https://orcid.org/0000-0003-0476-7527
Figures
Abstract
A nanoscale nonvolatile bidirectional reconfigurable field effect transistor (NBRFET) based on source /drain (S/D) self programmable floating gates is proposed. Comparing to the conventional reconfigurable field effect transistor (RFET) which requires two independently powered gates, the proposed NBRFET requires only one control gate. Beside, S/D floating gates are introduced. Reconfigurable function is realized by programming different types of charges into the S/D floating gates through biasing the gate at a positive or negative high voltage. The effective voltages of the S/D floating gates are determined jointly by the quantity of the charge stored in the S/D floating gates and the gate voltage. In addition, the charge stored in the floating gate has an effect of reducing the energy band bending near the source/drain regions when the gate is reversely biased, thereafter, the band to band tunneling (BTBT) leakage current can be largely decreased. The scale of the proposed NBRFET can be reduced to nanometer level. The device performances such as the transfer and output characteristics are verified by device simulation, which proves that the proposed NBRFET has very good performance in the nanometer scale.
Citation: Jin X, Zhang S, Liu X (2023) A nonvolatile bidirectional reconfigurable FET based on S/D self programmable floating gates. PLoS ONE 18(5): e0284616. https://doi.org/10.1371/journal.pone.0284616
Editor: Farooq Ahmad Khanday, University of Kashmir, INDIA
Received: January 1, 2023; Accepted: April 5, 2023; Published: May 24, 2023
Copyright: © 2023 Jin et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting information files.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The smallest unit of CMOS integrated circuit is MOSFET [1,2]. With the continuous reduction of MOSFET size and the increasingly physical limit of photolithography technology, in order to continue Moore’s law, new technologies must be tried to develop [3]. Although in recent years, research on quantum computing and integrated circuit devices based on new materials and the development of related integrated circuits has been ongoing [4,5], however, as the mainstream silicon-based technology, due to its mature process technology, various new ideas for improvement also emerge in endlessly. Reducing the number of devices required for logic function modules is a soft way to improve the integration of integrated circuits under the restriction that lithography technology cannot be further broken through. The reconfigurable field effect transistor (RFET) aroused the attention of the academic community in recent years. It can actually be seen as an expanded application of dopingless and electrostatic doped based devices [6–9]. However, compared with today’s mainstream technology, the reported RFET size is much larger. However, compared with the current technology, the reported RFET size is much larger. At nanoscale size, it is still unclear whether RFET can work properly like the FinFET technology [10–13]. RFET usually does not need to be doped, and its source and drain electrons or holes are generated by applying positive or negative voltage to the programming gate and through tunneling effect [14–16]. Therefore, comparing to the mainstream technology, RFET needs to add a program gate. The program gate makes the interconnection more complicated. For the device whose size has been reduced to nanometer level, if RFET is reversely biased, the energy band bending will be enhanced due to that the program gate is continuously positively or negatively biased, resulting in the band to band tunneling and leakage current is increased. To reduce the leakage current, we have proposed a nanoscale nonvolatile bidirectional RFET (NBRFET) based on S/D self programmable floating gates in this work. Comparing to conventional RFET which requires two independently powered gates, the proposed NBRFET requires only one control gate and S/D floating gates to achieve the nonvolatile function. Reconfigurable function is realized by programming different types of charges into the S/D floating gates through biasing the gate at a positive or negative high voltage. The effective voltages of the S/D floating gates are determined jointly by the amount of the charge stored in the S/D floating gates and the gate voltage. It should be noted that the gate voltage has a coupling effect on the effective voltage of the floating gates. When the gate electrode is positive biased and the floating gate is already programmed with appropriate amount of positive charges, the effective voltage in the S/D floating gates is higher than the gate voltage. Compared with conventional RFET, NBRFET can achieve higher conduction current under the same gate voltage due to this coupling effect. On the contrary, when the gate electrode is reversely biased, this coupling effect makes the effect voltage of the S/D floating gates lower than the forwardly biased case, then the potential differences between the gate electrode and the S/D floating gates are relatively reduced compared with RFET which has a fixed voltage of program gate, which is conducive to reducing the energy band bending and the corresponding band-to-band tunneling induced leakage current in the reversely biased state. The device scale can be reduced to nanometer level. The device performances such as the transfer and output characteristics and the reconfigurable function are verified by device simulation, which proves that the proposed NBRFET can work better in the nanometer scale.
