Electronic integrated circuits (ICs) have advanced their performance through the continuous dimensional scaling of transistors. Following Moore's law, complementary metal–oxide–semiconductor technology doubles the number of transistors in logic and memory circuits every 2 years.1,21. E. Mollick, IEEE Ann. Hist. Comput. 28, 62–75 (2006). https://doi.org/10.1109/MAHC.2006.452. R. R. Schaller, IEEE Spectrum 34, 52–59 (1997). https://doi.org/10.1109/6.591665 With the emergence of the Internet of Things, the functional requirements of ICs have expanded to a wider range of applications, such as biomedical electronics, sensor technology, communication, neuromorphic computing, and quantum computing. Accordingly, the paradigm in the semiconductor business is shifting from “more Moore” to “more than Moore.”3–53. S. Wang, X. Liu, M. Xu, L. Liu, D. Yang, and P. Zhou, Nat. Mater. 21, 1225–1239 (2022). https://doi.org/10.1038/s41563-022-01383-24. L. Meng, N. Xin, C. Hu, H. A. Sabea, M. Zhang, H. Jiang, and X. Guo, Nat. Commun. 13, 1410 (2022). https://doi.org/10.1038/s41467-022-28999-x5. J. Shalf, Philos. Trans. R. Soc., A 378, 20190061 (2020). https://doi.org/10.1098/rsta.2019.0061In the emerging more than Moore applications, ferroelectric thin-film transistors (FeTFTs) are gaining significant technological attention owing to their fast switching speed, low voltage operation, and low fabrication temperature.6–106. W. Shin, J. Yim, J. H. Bae, J. K. Lee, S. Hong, J. Kim, and J. H. Lee, Mater. Horiz. 9, 1623 (2022). https://doi.org/10.1039/D2MH00340F7. C. P. Chou, Y. X. Lin, Y. K. Huang, C. Y. Chan, and Y. H. Wu, ACS Appl. Mater. Interfaces 12, 1014–1023 (2020). https://doi.org/10.1021/acsami.9b162318. W. C. Y. Ma, IEEE Trans. Electron Devices 69, 6072 (2022). https://doi.org/10.1109/TED.2022.32088479. H. Joh, M. Jung, J. Hwang, Y. Goh, T. Jung, and S. Jeon, ACS Appl. Mater. Interfaces 14, 1326–1333 (2022). https://doi.org/10.1021/acsami.1c1687310. D. Lehninger, M. Ellinger, T. Ali, S. Li, K. Mertens, M. Lederer, and K. Seidel, Adv. Electron. Mater. 7, 2100082 (2021). https://doi.org/10.1002/aelm.202100082 FeTFTs have exhibited excellent performance as chemical sensor platforms66. W. Shin, J. Yim, J. H. Bae, J. K. Lee, S. Hong, J. Kim, and J. H. Lee, Mater. Horiz. 9, 1623 (2022). https://doi.org/10.1039/D2MH00340F and artificial synapses in neuromorphic computing.7,97. C. P. Chou, Y. X. Lin, Y. K. Huang, C. Y. Chan, and Y. H. Wu, ACS Appl. Mater. Interfaces 12, 1014–1023 (2020). https://doi.org/10.1021/acsami.9b162319. H. Joh, M. Jung, J. Hwang, Y. Goh, T. Jung, and S. Jeon, ACS Appl. Mater. Interfaces 14, 1326–1333 (2022). https://doi.org/10.1021/acsami.1c16873 Accordingly, extensive efforts have been made to improve the FeTFTs performance. From a material engineering perspective, various thermal or plasma treatments have been proposed to improve the ferroelectricity in FeTFTs.11,1211. Mohit, K. I. Haga, and E. Tokumitsu, Jpn. J. Appl. Phys., Part 1 59, SMMB02 (2020). https://doi.org/10.35848/1347-4065/ab86de12. M. M. Hasan, M. M. Islam, R. N. Bukke, E. Tokumitsu, H. Y. Chu, S. C. Kim, and