# Fileset

[20231130_JJAP_manuscript_namiki et al.pdf](https://mdr.nims.go.jp/filesets/36a4e53c-63cc-4404-911f-504f3a9e27a3/download)

## Creator

[Wataru Namiki](https://orcid.org/0000-0003-4053-7366), [Takashi Tsuchiya](https://orcid.org/0000-0002-6950-6160), [Daiki Nishioka](https://orcid.org/0000-0002-3369-7700), Tohru Higuchi, [Kazuya Terabe](https://orcid.org/0000-0003-3988-3456)

## Rights

This is the version of the article before peer review or editing, as submitted by an author to Japanese Journal of Applied Physics.  IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or any version derived from it.  The Version of Record is available online at https://dx.doi.org/10.35848/1347-4065/ad1fb0[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Asymmetric transition of electrical resistance in an all-solid-state redox device with Fe<sub>3</sub>O<sub>4</sub> and Li-ion electrolyte thin films for physical reservoir computing](https://mdr.nims.go.jp/datasets/a82332ce-b216-4050-a7dc-c44b1dba3659)

## Fulltext

Template for JJAP Regular Papers (Mar. 2022) 1 Asymmetric transition of electrical resistance in an all-solid-state 1 redox device with Fe3O4 and Li-ion electrolyte thin films for 2 physical reservoir computing 3 Wataru Namiki1, Takashi Tsuchiya1*, Daiki Nishioka1,2, Tohru Higuchi2, and Kazuya 4 Terabe1 5 1National Institute for Materials Science, Tsukuba, Ibaraki 305-0044, Japan 6 2Tokyo University of Science, Katsushika, Tokyo 125-8585, Japan 7 E-mail: TSUCHIYA.Takashi@nims.go.jp 8  9 In recent years, ion-gating devices have been used in artificial neuromorphic computing and 10 achieved high performance for time-series data processing. However, the origin of such 11 performance still needs to be clarified. In this study, we fabricated an all-solid-state redox 12 device with functional material Fe3O4 and Li-ion conducting solid electrolytes, and the 13 transient response of the electrical resistance of the Fe3O4 thin film to time-series data input 14 was investigated. The transition between high and low electrical resistance states was 15 asymmetry, and residual Li-ion in the thin film led to a hysteresis effect. These unique 16 features, which is induced by ion-electron dynamics coupling, contributes to the high 17 performance of physical reservoir computing utilizing ion-gating device. 18   19   Template for JJAP Regular Papers (Mar. 2022) 2 1. Introduction 1 There is a strong demand to develop artificial intelligence devices to process information 2 that increases explosively and efficiently. Several studies have been recently conducted to 3 realize artificial neural network systems that can process various information on terminal 4 devices using physical devices.1-7) Recently, reservoir computing, which can process time 5 series data with a small learning cost, has been attracting attention, and attempts to process 6 time series data with high accuracy using nonlinear responses and short-term memory of 7 physical devices have been vigorously pursued.8-26) Among these, an all-solid-state ion-8 gating device, which is one of the ionic devices,27-36) can control the physical properties (not 9 only electrical properties but also magnetic and optical properties) of functional materials by 10 transporting ions,37-53) and the modulated electrical properties can be used in artificial neural 11 networks and artificial retina.54-59) It has been demonstrated that ion-gating performs highly 12 in image recognition and time series data prediction tasks.22-24) However, the origin of the 13 high performance needs to be clarified, and the details