Tunable oxygen vacancy defects for high-performance electrolyte-gated synaptic transistors

Abstract

Synaptic transistors are regarded as promising components for advanced artificial neural networks and hardware-based learning systems because they can emulate the fundamental biological synapse functions. One-dimensional indium zinc oxide (InZnO) nanowires, owing to their excellent charge transport and trapping properties, demonstrate tremendous potential in synaptic transistors. However, the carrier concentration in InZnO nanowires is susceptible to oxygen vacancies, which can severely influence the performance of the synaptic transistors. Herein, we present a facile and reliable scheme to control the synaptic transistor properties via an Ar plasma-assisted oxygen vacancy defect-tunable strategy. This adjusting strategy is based on the thermal diffusion of oxygen atoms bombarded by Ar ions, which increases the oxygen vacancy concentration on the surface of InZnO nanowires and further regulates the carrier concentration in the device channel. Compared with the untreated devices, the responsivity of the Ar plasma-treated devices is increased by 400%, and the memory effect is also enhanced by 230%. This oxygen vacancy regulation strategy provides a new avenue for fabricating high-performance neuromorphic computing systems.

Graphical abstract

Keywords

electrolyte-gated synaptic transistor / Ar plasma treatment / oxygen vacancy / InZnO nanowires

1 Introduction

Brain-inspired synaptic devices, which can simulate the information processing of the biological nervous system and neural morphological computing with low-energy consumption, could provide important support for the development of artificial intelligence and neuromorphic computing [1−3]. Plasticity is critical for the performance of synaptic devices, which can be applied to emulate brain activities such as learning, forgetting, and memory by applying pulse signals and maintaining weights [4−6]. Therefore, it is essential to effectively adjust the plasticity to improve the synaptic performance. However, in the synaptic transistor, plasticity is usually affected by the choice of channel material, surface property and oxygen vacancy (O_Vac) concentration in the material, etc. [7−12]. Among them, surface modification distinguished by its characteristics of strong controllability, wide selection of material and less influence on materials structure is a preferred method to improve the plasticity of synaptic transistors. One-dimensional metal oxide nanowires have unique electronic transport channels, excellent thermal stability and chemical robustness, and tunable oxygen vacancies, making them highly advantageous for applications in nanotechnology and synaptic devices. Compared with other metal oxides, indium zinc oxide (InZnO) nanowires as a channel material have higher electron mobility, better environmental stability, adjustable electrical conductivity, etc. [13−15]. In this work, InZnO nanowires with a lower In content (atomic ratio of In:Zn = 1:3), which cannot only adjust the electrical conductivity and but also is conducive to environmental protection were selected as channel materials. However, in one-dimensional nanowires, the large specific surface area makes the carrier concentration susceptible to oxygen vacancies especially on the surface, which degrades the electrical properties [16, 17]. Thus, the regulation of oxygen vacancy on the surface of nanowires is an important factor in improving synaptic plasticity [18−20]. So far, in order to obtain high-quality one-dimensional semiconducting nanowires, multiple surface modification methods have been proposed, such as coating metal or polymer onto the surface of nanowires, producing a core-shell structure and plasma-assisted etching. Among them, the Ar-based plasma etching technique is regarded as an effective surface modification method because of its simplicity and precise control of the oxygen vacancy concentration on the surface of the materials [21].

Herein, we proposed an Ar plasma-treatment strategy to adjust the O_Vac defects in the InZnO nanowires to modulate the carrier concentration in the electrolyte-gated synaptic transistors (EGSTs). With the increasing of O_Vac concentration in InZnO nanowires, the synaptic properties of EGSTs are enhanced. In order to reveal the underlying mechanism for the Ar plasma treatment strategy, the quantitative relationship between Ar plasma treatment time and O_Vac concentration was established. Optimal synaptic performance is achieved when the O_Vac concentration reaches 33.7%, and the synaptic response amplitude is 400% higher than that of the untreated devices. The mechanism can be mainly attributed to the bombardment of Ar plasma on the surface of nanowires, which promotes the thermal diffusion of oxygen atoms from the nanowires to the vacuum environment due to the heating effect [22]. The Ar ion bombardment effect leads to an increase of O_Vac concentration on the surface of InZnO nanowires, thereby further increasing the carrier concentration [23]. This work can provide an effective strategy for the development of high-performance multifunctional neuromorphic computing systems.