J. Jang, IEEE Electron Device Lett. 43, 725–728 (2022). https://doi.org/10.1109/LED.2022.3162325 From a structural perspective, various stack structures have been proposed. The most widely used stack is the metal–ferroelectric–insulator–semiconductor (MFIS) structure. However, it suffers from poor reliability and endurance due to the thermal budget and excessive field applied to the FE–insulator interface.1313. T. Ali, K.Mertens, K. Kühnel, M. Rudolph, S. Oehler, D. Lehninger, L. M. Eng et al., Nanotechnology 32, 425201 (2021). https://doi.org/10.1088/1361-6528/ac146c To circumvent these problems, a metal–FE–metal–insulator–semiconductor (MFMIS) structure has been proposed.14,1514. C. Sun, K. Han, S. Samanta, Q. Kong, J. Zhang, H. Xu, and X. Gong, in Symposium on VLSI Technology ( IEEE, 2021).15. C. Sun, Z. Zheng, K. Han, S. Samanta, J. Zhou, Q. Kong, and X. Gong, IEEE Electron Device Lett. 42, 1786–1789 (2021). https://doi.org/10.1109/LED.2021.3121677 In the MFMIS stack, the capacitance ratio between the MFM and MIS could be controlled by changing the metal-to-FE area ratio (AM/AF). Thus, low-power operations and improvements in the reliability and endurance can be achieved.Low-frequency noise (LFN) plays a vital role when MFMIS FeTFTs are integrated for the purpose of diverse applications. The LFN determines the noise level of the signal in sensor applications and, thus, defines the lowest detectable signal.16,1716. W. Liao, W. Wei, Y. Tong, A. Wong, W. K. Chim, and C. Zhu, IEEE Electron Device Lett. 38, 1146–1149 (2017). https://doi.org/10.1109/LED.2017.271695917. W. Shin, G. Jung, S. Hong, Y. Jeong, J. Park, D. Jang, and J. H. Lee, Sens. Actuators, B 318, 128087 (2020). https://doi.org/10.1016/j.snb.2020.128087 In neuromorphic computing, 1/f noise and random telegraph noise significantly affect the learning accuracy of neural networks.18–2018. Institute of Electrical and Electronics Engineers, in 49th European Solid-State Device Research Conference (ESSDERC) ( IEEE, 2019).19. D. Kang, S. Jang, S. Choi, and S. Kim, Semicond. Sci. Technol. 36, 124001 (2021). https://doi.org/10.1088/1361-6641/ac251c20. M. Zhao, B. Gao, J. Tang, H. Qian, and H. Wu, Appl. Phys. Rev. 7, 011301 (2020). https://doi.org/10.1063/1.5124915 Therefore, it is crucial to understand the LFN characteristics of the MFMIS FeTFTs. In particular, the effects of phonon scattering caused by polarization on the LFN characteristics should be demonstrated. However, despite their relevance and significance, none of these features are investigated in MFMIS FeTFTs.

In this Letter, based on the above discussion, we fabricate MFMIS FeTFTs with hafnium zirconium oxide (HZO) as an FE layer and indium–gallium–zinc oxide (IGZO) as a conductive channel. Subsequently, the LFN characteristics of the fabricated FeTFTs are investigated. The power spectral densities (PSDs) of FeTFTs with different AM/AF's are measured under various bias conditions. The noise-generation mechanisms in each device are demonstrated through a rigorous analysis.