of the internal state change of the 14 functional material in response to time series data input, which is converted to voltage, need 15 to be investigated. In ion-gating, electronic carriers are injected into the functional material 16 via charge compensation along with the transport of ions.33-35,37-43,45-53) Therefore, measuring 17 electrical resistance is a powerful probe method to understand the internal state of functional 18 materials. Herein, we fabricated an all-solid-state reduction/oxidation (redox) device by 19 stacking a functional material and a solid electrolyte and focused on the transient response 20 of the electrical resistance of the functional materials to time-series data input duration to 21 clarify changes in the internal state (i.e., electrical resistance) of the functional materials. 22 In this study, we adopted Fe3O4, which is known to operate as an artificial neural network 23 device as well as to control the electrical and magnetic properties (magnetization and 24 magnetization direction) of Fe3O4 to a large extent by voltage when combined with a solid 25 electrolyte, for a functional material.43,48,60-62) Lithium-ions in the solid electrolyte diffuse 26 into the lattice of Fe3O4 by applying voltage, and such inserted ions modulate a number of 27 electronic carriers (>1021 cm-3) in Fe3O4 thin film.43,48, 63) Since Li-ions are driven by the 28 electric field that depends on the series resistance of the Fe3O4 thin film and the solid 29 electrolyte, it was found that by inputting time series data consisting of random voltages, the 30 effective electric field to Li-ion is modulated by the electrical resistance of Fe3O4 that 31   Template for JJAP Regular Papers (Mar. 2022) 3 changes sequentially and shows various transient responses. 1  2 2. Experimental methods 3 A 12-nm-thick magnetite thin film was deposited on a flat surface of a MgO (110) substrate 4 by pulsed laser deposition (PLD) using a magnetite polycrystalline target. The area of the 5 thin film was 500 ´ 700 μm2. The base pressure in the deposition chamber was 3.3 ´ 10-5 Pa. 6 The substrate temperature was kept at 573 K, A flow rate of the Ar (99 %) and O2 (1%) 7 mixture gas was fixed at 0.2 sccm, and the pressure was kept at 1.3 ´ 10-3 Pa during 8 deposition.48) A 5-nm-thick Ti adhesion layer and 50-nm-thick Pt electrode with a Hall-bar 9 shape were continuously deposited on the magnetite thin film by rf magnetron sputtering. 10 Then, A Li2O-SiO2-ZrO2 (LSZO) thin film was deposited by PLD, using a Li-excess LSZO 11 ceramic target, in a 3.4 Pa O2 atmosphere at a flow rate of 8.5 sccm. Finally, the Au gate 12 electrode was deposited on the LSZO electrolyte by rf magnetron sputtering. Metal shadow 13 masks were used to pattern each layer. 14  A micro-Raman spectroscope, which has incident light with a wavelength of 573 15 nm and a power of 0.5 mW to avoid thermal oxidation, was used to evaluate the quality of 16 the magnetite thin film. The exposure times and accumulation were set to 120 seconds and 17 20 times. 18  Electrical measurement was performed on the device in a high vacuum chamber (< 19 10-3 Pa) using a Keithley 4200-SCS parameter analyzer. Five pulse trains with different pulse 20 periods (50, 75, 100, 150, and 200 seconds) were prepared to investigate responses to various 21 inputs. Each pulse train had a base voltage of 1.0 V and an amplitude range of [0.5, 1.5 V]. 22 Five pulse trains were input one by one with time intervals in between. Drain current ID was 23 set to 3 μA during the measurement, and voltage drop VXX was acquired as information on 24 electrical resistance modulated by ion-gating. The following equation was used to calculate 25 the relaxation time t.57) 26 𝑉!! = 𝐴𝑒𝑥𝑝 '− )"#"!