2 Results and discussion

Fig.1(a) presents a schematic diagram of a biological synapse between two neurons. Neural signals transmit from the pre-synaptic to the post-synaptic neuron. During this transmission, neurotransmitters are released from synaptic vesicles into the synaptic cleft, and subsequently bound to specific receptors on the post-synaptic membrane, instigating the ion channels [inset in Fig.1(a)] [24]. Interestingly, the artificial neuromorphic transistors based on InZnO nanowires can emulate this biological synaptic signal conveyance via distinct electrical write/read processes. The EGST is composed of the top probe gate, LiClO₄/PEO electrolytic dielectric, source (S)/drain (D) electrodes, and Si/SiO₂ substrate [Fig.1(b)]. Detailed experimental section in supplementary material. The channel width and length are 1000 μm and 100 μm, respectively. The top gate mimics the pre-synaptic membrane to facilitate ion migration within the LiClO₄/PEO electrolyte, and the InZnO nanowires act as the post-synaptic membrane. Then, the consequent output signal is conducted through the S/D electrodes.

As illustrated in Fig.1(c), upon applying a pulse voltage (V_gs) to the top probe gate, the lithium ions (Li⁺) within the electrolytes migrate toward the electrolyte/InZnO interface, while {\mathrm{C}\mathrm{l}\mathrm{O}}_4^{-} anions move toward the electrolyte/top gate interface, forming an electric double layer (EDL) [25]. The majority of the external electric field potential is localized at the Helmholtz plane, generating a large capacitance in the EDL [26], which could lead to the reduction of operating voltage and a consequent decrease in energy consumption. In biological neural circuits, excitatory postsynaptic currents (EPSCs) are generated by opening and closing ion channels to regulate cell membrane potential. Similarly, in artificial synapses, when removing the pulse voltage, the Li⁺ near the channel will slowly diffuse back into the electrolyte due to the concentration gradient. The slow relaxation process induces a progressive diminution in electron concentration, resulting in a gradual decrease of the output current between the S/D electrodes. The channel conductance is affected by carrier concentration, which can effectively mimic the inherent synaptic weight modulation process of biological system [27, 28]. In this experiment, we utilized an Ar plasma etching technique to modify the surface of InZnO nanowires to modulate the channel carrier concentration. The flow rate of ultra-pure Ar gas was maintained at 35 standard cubic centimeters per min, and the power was 50 W. The bombardment treatment times for the devices were set as 15, 30, and 45 s respectively. It was found that the nanowires with a plasma treatment of 30 s exhibit the optimal electrical performance and will be presented in the following section.

To understand the physical mechanism of the synaptic electrical performance treated by the plasma treatment, the state of the oxygen atoms in the InZnO nanowires was investigated thoroughly. X-ray photoelectron spectroscopy (XPS) is considered as a vital tool to characterize the content of oxygen atoms and their relative concentration. As shown in Fig.2(a), three Gaussian-shaped fitted peaks of O1s are located at 529.8, 530.8, and 532.1 eV, which can be ascribed to M−O lattice oxygen, O_Vac, and absorbed surface oxygen (−OH), respectively. As the duration of Ar plasma treatment increases from 0 to 30 s, the O_Vac concentration of InZnO nanowires exhibits an increase from 22.6% to 33.7%. However, as the plasma treatment time increases to 45 s, the O_Vac concentration subsequently declined to 24.1% [Fig.2(b)−(d)]. The variation of oxygen vacancy concentration in InZnO nanowires under Ar plasma treatment can be divided by the following two processes: (i) in n-type semiconductor materials, the electron concentration is inherently associated with O_Vac. The generation of an O_Vac is accompanied by the release of two free electrons, leading to an increase in carrier concentration [29]. Upon Ar plasma treatment within 30 s, oxygen atoms are desorbed from the InZnO surface, thereby generating numerous O_Vac on the surface. (ii) When the plasma treatment time is extended to 45 s, the excessive Ar plasma treatment destroys the surface structure of the nanowires and reduces the O_Vac concentration of the nanowires. Specifically, Fig. S1(a) presents scanning electron microscopy (SEM) images of untreated nanowires, exhibiting uniform morphology with smooth surfaces. As shown in Fig. S1(b), the nanowires treated for 30 s maintain structural integrity with smooth surfaces and uniform diameters, indicating no significant mechanical degradation under moderate plasma treatment duration. However, when the treatment duration reached 100 s, the nanowires exhibited significant surface roughness [Fig. S1(c)]. Critical structural degradation occurs at 180 s treatment [Fig. S1(d)], where nanowire fragmentation severely compromises carrier mobility and device performance. These results establish that excessive plasma treatment not only fails to increase O_Vac concentrations but also detrimentally impacts device functionality through irreversible structural damage. The above results demonstrate that suitable Ar plasma treatment time can effectively control the O_Vac concentration of InZnO nanowires, thus realizing the modulation of carrier concentration.