Methods
Fig 1(a) is a top view of NBRFET, Fig 1(b) is a cross view along the line A in Fig 1(a) and 1(c) is a cross view along line B in Fig 1(a). Fig 1(d) is a top view of NBRFET along line A in Fig 1(b). NiSi is adopted to form Schottky barriers on the interfaces between the source / drain and the silicon. The barrier height between the S/D electrodes and conduction band of silicon is 0.6eV, the barrier height between the S/D electrodes and the valance band of silicon is 0.48eV [17]. Lsi is the length of silicon between the S/D electrodes, Hsi is the height of silicon, and Wsi is the width of silicon. LCG is the length of the control gate in the central region of the device, tFG is the thickness of the floating gate, tox1 is the thickness of the gate oxide between the silicon and control gate / floating gate, tox2 is the thickness of the insulating layer between the control gate and the floating gate on both sides, and ttunnel is the thickness of the tunneling layer between the S/D electrodes and gate oxide.
Device manufacturing is fully compatible with the current CMOS technology. A brief fabrication flow of the proposed NBTFET is shown in Figs 1 and 2. As shown in Fig 2(a), 2(b) and 2(c), prepare a SOI wafer, after removing the surrounding silicon through the photolithography and etching processes, a rectangular silicon area is formed. As shown in Fig 2(d), 2(e) and 2(f), through the deposition process, the insulating dielectric material for the formation of gate oxide layer is deposited. After flattening the surface of the insulating dielectric material through CMP process, the silicon film is exposed again. As shown in Fig 2(g), 2(h), 2(i) and 2(j), remove parts of the insulating dielectric material through photolithography and etching processes, then deposit metal. After flattening the surface of the metal layer to expose the silicon film and gate oxide, both the FG and Gate are initially formed. As shown in Fig 2(k), 2(l) and 2(m), deposit the insulating dielectric material again to further form the gate oxide layer. Then remove insulating dielectric material around through photolithography and etching processes to expose the gate electrode. Then deposit metal again and flatten the surface to expose gate oxide layer, and the gate and gate oxide are further formed. As shown in Fig 1(a), 1(b), 1(c) and 1(d), partially remove the central part of the insulating dielectric material layer, then deposit metal again and flatten the surface to expose the insulating dielectric material layer. The gate electrode is finally formed. Then remove some parts of gate oxide layer and silicon film to expose buried oxide through photolithography and etching processes for reserving space of source and drain electrodes. Then deposit Ni, after annealing, the NiSi interfaces and the Schottky barrier between Ni and silicon are formed on both source and drain sides.
