Next Article in Journal
Water Kefir Grains—Microbial Biomass Source for Carbonaceous Materials Used as Sulfur-Host Cathode in Li-S Batteries
Previous Article in Journal
The Effect of Pyrolysis Temperature and the Source Biomass on the Properties of Biochar Produced for the Agronomical Applications as the Soil Conditioner
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Materials
Volume 15
Issue 24
10.3390/ma15248858
materials-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Forming-Free Tunable Analog Switching in WOx/TaOx Heterojunction for Emulating Electronic Synapses

by
Chandreswar Mahata
1,
Juyeong Pyo
1,
Beomki Jeon
1,
Muhammad Ismail
1,
Myounggon Kang
2,* and
Sungjun Kim
1,*
1
Division of Electronics and Electrical Engineering, Dongguk University, Seoul 04620, Republic of Korea
2
Department of Electronics Engineering, Korea National University of Transportation, Chungju-si 27469, Republic of Korea
*
Authors to whom correspondence should be addressed.
Materials 2022, 15(24), 8858; https://doi.org/10.3390/ma15248858
Submission received: 22 November 2022 / Revised: 6 December 2022 / Accepted: 9 December 2022 / Published: 12 December 2022
(This article belongs to the Section Electronic Materials)
Browse Figures
Versions Notes

Abstract

:
In this work, the sputtered deposited WOx/TaOx switching layer has been studied for resistive random-access memory (RRAM) devices. Gradual SET and RESET behaviors with reliable device-to-device variability were obtained with DC voltage sweep cycling without an electroforming process. The memristor shows uniform switching characteristics, low switching voltages, and a high RON/ROFF ratio (~102). The transition from short-term plasticity (STP) to long-term potentiation (LTP) can be observed by increasing the pulse amplitude and number. Spike-rate-dependent plasticity (SRDP) and paired-pulse facilitation (PPF) learning processes were successfully emulated by sequential pulse trains. By reducing the pulse interval, the synaptic weight change increases due to the residual oxygen vacancy near the conductive filaments (CFs). This work explores mimicking the biological synaptic behavior and further development for next-generation neuromorphic applications.

1. Introduction

To improve beyond the conventional memory storage device for high-density applications, resistive random-access memory (RRAM) consisting of memory cells with a simple metal-insulator-metal structure has attracted considerable attention [1,2,3]. Several researchers have extensively studied the resistive switching effects of next-generation non-volatile memory applications due to their low power consumption, high operating speed, and compatible fabrication steps [2,4,5]. RRAM structure consists of a switching layer sandwich between two electrodes, and a transition from a high-resistive state (HRS) to a low-resistive state (LRS) due to the oxygen vacancy (VO), and the migrations and recombination of oxygen ions (O−) [6]. The rupture and formation of the conductive filaments (CFs) that form during electroforming are controlled mainly by the oxygen vacancy creation in the switching materials [7,8]. Gradual resistance changes in the memristor with a high ON/OFF ratio can be controlled under DC and pulse conditions for the applications of analog neuromorphic devices [9,10,11]. Therefore, the motivation for implementing the synaptic function in resistive switching devices is to mimic artificial synapses by a graded oxygen content switching layer. Compared to other transition metal oxides such as Al2O3, TiO2, HfO2, and ZrO2, the oxygen vacancy formation is well controlled in WOx and TaOx [12,13,14,15,16,17,18,19].
Kim et al. have described uniform switching ability in the Pt/WOx/W memristor with multi-bit memory applications by the movement of controlled oxygen vacancies due to the intrinsic property of WOx [20]. Moreover, it was earlier reported that the rectification of the resistive switching performance was achieved in Pt/WOx/Au RRAM due to the work function asymmetry. Electric-field-induced oxygen vacancy electromigration was found to be responsible for SET and RESET behavior [21].
It has been reported that TaOx can generate oxygen vacancies easily, which is beneficial for forming strong conductive filaments for non-volatile memory applications [16,22]. However, an additional oxygen exchange layer can help gradual switching due to the redox reaction under an external electric field [17]. Modified graded oxygen content in the TaOx switching layer has demonstrated that the conical shape of CFs can control the formation and rupture in an effective way for high-performance RRAM devices [18]. Ma et al. predicted that in TiN/TaO2.0/TiN RRAM structures, the electroformation of CFs is less dominated by the oxygen ions reacted on the electrode [19]. Sugawara et al. have described a TaOx bilayer memristor with different concentrations of oxygen, where the movement of oxygen ions define the resistance transition, which leads to analog resistive switching [6]. Park et al. also described a TaO2−x/Ta2O5−x-based resistive switching device where the SET process is controlled by the VO migration from the bottom TaO2−x layer [8].
Bilayer Pt/HfO2/TaOx/Pt memristor devices show hysteresis behavior, and multiple resistance states are achieved by controlling the conductive filaments at the HfO2/TaOx interface [2]. In the Ta/WOx/TaOx/Pt bilayer, WOx was used as a redox layer, and the WOx/TaOx interface controls the movement of oxygen vacancies in the TaOx layer that results in gradual resistive switching [17]. Overgrowth CFs were also reduced due to the non-existence of redox reactions (because of the presence of the Al2O3 layer in the W/Al2O3/TaOx/TiN memory structure).
In this study, the gradual conductance change was controlled in the sputtered deposited WOx (50 nm)/TaOx (5 nm) bilayer thin film. Therefore, a detailed understanding of the memory performance is associated with the multilevel resistance and modulation of conductance under DC and pulse bias conditions. Critical synaptic functions were mimicked with stable multistate memory characterizations. The WOx redox reactive layer was inserted between the ITO electrode and TaOx switching layer. Thus, the biological synaptic function has been demonstrated, including the spike-rate-dependent plasticity (SRDP), short-term plasticity (STP), long-term plasticity (LTP), and paired-pulse facilitation (PPF). The improved synaptic properties of the ITO/WOx/TaOx/TiN memristor device makes it a promising candidate for neuromorphic computing.