Next, underlying physical mechanism of the Ar plasma-assisted regulation on oxygen vacancy concentration in metal oxide nanowires was proposed, where there are two main supposed mechanisms: one possible mechanism is that the high-energy Ar ions bombardment causing oxygen atoms to overflow from the surface of the material, and the other mechanism is the thermal diffusion of oxygen atoms induced by Ar plasma treatment under surface heating [21]. Fig.3(a) illustrates a schematic diagram of the Ar plasma treatment device. During the Ar plasma treatment process, high-energy Ar ions selectively bombard the lighter oxygen atoms located on the surface of InZnO nanowires, which can reduce the hanging bonds and carbon contamination on the surface and generating substantial oxygen vacancies [19]. This phenomenon is attributed to the physical momentum transfer between the Ar ions and the surface oxygen atoms of the InZnO nanowires. As shown in Fig.3(b) and (c), the kinetic energy of the Ar ions produces a certain amount of heat on the surface of the InZnO nanowires. Because the binding energy between indium atom and oxygen atom is relatively weak, heat can activate oxygen atoms and promote the migration of oxygen atoms from the surface of nanowires to the vacuum environment, resulting in a large number of O_Vac. Furthermore, a substantial oxygen concentration gradient difference between the Ar plasma atmosphere and the InZnO nanowires can enhance the thermal diffusion of oxygen atoms from the nanowires to the external environment. As shown in Fig.3(d) and (e), in the EGST, the ion equilibrium state in the electrolyte is altered upon applying a positive spike voltage to the top gate, with anions ({\mathrm{C}\mathrm{l}\mathrm{O}}_4^{-}) in the electrolyte being attracted toward the electrolyte/top gate interface, and cations (Li⁺) being pushed toward the electrolyte/InZnO interface. The anions can induce an adjacent hole layer on the top gate side, whereas the cations can induce electrons on the surface of InZnO nanowires, thereby forming EDLs at the above-mentioned two interfaces. Following Ar plasma treatment, the O_Vac concentration on the surface of InZnO nanowires exhibits a significant increase, which consequently elevates the free electron concentration, and effectively modulates the carrier concentration within the device channel [29]. Additionally, we investigated the impact of Ar plasma treatment on the lattice structure of InZnO nanowires by X-ray diffraction (XRD) technique. It was observed that the lattice structure remains unchanged before and after Ar plasma treatment, exhibiting the same diffraction peaks of hexagonal ZnO (PDF#76-0704) and cubic In₂O₃ structure (PDF#71-2194) (Fig. S2), which proved that this Ar plasma treatment scheme can effectively modulate the properties of the synaptic devices, and has no obvious effect on the lattice structure of the nanowires.