Analysis and discussions
The performances of the NBRFET and the comparison with conventional BRFET are verified through TCAD simulation [18]. All physics models such as CVT mobility, auger recombination, band gap narrowing, standard BTBT, Fnord tunneling and a compact density gradient quantum confinement model are all turned on. The operation of writing / erasing charge in floating gate is realized by grounding the source electrode and the drain electrode, and simultaneously applying a higher positive or negative voltage to the gate electrode. Fig 3(a) shows the dependence between the charge stored in the floating gate near the source or the floating gate near the drain (QSFG or QDFG) and the programming time under different gate voltage VGs. QSFG or QDFG is roughly proportional to the programming time, and the programming speed is roughly proportional to VG. Therefore, the programming time can be shortened by applying a higher negative or positive VG. Fig 3(b) shows the dependence between the charge stored in the floating gate near the source or the floating gate near the drain with an initial QSFG / QDFG and the erasing time under different VGs. From the perspective of quantum mechanics, the erasure of electric charges is a probability problem, and increasing VG increases the probability of electric charges being erased. Similar to the programming process, the time required for the first erased charge is also roughly proportional to the VG. As a voltage is also required to be applied to the gate electrode during the reading and writing process of the proposed NBRFET, which is often lower than 1V, in order to analyze the impact of the reading voltage on QSFG / QDFG, we show the dependence between the charge stored in the floating gate near the source or the floating gate near the drain with an initial QSFG / QDFG and the erasing time under a low VG equals 1V with different tox1 and tox2s in Fig 3(c). When the thicknesses of tox1 and tox2 are increased, the average time required to erase the first charge in the floating gates will also increase. As Fig 3(c) shows, when tox1 equals to 1 nm, the minimum time required for the erased charge is about 10−8 seconds, while when tox1 equals to 2 nm, the minimum time required for the erased charge increases to about 10−6 seconds. This means that if the clock frequency of the read signal is greater than 100 MHz for the case that tox1 equals 1 nm, the charge stored in the floating gates will not be easily erased, while if the clock frequency of the read signal is greater than 1 MHz for the case that tox1 equals 2 nm, the charge stored in the floating gates will not be easily erased. For today’s technology, the read frequency is above 1GHz, so even if the thickness of tox1 is as low as 1nm, as long as the read signal clock frequency is high enough, the stored charge in the written floating gate is nonvolatile.
(a) the dependence between the charge stored in the floating gate near the source or the floating gate near the drain (QSFG or QDFG) and the programming time under different gate voltage VGs. (b) the dependence between the charge stored in the floating gate near the source or the floating gate near the drain with an initial QSFG / QDFG and the erasing time under different VGs. (c) the dependence between the charge stored in the floating gate near the source or the floating gate near the drain with an initial QSFG / QDFG and the erasing time under a low VG equals 1V with different tox1 and tox2s.
Fig 4(a) shows the relationship between the effective voltage of S/D floating gates (VSFG / VDFG) and gate voltage VG, and Fig 4(b) shows the relationship between drain to source current IDS and gate voltage VG with different QSFG / QDFG, respectively. VSFG represents the effective voltage of the source floating gate and VDFG represents the effective voltage of the drain floating gate. In general, the VSFG and VDFG both increase with the increase of VG and decreases with the decreasing of VG. Increasing QSFG / QDFG is helpful to increase the effective voltages of S/D floating gates in both conditions of forwardly and reversely biased states. For the forward bias state, the VSFG and VDFG both increases due to the increasing of QSFG and QDFG, leading to more band bending and stronger tunneling effect near the S/D regions in semiconductor. Therefore, the source electrode can provide more carriers from BTBT, so the forward conduction current increases. Due to that there is a coupling effect between VSFG / VDFG and VG. Therefore, the VSFG / VDFG can continuously increase with the increasing of VG. Therefore, the carriers near the source/drain can be continuously increased. For the reverse bias state, the increase of QSFG and QDFG reduces the effective potential difference between the drain electrode and the floating gate electrode, as well as the effective potential difference between the source electrode and the floating gate electrode, so that the energy band bending in the reverse state is greatly reduced. Therefore, it is helpful to eliminate the BTBT effect in the reverse state and reduce the generation of leakage current. As Fig 4(b) shows, there is an optimal QSFG/ QDFG, it can bring comprehensive improvements in forward current, reverse leakage current, and subthreshold characteristics.