Figures 1(a) and 1(b) show the schematic cross section of the fabricated MFMIS FeTFTs and the corresponding transmission electron microscopy (TEM) image, respectively. The FeTFTs were fabricated on a buffer oxide (SiO2, 300 nm). A 35 nm of Mo is deposited as a bottom gate (BG) via DC sputtering process and patterned by dry etching. Subsequently, the FE layer (HZO, 7.5 nm) and a floating gate (FG, TiN, 25 nm) were deposited by thermal atomic layer deposition (ALD) and plasma-enhanced ALD, respectively. Thermal annealing was then performed at 500 °C for 30 s under N2 ambient conditions to induce ferroelectricity in HZO. A dielectric layer (ZrO2, 8 nm) was deposited by thermal ALD. Amorphous IGZO was deposited via radio frequency sputtering as a conductive channel and then patterned by wet etching with a diluted HCl solution. Mo was deposited as the source and drain by DC sputtering and etched using an SC-1 solution. Finally, the FeTFTs are annealed at 350 °C for 1 h under O2 ambient conditions to stabilize the IGZO channel. The channel width (W)/length (L) of the device used in this study is 20/20 μm. Note that the AM (area of TiN gate) is fixed, and the AF (area of HZO) is modulated by changing the bottom gate area. As shown in Fig. 1(c), the ferroelectricity of HZO is investigated using positive-up-negative-down (PUND) measurements in the MFM stack with a frequency ( f ) of 100 kHz. Figure 1(d) shows the polarization vs voltage of the HZO, exhibiting a remnant polarization of 21.1 μC/cm2. For electrical and LFN measurements, a semiconductor device parameter analyzer (Agilent, B1500A), a low-noise current preamplifier (Stanford Research System, SR560), and a dynamic signal analyzer (Keysight, 35670A) were used.Figure 2(a) shows the double-sweep transfer characteristics (ID–VBG) of the fabricated FeTFTs with different AM/AF's. The bottom gate bias (VBG) is swept from –4.5 to 4.5 V at a fixed drain-to-source bias (VDS) of 0.1 V. In all the cases, the devices exhibit anticlockwise hysteresis originating from the polarization of HZO. With an increase in AM/AF from 2 to 6, the memory window at 10 nA is increased from 0.34 to 2.05 V. This is because a higher voltage is applied to HZO with an increase in AM/AF.Figure 2(b) shows the drain current (ID) PSD (SID) vs f of the FeTFTs with different AM/AF's. The PSD is measured at the erase state [high threshold voltage (Vth) state] to avoid the Vth shift during the PSD measurement. The VBG is changed to generate the ID that ranges from 10 nA to 1 μA in each device. Note that the two orders of the current range are enough to verify the noise source of the devices. In this study, the PSD is measured in the f range of 10 and 1.60 × 103 Hz. It took a total of 20 s to obtain the PSD in each bias condition. The PSD result shown is the average of 200 times PSD measurements, with one PSD measured for 0.1 s. In the measured f ranges, the devices exhibit 1/f γ noise behavior (γ = − lnSID/lnf ). It has been reported that the 1/f noise in TFTs originates from the random fluctuation of the carrier number owing to the trapping/de-trapping process to/from the defects inside the gate oxide. Such behavior can be explained by the carrier number fluctuation (CNF) model. The CNF model is expressed as follows:2121. G. Ghibaudo, Solid-State Electron. 32, 563–565 (1989). https://doi.org/10.1016/0038-1101(89)90113-5 SIDID2=(gmID)2q2kBTNTλWLCox2f,(1)where gm is the transconductance, q is the electron charge, kB is the Boltzmann constant, T is the temperature, NT is the volume trap density, λ is the tunneling attenuation coefficient, and Cox is the gate oxide capacitance per unit area. To verify the explanation of the LFN characteristics of the FeTFTs by the CNF model, the PSD is measured at various VBGs. Figures 3(a)–3(c) show the ID normalized PSD (SID/ID2) of the FeTFTs with AM/AF of 2, 4, and 6. Note that VDS is fixed at 0.1 V, and VBG is changed to generate IDs ranging from 20 nA to 1 μA.It is necessary to investigate the correlation between SID/ID2 and (gm/ID)2 to verify the origin of 1/f noise. Figures 4(a)–4(c) show SID/ID2 sampled at 10 Hz and (gm/ID)2 vs ID of the FeTFTs with AM/AF's of 2, 4, and 6, respectively. Note that the