2. Materials and Methods

The memristors were fabricated with titanium nitride (TiN) as the bottom electrode, deposited by DC sputter using a Ti (99.99%) target. The working pressure was kept at 1 mTorr (Ar, 19 sccm, and N2, 1 sccm) with 350 W power. First, an insulating layer of ~5 nm TaOx was formed using DC sputter at 5 mTorr working pressure under a gas mixture of Ar, 20 sccm, and O2, 6 sccm. The second layer of WOx (~50 nm) was deposited on TaOx by RF sputter (at 80 W) from a W target using a gas mixture of Ar, 20 sccm, and O2, 6 sccm. Using photolithography, square patterns of 100 µm × 100 µm were made. Finally, indium tin oxide (ITO) top electrodes were achieved by the lift-off process. Using RF sputter at 60 W, ITO was deposited from a commercial ITO target (99.99% purity), under the gas pressure of 3 mTorr, using Ar. The electrical characteristics were obtained by connecting the bottom electrode to the ground, and external electric fields were applied to the top electrode. The Keithley 4200 SCS semiconductor parameter analyzer and 4225-PMU pulse module were used for DC I-V cycling for multistate memory characterization and to observe synaptic behavior of the ITO/WOx/TaOx/TiN memristor devices.

3. Results and Discussion

3.1. Gradual Bipolar Resistive Switching of the ITO/WOx/TaOx/TiN Device

Under a positive bias voltage with the current compliance (ICC) of 10 µA, the SET process of the ITO/WOx/TaOx/TiN memristor was observed, as shown in Figure 1a. In step-I, the current increases gradually and reaches the ICC. The device is then in a low-resistance state (LRS) in step-II under the sweepback voltage. The device shows forming-free bipolar switching, which is typical intrinsic TaOx behavior for the presence of abundant oxygen vacancies [23]. Generating a large amount of VOs during applications of external electric fields helps the forming-free SET process, which is advantageous for RRAM applications [24]. As electroforming is a destructive process due to chemical reactions, it can change the switching layer’s intrinsic properties and uniformity for device-to-device performance [25]. At negative bias, the RESET process occurs, as shown in step-III and step-IV. The overall gradual transition of the SET and RESET operation in the device predicts the gradual change of conductance, which is promising for multilevel memory storage. Forming-free resistive switching may be attributed to uniformly distributed oxygen vacancies that accumulate and form the CFs under external positive bias [26]. The switching mechanism of the SET and RESET process is described in Figure 1b. In this case, the top WOx layer is believed to act as an oxygen ion reservoir, which further helps the redox reaction at the WOx/TaOx interface. The CFs are preferably ruptured near that point, and the gradual switching may be due to trap-assisted-tunneling (TAT) conduction [27]. Therefore, at positive bias, the conduction path is created in the TaOx layer, which consists of oxygen vacancies [2]. In the RESET process, the recombination of oxygen ions that have drifted from the WOx and TaOx layers leads to the rupture of CFs at the WOx/TaOx interface [28]. Thus, unlike other transition metal oxides such as Al2O3, in which redox reactions are absent, WOx benefits the gradual SET process [29]. Forming-free resistive switching in the amorphous TaOx/WOx layer is described by Prakash et al., where the switching mechanism is dominated by the CFs formation and rupture [30]. Furthermore, Wang et al. have described a large number of oxygen vacancies contained in sputtered TaOx [31]. The low resistive TaOx layer shows the bipolar resistive switching characteristics without any forming process. Cyclic DC endurance characteristics up to 100 cycles show a stable memory window, as shown in Figure 1c. Endurance properties for 10 ITO/WOx/TaOx/TiN memristor devices are shown in Figure 1d. A device-to-device reliability with a RON/ROFF ratio >10 predicts that the memristor is suitable for high-density and gradual resistive switch synaptic devices. Figure 1e shows the I-V characteristics of ITO/WOx/TiN devices. As mentioned above, the leakage current was found to be high due to a significant concentration of intrinsic oxygen vacancies in the WOx layer. The ITO/WOx/TiN device was set at a very high current compliance of 10 mA. However, due to the formation of very thick CFs, the RESET process cannot be achieved. Multiple ITO/WOx/TiN devices were measured to confirm the phenomenon, as shown in Figure 1e. A histogram of LRS and HRS distribution from the resistive switching characteristics of the ITO/WOx/TaOx/TiN memristor has been plotted in Figure 1e. The average LRS and HRS at the read voltage of −0.2 V are 1.05 MΩ and 110 MΩ, respectively. The well-uniform distribution of 30 memristors shows the device-to-device reliability of the present device.
A desirable high-density memory for synaptic devices is achieved by controlling the ICC value during SET operation and the RESET stop voltage, which can modulate the CFs’ size and diameter [32]. Multiple resistance states were obtained by modulating ICC from 1 × 10−7 A to 7 × 10−5 A during the SET process in the ITO/WOx/TaOx/TiN memristor, as shown in Figure 2a. The forming and rupture of oxygen vacancies near the CFs at the WOx/TaOx interface is controlled by the O− ions’ migrations through redox reactions, as described by Prakash et al. [33]. The low resistance state (LRS) varied from 17.2 MΩ to 0.15 MΩ, as shown in Figure 2b. The gradual increase of the CFs’ diameter caused by increasing ICC was near the WOx/TaOx interface, initiating multiple LRSs. Therefore, the ITO/WOx/TaOx/TiN memristor is suitable for the application of multistate memory devices.
A gradual RESET process was observed when the memristor was subjected to an increasing negative voltage from −0.8 V to −2.0 V at the SET ICC value of 10 µA, as shown in Figure 2c. The ratio of HRS to LRS (ION/IOFF) induced by increasing VRESET was increased from 2.2 to 73.8. The gradual recombination of oxygen ions near the CFs controls the thinning of CFs inside the TaOx layer [34]. Linearly increasing the VRESET, a uniform change in the HRS, indicates a stable distribution of oxygen vacancies inside the switching layer, which is essential for gradual resistive switching in synaptic devices. The endurance characteristics of each HRS were confirmed by 25 switching cycles read at −0.2 V. The migration of oxygen ions from the WOx layer helps gradual recombination at the WOx/TaOx interface.