$*+ + 𝑉%  (1). 27 where A, t, t0, and V∞ denote the amplitude of VXX response, retention time, the pulse-on 28 finished time, and the base level of VXX at the corresponding pulse-on period. 29  30   Template for JJAP Regular Papers (Mar. 2022) 4 3. Results and discussion 1 3.1 Results 2 Figure 1(a) shows a schematic illustration of a fabricated device consisting of a magnetite 3 channel, electrodes for electrical measurement, LSZO electrolyte film, and gate electrodes. 4 Raman peaks, which are located at 669 cm-1, 542 cm-1, and 310 cm-1, of the magnetite thin 5 film were assigned by A1g, T2g(2), and T2g(3), and there is no impurity (e.g., FeO, Fe2O3, and 6 so on), as shown in Fig. 1(b).43,48) Figure 1(c) shows a transmission electron microscope 7 (TEM) image of cross-section of a fabricated device and its fast Fourier transformation 8 (FFT). It was successfully confirmed that the thin film was grown epitaxially on the surface 9 of MgO (110) substrate.48) 10  The electrical property of said device was evaluated by measuring drain current ID, 11 which was modulated by gate voltage VG sweeping, as shown in Fig. 2(a). ID flowing in the 12 channel thin film was enhanced as VG increased from 0.0 V to 2.0 V, resulting from electronic 13 carrier doping induced by Li-ion migrating to the Fe3O4 side. On the other hand, ID was 14 reduced as VG decreased from 1.0 V to 0.0 V, resulting from removing the electronic carrier 15 induced by Li-ion migrating to the gate electrode side. On this ID modulation, large hysteresis 16 was observed due to relatively slow migration in the Fe3O4, indicating that Li-ion was 17 inserted in and deserted from Fe3O4. Here, magnetite has 24 Fe ions, which are 8 tetrahedral 18 Fe ions (Fe8a) and 16 octahedral Fe ions (Fe16d), and 32 O ions (O32e) in a unit cell. Li-ion 19 inserted in the magnetite is located at the 16c site, which has an octahedral vacant center 20 surrounded by 6 oxygen ions, as shown in Fig. 2(b).64) While Fe8a ion is trivalent and has 21 five down-spins in d orbit, Fe16d ion has a mixed valence state due to the co-existence of 22 trivalent and divalent Fe ions, and trivalent Fe ion has five up-spins and divalent Fe ion has 23 a down-spin in addition to five up-spins. Down spin at Fe16d contributes to electrical 24 conduction.65) Thus, when the ID was enhanced (reduced), electron was doped (removed) in 25 trivalent (divalent) Fe16d through reduction (oxidation), increasing the number of down-spin 26 of Fe16d.43) Such electronic carrier tuning induced by insertion and desertion of Li-ion 27 showed large hysteresis with slow relaxation. This hysteresis effect indicates that the present 28 state has information based on the past state induced by input at the past step.23,24) We will 29 examine how this hysteresis affects the dynamics of ions and electrons during input of the 30 random wave used in the time-series data processing task. 31   Template for JJAP Regular Papers (Mar. 2022) 5  Figure 3(a) shows a measurement configuration of VXX response in proportion to 1 electrical resistance for time-series data inputs. A time-series voltage has a pulsed shape with 2 50 steps and is input as VG. ID was fixed at 3 µA, and potential drop VXX was acquired during 3 the VG inputs. The horizontal axis of the VXX signal was normalized to compare the shape of 4 the signals measured at various pulse periods. Figure 3(b) shows the VXX response for the 5 VG input in a cropped region of