different values of the constant are multiplied by (gm/ID)2 in each device for fitting. In the case of the FeTFT with an AM/AF of 2, SID/ID2 and (gm/ID)2 exhibit the same tendency as ID in all operating regions, demonstrating that the 1/f noise of the device stems from the CNF. However, in FeTFTs with AM/AF's of 4 and 6, the deviation of SID/ID2 to (gm/ID)2 is observed in the low ID region, as indicated by the ellipses in Figs. 4(b) and 4(c). Note that the five devices for each case are measured and the results were reproducible. The deviation of SID/ID2 and (gm/ID)2 could be explained by the correlated mobility fluctuation (CMF) model by multiplying (1 ± αμeffCoxIDgm)2 to Eq. (1), where α is the scattering parameter that reflects how a variation in the oxide charge couples to the mobility and μeff is the effective mobility.2121. G. Ghibaudo, Solid-State Electron. 32, 563–565 (1989). https://doi.org/10.1016/0038-1101(89)90113-5 According to the CMF model, the trapped carriers at the defects in gate oxide additionally fluctuate the carrier mobility through the Coulombic force. However, the CMF model cannot be used to explain the LFN characteristics of the devices. Because the CMF term is proportional to ID/gm, the model can account for the deviation of SID/ID2 and (gm/ID)2 in the high ID region. However, as shown in Figs. 4(b) and 4(c), the deviation is observed in the low ID region.In IGZO TFTs, 1/f noise is not only attributed to the CNF at the gate oxide-conductive channel interface but also to the bulk carrier mobility fluctuation (BMF) in the IGZO channel.22,2322. W. Ye, Y. Liu, B. Wang, J. Huang, X. Xiong, and W. Deng, IEEE Trans. Electron Devices 69, 6154–6156 (2022). https://doi.org/10.1109/TED.2022.320627423. C. G. Theodorou, A. Tsormpatzoglou, C. A. Dimitriadis, S. A. Khan, M. K. Hatalis, J. Jomaah, and G. Ghibaudo, IEEE Electron Device Lett. 32, 898–900 (2011). https://doi.org/10.1109/LED.2011.2143386 When the TFT operates in the low ID region and a conductive channel is formed broadly across the channel, bulk conduction plays a significant role.2323. C. G. Theodorou, A. Tsormpatzoglou, C. A. Dimitriadis, S. A. Khan, M. K. Hatalis, J. Jomaah, and G. Ghibaudo, IEEE Electron Device Lett. 32, 898–900 (2011). https://doi.org/10.1109/LED.2011.2143386 In particular, the remote phonon scattering should be considered when the charge density in the conductive channel is low.2424. K. Maitra, M. M. Frank, V. Narayanan, V. Misra, and E. A. Cartier, J. Appl. Phys. 102, 114507 (2007). https://doi.org/10.1063/1.2821712 When HZO is polarized, the carrier mobility at the bulk of conductive channel fluctuates significantly owing to the remote phonon scattering. This is because the soft bonds in polarized FE materials, related to low-energy optical phonons, fluctuate carrier mobility.2525. W. Shin, J. H. Bae, S. Kim, K. Lee, D. Kwon, B. G. Park, and J. H. Lee, IEEE Electron Device Lett. 43, 13–16 (2022). https://doi.org/10.1109/LED.2021.3127175 When the BMF is dominant, SID/ID2 is expressed as follows:1717. W. Shin, G. Jung, S. Hong, Y. Jeong, J. Park, D. Jang, and J. H. Lee, Sens. Actuators, B 318, 128087 (2020). https://doi.org/10.1016/j.snb.2020.128087 SIDID2=αHμeff2kTfL2I,(2)where αH is Hooge's parameter that represents the magnitude of mobility scattering. According to (2), the SID/ID2 decreases with a power of −1 in the log–log plot of SID/ID2 vs ID, which is the case in the FeTFTs with AM/AF's of 4 and 6 operating in the low ID regions [Figs. 4(b) and 4(c)]. The polarization of HZO in the FeTFT increases with AM/AF, which can be confirmed by the increase in the memory window of FeTFTs [Fig. 2(a)]. Thus, a greater mobility fluctuation is induced in the low ID region in the cases of the FeTFTs with larger AM/AF's, increasing the 1/f noise. However, as the ID increases, the number of electrons at the IGZO channel increases, screening the remote phonon scattering.2424. K. Maitra, M. M. Frank, V. Narayanan, V. Misra, and E. A. Cartier, J. Appl. Phys. 102, 114507 (2007). https://doi.org/10.1063/1.2821712 This is why the noise generation mechanism is changed from BMF to CNF with an increase in ID, even in the FeTFTs with larger AM/AF's.One might speculate that