3.2. Synaptic Characteristics under an Identical Pulse Sequence

Long-term plasticity (LTP) and short-term plasticity (STP) are the critical features of synaptic function [35]. The spike-amplitude-dependent plasticity (SADP) behavior shown in Figure 3a demonstrates that the pulse amplitude highly influences synaptic weight change. Similar characteristics were reported by Hao et al. [36]. Figure 3b–d depict the transition from STP to LTP, depending on the pulse amplitude and spike numbers. As shown in Figure 3b, by applying a pulse amplitude of 1.25 V/100 µs, a slow transition from STP to LTP can be obtained. Whereas, at higher pulse amplitude, a sharp transition occurs, as shown in Figure 3c,d. The EPSC obtained by higher voltage stimuli showed a permanent potentiation behavior with a saturated value up to the applied ICC. The currents at the base voltage (+0.2 V) also confirm the phenomenon. This experiment demonstrated that the STP–LTP transition occurred due to repeated pulse stimulations [37]. The EPSC obtained under the repeated pulse sequence and at different amplitudes confirms that the transition occurred after sufficiently thicker CFs were formed. STP behavior is due to the rupture of weak conductive filaments near the WOx/TaOx interface. A small number of pre-synaptic pulses produce unstable thin filaments that dissolve quickly after the withdrawal of pulse trains, and the EPSC recovers its original value at the base (read) voltage, as shown in Figure 3b–d. On the other hand, a significant number of oxygen vacancies are created (strong CFs are formed) after increasing the number of stimuli, leading to LTP behavior [38]. The overall EPSC increment of the memristor at LTP is due to the repeated rehearsal of the stimuli, analogous to the biological memory functions. This transition behavior was reported by Chang et al. previously, and is crucial for neuromorphic systems applications [39].

3.3. Pulse Frequency-Dependent Synaptic Properties

In neuroscience, tunable synaptic weight change, which depends on the pulse sequence frequency as the spike-rate-dependent plasticity (SRDP), is a necessary learning process [40]. The modulation of EPSC in the ITO/WOx/TaOx/TiN memristor was monitored by varying the pulse interval, as shown in Figure 4a. The current response was monitored under pulse train frequencies from 2 Hz to 100 Hz with a fixed amplitude and width of +1.5 V/200 µs. Each frequency response was measured after resetting the memristor in DC mode to bring the device into the initial HRS. Under all pulse frequencies, the current increment indicates that the frequencies are less than the VO and O− recombination time [41]. The higher current was achieved with a continuous application of a pulse train with shorter intervals (higher pulse frequency). The increased current response at the higher spiking rate is due to the residual oxygen vacancies near the CFs created by the previous spike influence [40,42]. The short-term plasticity at different frequencies for the ITO/WOx/TaOx/TiN memristor can act as dynamic filters [42,43,44]. The creation of high-density oxygen vacancies at the WOx/TaOx interface with a higher pulse frequency act as the high-pass filtering characteristic. This is similar to parallel fiber synapses in a biological system, as the pre-synaptic pulse train can control the post-synaptic output characteristics [42,45]. As shown in Figure 4a, the memristor can act as a high-pass dynamic filter because a higher EPSC gain is obtained under shorter pulse intervals. The EPSC gain at each frequency was calculated and plotted in Figure 4b. It was calculated from the ratio of A20/A1, where A20 and A1 are the peak amplitude of currents after the 20th and 1st pulse, respectively. The EPSC gain of 1.24 at 2 Hz was increased to 1.86 at 100 Hz, indicating that the ITO/WOx/TaOx/TiN memristor exhibits frequency-dependent synaptic weight modulation. Experience-dependent short-term plasticity was observed, where the post-synaptic current increases and decreases depending on the immediate applications of higher- and lower-frequency pulse trains. The experimental results show that the change in synaptic weight depends on the interval of the spikes. Similar to biological memory, the post-synaptic currents increase at the higher frequency of spike trains. Whereas, at the immediate application of a lower input spike frequency, a synaptic depression was observed without the application of an opposite-polarity pulse sequence [46]. As shown in Figure 4c, a potentiation behavior was observed at a 20 Hz pulse train frequency. Afterward, at a 2 Hz pulse frequency, the currents decrease (depression behavior) at the same pulse amplitude at +1.5 V/200 µs. Similarly, depression behavior was observed at a 2 Hz pulse frequency immediately after a 5 Hz pulse frequency. Unlike the result presented in Figure 4a, the EPSC increases at a 2 Hz pulse frequency due to the application of an individual set of pulse trains after bringing the memristor to its RESET (HRS) state. Therefore, this phenomenon mimics the bio-realistic function where the synaptic weight changes depend on the previous pulse stimulation frequency. The migrated oxygen ions from the WOx layer in the structure also trigger the recombination of oxygen vacancies and oxygen ions. Thus, the ITO/WOx/TaOx/TiN memristor can mimic the history-dependent plasticity where synaptic weight change depends on the previous experience. This behavior is similar to the paired-pulse depression phenomenon and was also observed in earlier studies [17,36,47].