normalized time, corresponding to a time step k, ranging 6 from 559 to 579. Here, k is a pulse period with pulse-on and pulse-off. As the pulse period 7 of VG was extended, the degree of modulation in VXX became larger since the amount of 8 doped electronic carrier was increased during its application time. Then, VXX decreased 9 while maintaining said trend in the pulsed-off interval. In addition to such modulation rate 10 variation at various pulse periods, we evaluated the relaxation time of VXX response under 11 pulse voltage application. Figures. 3(c) and 3(d) show relaxation time variations in pulse-on 12 and pulse-off durations at various k. Although relaxation time in pulse-on duration was 13 longer than that in pulse-off duration (asymmetric transient response of VXX at each k) in the 14 entire range, the relaxation times of both durations increased as the pulse period extended. 15 Furthermore, relaxation time variation resulted from the input of time-series voltage, and it 16 was found that the relaxation time of electronic carrier tuning, induced by ion and electron 17 migrations, positively correlates with the amplitude of voltage input. Thus, an ion-gating 18 device shows a non-single transient response with various relaxation times by the input with 19 various intensities, and a variety of outputs can be obtained by inputting the time-series 20 voltage. 21  Fig. 4(a) shows the VXX response for input pulse voltages train in a cropped region 22 of normalized time ranging from 750 (time step of k39) to 850 (time step of k42) to investigate 23 a detail of relaxation time, which depends on input amplitude. The difference is that the 24 amount of change in VXX was more significant as the pulse period became longer and resulted 25 from the more considerable amount of tuned electronic carrier density due to the longer 26 application time of VG. VXX amplitude variation at each k was not in good agreement with 27 input amplitude variation, shown as red and blue dashed lines described in the figures for 28 pulse-on and pulse-off. This inconsistency results from said asymmetric transition response, 29 induced by ion-gating, of the electrical resistance state. Those relaxation times in pulse-on 30 and pulse-off durations are plotted in the upper and lower panels of Fig. 4(b). The extension 31   Template for JJAP Regular Papers (Mar. 2022) 6 of the pulse period not only lengthened the relaxation time but also changed the shape of the 1 k-dependence of the relaxation time as follows. For a pulse-on time, the variation of 2 relaxation time was slight for short pulse periods of 100 seconds or less, but the variation of 3 relaxation time increased for more extended pulse periods of 150 seconds or more, as shown 4 in an upper panel of Fig. 4(b). Then, the relaxation time of k39 was considerably expanded, 5 and the variation of relaxation time after k40 became small, as shown in the upper panel of 6 Fig. 4(c). 7 On the other hand, the change in the distribution of relaxation times for pulse-off 8 duration was different from that of the pulsed-on one, as shown in a lower panel of Fig. 4(b). 9 While up-turn convexity was shown below shorter intervals of 75 seconds, down-turn 10 convexity was shown above longer intervals of 150 seconds via approximate flat distribution 11 at the interval of 100 seconds. This is because the increase in relaxation time at k40 was 12 smaller than those at other ks, as shown in a lower panel of Fig. 4(c). From these results, 13 dynamics driven by voltage application in ion-gating devices showed the asymmetric 14 transition between high VXX and low VXX states. 