Next, we will discuss the electrical performance of the EGSTs with different O_Vac concentrations on the basis of bioinspired operational mechanism. Figure S3 describes the transfer curves of the FETs/EGSTs with different O_Vac concentrations of 22.6%, 26.1%, 33.7% and 24.1%, respectively, under a source-drain voltage (V_ds) of 10 /1 V. It was observed that the transistor has the maximum switching current ratio when the O_Vac concentration reaches 33.7%, and the threshold voltage is close to 0 V. Typical EPSCs of InZnO EGSTs with different O_Vac concentrations were measured to determine the intrinsic relationship between the synaptic characteristics and the O_Vac concentration of the InZnO nanowires. As shown in Fig.4(a), when the O_Vac concentration reaches 33.7%, the EPSC shows the maximum peak (compared to the other EGSTs) and concurrently manifests a short-term memory effect. Specifically, when the pulse voltage is applied, the EPSC rises rapidly to a peak value. While, when the pulse voltage is removed, the EPSC gradually reduces to the resting value. The effect is triggered at a V_gs of 1 V, a pulse width of 50 ms, and a V_ds of 0.5 V. Figure S4(a) demonstrates comparable oxygen vacancy concentrations of ~26.0% at both 15 and 40 s plasma treatment durations. To investigate this phenomenon, a comparative analysis of EGST performance under these two distinct conditions was conducted. As illustrated in Fig. S4(b), systematic investigations were performed on the essential EPSC synaptic characteristics of InZnO EGSTs with varying oxygen vacancy concentrations under identical experimental parameters. The experimental results demonstrate that the EPSC peak values are 1.86 and 1.74 μA for treatment durations of 15 and 40 s, respectively. This performance discrepancy can be mainly attributed to structural degradation of individual nanowires (e.g., surface roughness) induced by excessive plasma treatment, which consequently reduce EGSTs performance. In order to further assess the response of EPSC, Fig.4(b) and (c) explore the relationship between EPSCs and V_gs. It can be seen that the current gain increases as the V_gs extends from 0.5 to 1.5 V [Fig.4(b)] and the pulse width ranges from 30 to 150 ms [Fig.4(c)]. The EPSC current gain is defined as the ratio of the amplitude of the N_th EPSC to the first EPSC amplitude. Remarkably, under an O_Vac concentration of 33.7%, the current gain exhibits the fastest rapid growth rate and attains the highest value (V_ds of 0.5 V). Specifically, when the gate voltage (V_gs) was set to 1.5 V, the EGST plasma-treated duration of 30 s demonstrated a current gain of 445% (G₁), significantly higher than that of 111% (G₂) for the untreated EGST, where G₁/G₂ ≈ 400% indicates the high efficiency of Ar plasma treatment for the artificial synaptic device.

The paired-pulse facilitation (PPF) is a critical figure-of-merit to determine the short-term synaptic plasticity, which can be enhanced by two consecutive signals with a short interval [30, 31]. Typically, a high PPF index represents strong synaptic plasticity, which holds significance for the adaptation and learning processes of neural networks, contributing substantially to the establishment and retention of memory traces. The PPF index ∆t can be fitted using the following exponential function:

(1)PPF=1+C_1\mathrm{e}\mathrm{x}\mathrm{p}\left(-{\displaystyle \frac{\mathrm{\Delta }t}{{\tau }_1}}\right)+C_2\mathrm{e}\mathrm{x}\mathrm{p}\left(-{\displaystyle \frac{\mathrm{\Delta }t}{{\tau }_2}}\right),

where C₁ and C₂ represent the initial amplitudes of the first and second EPSC, τ₁ and τ₂ are the corresponding characteristic relaxation times, and ∆t is the time interval between two pulses. As shown in Fig.4(d), the InZnO EGSTs exhibit the highest PPF index of 167% at an O_Vac concentration of 33.7%, exhibiting excellent performance (Table S1), and the index decreases exponentially with the increase of interval time. Due to the PPF being highly correlated with the time interval between the two successive pulses, the EPSC response manifests discernible selectivity for continuous signals across varied frequencies, which is of great significance for the realization of high-pass filtering [32, 33]. In addition, energy consumption is another important parameter for evaluating the performance of synaptic devices. The power consumption can be calculated by the following formula:

(2)P=I_{\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}\times V \times t,

where I_peak is the peak current of the EPSC, V is the read voltage, and t is the applied pulse width. As depicted in Fig.4(f), the energy consumption decreases as the O_Vac concentration increases from 22.6% to 33.7%, and it reaches the minimum of 53 PJ when the O_Vac concentration is 33.7%. While, the energy consumption further increases to 79 PJ as the treatment time increases to 45 s. It is evident that the device achieves its lowest energy consumption when the O_Vac concentration is 33.7%.