3.4. Paired-Pulse Facilitation (PPF) Emulation

In the short-term plasticity phenomenon, PPF, two consecutive pulse stimuli are applied, and the EPSCs are monitored due to the change in post-synaptic weight. The application of a pre-synaptic paired pulse increased the EPSC amplitude, which was triggered by the second pulse that was higher than the previous one [41,47,48]. As shown in Figure 5a, consecutive paired pulses were applied to the ITO/WOx/TaOx/TiN memristor with different pulse intervals (Δt) ranging from 0.1 ms to 1 ms with a pulse amplitude of +1.2 V/100 µs. It was found that with increasing the pulse interval, EPSC responses decrease gradually. At the first pulse, the oxygen vacancies are accumulated near the CFs. However, due to the shorter spike interval of 0.1 ms, the oxygen vacancies do not have enough time to recombine with oxygen ions immediately after removing the first pulse. Therefore, the remaining oxygen vacancies that accumulated near the CFs and produced a higher EPSC appeared at the second applied pulse, which induced the PPF. Although at a higher spike interval time, the reserved oxygen vacancies have a longer time to recombine with O−, resulting in a reduced EPSC amplitude, as shown in Figure 5a. Thus, at the larger time interval, the enhancement of the post-synaptic current at the second pulse is likely to fade away, which is similar to the biological synapse. The PPF index was calculated from (A2/A1) × 100, which reached 112.6% at a Δt of 0.1 ms, and decays gradually with the increasing pulse interval, as shown in Figure 5b. The experimental data have been fitted with the decay factor to determine the relaxation time constant using Equation (1) [41]:
PPF = C1 exp(−Δt/τ1) + C1 exp(−Δt/τ2),
where τ1 and τ2 are the characteristic relaxation times, Δt is the time interval between two successive pre-synaptic spikes, and C1 and C1 are the initial facilitation magnitudes. The relaxation time constants τ1 and τ2 estimated from the simulations were 48 µs and 255 µs, respectively. Therefore, the synaptic weight change due to the short-term plasticity was successfully emulated in the bilayer of the WOx/TaOx artificial synapse, confirming potential applications in neuromorphic computing. In Table 1, a detailed comparison of DC electrical parameters is presented, showing that the ITO/WOx/TaOx/TiN memristor has improved gradual SET and RESET characteristics with reliable device-to-device variability and memory window. Also, the sputtered deposited WOx/TaOx switching layer has promising synaptic properties suitable for neuromorphic applications.

4. Conclusions

In conclusion, fabricated memristors demonstrated gradual SET/RESET characteristics due to the presence of the WOx oxygen ion reservoir, which promotes the redox reaction at the WOx/TaOx interface. The multilevel memory characteristics were observed by varying the SET current compliance and VRESET with a switching ratio >100 under multiple cycling processes. Synaptic behavior was obtained with potentiation characteristics under different pulse amplitudes. Moreover, oxygen migration and redox reactions under pulse application control reliable SRDP and SADP. Therefore, the bilayer-WOx/TaOx--based memristor is promising for the use in high-density memory integration in low-power synaptic devices for neuromorphic computing.