15  16 3.2 Discussion 17  Here, we discuss the origins of asymmetric transition between high VXX and low 18 VXX states. Figure 4(d) shows schematic illustrations of transitions between the low electrical 19 resistance (R) state and the high R state of the channel of the ion-gating device. Ion-gating 20 leads to R modulation of the Fe3O4 thin film through redox reaction. The Fe3O4 thin film 21 possesses a large amount of Li-ions in its lattice in a low R state, and those Li-ions are 22 removed from the lattice by VG application.43,48) The Li-ion was driven by an electric field, 23 and the electric field depends on R of the thin film (RC) and R of the solid electrolyte (RSE). 24 This is because the electrical equivalent circuit between the gate and source electrodes is 25 described as a series circuit of RC and RSE. Here, the total R (Rtotal) can be described as 26 follows: 27 𝑅&'&() = 𝑅* + 𝑅+,,         (2) 28 Moreover, voltage potential between gate and source electrodes (i.e., VG) is expressed using 29 eq. (2) as shown below, 30 𝑉- = 𝐼-𝑅&'&() = 𝐼-(𝑅* + 𝑅+,).         (3) 31   Template for JJAP Regular Papers (Mar. 2022) 7 Then, eq. (3) can be transformed as follows. 1 𝐼- =./."= 0"(2#32$%),         (4) 2 where IG and Q denote electrical current flowing in the equivalent circuit, and the amount of 3 electrical charge, which corresponds to Li-ion of the solid electrolyte and electronic carrier 4 modulated by redox reaction, and dQ/dt shows the amount of flowing Q per a unit of time t. 5 When RC is reduced by VG application, IG and dQ/dt become large since VG and RSE are 6 constants. In short, the effective electric field affecting Li-ion varied with the RC state and 7 said the electric field was large (small) when RC was low (high) by Li-ion insertion 8 (desertion) due to a slight potential drop in the Fe3O4 and a significant potential drop in the 9 solid electrolyte. This variation in the electric field leads to variation in the transition speed 10 of the RC state, meaning that the transition from low RC to high RC is faster, and the transition 11 from high RC to low RC is slower. This is the origin of asymmetric transition response of VXX. 12 Furthermore, some of the inserted Li-ions remained in the thin film, resulting in the 13 inconsistency between VXX amplitude and input amplitude in Fig. 4(a). These ions contribute 14 to the hysteresis effect observed in the VXX response for time-series VG input. This study 15 found that ion-gating leads to electronic carrier density modulation accompanied by unique 16 features, such as asymmetric transition and hysteresis effect. These are critical features for 17 implementing physical reservoir computing, which requires nonlinearity (asymmetric 18 transition response) and short-term memory (hysteresis effect).22-24) 19  20 4. Conclusions 21 We fabricated an all-solid-state redox device with Fe3O4 and Li-ion electrolyte thin films 22 and measured the VXX response of the said device under a random VG train input to 23 investigate the origin of a performance as an artificial neuromorphic device through the 24 internal state (i.e., electrical resistance) of the Fe3O4 thin film. It was found that the transition 25 of the R state was asymmetric, and the information of VG input at the past step remained as 26 residual Li-ions in the thin film. This asymmetric behavior and hysteresis can satisfy 27 essential requirements (i.e., nonlinearity and short-term memory) for reservoir computing. 28 This study indicates that the asymmetry and hysteresis contribute to such performance. 