Fig 5(a) shows the transfer characteristics comparison between the proposed NBRFET, dopingless (DL) TFET in [7] and the conventional BRFET with similar geometrical scales. According to the analysis of Fig 4(a), the effective voltage of VSFG/VDFG changes with the amount of charge in SFG and DFG, and they also change with the change of VG. There is a coupling effect between VSFG/VDFG and VG. Therefore, compared to the BRFETs with fixed VPGs, the VSFG and VDFG of NBRFET will increase with the increasing of VG and also decrease with the decreasing of VG. Therefore, when the gate of NBRFET is reversely biased, VSFG and VDFG have a lower effective voltage. As shown in Fig 5(a), when VG is less than -0.4V, the reverse leakage current of the NBRFET is equivalent to that of BRFET with a VPG of 0.6V. As VG increases, the VSFG and VDFG also increase, the current generated by NBRFET is equivalent to the current generated by BRFETs with larger VPG s. In Fig 5(a), the curves of NBRFET and the ones of BRFETs with different VPG s have an intersection point. Referring to Fig 4(a), the VSFG and VDFG of the NBRFET corresponding to this intersection point are almost equal to the VPG of the BRFET corresponding to the intersection point. It can be clearly seen that the NBRFET with S/D floating gates charged to 2.4×10-16C has better forward conduction characteristics and lower reverse leakage current than ordinary BRFET with programming gate with a fixed VPG. Fig 5(b) shows the leakage current comparison between the proposed NBRFET and conventional BRFET with different drain voltage VDs. It can be seen that the reverse leakage current generated by NBRFET is smaller than that generated by BRFET in a large range of drain voltage.
(a) The transfer characteristics comparison between the proposed NBRFET and conventional BRFET with similar geometrical scales. (b) The leakage current comparison between the proposed NBRFET and conventional BRFET with different drain voltage VDs.
Fig 6(a) shows the comparison of electron concentration distribution between NBRFET and BRFET with positively biased gate electrode and positively charged floating gates. A positive drain voltage VD is also applied. Since the effective voltage in the floating gate is determined by both the type and amount of charges in the floating gate, as well as the gate voltage, the effective voltages of the S/D floating gates increase with the increasing of gate voltage due to coupling effect, thereafter the positively biased gate voltage increase the effective voltage of the S/D floating gates, and the S/D floating gates pulls down the energy band of the silicon regions on both sides of source and drain, and makes the band bending happen through electric field effect near both the source side and the drain side of the silicon region. The band bending induces the BTBT phenomena happen and carriers are generated. The generated holes can be filled by the electrons supplied by the source electrode, the generated electrons flows from the source side to the drain side along the electric field lines generated by the positive VD. Thus, the continuous drain to source current is formed. The proposed NBRFET thus works in turn on state. Fig 6(b) shows the comparison of electron concentration distribution between NBRFET and BRFET with negatively biased gate electrode and positively charged floating gates. Fig 6(c) shows the comparison of hole concentration distribution between NBRFET and BRFET with negatively biased gate electrode and positively charged floating gates. The negatively biased gate voltage decreases the effective voltage of the floating gates. The concentration of electron or hole on both sides of source and drain of NBRFET is much smaller than that of BRFET which PG is fixed at a relatively high voltage. Therefore, the source/drain resistance is increased with the decreasing of VG. Therefore, as Fig 4(a) shows, the reverse leakage current of NBRFET is as small as that of BRFET with a low PG voltage, while the on-state current of NBRFET is as large as that of BRFET with a high PG voltage.
(a) the comparison of electron concentration distribution between NBRFET and BRFET with positively biased gate electrode and positively charged floating gates (b) the comparison of the electron concentration distribution between NBRFET and BRFET with negatively biased gate electrode and positively charged floating gates. (c) the comparison of hole concentration distribution between NBRFET and BRFET with negatively biased gate electrode and positively charged floating gates.
Fig 7(a) shows the comparison of conduction band energy distribution between the proposed NBRFET and conventional BRFET in forwardly biased state. It can be clearly seen that the energy band contour lines in NBRFET is more intensive. Especially in the tunnel layer, the band bending amplitude reaches 0.45eV. For the BRFET, because its fixed programmed gate voltage does not change with the change of the gate voltage, the band bending amplitude is maintained at 0.35eV. This is the fundamental reason why the forward conduction current of NBRFET can be continuously increased by increasing VG, while the forward conduction current of BRFET will saturate for a large VG. Fig 7(b) shows the comparison of conduction band energy distribution between the proposed NBRFET and conventional BRFET in reversely biased state. It is also clear that due to the effective voltage of S/D floating gate decreases with the decreasing of gate voltage, the energy band contour distribution near the source/drain region in NBRFET becomes very sparse compared to the conventional BRFET, therefore, NBRFET can achieve much lower reverse leakage current under the same reversely biased gate voltage.