The efficiency of information transfer between two neurons in biology is defined as synaptic weights, and synaptic plasticity has been regarded as the capability to regulate the synaptic weights [34, 35]. Based on the above comparison, when the O_Vac concentration is 33.7%, the InZnO EGSTs exhibit optimal synaptic performance. Therefore, we conducted detailed research on the synaptic plasticity of this device. Figure S5 illustrates the dependence of EPSC on V_gs and pulse width, respectively. When the V_gs increases from 0.5 to 1.5 V with a step of 0.2 V, the EPSC increases from 2.1 to 8.2 μA. As the pulse width increases from 20 to 150 ms in the step of 30 ms, the EPSC increases from 2.7 to 8.1 μA, where the V_ds are fixed at 0.5 V. This is because as the pulse amplitude and width increase, more ions move to the interface, which induces more carriers and produces higher EPSC. As shown in Fig.4(g), when applying a sequential pulse with ∆t of 30 ms on the probe gate, two EPSC peaks can be observed, where the first peak A₁ and the second peak A₂ are 3.17 and 4.73 μA, respectively. The PPF is approximately 149% calculated as A₂/A₁. To verify high-pass filtering characteristics, the measurement is performed with V_gs and V_ds of 1 and 0.5 V, and the pulse width is set as 10 ms. As expected, the InZnO EGSTs demonstrates totally different EPSC responses to the signal stimulations of different frequencies [Fig.4(h)]. The EPSC decreases from 9.31 μA at 50 Hz to 4.8 μA at 1 Hz, showing a high selectivity for high-frequency signals. The amplitude of 4.87 μA is set as a low frequency threshold current, while the 7.14 μA is regarded as a high frequency threshold current, suggesting that the InZnO EGSTs can be implemented as a high-frequency pass filter [36−38]. Long-term synaptic plasticity plays an essential role in learning, forgetting, and storing memory, which can be realized by applying a strongly stimulated pulse signal. In Fig.4(i), as the V_gs increases from 0.5 to 2.5 V in steps of 0.5 V, the final resting potential of EPSC can be increased from 1.20 to 2.53 μA in comparison with the initial potential of 1.02 μA. As the time further increases, the resting EPSC does not undergo substantial changes, suggesting the formation of relatively long-term memory obtained in the InZnO EGSTs. The observed phenomenon is similar to the biological function, where the learning and memory effect of living organisms is closely related to learning time and frequency [39]. The relative retention characteristics by comparing the time required for current decay from peak values to 80% retention under identical 1.5 V pulse conditions. Specifically, the device with 33.7% oxygen vacancy concentration demonstrates a decay time of 97 s (T₁), whereas the untreated sample with 22.6% oxygen vacancy concentration requires only 43 s (T₂) to reach the same retention threshold (Fig. S6). The calculated ratio T₁/T₂ ≈ 230% highlights the memory retention property achieved through oxygen vacancy optimization.

To demonstrate the potential of the synaptic transistor for practical neural network applications, a three-layer artificial neural network (ANN) was implemented using NumPy, and successfully recognized 28 × 28-pixel handwritten digits (MINST DATA) through backpropagation. As shown in Fig.5(a), the ANN comprises three hierarchical layers: an input layer with neuron counts determined by the input categories, a hidden layer, and an output layer corresponding to the 10 digits (0–9). In Fig.5(b), five cycles curves of potentiation/depression based on the InZnO EGST driven by 100 identical pulses (V_gs = 0.5 V for potentiation, V_gs = −0.2 V for depression, pulse width of 10 ms and V_ds of 0.5 V), the recognition accuracy of the handwritten digits can be derived via ANN [40, 41]. Then the accuracy of the handwritten recognition applications for two versions were investigated, i.e., small digits and large digits. As presented in Fig.5(c) and (d) the maximum ideal values are 98.3% and 95.7% for small digits and large digits, respectively. In comparison, the accuracy based on experimental data is as high as 98.2% and 94.2% after 70 training epochs, respectively, which is close to the simulation value. Furthermore, the recognition accuracy of large digits can be envisioned with improved linearity and symmetry of conductance updates in future work. After the Ar plasma treatment, the same electrical stimulation could cause more changes in the synaptic weight of the device, which is similar to human synapses, and equivalent to increased plasticity. Further performance enhancements can also be achieved by optimizing the plasma processing power and processing time.