29 Furthermore, the performance can be improved by modulation of these features appearing 30 in the device through material engineering to optimize its response speed.66) 31   Template for JJAP Regular Papers (Mar. 2022) 8  1 Acknowledgments 2 This work was supported in part by the Japan Society for the Promotion of Science (JSPS) 3 KAKENHI Grant numbers, JP22H04625 (Grant-in-Aid for Scientific Research on 4 Innovative Areas “Interface Ionics”) and JP22KJ2799 (Grant-in-Aid for JSPS Fellows), 5 and JST PRESTO (Grant number, JPMJGX23S2). Part of this work was supported by the 6 Iketani Science and Technology Foundation and JFE 21st century foundation. Part of this 7 work was supported by the Electron microcopy unit of the National Institute for Materials 8 Science. 9   10   Template for JJAP Regular Papers (Mar. 2022) 9 References 1 1) T. Ohno, T. Hasegawa, T. Tsuruoka, K. Terabe, J. K. Gimzewski, and M. Aono, Nat. Mater. 2 10, 591 (2011). 3 2) H. O. Sillin, R. Aguilera, H. -H. Shieh, A. V. Avizienis, M. Aono, A. Z. Stieg, and J. K. 4 Gimzewski, Nanotechnology 24, 384004 (2013). 5 3) R. Nur, T. Tsuchiya, K. Toprasertpong, K. Terabe, S. Takagi, and M. Takenaka, Nanoscale 6 14, 2013 (2022). 7 4) H.-M. Huang, Z. Wang, T. Wang, Y. Xiao, and X. Guo, Adv. Intell. Syst. 2, 2000149 (2020). 8 5) A. M. Hammond and R. M. Camacho, Opt. Exp. 27, 29620 (2019). 9 6) S. Jung, H. Lee, S. Myung, H. Kim, S. K. Yoon, S.-W. Kwon, Y. Ju, M. Kim, W. Yi, S. Han, 10 B. Kwon, B. Seo, K. Lee, G.-H. Koh, K. Lee, Y. Song, C. Choi, D. Ham and S. J. Kim, 11 Nature 601, 211 (2020). 12 7) J. M. Goodwill, N. Prasad, B. D. Hoskins, M. W. Daniels, A. Madhavan, L. Wan, T. S. 13 Santos, M. Tran, J. A. Katine, P. M. Braganca, M. D. Stiles, and J. J. McClelland, Phys. 14 Rev. Appl. 18, 014039 (2022). 15 8) H. Jaeger and H. Haas, Science 304, 78 (2004). 16 9) W. Maass, T. Natschläger, and H. Markram, Neural Comput. 14, 2531 (2002). 17 10) D. Verstraeten, B. Schrauwen, M. D’Haene, and D. Stroobandt, Neural Networks 20, 391 18 (2007). 19 11) L. Appeltant, M. C. Sorano, G. van der Sande, J. Danckaert, S. Massar, J. Dambre, B. 20 Schrauwen, C. R. Mirasso, and I. Fischer, Nat. Commun. 2, 468 (2011). 21 12) C. Du, F. Cai, M. A. Zidan, W. Ma, S. H. Lee, and W. D. Lu, Nat. Commun. 8. 2204 (2017). 22 13) S. Kan, K. Nakajima, T. Asai, and M. A. Kasaya, Adv. Sci. 9, 2104076 (2022). 23 14) M. A. Kasaya, Y. Takeshima, S. Kan, K. Nakajima, T. Oya, and T. Asai, Neuromorph. 24 Comput. and Eng. 2, 014003 (2022). 25 15) J. Torrejon, M. Riou, F. A. Araujo, S. Tsunegi, G. Khalsa, D. Querlioa, P. Bortolotti, V. Cros, 26 K Yakushiji, A. Fukushima, H. Kubota, S. Yuasa, M. D. Stiles, and J. Grollier, Nature 547, 27 428 (2017). 28 16) R. Nakane, G. Tanaka, and A. Hirose, IEEE Access 6, 4462 (2018). 29 17) R. Nakane, A. Hirose, and G. Tanaka, Phys. Rev. Res. 3, 033243 (2021). 30 18) R. Nakane, A. Hirose, and G. Tanaka, Phys. Rev. Appl. 19. 034047 (2023). 31   Template for JJAP Regular Papers (Mar. 2022) 10 19) W. Namiki, D. Nishioka, Y. Yamaguchi, T. Tsuchiya, T. Higuchi, and K. Terabe, Adv. Intell. 1 Syst. 2300238 (2023). 2 20) Q. Vinckier, F. Duport, A. Smerieri, K. Vandoorne, P. Bienstman, M. Haelterman, and S. 3 Massar, Optica 2, 438 (2015). 4 21) K. Nakajima, H. Hauser, T. Li, and R. Pfeifer, Sci. Rep. 5, 10487 (2015). 5 22) D. Nishioka, T. Tsuchiya, W. Namiki, M. Takayanagi, M. Imura, Y. Koide, T. Higuchi, and 6 K. Terabe, Sci. Adv. 8, eade1156 (2022). 7 23) T. Wada, D. Nishioka, W. Namiki, T. Tsuchiya, T. Higuchi, and K. Terabe, Adv. Intell. Syst. 8 5, 2300123 (2023). 