(a) the comparison of conduction band energy distribution between the proposed NBRFET and conventional BRFET in forwardly biased state. (b) The comparison of conduction band energy distribution between the proposed NBRFET and conventional BRFET in reversely biased state.
Fig 8(a) and 8(b) show the output characteristics between IDS and VD, and the reconfigurable characteristics of the proposed NBRFET, respectively. As shown in Fig 8(a), the output characteristics IDS is increasing with the increase of QSFG and QDFG. Compared with the conventional BRFET, the proposed NBRFET can obtain better output characteristics, when the S/D floating gates are stored with appropriate amount of charge. Fig 8(b) shows that the charge type stored in the floating gate can determine the conduction mode of the proposed NBRFET. When the S/D floating gates are stored with positive charge and the gate is also forwardly biased, electrons in the conduction band generated by the BTBT effect can flow to the drain electrode and the device operates in the turn-on state of n-mode. Similarly, when the S/D floating gates are stored with negative charge and the gate is reversely biased, the valence band holes generated by the BTBT effect can be allowed to pass through, and the device operates in p-mode.
(a) The output characteristics between IDS and VD. (b) The reconfigurable characteristics of the proposed NBRFET.
Fig 9(a) shows the IDS-VG dependence of NBRFET with different Lsis. Fig 9(b) shows the dependence between subthreshold swing and Lsi of NBRFET. As the distance between FG and gate decreases with the decreasing of Lsi, when the gate is reversely biased, the energy band bending between FG and Gate is also enhanced, therefore the reverse leakage current increases with the decreasing of Lsi. However, as the Lsi decreases, the gate controllability to the whole channel from source to drain is also increased. Therefore, as Fig 9(b) shows, the SS decreases with the decreasing of Lsis. There is a Compromise relationship between the reverse leakage and SS. The recommend optimal Lsi is about 14nm.
(a) the IDS-VG dependence of NBRFET with different Lsis. (b) The dependence between subthreshold swing and Lsi of NBRFET.
Fig 10 shows the IDS-VG dependence of NBRFET with different LCGs and Lsis. As LCG decreases, the controllability of Gate to the channel potential is weakened. Lower reverse VG is needed to cut off the channel of electrons. Therefore, changes in the LCG have an impact on the threshold voltage. However, this change has little effect on the forward conduction current and the reverse leakage current.
Conclusions
In this paper, we propose a nanoscale nonvolatile bidirectional reconfigurable field effect transistor (NBRFET) based on source /drain (S/D) self programmable floating gates. Different from the conventional RFET which requires two independently powered gates, the proposed NBRFET requires only one control gate and introduces S/D floating gates to realize the nonvolatile function. Reconfigurable function is realized by programming different types of charges into the S/D floating gates through biasing the gate at a positive or negative high voltage. The effective voltages of the S/D floating gates are determined jointly by the amount of the charge stored in the S/D floating gates and the gate voltage. In addition, the charge stored in the floating gate has an effect of reducing the energy band bending near the source/drain regions when the gate is reversely biased, thereafter, the BTBT leakage current can be largely decreased. The device scale can be reduced to nanometer level. The device performances such as the transfer and output characteristics are verified by device simulation, which proves that the proposed NBRFET has very good performance in the nanometer scale.
References
- 1.
Auth C, Allen C., Blattner A., et al. A 22nm High Performance and Low-Power CMOS Technology Featuring Fully-Depleted Tri-Gate Transistors, Self Aligned Contacts and High Density MIM Capacitors[C]. IEEE VLSI Symposia—Technology, IEEE, Hawaii, USA, 2012: 131–132.
- 2. Kuhn K. Considerations for Ultimate CMOS Scaling[J]. IEEE Transactions on Electron Devices, 2012, 59(7): 1813–1828.