Author Contributions

Conceptualization, experiments and data analysis, writing, C.M.; experiments, J.P., B.J. and M.I.; review and editing, M.K. and S.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2018R1A6A1A03023788), and the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korea government (MOTIE) under Grant 20224000000020.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Wang, T.; Brivio, S.; Cianci, E.; Wiemer, C.; Perego, M.; Spiga, S.; Lanza, M. Improving HfO2-Based Resistive Switching Devices by Inserting a TaOxThin Film via Engineered In Situ Oxidation. ACS Appl. Mater. Interfaces 2022, 14, 24565–24574. [Google Scholar] [CrossRef] [PubMed]
  2. Sung, J.H.; Park, J.H.; Jeon, D.S.; Kim, D.; Yu, M.J.; Khot, A.C.; Dongale, T.D.; Kim, T.G. Retention enhancement through capacitance-dependent voltage division analysis in 3D stackable TaOx/HfO2-based selectorless memristor. Mater. Des. 2021, 207, 109845. [Google Scholar] [CrossRef]
  3. Luo, Q.; Xu, X.; Liu, H.; Lv, H.; Gong, T.; Long, S.; Liu, Q.; Sun, H.; Banerjee, W.; Li, L.; et al. Demonstration of 3D vertical RRAM with ultra low-leakage, high-selectivity and self-compliance memory cells. In Proceedings of the IEEE International Electron Devices Meeting (IEDM), Washington, DC, USA, 7–9 December 2015; pp. 10.2.1–10.2.4. [Google Scholar]
  4. Ielmini, D. Resistive switching memories based on metal oxides: Mechanisms, reliability and scaling. Semicond. Sci. Technol. 2016, 31, 063002. [Google Scholar] [CrossRef]
  5. Yu, S.; Gao, B.; Fang, Z.; Yu, H.; Kang, J.; Wong, H.S.P. A low energy oxide-based electronic synaptic device for neuromorphic visual systems with tolerance to device variation. Adv. Mater. 2013, 25, 1774–1779. [Google Scholar] [CrossRef] [PubMed]
  6. Sugawara, K.; Shima, H.; Takahashi, M.; Naitoh, Y.; Suga, H.; Akinaga, H. Low-Frequency-Noise Spectroscopy of TaOx-based Resistive Switching Memory. Adv. Electron. Mater. 2022, 8, 2100758. [Google Scholar] [CrossRef]
  7. Celano, U.; Goux, L.; Degraeve, R.; Fantini, A.; Richard, O.; Bender, H.; Jurczak, M.; Vandervorst, W. Imaging the Three-Dimensional Conductive Channel in Filamentary-Based Oxide Resistive Switching Memory. Nano Lett. 2015, 15, 7970–7975. [Google Scholar] [CrossRef] [PubMed]
  8. Park, G.-S.; Kim, Y.B.; Park, S.Y.; Li, X.S.; Heo, S.; Lee, M.-J.; Chang, M.; Kwon, J.H.; Kim, M.; Chung, U.-I.; et al. In situ observation of filamentary conducting channels in an asymmetric Ta2O5−x/TaO2−x bilayer structure. Nat. Commun. 2013, 4, 2382. [Google Scholar] [CrossRef] [Green Version]
  9. Zhang, Z.; Wang, Z.; Shi, T.; Bi, C.; Rao, F.; Cai, Y.; Liu, Q.; Wu, H.; Zhou, P. Memory materials and devices: From concept to application. InfoMat 2020, 2, 261–290. [Google Scholar] [CrossRef] [Green Version]
  10. Majumdar, S.; Tan, H.; Qin, Q.H.; van Dijken, S. Energy-Efficient Organic Ferroelectric Tunnel Junction Memristors for Neuromorphic Computing. Adv. Electron. Mater. 2019, 5, 1800795. [Google Scholar] [CrossRef] [Green Version]
  11. Lim, S.; Kwak, M.; Hwang, H. One transistor-two resistive RAM device for realizing bidirectional and analog neuromorphic synapse devices. Nanotechnology 2019, 30, 455201. [Google Scholar] [CrossRef]
  12. Wu, Y.; Yu, S.; Lee, B.; Wong, P. Low-power TiN/Al2O3/Pt resistive switching device with sub-20 A switching current and gradual resistance modulation. J. Appl. Phys. 2011, 110, 3–8. [Google Scholar] [CrossRef]
  13. Huang, R.; Yan, X.; Ye, S.; Kashtiban, R.; Beanland, R.; Morgan, K.A.; Charlton, M.D.B.; de Groot, C.H. Compliance-free ZrO2/ZrO2−x/ZrO2 resistive memory with controllable interfacial multistate switching behaviour. Nanoscale Res. Lett. 2017, 12, 384. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  14. Yu, S.; Guan, X.; Wong, H.S.P. Conduction mechanism of TiNHfO xPt resistive switching memory: A trap-assisted-tunneling model. Appl. Phys. Lett. 2011, 99, 16–19. [Google Scholar] [CrossRef]
  15. Bousoulas, P.; Karageorgiou, I.; Aslanidis, V.; Giannakopoulos, K.; Tsoukalas, D. Tuning Resistive, Capacitive, and Synaptic Properties of Forming Free TiO2-x-Based RRAM Devices by Embedded Pt and Ta Nanocrystals. Phys. Status Solidi Appl. Mater. Sci. 2018, 215, 1700440. [Google Scholar] [CrossRef]
  16. Chen, Y.C.; Chung, Y.L.; Chen, B.T.; Chen, W.C.; Chen, J.S. Revelation on the interrelated mechanism of polarity-dependent and multilevel resistive switching in TaOx-based memory devices. J. Phys. Chem. C 2013, 117, 5758–5764. [Google Scholar] [CrossRef]
  17. Chen, W.J.; Cheng, C.H.; Lin, P.E.; Tseng, Y.T.; Chang, T.C.; Chen, J.S. Analog Resistive Switching and Synaptic Functions in WOx/TaOxBilayer through Redox-Induced Trap-Controlled Conduction. ACS Appl. Electron. Mater. 2019, 1, 2422–2430. [Google Scholar] [CrossRef]
  18. Wang, B.; Xue, K.H.; Sun, H.J.; Li, Z.N.; Wu, W.; Yan, P.; Liu, N.; Tian, B.Y.; Liu, X.X.; Miao, X.S. Performance enhancement of TaO x resistive switching memory using graded oxygen content. Appl. Phys. Lett. 2018, 113, 183501. [Google Scholar] [CrossRef]