the remote phonon scattering would be screened by the FG (TiN), and thus, the impact of remote phonon scattering in the devices should be insignificant. However, recent studies reported that the phonon scattering could be enhanced by the presence of a metal gate.2424. K. Maitra, M. M. Frank, V. Narayanan, V. Misra, and E. A. Cartier, J. Appl. Phys. 102, 114507 (2007). https://doi.org/10.1063/1.2821712 The boundary condition for electric potential at the channel–barrier interface can characterize the effect of the gate and the oxide on the channel. In this instance, the effects of the metal gate on the phonon scattering at a distance can be explained as follows: Scattering of electrons in the channel is caused by vibrational modes of the dielectric involving ionic displacements that generate large oscillating electric dipole fields. When a metal is present behind a thin dielectric, electrons respond quickly to the electric field generated by vibration. This generates a new field coming from image charges with the opposite sign. The modes whose dynamic dipole moments are aligned along the surface normal, which is the case for the device in this study, generate identical image dipoles, i.e., the total dipole moments are doubled, increasing scattering and thus reducing mobility.24,2624. K. Maitra, M. M. Frank, V. Narayanan, V. Misra, and E. A. Cartier, J. Appl. Phys. 102, 114507 (2007). https://doi.org/10.1063/1.282171226. B. Laikhtman and P. M. Solomon, J. Appl. Phys. 103, 014501 (2008). https://doi.org/10.1063/1.2826951The transition of the dominant noise source from carrier mobility fluctuation to CNF can be further confirmed by the different behavior of γ in relation to ID. Figure 4(d) shows the γ vs ID of the FeTFTs with AM/AF's of 2, 4, and 6. The γ of the FeTFT with AM/AF of 2 has the same value regardless of the change in ID. However, the γ value of the FeTFTs with AM/AF's of 4 and 6 changes with respect to ID. In the low ID region, the γ value is much smaller than one. This is because the scattering rate of mobility fluctuation is more significant at the higher frequency.2727. R. Jayaraman and C. G. Sodini, IEEE Trans. Electron Devices 36, 1773–1782 (1989). https://doi.org/10.1109/16.34242 However, as the dominant noise generation mechanism is changed to CNF with an increase in ID, the γ value increases and approaches that of the FeFET with AM/AF of 2.In this study, we fabricated IGZO FeTFTs with an MFMIS structure and investigated the LFN characteristics of the devices with different AM/AF's. It was shown that the noise generation mechanisms differ depending on the AM/AF and ID regions. In the low ID regions, excess noise generated by the polarized HZO is observed in the FeTFTs with larger AM/AF's owing to the increased mobility fluctuation caused by remote phonon scattering. In the high ID regions, 1/f noise is generated from the CNF regardless of the AM/AF. Considering that the ferroelectric TFTs with IGZO channels are seen as promising for new technologies in the future computing paradigm, understanding the LFN characteristics of this emerging devices should be crucial for its practical implementation.28–3228. W. Shin, J. H. Bae, J. Kim, R. H. Koo, J. J. Kim, D. Kwon, and J. H. Lee, Appl. Phys. Lett. 121, 163501 (2022). https://doi.org/10.1063/5.011130929. G. Liu, S. Rumyantsev, M. S. Shur, and A. A. Balandin, Appl. Phys. Lett. 102, 093111 (2013). https://doi.org/10.1063/1.479484330. W. Shin, R. H. Koo, K. K. Min, D. Kwon, J. J. Kim, D. Kwon, and J. H. Lee, IEEE Electron Device Lett. 44, 345–348 (2023). https://doi.org/10.1109/LED.2022.323180931. S. L. Rumyantsev, C. S. Jiang, R. Samnakay, M. S. Shur, and A. A. Balandin, IEEE Electron Device Lett. 36, 517–519 (2015). https://doi.org/10.1109/LED.2015.241253632. W. Shin, J. Kim, G. Jung, S. Ju, S. H. Park, Y. Jeong, and J. H. Lee, “ In‐memory‐computed low‐frequency noise spectroscopy for selective gas detection using a reducible metal oxide,” Adv. Sci. 10, 2205725 (2023). https://doi.org/10.1002/advs.202205725 We believe that the results in this study based on their extensive understanding of electronics and applied physics paves the way for the further development for ferroelectric-based TFTs.