- 3.
Narasimha S, Jagannathan B, Ogino A, et al. A 7nm CMOS Technology Platform for Mobile and High Performance Compute Application[C]. IEEE International Electron Device Meeting, IEEE, San Francisco, USA, 2017: 29.5.1–29.5.4.
- 4. Shulaker M-M, Hills G, Patil N, et al. Carbon Nanotube Computer[J]. Nature, 2013, 501: 526–530.
- 5.
Meunier T, Urdampilleta M, Niegemann D, et al. Qubit Read-Out in Semiconductor Quantum Processors: Challenges and Perspectives[C]. IEEE International Electron Device Meeting, IEEE, San Francisco, USA, 2019: 741–744.
- 6. Kumar Naveen, Raman Ashish, performance assessment of the charge-plasma-based cylindrical GAA vertical nanowire TFET with impact of interface trap charges, IEEE TRANSACTIONS ON ELECTRON DEVICES, Vol.66, Issue 10, Page 4453–4460, OCT, 2019
- 7. Ashok Kumar Gupta Ashish Raman, Performance analysis of electrostatic plasma-based dopingless nanotube TFET, APPLIED PHYSICS A-MATERIALS SCIENCE & PROCESSING, Volume 126, Issue 7, Article Number 573, JUN 30 2020,
- 8. Anand Sunny Amin S. Intekhab, Sarin R. K., Analog performance investigation of dual electrode based doping-less tunnel FET, JOURNAL OF COMPUTATIONAL ELECTRONICS, Volume 15, Issue 1, Page 94–103, MAR, 2016
- 9. Kumar Naveen, Raman Ashish, Novel Design Approach of Extended Gate-On-Source Based Charge-Plasma Vertical-Nanowire TFET: Proposal and Extensive Analysis, IEEE TRANSACTIONS ON NANOTECHNOLOGY, Volume 19, Page 421–428, 2020,
- 10. Trommer J., Heinzig A., Slesazeck S., Mikolajick T., and Weber W. M., “Elementary aspects for circuit implementation of reconfigurable nanowire transistors,” Electron Device Letters, IEEE, vol. 35, no. 1, pp. 141–143, Jan. 2014.
- 11. Marchi M.D., Zhang J., Frache S., Sacchetto D., Gaillardon P., Leblebici Y., et al, “Configurable logic gates using polarity-controlled silicon nanowire gate-all-around FETs,” IEEE Electron Device Letters, vol. 35, no. 8, pp. 880–882, Aug. 2014.
- 12. Bhattacharjee A., Saikiran M., Dutta A., Anand B., and Dasgupta S., “Spacer engineering-based high-performance reconfigurable FET with low OFF current characteristics,” IEEE Electron Device Letters., vol. 36, no. 5, pp. 520–522, May 2015.
- 13. Bhattacharjee A. and Dasgupta S., “Impact of gate/spacer-channel underlap, gate oxide EOT, and scaling on the device characteristics of a DGRFET,” IEEE Trans. Electron Devices, vol. 64, no. 8, pp. 3063–3070, Aug. 2017.
- 14. Heinzig A., Slesazeck S., Kreupl F., Mikolajick T., and Weber W. M., “Reconfigurable silicon nanowire transistors,” Nano Letters, vol. 12, no. 1, pp. 119–124, Jan. 2012. pmid:22111808
- 15.
De Marchi M et al 2012 Polarity control in double-gate, gate all-around vertically stacked silicon nanowire FETs IEEE Int. Electron Devices Meeting (IEDM) (10–13 December) (IEEE).
- 16.
(2018). International Roadmap for Devices and Systems 2018 Edition. [Online]. https://irds.ieee.org/.
- 17. Hauenstein R. J., Schlesinger T. E., and McGill T. C., Schottky barrier height measurements of epitaxial NiSi2 on Si, Appl. Phys. Lett. 47, 853 (1985);
- 18.
https://www.silvaco.com/products/tcad/device_simulation/device_simulation.html.