  19. Ma, Y.; Cullen, D.A.; Goodwill, J.M.; Xu, Q.; More, K.L.; Skowronski, M. Exchange of Ions across the TiN/TaO xInterface during Electroformation of TaO x-Based Resistive Switching Devices. ACS Appl. Mater. Interfaces 2020, 12, 27378–27385. [Google Scholar] [CrossRef]
  20. Kim, S.; Biju, K.P.; Jo, M.; Jung, S.; Park, J.; Lee, J.; Lee, W.; Shin, J.; Park, S.; Hwang, H. Effect of scaling WOx-Based RRAMs on their resistive switching characteristics. IEEE Electron. Device Lett. 2011, 32, 671–673. [Google Scholar] [CrossRef]
  21. Won, S.; Lee, S.Y.; Park, J.; Seo, H. Forming-less and Non-Volatile Resistive Switching in by Oxygen Vacancy Control at Interfaces. Sci. Rep. 2017, 7, 10186. [Google Scholar] [CrossRef]
  22. Luo, Q.; Xu, X.; Gong, T.; Lv, H.; Dong, D.; Ma, H.; Yuan, P.; Gao, J.; Liu, J.; Yu, Z.; et al. 8-Layers 3D vertical RRAM with excellent scalability towards storage class memory applications. In Proceedings of the IEEE International Electron Devices Meeting (IEDM), San Francisco, CA, USA, 2–6 December 2017; pp. 2.7.1–2.7.4. [Google Scholar]
  23. Gao, S.; Zeng, F.; Li, F.; Wang, M.; Mao, H.; Wang, G.; Song, C.; Pan, F. Forming-free and self-rectifying resistive switching of the simple Pt/TaOx/n-Si structure for access device-free high-density memory application. Nanoscale 2015, 7, 6031–6038. [Google Scholar] [CrossRef] [PubMed]
  24. Ricci, S.; Mannocci, P.; Farronato, M.; Hashemkhani, S.; Ielmini, D. Forming-Free Resistive Switching Memory Crosspoint Arrays for In-Memory Machine Learning. Adv. Intell. Syst. 2022, 4, 2200053. [Google Scholar] [CrossRef]
  25. Shuai, Y.; Ou, X.; Luo, W.; Mücklich, A.; Bürger, D.; Zhou, S.; Wu, C.; Chen, Y.; Zhang, W.; Helm, M.; et al. Key concepts behind forming-free resistive switching incorporated with rectifying transport properties. Sci. Rep. 2013, 3, 2208. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  26. Baek, G.H.; Lee, A.R.; Kim, T.Y.; Im, H.S.; Hong, J.P. Oxide stoichiometry-controlled TaOx-based resistive switching behaviors. Appl. Phys. Lett. 2016, 109, 143502. [Google Scholar] [CrossRef]
  27. Sakellaropoulos, D.; Bousoulas, P.; Nikas, G.; Arvanitis, C.; Bagakis, E.; Tsoukalas, D. Enhancing the synaptic properties of low-power and forming-free HfOx/TaOy/HfOx resistive switching devices. Microelectron. Eng. 2020, 229, 111358. [Google Scholar] [CrossRef]
  28. Zhou, Q.; Zhai, J. The improved resistive switching properties of TaOx-based RRAM devices by using WNx as bottom electrode. Phys. B Condens. Matter 2013, 410, 85–89. [Google Scholar] [CrossRef]
  29. Chakrabarti, S.; Panja, R.; Roy, S.; Roy, A.; Samanta, S.; Dutta, M.; Ginnaram, S.; Maikap, S.; Cheng, H.M.; Tsai, L.N.; et al. Evolution of resistive switching mechanism through H2O2 sensing by using TaO x-based material in W/Al2O3/TaOx/TiN structure. Appl. Surf. Sci. 2018, 433, 51–59. [Google Scholar] [CrossRef]
  30. Prakash, A.; Jana, D.; Samanta, S.; Maikap, S. Self-compliance-improved resistive switching using Ir/TaOx/W cross-point memory. Nanoscale Res. Lett. 2013, 8, 527. [Google Scholar] [CrossRef] [Green Version]
  31. Wang, Z.; Yin, M.; Zhang, T.; Cai, Y.; Wang, Y.; Yang, Y.; Huang, R. Engineering incremental resistive switching in TaO: X based memristors for brain-inspired computing. Nanoscale 2016, 8, 14015–14022. [Google Scholar] [CrossRef]
  32. Li, Z.; Tian, B.; Xue, K.H.; Wang, B.; Xu, M.; Lu, H.; Sun, H.; Miao, X. Coexistence of digital and analog resistive switching with low operation voltage in oxygen-gradient HfOx Memristors. IEEE Electron. Device Lett. 2019, 40, 1068–1071. [Google Scholar] [CrossRef]
  33. Prakash, A.; Park, J.; Song, J.; Woo, J.; Cha, E.-J.; Hwang, H. Demonstration of Low Power 3-bit Multilevel Cell Characteristics in a TaOx-Based RRAM by Stack Engineering. IEEE Electron. Device Lett. 2015, 36, 32–34. [Google Scholar] [CrossRef]
  34. Mehonic, A.; Buckwell, M.; Montesi, L.; Garnett, L.; Hudziak, S.; Fearn, S.; Chater, R.; McPhail, D.; Kenyon, A.J. Structural changes and conductance thresholds in metal-free intrinsic SiO x resistive random access memory. J. Appl. Phys. 2015, 117, 124505. [Google Scholar] [CrossRef] [Green Version]
  35. Zhou, L.; Yang, S.; Ding, G.; Yang, J.Q.; Ren, Y.; Zhang, S.R.; Mao, J.Y.; Yang, Y.; Zhou, Y.; Han, S.T. Tunable synaptic behavior realized in C 3 N composite based memristor. Nano Energy 2019, 58, 293–303. [Google Scholar] [CrossRef]
  36. Hao, S.; Ji, X.; Liu, F.; Zhong, S.; Pang, K.Y.; Lim, K.G.; Chong, T.C.; Zhao, R. Monolayer MoS2/WO3Heterostructures with Sulfur Anion Reservoirs as Electronic Synapses for Neuromorphic Computing. ACS Appl. Nano Mater. 2021, 4, 1766–1775. [Google Scholar] [CrossRef]
  37. Kim, M.K.; Lee, J.S. Short-Term Plasticity and Long-Term Potentiation in Artificial Biosynapses with Diffusive Dynamics. ACS Nano 2018, 12, 1680–1687. [Google Scholar] [CrossRef]
  38. La Barbera, S.; Vincent, A.F.; Vuillaume, D.; Querlioz, D.; Alibart, F. Interplay of multiple synaptic plasticity features in filamentary memristive devices for neuromorphic computing. Sci. Rep. 2016, 6, 39216. [Google Scholar] [CrossRef]