3 Conclusions and outlook

In summary, we proposed a reliable and facile approach to tailor an EGST via an Ar plasma-assisted oxygen vacancy defect-tunable strategy, which can effectively regulate the synaptic response amplitude by simply modulating the carrier concentration in the InZnO nanowires. The optimal synaptic performance is achieved at an O_Vac concentration of 33.7% and the synaptic response amplitude is four times higher than that of the untreated ones, which can be explained by the thermal diffusion of oxygen atoms bombarded by Ar ions. All these results have indicated the great potency of this Ar plasma-assisted oxygen vacancy defect-tunable strategy for fabrication of future high performance of synaptic devices, and their multifunctional neuromorphic computing systems.

References

[1] J. Li,Z. Shen,Y. Cao,X. Tu,C. Zhao, Y. Liu,Z. Wen, Artificial synapses enabled neuromorphic computing: From blueprints to reality, Nano Energy. 103(Part A), 107744 (2022)

[2] F. Zhang, C. Li, Z. Li, L. Dong, and J. Zhao, Recent progress in three-terminal artificial synapses based on 2D materials: From mechanisms to applications, Microsyst. Nanoeng. 9, 16 (2023)

[3] X. Liu, F. Wang, J. Su, Y. Zhou, and S. Ramakrishna, Bio‐inspired 3D artificial neuromorphic circuits, Adv. Funct. Mater. 32(22), 2113050 (2022)

[4] H. F. Ling, D. A. Koutsouras, S. Kazemzadeh, Y. van de Burgt, F. Yan, and P. Gkoupidenis, Electrolyte-gated transistors for synaptic electronics, neuromorphic computing, and adaptable biointerfacing, Appl. Phys. Rev. 7(1), 011307 (2020)

[5] J. W. Park,B. H. Kang,H. J. Kim, A review of low‐temperature solution‐processed metal oxide thin‐film transistors for flexible electronics, Adv. Funct. Mater. 30(20), 1904632 (2020)

[6] S. Seo, B. Kim, D. Kim, S. Park, T. R. Kim, J. Park, H. Jeong, S. O. Park, T. Park, H. Shin, M. S. Kim, Y. K. Choi, and S. Choi, The gate injection-based field-effect synapse transistor with linear conductance update for online training, Nat. Commun. 13(1), 6431 (2022)

[7] Y. Yang, H. Cui, S. Ke, M. Pei, K. Shi, C. Wan, and Q. Wan, Reservoir computing based on electric-double-layer coupled InGaZnO artificial synapse, Appl. Phys. Lett. 122(4), 043508 (2023)

[8] S. Zhang, J. Guo, L. Liu, H. Ruan, C. Kong, X. Yuan, B. Zhang, G. Gu, P. Cui, G. Cheng, and Z. Du, The self-powered artificial synapse mechanotactile sensing system by integrating triboelectric plasma and gas-ionic-gated graphene transistor, Nano Energy 91, 106660 (2022)

[9] H. Nakazawa, H. Ishii, and Y. Takamura, Electrical properties and parameter analysis of all-solid-state LaZrO/InO electric double-layer transistors, J. Appl. Phys. 134(2), 024502 (2023)

[10] C. Du, Y. Ren, Z. Qu, L. Gao, Y. Zhai, S. T. Han, and Y. Zhou, Synaptic transistors and neuromorphic systems based on carbon nano-materials, Nanoscale 13(16), 7498 (2021)

[11] X. Hu, H. Jiang, L. X. Lu, S. X. Zhao, Y. Li, L. Zhen, and C. Y. Xu, Revisiting the hetero‐interface of electrolyte/2D materials in an electric double layer device, Small 19(43), 2301798 (2023)

[12] F. Léonard and A. A. Talin, Electrical contacts to one- and two-dimensional nanomaterials, Nat. Nanotechnol. 6(12), 773 (2011)

[13] N. Gogurla and S. Kim, Self‐powered and imperceptible electronic tattoos based on silk protein nanofiber and carbon nanotubes for human–machine interfaces, Adv. Energy Mater. 11(29), 2100801 (2021)

[14] Y. Cai, J. Yang, F. Wang, S. Li, Y. Wang, X. Zhan, F. Wang, R. Cheng, Z. Wang, and J. He, Ultrasensitive solar-blind ultraviolet detection and optoelectronic neuromorphic computing using α-In₂Se₃ phototransistors, Front. Phys. (Beijing) 18(3), 33308 (2023)