This research was supported in part by the National Research Foundation (NRF) funded by the Korean Ministry of Science and ICT under Grant No. 2022M3I7A2085479, in part by Samsung Research Funding & Incubation Center of Samsung Electronics under Project No. SRFC-TA2103-02, and in part by the BK21 FOUR program of the Education and Research Program for Future ICT Pioneers, Seoul National University in 2023.

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

W. Shin and E. C. Park contributed equally to this work.

Wonjun Shin: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Writing – original draft (equal). Eun Chan Park: Conceptualization (equal); Formal analysis (equal); Methodology (lead). Ryun-Han Koo: Data curation (equal). Dongseok Kwon: Investigation (equal). Daewoong Kwon: Formal analysis (equal); Funding acquisition (equal); Supervision (equal). Jong-Ho Lee: Resources (equal); Supervision (equal); Writing – original draft (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

REFERENCES

1. E. Mollick, IEEE Ann. Hist. Comput. 28, 62–75 (2006). https://doi.org/10.1109/MAHC.2006.45, Google ScholarCrossref2. R. R. Schaller, IEEE Spectrum 34, 52–59 (1997). https://doi.org/10.1109/6.591665, Google ScholarCrossref3. S. Wang, X. Liu, M. Xu, L. Liu, D. Yang, and P. Zhou, Nat. Mater. 21, 1225–1239 (2022). https://doi.org/10.1038/s41563-022-01383-2, Google ScholarCrossref4. L. Meng, N. Xin, C. Hu, H. A. Sabea, M. Zhang, H. Jiang, and X. Guo, Nat. Commun. 13, 1410 (2022). https://doi.org/10.1038/s41467-022-28999-x, Google ScholarCrossref5. J. Shalf, Philos. Trans. R. Soc., A 378, 20190061 (2020). https://doi.org/10.1098/rsta.2019.0061, Google ScholarCrossref6. W. Shin, J. Yim, J. H. Bae, J. K. Lee, S. Hong, J. Kim, and J. H. Lee, Mater. Horiz. 9, 1623 (2022). https://doi.org/10.1039/D2MH00340F, Google ScholarCrossref7. C. P. Chou, Y. X. Lin, Y. K. Huang, C. Y. Chan, and Y. H. Wu, ACS Appl. Mater. Interfaces 12, 1014–1023 (2020). https://doi.org/10.1021/acsami.9b16231, Google ScholarCrossref, ISI8. W. C. Y. Ma, IEEE Trans. Electron Devices 69, 6072 (2022). https://doi.org/10.1109/TED.2022.3208847, Google ScholarCrossref9. H. Joh, M. Jung, J. Hwang, Y. Goh, T. Jung, and S. Jeon, ACS Appl. Mater. Interfaces 14, 1326–1333 (2022). https://doi.org/10.1021/acsami.1c16873, Google ScholarCrossref10. D. Lehninger, M. Ellinger, T. Ali, S. Li, K. Mertens, M. Lederer, and K. Seidel, Adv. Electron. Mater. 7, 2100082 (2021). https://doi.org/10.1002/aelm.202100082, Google ScholarCrossref11. Mohit, K. I. Haga, and E. Tokumitsu, Jpn. J. Appl. Phys., Part 1 59, SMMB02 (2020). https://doi.org/10.35848/1347-4065/ab86de, Google ScholarCrossref12. M. M. Hasan, M. M. Islam, R. N. Bukke, E. Tokumitsu, H. Y. Chu, S. C. Kim, and J. Jang, IEEE Electron Device Lett. 43, 725–728 (2022). https://doi.org/10.1109/LED.2022.3162325, Google ScholarCrossref13. T. Ali, K.Mertens, K. Kühnel, M. Rudolph, S. Oehler, D. Lehninger, L. M. Eng