, S. Khandelwal, A.I. Khan, S. Salahuddin, C. Hu, S.K. Lim, in 2017 IEEE/ACM international symposium on low power electronics and design (ISLPED)Google Scholar
- 85.J. Frougier, N. Shukla, D. Deng, M. Jerry, A. Aziz, L. Liu, G. Lavallee, T.S. Mayer, S. Gupta, S. Datta, in IEEE Symposium on VLSI Tech. 228–229(2016)Google Scholar
- 90.S. Lim, J. Yoo, J. Song, J. Woo, J. Park, H. Hwang, IEEE IEDM 34, 7 (2016)Google Scholar
- 91.N. Shukla, B. Grisafe, R.K. Ghosh, N. Jao, A. Aziz, J. Frougier, M. Jerry, S. Sonde, S. Rouvimov, T. Orlova, S. Gupta, S. Datta, IEEE IEDM 34, 6 (2016)Google Scholar
- 92.J. Park, D. Lee, J. Yoo, H. Hwang, IEEE IEDM 23, 7 (2017)Google Scholar
- 96.J. Song, J. Park, K. Moon, J. Woo, S. Lim, J. Yoo, D. Lee, H. Hwang, IEEE IEDM 25, 3 (2016)Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.