[15] P. H. Chu, N. Kleinhenz, N. Persson, M. McBride, J. L. Hernandez, B. Fu, G. Zhang, and E. Reichmanis, Toward precision control of nanofiber orientation in conjugated polymer thin films: Impact on charge transport, Chem. Mater. 28(24), 9099 (2016)

[16] Y. Chang, H. Cong, R. Zhou, W. Zhang, Y. Qin, X. Liu, and F. Wang, Enhanced artificial synaptic properties enabled by arrays of electrolyte-gated electrospun InZnO nanowires, ACS Appl. Electron. Mater. 4(5), 2570 (2022)

[17] J. Oh, S. Park, S. H. Lee, S. Kim, H. Lee, C. Lee, W. Hong, J. H. Cha, M. Kang, J. H. Jin, S. G. Im, M. J. Kim, and S. Y. Choi, Ultrathin all-solid-state MoS₂-based electrolyte gated synaptic transistor with tunable organic−inorganic hybrid film, Adv. Sci. (Weinh.) 11(23), 2308847 (2024)

[18] S. Sun, M. Zhang, J. Bian, T. Xu, and J. Su, In₂O₃/ZnO heterojunction thin film transistor for high recognition accuracy neuromorphic computing and optoelectronic artificial synapses, Nanotechnology 35(36), 365602 (2024)

[19] L. Zheng, Z. Liu, S. Xin, Q. Chen, J. Ming, L. Wu, J. Xu, P. Xu, K. Liu, R. Seeram, and F. Wang, Flexible electrolyte-gated transistor based on InZnSnO nanowires for self-adaptive applications, Appl. Mater. Today 41, 102424 (2024)

[20] S. K. Kim, S. W. Choi, M. Kim, K. R. Yun, G. Wang, and T. Y. Seong, Dual‐electrolyte neuromorphic transistor for risk detection and image processing, Adv. Mater. Technol. 10(7), 2401617 (2024)

[21] Y. Du, W. Xiao, G. Miao, Z. Yin, R. Ci, G. Liu, and F. Shan, Solution-processed LiGdO solid electrolyte for an In₂O₃ synaptic transistor and its application in neuromorphic computing, J. Mater. Chem. C 13(3), 1318 (2025)

[22] W. Zeng, X. Luo, J. Xu, M. Zhang, S. Liu, Q. Zhang, and G. Zhu, Ferroelectric/electric-double-layer-modulated synaptic thin film transistors toward an artificial tactile perception system, ACS Appl. Mater. Interfaces 17(3), 5086 (2025)

[23] M. Zahid, M. I. Sadiq, C. Jin, J. Wang, X. Shi, W. Liu, F. Aslam, Y. Xu, M. Tahir, J. Yang, and J. Sun, Nitrogen-doped carbon quantum dot-decorated In₂O₃ synaptic transistors for neuromorphic computing, Appl. Phys. Lett. 125(19), 193503 (2024)

[24] H. Yu, Y. Xu, Z. Deng, C. Jin, W. Liu, X. Shi, J. Liu, J. Sun, and J. Yang, Self-supported flexible organic electrochemical synaptic transistors on self-healing composite electrolyte membranes, Appl. Phys. Lett. 124(17), 173703 (2024)

[25] Y. Ding, T. Li, B. Yan, G. Liu, and F. Shan, Hydroxyl-induced stability and mobility enhancement for field effect transistor based on In₂O₃ nanofiber, Appl. Phys. Lett. 121(26), 263301 (2022)

[26] C. F. Hu, J. Y. Feng, J. Zhou, and X. P. Qu, Investigation of oxygen and argon plasma treatment on Mg-doped InZnO thin film transistors, Appl. Phys. A 122(11), 941 (2016)

[27] J. Jadwiszczak, G. Li, C. P. Cullen, J. J. Wang, P. Maguire, G. S. Duesberg, J. G. Lunney, and H. Zhang, Photoresponsivity enhancement in monolayer MoS₂ by rapid O₂: Ar plasma treatment, Appl. Phys. Lett. 114(9), 091103 (2019)