et al., Nanotechnology 32, 425201 (2021). https://doi.org/10.1088/1361-6528/ac146c, Google ScholarCrossref14. C. Sun, K. Han, S. Samanta, Q. Kong, J. Zhang, H. Xu, and X. Gong, in Symposium on VLSI Technology ( IEEE, 2021). Google Scholar15. C. Sun, Z. Zheng, K. Han, S. Samanta, J. Zhou, Q. Kong, and X. Gong, IEEE Electron Device Lett. 42, 1786–1789 (2021). https://doi.org/10.1109/LED.2021.3121677, Google ScholarCrossref16. W. Liao, W. Wei, Y. Tong, A. Wong, W. K. Chim, and C. Zhu, IEEE Electron Device Lett. 38, 1146–1149 (2017). https://doi.org/10.1109/LED.2017.2716959, Google ScholarCrossref17. W. Shin, G. Jung, S. Hong, Y. Jeong, J. Park, D. Jang, and J. H. Lee, Sens. Actuators, B 318, 128087 (2020). https://doi.org/10.1016/j.snb.2020.128087, Google ScholarCrossref18. Institute of Electrical and Electronics Engineers, in 49th European Solid-State Device Research Conference (ESSDERC) ( IEEE, 2019). Google Scholar19. D. Kang, S. Jang, S. Choi, and S. Kim, Semicond. Sci. Technol. 36, 124001 (2021). https://doi.org/10.1088/1361-6641/ac251c, Google ScholarCrossref20. M. Zhao, B. Gao, J. Tang, H. Qian, and H. Wu, Appl. Phys. Rev. 7, 011301 (2020). https://doi.org/10.1063/1.5124915, Google ScholarScitation, ISI21. G. Ghibaudo, Solid-State Electron. 32, 563–565 (1989). https://doi.org/10.1016/0038-1101(89)90113-5, Google ScholarCrossref22. W. Ye, Y. Liu, B. Wang, J. Huang, X. Xiong, and W. Deng, IEEE Trans. Electron Devices 69, 6154–6156 (2022). https://doi.org/10.1109/TED.2022.3206274, Google ScholarCrossref23. C. G. Theodorou, A. Tsormpatzoglou, C. A. Dimitriadis, S. A. Khan, M. K. Hatalis, J. Jomaah, and G. Ghibaudo, IEEE Electron Device Lett. 32, 898–900 (2011). https://doi.org/10.1109/LED.2011.2143386, Google ScholarCrossref, ISI24. K. Maitra, M. M. Frank, V. Narayanan, V. Misra, and E. A. Cartier, J. Appl. Phys. 102, 114507 (2007). https://doi.org/10.1063/1.2821712, Google ScholarScitation, ISI25. W. Shin, J. H. Bae, S. Kim, K. Lee, D. Kwon, B. G. Park, and J. H. Lee, IEEE Electron Device Lett. 43, 13–16 (2022). https://doi.org/10.1109/LED.2021.3127175, Google ScholarCrossref26. B. Laikhtman and P. M. Solomon, J. Appl. Phys. 103, 014501 (2008). https://doi.org/10.1063/1.2826951, Google ScholarScitation, ISI27. R. Jayaraman and C. G. Sodini, IEEE Trans. Electron Devices 36, 1773–1782 (1989). https://doi.org/10.1109/16.34242, Google ScholarCrossref, ISI28. W. Shin, J. H. Bae, J. Kim, R. H. Koo, J. J. Kim, D. Kwon, and J. H. Lee, Appl. Phys. Lett. 121, 163501 (2022). https://doi.org/10.1063/5.0111309, Google ScholarScitation29. G. Liu, S. Rumyantsev, M. S. Shur, and A. A. Balandin, Appl. Phys. Lett. 102, 093111 (2013). https://doi.org/10.1063/1.4794843, Google ScholarScitation, ISI30. W. Shin, R. H. Koo, K. K. Min, D. Kwon, J. J. Kim, D. Kwon, and J. H. Lee, IEEE Electron Device Lett. 44, 345–348 (2023). https://doi.org/10.1109/LED.2022.3231809, Google Scholar