  39. Chang, T.; Jo, S.H.; Lu, W. Short-term memory to long-term memory transition in a nanoscale memristor. ACS Nano 2011, 5, 7669–7676. [Google Scholar] [CrossRef]
  40. Li, J.; Ge, C.; Lu, H.; Guo, H.; Guo, E.J.; He, M.; Wang, C.; Yang, G.; Jin, K. Energy-Efficient Artificial Synapses Based on Oxide Tunnel Junctions. ACS Appl. Mater. Interfaces 2019, 11, 43473–43479. [Google Scholar] [CrossRef]
  41. Yan, X.; Zhao, Q.; Chen, A.P.; Zhao, J.; Zhou, Z.; Wang, J.; Wang, H.; Zhang, L.; Li, X.; Xiao, Z.; et al. Vacancy-Induced Synaptic Behavior in 2D WS2 Nanosheet–Based Memristor for Low-Power Neuromorphic Computing. Small 2019, 15, 1901423. [Google Scholar] [CrossRef]
  42. Li, X.; Yu, B.; Wang, B.; Bao, L.; Zhang, B.; Li, H.; Yu, Z.; Zhang, T.; Yang, Y.; Huang, R.; et al. Multi-terminal ionic-gated low-power silicon nanowire synaptic transistors with dendritic functions for neuromorphic systems. Nanoscale 2020, 12, 16348–16358. [Google Scholar] [CrossRef]
  43. Parmar, V.; Kingra, S.K.; Suri, M. Short-term plasticity circuit device exploration in the MASTISK neuromorphic framework. J. Phys. D Appl. Phys. 2018, 51, 454004. [Google Scholar] [CrossRef]
  44. Zhu, L.Q.; Wan, C.J.; Guo, L.Q.; Shi, Y.; Wan, Q. Artificial synapse network on inorganic proton conductor for neuromorphic systems. Nat. Commun. 2014, 5, 3158. [Google Scholar] [CrossRef] [Green Version]
  45. Abbott, L.F.; Regehr, W.G. Synaptic computation. Nature 2004, 431, 796–803. [Google Scholar] [CrossRef]
  46. Khanas, A.; Hebert, C.; Becerra, L.; Portier, X.; Jedrecy, N. Second-Order Memristor Based on All-Oxide Multiferroic Tunnel Junction for Biorealistic Emulation of Synapses. Adv. Electron. Mater. 2022, 8, 2200421. [Google Scholar] [CrossRef]
  47. Guo, J.; Liu, Y.; Li, Y.; Li, F.; Huang, F. Bienenstock-Cooper-Munro Learning Rule Realized in Polysaccharide-Gated Synaptic Transistors with Tunable Threshold. ACS Appl. Mater. Interfaces 2020, 12, 50061–50067. [Google Scholar] [CrossRef]
  48. Wu, Y.; Wang, X.; Lu, W.D. Dynamic resistive switching devices for neuromorphic computing. Semicond. Sci. Technol. 2022, 37, 024003. [Google Scholar] [CrossRef]
  49. Kyu Yang, M.; Ju, H.; Hwan Kim, G.; Lee, J.K.; Ryu, H.C. Direct evidence on Ta-metal phases igniting resistive switching in TaOx thin film. Sci. Rep. 2015, 5, 14053. [Google Scholar] [CrossRef] [Green Version]
  50. Wang, R.; Shi, T.; Zhang, X.; Wu, Z.; Liu, Q. A dual-functional Ta/TaO: X/Ru device with both nonlinear selector and resistive switching behaviors. RSC Adv. 2021, 11, 18241–18245. [Google Scholar] [CrossRef]
  51. Kim, S.; Abbas, Y.; Jeon, Y.R.; Sokolov, A.S.; Ku, B.; Choi, C. Engineering synaptic characteristics of TaOx/HfO2 bi-layered resistive switching device. Nanotechnology 2018, 29, 415204. [Google Scholar] [CrossRef]
  52. Jung, P.Y.; Panda, D.; Chandrasekaran, S.; Rajasekaran, S.; Tseng, T.Y. Enhanced Switching Properties in TaOx Memristors Using Diffusion Limiting Layer for Synaptic Learning. IEEE J. Electron. Devices Soc. 2020, 8, 110–115. [Google Scholar] [CrossRef]
  53. Lee, S.H.; Moon, J.; Jeong, Y.J.; Lee, J.; Li, X.; Wu, H.; Lu, W.D. Quantitative, Dynamic TaOxMemristor/Resistive Random Access Memory Model. ACS Appl. Electron. Mater. 2020, 2, 701–709. [Google Scholar] [CrossRef]
Materials 15 08858 g001 550
Figure 1. (a) Bipolar resistive switching characteristics of the ITO/WOx/TaOx/TiN device. (b) Schematic description of the SET and RESET process showing the formation and annihilation of conductive filament. (c) DC endurance characteristics with 100 stable cycles. (d) Multiple devices show promising device-to-device reliability with a uniform RON/ROFF ratio. (e) I–V characteristics of the ITO/WOx/TiN device at the current compliance of 10 mA without RESET characteristics. (f) The distribution histogram of HRS and LRS for 30 ITO/WOx/TaOx/TiN memristors.
Figure 1. (a) Bipolar resistive switching characteristics of the ITO/WOx/TaOx/TiN device. (b) Schematic description of the SET and RESET process showing the formation and annihilation of conductive filament. (c) DC endurance characteristics with 100 stable cycles. (d) Multiple devices show promising device-to-device reliability with a uniform RON/ROFF ratio. (e) I–V characteristics of the ITO/WOx/TiN device at the current compliance of 10 mA without RESET characteristics. (f) The distribution histogram of HRS and LRS for 30 ITO/WOx/TaOx/TiN memristors.
Materials 15 08858 g001
Materials 15 08858 g002 550
Figure 2. (a) Increasing LRS behavior of the ITO/WOx/TaOx/TiN memristor device by varying the SET current compliance from 1 × 10−7 A to 7 × 10−5 A. (b) Variations of multiple LRSs at each current compliance from 17.2 MΩ to 0.15 MΩ. (c) Multilevel memory state achieved by increasing VRESET from −0.8 V to −2.0 V at the current compliance of 1 × 10−5 A. (d) Endurance characteristics of 7 HRSs under different VRESET values where the ION/IOFF ratio increases gradually from 2.2 to 73.8.