[28] D. K. Hwang, M. Misra, Y. E. Lee, S. D. Baek, J. M. Myoung, and T. I. Lee, The role of Ar plasma treatment in generating oxygen vacancies in indium tin oxide thin films prepared by the sol-gel process, Appl. Surf. Sci. 405, 344 (2017)

[29] J. Park, J. E. Huh, S. E. Lee, J. Lee, W. H. Lee, K. H. Lim, and Y. S. Kim, Effective atmospheric-pressure plasma treatment toward high-performance solution-processed oxide thin-film transistors, ACS Appl. Mater. Interfaces 10(36), 30581 (2018)

[30] A. Charnas, M. Si, Z. Lin, and P. D. Ye, Enhancement-mode atomic-layer thin In₂O₃ transistors with maximum current exceeding 2 A/mm at drain voltage of 0.7 V enabled by oxygen plasma treatment, Appl. Phys. Lett. 118(5), 052107 (2021)

[31] J. T. Ye, S. Inoue, K. Kobayashi, Y. Kasahara, H. T. Yuan, H. Shimotani, and Y. Iwasa, Liquid-gated interface superconductivity on an atomically flat film, Nat. Mater. 9(2), 125 (2010)

[32] Y. Yang, H. Hua, Z. Lv, M. Zhang, C. Liu, Z. Wen, H. Xie, W. He, J. Zhao, and C. C. Li, Reconstruction of electric double layer for long‐life aqueous zinc metal batteries, Adv. Funct. Mater. 33(10), 2212446 (2023)

[33] X. Liu, J. Miao, Q. Fan, W. Zhang, X. Zuo, M. Tian, S. Zhu, X. Zhang, and L. Qu, Recent progress on smart fiber and textile based wearable strain sensors: Materials, fabrications and applications, Adv. Fiber Mater. 4(3), 361 (2022)

[34] G. Zhang, H. Yu, Y. Shi, W. Liu, X. Shi, C. Jin, Y. Xu, Z. Li, S. Huang, J. Sun, and J. Yang, Flexible and low-voltage ITO synaptic transistors for biotic tactile sensing, Appl. Phys. Lett. 123(2), 023702 (2023)

[35] R. Fang, X. Li, K. Ren, W. Zhang, H. Xu, L. Wang, and D. Shang, Improved dynamic characteristics of oxide electrolyte-gated transistor for time-delayed reservoir computing, Appl. Phys. Lett. 124(5), 053505 (2024)

[36] Y. Li, Y. J. Huang, X. L. Chen, W. S. Wang, X. Huang, H. Xiao, and L. Q. Zhu, Multi-terminal pectin/chitosan hybrid electrolyte gated oxide neuromorphic transistor with multi-mode cognitive activities, Front. Phys. (Beijing) 19(5), 53204 (2024)

[37] H. Cong, Y. Chang, R. Zhou, W. Zhang, G. Sun, P. Xu, Y. Qin, S. Ramakrishna, X. Liu, and F. Wang, Rational tuning of the cation ratio in metal oxide semiconductor nanofibers for low-power neuromorphic transistors, Sci. China Mater. 66(8), 3251 (2023)

[38] S. Xin, Y. Chang, R. Zhou, H. Cong, L. Zheng, Y. Wang, Y. Qin, P. Xu, X. Liu, and F. Wang, Ultraviolet-driven metal oxide semiconductor synapses with improved long-term potentiation, J. Mater. Chem. C 11(2), 722 (2023)

[39] J. T. Yang, C. Ge, J. Y. Du, H. Y. Huang, M. He, C. Wang, H. B. Lu, G. Z. Yang, and K. J. Jin, Artificial synapses emulated by an electrolyte-gated tungsten-oxide transistor, Adv. Mater. 30(34), 1801548 (2018)

[40] Y. J. Huang, L. F. Wu, J. K. Di, X. Huang, W. S. Wang, S. Y. Zhou, B. C. Gong, and L. Q. Zhu, Solidum alginate gated oxide dendritic transistor for spatiotemporal arithmetic application, Front. Phys. (Beijing) 20(3), 34205 (2025)

[41] Z. Fan, H. Fan, C. He, and X. B. Yan, Implementation of an artificial neuron circuit model based on high speed AlN stacked threshold switching devices, Front. Phys. (Beijing) 20(3), 34201 (2025)

RIGHTS & PERMISSIONS

Higher Education Press