9 24) K. Shibata, D. Nishioka, W. Namiki, T. Tsuchiya, T. Higuchi, and K. Terabe, Sci. Rep. 13, 10 21060 (2023). 11 25) B. Brazani, G. Dion, J.-F. Morissette, L. Beaudoin, and J. Sylvestre, and J. 12 Microelectromech. Sys. 29, 338 (2020). 13 26) J. C. Gartside, G. D. Stenning, A. Vanstone, H. H. Holder, D. M. Arroo, T. Dion, F. Caravelli, 14 H. Kurebayashi, and W. R. Branford, Nat. Nanotech. 17, 460 (2022). 15 27) D. Y. Kim, S. Miyoshi, T. Tsuchiya, and S. Yamaguchi, ECS Trans. 45, 161 (2012). 16 28) W. Namiki, M. Takayanagi, K. Horiba, M. Kobayashi, M. Minohara, H. Kumigashira, and 17 T. Higuchi. ECS Trans. 75, 83 (2017). 18 29) W. Namiki, T. Tsuchiya, M. Takayanagi, S. Furuichi, M. Minohara, M. Kobayashi, K. 19 Horiba, H. Kumigashira, and T. Higuchi, J. Phys. Soc. Jpn. 86, 074704 (2017). 20 30) W. Namiki, T. Tsuchiya, M. Takayanagi, K. Kawamura, T. Kawaguchi, and T. Higuchi, 21 Trans. Mater. Res. Soc. Jpn. 43, 175 (2018). 22 31) T. Tsuchiya, S. Miyoshi, Y. Yamashita, H. Yoshikawa, K. Terabe, K. Kobayashi, S. 23 Yamaguchi, Solid State Ionics 253, 110 (2013). 24 32) D. Nishioka, T. Tsuchiya, W. Namiki, M. Takayanagi, K. Kawamura, T. Fujita, R. Yukawa, 25 K. Horiba, H. Kumigashira, and T. Higuchi, Nanoscale Res. Lett. 15, 42 (2020). 26 33) K. Terabe, T. Tsuchiya, and T. Tsuruoka, Adv. Electron. Mater. 8, 2100645 (2022). 27 34) T. Tsuchiya, K. Terane. R. Yang, and M. Aono, Jpn. J. Appl. Phys. 55, 1102A4 (2016). 28 35) T. Tsuchiya, M. Imura, Y. Koide, and K. Terabe, Sci. Rep. 7, 10534 (2017). 29 36) D. Kan, T. Hatano, A. Abe, H. Ikuta, and Y. Shimakawa, Appl. Phys. Lett. 117, 231602 30 (2020). 31   Template for JJAP Regular Papers (Mar. 2022) 11 37) T. Tsuchiya, K. Terabe, and M. Aono, Appl. Phys. Lett. 103, 073110 (2013). 1 38) T. Tsuchiya, K. Terabe, and M. Aono, Adv. Mater. 26, 1087 (2014). 2 39) T. Tsuchiya, K. Terabe, and M. Aono, Appl. Phys. Lett. 105, 183101 (2014). 3 40) T. Tsuchiya, T. Tsuruoka, K. Terabe, and M. Aono, ACS Nano 9, 2102 (2015). 4 41) T. Tsuchiya, M. Ochi, T. Higuchi, K. Terabe, and M. Aono, ACS Appl. Mater. Inter. 7 12254 5 (2015). 6 42) T. Tsuchiya, S. Moriyama, K. Terabe, and M. Aono, Appl. Phys. Lett. 107, 013104 (2015). 7 43) T. Tsuchiya, K. Terabe, M. Pchi, T. Higuchi, M. Osada, Y. Yamashita, S. Ueda, and M. 8 Aono, ACS Nano 10, 1655 (2016). 9 44) K. Terabe, T. Tsuchiya, R. Yang, and M. Aono, Nanoscale 8, 13873 (2016). 10 45) T. Tsuchiya, M. Ochi, T. Higuchi, and K. Terabe, Jpn, J. Appl. Phys. 55, 06GJ03 (2016). 11 46) T. Tsuchiya, T. Tsuruoka, S.-J. Kim, K. Terabe, and M. Aono, Sci. Adv. 4, eaau2057 (2018). 12 47) D. Nishioka, T. Tsuchiya, T. Higuchi, and K. Terabe, Jpn. J. Appl. Phys. 59, SIIG09 (2020). 13 48) W. Namiki, T. Tsuchiya, M. Takayanagi, T. Higuchi, and K. Terabe, ACS Nano, 14, 16065 14 (2020). 15 49) T. Tsuchiya, M. Takayanagi, K. Mitsuishi, M. Imura, S. Ueda, Y. Koide, T. Higuchi, and K. 16 Terabe, Commun. Chem. 4, 117 (2021). 17 50) Y. Isoda, D. Kan, Y. Ogura, T. Majima, T. Tsuchiya, and Y. Shimakawa, Appl. Phys. Lett. 18 120, 091601 (2022). 19 51) Y. Isoda, D. Kan, T. Majima, and Y. Shimakawa, Appl. Phys. Exp. 16, 015506 (2023). 20 52) L. Xie, Y. Ishoda, T. Majima, Y. Shen, D. Kan, and Y. Shimakawa, J. Solid State 21 Electrochem. https://doi.org/10.1007/s10008-023-05759-5 22 53) T. Wada, W. Namiki, T. Tsuchiya, D. Kan, Y. Shimakawa, T. Higuchi, and K. Terabe, Jpn. 23 J. Appl. Phys. 61, SM1002 (2022). 24 54) E. J. Fuller, F. E. Gabaly, F. Léonard, S. Agarwal, S. J. Plimpton, R. B. Jacobs-Gedrim, C. 25 D. James, M. J. Marinella, and A. A. Talin, Adv. Mater. 29, 1604310 (2017). 26 55) C. -S. Yang, D.-S. Shang, N. Liu, E. Fuller, S. Agrawal, A. A. Talin, Y.-Q. Li, B.-G. 27 Shen, and Y. Sun, Adv. Funct. Mater. 28, 1804170 (2018). 28 56) D. Nishioka, T. Tsuchiya, T. Higuchi, and K. Terabe, Neuromorph. Comput. and Eng. 3, 29 034008 (2023). 30 57) M. Takayanagi, D. Nishioka, T. Tsuchiya, M. Imura, Y. Koide, T. Higuchi, and K. Terabe, 31   Template for JJAP Regular Papers (Mar. 2022) 12 Mater. Today Adv. 18, 100393 (2023). 1 58) H. N. Mohanty, T. Tsuruoka, J. R. Mohanty, and K. Terabe, ACS Appl. Mater. Interfaces 2 15, 19279 (2023). 3 59) X. Wan, T. Tsuruoka, and K. Terabe, Nano Lett. 21, 7938 (2021). 4 60) P. Monalisha, S. Li, S. G. Bhat, T. Jin, P. S. Ail Kumar, and S. N. Piramanayagam, J. Appl. 5 Phys. 133, 084901 (2023). 6 61) T. Miao, W. Liu, C. Huang, B. Cui, R. Chu, X. Zhao, X. Wu, S. Wu, J. Xie, H. Liu, J. Chen, 7 B. Cheng, and G. Hu, J. Mater. Chem. C 11, 7732 (2023). 8 62) V. P. Singh, C. P. Singh, H. Ranjan, and S. K. Pandey, Mater. Lett. 344, 134431 (2023). 9 63) A. F. Pacheco, J. Orna, J. M. De Teresa, P. A. Algarabel, L. Morellon, J. A. Pardo, M. R. 10 Ibarra, E. Kampert, U. Zeitler, Appl. Phys. Lett. 95, 262108 (2009). 11 64) C. N. Lininger, C. A. Cama, K. J. Takeuchi, A. C. Marschilok, E. S. Takeuchi, A. C. West, 12 M. S. Hybertsen, Chem. Mater. 30, 7922 (2018). 13 65) A. Yanase and K. Siratori, J. Physic. Soc. Jpn. 53, 312 (1984). 14 66) Xie, L., Isoda, Y., Majima, T. et al. Orientation-dependent electrochemical response of 15 LaSrNiO4 epitaxial films. J Solid State Electrochem (2023). 10.1007/s10008-023-05759-16 5 17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33   Template for JJAP Regular Papers (Mar. 2022) 13  1  2  3  4  5 Figure Captions 6 Fig. 1. (a) A schematic illustration of a fabricated all-solid-state redox device. 7 LSZO denotes a Li2O-SiO2-ZrO2 thin film. (b) A Raman spectrum of a Fe3O4 thin 8 film deposited on MgO single crystal. (c) TEM image and its fast Fourier transform 9 of a cross-section of the fabricated device shown in (a). 10  11  12 Fig. 2. (a) Normalized drain current ID variation at various gate voltage VG, measured with 13 sweeping VG. The sweeping rate is fixed at 0.9 mV/sec. (b) A part of a unit cell of Fe3O4 and 14 spin configuration in 3d orbits of Fe8a and Fe16d. The red arrow denotes doped down spin 15 with Li-ion (Li+) insertion into the lattice. 16  17  18 Fig. 3. (a) A measurement configuration of an all-solid-state redox device to acquire voltage 19 drop in Fe3O4 thin film VXX. S and D denote source and drain electrodes. (b) VXX response 20 for a voltage pulse at a time step k, corresponding to a region colored gray surrounded by a 21 dashed line. Relaxation time variations at various k in (c) pulse-on duration and (d) pulse-22 off duration. 23  24  25 Fig. 4. (a) VXX response at the cropped region ranging from 750 (k39) to 850 (k42). Red (Blue) 26 arrows denote VXX amplitude and VG amplitude in pulse-on (pulse-off) duration. (b) 27 Relaxation time variation at the region ranging from k39 to k42. (c) Pulse period dependence 28 of relaxation time at various time steps. (d) Schematic principle of transition speed variation 29 of redox reaction induced by pulse-on and pulse-off. 30  31  32   Template for JJAP Regular Papers (Mar. 2022) 14  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15 Fig.1. 16  17  18  19  20  21   Template for JJAP Regular Papers (Mar. 2022) 15  1  2  3  4  5  6  7  8  9  10 Fig. 2. 11  12   Template for JJAP Regular Papers (Mar. 2022) 16  1  2  3  4  5  6  7  8  9  10  11  12  13 Fig. 3 14  15  16  17  18  19  20   Template for JJAP Regular Papers (Mar. 2022) 17  1  2  3  4  5  6  7  8  9  10  11  12 Fig.4 13  14