Figure 2. (a) Increasing LRS behavior of the ITO/WOx/TaOx/TiN memristor device by varying the SET current compliance from 1 × 10−7 A to 7 × 10−5 A. (b) Variations of multiple LRSs at each current compliance from 17.2 MΩ to 0.15 MΩ. (c) Multilevel memory state achieved by increasing VRESET from −0.8 V to −2.0 V at the current compliance of 1 × 10−5 A. (d) Endurance characteristics of 7 HRSs under different VRESET values where the ION/IOFF ratio increases gradually from 2.2 to 73.8.
Materials 15 08858 g002
Materials 15 08858 g003 550
Figure 3. (a) Potentiation behavior under an identical pulse sequence of amplitude 1.1 V/100 µs to 1.25 V/100 µs. (b–d) EPSC obtained in the TO/WOx/TaOx/TiN memristor during the transition from STP to LTP was observed by increasing the pulse amplitude and number from 1.1 V/100 µs to 1.25 V/100 µs.
Figure 3. (a) Potentiation behavior under an identical pulse sequence of amplitude 1.1 V/100 µs to 1.25 V/100 µs. (b–d) EPSC obtained in the TO/WOx/TaOx/TiN memristor during the transition from STP to LTP was observed by increasing the pulse amplitude and number from 1.1 V/100 µs to 1.25 V/100 µs.
Materials 15 08858 g003
Materials 15 08858 g004 550
Figure 4. (a) Pulse-frequency-dependent current response of the ITO/WOx/TaOx/TiN memristor with trains of 20 pulses at a frequency range of 100 Hz to 2 Hz with an amplitude of +1.5 V/200 µs. (b) EPSC gain was obtained as A20/A1 by the pre-synaptic pulse sequence at different frequencies. (c) Post-synaptic current response at the same voltage amplitude of +1.5 V/200 µs with decreasing frequency.
Figure 4. (a) Pulse-frequency-dependent current response of the ITO/WOx/TaOx/TiN memristor with trains of 20 pulses at a frequency range of 100 Hz to 2 Hz with an amplitude of +1.5 V/200 µs. (b) EPSC gain was obtained as A20/A1 by the pre-synaptic pulse sequence at different frequencies. (c) Post-synaptic current response at the same voltage amplitude of +1.5 V/200 µs with decreasing frequency.
Materials 15 08858 g004
Materials 15 08858 g005 550
Figure 5. (a) Post-synaptic current response observed by applying a paired pulse with the different intervals of 0.1 ms to 1 ms on the ITO/WOx/TaOx/TiN memristor device. Pre-synaptic pulses were applied with an amplitude of +1.2 V/100 µs. (b) Experimental PPF ratios were calculated as a function of the pulse interval plotted as (A2/A1) × 100% with a simulated curve using the model in Equation (1).
Figure 5. (a) Post-synaptic current response observed by applying a paired pulse with the different intervals of 0.1 ms to 1 ms on the ITO/WOx/TaOx/TiN memristor device. Pre-synaptic pulses were applied with an amplitude of +1.2 V/100 µs. (b) Experimental PPF ratios were calculated as a function of the pulse interval plotted as (A2/A1) × 100% with a simulated curve using the model in Equation (1).
Materials 15 08858 g005
Table
Table 1. Comparison of resistive switching and synaptic properties of the TaOx and TaOx-bilayer-based memristor.
Table 1. Comparison of resistive switching and synaptic properties of the TaOx and TaOx-bilayer-based memristor.
Device
Structure		VSET
(V)		VRESET (V)ON/OFF
Ratio		Switching
Characteristics		Device-to-
Device
Reliability		Synaptic
Properties		Ref.
Ta/TaOx/Pt+0.65 V–0.6 V>10AbruptN/AN/A[49]
Ir/TaOx/W+2.5 V–2.5 V>10GradualN/AN/A[30]
W/Ta/TaOx/
Pt		+0.75 V–1.5 V>10GradualN/AN/A[33]
Ta/TaOx/Ru–1.5 V+2.0 V>100AbruptN/AN/A[50]
Pt/TaOx/n-Si–2.5 V+1.5 V>10AbruptN/AN/A[23]
Pt/TaOx/Ta–0.8 V+1.0 V>10AbruptN/AN/A[26]
Pt/TaOx/Ta–0.8 V+1.0 V>10GradualN/AN/A[18]
Ti/TaOx/HfO2/Pt+1.5 V–1.7 V<10GradualN/APD, STDP[51]
TaN/Ta/TaOx/
Al2O3/Pt		+1.5 V–1.8 V<10GradualN/APD[52]
TiN/SiO2/
TaOx/Pt		+0.5 V–1.5 V>10GradualN/APD, STDP[31]
TaN/Ta/HfO2/
TaOx/Pt		+0.7 V–0.9 V>10GradualN/APD[1]
TiN/Ti/TaOy/HfOx/
Au		–5.0 V+5.0 V>10GradualN/APD, STDP[27]
W/Al2O3/TaOx/
TiN		+1.5 V–2.2 V<10GradualN/AN/A[29]
Pt/TaOx/HfO2/Pt+10 V–5.0 V>10GradualN/AN/A[2]
Ta/WOx/TaOx/Pt–4.0 V+4.0 V>10GradualN/APD, PPF, SRDP[17]
Pd/Ta2O5/TaOx/Pd–1.8 V+1.0 V>10GradualN/APD[53]
ITO/WOx/TaOx/
TiN		+1.5–1.8 V>10GradualGood
(Figure 1d)		PD, STP, LTP, SADP, SRDP, PPFThis work
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Mahata, C.; Pyo, J.; Jeon, B.; Ismail, M.; Kang, M.; Kim, S. Forming-Free Tunable Analog Switching in WOx/TaOx Heterojunction for Emulating Electronic Synapses. Materials 2022, 15, 8858. https://doi.org/10.3390/ma15248858

AMA Style

Mahata C, Pyo J, Jeon B, Ismail M, Kang M, Kim S. Forming-Free Tunable Analog Switching in WOx/TaOx Heterojunction for Emulating Electronic Synapses. Materials. 2022; 15(24):8858. https://doi.org/10.3390/ma15248858

Chicago/Turabian Style

Mahata, Chandreswar, Juyeong Pyo, Beomki Jeon, Muhammad Ismail, Myounggon Kang, and Sungjun Kim. 2022. "Forming-Free Tunable Analog Switching in WOx/TaOx Heterojunction for Emulating Electronic Synapses" Materials 15, no. 24: 8858. https://doi.org/10.3390/ma15248858

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop